PROPERTIES

Background of the Invention

Proteins play a key role as biomarkers in every stage of disease: prognosis, diagnosis and treatment. This complex arrangement of amino acids is our exclusive insight in the physiological state of a cell as usually— there is no direct correlation between m NA levels and protein expression[j_, 2]. Protein activity often depends on interactions with several other proteins (e.g. cytokines)[3, 4], generating a need for highly multiplexed detection systems. Protein microarrays, consisting of spatially addressable test sites, allow us to monitor many biorecognition reactions in parallel. This approach to protein detection provides a wealth of information in a short period of time, which continues to improve our understanding of disease pathways[5], and to aid drug target[6, 7] and biomarker discovery[3, 5, 6].

Despite the desire to explore protein interactions, several obstacles have prevented protein microarrays from achieving the same popularity as their DNA counterparts [i, 3, 4, 7]. Proteins are fragile and environmental stress alters their conformation and compromises their bioactivity[I, 8].

AdditionallyEven though early detection of the few biomarker copies is linked to higher patient survival rates [3, 5], clinical sample volumes are small (e.g. biopsy material or blood from an infant heel prick can be <10 μ1)[2, 6, 9] and interesting proteins are present in low concentrations [4, 6, 10 ^" ). In plasma, 99% of the protein mass is from only 22 high abundance proteins[4] but most of the diagnostically relevant protein markers are in low abundance and, unlike polymerase chain reaction for DNA[I, 3, 8], although early detection of the few biomarker copies is linked to higher patient survival rates [3, 5].

Protein microarrays can be classified based on the test site biomolecule. Forward phase, or capture, arrays are a pattern of several different probes (e.g. antibodies). The array is incubated in the sample solution to simultaneously test for multiple targets. In contrast, a reverse phase array consists of a pattern of samples; a fraction from multiple protein samples is spotted onto the substrate and tested for the presence of one protein. Forward phase arrays are typically employed in biomarker discovery[4] or clinical diagnosisQ _. , 5], while reverse phase arrays, originally designed for biomarker detection[i, 5], additionally feature in cell pathway analysis for therapy response or disease progression[5, 6].

When detecting proteins, both the forward and reverse phase arrays present problems. Labels (e.g. enzymes, fluorescent dyes, quantum dots, or metallic nanoparticles) are commonly used to visualise the presence of the target protein|T l ^" |. In a forward phase array, the target is either directly labelled or detected with a second, labelled antibody (i.e. sandwich assay). The fragile and complex nature of proteins discourages direct labelling; variations in labelling efficiency hinders quantification and adding a tag could affect protein binding[J_, 3, II] . The sandwich assay has higher specificity and sensitivity, but suffers from additional cross reactivity between detection antibodies [i, 3]. As a result, multiplexing in the system is limited to detecting 30-50 antigens in parallel[3, 5J. Additionally, relatively large sample volumes, i.e. up to several hundred microliters, are needed to submerge the test sites, up to several hundred microlitres. The reverse phase array requires very little sample, which also means few target proteins per spot[8, 12]. The protein solution can be pre-fractionated to remove large amounts of undesirable proteins[8, 10], but any biomarkers complexed to these proteins will also be removed 10, 13]. Proteins are sticky and the low abundance targets can be lost while transferring the purified solution to the test spots.

Specification of the Invention

The invention proposes a device comprising a simple and low cost three-dimensional (3D) microarray to simultaneously sort and pattern fractions of biological samples. It addresses the issues present in either forward or reverse phased arrays, alone, by combining forward and reverse phase arrays. The samples are distributed into microchannels in the x-y plane and sorted in the z-direction by

multiplexed affinity columns (microchannels). Capture probes are immobilized systematically along the length of the separable columns. The columns comprise in z-direction a stack of layers, whereas each layer in the stack captures a specific target component.

Thus, a layered vertical- flow system is obtained, capable of simultaneously fractionating and spotting several protein samples in parallel. The samples could be, but is not limited to, blood, urine, cell lysate or tissue lysate.

The spotting and/or detection process carried out by the apparatus is not limited to proteins, but with appropriate markers in the corresponding capture layers, it could also be applied to DNA, peptides, other small molecules or a combination of the proposed ligands. Furthermore, it is not limited to assemblies where the sample comprises ligands that bind directly to the capture layer, but the capture or the binding could be achieved by placing specific carriers and/or microparticles in the capture layers. For example, the layers could be a non-permeable support containing a microarray of holes. The holes are filled with a porous material (e.g. a hydrogel) with a pore size large enough to let proteins diffuse through and small enough to physically trap biofunctionalised microparticles within the matrix. The hydrogel is inserted into the holes and crosslinked (e.g by UV, temperature change, or ions), creating individually functionalised layers. The 3D microarray is not limited to affinity-based sorting; target proteins could also be directly attached to the layers. Unlike a traditional reverse phase microarray, the proteins would be sorted by size or charge as they are patterned onto the substrate. This could be achieved by stacking membranes with different pore sizes or by filling the channels with a conductive medium (e.g. a hydrogel) and applying an electric field. The layers would then be separated, as before, and analysed individually.

