METHOD AND DEVICE FOR DETECTION OF WHOLE ORGANISM BACTERIA

BACKGROUND OF THE INVENTION

The present invention relates to a point-of-care device for detecting whole- organism bacteria in a water source, a method of detecting whole-organism bacteria from a water source with the use of the device and a method of making the device.

According to the World Health Report (WHO, 2002) 3.4 million people die each year from water, sanitation, and hygiene-related causes, 99% of which occur in the developing world. The most common disease causing contaminant found in water supplies is the faecal bacterium, E. coli although other pathogenic bacteria, such as Enterobacter, Salmonella, Shigella and Pseudomonas, are also commonly found. As a result of sewage contamination, or due to inadequate disinfection at waste water treatment plants, these pathogenic bacteria enter our water and food supplies. When consumed, these bacteria can cause various diseases which include diarrhoea, urinary tract infections, and meningitis.

In particular, the presence of E. coli has been used for many years to indicate the microbial quality of water supplies. As a result, water service providers and water service authorities regularly monitor water supplies for the presence of this bacterium. When appropriately treated, there should be little to no E. coli in the water. Acceptable levels of this bacterium in drinking water is 0 colony forming units (cfu) in a 100 ml (SANS 241 2011 , WHO, 2011), and 1000 cfu in 100 ml of treated waste water (Tshwane Water, South Africa).

Conventional detection methods for pathogenic water-borne bacteria include membrane filtration and multiple tube fermentation (Rompre et al. 2002) and require a 24-48 hour enrichment period. This prolonged detection time leaves water users exposed to potentially harmful water supplies, and users are often only notified of contamination when it's already too late. Furthermore, laboratory facilities and trained microbiologists are required to perform these tests, which are expensive and often confined to urban based laboratories, requiring transport of samples to the laboratory, resulting in testing of water sources in outlying areas only occurring rarely.

Traditional lateral flow devices are well known to those skilled in the art. To date, these devices have been used to detect, among others, pathogenic bacteria and toxins, infectious viruses, drugs and pesticides. However, to apply the antibodies used for detection onto the lateral flow test strip in a reliable, reproducible manner, very expensive, low volume dispensing machines are required. Such dispensing machines are not easily implemented in low resource settings. As a result, many research groups with biological sensing expertise, but with limited funding, cannot easily delve into the development of low cost lateral flow based diagnostics. This impacts the rate at which life-saving diagnostics are developed. To make sensing using paper substrates more amenable to research and development in under resourced areas, paper based microfluidic devices may offer a solution. These do not utilise various specialised papers (like membranes and conjugate pads). They are manufactured using a single piece of paper (usually chromatography paper), and are either cut into a specific pattern or utilise hydrophobic barriers in order to guide the sample towards the detection zones. These zones are preloaded with sensing reagents that are spotted into place using a pipette. As a result, these sensors do not need expensive dispensing machines for manufacture, and since a single piece of paper forms the device rather than three or four, manufacturing is made easier.

However, little work has been done to characterise and understand the interaction between chromatography paper, the paper commonly used for device construction, and the biological or chemical reagents used for sensing on board the device. Hence little is known about the effect chromatography paper has on the sensitivity of the device. Nitrocellulose membranes on the other hand, the most common type of paper used on lateral flow tests, has been well characterised.

The applicant therefore started with the development of a traditional lateral flow test first and gauged the performance of the detection mechanism on this more well established platform, before applying it to a paper based microfluidic device. In this way, any change in terms of sensing capabilities due to the use of chromatography paper rather than specialised membranes, could be identified immediately. There currently exists no truly low cost method for the fabrication of lateral flow tests. However, the applicant has demonstrated that rather than requiring the use of expensive equipment to deposit of lines of reagent onto the test substrate, as is traditionally performed, a pipette may simply be used to deposit spots of reagent.

Researchers working with limited resources can therefore use lateral flow devices in this fashion to demonstrate proof of concept in a low cost manner. Furthermore, researchers in the field of paper based microfluidics may use this technique for initial development and optimisation of sensing mechanisms before transferring the methodology to a paper based microfluidic device. In this way, any problems that arise during the development of the paper based microfluidic device can be attributed immediately to the paper or the new sensor format itself, since the sensing mechanism has already been proven operational.

The present invention describes the development and optimisation of a lateral flow test for E. coli detection in water sources and its incorporation of the optimised and well characterised lateral flow test into a paper based microfluidic device. The sensing mechanism of the testing system (based on immunoassay technology) was optimised first on the lateral flow test, and then only transferred to the paper based microfluidic device due to it being a more well established and better understood testing platform. In contrast to paper based microfluidic devices, lateral flow tests are constructed using paper types specially developed to promote immunoassay function, analyte transfer, etc. As a result, these paper types will enable the optimisation of the sensing mechanism, without any negative influence from the paper itself. Paper based microfluidic devices on the other hand make use of lower cost, less specialised paper types, such as chromatography paper. These types of papers are more robust, are less sensitive to environmental conditions, and can withstand the printing and heating processes used during the manufacture of paper based microfluidic devices. As these papers were not designed to promote immunoassay functionality, they can influence the sensing mechanism in a negative manner. However, the optimal performance of the sensing mechanism, as determined using the lateral flow testing format, will serve as a guideline on how the paper based microfluidic device should perform. Should the switch over influence the sensor systems negatively, the reason could almost be exclusively due to the change in paper type, since no other changes were made to the sensing system. As a result, steps can be taken to overcome this and restore the performance of the sensor back to that obtained on the optimised lateral flow test.

Various paper based microfluidic device fabrication techniques exit. These include photolithography, polydimethylsiloxane (PDMS) plotting, wax printing, laser cutting, and plasma and inkjet etching (Dungchai et al. 2011). The photolithography technique, while expensive, creates devices of high resolution. This method requires the use of clean room facilities which include UV light sources, expensive photoresists, and oxygen plasma. The PDMS manufacturing technique makes use of a standard desktop plotter to deposit PDMS droplets onto paper in a desired pattern. It is less expensive than photolithography, but has a lower resolution. It produces devices that are flexible, an important requirement since these devices are often exposed to mechanical stresses such as bending and folding. Wax printing uses a solid ink printer which deposits wax onto the surface of paper. The wax is then melted into the depth of the paper using a hot plate or an oven. Total fabrication time is less than 5 minutes and does not require the use of organic solvents or any pre- treatment of the paper. A solid wax printer, while expensive, is capable of manufacturing devices at approximately $0,001 per device, which is less expensive than all other techniques mentioned. This printing technique is also more easily up- scaled for mass production compared to the other processes. Large scale production is envisioned to be similar to newspaper printing techniques. Wax is also more environmentally friendly, readily available, and stable at high temperatures (60°C), temperatures which are common in developing countries. The resolution of wax printed devices is typically lower than other manufacturing methods, and this is due to the melting step required in the process. Other groups have developed screen printing methods to remove the need for an expensive solid ink printer. However, up scaling such a production technique has proved difficult.

The applicants explored the use of the wax printing manufacturing technique in order to create a paper based microfluidic device for the detection of E. coli in water. Optimal fabrication parameters, such as the optimal wax melting times, melting temperatures and optimal line thicknesses were investigated. Only a few studies reporting the optimal melting time and melting temperatures, as well as the optimal heating source have been found to date.

Finally, a paper based microfluidic device was created using all the optimal parameters determined in the aforementioned studies. This device was then combined with the immunoassay sensing mechanism optimised in Examples 1 and 2, in order to create a low cost E. coli sensor.

Hossain et al. (2012) developed a paper sensor to detect E. coli and coliforms in recreational water. This device was fabricated by cutting Whatman no. 1 chromatography paper into individual test strips and loading them with sensing reagents. The wax printing technique was not used. As a result, fluid movement towards the detection zones was controlled purely by the shape and dimensions of the test strip. To apply the test to drinking water analysis, (where as little as 1 cfu/100 ml must be detected), a combined immuno-magnetic separation (IMS) and an 8 hour culturing step was used. As enzyme-substrate technology is employed on the sensor, chemical lysis of the sample is also required. Due to these specialised sample preparation requirements, the sensor cannot be employed in the field by semi-skilled workers, and the cost of testing rises, negating the benefit of using a low cost paper substrate. Furthermore, IMS contributes significantly to the cost, so identifying another means of pre-concentration is vital to reduce the cost of testing.

Jokerst et al. (2011) developed a colorimetric paper based test that utilises enzyme-substrate technology to detect microbial pathogens in food samples. The device was fabricated by printing and then melting wax circles into Whatman no. 1 chromatography paper. These circles serve as the wells in this paper based ELISA (enzyme-linked immunosorbent assay)-like plate, into which reagents and sample have to be manually added for detection to occur. As a result, the device does not enable autonomous delivery of the sample into the detection zones, like the device being developed in this paper. To detect E. coli, the sensor developed by Jokerst ei al targets B-galactodisase, the enzyme common to all coliforms and hence E. coli. By targeting this enzyme, the sensor will detect all coliforms in a sample, and therefore cannot be considered an E. co//-selective sensor. Since an enzyme is targeted, bacteria lysis is further required prior to analysis. To detect low numbers of E. coli (10 cfu), a 12 hour enrichment period is required. As the author was unable to show the detection of a single E. coli, the device cannot be used for drinking water analysis. The sensor is structured as an ELISA plate, such that sample and reagents must be added into the detection zones manually. As a result, the sensor can only be used by trained lab personnel. Due to the need for lysis, enrichment, and manual addition of reagents, the sensor becomes unsuitable for use in the field by unskilled workers.

A simple, point of care testing device that can be used at the water source, negating the need for sample transportation and laboratory testing by skilled personnel would therefore be greatly beneficial.

SUMMARY OF THE INVENTION

According to a first aspect of the invention, there is provided a point-of-care device for detection of whole-organism bacteria in a fluid sample, in particular a water sample.

The point-of-care device may be a lateral flow sensor or a paper based microfluidic device.

The point-of-care device may further comprise one or more polyclonal antibodies specific for a species of whole-organism bacteria to be detected.

The whole-organism bacteria may be selected from the group Enterobacteriaceae, comprising any one or more of Salmonella sp., Shigella sp., Klebsiella sp., Escherichia coli (E. coli) sp. or Pseudomonas sp.). Preferably, the whole-organism bacteria are £ coli sp.

Preferably, the species of whole-organism bacteria comprise multiple strains of bacteria within the species.

The one or more polyclonal antibodies specific for the whole-organism bacteria may be a capture antibody (Ab). The one or more polyclonal antibodies may be a detection antibody. Preferably, the one or more polyclonal antibodies are both a capture and a detection antibody. The capture and detection polyclonal antibodies may be the same polyclonal antibody or different polyclonal antibodies specific for the whole-organism bacteria.

The point-of-care device may further comprise a monoclonal antibody or antibody fragment specific for the whole-organism bacteria. The monoclonal antibody or antibody fragment may be a capture antibody. The monoclonal antibody or antibody fragment may be a detection antibody. For example, where the polyclonal antibody is a capture antibody, the monoclonal antibody or antibody fragment is a detection antibody. Alternatively, where the polyclonal antibody is a detection antibody, the monoclonal antibody or antibody fragment is a capture antibody.

Typically, the detection polyclonal antibody, monoclonal antibody or antibody fragment is labelled with an indicator molecule.

The indicator molecule may be any one of a chromatographic, optical, fluorescent, electron transfer-based, radio-labeled or other indicator known to those skilled in the art. Preferably, the indicator molecule is a gold nanoparticle.

The point-of-care device may further comprise a control antibody. The control antibody may be a polyclonal or monoclonal antibody or antibody fragment thereof. For example, the control antibody may be an antibody to the detection antibody.

For example, an antibody fragment may be a F(ab')2 fragment, Fab fragment or chimeric Fab fragment (cFab).

Typically, the lateral flow device comprises:

(i) a sample pad, for loading a fluid test sample; (ii) a conjugate pad, comprising a polyclonal or monoclonal detection antibody or antibody fragment, wherein the detection antibody or antibody fragment binds to an O and/or K antigen on the surface of the whole organism bacteria, wherein the detection antibody is labelled with an indicator molecule;

(iii) a test membrane, having two regions, wherein the first region

comprises a polyclonal or monoclonal capture antibody or antibody fragment, wherein the capture antibody or antibody fragment binds to an O and/or K antigen on the surface of the detection antibody- labelled whole organism bacteria and wherein the second region comprises a polyclonal or monoclonal control antibody or fragment specific for the detection of the detection antibody; and

(iv) an absorbent wick,

wherein the four overlapping sections are in fluid connection with each other; wherein where conjugate pad comprises a polyclonal antibody, the test membrane may comprise a different or the same polyclonal antibody, and wherein where the conjugate pad comprises a monoclonal antibody or fragment, the test membrane comprises a polyclonal antibody and where the test membrane comprises a monoclonal antibody or fragment, the conjugate pad comprises a polyclonal antibody.

For example, the conjugate pad may be a glass-fiber membrane, or any other suitable conjugate pad known to those skilled in the art. For example, the test membrane may be a nitrocellulose membrane known to those skilled in the art.

Preferably, the test membrane further comprises the control antibody.

Typically, the paper based microfluidic device comprises:

(i) a paper substrate; and

(ii) a microfluidic conduit defined by a hydrophobic barrier on the paper substrate,

wherein the microfluidic conduit comprises a sample inlet zone, a conjugate zone, a test zone and a wick;

wherein the conjugate zone, comprises a polyclonal or monoclonal detection antibody or antibody fragment and wherein the detection antibody or antibody fragment binds to an O and/or K antigen on the surface of the whole-organism bacteria, wherein the detection antibody is labelled with an indicator molecule; and wherein the test zone comprises a first region comprising a polyclonal or monoclonal capture antibody or antibody fragment, and wherein the capture antibody or antibody fragment binds to an O and/or K antigen on the surface of the detection antibody-labelled whole-organism bacteria and a second region comprising a polyclonal or monoclonal control antibody or fragment specific for the detection of the detection antibody.The one or more polyclonal antibodies may be a capture antibody or a detection antibody. For example, the capture polyclonal antibody may be the same or a different polyclonal antibody to the detection polyclonal antibody.

The microfluidic conduit may be defined by a melted wax border forming a hydrophobic wax barrier penetrating the cross section of the paper substrate.

The paper substrate may be a chromatography paper. For example, the paper substrate may be Whatman no. 1 chromatography paper. The test area of the device may further comprise a monoclonal antibody or antibody fragment to the whole-organism bacteria. The monoclonal antibody or antibody fragment may be a capture antibody. The monoclonal antibody or antibody fragment may be a detection antibody. For example, where the polyclonal antibody is a capture antibody, the monoclonal antibody or antibody fragment is a detection antibody. Alternatively, where the polyclonal antibody is a detection antibody, the monoclonal antibody or antibody fragment is a capture antibody.

Preferably, the test area of the device comprises the control antibody.

In particular, the microfluidic conduit may define a plurality of zones within the test area. For example, the test area may comprise a sample inlet zone, a detection zone and a wick zone in fluid communication within the test area. Preferably, the detection zone comprises a test zone and a control zone. Further preferably, the sample inlet zone is in fluid communication with a conjugate zone which is further in fluid communication with the detection zone.

In particular, the microfluidic conduit may be defined by a printed wax border having a width of between about 300 pm and about 1000 μητι. Preferably, the printed wax border has a width of about 500 pm.

The paper based microfluidic device or lateral flow device may further comprise a calibration scale for quantitation of colorimetric results printed on the paper substrate.

The presence of a positive signal on the lateral flow device or microfluidic device from a fluid sample which has been pre-cultured in a nutrient broth for between 5 and 6 hours is indicative that the fluid sample contains 9000 cfu or less of the whole-organism bacteria.

Similarly, the presence of a positive signal from a fluid sample which has been pre-cultured in a nutrient broth for 15 hours is indicative that the fluid sample contains between 1 and 9 cfu of the whole organism bacteria.

According to a further embodiment of the invention, there is provided a method of detecting whole-organism bacteria in a fluid sample with the use of the point-of-care device of the invention.

The lateral flow point-of-care device may be a lateral flow sensor or a paper based microfluidic device.

The method may comprise a step of contacting one or more polyclonal antibodies specific for a species of whole-organism bacteria to be detected with the whole-organism bacteria.

The whole-organism bacteria may be selected from the group Enterobacteriaceae, comprising any one or more of Salmonella sp., Shigella sp., Klebsiella sp., Escherichia coli (E. coli) sp. or Pseudomonas sp. Preferably, the whole-organism bacteria are E. coli sp.

Preferably, the species of whole-organism bacteria comprise multiple strains of bacteria within the species.

In particular, the method comprises the steps of:

(i) loading a fluid sample comprising the whole-organism bacteria onto the device;

(ii) contacting the whole-organism bacteria in the sample with a polyclonal or monoclonal detection antibody or antibody fragment to produce detection antibody-labelled whole-organism bacteria and remaining detection antibody;

(iii) contacting the detection antibody-labelled whole-organism bacteria in the sample with a capture polyclonal or monoclonal antibody or antibody fragment in a test zone or membrane of the device; and

(iv) detecting whether or not a positive signal is produced in the test zone or membrane of the device,

wherein where the detection antibody is a monoclonal antibody or fragment, the capture antibody is a polyclonal antibody and where the capture antibody is a monoclonal antibody or fragment, the detection antibody is a polyclonal antibody.

Preferably, the detection and capture antibodies are both polyclonal antibodies. The polyclonal detection and capture antibodies may be the same or different polyclonal antibodies.

The method may further comprise a step of contacting the remaining detection antibody from step (ii) with a control antibody in a test zone or membrane of the device, in which case a positive signal in the test and control zone of the device indicates that the fluid sample is positive for the whole-organism bacteria and a negative signal in the test and positive signal in the control zone of the device indicates that the fluid sample is negative for the whole-organism bacteria.

The positive signal may be a chromatographic, optical, fluorescent, electron transfer-based, radio-labeled or other signal.

The method may further comprise a step of comparing an intensity of the positive signal with a calibration scale thereby to quantify the amount of bacteria present in the fluid sample. For example, where the device is a paper based microfluidic device the calibration scale may be printed on the paper substrate.

The method may further comprise a step of pre-culture of the bacteria. For example, the bacteria may be incubated for a period of from about 5 to about 18 hours. The method may further comprise a step of quantifying the number of whole organism bacteria in the fluid sample by pre-culturing the fluid sample in a nutrient broth and testing the nutrient broth for the presence or absence of the whole- organism bacteria over the incubation period.

The presence of a positive signal on the lateral flow device or microfluidic device when using the method of the invention from a fluid sample which has been pre-cultured in a nutrient broth for between 5 and 6 hours is indicative that the fluid sample contains 9000 cfu or less of the whole-organism bacteria.

Similarly, the presence of a positive signal on the lateral flow device or microfluidic device when using the method of the invention from a fluid sample which has been pre-cultured in a nutrient broth for 15 hours is indicative that the fluid sample contains between 1 and 9 cfu of the whole organism bacteria.

The method may further comprise a step of vortexing the sample of bacteria prior to testing.

Preferably, when a step of pre-culture is used, the method is capable of detecting only live bacteria. For example, when a step of pre-culture is not used and the level of bacteria in the sample is equal to or higher than the detection limit of the test, the test is capable of detecting dead, live and viable-but-non-culturable bacteria. According to a further aspect of the invention, there is provided a method of making a point-of-care device of the invention for detection of whole-organism bacteria in a fluid sample.

In one embodiment, the point-of-care device is a lateral flow sensor.

In particular, the method of making the device comprises the following steps:

(i) providing a sample pad, conjugate pad, test membrane and an absorbent wick in fluidic connection with each other;

(ii) depositing a polyclonal or monoclonal detection antibody or antibody fragment specific for the whole-organism bacteria onto the conjugate pad;

(iii) depositing a capture polyclonal or monoclonal antibody or antibody fragment specific for the whole-organism onto the test membrane, wherein where the detection antibody is a monoclonal antibody or fragment, the capture antibody is a polyclonal antibody and where the capture antibody is a monoclonal antibody or fragment, the detection antibody is a polyclonal antibody.