To achieve the appropriate capture by the layers, each layer in the microarray stack is functionalised with a different capture probe. Functionalization can be physical adsorption or covalent coupling. The probe can be directly attached to the matrix or linked to a carrier (e.g. a microparticle) which is then embedded in the porous material.

The patterned and functionalized microarrays are held together in a stack with appropriate non- permanent means, so that they can be peeled apart for analysis. The adhesion can be realized, for instance,,by a non-permanent, double-sided adhesive, or just by appropriate pressure over the whole surface by mechanical clamping layers from top and bottom. In a preferred embodiment of the device, the top and/or bottom mechanical clamping layer contains holes that serve as the entry channel to the multiplexed affinity columns, such as a bottomless wellplate. In an even more preferable environment, a spacer layer, is inserted between the top and/or bottom patterned nitrocellulose layer and the respective top and/or bottom mechanical clamping layer, more preferably each of those spacer layers also containing a hole just above or below each of the channels, that serves as a reservoir to hold the sample and to prevent leakage between the mechanical clamping layer and the capture layer.

To carry out the sample characterization with the device, an essentially liquid sample is introduced into an array of separable affinity columns obtained by layering, as described, above. Preferably, the liquid samples are in the order of one or few μΐ. As the samples pass through the layers, target proteins bind to specific locations throughout the microchannel. Similar to a reverse phase array, this approach analyses small volumes from multiple samples; however, it has the added advantage of testing each 90 sample for multiple targets, corresponding to the multiple layers, as is done in a forward phase array.

Preferable there is some active support mechanism to actively pull the sample through the

microchannels, i.e. a force is exerted onto the sample liquid in parallel to the microchannel axis, most preferably be centrifuging the device. This leads to a reduction in assay time, ensures the entire sample is pulled through the channel, and thus increases signal intensity and reduces variability.

95 The layers in the 3D stack are peeled apart or otherwise separated, revealing 2D microarrays that contain only a specific fraction from multiple samples. Preferably, these arrays are then incubated in a labelled antibody specific to the captured protein, which increases the multiplexing by removing cross- reactivity between detection antibodies.

In a preferred embodiment, the capture layers are essentially made from a porous material, most a 100 hydrogel or a polymer, whereas the mechanical clamping and spacer layers are non-porous.

In the above or other preferred embodiments, the capture layers contain furthermore a spacing material, preferable a hydrophobic spacing material to ensure separation between the channels.

Advantages of the Invention over the State of the Art

Sorting and capturing the proteins at the same time eliminates loss associated with pre-fractionation.

105 The high cost of protein microarray technology contributes to its limited popularity [7]. The invention aims to make microarrays more accessible by keeping the design flexible, simple and inexpensive. The layers in the proposed 3D stack are made of nitrocellulose and the hydrophobic barriers, which isolate the columns in the microarray, are patterned with wax. Nitrocellulose is a porous substrate with a high protein binding capacity; it is widely used in protein immobilization applications, such as lateral

110 flow tests, Western blotting and dot ELISA[14]. Proteins are irreversibly bound to nitrocellulose

membranes by electrostatic interactions [6]. The pores are small (0.45 μπι) and uniform, which is advantageous for both high resolution wax printing and vertical flow assays [14].

Xu et al. previously reported a filtration assay composed of nitrocellulose layers spotted with an array 115 of capture antibodies [ 15]. In this vertical flow device, a single sample was cycled through the stack, overcoming the diffusion limit to enhance assay sensitivity and specificity. Adding layers to the stack decreased the array footprint for the same number of tests. More recently, Ramachandran et al.

presented a flow-through membrane immunoassay (FMIA), where a single layer of nitrocellulose was pressed between two bottomless well plates with four test spots in each well [16]. The system detected 120 IgM from plasma samples pumped through the membrane. This approach has the advantage of

analysing several samples in parallel; however, both techniques require relatively large sample volumes (200 to 500 μΐ), expensive and complex robotic spotters, and cross-reactivity between detection antibodies remains a concern for high throughput applications. Chung et al. presented a unique way to transfer a microarray of tissue samples to a stack of ten membranes [17]. The protein 125 content is linearly distributed throughout the stack, and each layer can be tested individually. This approach is not compatible with densely arrayed liquid samples (e.g. blood), because only spacing isolates the samples from each other. Protein content is distributed across the membranes, meaning that each layer only represents a portion of the sample and protein detection is not multiplexed[4J. 130 An alternative method for monitoring biorecognition reactions in parallel is to pattern channels into the layers of a vertical- flow system. This approach was first demonstrated with layers of

chromatography paper for low-cost analytical devices. There are several techniques for patterning microfluidic structures in paper; for example, laser cutting[l_8], photolithography [19], chemical patterning[2Q], plasma treatment [21], wax printing[22, 23], or PDMS printing[24], many of which

135 also work for nitrocellulose membranes. We chose the straightforward and inexpensive method of wax printing with a solid ink printer. The protocol was originally developed for paper[22, 23], and was found to wet the hydrophilic zones better than paper patterned with a photoresist[9]. Recently, Lu et al. demonstrated wax printing on nitrocellulose. They reported 100-μηι wide channels, surpassing the minimum feature size in wax-printed paper and approaching what can be achieved with paper-based

140 photolithography [14, 25].

The first three-dimensional microfluidic paper analytical devices (3D μΡΑϋβ) distributed samples into multiple detection zones[23, 26]. To achieve this, they used double-sided tape resulting in an irreversible stack of patterned paper. Biorecognition in these systems is not technically multiplexed[4],

145 as each test site is exposed to only a portion of the original sample. Vella et al. showed that the 3D assembly can be used to assess liver function by detecting enzymes and total protein content from a fmgerstick of blood[9]. Origami 3D microarrays are an alternative design that does not require laser- patterned tape and allows the user to access the layers by unfolding the paper after performing the assay[27, 28]. Ge et al. incorporated several immunoassay steps into their origami μΡΑΟ by

150 systematically refolding and tearing away tabs from the paper [27]. Their device separated serum from blood to detect four cancer markers in parallel. Recently, several groups have integrated

electrochemical detection systems into origami 3D μΡΑΟβ for sensing enzymes [29^31]. Even though origami exposes the layers, nitrocellulose membranes are too brittle to fold. Like in the previous design, the samples are distributed among the test sites and therefore this system is not multiplexed.

155 While the layers are accessible, they are attached; this makes it more difficult to incubate each one in a different detection antibody. The layers in 3D μΡΑΟ≤ have also been used to introduce timing delays [32], or to quantify a reagent {i.e. hydrogen peroxide) based on the time it takes to wick through layers[33J. However, the known state of the art does not reveal any process to realize an array of multiplexed channels, as in the proposed invention.

160 Different Embodiments of the Invention

In the following different embodiments are described. Those embodiments only examples of embodiments. In particular, certain aspects of the embodiment can be combined to generate a different embodiment that is not explicitly described, below.

In one embodiment, the hydrophobic spacing material between the multiplexed affinity channels is 165 realized by wax barriers around each antibody-loaded spot or microchannel that extend through the thickness of the porous capture layer material; this allows liquid to pass through vertically while isolating samples from each other laterally. To receive liquid samples in the magnitude of few μΐ, the preferred diameter of the samples is between 10 μιη and 5000 μπι, most preferably between 100 μ ^η ι and 1000 μπι. The pore size of the porous membranes are preferably in the order of some (1-10) tenths 170 of μ-rn. The patterning of each layer can be realized by a solid ink printer, as proposed, for example in

[14] by Lu et al. The porous material could for instance be realized by nitrocellulose.

After printing alone, the wax is only on the surface; liquid would initially be confined to the spots, but would quickly diffuse out laterally as it passed through the membrane. Melting the wax pulls the pattern though the thickness of the nitrocellulose, isolating the liquid channels. 175 The melting can for instance be carried out by placing the membranes into a pre-heated oven to the melting point of the wax for some minutes. During the design of the channel size, it should preferably considered that the channels will shrink due to lateral spread of the wax. Preferably, the holes should thus be printed 30-50% larger than they have to be in the final design, when the process according to [14] is used for printing.

180 The wax patterned nitrocellulose layers are then submitted to biofunctionalization.- Whereas each square can be customized with a different target. The wax can be used to carry a label to keep track on which layer was exposed to which target.

Unused capture layers may be functionalized with blocking agents, such as, for example, bovine serum albuminum, to prevent non-specific adsorption to the porous material. To remove loosely bound 185 material, the slices can be rinsed in tris-buffered saline. The arrays could be stored dry by rinsing in ultrapure water briefly before drying with a stream of nitrogen.