Preferably, the detection and capture antibodies are both polyclonal antibodies. The polyclonal detection and capture antibodies may be the same or different polyclonal antibodies.

The device may further comprise depositing a control antibody onto the test membrane of the device. For example, the control antibody may be specific for the detection antibody. The control antibody may be a polyclonal or monoclonal antibody or antibody fragment.

The means of depositing may be by a pipette or with the use of an automated reagent striping machine.

In another embodiment, the point-of-care device is a paper based microfluidic device.

In particular, the method of making the device comprises the following steps:

(i) providing a paper substrate;

(ii) defining a microfluidic conduit on the paper substrate for directing the sample through a test area of the device; and

(iii) depositing one or more polyclonal antibodies specific for the whole- organism bacteria onto the device test area.

The step of defining a microfluidic conduit may be performed by a microfluidic device fabrication technique such as photolithography, polydimethylsiloxane (PDMS) plotting, wax printing, wax dipping, wax screen printing, laser cutting, plasma or inkjet etching or the like.

Preferably, the step of defining a microfluidic conduit is performed by wax printing, followed by melting the wax to form a hydrophobic wax barrier penetrating the cross section of the paper substrate.

The paper substrate may be a chromatography paper. For example, the paper substrate may be Whatman no. 1 chromatography paper.

The one or more polyclonal antibodies may be a capture antibody. The one or more polyclonal antibodies may be a detection antibody. Preferably, the one or more polyclonal antibodies comprise both a capture and a detection antibody. The capture and detection polyclonal antibody may be the same polyclonal antibody or different polyclonal antibodies specific for the whole-organism bacteria. The method may further comprise a step of depositing a monoclonal antibody or antibody fragment specific for the whole-organism bacteria onto the device test area. The monoclonal antibody or antibody fragment may be a capture antibody. The monoclonal antibody or antibody fragment may be a detection antibody. For example, where the polyclonal antibody is a capture antibody, the monoclonal antibody or antibody fragment is a detection antibody. Alternatively, where the polyclonal antibody is a detection antibody, the monoclonal antibody or antibody fragment is a capture antibody.

Typically, the detection polyclonal antibody, monoclonal antibody or antibody fragment is labelled with an indicator molecule.

The indicator molecule may be any one of a chromatographic, optical, fluorescent, electron transfer-based, radio-labeled or other indicator known to those skilled in the art. Preferably, the indicator molecule is a gold nanoparticle.

The method may further comprise a step of depositing a control antibody onto the device test area. The control antibody may be a polyclonal or monoclonal antibody or antibody fragment thereof. For example, the control antibody may be specific for the detection antibody.

For example, the antibody fragment may be a F(ab')2 fragment, Fab fragment or chimeric Fab fragment (cFab).

Typically, at least the capture and control antibodies are deposited onto the device test area as spots or the capture and control antibodies are printed onto the device. The means of depositing may be by a pipette.

The microfluidic conduit may define a plurality of zones within the test area. For example, the test area may comprise a sample inlet zone, a detection zone and a wick zone in fluid communication within the test area. Preferably, the detection zone comprises a test zone and a control zone. Further preferably, the sample inlet zone is in fluid communication with a conjugate zone which is further in fluid communication with the detection zone.

The microfluidic conduit when fabricated by wax printing may be defined by a printed wax border having a width of between about 300 Mm and about 1000 pm. Preferably, the printed wax border has a width of about 500 pm.

The wax printing of the microfluidic conduit may be performed by any suitable solid wax printer known to those skilled in the art. The heat source for melting the wax into a hydrophobic wax barrier may be an oven or a hot plate, or other suitable heat source known to those skilled in the art. Preferably, the heat source is a hot plate.

The melting temperature for heating the printed wax border into the hydrophobic wax barrier is between about 100°C to about 250°C. Preferably, the melting temperature is from about 100°C to about 200°C. Even more preferably, the melting temperature is about 200°C. The desired melting time is dependent on the width of the printed wax border and the melting temperature, but is typically from about 1 to 2 minutes.

The method may further comprise a step of printing a calibration scale for quantitation of colorimetric results on the paper substrate. BRIEF DESCRIPTION OF THE FIGURES

Non-limiting embodiments of the invention will now be described by way of example only and with reference to the following figures:

Figure 1 : shows an absorbance plot indicating the minimum concentration of antibody that stabilises the gold conjugate when antibody concentrations from 0.01 mg/ml to 0.128 mg/ml were tested.

Figure 2: shows the RGB plot that was used to ascertain the amount of red and purple colour in each of the vials used for this study.

Figure 3: shows a graph of the optimal control spot antibody concentration based on greyscale value measure with Image J and cost in Rands. Figure 4: shows a graph of the optimal test spot antibody concentration based on greyscale value measure with Image J and cost in Rands.

Figure 5: shows a graph of the optimal conjugate volume based on greyscale value measure with Image J and cost in Rands.

Figure 6: shows a graph of the trend from elution studies indicating decreasing

E. coli counts with furthering distance from sample pad.

Figure 7: shows a graphical illustration of a paper based microfluidic device for

E. coli detection.

Figure 8: shows a graph indicating the optimum printed line width based on greyscale value measure with Image J.

Figure 9: shows a graph of the comparison of effectiveness between an oven and hot plate at each line width based on greyscale value measure with Image J.

Figure 10: shows a graph of the optimum melting temperature for each line width based on greyscale value measure with Image J.

Figure 11: shows a graph of the optimum melting time for each line width based on greyscale value measure with Image J.

Figure 12: shows a graph depicting the effect of melting temperature on line resolution based on greyscale value measure with Image J, and

Figure 13: shows a graph depicting the effect of melting time on line resolution based on greyscale value measure with Image J.

Figure 14: shows the methods used for analysis of waste water samples; a) assessment of E. coli in 100ml of effluent, b) assessment of E. coli in

1ml of sediment pond, c) spiking technique used to create suspensions containing various bacteria counts for use in timing tests.

Once loaded, all petri dishes are incubated in a 37°C oven.

Figure 15: shows results of the 24-hour analysis of the effluent and sediment pond.

Figure 16: shows the time required for detection on lateral flow test based on bacteria count in sample (a) best case scenario, b) worst case scenario. PB= phosphate buffer; SP = sediment pond, E = effluent.

Figure 17: shows a comparison between the Colilert* test and the lateral flow test of the present invention (ColiSpot). *Colilert was performed in petri dishes and assessed for colour change Colilert trays were not used. DETAILED DESCRIPTION OF THE INVENTION

The present invention will now be described more fully hereinafter with reference to the accompanying drawings, in which some, but not all embodiments of the invention are shown.

The invention as described should not be limited to the specific embodiments disclosed and modifications and other embodiments are intended to be included within the scope of the invention. Although specific terms are employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.

As used throughout this specification and in the claims which follow, the singular forms "a", "an" and "the" include the plural form, unless the context clearly indicates otherwise.

The terminology and phraseology used herein is for the purpose of description and should not be regarded as limiting. The use of the terms "comprising", "containing", "having" and "including" and variations thereof used herein, are meant to encompass the items listed thereafter and equivalents thereof as well as additional items.

The present invention provides a point-of-care device for detecting whole- organism bacteria in a water source, a method of detecting whole-organism bacteria from a water source with the use of the device and a method of making the device.

While lateral flow and paper-based sensors have been developed for the detection of bacteria such as E. coli, most of the sensors can only detect either i) a single strain of the bacterium, ii) the toxins produced by the bacterium or iii) bacterium DNA (Alocilja et al. 2004; Zhao et al. 2010; Nakasonea et al. 2006). In particular, these sensors are not well suited for analysis at the source, for example at a water source, because i) many strains of bacteria exist in such sources, so merely detecting a single strain is insufficient, ii) to enable speed of testing, whole-bacteria should be detected to negate the need for DNA and toxin extraction. Furthermore, the tests detecting toxins can only detect pathogenic strains of bacteria. For water quality monitoring, both pathogenic and non-pathogenic types of bacteria must be detected. As a result, lateral flow tests (LFTs) based on the detection of toxins cannot be applied for water quality monitoring. LFTs detecting the DNA of bacteria are difficult to use in the field due to the need to extract, isolate, and sometimes even purify the DNA used for testing. Such techniques cannot be used by semi-skilled lab or field workers.

The benefits of the present invention include the ability detect most if not all E. coli occurring in environmental water sources, compared to the detection of a single strain of E. coli in food samples. A lateral flow test (LFT) for the detection of E. coli in water has been developed (MERCK), but this LFT detects only one strain of E. coli (0157:H7), and requires 24 hrs of pre-culturing. On the other hand the present invention allows for the detection of low £ coli counts (1 cfu) in under 15 hours, compared with 18 hours required for the Colilert £. coli detection system, and 24-48 hours required for the membrane filtration method (golden standard for E. coli detection). Further the present invention provides for the detection of whole organism £. coli, without the need to lyse the cells.

The applicant has therefore investigated the development of a point-of-care device and method for detection of whole-bacteria by means of polyclonal antibody- antigen reactions in order to capture and detect multiple strains of whole-bacteria in a fluid source. The device was developed to detect whole-organism bacteria consisting of the group Enterobacteriaceae, comprising any one or more of Salmonella sp., Shigella sp., Klebsiella sp., Escherichia coli (E. coli) sp. or Pseudomonas sp. Preferably, E. coli are detected by the device.

In particular, the point-of-care devices developed by the applicant aim to enable low cost detection of whole-organism bacteria, including at the source, such as a water source, in a manner that is faster than what is currently achievable using the current available detection technologies. Furthermore, with the use of polyclonal antibodies as either the detection or capture antibody, or both, the device is aimed at detection of multiple strains of the bacteria to be detected, since multiple epitopes may be bound by the polyclonal antibody. Such a device will allow field workers to test various sources for bacteria and make decisions faster. This is especially important during emergency situations, like bacterial outbreaks, where remediation efforts should occur as fast as possible to prevent the spread of disease.

Preferably, the reporting method is a visual reporting method to ensure that no additional read-out instrumentation is required and little to no user training is needed to interpret the test results. In particular, in the case of the paper-based microfluidic device, the device is made portable and low cost by miniaturising its components and combining them onto a lightweight paper-based substrate. This, combined with simple result read-out technology, makes the device well suited for point-of-care or field analysis.

The term "antibody" includes monoclonal antibodies, polyclonal antibodies, multispecific antibodies (for instance, bispecific antibodies and polyreactive antibodies), and antibody fragments. Accordingly, the term "antibody" as used in this specification includes, but is not limited to, any specific binding member, immunoglobulin class and/or isotype (for instance: lgG1 , lgG2, lgG3, lgG4, IgM, IgA, IgD, IgE and IgM) or an antibody fragment thereof.

It is understood in the art that an antibody is a glycoprotein comprising at least two heavy (H) chains and two light (L) chains which are inter-connected by disulfide bonds, or an antigen binding portion thereof. A heavy chain comprises a heavy chain variable region (VH) and a heavy chain constant region (CH1 , CH2 and CH3). A light chain comprises a light chain variable region (VL) and a light chain constant region (CL). The variable regions of both the heavy and the light chains comprise framework regions (FR's) and complementarity determining regions (CDR's). The four FR's are relatively conserved while the CDR regions (CDR1 , CDR2 and CDR3) comprise hypervariable regions. The FR's and CDR's are arranged from the NH ₂ terminus to the COOH terminus as follows: FR1 , CDR1 , FR2, CDR2, FR3, CDR3, FR4. The variable regions of the heavy and light chains contain a binding domain that interacts with an antigen. Further, the constant regions may mediate the binding of the immunoglobulin to host tissues or factors.

Also included in the definition of "antibody" are chimeric antibodies, humanized antibodies, recombinant antibodies, human antibodies generated from a transgenic non-human animal and antibodies selected from libraries using enrichment technologies available to those skilled in the art. The term "epitope" as used herein means any antigenic determinant on an antigen to which the paratope of an antibody can bind. Epitopic determinants usually consist of chemically active surface groupings of molecules such as amino acids or sugar side chains and usually have specific three dimensional structural characteristics, as well as specific charge characteristics.

An "antibody fragment" comprises a portion of an intact antibody, such as the antigen binding or variable region of the intact antibody. Examples of antibody fragments include, but are not limited to, Fab, Fab', F(ab')2, Fv fragments, scFV fragments; diabodies; or linear antibodies.

Papain digestion of antibodies produces two identical "Fab" fragments or antigen-binding fragments, each with a single antigen-binding site, and a residual "Fc" fragment, whose name reflects its ability to crystallize readily. Pepsin treatment of antibodies yields an F(ab')2 fragment that has two antigen-combining sites and which retains its ability to cross-link an antigen.

The term "Fv" refers to the minimum antibody fragment that contains a complete antigen-recognition and antigen-binding site. This fragment contains a dimer of one heavy- and one light-chain variable region domain in tight, non-covalent association. The folding of these two domains results in the formation of six hypervariable loops (three loops each from the H and L chain) that contribute the amino acid residues for antigen binding and confer antigen binding specificity to the antibody. However, even a single variable region (or half of an Fv comprising only three CDRs specific for an antigen) has the ability to recognize and bind an antigen, although at a lower affinity. "Single-chain Fv" ("sFv" or "scFv") are antibody fragments that comprise the VH and VL antibody domains connected into a single polypeptide chain. The sFv polypeptide can further comprise a polypeptide linker between the VH and VL domains that enables the sFv to form the desired structure for antigen binding.

The "Fab" fragments contain the constant domain of the light chain and the first constant domain (CH1) of the heavy chain. Fab' fragments differ from Fab fragments by the addition of a few residues at the carboxy terminus of the heavy chain CH1 domain including one or more cysteines from the antibody hinge region. Fab'-SH is the designation herein for Fab' in which the cysteine residue(s) of the constant domains bear a free thiol group. F(ab')2 antibody fragments originally were produced as pairs of Fab' fragments which have hinge cysteines between them. Other chemical couplings of antibody fragments are also known in the art.

Variant antibodies also are included within the scope of the invention. Thus, variants of the sequences recited in the application also are included within the scope of the invention. Further variants of the antibody sequences having improved affinity can be obtained using methods known in the art and are included within the scope of the invention. Those skilled in the art can modify the amino acid sequences of a polypeptide utilizing recombinant methods and/or synthetic chemistry techniques for the production of variant polypeptides. For example, amino acid substitutions can be used to obtain antibodies with further improved affinity. Alternatively, codon optimization of the nucleotide sequence can be used to improve the efficiency of translation in expression systems for the production of the antibody. Such variant antibody sequences will share 70% or more (i.e., 75%, 80%, 85%, 90%, 95%, 96%, 97%, 98%, 99% or greater) sequence identity with the sequences recited in the application. Such sequence identity is calculated with regard to the full length of the sequence recited in the application.

The following examples are offered by way of illustration and not by way of limitation.

EXAMPLE 1

Detection of E. coli using lateral flow and paper based microfluidic technology: Manufacturing lateral flow tests on a budget

Working principle

The proof-of-concept lateral flow sensor developed consists of four sections, each of which is prepared on a specific type of paper, and assembled together on an adhesive backing card. Through overlapping, a fluidic connection between each section is created, enabling the flow of the sample towards the detection regions. The four sections are the sample pad, conjugate pad, a nitrocellulose test membrane and an absorbent pad/wick. The sample pad serves to absorb and transfer the sample fluid onto the testing device. The conjugate pad (a glass-fibre membrane) is loaded with gold labelled, anti-E. coli antibodies, which are stored in a dehydrated state. The test membrane is loaded with two different antibodies. An anti-E coli antibody forms the test line/spot and an antibody specifically designed to detect the antibody in the conjugate forms the control line/spot. The control line/spot is used to indicate whether the sample has migrated through the entire length of the device. The purpose of the absorbent pad or wick is to help draw the sample fluid up the length of the test strip. The sandwich assay format was selected for this test as the analyte (E. coli) exhibits several epitopes, and therefore enables the binding of two antibodies onto each E. coli simultaneously. The sandwich assay format also lends itself to quantitation due to the proportional relationship that exists between the analyte concentration in the sample and the colour intensity of the test line/spot. When a sample is applied to the sample pad, the liquid migrates up the test by capillary forces. The analyte (E. coli) comes into contact with the detection antibodies in the conjugate pad, and become "labelled" with gold. Not all the detection antibodies bind to the E. coli however, and continue to flow unbound up the test. The labelled-E. coli is captured at the test line, resulting in the development of a red test signal. Both the labelled and unlabelled E. coli often compete for binding sites on the test line, and this sometimes results in the "hook effect". The gold labelled antibodies that have not bound to E. coli pass over the test line/spot to the control line/spot and become bound. This occurs as the control line/spot antibodies are raised to target the anti-E. coli antibodies present in the conjugate. This results in the formation of a control line signal. Hence the presence of two lines/spots on the test indicates the presence of E. coli in the sample, while the formation of only a control line/spot indicates the absence of E. coli in the sample. Any other combinations of signal that form are regarded as error. Materials

A polyclonal, rabbit anti-E coli antibody (Thermo Fischer Scientific, USA) was used in the conjugate and a polyclonal rabbit anti-E. coli antibody (ABD Serotec, UK) was used on the test line/spot. Both these antibodies detect all Ό" and "K" antigens on the E. coli bacterium. Goat-anti-rabbit antibodies (Kirkegaard & Perry Laboratories, Inc) were used on the control line/spot. Gold nanoparticle solution, with a particle size of 40 nm and an optical density (OD) of one was purchased from the Diagnostic Consulting Network (California, USA). Phosphate buffer powder was purchased from Sigma Aldrich. The powder was added to deionised water and used at a concentration 6.92 mg/ml. Bovine serum albumin (BSA) flakes (Sigma Aldrich), were dissolved in phosphate buffer and used at a concentration of 7 mg/ml.

Dialysis cassettes were purchased from Thermo Fisher Scientific and were used to dialyse the conjugate antibodies in borate buffer (0.67 g/L) overnight. This was done to remove excess salt contained in the storage buffer. The dialysed antibody was stored in aliquots in a -18°C freezer until used. The paper used for the development of the lateral flow test included the following. For the conjugate and sample pad, G041 Millipore glass fiber was used. A nitrocellulose test membrane (HF80HP, GE Health) and a Surewick CFSP wicking pad were also used. These papers were assembled on a PVC adhesive backing card purchased from Kenosha CV (Netherlands). A blocking solution was purchased from Invitrogen, and was used to block the test membrane.

The instrumentation used for this work included a Beckman Coulter Microfuge 16 desktop centrifuge, an Ecotherm oven (Model 22, Labotec), a pH meter (Eutech pH6+), a desktop scanner (HP Scanjet 2400) and a Canon powershot G-11 digital camera.

Samples

ATCC strains of bacteria were kindly provided by the Natural Resources and Environment Unit (CSIR, South Africa). These bacteria included Escherichia coli KM, E. coli 0157:H7, Enterobacter cloacae, Pseudomonas aueruginosa, Salmonella enteritidis and Acinetobacter. Bacteria were cultured using nutrient agar and nutrient broth, both of which were purchased from Oxoid. Heat killed E. coli 0157:H7 was purchased from Kirkegaard & Perry Laboratories Inc. All the above bacteria were handled in a biosafety laboratory, within a fume hood and in close proximity to a burning flame. Samples were created by spiking the bacteria into phosphate buffer (PB). Once the suspension turned turbid, its turbidity was compared to the McFarland standards, after which the absorbance of the suspension was determined using a UVA IS spectrophotometer (Perkin Elmer). In this way the concentration of bacteria in the sample was determined. The serial dilution technique was used to prepare samples of different bacterial counts. Methods

Conjugate preparation

Minimum amount of antibody required to stabilise the gold nanoparticles The gold nanoparticles used to prepare the conjugate were citrate capped and therefore exhibited a negative charge. When in suspension, electrostatic repulsion ensures that the gold nanoparticles remain well dispersed and maintain a specific distance between each other. This dispersion is what causes the gold solution to appear red. When sodium chloride is added to the gold solution, the positively charged sodium ions are attracted to the negatively charged gold particles. This forces the adjacent gold nanoparticles to aggregate, causing the optical properties of the gold solution to change. The colour of the gold solution changes from red to purple/blue. To prevent such aggregation, the gold nanoparticles can be coated with a protein layer, such as with antibodies (Abs) or BSA. These proteins act as an electrostatic shield, preventing the nanoparticles from interacting with the sodium ions. In this way, proteins serve to stabilise the nanogold in the presence of salt, and prevents any colour change from occurring. Using this phenomenon, the minimum amount of antibody required to stabilise the gold was determined. Antibody (Ab) solutions ranging in concentrations from 0.01 mg/ml to 0.1 mg/ml were prepared in 670 μΙ of phosphate buffer (PB). The antibodies were rabbit anti E. coli antibodies (Thermo Fisher Scientific). Each Ab solution was added to 3.33 ml of gold. Hence, an antibody to gold volume ratio of 1:5 was used. After allowing the antibody-gold solution to incubate for 30 minutes, a 10% (w/v) salt solution was added and the solution was monitored for any colour change. To confirm the colour changes observed, UV VIS spectrophotometry was used.

Optimal antibody to gold ratio

The optimal gold to antibody (Ab) volume ratio was determined using a similar technique to that used above. Varying volumes of OD1 gold were added to a set volume of antibody. The antibody concentration in that set volume was kept constant at the optimal concentration determined above. The antibody to gold volume ratio was varied from 1 :1 to 1 :20. After allowing the gold solution to incubate with the antibody for 30 minutes, a 10% w/v salt solution was added and the solution was monitored for any colour change. To confirm any colour changes, Image J (image analysis software) was used. An image of each solution was captured, after the addition of salt. Using image J, the RGB values of each solution was obtained and plotted alongside each other. In this way, the amount of red colour in each vial was semi-quantified and used to determine the maximum gold volume that could be used without causing the Ab-gold solution to become unstable.

Conjugate pad preparation

Millipore G041 conjugate paper was used as the conjugate pad. The conjugate pad was supplied as strips that are 30 cm long with a width of 1 cm. Before the conjugate was loaded onto the conjugate pad, the paper was first blocked. It is immersed into a membrane blocking solution, pat dried using paper towels, and then oven dried at 37°C for 1 hour. Blocking the paper helps prevent non-specific binding of the bacteria on the paper and also helps prevent the conjugate from becoming permanently fixed onto the conjugate pad. The OD 10 conjugate was then loaded onto the blocked conjugate pad. Each test will contain approximately 6 μΙ spots of conjugate. To guide the pipetting process, a marker pen was used to mark the position where the conjugate is to be loaded. Each strip of conjugate pad was placed alongside a ruler, to ensure that the markings are spaced 1 cm apart. The markings were located at the upper or lower edge of the conjugate paper strip so that they could be cut off before the conjugate pad is assembled onto the lateral flow test strips. Once loaded, the conjugate was dried into place in a 37°C oven for 30 minutes. Once dried the conjugate pad was ready to be assembled onto the final device. Normally, expensive spraying/striping machines are used to load the conjugate onto the conjugate pad. Using the method described above, the conjugate can be loaded onto the conjugate pad without the use of expensive machines, enabling the demonstration of proof of concept in a low cost manner.

Loading of test and control spots

Rabbit-anti E. coli antibodies (ABD Serotec) were deposited onto the test membrane as the test signal reagent. Goat-anti rabbit antibodies (KPL) were used as the control spot reagent.

Test and control signal reagents were loaded onto the nitrocellulose test membrane, such that each test contains approximately 1 μΙ of each reagent. As was done with the conjugate pad, a marker pen and ruler were used to mark and guide the positioning of these spots on the membrane. Using the ruler, horizontally adjacent test spots were spaced 1 cm apart. The same spacing was used for horizontally adjacent control spots. The control spots were positioned ±1 cm vertically above the test spots. The control spot was therefore located closer to the wick in the final lateral flow test strip. Maintaining this 1 cm spacing between horizontally adjacent test spots, horizontally adjacent control spots and horizontally adjacent conjugate spots allows for easy alignment of all three reagents on the backing card. This is especially important when the individual test strips are cut manually (using a pair of scissors), and one needs to ensure that each test strip contains all three reagents. When an automated guillotine is used to cut out individual test strips, the guillotine can be programmed to ensure that each strip is 1 cm. However, as this specification is focusing on low cost fabrication techniques, the markings or common spacing between adjacent reagent spots is important. When a striping or spraying machine is used to deposit all three reagents, the spacing of adjacent spots becomes irrelevant since the entire conjugate pad or test membrane contains reagent, i.e. there are no gaps or empty sections in between. Once loaded onto the membrane, the test and control spots are dried in a 37°C oven for 30 minutes.

As the nitrocellulose test membrane is flimsy and therefore easily prone to damage, it was first placed onto a PVC adhesive backing card before the test spots and control spots were added. The entire lateral flow test (LFT) is later assembled onto this backing card. Once the control spot and test spot antibodies were dried onto the test membrane, the membrane was immersed into a blocking solution, towel dried and thereafter dried for 1 hour in a 37°C oven. Thereafter the membrane was ready for final assembly. Assembly and preparation of LFTs

Once the test membrane and conjugate pads are loaded with their respective reagents, the LFT is assembled. The test membrane is applied to the middle section of the backing card. The wicking pad is positioned onto the upper most section of the backing card, and overlaps the test membrane by at least 2 mm. This overlapping creates a fluidic connection between the various sections of the LFT, and ensures that the sample is drawn from the sample pad to the wicking pad. The conjugate pad is placed so as to overlap the bottom edge of the test membrane by at least 2 mm. Using the markings on the conjugate pad and the nitrocellulose membrane, the conjugate spots are aligned with the test and control spots. Finally the sample pad is placed at the lowest position of the backing card, overlapping the bottom edge of the conjugate pad. The entire test card is then cut into individual test strips using a pair of sharp scissors. Each LFT will therefore contain a test, control and conjugate spot that are all well aligned with each other. To date, the inventors have found that manual cutting has no significant effect on the flow properties of the sample fluid, such that the performance of the sensor is unaffected by manual cutting. As a result, automated guillotines, like striping or spraying machines, may not be required during the proof of concept developmental phase. These findings prove that low resource diagnostic labs can successfully develop lateral flow tests without the use of expensive fabrication equipment. Such equipment may only be required during prototype development, once there is proof of concept. This avoids unnecessary investment before the detection system has actually been shown to work.

Running of lateral flow tests

To test each LFT, 200 μΙ of the spiked sample is transferred to a glass test tube, into which the LFT is placed with its sample pad immersed. A 200 μΙ volume of sample proved sufficient to wick through the entire length of the LFT. This volume also ensured that only the sample pad (and not the conjugate pad) makes direct contact with the fluid. The LFT absorbs the sample, and takes approximately 10-15 minutes to form a strong control and test line/spot signal.

Determination of the optimum control spot and test spot antibody concentration

A series of LFTs were made, each of which having varying concentrations of antibody in either the control spot or the test spot. All other fabrication parameters were kept constant when either of these concentrations was varied. The concentration of the control spot was varied from 0.2 mg/ml to 1 mg/ml, while keeping the test spot concentration constant. The aim of this was to identify how the concentration of the antibody influences the signal intensity of the control spot. To perform this two types of samples were used, a 10 ² cfu/ml heat killed E. coli 0157:H7 sample and a phosphate buffer sample. In order to determine the optimum test spot antibody concentration, a series of LFTs were made where the test spot antibody concentration was varied from 4 mg/ml to 0.5 mg/ml, the control spot antibody concentration was kept constant. A 10 ⁷ cfu/ml E. coli sample was used for this study. The change in colour intensity of the test spot in relation to the test spot antibody concentration was monitored. Image J computer software was used to analyse the signal intensities. Image J analyses a captured image and uses grey scale analysis to assess the colour intensity of that image. A high grey scale value indicates a light (or white) image while a low grey scale value indicates a dark (or black) image. By monitoring the grey scale value of the test and control spot signals with relation to their antibody concentration, the optimum test and control spot antibody concentrations could be determined.

Determination of the optimum conjugate volume

The inventors then determined how the volume/amount of conjugate used on an LFT influences the intensity of the test spot signal, as well as the detection limit of the device. Conjugate volumes ranging from 1 μΙ to 10 μΙ were loaded onto separate LFTs which were run with a 10 ⁷ cfu/ml heat killed E. coli 0157:1-17 sample. The test spot colour intensity was monitored both visually and using Image J. To ascertain if using a higher conjugate volume would improve the detection limit of the test, a set of LFTs containing 10 pl of conjugate (instead of the usual 6 μΙ) were made and run with varying concentrations of E. coli 0157:H7 samples. The sample concentrations ranged from 10° cfu/ml to 10 ⁹ cfu/ml. The lowest sample concentration that produced a positive test signal is considered the detection limit of the device. The detection limit obtained when using 10 μΙ of conjugate is then compared to the detection limit obtained when using 6 μΙ of conjugate. In this way, any improvement in the detection limit of the test as a direct result of an increased conjugate volume would be observed.

The influence of the test membrane on the signal intensity

Different types of membranes were placed in the test region of the device. Test and control spots were deposited onto each of these membranes. Once assembled, each test was run with 10 ⁷ cfu/ml heat killed £. coli 0.157:H7 samples. The aim was to examine the effect that different types of membranes have on the colour intensity of the test and control spots, and to select the optimal type.

Table 1: Nitrocellulose membranes used on the test region of the LFT

Paper Name Abbreviation

Whatman AE 100 AE100

Whatman FF85 FF85

Sartorius CN 140 CN140

Whatman Fusion Five F5

Millipore NF240 240

Whatmanlmmunopore FP FP

Whatman Prima 125 125

Whatman FF60/100 FF60

Millipore HF075 075

Whatman AE98 AE98

Vivid 70 Pall nitrocellulose Pall

Whatman AE 99 AE99

Millipore HF135 135

Millipore HF090 090 Optimal testing and drying times

The inventors consequently determined: a) how long tests should be run for, in order to determine the time duration for which the tests should be kept immersed in their sample; and b) how long after removing the LFT from its sample, should the test results be read.

When tests are removed from their sample and left exposed to ambient conditions for some time, sample backflow can influence the signal intensity, and can influence the test result. Sample backflow occurs when the sample pad becomes drier than the wick, forcing liquid to migrate from the wick back down towards the sample pad. During such backflow, the gold nanoparticles can deposit non- specifically on and around the test spot, creating a false positive signal. Furthermore, when left exposed to ambient conditions for too long, the test membrane dries out and alters the true colour of the test signals. The signal has been found to change from red to purple. This change in signal colour and intensity can lead to misinterpretation of test results, and can become especially problematic when a correlation between signal intensities and sample concentration must be determined.

For the drying time investigation, tests were left in their sample (E coli, 10 ⁷ cfu/ml) to run for approximately 15 minutes. Once they were removed from the sample, the LFTs were exposed to ambient conditions. Photos of the tests were taken every 2 minutes after they were removed from the sample and using image J software, were monitored for changes in signal colour and intensity.

For the testing time study, LFTs were left immersed in E. coli samples (10 ⁷ cfu/ml) and phosphate buffer samples and run for 30 seconds to 30 minutes. The signal intensities were examined as soon as the LFTs were removed from their samples. Control spot signals were examined on those tests run with PB only, while test spot signals were examined on those tests run with E. coli rich samples.

Test line position

The aim of this investigation was to determine how the vertical position of the test spot on the length of the membrane influences the signal strength. Several LFTs were made as per usual, however no control spots were used. This is to enable the movement of the test spot position along the entire length of the membrane section. Four devices were made. The test spot on each device was positioned approximately 4 mm, 7 mm, 12 mm and 13 mm from the upper edge of the conjugate pad. Heat killed E coli 10 ⁷ cfu/ml samples were used to run these tests. The difference in signal intensity between the various test spot positions was used to a) determine whether the test spot position influences the signal intensity, and b) if it does, determine the optimal test spot position.

The effect an increased test spot antibody concentration, and an increased conjugate volume and concentration has on the limit of detection

The inventors determined whether an increased concentration of antibody in the test spot and conjugate would influence the detection limit of the LFTs. To examine the influence that an increased test spot concentration will have, LFTs were made with a test spot concentration of 4 mg/ml. This is the highest concentration at which this particular antibody is supplied. E. coli 0157:1-17 at concentrations ranging from 10 ⁸ cfu/ml to 10° cfu/ml were used to test these LFTs. To determine the effect an increased Ab concentration in the conjugate will have, the concentration of the antibodies in the conjugate was increased to 0.12 mg/ml. This is almost 4 times the amount of antibody found sufficient for gold stabilisation. These tests were run with heat killed E. coli 0157:H7 samples ranging from 10 ⁸ cfu/ml to 10 ³ cfu/ml. To perform the increased conjugate volume studies, LFTs were manufactured with an increased conjugate volume of 10 μΙ. This is well above the conjugate volume determined as optimal. The LFTs were run with E. coli 0157:H7 suspensions ranging in concentrations from 10 ⁸ cfu/ml to 10° cfu/ml. In both conjugate studies it is expected that as more antibodies are available to label the E. coli in the sample, more labelled E. coli should be captured at the test spot. Ideally, this should enable the detection of E. coli at lower sample concentrations. The same applies to the increased test spot concentration studies. Since more antibodies are available to capture the labelled bacteria at the test spot, the detection limit of the test may be improved.

Results and Discussion

General comments on the low cost manufacturing process

Usually lateral flow tests have test and control lines and not spots. These lines are deposited onto the nitrocellulose membrane using expensive striping machines. As this invention focuses on low cost fabrication methods for paper sensors, a pipette, instead of a striping machine, was used to deposit the test and control antibodies. As a result, the signals on the LFT have a circular shape, and are hence referred to as test and control spots. While unusual, these rounded spots still form strong, clearly visible signals. Visualisation of these signals is all that is required in order to confirm the functionality of the sensing mechanism, and hence the functionality of the LFT. As a result, no striping machines are required. To determine if the use of spots instead of lines negatively influences the performance of lateral flow tests, a set of LFTs were made externally, using state of the art striping machines. The detection limit of these tests (with lines) was the same as that of the tests with spots. Hence it is believed that lateral flow devices can be optimised to a far extent without ever having to purchase striping machines. The spots only become difficult to work with when quantification of the signals is required. While the signal intensities of the line and spot look similar to the naked eye, the crest shape of the spotted signal makes it difficult to quantify. This is due to the difficulty associated with selecting an irregularly shaped area for image analysis. A pipette was also used to load the conjugate onto the conjugate pad in a low cost manner. Normally a spraying machine is used for this purpose. Using a pipette for conjugate deposition appears to have no effect on the extent of conjugate release from the conjugate pad. Little to no residue is left behind on the conjugate pad once the test is run. To make a comparison, LFTs were manufactured by depositing the conjugate using a spraying machine and then assessing the detection limit of these LFTs. The tests had the same limit of detection as those tests made with conjugate spots. However, the distribution of the conjugate across the width of the test strip as it flowed upwards, differed between the two. When a pipette is used for conjugate deposition, the conjugate is positioned at the centre of the conjugate pad. As a result, when released, the conjugate flows up mostly along the centre of the strip width. To reduce this effect, the conjugate can be loaded across the entire width of the conjugate pad. Nonetheless, should the conjugate spot be well aligned with the test and control spot, this issue becomes irrelevant. Hence spotting of the conjugate is sufficient during proof of concept development. In summary, it is believed that striping/spraying machines are not essential for manufacturing LFTs during the proof of concept phase. We recommend that low resource laboratories that are developing paper sensors need not invest in striping/spraying equipment prior to proof of concept. This helps to reduce the financial risk associated with the development of any new sensing device.

One of the biggest drawbacks associated with the use of test and control spots is the development of "crest-shaped" signals. Fully rounded spots are usually expected, however crests form as a consequence of two factors. The first is the coffee ring affect which is experienced by the test spot during the drying process. In physics, this ring pattern is left by a particle-containing-droplet once the liquid has evaporated. The pattern is due to the capillary flow induced across the drop as a result of differential evaporation rates. Liquid evaporating from the edge of a droplet is replenished by liquid from the interior. This flow towards the edge of the drop carries the dispersed particles in the drop to the edges. In the case of the LFT test spot, this implies that most of the antibody concentrates at the outer edges of the spot.

Another type of flow induced inside a droplet during evaporation is called Marangoni flow. If the Marangoni flow in the droplet is strong enough, it can help redistribute the particles back towards the center of the droplet. Water, the main constituent of the buffer used to dilute the test and control spot antibodies, has a weak Marangoni flow, exacerbating this problem further.

The second factor is related to the lateral movement of the bacteria up the test. As the bacteria move upward, they make contact first with the lower region of the coffee ring, and become bound. As more bacteria contact this region, a build-up of Ab-bacteria-Ab-gold forms. The greater the affinity the antibody has for the bacteria, the faster and denser this build up forms. As more bacteria reach this area, they either add to this build up, or are diverted away from the test spot completely, moving towards the outer edges of the LFT. As a result, some of the bacteria from the sample move up the LFT along the edges of the device, completely bypassing the test spot. As a result, only a dark, crest shaped test signal is observed since only the bottom of the coffee ring binds the gold labelled E. coli. A small amount of bacteria might reach other areas of the test spot, albeit to a small extent. This is evident by the light red appearance of other regions of the test spot. This uneven distribution of the signal makes it difficult to quantify spotted signals. Nonetheless, it was found that the coffee ring effect does not prevent the colour intensity of the crests from showing a dependence on the E. coli counts in the sample, the Ab concentration of the test and control spots, the Ab volume used, etc. This colour dependence is expected when the sandwich assay format is used. To quantify these dependencies and thereby quantify the influence that different fabrication parameters have on the test performance, grey scale measurements of the darkest region of the test and control spots were used. In this way, the design of the paper sensor can be optimised in a low cost manner. All the above techniques enable the development of proof of concept devices and significant optimisation of the device performance without having to make large financial investments in new equipment.

Conjugate preparation

The minimum amount of antibody required to stabilise the gold nanoparticles Based on the visible colour change reactions, the lowest concentration of antibody that was able to stablise the gold and maintain its red appearance after the addition of salt, is 0.03 mg/ml. Those solutions spiked with an antibody concentration of less than 0.03 mg/ml turned purple after the addition of salt. Those antibody solutions with a concentration above 0.03 mg/ml however, remained red. Absorbance spectroscopy confirms this. Gold nanoparticles usually exhibit a characteristic absorbance peak between a wavelength of 500 nm and 600 nm. When there are changes in the optical properties of the gold (such as when it turns purple due to the addition of salt), its characteristic absorbance peak disappears. Figure 1 shows that the absorbance peaks of those curves with antibody concentrations below 0.03 mg/ml are absent. This implies that when these Ab concentrations were used, the gold solutions turned purple. Only those curves with antibody concentrations above or equal to 0.03 mg/ml retain their red appearance and hence display absorbance peaks. This confirms that an antibody concentration of 0.03 mg/ml is the minimum concentration required to stabilise the gold solution. To compensate for any pipetting error when creating the conjugate, and to serve as a safety measure, an antibody concentration of 0.05 mg/ml, slightly higher than the minimum concentration required, will be used to develop the conjugate.

Optimisation of antibody to gold volume ratio

Figure 2 shows the RGB plot that was used to ascertain the amount of red and purple colour in each of the vials used for this study. Each vial contained a constant amount of antibody, and increasing volumes of gold. Once salt is added to each of these vials, only those solutions that are unstable change from red to purple. Figure 2 shows how using RGB values, the colour change in each of these vials can be quantified by indicating the red, blue and green colour intensities prevalent in each tube. In Figure 2, the top curve indicates the red colour intensity in the solutions, the intermediate curve indicates the blue colour intensity in the solution, and the bottom curve the green colour intensity. To indicate the development of any purple colour in these vials, the blue and red curves overlap each other. Based on Figure 2 antibody to gold volume ratios between 1 :1 and 1 :13 will enable the development of stable conjugates. This is evident by the higher red colour intensity (than blue and green) at these volume ratios. Hence the gold solution maintains its red appearance from volume ratios of 1 :1 to 1 :13.

When the gold volume in the conjugate increases to more than 13x that of the antibody, the conjugate becomes unstable. This is a result of there being insufficient antibody available to completely attach to all the gold nanoparticles in solution. The optimal Ab to gold volume ratio lies where the red curve is positioned well above the blue curve. At volume ratios of 1 :13 and above, these two curves intersect indicating that a colour change to purple occurs. At an Ab-gold volume ratio of 1 :1 , the red and blue curves are positioned well away from each other. This distance is steadily maintained until a volume ratio of 1 :10 is reached.

Between ratios of 1 : 10 and 1 :13 the curves start drawing closer together, hinting towards the start of instabilities in the gold solution. The optimal Ab:gold volume ratio is therefore selected as 1 :10, as by using this this ratio the maximum quantity of conjugate per volume of antibody can be produced, while still ensuring the stability of the conjugate. Hence, an Ab to gold volume ratio of 1 :10 will be used for the development of the conjugate.

Antibody labelling

As determined earlier, an antibody concentration of 0.05 mg/ml and an antibody to gold volume ratio of 1 : 10 was used to prepare the conjugate. Once the antibody was added to the gold, the two were contacted for 30 minutes to promote efficient labelling of the antibody by the gold. After 30 minutes, bovine serum albumin (7 mg/ml) was added, in a process referred to as blocking. The BSA was left to incubate with the gold-Ab solution in the refrigerator overnight. During the blocking process the BSA is free to bind any unoccupied sites on the gold nanoparticles. This process prevents the gold from binding non-specifically to the paper itself, to the test line antibodies, or to the E. coli in the sample. The volume of BSA added was equal to the volume of Ab used, as this amount has proven sufficient for this purpose. The BSA-gold-Ab solution was then centrifuged at 13000 g for 15 minutes in order to separate the labelled and unlabelled antibodies from each other. The pellet (the conjugate) that was obtained after centrifugation was re-dispersed into a BSA solution, but was re-dispersed into a volume that ensured a 10x increase in the gold concentration. To do this, the pellet was re-suspended into a volume that was approximately 10x lower than the original volume of gold used to manufacture the conjugate. This increases the optical density (OD) of the gold from OD 1 to OD 10. Using a conjugate with a higher optical density on the test will produce test signals that are darker, stronger and hence clearer.

To ensure the gold concentration is in fact increased in this manner, a UV VIS spectrophotometer was used to measure the absorbance of the conjugate before and after the gold concentration was increased. The difference in the absorbency between the two indicates the extent of concentration. The concentrated gold (±OD 10) is first diluted by a factor of 100 before being placed into the spectrophotometer. As a result the absorbance value obtained must be multiplied by 100 to determine the actual absorbency of the concentrated gold. The absorbency of phosphate buffer and of the supernatant obtained after centrifuging was also measured. The results are shown in Table 2.

Table 2: Absorbance values for gold at OD 1 , OD 10 and the supernatant and phosphate buffer

Sample Absorbance

OD 1 1. 094

OD 10(1 OOx diluted) 0.095

Supernatant 0.031

Phosphate buffer 0.0365 The absorbance value of the conjugate increases from 1.094 (at OD1) to 9.5 (0.095 x 100, at OD10). This indicates an approximate 10x increase in the optical density of the gold. These results prove that by re-dispersing the conjugate pellet into a 10x lower volume, a 10x increase in the gold concentration can be obtained. The absorbency of the supernatant (removed from the tube after centrifuging) and that of the plain phosphate buffer is the same. This indicates that little to no gold is lost to the supernatant during the centrifugation process, and that all the gold was in fact used in the preparation of the final conjugate.

Determination of the optimum control spot antibody concentration

According to Figure 3, the intensity of the control spot signal increases as the Ab concentration increases. This is observed by the decreasing grey scale value of the signals with an increase in Ab concentration. At a concentration of 1 mg/ml, the signal is at its clearest and darkest. It is assumed that the signal intensity can be improved further by increasing the concentration of Ab above 1 mg/ml. However, at this time, the maximum concentration at which the control spot Ab is supplied is 1 mg/ml. As expected, the cost of the control spot reagent and hence the LFT rises as the concentration of the antibody in the control spot increases. Below a concentration of 0.6 mg/ml, the control signal is almost invisible to the naked eye and can therefore not be used. Hence, increasing the control spot concentration from 0.2 to 0.6 mg/ml is justified, regardless of the 33 cents increase in antibody cost. At a concentration of 0.8 mg/ml, the control signal is clear and visible. However the signal intensity is less than that obtained when an Ab concentration of 1 mg/ml is used. This is evident by the 12.2 unit decrease in the grey scale value as the concentration of Ab increases from 0.8 mg/ml to 1 mg/ml. A device manufactured using a control spot (CS) concentration of 1 mg/ml will cost 17 cents more than a device made using an Ab concentration of 0.8 mg/ml. Considering the fact that the final device should cost ±R10.00 (10 South African Rands), using an Ab concentration of 1 mg/ml rather than 0.8 mg/ml will contribute only 1.7% more to the final cost of the device. An antibody concentration of 1 mg/ml will produce stronger signals, reducing any ambiguities or misinterpretation of the test results. This helps improve the reliability of the test. Hence a control spot antibody concentration of 1 mg/ml is justified, and was selected for the development of the LFT.

Determination of the optimum test spot antibody concentration

Test signals were observed over the entire range of test spot concentrations investigated (0.5 mg/ml to 4 mg/ml). The colour intensity of the test spots, as judged by eye, remained seemingly constant between antibody concentrations of 4 and 2.5 mg/ml. The grey scale values of these signals fluctuate within a mere 1-3.5% of each other (Figure 4). This indicates that when using an Ab concentration of 2.5 mg/ml in the test spot, the same signal intensity as when using 4 mg/ml is achievable. Using an Ab concentration of 2.5 mg/ml will also reduce the cost of the test (by ±R1.39 each). The grey scale value of the test signals increase by ±13% when the Ab concentration decreases below 2.5 mg/ml. This implies that the test signal becomes weaker. However, to the naked eye, these signals are still clearly visible. Therefore, an Ab concentration below 2.5 mg/ml, say 1.5 mg/ml, can be used on the lateral flow test. Doing this would help reduce the cost per test by approximately 92 cents. However, according to Figure 4, the grey scale values begin to fluctuate significantly when the concentration of the Ab is reduced below 2.5 mg/ml. This hints towards the introduction of some type of instability on the device, which could eventually affect the reproducibility of the test signals. Therefore, while using antibody concentrations below 2.5 mg/ml can lower the cost of the device, it may reduce the quality and reliability of the tests. Hence a test spot Ab concentration of 2.5 mg/ml was selected for the development of this device as it provides the correct balance between cost and performance.

Determination of optimal conjugate volume

This study was performed in order to determine the optimal volume of conjugate that should be used on the LFTs. In Figure 5, the grey scale value of the test signals fluctuate as the volume of the conjugate used on the tests increases. These values fluctuate within 2.5%-10% of each other. As a result of this fluctuation, the influence of increased conjugate volumes on the intensity of the test signals is difficult to determine. Fitting a polynomial trend line to the grey scale value curve however, does help in identifying a trend. The trend line indicates an overall decrease in the grey values with an increase in conjugate volume. Hence, the test spot signal intensity increases as the amount of conjugate used on the LFT increases. This indicates that by applying more labelled antibodies onto the test, more gold-labelled E. coli reach the test line, producing stronger test signals. The trend line slope remains more or less constant as the conjugate volume increases to 7 μΙ. The trend-line becomes more linear between 7 μΙ and 10 μΙ of conjugate. This implies that the intensity of the test spot reaches its peak at 7 μΙ, and using any more conjugate on the test will do little to improve the intensity of the signal and will increase the cost of the device by 45 cents. However, by increasing the volume of the conjugate above 7 μΙ, the intensity of the control spot could be improved. This is due to more labelled antibody becoming available to bind to this region.

To test this theory, LFTs having both test and control spots were made, but loaded with varying amounts of conjugate. Conjugate volumes were varied from 5 μΙ to 10 μΙ, and tests were run using E. coli suspensions of 10 ⁸ , 10 ⁷ and 10 ⁶ cfu/ml. When 5 μΙ of conjugate is used, faint to no control spots are observed across all the bacterial concentrations investigated. When 6 μΙ to 10 μΙ of conjugate is used however, control spots (of similar colour intensity) form for all the bacterial concentrations. Hence to form clear control spots, a conjugate volume of 6 μΙ should be used.

Increasing the conjugate volume from 6 μΙ to 7ul however, can only be warranted if it results in an improvement in the detection limit of the device. This is because the intensity of the test spots that form when 6 μΙ and 7 μΙ of conjugate is used (based on the trend line), are not significantly different. By eye, this difference is insignificant. Hence, if the increased conjugate volume does not improve the colour intensity of the test spot, nor improve the detection limit of the device, the increased cost of using 7 μΙ of conjugate per test cannot be justified. Testing of different types of membranes on the LFT

Different types of nitrocellulose membranes were applied on the lateral flow tests. Test spots were deposited onto the membranes which were later assembled and run using spiked E. coli samples. The speed of testing was monitored. This included the time required for the sample to wick to the opposite end of the strip, and the time it takes for a clear signal to develop.

Most importantly, the intensity of the test spot was examined. All paper types, excluding Whatman's fusion five, produced test spots. Some paper types developed fully rounded test spots, while some produced crests rather than spots. Those papers that produced fully rounded test spots included CN140, FF60/100 HF075, AE99 and HF090. The most striking difference between these papers and those that produced crests, is its wicking speed. The wicking speeds of these papers ranged from 0.025 to 0.06cm/s. The membranes that produced crests were 125, FP, 240 and AE98 and have wicking speeds that range from 0.017 to 0.036 cm/s. It is possible that papers with a higher wicking speed (and hence stronger capillary force) are able to draw the E. co//-Ab-gold complex across the bottom edge of the test spot faster, reduces its contact time with these Abs and thereby limited the chance of binding occurring at this point. Furthermore, due to the coffee ring effect experienced during drying, the antibodies are more concentrated at the edges of the spots. It could be that membranes with slower wicking speeds allow more contact time between the E. coli- Ab-gold complex and the lower edge of the spot. This increased contact time could be just long enough to permanently bind the E. coli in place in this area. Hence, the formation of a crest shaped signal. Furthermore, the crest shaped lines have a much darker intensity than the fully rounded spots. This can be explained by the fact that the gold nanoparticles are, in the case of the crest shaped lines, concentrating in a much smaller area as compared to those in the fully rounded spots. As a result of these tests it has been decided that membranes with a larger pore size will be used for the development of this device. These types of papers allow the test to run faster without compromising its sensitivity, ensures that the 2 Mm E. coli bacteria will travel through the paper without getting trapped or stuck, and offer a greater chance of developing fully rounded test spots.

Background staining is another important aspect to consider when selecting a membrane. Two types of staining occur, overall staining of the membrane by gold nanoparticles, and non-specific deposition of gold in the test spot. To prevent such staining, the gold nanoparticles and the test membranes were adequately blocked. However, it was found that the type of membrane used on the LFT influences the extent of staining. Each type of paper undergoes a specific type of treatment. Hence each paper type contains varying amounts and types of surfactants and other chemicals. As a result, properties such as electrostatic charge, wettability, etc., differ for each membrane.

Whatman's Prima 125 membrane was found most optimal for this application. It produces the least background noise, possesses a fast wicking speed and produces fully rounded test spots. However, it was later found that this membrane is no longer being manufactured. A replacement membrane (HF80HP from GE Health) was introduced and is considered the upgraded version of Prima 125. This membrane was purchased, and found optimal for the particular application. Optimal testing time

The grey scale values of the test spots that formed after allowing several LFTs to lay in their sample and run for 30 seconds to 30 minutes were determined (data not shown). The grey scale values were found to decrease from 142 to 110 units as the testing time increased from 0.5 to 8 minutes. This indicates an increase in the test spot colour intensity. The test spot on the LFT tested at 8 minutes appears smaller in size than the test spots on the other LFTs. As a result, the test spot colour intensity on this LFT is higher (having a lower grey scale value). The 8-minute point is therefore neglected in this analysis since it cannot be fairly compared to the other test signal intensities. Between testing times of 6 and 16 minutes, the grey scale values average at around 119, fluctuating by ±10 grey scale units. The grey scale value of the test spot reaches its lowest value, and hence greatest intensity after the LFT was run for 18 and 20 minutes. The LFTs that were run for longer than 20 minutes produced weaker test signals, the grey scale values of which reached similar levels to that obtained at 6 to 16 minutes. This could have occurred as a result of evaporation or backflow due to the longer running times. Hence, a running time of 20 minutes appears optimal.

Optimal drying time

This study was performed to determine how long after running the test, should the results be viewed and interpreted. If read too soon, the test spot on the lateral flow test may be unclear (or underdeveloped) as the Abs and sample analyte may have had insufficient time to interact. Alternatively, tests should not be read too long after being run, as sample backflow can become problematic. Test spot intensity was found to increase after 2 minutes of drying. This is indicated by a decrease in the grey scale value from 218.6 at 0.5 minutes to 206 at 2 minutes. Thereafter, the grey value remains steady, fluctuating between 206 and 201 grey scale units, until a drying time of 20 minutes. At 22 minutes, the signal intensity increases sharply, as depicted by a sudden decrease in the grey scale value. Between 22 and 30 minutes of drying, the grey scale value averages at approximately 94.5 units, the lowest grey scale value observed during this study. The intensity of the test signal reduces after an hour of drying, indicating that the LFTs should not be read more than an hour after it had been run. The weakest signal intensity was obtained after less than 2 minutes of drying. Hence, test results should not be read prior to 2 minutes after the LFT is removed from the sample. The strongest signal intensity was obtained after 20 and 30 minutes of drying. This signal however, had grey scale values that were only 6% lower than the grey scale values obtained after 2 minutes of drying. This difference in grey scale values cannot translate into a better visual signal as the difference is too small to be noticed by the naked eye. Hence for qualitative results, the test signal can be read 2 minutes after being removed its sample. When quantitation is required however, the test signal should only be read after 25 minutes of drying. In this way the true test signal intensity would be determined and can be compared to a calibration curve in order to gauge the analyte concentration in the sample. Test line position

This experiment was performed to determine how the distance between the conjugate pad and the test spot influences the intensity of the final test signal. The test spots were positioned to lay 4 mm, 7 mm, 12 mm and 13 mm away from the conjugate pad. By eye, the intensity of the test signals appear to increase as the distance between the test spot and the conjugate pad increases. The grey scale value curve confirms this. The grey values obtained when the test spots are positioned at 13 mm and 12 mm are lower than that obtained when the test spots are located at 7 mm and 4 mm. When the test spot is positioned further away from the conjugate pad, there is more time available for the bacteria and conjugate to interact and bind to each other, before becoming trapped at the test spot. This increased reaction time ensures that more E. coli becomes labelled and hence more labelled E. coli are trapped at the test spot. This produces a darker and stronger test signal. Hence, positioning the test line further away from the conjugate pad helps enhance the colour intensity of the test spots. However, considering the small size of the device and the need to make room for a control spot on the membrane, there is a limitation to how far the test spot can be positioned from the conjugate pad. Nonetheless, this information helps make developers aware that maximising the distance between the conjugate and the test spot can help improve test signal intensities.

Effect an increased test spot and conjugate antibody concentration and an increased conjugate volume has on the device detection limit

Grey scale analysis performed during the determination of the optimal test spots Ab concentration indicates that the signal intensities of test spots can be enhanced by increasing the test spot antibody concentration. However, while grey scale values indicate that the test spots formed at 4 mg/ml are darker than those formed at 2.5 mg/ml, by eye, the difference goes unoticed. Hence both signals appear equally visible to the naked eye. The use of an Ab concentration higher than 2.5 mg/ml was therefore not justified, since the benefit does not outweigh the increased cost of the sensor. However, should the increased test spot concentration improve the detection limit of the LFTs, by even a single fold, its use will be justified. Earlier it was shown that increasing the conjugate volume on the LFTs increases the test spot signal intensity. However, using a higher volume of conjugate will, like in the case of increasing the test spot Ab concentration, only be justified if it can translate into an improved detection limit.

LFTs were manufactured with an increased conjugate volume and test spot antibody concentration in order to determine the influence this will have on the detection limit of the device. The antibody concentration in the conjugate was also increased in order to determine if this too will have any effect on the sensitivity of the device. The sample concentration tested on the LFT started at 10 ⁸ cfu/ml, and decreased in 10 ¹ cfu/ml increments to 10° cfu/ml. A negative control test was run using only phosphate buffer. It was determined that when a higher test spot concentration was used, the detection limit of the device remained at 10 ⁶ cfu/ml. This is the same detection limit obtained when a test spot concentration of 2.5 mg/ml was used. Hence using a higher concentration of test line Ab is not justified even though the signal intensity itself is improved.

LFTs manufactured with an increased Ab concentration in the conjugate were also tested where the sample concentration used on each LFT increased from 10 ³ cfu/ml to 10 ⁸ cfu/ml. A negative control test was run with phosphate buffer. These tests were run with heat killed E. coli 0157:H7 and the detection limit found to be 10 ⁵ cfu/ml. Hence, no change in the detection limit occurs with an increase in conjugate volume. The reason for the lowered detection limit of LFTs run with heat killed E. coli compared to live E. coli, is the increased affinity the antibodies on the test have for the heat killed organisms. The animals used to produce the antibodies used on the LFT are actually immunised with the heat killed E. coli. As a result, these antibodies were produced in response to the heat killed organisms and will therefore always bind these organisms more sensitively.

Tests were also run with an increased conjugate volume. The sample concentration was decreased from 10 ⁸ cfu/ml to 10° cfu/ml. Negative control tests were run with phosphate buffer.

The LFTs manufactured using an increased conjugate volume showed no improvement in detection limit, even though an improvement in the test spot intensity occurs. In a final attempt to improve the detection limit of the sensor, LFTs were made with a combination of an increased conjugate volume and antibody concentration and an increased test spot Ab concentration. These LFTs also demonstrated no improvement in the detection limit. Hence, it appears that the sensor has been fully optimised and nothing more can be done to improve the sensitivity of the sensor. An ELISA was done on the antibodies used in both the conjugate and test spot of the sensor. Both Abs were found to have a detection limit of 10 ⁶ cfu/ml. It can therefore be concluded that so long as this particular Ab pair is used, and is used in the standard LFT format with gold nanoparticles, the detection limit cannot be changed.

Conclusion

A low cost method for the fabrication of paper based lateral flow sensors has been demonstrated. It is believed that sensors developed using this method serve well for the demonstration of proof of concept. Using initial optimisation studies, covered in this example, the eventual performance of the sensor can be gauged. It was found that increasing the concentration and volumes of bio-reagent used on the LFTs indefinitely, does not translate into improved sensor performance. Optimal values for these parameters are those that strike the correct balance between performance and fabrication cost.

EXAMPLE 2

Detecting E. coli using lateral flow and paper based microfluidic technology: Assessment of the lateral flow test performance

Sensitivity and specificity

Specificity is a statistical estimate of the ability of the test to identify a true negative sample. A specific test will not produce a positive test signal in the presence of bacteria other than its target, which in this case, is E. coli. Specificity therefore provides an indication of the cross reactivity of the test. Sensitivity is a statistical estimate of the ability of the test to accurately identify true positive samples. Therefore, a sensitive test will always produce positive test results whenever target bacteria are present and are within the detection limits of the test. The sensitivity and specificity are calculated as follows:

Sensitivity = True Positives/(True positives + False Negatives) x 100%

Specificity = True Negatives /(True Negatives + False Positives) x 100%

For specificity analysis of the lateral flow test, six different types of bacteria were used. These bacteria were kindly provided by the Natural Resources and Environment (NRE) unit at the CSIR and include: E. coli K12, E. coli 0157:H7, Enterobacter cloacae, Pseudomonas aueruginosa, Salmonella enteritidis and Acinetobacter.

Each type of bacteria was suspended in phosphate buffer to create a 10 ⁸ cfu/ml solution and then tested on lateral flow tests, which were monitored for the development of a positive test signal. A total of 120 tests were performed. To determine the detection limit of the test, serial dilution was used to generate samples having E. coli 0157:H7 concentrations ranging from 10 ⁸ to 10° cfu/ml. These samples were then tested on the LFTs. The lowest bacterial concentration that produces a positive test signal is considered the detection limit of the device. To determine the test sensitivity, E. coli 0157:H7 and K12 suspensions were used at varying sample concentrations. A total of 300 tests were performed.

Detection of dead bacteria

To determine if dead bacteria can be detected on the LFTs, E. coli K 2 and 0157:H7 samples, suspended at 10 ⁸ cfu/ml, were doused with varying amounts of chlorine to kill the bacteria, and then tested on the LFT. As liquid chlorine was unavailable, household bleach was used for these experiments. Bleach contains 2% sodium hypochlorite, which translates into 20000 ppm of available chlorine. To vary the chlorine concentrations to which the bacteria were exposed, equation (1) was used.

C = C ₂ V ₂ (1 )

g. 20000 p

Therefore, to expose the bacteria to a chlorine concentration of 8 ppm, 0.5 μΙ of bleach is added to 1249.5 μΙ of the E. coli sample. In this way, bacterial suspensions were exposed to chlorine concentrations ranging from 2 ppm to 1000 ppm, and in some cases even higher. The bacteria were contacted with the different chlorine concentrations for 30 minutes to render them dead. To confirm that the bacteria were in fact killed, the chlorine doused bacteria were cultured in nutritive media overnight at 37°C. The absence of growth after 24 hours confirmed that the bacteria were dead. Various contact times were also investigated. This was done to determine how long the bacteria and chlorine need to be incubated together to ensure the bacteria are killed. Contact times investigated were varied from 1 minute to 60 minutes, following which each suspension was grown on nutritive media overnight. Those chlorine concentrations and contact times that proved effective in killing the bacteria were then used to kill the bacteria used to perform this study. To perform these tests, 200 μΙ of the chlorine-E coli solutions was tested on the LFT.

The aim of this investigation is to firstly identify whether the LFTs detect dead bacteria, i.e. to determine whether the LFTs are able to distinguish dead from live bacteria. Secondly, should the LFTs be able to distinguish dead from live bacteria, the aim is to determine the chlorine dosage at which this occurs. This concentration can then be compared to that currently used for disinfection in the water treatment industry. This will help determine whether only live bacteria will be detected when the LFT is used to analyse such environmental samples.

Movement of bacteria through the paper matrix

This study was performed to better understand how bacteria migrate from the sample pad to the test spot of the lateral flow test. The different types of paper making up the LFT have complex structures made up of pores and elongated cellulose fibers, and differ from each other in various ways. The chance of bacteria become trapped in this complex matrix is high, however the extent to which they become trapped is unknown. The information from this study will help provide insight into whether all of the bacteria contained in the sample actually reach the test spot and undergo detection. If many bacteria become trapped prior to reaching the test spot, fewer bacteria become available for detection. This can result in the reported detection limit of the test being poorer than what it actually is. For example, take the case of an LFT producing a signal only when a minimum of 10 ⁵ cfus is contained in the sample. Assuming bacteria losses are significant and only 10 ³ cfus reach the test spot, the actual detection limit of the test is 10 ³ cfus and not 10 ^s cfus, as will be reported. Ideally, when 10 ³ cfus are loaded into the sample, the entire 10 ³ cfus will undergo detection and indicate the true potential of the sensor being analysed. Characterising the extent of bacteria trapping in paper will therefore help developers of paper sensors better understand analyte losses in a device, enabling them to construct or design their devices in such a way so as to minimise such losses.

Determining the exact number of bacteria trapped in the lower regions of the LFT once sample movement has ceased is somewhat difficult. This is mostly due to the complex structure of paper. One method is to release the trapped bacteria from the different regions of the paper into buffer through elution, culture the elution buffer in nutritive agar and count the number of bacteria that grows. This method will however, only give an idea of the amount of bacteria trapped in the various regions of the test. To determine the exact number is difficult as some bacteria may not be released from the paper. Another technique that may indicate the extent of trapping is to stain or colour the E. coli (either with a fluorescent or colorimetric label) and then flush this through the lateral flow test. In this way, the trapping may be visualised. However, even this method is only qualitative. A disadvantage of the visual analysis is that signals from labelled bacteria trapped on the bottom surface of the paper and on the adhesive backing, may be difficult to see and are hence lost. This is because visual colorimetric signals only arise from the top few microns of the paper surface. Hence, determining the exact number of bacteria entrapped in the paper is difficult. However, using these techniques, it may be possible to obtain an idea of the number of bacteria becoming entrapped in the LFT and determine the effect this has on sensor performance. In this way those regions of the sensor resulting in the most trapping can be identified so that steps can be taken to reduce this. This information is useful especially of late, when the paper sensing industry is receiving much attention. To date, we have found no information illustrating the extent of analyte trapping in paper sensors, and the effect this has on the performance of sensor. This information contributes to the development of high quality sensors.

For the elution studies, LFTs were run with E. coli K12 suspensions having concentrations of 10 ⁸ and10 ³ cfu/ml. Once the LFTs were run, its different sections were aseptically detached from each other. For the 10 ³ cfu/ml test, the sample pad, conjugate pad, test spot, control spot and wick were used. For the 10 ⁸ cfu/ml test, the same sections were used, besides the wick. The separated sections were placed into separate glass test tubes containing 600 μΙ of PB and vortexed gently to promote the release of the bacteria from the paper into the buffer. Once this elution step was complete, the buffer was inoculated into petri dishes filled with nutritive media. Since a single plate can only be inoculated with 100 μΙ, 6 petri dishes were used per section of the LFT in order to culture the entire 600 μΙ of elution buffer. To determine the total number of colony forming units released from each section of the LFT, the colonies that form in all 6 plates were added.

To confirm that E. coli was in fact present in the samples used for this analysis, 100 μΙ of the original sample was cultured and demonstrated positive growth. To serve as a negative control, phosphate buffer (with no E. coli) was run on a lateral flow test. This LFT was also separated into its constituent sections which were passed through the elution process, after which the elution buffer was cultured. As expected, no growth was observed.

For all the aforementioned studies, £. coli K12 suspensions (10 ^s and 10 ³ cfu/ml) were run on a lateral flow device loaded with anti-E. co//-0157:H7 antibodies in the conjugate and anti-E co//-"all O and K" antibodies on the test line. Since the conjugate Abs are specific for the 0157:H7 strain, they would not bind to E. coli K12. The test spot (TS) Abs detect a wider range of E. coli and are therefore likely to bind the E. coli at the test spot. However, since the E. coli is unlabelled, a test spot signal will not be observed. The possibility of E. coli binding at the test spot must be kept in mind when accounting for bacteria during the elution studies.

For the colorimetric staining studies, E. coli K12 was cultured on McConkey agar. This agar grows the E. coli as red colonies. The red E. coli colonies were suspended in phosphate buffer to a concentration of approximately 10 ⁷ cfu/ml and were run on the lateral flow test. The red bacteria were monitored as it moved up the length of the test strip.

E. coli 0157:H7 (10 ⁵ cfu/ml) was also fluorescently labelled using anti-E coli antibodies tagged with polyacrylonitrile chromeon 470 nanoparticles (Sigma Aldrich). These fluorescently labelled bacteria were allowed to flow up a half-strip lateral flow test. Half strip tests are assembled with only a nitrocellulose membrane (loaded with test and control spots) attached to an absorbent wick, both of which are assembled together on an adhesive backing card. Conjugate and sample pads are not included. To visualise the fluorescence after running the half strips, the strips were dried for 10 minutes in a 37°C oven, placed on an inverted fluorescence microscope and excited using wavelength of 470 nm. The dye emits at a wavelength of 611 nm, which was visualised on the microscope. Using all of the above techniques, a trend depicting the extent of bacteria trapping on a lateral flow test was obtained.

Flow-through tests

Flow-through tests are another form of rapid point of care tests. In this testing format, a section of nitrocellulose membrane is loaded with test and control spots and then placed on top of a piece of absorbent pad. The sample fluid is loaded onto the membrane, contacts the test and control spots, then continues to flow downward into the absorbent pad. Flow-through tests therefore have no sample and conjugate pads, implying that the trapping of bacteria in these regions cannot influence the performance of the sensor.

A flow through test was done to identify the influence that bacteria trapping has on the performance of the lateral flow tests. To manufacture the flow through test, a 1 cm by 1 cm unbacked nitrocellulose membrane was used. Rabbit anti E. coli antibody (ABD serotec) was diluted to 2.5 mg/ml and loaded onto the membrane, serving as the test spot. The test spot was dried in place for 30 minutes at 37°C, blocked using a membrane blocking solution and then dried for 1 hour at 37°C. 1 cm by 1 cm pieces of absorbent pad were prepared. To assemble the flow-through test, the absorbent pad was positioned directly behind the membrane, and sandwiched between 2 large paper clips. The paper clips are elevated to prevent the absorbent pad from making contact with any other surface that may influence the performance of the sensor.

The samples were prepared as follows. E. coli 0 57:H7 suspensions were made, ranging in concentration from 10 ⁸ cfu/ml to 10 cfu/ml. A volume of 200 μΙ of each dilution was added into glass test tubes, into which 6 μΙ of gold conjugate was added and allowed to contact for 7 minutes. The detection limit of these flow through tests were examined and are discussed in Example 2. It must be kept in mind however, that the membrane used for this study was different to the membrane used on the lateral flow tests. This is because the lateral flow test membrane is backed and is unsuitable for use in a flow through system.

Scanning electron microscopy analysis

To further understand how bacteria move through the paper matrix of lateral flow tests, scanning electron microscopic (SEM) images of the different sections of the LFT were taken. LFTs were run with heat killed E. coli 0157:H7 samples (10 ⁷ cfu/ml) and with non-spiked buffer (this served as a negative control). Once the test had run, the LFTs were aseptically separated into their constituent parts. These include the sample pad, conjugate pad, test spot region and control spot region. Each section was sputter coated with gold to improve its visibility under the SEM and then analysed. Besides just examining the bacteria present on the different sections of the LFT, the structural and morphological characteristics of the different types of paper used to create the LFT were also analysed. The aim of this work was to confirm whether or not bacteria do remain left behind in the lower regions of the LFT once the test is run..

Detection of a single E. coli organism on the lateral flow test

According to the World Health Organisation (WHO) and the South African National Standards (SANS 241 , 2011), there should be no E. coli present in drinking water. As a result, in order to be applicable for drinking water analysis, E. coli detection systems must be able to detect as little as 1 cfu in a 100 ml water sample. For waste water analysis however, this can change. The level of microbes allowed in the effluent leaving waste water treatment facilities is controlled by government legislation, and is often specific for each waste water treatment facility. The acceptable levels are usually guided by the location of the plant, and the water body into which the effluent enters. The level of E. coli generally allowed in waste water effluent is 1000 bacteria per 100 ml of water. This implies that for waste water analysis, E. coli detection systems must have the ability to detect at least 1000 E. coli per 100 ml sample. However, most systems employed for waste water analysis still have the ability to detect a single organism.

For this study, the ability of the lateral flow test to detect a single E. coli bacterium was investigated. The detection limit of the LFT was found to be 10 ⁶ cfu/ml. On its own, the LFT can therefore only be applied for water analysis during emergency outbreaks, when large amounts of bacteria are prevalent. It was decided that by incorporating a pre-enrichment step, the LFT could potentially detect a single E. coli. During pre-enrichment, bacteria are placed into nutrient rich media, exposed to temperatures favourable for growth and thereby cultivated to levels that are detectable on the LFT (10 ⁶ cfu/ml). Due to the incorporation of this step however, the duration of the testing procedure becomes longer. However, if the time required to detect a single bacterium is less than that required during conventional E. coli detection methods, the test will still be advantageous to the water monitoring industry. A faster time to result implies that information regarding the safety of drinking water sources is obtained earlier, helping to prevent the spread of disease. To determine whether pre-enrichment does enable the LFT to detect one E. coli, the following was done.

A 2.7 x 10 ^s cfu/ml E. coli 0157:H7 suspension was serially diluted to concentrations of 2.7 x 10 ² cfu/ml, 2.7 x 10 ¹ cfu/ml and 2.7 x 10° cfu/ml. 1 ml of each of the three suspensions was inoculated into 11 petri dishes each of which contained 5 ml of non-selective nutrient broth. The petri dishes were then incubated at 37°C for 11 hours. After every 1 hour of incubation, 1 petri dish for each concentration was tested using 3 lateral flow tests. This was done until the LFT developed a positive test signal. Three additional petri dishes containing 5 ml of broth were inoculated with 1 ml of each sample concentration, and tested without having gone any incubation. This served as an experimental control. None of these LFTs produced positive test signals.

To ensure that the 10 ² , 10 and 10° cfu/ml samples did actually contain these amounts of bacteria, 1 ml of each sample was cultured in nutrient agar. As 1 petri dish can only be inoculated with 100 μΙ, 10 plates were used per sample. Therefore, in order to determine the total number of colonies per ml of sample (for each concentration), the colonies that form in all 10 plates are added. .

Analyses of environmental water samples

To determine the ability of the lateral flow test to handle environmental water samples, effluent samples were collected from a local waste water treatment facility (Tshwane Water) and analysed. As the E. coli content of these samples are low (less than 100 cfu per 100 ml of sample), the pre-enrichment step was included during this analysis.

To ensure that the effluent sample flows through the lateral flow test and produces a test spot should E. coli be present, effluent samples were spiked with E coli 0157:H7 and tested. E. coli suspensions with concentrations ranging from ±10 ⁸ cfu/ml to 10° cfu/ml were prepared. This study was done to ensure that the effluent flows through the paper without clogging it and that any chemical and/or biological compounds in the effluent do not hamper signal development or cause high background noise.

To ensure that the strains and serotypes of E. coli in the effluent can be detected on the LFT, a sample from the plant inlet was obtained. The inlet will contain all the different types of E. coli that could possibly occur in the treated effluent, albeit at a much higher concentration. In this Example, the LFTs were only tested using 2 strains of E. coli, 0157:H7 and K12. Example 4 sets out whether common environmental strains of E coli could be detected on the LFT. There is little likelihood of there being large amounts of E. coli 0 57:H7 (a strain normally found in food), or E coli K12 (a lab strain) in the inlet sample. Therefore, any positive test signals that do arise would be due to the detection of strains other than these two.

The combined pre-enrichment and LFT testing procedure was used to analyse the effluent sample. As is standard in the water industry, 100 ml effluent samples were tested. Twenty four, 100 ml effluent samples were filtered through separate 0.45 Mm cellulose filters. Each filter was placed into its own petri dish filled with nutrient broth and incubated over several hours. During incubation, the broth was tested on a LFT every hour until a positive test signal was obtained. 12 filters were placed in general nutrient media, and 12 were placed in coliform and E. coli selective media.

Comparison with other commercial kits: sensitivity, specificity, dead bacteria A performance comparison between the lateral flow test of the present invention and 3 other commercially available E. coli detection systems was donE. coliled (IDEXX Laboratories, 2011), Microsnap (Hygiena Inc, 2013) and Cell Biolab's Rapid test kit for E co// 0157:H7 (Cell Biolabs Inc, 2011) was evaluated with respect to the following: (1) ability to distinguish dead from live bacteria, (2) detection limit/sensitivity and (3) specificity/cross reactivity

This analysis was done to determine how well the developed LFT performs in relation to systems that have undergone market validation. Some of these systems have in fact been approved for the detection of E. coli in water sources globally. The information gathered from this investigation will provide an indication of how marketable the lateral flow test is, and what aspects of the test still require improvement.

Colilert and Microsnap rely on the ability of two enzymes, β-glucuronidase (found in E. coli) and β-galactosidase (found in all coliforms) to cleave specific substrates that release chromogenic and fluorogenic compounds. Consequently, colour changes are used to indicate the presence of these bacteria. The Cell Biolabs lateral flow test, like the developed lateral flow test, employs antibody-antigen binding for detection.

For the Colilert tests, 100 ml samples of Pseudomonas aueruginosa, Enterobacter cloacae, Salmonella enteritidis and E. coli 0157:H7 was prepared to a concentration of 1.5 x 10 ¹ cfu/100 ml. The Colilert substrate was dissolved into each 100 ml sample and then aseptically transferred into Colilert quanti-trays. Following 18-24 hours of incubation at 37°C, the results were read. A similar procedure was followed when testing dead bacteria samples.

To determine the detection limit of Cell Biolab's lateral flow tests for E. coli 0157:H7, suspensions of this bacteria were made up to concentrations ranging from 10 ^s cfu/ml to 10° cfu/ml. 150 pi of each sample was diluted with 150 μΙ of the manufacturers buffer. The samples were then ready for use. A similar procedure was followed when testing dead bacteria samples. When used with samples having a low bacteria concentration, the samples were first filtered to capture the bacteria on a membrane which is then incubated for the prescribed 18-24 hours. 150 μΙ of the enriched sample is then extracted and diluted in 150 μΙ of buffer before being run on the test strips. When these LFTs were tested for specificity, non-E coli 0157:H7 samples were made up to suspensions of 10 ⁷ cfu/ml. 150 μΙ of these suspensions were diluted in buffer before being run on the LFTs.

To analyse the Microsnap system, 10 ⁴ cfu/100 ml bacterial suspensions of Pseudomonas aueruginosa, Enterobacter cloacae, Salmonella enteritidis and E. coli K12 were prepared. 00 ml of each suspension was filtered through a 0.45 pm filter which was incubated in the nutrient broth supplied by the manufacture for 8 hours. Thereafter, 200 μΙ of each enriched sample was transferred to individual Microsnap tests, incubated for 10 minutes at 37°C (as per the manufactures instructions) and then the results were read using the provided Ensure Luminometer (reader). The same procedure was followed when dead bacteria samples were analysed.

Results and discussion

Sensitivity and specificity

The detection limit of the lateral flow test was determined to be 10 ⁶ cfu/ml. A review by Peruski et al. (2003) states that lateral flow tests for microorganisms, based on immunoassay technology, have detection limits ranging from 10 ⁸ to 10 ⁵ cfu/ml. This indicates that the LFT is performing within acceptable limits.

For specificity analysis, the bacteria listed above were tested on the LFTs. The production of a positive test signal on the LFT indicated cross reactivity between the antibodies and the non-target bacteria. To determine the sensitivity of the lateral flow test, E. coli K12 and E. coli 0157:H7 suspensions were tested. The number of false negative results obtained was used as an indication of the tests sensitivity. The results of this analysis are shown in Table 3. Table 3: Sensitivity and specificity testing results

Table 4: Summary of Sensitivity and specificity results

TP/(TP+FN) x 100

300/(300+0) x 100

100%

Specificity = TN/(TN+FP) x 100

= 80/(80+40) x 100

= 67%

These results indicate that the lateral flow test offers 100% sensitivity when the E coli count in the sample is above its detection limit of 10 ⁶ cfu/ml. At the time of this analysis only two serotypes of E. coli, K12 and 0157:H7 were available for testing. The antibodies used on the lateral flow test are able to detect all O and K strains of E. coli. These are the most common strains of E. coli that exist in the environment. All known types of E. coli to date, contain the O-antigen. This implies that the LFT could potentially detect all types of E coli, depending on the ability of the antibody to detect both the long and short length O-antigen chains. As many of these E. coli types therefore need to be tested on the LFT in order to more realistically predict the true sensitivity of the device. This is important as it would indicate the extent of sensitivity that can be expected when testing environmental water samples.

A specifcity result of 67% was obtained and is mostly due to the non-specific interaction of the LFT with the bacterium, Pseudomonas aueruginosa. Cross reactivity was observed with Salmonella enteritidis, occurring in 33.3% of the tests performed. This may be due to there being fewer antigenic binding sites on the bacterium which reduces the chance of it binding each time, or the epitopes are becoming hidden in some instances. No other cross reactivity was observed. To ensure that most strains of E. coli are detected by the LFT, polyclonal antibodies were employed for sensing. Due to their inherent nature, polyclonal antibodies tend to display cross reactivity, and according to the manufacturer, this particular antibody is known to cross react with other enterobacteriaceae. This is probably due to these bacteria sharing common antigenic sites with E. coli, and serves as an explanation for the cross reactivity with Salmonella enteritidis (an enterobacteriaceae). It is known that anti-E. coli antibodies can cross react with Pseudomonas aeruginosa due to the common surface antigens shared between these two bacteria. This could be the reason for the weak test signals obtained with Pseudomonas aeruginosa on the LFT. Cross reactivity with enterobacteriaceae may not negatively influence the tests performance, as a positive test signal will still indicate that the water source is unsafe. This is the main intent of this device.

A recent study by Luyt et al (2012) states that the specificity of the Colilert detection system is 92.8%, however Colilert can only detect 56.4% of all known E. coli strains. As mentioned by Guo et al (2005), the O-antigens which are targeted and used for detection by the LFT exists on most, if not all E. coli types. This implies that the developed LFT has a great potential to detect more than 56.4% of all E. coli found in the environment. To further examine the performance of the lateral flow test, it will undergo field trials at Tshwane Water (South Africa). During field trials, the performance of the lateral flow test, with respect to its sensitivity, specificity, and speed of analysis will be compared to conventional and latest E. coli detection methods.

Detection of dead bacteria on LFTs

At water treatment facilities, adequate disinfection is achieved when a chlorine dosage between 5 ppm to 7 ppm is used in conjunction with a contact time of 30 minutes. In the lab, E. coli solutions (10 ⁸ cfu/ml) were dosed with chlorine concentrations ranging from 1 ppm to 8.5 ppm, incubated for 60 minutes and then grown in nutrient media. No E. coli growth was observed on any of the plates, indicating that even at chlorine concentrations as low as 1 ppm, the bacteria are rendered metabolically inactive (dead). The original E. coli solution used for this study (not doused with chorine) was to confirm live bacteria were used in this study. These plates demonstrated positive growth.

To determine the contact time required, E. coli suspensions were doused with 5 ppm of chlorine and contacted for 2 to 60 minutes. A chlorine concentration of 5 ppm was selected as it is commonly used in water treatment plants and has been confirmed to render bacteria dead. . After culturing, all samples demonstrated no growth. This implies that at a chlorine concentration of 5 ppm, 2 minutes of contact time is sufficient to render bacteria dead. The positive control (the E. coli sample used for this study), demonstrated positive growth.

These parameters were therefore used to kill the E. coli used to evaluate the LFTs ability to distinguish live from dead E. coli. Dead bacteria could be in a viable but non culturable state (VBNC) or completely disintegrated. E. coli. This study was done by monitoring whether the LFT develops a positive test signal when dead bacteria are tested.

Live E. coli 0157:H7 samples were also doused with the following chlorine concentrations: 10 ppm, 1000 ppm, 2000 ppm, 2500 ppm and 3000 ppm. These concentrations render the bacteria non-culturable, however it is not known whether these concentrations are adequate to disintegrate the bacteria. For the antibodies on the LFT to not detect the E. coli, their surface antigens must be absent. This is because the antibodies used on the LFT target antigens or epitopes that lie on the external surface of the E. coli bacterium. Since it was determined that 1-8.5 ppm of chlorine renders the bacteria non-culturable, any concentration higher than this should do the same. These high concentrations of chlorine were used to identify a) the chlorine dosage at which the LFT no longer detects the E. coli (implying that the bacteria may have completely disintegrated), and b) how this dosage compares to that used in the water treatment industry.

After contacting the E. coli and chlorine for 30 minutes, the samples were tested on the LFT. None of the tests produced positive test signals, including the sample inoculated with 10 ppm of chlorine. While the test spot on the 10 ppm test demonstrated a pink haze, it cannot be regarded as a true test signal especially when compared to the dark signal intensity of the LFT used as a positive control (run with a non-chlorinated, E. coli rich sample). This is further observed by the high grey scale value of the 10 ppm test in contrast to the lower grey scale value of the positive control LFT (results not shown). No test spots were observed on the remaining LFTs.

These results indicate that chlorine concentrations at or below 10 ppm may cause the E. coli bacteria to rupture and not be detected on the LFT.

Next, live £ coli 0157:H7 samples (10 ^s cfu/ml) were then doused with 2 ppm to 30 ppm of chlorine. Strong positive test signals were observed for chlorine concentrations between 2 ppm and 10 ppm.. At 12 ppm, the intensity of the test signal significantly weakens, and remains that way until a chlorine concentration of 20 ppm. At chlorine concentrations of 30 ppm and higher, the test signal disappears completely. The decrease in signal intensity at 12 ppm indicates that at this concentration the bacteria may begin to disintegrate, providing an explanation for the weakened signal intensity.

Hence, the LFTs are only able to convincingly distinguish dead from live bacteria at chlorine concentrations above 30 ppm. The change in signal intensity at 12 ppm could be used in conjunction with a calibration curve (that indicates what a positive signal should look like), however, 12 ppm while close, is still above the chlorine concentrations used during water treatment. Nonetheless, since the E. coli counts in treated water will always be well below the detection limit of the device (10 ⁶ cfu/ml), pre-enrichment will always be required. Through pre-enrichment, the LFT will always detect only viable (live) bacteria, as only these bacteria can be cultured and hence detected. Therefore, the risk of dead bacteria producing false positive test results is prevented. Pre-enrichment is a requirement of most microbiological tests due to the need to detect as little as 1 cfu in 100 ml of water. This pre-enrichment step will ensure that the LFT detects only live E. coli in a sample.

Heat killed bacteria was purchased as a positive control from the antibody supplier. These bacteria are confirmed to retain their whole cell structure but are completely metabolically inactive. At concentrations equal to or above the detection limit of the LFT, intense positive test signals are observed when these bacteria are tested. Out of interest, these heat killed bacteria were doused with 1 ppm to 3500 ppm of chlorine, incubated for 30 minutes, and then tested on the LFTs. Only when the sample is doused with chlorine levels above 1000 ppm, does the formation of positive test signals stop. As heat killed bacteria undergo several treatments in order to lose its pathogenicity but retain its whole-cell structure, the bacterium may be less susceptible to disintegration than a normal E. coli bacterium. As a result, a higher chlorine concentration is required to completely disrupt the cells. This provides further evidence that even if the bacteria are metabolically dead, the antibodies will only stop detecting them once they have completely disintegrated.

Movement of E. coli through the LFT

To understand E. coli growth patterns, several petri dishes were inoculated with 10 ¹ to 10 ⁷ cfu of E. coli. Plates with E. coli counts between 10 ⁷ to 10 ⁵ cfu exhibited "lawn growth". When this growth pattern occurs, individual colonies cannot be counted. Individual colonies were only observed on plates inoculated with 10 ⁴ to 10° cfu.

These growth patterns were used as a guide in estimating the number of £. coli eluted from the different sections of the LFT. This is important especially when lawn growth is obtained and individual colonies cannot be counted.

As 200 μΙ of a 10 ⁸ cfu/ml sample was run on the LFT, approximately 10 ⁷ cfu was contained in its sample. After cultivation, all the elution buffers for each section of the LFT displayed "lawn growth". This is similar to the growth pattern observed on the petri dishes inoculated with between 10 ⁷ to 10 ⁵ cfu. This result therefore implies that between 10 ⁷ (100% of the original sample) to 10 ⁵ cfus (1% of the original sample) was eluted from each section of the LFT. Had less than 10 ^s cfu been released from any the sections, individual colonies would have been observed. This result implies that either between 1% to 100% of the bacteria in the sample contact the test spot and undergo detection, indicating that either extensive or little entrapment of E. coli occurs along the length of the test. To get a better indication of trapping, a 10 ³ cfu/ml sample was used, such that 10 ² cfu of E. coli was contained in the 200 μί sample run on the LFT.

The elution buffer from all sections of this LFT demonstrated single colony growth. The results are listed in Table 5, and graphically displayed in Figure 6. The least number of E. coli were released from the section of the LFT furthest away from the sample pad (the wick), while most bacteria were released from the sample pad. This indicates that most E. coli becomes entrapped in the bottom regions of the LFT (sample and conjugate pads), or because most of the bacteria were bound in place by the test spot antibody.

Using spectrophotometry, it was determined that the original sample contained ±1200 colony forming units. A total of 168 cfu were released from the entire LFT, implying that just 14% of the inlet bacteria were accounted for. The remaining bacteria were probably not eluted from the paper. If harsh elution (via vortexing) cannot force the bacteria out of the paper matrix severe entrapment may be occurring. This could be reducing E. coli movement and hence preventing it from reaching the test spot. From the bacteria that were eluted, 44.6% and 32.1% of it were released from the sample and conjugate pad respectively, the sections of the LFT in closest contact to the sample. This is of concern as it implies that 10.75% of the inlet bacteria definitely do not reach the test spot. However, since the material in the conjugate and sample pads are designed to release conjugate, they could be releasing more bacteria during elution. So while the results indicate that more bacteria are trapped in the sample and conjugate pad, it may just be that the nitrocellulose membrane traps the same amount of bacteria, but just does not release them as well during elution. Furthermore, the test spot itself could have bound E. coll, preventing it from being eluted and thereby accounting for some of the non-eluted bacteria. The antibody -antigen binding appears strong enough to withstand the type of vortexing used for elution. When an LFT with a visible test spot was vortexed, the test spot remained visible after vortexing, implying that the labelled bacteria remained bound.

Colorimetric and fluorescent staining studies were also performed. Colorimetric staining proved that E. coli certainly do become trapped in the sample and conjugate pads of the LFT. This is observed by the red staining of these LFT sections compared to the rest of the test. The sample pad is immersed directly into the sample, and is therefore the region where the most trapping probably occurs. Red staining, like that observed in the sample and conjugate pads, was not as predominant on the rest of the test. A build-up of red colour was observed at the beginning of the membrane, indicating that some bacteria may have become trapped here. The development of a test signal however, implies that labelled E. coli still reaches the test spot. However due to the apparent losses at the beginning of the LFT, the number of E coli reaching the test spot is less than that contained in the sample. The wick was also scanned for red staining, and none was observed. This could imply that little to no E coli reaches this section of the LFT.

A half strip test was used for the fluorescence study. The bottom of the test membrane was immersed into the sample itself (due to the absence of a sample pad). Insignificant trapping was found to occur on the membrane as bulk of the fluorescence was observed only on the test line and upper edge of the membrane, close to where the wick was positioned. The wick itself was not analysed. Some fluorescence was observed at the sample edge of the membrane, but was minimal. These results indicate that the sample and conjugate pads result in the most trapping, since both the colorimetric and fluorescence analysis demonstrated no staining in the lower region of the membrane. However, to confirm this is difficult as if bacteria are trapped in the membrane, they are more difficult to visualise due to the dense structure of the membrane. Seeing as visual signals only arise from the upper surface, bacteria lost in the depth of the membrane may be more difficult to see, since these fibers are packed tightly together.

Nonetheless the purpose of this experiment was to show that a) bacteria trapping does occur, and b) the number of E coli available for detection at the test spot is lower than that which enters at the sample pad. These results indicate exactly this. The question still left unanswered is how much of a lower concentration reaches the test spot, and is it significant enough to be affecting the detection limit of the sensor.

The method of fluid movement on a paper substrate is known. It is known that bulk liquid films move through the paper by passing through the channels formed by overlapping fibres and by the bulk filling of paper pores. While E. coli are motile due to the presence of flagella, it is unlikely that their movement through the lateral flow sensor occurs by swimming. It has been illustrated that bacteria move through a series of "runs and tumbles". When moving through porous structures however, the limited diameter of the pores is sometimes less than the length of their "runs". This reduces their diffusivity through the medium. When the pore size is below 10μΐη, their motion forward can completely stop. Since the pore sizes of the paper used on the LFT range between 12 to 15 μΐη, it is safe to conclude that the E. coli cannot swim up the LFT. It has also been shown that bacteria prefer to move in swarms instead of individually. Taking all the above information into consideration, it is believed that the bacteria pass through the LFT in swarms, moving along the LFT with the bulk fluid (buffer). As the fluid spreads through the paper, the suspended bacteria do so as well and manage to reach the test spot and undergo detection. Once the sample is depleted or the paper cannot absorb more fluid (whichever occurs first), the movement of the sample fluid stops, as does the bacteria. The bacteria will therefore remain wherever they were last positioned which could be anywhere along the length of the test strip. If this method of movement is true, then flushing the test with additional buffer may help transfer the remaining bacteria to the test spot, allowing them to contact the antibodies and undergo detection. Hence bacteria which would have otherwise been "wasted" can now contribute to the detection signal. However, whether or not free antibodies are available at the test spot to bind and detect even more bacteria, is unknown. Only if the antibodies can detect more bacteria, will additional flushing of the LFT help improve the detection limit.

To determine this, a flow through test was performed. In this testing format, the volume of reagents used for testing on the LFT remains constant, but the regions where trapping occurs, i.e. the sample and conjugate pads were removed. Should the flow through test offer an improvement in detection limit, trapping can be confirmed as influencing the performance of the lateral flow test.

Table 5: Number of E. coli found in the different sections of the LFT during elution studies

It was mentioned earlier that the detection limit of the LFT is 10 cfu/ml when 200 μΙ of the bacteria suspension is used for testing. In 200 μΙ, there are approximately 10 ^s cfu of E. coli.

The flow through test consists of a test membrane loaded with the test spot reagent, which is supported by an absorbent pad or wick. 1 μΙ of the test spot antibody is loaded onto the membrane and dried into place, as is done on a regular LFT. The absorbent pad supports the membrane (from behind), and helps draw the sample through the membrane. E. coli samples, ranging in concentration from 10 ⁸ to 10° cfu/ml were made. 200 μΙ of each sample was mixed with 6 μΙ of conjugate, allowed to contact for 7 minutes, and then loaded onto the test membrane. On a regular LFT, 6 μΙ of conjugate is used, and the time for which the bacteria is contacted with the conjugate before reaching the test spot, is approximately 7 minutes. This ensures a fair comparison between the LFT and the flow through test. The contact time between the bacteria and conjugate on a LFT is controlled by the absorption rate of the wicking pad. This may be different on the flow through test.

The flow through tests were found to offer the same detection limit as the LFTs, namely 10 ⁶ cfu/ml (10 ^s cfu). This indicates that the number of E. coli that actually reach the test region of the LFT may actually be that contained in the sample, and the trapping that does occur has no influence on the detection limit of the LFT. When 10 ⁴ cfu was loaded onto the flow-through device, no test signal was produced. This implies that even if the losses on the LFT (when run with 10 ⁵ cfu) are large enough to decrease the number of E. coli reaching the test spot to 10 ⁴ cfu, no test signal would be observed. Since a signal on the LFT does form when 10 ⁵ cfu are placed in the LFT, it can be confirmed that 10 ⁵ cfu reach the test spot. This indicates that even though bacteria entrapment occurs, it does so to a limited extent.

This study was performed on a lateral flow test manufactured using industry standard paper. When larger paper sensors are used, or the type of paper changes, bacteria/analyte trapping may be significant enough to negatively influence the detection limit of the sensor. For this lateral flow test, the bacteria losses would have only affected the test performance should the detection limit have been less than 1000 cfu. For example, had the detection limit of the LFT been 100 cfu and 100 cfu were loaded into the LFT, a positive test signal may not have been produced. This is because the entrapment study proved that +129 cfu were trapped in the lower sections of the LFT. It is therefore recommended that an analysis such as this be performed by developers of paper sensors to ensure that test performance is not being limited as a result of bacteria losses in the sensor itself.

Hence it was shown that the complex matrices of paper substrates can reduce the amount of analyte that actually undergoes detection, but may not necessarily reduce the performance of a sensor, depending on its size and detection limit. This proves that paper sensors offer good potential as low cost sensing substrates.

Scanning electron microscopy analysis

The sample pad, conjugate pad, test spot and control spot of an LFT run with heat killed E. coli 0157:H7 was analysed in this study. The test spot of a lateral flow test run with only buffer and a piece of nitrocellulose membrane as supplied by the manufacture, were also analysed. The manufactures membrane (used on the test region of the LFT), was not exposed to E. coli, blocking reagents or buffer and is therefore referred to as the "untreated membrane" in this discussion.

The structural differences between the different types of paper used to manufacture the LFT are apparent. Each paper type displayed a complex structure consisting of solid fibrous rods and hollow pores. The conjugate and sample pads showed a more porous structure, while the membrane fibers were more closely packed together. The untreated membrane demonstrated an organic residue on its surface, which presumably results from the manufacturing process. Such membranes undergo various treatments during manufacturing in order to ensure optimal functionality for specialised applications, such as medical diagnostics. The residue appears as a combination of spherical structures and a thin, fibrous mesh. On its surface, these spherical structures have even smaller (±330 nm) oval structures on them. The presence of the residue makes it is difficult to visualise single E. coli organisms. The residue however, is slightly larger than an £. coli organism. They are about 4 pm by 2 pm in size, compared to an E. coli which is about 2 μιτι by 0.5 pm. The residue also differs from an E. coli structure in that it has a highly irregular, rough, jagged-edged surface, in contrast to the smoother, more regular surface of an £. coli. Based on these structural differences, using SEM, E. coli was identified on the various sections of the LFT.

No E. coli was identified on the test spot of the LFT run with only phosphate buffer. The residue however, was still prevalent. This indicates that even by flushing the test with buffer or gently swirling the membrane in a blocking solution, the residue cannot be washed away. E. coli was also not observed on the untreated membrane, but the residue was still apparent.

During the colorimetric staining studies, some E. coli was shown to become trapped in the sample and conjugate pads of the LFT, and therefore do not reach the test spot. During SEM analyses, E. coli were identified in these regions, more so than in any other section of the LFT. This could have been due to the fact that fibers in the sample and conjugate pads are more dispersed and porous, making visualisation of individual organisms easier. Bacteria were also found trapped in other regions of the LFT, such as the test spot, and even in the control spot.

Some of the E. coli observed on the test spot of the LFT (run with E. coli) appear to have lodged in this region randomly, while many seem to have bound to the antibodies. However, confirming that the E. coli is bound to antibody is difficult due to the presence of the organic residue. The bacteria that were observed in the control spot region of the LFT appeared to have lodged there randomly. Most of the randomly located bacteria are either entangled between paper fibers or just lie on the paper surface. They seem to have been positioned here when the sample fluid stopped flowing. This confirms the theory that once fluid flow stops, bacteria movement stops as well. It may therefore be possible that additional flushing of the test (once the entire sample has been wicked) will help transport the bacteria left behind in the sample and conjugate pads to the test spot. This however, may not improve the detection limit of the developed LFT but may do so for sensors with lower detection limits.

Some E. coli aggregates were observed in the sample and conjugate pads. These aggregates may have formed in the original sample, prior to being loaded on the LFT. Unlike the individual organisms identified earlier, these aggregates actually appear entangled or between the cellulose fibers. The aggregates are large in size and therefore are at greater risk of becoming entangled prior to contacting the test spot. To reduce such aggregation, the sample can be sonicated for a short while before being run.

The cross section of the conjugate pad was examined by titling the microscope stage. This was done to confirm that bacteria also flow through the cross section of the paper, and not only along the surface. While titling the stage, the cellulose fibers were confirmed to be solid cylindrical structures. Therefore there is no chance of £. coli being lost to the inner section of the fibers.

E. coli was observed within the cross section of the paper. This implies that some E. coli do not travel along the surface of the paper, and bypass the antibodies. It is possible that some antibodies may penetrate into the cross section of the paper. However, even if these antibodies bind the bacteria flowing through the cross section of the LFT, the signal is lost as it cannot be observed from below the membrane surface. Hence, many bacteria do go undetected. The large number of £. coli in the cross section of the membrane (compared to its surface) indicates that bacteria may prefer to travel within the cross section of the paper. This may be due to the large porous structure of the cross section, serving as the path of least resistance.

An image of the membrane pores of the LFT run with £. coli was analysed. Many cylindrical/rod shaped structures were observed in these pores, implying that they lied within the cross section of the paper. Because their entire surface cannot be seen, it is difficult to confirm whether these structures are £. coli. Similar structures were observed in the membrane pores around the control spot on the LFT run with £. coli. For comparison, the test spot on the LFT run with buffer was analysed. No cylindrical structures were observed. Similarly, none of these structures were found on the untreated membrane. This confirms that the structures within the pores are E coli. Furthermore, like £. coli, these structures have a smooth surface. This is further evidence of how £. coli transfers through the cross section of the LFT

The above information will help prompt paper sensor developers to think of ways to reduce analyte transfer through the cross section of the paper substrate by either blocking off parts of the paper's cross section, or by simply reducing the thickness of the paper used. This will ensure that more analyte is exposed and available for detection at the upper surface of the membrane, which is especially important when colorimetric signals are used.

All the above results confirm that whole £. coli bacteria can be transferred though the LFT and do reaching the detection areas of the sensor (the test and control spots). While some £. coli is lost in the cross section of the paper, and become "stuck" in the sample and conjugate pads, this does not significantly reduce the number of £. coli reaching the test spot. As a result, it does not influence the detection limit of the developed sensor.

Detection of one £. coli organism

Pre-enrichment or pre-concentration of water samples prior to microbiological analysis is common practice in the water industry. This is because the bacterial count of most water sources is far too low for sensors to detect directly. When conventional detection methods are used, the enrichment periods required range between 18 to 24 hrs in order for one organism to be detected. The aim of this investigation was to determine whether sample pre-enrichment would enable the LFT to detect a single £. coli, and if so, determine the duration of enrichment required. The hope is that the enrichment time will be well below the current required time of 18 to 24 hours.

A single bacterium requires between 20 and 30 minutes of incubation to replicate itself (Spellberg et al. 2008), i.e. the time required for one £. coli to become two. By selecting an average replication time of 25 minutes, a spread sheet was created to approximate the incubation time required for each broth-bacteria solution being tested to reach a concentration of 10 ⁷ cfu/ml solution. A concentration of 10 ⁷ cfu/ml was selected as it develops a strong, unambiguous positive signal on the LFT. The computed results indicate that it could take ±7.5 hours of incubation to detect the 10 ² cfu/ml sample on the LFT, ±8.75 hours to detect the 10 ¹ cfu sample, and about 10.42 hours to detect the 10° cfu/ml sample. Using these approximations, this study was conducted, wherein E. coli was spiked into phosphate buffer and incubated. After 7 hours of incubation, the 10 ² cfu/ml sample produced a positive test signal. This result corresponds closely to the computed calculations. After 9 hours of incubation, the 10 ¹ cfu/ml sample produced a positive test signal and the 10° cfu/ml sample did the same at T=10 hours. Each concentration was retested from a different petri dish an hour after the positive signals were obtained. This was done to confirm that the positive signals were in fact, true test results. The intensity of the signals produced by the incubated samples was compared to that produced by a known 10 ⁷ cfu/ml sample.

The signal intensities were found comparable and therefore confirms that the incubated samples produced true positive test signals. The samples used for this study were confirmed to contain 10 ² , 10 ¹ and 10° cfu/ml by incubating 1 ml of each sample in agar overnight. The results are in listed in Table 6. To reduce the incubation time further, a study was done wherein the volume of broth used was reduced from 5 ml to 1 ml. By using less broth, the concentration of bacteria inoculated into the broth becomes higher, implying a shorter incubation period may be required to reach 10 ⁷ cfu/ml.

Table 6: Actual E. coli counts determined from plate count

For example, when 1 ml of broth is used with 1 ml of a 10 cfu/ml sample, than the starting concentration would be:

d , = C ₂ V ₂

(3 x 100)(1) = C ₂ (2)

C ₂ = 1. 5 cfu/ml

When 5 ml of broth is used, the initial concentration of bacteria is 0.5 cfu/ml. The expected reduction in incubation time might however, be counteracted by the effects of slower growth due to the availability of less nutrients per organism. As was expected however, the incubation time reduced from 10 hours to detect a single E. coli, to 9 hours. Similarly, an hour reduction in detection time is expected when analysing 10 ¹ and 10 ² cfu/ml samples.

Hence the LFTs do have the potential to be applied for drinking and waste water analysis, since a single £. coli can be detected. Furthermore, the time required for the LFTs to detect a single E. coli is less than half the time required by conventional detection methods. This system therefore offers better protection for water consumers as they can be warned of contaminated water supplies sooner, thereby reducing their risk of contracting water borne disease. By employing the use of a low cost colorimetric reader, the time for detection can be reduced even further due to the reader's ability to identify the presence of signals earlier. These earlier signals may be too weak to be noticed by the naked eye.

Analysis of environmental water samples

To determine whether the LFT can handle unprocessed, raw environmental samples, effluent samples spiked with varying amounts of E. coli 0157:H7 were tested on the LFT. Phosphate buffer powder was added to these samples. The samples wicked through the entire length of the test with no evidence of clogging. The minimum spiked E. coli concentration that produced a positive test signal was 10 ⁶ cfu/ml, same as that obtained when deionised-water is spiked with E. coli. This implies that the LFT can be used to test effluent samples directly, without the need to pretreat the sample. Furthermore, any chemicals that may have been contained in the sample were not found to interfere with the running of the test. Should the effluent on some occasions be found to contain large pieces of dirt and debris, pre-filtration may be required.

An inlet sample from Tshwane water was also analysed. The sample was mostly free of large pieces of debris as it was collected from a point in the inlet line following screen-filtration. Since it had undergone no treatment, this sample contains large amounts of bacteria. The sample was tested before and after a 1 to 10 fold serial dilution. The sample was diluted in case the level of E. coli was so high that competitive binding at the test spot may produce false negative results on the LFT. None of the samples however, even those prior to dilution, produced positive test results. This implies that the level of E. coli in the sample may in fact be lower than the detection limit of the test. This is an unexpected finding, as the inlet was thought to contain very high levels of bacteria.

To verify this, ten 100 ml volumes of the inlet sample were filtered, after which each filter was placed into nutritive media and incubated for several hours. One sample was tested every hour, and after three hours, a positive test signal developed on the LFT. This indicates that the E. coli count in the inlet was slightly below 10 ⁶ cfu/ml. To determine the concentration of the inlet, and therefore determine exactly why it requires 3 hours of incubation to be detected, the inlet sample was tested using Colilert. It was diluted and then tested, since Colilert can only detect≥2400 cfus per 100 ml. The inlet was found to contain 10 ⁷ cfu/100 ml, which equates to 10 ⁵ cfu/ml. This is slightly below the detection limit of the LFT, and explains the need for the 3 hour incubation. This inlet concentration was confirmed by independent testing at Tshwane water.

As the inlet sample was placed in E. coli and coliform selective broth, only these bacteria would have been cultured to levels detectable on the LFT. To ensure the positive test signal arose from the detection of E. coli and not due to the antibodies cross reacting with other coliforms, the following was done. The inlet sample was cultivated on E. coli specific agar, after which the E. coli colonies were isolated and used in an ELISA with the LFT antibodies. Both the conjugate and test spot antibodies bound to the environmental strains of E. coli, with a detection limit of 10 cfu/ml. This confirms that the signal produced on the LFT after incubating the inlet sample for 3 hours was a result of E. coli binding. This proves the ability of the LFT to detect environmental strains and serotypes of E. coli, other than K12 and 0157:H7. This indicates the tests applicability for environmental water analysis.

An effluent sample was also analysed. After 24 hours of incubation, positive test signals were not obtained when general nutrient broth was used. While growth was apparent (the broth turned turbid), no positive test signals formed on the LFT. When using selective broth however, a positive test signal was observed after 18 hours of incubation. According to Tshwane water, the effluent sample contained ±20 cfus/100 ml on that particular day. Based on earlier results, 20 cfus should have been detected in ±9 hours when non-selective broth is used. The reason for E. coli being detected in the selective broth and not in the non-selective broth may be a result of competitive growth. Non-selective broth would have promoted the growth of all bacteria in the sample, including all non-E. coli bacteria. As a result, these non-E. coli bacteria may have outcompeted E. coli for nutrients, preventing it from multiplying to levels detectable on the LFT. The effect of crowding may have reduced its extent of growth further. As a result, selective broth may always be required when analysing effluent samples, just to ensure that E. coli is cultivated, and can therefore be detected.

The long detection time required by selective broth however poses a significant disadvantage. The long enrichment period may be due to the fact that such media contain additives that suppress the growth of some non-target bacteria. These suppressants could be reducing the replication rate of the targeted bacteria.

Taking the above into consideration, various selective media need to be tested for their ability to promote faster replication. In this way the enrichment time to detect a single bacterium can hopefully be reduced to that obtained when laboratory samples were analysed. Alternatively, immuno-magnetic separation can be used to extract only E. coli from an effluent sample. The extracted E. coli can then be enriched in non-selective nutrient broth since non-E. coli bacteria will not be present. Non-selective media will help reduce the total enrichment time required for detection. This method is currently under investigation. In conclusion, the LFT has great potential for use in the water monitoring industry. The LFT is easy to use and interpret, low cost, and can detect the common strains of E. coli that exist in environmental samples.

Comparison with commercially available tests

The performance of Colilert, Microsnap and Cell Biolab's lateral flow tests for E. coli 0157:H7 were evaluated. To determine the detection limit of these tests, E. coli 0157:H7 suspensions ranging from 10 ⁸ to 10° cfu/ml were prepared. Samples were prepared according to the supplier's instructions and tested. The detection limit of the commercial LFTs was found to be the same as that of the developed LFTs, namely 10 ^s cfu/ml. This implies that the developed LFT performs equally as well as a commercially available (and hence optimised) lateral flow test for E. coli 0157:H7. The antibodies used on the commercial LFT only targets the E. coli strain, 0157.Ή7, while the antibodies on the developed LFT targets a much broader group, namely, all O and K serotypes. Interestingly, the broader antibody achieves the same detection limit as the more specific one. When tested with £. coli K12, the developed LFT still offered a detection limit of 10 ^s cfu/ml, although there was a reduction in the intensity of the test signals. As a result, the possibility of the broader antibody having a higher affinity for the common 0157:H7 strain can be ruled out. To ensure that the detection limit of 10 ⁶ cfu/ml on the developed LFT holds true for all strains of E. coli, a more comprehensive list of strains should be tested. Nonetheless, the above result is a good indication of the performance of the LFT of this invention. By incorporating a pre-enrichment step, both the commercial and the developed LFT can detect a single E. coli bacterium. Cell Biolab's LFT can detect 1 E. coii in 18 to 24 hours. However, as mentioned earlier, the developed LFT could potentially detect one E. coli in under 15 hours, however further work is required with real water samples before this can be confirmed. The commercial LFT cannot be applied for water analysis due to its inability to detect multiple strains of E. coli, and is in fact, currently marketed for food analysis only.

Two other studies were also performed on Cell Biolab's lateral flow tests. A specificity analysis was done and the tests ability to detect dead bacteria was also investigated. For the specificity analysis, the different types of bacteria listed in Example 1 were used. No cross reactivity was observed on the commercial LFTs, including E. coli K12. The developed LFTs were found to display cross reactivity with Salmonella enteritidis and Pseudomonas aueruginosa. The cell biolabs test therefore proves to be more specfic than the developed LFT and this can be attributed to the higher specficity of the antibody used on these tests. This antibody targets only a single strain of E. coli, compared to mulitple strains and serotypes targeted by the antibody on the developed LFT. There could be bacteria with which the Cell Biolab's tests cross reacts, but this can only be determined by testing with more types of bacteria.

The commercial LFTs were also tested using E. coli suspensions that have been heat and chlorine killed. Household bleach was used to obtain the chlorine killed bacteria. In the water industry, dead bacteria should go undetected as only live bacteria pose a threat to water consumers. The commercial test strips produced postive test signals when tested with the dead E. coli samples. Hence, these tests cannot distinguigh dead from live bacteria, similar to the developed LFT. This is mostly due to the fact that when rendered metabolically dead, the bacteria may still remain intact or large fragments of the membrane with its surface antigens are still accesible to the antibody for binding. However, both test strips can overcome this issue via the use of the pre-enrichment step.

The performance of the Colilert system was also analysed. For specificity testing, suspensions of Pseudomonas aeruginosa and Salmonella enteritidis were prepared. Enterobacter cloacae and E. co// 0157:H7 suspensions were also prepared and served as positive controls. The formation of yellow wells in the Colilert tray after 18 hours of incubation indicates the presence of conforms in the sample. To indicate the presence of E. coli in the sample, these wells fluoresce when exposed to UV light.

As expected, Enterobacter cloacae and E. coli samples produced yellow wells on the Colilert tray. When exposed to UV light, the wells on the E. coli test fluoresced, confirming the presence of E. coli. The test using Salmonella enteritidis suspensions produced no coloured or fluorescing wells, as expected. The Pseudomonas aeruginosa test produced false positive results, albeit to a limited extent wherein just two wells turned yellow. Pseudomonas aeruginosa is neither part of the E. coli species or the coliform/enterobacteriaceae genera and should therefore go undetected on the Colilert test. The wells on the Pseudomonas aeruginosa test did not fluoresce, confirming the result as a false positive test for the presence of conforms. The false result indicates the possibility that Pseudomonas aeruginosa may contain/secrete the enzyme responsible for producing the yellow colour, B- galactosidase. The developed lateral flow test also displayed cross reactivity with Pseudomonas aeruginosa. However, since the lateral flow test is not based on enzymatic reactions, B-galactosidase cannot be responsible for the non-specific interaction. Upon further research it was found that certain strains and serotypes of Pseudomonas aeruginosa possess the enzyme B-galactosidase (Kong et al. 2005; Marne C et al. 1988; Rohlfing 1968). This explains the cross reactivity observed on the Colilert test. It was also discovered that Pseudomonas aeruginosa shares common surface antigens with some strains of E. coli (King et al. 2008), which could serve as an explanation for the cross reactivity observed on the lateral flow test. Seeing as Colilert, a method considered a new standard for E. coli detection is not free of cross reactivity; the issue of non-specific interactions on the LFT may not be detrimental to the test's success, if it occurs to a limited extent. A study by Luyt et al (2012) states that the specificity of Colilert is 92.8%, however Colilert can only detect 56.4% of all E. coli strains. As mentioned by Guo ef al (2005), the O-antigens which are targeted and used for detection by the LFT exists on most, if not all E. coli types. This implies that the developed LFT has a great potential to detect more than 56.4% of all E. coli found in the environment. Furthermore, should cross reactivity with other pathogenic or sewage-related bacteria occur on the developed LFT, the intent of the test, to identify unsafe water sources, will still hold true.

The Colilert system can detect as little as 1 cfu in a 100 ml water sample, but requires a prolonged 18 hour pre-enrichment period to achieve this. This low detection limit makes the Colilert system well suited for water testing, however its long detection time is not ideal.

The LFT of the present invention has demonstrated its potential to detect 1 cfu in 9 hours, and since most environmental strains of E. coli possess the O antigen (Guo et al. 2005), the LFT should potentially detect more than 60% of all E. coli. The 9 hour pre-enrichment step also enables the LFT to distinguish dead from live bacteria.

Finally, the Microsnap E. coli detection system was analysed. The Microsnap system is fairly new, and no information regarding its implementation at water treatment plants or environmental water testing authorities has been found. The MicroSnap Coliform and E. coli detection system is primarily designed to give a rapid and semi-quantitative assessment of E coli and coliforms in food samples (Meighan et al., 2014). Hence, while the system may be applied for water testing, further validation with environmental water samples is required. Thus far, validation studies have only been performed using spiked bottled water (Meighan et al., 2014). The product is also sold as a system for quantitation and detection at low to medium levels of both E. coli and coliforms. This implies that during environmental bacterial outbreaks (when high bacterial counts occur), this system may not be suitable to serve as an early warning system. The developed LFT system is well suited for such cases, providing results within 15 minutes or in less than 3hrs. Microsnap's method of using enzymes to detect the targeted bacterial species, and even further, correlating the quantity of enzyme to the quantity of bacteria present, is still a developing field. Further validation of the system is needed before it can be implementing for water testing globally.This Microsnap system detects E. coli based on the activity between the enzyme B-glucoronidase (found in E. coli) and a bioluminogenic substrate. The product of the reaction is read using a luminometer that correlates the relative light units of the sample with the number of colony forming units it contains. The system was analysed for its detection limit, specificity (using the bacteria listed in Example 1) and its ability to differentiate between dead and live bacteria. For the latter study, an E. coli suspension was rendered dead by contacting it with chlorine. The results are shown in Table 7.

Table 7: Microsnap test results

The results indicate that Microsnap exhibits slight cross reactivity with non-E. coli bacteria. Salmonella contains the targeted enzyme, B-glucuronidase (Tryland et al., 1997), which explains the non-specific result. Cross reactivity with Enterobacter cloacae occurs to a lesser extent, with only 1 % of the bacteria in the original sample being reported as E. coli. Pseudomonas aeruginosa displays no cross reactivity. The chlorine killed E. coli sample was reported as containing <100 cfus. The original sample actually contained twice this amount. This could imply that live bacteria were present in the tested sample, however when cultured in agar, no bacteria grew confirming that all the bacteria were dead.

The negative control sample (buffer) was reported as containing <50 cfu of E. coli. This occurred on each of the separate occasions that the system was tested. The buffer was assessed for contamination by cuituring and was confirmed E. coli free. It is possible that the filter holders used for the study may have been contaminated when E. coli rich samples were filtered through them. However, this is unlikely as they are rinsed and washed with ethanol each time before use. Furthermore should this be the case, it is unlikely that it would occur each and every time this test was performed, which in this case, was on different days. When tested with 10°, 10 ² and 10 ³ cfu, the Microsnap system reported it as such. Microsnap therefore appears able to detect less than 10 cfu in a sample within 6-8 hours. Additionally, its ability to detect only E. coli (specificity) and its ability to distinguish dead from live bacteria requires further analysis. Furthermore, the false positive results obtained with buffer indicate that more analyses and verification of this system is required to completely ascertain its reliability and suitability for water analysis

Table 8: Summary of system comparison

Conclusion

A lateral flow test for E. coli has been developed. The LFT epitomises many of the characteristics desired in a modern E. coli detection system for water safety monitoring. It is low cost, simple to operate, and results can be interpreted by semiskilled field workers. This empowers them to make decisions regarding the safety of water sources, the efficiency of treatment processes, and serves as a method of quick testing in cases of bacterial outbreaks when £ coli levels become very high.

The test has a detection limit of 10 ⁶ cfu/ml, but when combined with an enrichment step, can detect a single organism. The test has the potential to detect a single £ coli in less than half the time required by conventional methods. The differences in enrichment time required for the detection of different concentrations of the bacterium can potentially serve as a method of quantifying the test results. Furthermore, enrichment also helps to ensure that the LFTs only detect live bacteria, an important requirement in the water industry.

While LFTs for whole-organism £ coli do exist, many of them only detect a specific strain of the bacterium, such as Merck's Singlepath test for £ coli 0157:H7. As a result, they cannot be applied for water analysis. EXAMPLE 3

Development of a paper based sensor for the detection of E. coli in water sources: From lateral flow to paper based microfluidics

Working principle

Figure 7 displays an image of the developed paper based microfluidic device. An E. coli rich aqueous sample enters the device at the sample inlet. This inlet was made slightly wider than the rest of the device, in order to accommodate and facilitate mixing of buffer powder into the water sample. Buffer helps to prevent the production of false positive test results. When the sample enters the device, its flow rate is reduced (due to the larger area of the mixing chamber) to allow for sufficient time for the buffer powder to dissolve into the sample. After mixing, hydrophobic wax barriers guide the sample towards the detection zones, which consist of the conjugate, test spot and control spot reagents. The E. coli in the sample first interacts with the conjugate and becomes labelled by gold tagged rabbit-anti-E. coli antibodies. The labelled E. coli flows towards the narrowing region of the device, which serves to direct all the sample fluid over the test and control spot regions. This narrowing region therefore ensures that little to no gold-Ab-E. coli complex bypasses the test and control spot of the device. In this way, no gold labelled bacteria (which contribute to the signal colour intensity) are lost. The test spot contains rabbit anti E. coli antibodies and the control spot contains goat anti rabbit antibodies. The E. coli- gold-Ab and gold-Ab complexes are then forced into the test and control spot region of the device, where they are bound in place. As a result of the Ab-E. coli-No sandwich formation in the test spot, a red colour signal is produced. The remaining conjugate antibodies are then captured in the control spot, forming a second red colour signal. Since sample continuously wicks through the device, even after the test and control spot signals have formed, it serves as a washing mechanism that removes any free conjugate lying randomly on the test, especially on the test and control spots. This enables easier visualisation of the test signals. Hence the presence of a test and control spot indicates a positive test result for E. coli.

Materials

All devices were designed using a computer aided drawing programme, Design CAD 3D MAX 21. To print the designs on paper, a Xerox Color-Qube 8870 solid wax printer was used. This printer operates by heating resin based solid wax cartridges (these replace the ink cartridges found in standard desktop printers) until they are melted and then deposits the molten wax droplets onto the paper surface. Once the wax comes into contact with the paper, it cools quickly, preventing any further spreading of the wax on the paper. Designs were printed onto 20 cmx20 cm sheets of Whatman chromatography paper (No. 1). Chromatography paper was selected as it is readily available, relatively inexpensive and can be reproducibly manufactured.

To create the flow channels on paper, the printed device is placed onto a heating source (either an oven or hot plate) to enable the wax to melt and penetrate into the cross section of the paper. An Ecotherm oven (Model 22, Labotec) and a StableTemp, DLM 51806-15, (Cole Parmer) hot plate was used. Once melted, the device is allowed to cool for <10s, and is then ready for use. The sensing reagents used for E. coli detection included the following. A polyclonal, rabbit anti-E coli antibody (Thermo Fischer Scientific, USA) was used in the conjugate and a polyclonal rabbit anti-E. coli antibody (ABD Serotec, UK) was used on the test spot. Both these antibodies detect all "0" and "K" antigens on the E. coli bacterium. Goat- anti-rabbit antibodies (Kirkegaard & Perry Laboratories, Inc) were used on the control spot. Gold nanoparticle solution, with a particle size of 40 nm and an optical density (OD) of one was purchased from the Diagnostic Consulting Network (California, USA). This gold was used to label the conjugate antibodies. For more details on the conjugate preparation refer to Example 1. Phosphate buffer powder was purchased from Sigma Aldrich. A blocking solution was purchased from Invitrogen, and is used to block the paper to prevent non-specific binding.

ATCC strains of E. coli 0157:H7 were kindly provided by the Natural Resources and Environment Unit (CSIR, South Africa). Bacteria were cultured using nutrient agar and nutrient broth, both of which were purchased from Oxoid. To determine the bacteria concentration of samples, the McFarland standards, in conjunction with UV/VIS spectrophotometry (Perkin Elmer) was used. Serial dilution was used to prepare samples of different bacterial counts for testing.

Methods

Determination of the optimum printed line width

Lines of different design width (the width inputted in Design CAD), ranging from 30 pm to 1000 pm were printed and then melted on a hot plate at the same temperature and for the same time period. The image formed on the reverse or under side of the printed surface (called the melted surface) was examined. The optimal line width is that which forms the darkest and clearest image on the melted surface. This indicates that wax from the printed surface has successfully penetrated through the depth of the paper. To further analyse wax penetration, a cross sectional analysis of the paper was done. The smallest line width that resulted in complete vertical penetration of the wax through the cross section of the paper was considered the optimal line width.

Comparison of the melting effectiveness between an oven and a hot plate Printed lines were melted using both an oven and hot plate at temperatures ranging from 50°C to 250°C, in increments of 50°C. The images formed on the melted surface at each temperature were compared for each of the two melting processes. The optimal heating source was selected as that which forms the darkest image on the melted surface at each investigated temperature.

Determination of the optimum melting temperature

Lines of different width were melted at temperatures ranging from 50°C to 250°C, in increments of 50°C. Although the optimal line width was already identified, using a range of line widths helps to better understand the influence of temperature on the formation of barriers. The optimal melting temperature was selected after analysing the darkness of the image formed on the melted surface for each of the investigated temperatures. Determination of the optimum melting time

Printed lines of varying widths were melted at the same temperature for different time intervals ranging from 1 minute to 7 minutes. The optimal melting time period was selected by probing the image formed on the melted surface for each melting time period investigated.

Determining the effectiveness of barriers

5 mm by 5 mm squares with different line widths ranging from 50 pm to 1000 Mm were printed. Each square was then melted at 200°C for 1 minute on a hot plate to form hydrophobic barriers in the cross section of the paper. 10 μΙ of diluted food dye was pipetted into the printed square chambers and monitored for any leakage for 10 minutes.

Development of a Preliminary Paper Based Microfluidic Device (PBM) device A preliminary paper based microfluidic device was fabricated based on the results of the optimisation studies described earlier. The developed device showed no signs of fluid leakage across the hydrophobic barriers, and contained fluid for more than 30 minutes. This indicates the formation of functional hydrophobic barriers and functional hydrophilic channels. The antibodies used to develop the lateral flow test (described in Example 1) were then added to the paper based device as follows.

The conjugate zone of the device was first blocked using a commercial membrane blocker. Excess blocker was patted off with a towel, and then dried for 1 hour at 37°C. The conjugate was then added and dried in place for 30 minutes at 37°C. Next, 1 μΙ of the test and control signal antibodies were loaded into their respective detection regions and dried for 30 min at 37°C. Thereafter they were blocked and dried for 1 hour at 37°C. All antibody concentrations were kept constant to that used on the lateral flow test. Before use, phosphate buffer powder was added onto the device, which was then sandwiched between 2 pieces of plastic adhesive. The detection limit of the PBM device was determined by testing it with various concentrations of E. coli 0157:1-17.

Results and Discussion

Development of a paper based microfluidic device for E. coli detection in water

Using an optimal line width of (500 μιτι), an optimum melting time of 1 min and optimal melting temperature of 200°C, a PBM device was manufactured. This is demonstrated in Figure 7.

A calibration scale was developed based on the signal intensities obtained when testing various concentrations of bacteria. This colour scale was printed onto the device itself, to demonstrate how simple quantitation of colorimetric results can be achieved.

PBM devices were tested with suspensions of E. coli ranging from 10 ^s to 10 ^s cfu/ml. Positive test signals were obtained from 10 ⁸ cfu/ml to 10 ⁶ cfu/ml. Phosphate buffer was used as a negative control, and no test signal was observed. These results are similar those obtained on the lateral flow test, where a detection limit of 10 ⁶ cfu/ml was observed. This indicates that the PBM device does not reduce the functionality or performance of the optimised sensing mechanism in any way. However, the intensities of the positive test signals produced on the PBM device are weaker than those obtained on the lateral flow test. To determine if this is due to the change in paper type, a lateral flow test was made using chromatography paper as the test membrane. When tested with varying concentrations of E. coli, weak test signals were still obtained. Upon close examination, it was determined that when the test and control spots are applied to chromatography paper, they are wicked into the paper faster and therefore spread out to greater extent than that which occurs when using regular nitrocellulose membranes. As a result the test and control spots appear larger in size, implying that there is a decreased antibody concentration per mm ² of paper compared to when nitrocellulose is used. This could be why weaker test spots are observed on chromatography paper. To reduce this effect, the hydrophobic barriers in the test spot region can be positioned closer together to reduce the area available in which the reagents can spread. This could increase the antibody concentration per mm ² of paper area, and improve the intensity of the signals.

Conjugate release from the chromatography paper was also poor. A significant quantity of conjugate did not release completely, implying that they did not contribute to signal formation. This could be another reason for weakened test signal intensity. The conjugate pad material used on the LFTs is made of glass fiber and was developed specifically to release conjugate in lateral flow tests. In fact, the glass fibres have a release factor of over 90%. Chromatography was developed to perform chromatographic separations, and will therefore, instead of releasing conjugate, try to slow down its movement up the test. Increasing the extent of blocking may help improve conjugate release, however this requires further investigation.

Conclusion

A paper based microfluidic device was developed for the detection of E. coli in water sources. Due to the inherent nature of the paper used to create this device, the intensity of the test signals obtained were weaker than that observed on the lateral flow tests. Nonetheless, the PBM device offered the same detection limit as the lateral flow test, i.e. 10 ⁶ cfu/ml.

The fabrication studies demonstrated that the original line width is the most important fabrication parameter in creating effective wax barriers. An optimal line width of 500 Mm was selected, however any line width above 300 pm proved capable of forming impermeable wax barriers. The effect of temperature on melting effectiveness was only observed for lines narrower than 300 pm, but it is believed that temperature also influences wider lines as well. An optimal melting temperature of 200°C was selected, as it gave the best wax penetration for narrow lines. However, should lines wider than 300 pm be used, a lower melting temperature can be used. Optimal melting times depend largely on the melting temperature. An optimal time of 1 minute was selected when using a melting temperature of 200°C. Melting times above 1 minute did not improve wax penetration, however shorter melting periods do significantly decrease the extent of wax penetration. A hot plate was found to be a more effective heating source as it allows for direct contact between the heating plate and the wax. The barrier effectiveness test confirms that line widths below 300 pm cannot form impermeable hydrophobic barriers in paper. It was shown that line resolution decreases as the devices are manufactured, especially after the melting stages. To create a device with a desired dimension, one must take into account wax spreading, which could produce lines that are 13% to 25% wider. This spreading also dictates the minimal distance that should exist between two parallel wax lines during the design phase. Furthermore, lines narrower than 300 μιη are difficult to print, due the resolution capabilities of the printer. It was shown that higher melting temperatures and longer melting times can negatively affect line resolution, and does not significantly improve vertical wax penetration. A melting temperature above 200°C cannot be used to create barriers narrower than 300 pm as lateral spreading dominates vertical spreading. Fibre orientation within the paper was found to influence wax spreading both vertically and laterally through the paper, with its affect becoming more pronounced for lines narrower than 300 pm. Vertical lines were found to form more effective wax barriers, even though they exhibit higher lateral spreading. Hence, all the fabrication stages used to create paper based microfluidic devices were fully optimised, and use of these findings will ensure rapid, development of fully functional, low cost, diagnostics devices.

EXAMPLE 4

Analysis of samples obtained from Tshwane waste water treatment works (Daspoort, South Africa)

Samples were collected from Daspoort Waste Water Treatment Works (WWTW) for microbial analysis. The samples were collected at the outlet of the secondary sediment pond (SP) and from the treated effluent (E) which exits the plant. Daspoort's Micro Laboratory tests the effluent for E. coli and coliform content daily. The developed method seeks to replace the method they currently use for this analyses (membrane filtration), with a more rapid, low cost, and simple test.

The sediment pond sample was analysed to prove that the LFT can detect common environmental strains of £. coli. The water from the sediment pond is not treated with UV, but has undergone some level of chlorine treatment. As the effluent undergoes both UV and chlorine dosing, the sediment pond should contain higher counts of bacteria (especially E. coli). Preliminary results at the CSIR indicate that the sediment pond does contain higher counts of E. co///coliforms, and may have been treated with less chlorine than the effluent. The sediment pond samples were also analysed in order to determine whether a link between bacteria content and chlorine dosing can be developed.

The sediment pond and effluent were analysed using three methods (Figure 14). To assess whether the LFT can detect E. coli types occurring in both the effluent and the sediment pond, "24-hour analyses" were performed. Petri dishes were loaded with effluent or sediment pond samples (Figure 14a, b), incubated for 24 hours and thereafter tested on a LFT. Any bacteria contained in these samples would grow to levels detectable on the LFT within this time.

A "timing test" was also performed in order to determine the length of time required to detect different E. coli counts in both the effluent (E) and sediment pond (SP) matrices. To do this, water collected from the sediment pond and effluent were spiked with various counts of E. coli 0157.H7 (Figure 14c), incubated for 24 hours, during which time a lateral flow test was run hourly. For all these studies, an E. coli and coliform selective broth (Colilert, IDEXX) was used for culturing. The broth was made up in dibasic phosphate buffer (pH 8.5). Colilert broth turns yellow in the presence of E. coli and coliforms, typically after 18 hours of incubation. The LFTs used for these studies were made using Rabbit anti E. coli "all O and K" antibodies which were labelled with the colorimetric reporting element, nano-gold.

Results from 24 Hour Analyses

Figure 15 confirms that the LFT can detect the types/strains of E. coli found in both the sediment pond and the effluent. In 82% of sediment pond analyses, E. coli was detected on the LFT. In 60% of the effluent analyses, E. coli was detected. This is as expected since at Daspoort WWTW, the sediment pond lays upstream of several treatment processes and therefore will contain more E. coli than the effluent. The effluent, after having passed through the complete treatment works, is usually said to contain less than 30 cfu/100ml, and on several occasions, contained 0 cfu/100ml (as per results obtained from Daspoort's laboratory). It is probable that in the 40% of cases E. coli was not detected in the effluent, the effluent contained 0 cfu/100ml on that day. In fact, during sample collection on certain days, strong chlorine fumes were smelt. High chlorine concentrations would kill any bacteria contained in the effluent. Hence, the LFT results show strong correlation with what is expected from each sample type, and with results obtained from Daspoort labs. Overall, these results prove that current method is capable of detecting E. coli types prevalent in South African waters.

Results from Timing Studies

The next analysis was performed in order to determine if method and device of the present invention can detect £ coli in time periods shorter than that which is currently achieved. This was done by spiking waste water samples [from the sediment pond (SP) and effluent (E)] with £ coli 0157:H7, with counts between 10° and 10 ⁵ cfu. As a control, phosphate buffer (PB), was also spiked with these E. coli concentrations. The spiked samples were incubated and tested hourly. The results are shown in the Figure 16. Figure 16a is considered the best case scenario, wherein the shortest time to detection recorded for each concentration is displayed. Figure 16b is considered the worst case scenario, wherein the longest time to detection recorded for each concentration is displayed. The differences in the time required for detection for each concentration is due to interfering compounds contained in the samples themselves. These compounds are different on different days, and its impact on the time required for detection is discussed later.

Figure 16 demonstrates how an increase in bacteria count decreases the total time required for detection on the LFT. It also indicates that the time required to detect a specific bacteria count is independent of the sample matrix, provided that sample matrix contains no additives that can kill the spiked bacteria, or hamper its growth in any way. As an example, the 10 ³ cfu count in all 3 matrices (PB, SP and E) requires ±8.5 hours for detection in the worst case scenario. Table 9 summarises the average time required to detect the different £ coli counts. These times are averaged across all three sample matrices investigated (SP, E and PB). Table 9: Total detection time required by LFT in order to detect different bacteria counts. Bacteria Time required for detection: . Time required for detection: count (cfu) best case (hr) worst case (hr)

10 ⁵ 5 5

10 ³ 6 8.5

10 ¹ 12 13

10° 15 17

The results of a comparison for time required for detection on the LFT to the current state of the art in E. coli detection test (Colilert) are set out in Figure 17. This analysis was done by comparing the time required for the Colilert broth to turn yellow (indicating the presence of E. coli and coliforms), to the time required for a positive test signal to be obtained on the LFT. Figure 17a shows the shortest times recorded for detection for each bacterial concentration on both the LFT of the present invention and Colilert test, while Figure 17b indicates the longest times recorded for detection for both tests.

In the best case scenario (Figure 17a), the lateral flow test provided results earlier than Colilert over majority of the E. coli counts investigated. In the worst case scenario, the LFT still out-competes the Colilert test over most bacteria counts. There were instances however, in which Colilert test provided results earlier. The results obtained at high bacterial counts appear promising (see box in Figure 17). The LFT provided positive test results in a significantly shorter time period than Colilert test, a promising indication of its ability to serve as an early warning system. An early warning system such as this currently does not exist.

Hence the results of the present lateral flow test appear promising, demonstrating the tests ability to serve as a low cost, rapid and simple E. coli detection system. The test has proven capable of detecting environmental strains of E. coli, and various counts of the bacterium in time periods shorter than that achievable by current E. coli detection systems.

REFERENCES

Byrne, B., Stack, E., Gilmartin, N., & O'Kennedy, R. (2009). Antibody-based sensors: principles, problems and potential for detection of pathogens and associated toxins. Sensors (Basel, Switzerland), 9(6), 4407-45. doi:10.3390/s90604407

Dungchai, W., Chailapaku, O. and Henry, C. S. (2011). A low-cost, simple, and rapid fabrication method for paper-based microfluidics using wax screen-printing. Analyst, Vols. 136 Pg 77-82.

Govindasamy, K., Potgieter, S., Land K., and Muzenda. E., Fabrication of Paper Based Microfluidic Devices. In: Proceedings of the World Congress on Engineering, London, UK, 4-6 July 2012.

Guo, H., Yi, W., Shao, J., Lu, Y., Zhang, W., Song, J., & Wang, P. G. (2005). Molecular Analysis of the O-Antigen Gene Cluster of Escherichia coli 086 : B7 and Characterization of the Chain Length Determinant Gene (wzz), 71 (12), 7995-8001. doi:10.1128/AEM. 71. 12. 7995. Hossain, S., Ozimok, C, Sicard, C, Aguirre, S., Ali, M., Li, Y. and Brennan, J. D. (2012). Multiplexed paper test strip for quantitative bacterial detection. Analytical and Bioanalytical Chemistry, 403, 1567.

Jokerst, J. C, Adkins, J. A. Bisha, B., Mentele, M. M., Goodridge, L. D. and Henry, C. S. A Paper-Based Analytical Device for the colorimetric detection of foodbome pathogenic bacteria. In: Proceedings of the 15th International Conference on Miniaturized Systems for Chemistry and Life Sciences, 2116 (2011).

King, J. D., Mulrooney, E. F., Vinogradov, E., Kneidinger, B., Mead, K., & Lam, J. S. (2008). IfnA from Pseudomonas aeruginosa 012 and wbuX from Escherichia coli 0145 encode membrane-associated proteins and are required for expression of 2,6-dideoxy-2-acetamidino-L-galactose in lipopolysaccharide O antigen. Journal of Bacteriology, 190(5), 1671-9. doi:10.1128/JB. 01708-07.

Kong, K., Jayawardena, S. R., Dayaram, S., Puerto, A., Koh, C, H0iby, N., Kong, K., et al. (2005). Pseudomonas aeruginosa AmpR Is a Global Transcriptional Factor That Regulates Expression of AmpC and PoxB β and Other Virulence Factors Pseudomonas aeruginosa AmpR Is a Global Transcriptional Factor That Regulates Expression of AmpC and PoxB β-Lactamases. American Society for Microbiology, 49( 1), pp. 4567-4575.