In one particular embodiment, Protran B85 nitrocellulose membranes were used, which have a pore size of 0.45 μπι and 80 μ§/αη ² protein binding capacity [34] - h this embodiment, a Xerox X8560 printer was used to deposit multiple copies of the array on each sheet of nitrocellulose, to prepare 190 patterned layers for several experiments iri parallel. The print quality was set to the maximum

1200 x 1200 dpi.

In one particular embodiment, the membrane functionalisation is by passively adsorbing antibodies. This could be achieved using a simple well formed from a microscopy slide, an elastomeric ring and a metallic washer. The membrane is pressed between the slide and the washer. The elastomeric ring is 195 placed between the membrane and the washer to create the walls of the well and prevent leakage. The elastomer ring can for instance be cut from a ~ 2.5 mm thick slab of PDMS using the metal washer as a guide. The chamber reduces the amount of liquid needed to submerge the arrays and the

hydrophobic wax prevents protein adhesion outside of the channels. Protein adhesion could be improved by leaving the arrays in an oven pre-heated to the optimal temperature (e.g. 37°C).

200 In one particular embodiment, the stack is used for a direct-labelled immunoassay. The protein

solution is labelled (e.g. with a fluorescent dye, quantum dots, enzymes, or metallic nanoparticle) before injecting in the channels. Another option is to use the stack for a sandwich assay. In this case the sample would not be labelled. After injecting the protein solution, the layers are separated and each is incubated in the labelled antibody specific to the protein captured on that layer.

205 In one particular embodiment, the layers of biofunctionalised nitrocellulose are aligned with four corner pins. A hole (1 mm in diameter) is punched from a marked position in the corner of each layer. The stack is assembled by inserting the four pins into four holes in a micromachined piece of solid plastic (e.g. PMMA). The membranes are slide onto the pins, followed by an micromoulded elastomer (e.g. PDMS). In addition to four holes for the alignment pins, the l-nim thick elastomer contains an

210 array of holes matching the wax pattern on the nitrocellulose. Several elastomer layers can be stacked to adjust the reservoir volume of the system. The final layer is another micromachined piece of solid plastic. In one embodiment, this sheet has tapered inlet holes to facilitate pipetting samples into the channels. The sample could be pulled through the channels with a centrifuge (129 x g). After sample incubation, the stack is disassembled and the layers can be handled with tweezers.

215 In one particular embodiment the detection antibodies (or directly- labelled sample) are imaged with a fluorescence microscope. The membrane layers could be clamped between two microscopy slides, keeping them flat for automated imaging (e.g. with a microarray reader). Description of the Drawings

220 Figure 1 : A schematic of one embodiment of the device, the 3D microarray. A) Capture antibodies are passively adsorbed by irreversible electrostatic attraction to the nitrocellulose membranes. The nitrocellulose is patterned with wax, creating hydrophobic barriers between the antibody loaded spots. B) Layers of patterned and biofunctionalised nitrocellulose are stacked to form an array of multiplexed affinity columns. C) The 3D microarray is assembled within a device to aid alignment and sample

225 injection. Capture layers are exemplarily clamped between two micromachined pieces of PMMA.

Spacer layers prevent leakage between the PMMA and the nitrocellulose and act as a reservoir for the sample. Samples are pipetted into the channels using the tapered inlets. The device as exemplarily shown can analyse 25 samples simultaneously; the cut away side view shows one row of 5 channels.

Figure 2: Images of wax printing on nitrocellulose for a particular embodiment of the invention. A) A 230 section of the Adobe Illustrator file for preparing patterned nitrocellulose layers in parallel. B) Images of the front and back of the nitrocellulose membrane before and after melting the wax. Initially the wax is only on the surface of the membrane and as it melts the hydrophobic barriers extend through to the backside. C) Microscopy images of one spot before (left) and after (right) melting. Lateral diffusion causes the spots to shrink as the wax melts. Scale bar = 200 μπι.

235 Figure 3: Picture of the assembly of one embodiment of the invention: A) Capture Layers are

functionalised in an incubation chamber formed by clamping a doughnut-shaped piece of PDMS between a metallic washer and a microscope slide. B) After functionalisation the arrays are stacked to form the affinity columns. Four circles, printed in the corners of the nitrocellulose, are manually removed with a 1 -mm punch. The arrays are then slid over alignment pins with the same diameter,

240 which are inserted into the bottom piece of PMMA. C) The PDMS layers and top piece of PMMA are aligned with the pins and the system is held together with two binder clips.

Figure 4: Stack-and-Separate immunoassay. A) A schematic of the four-layer stack for evaluating the design. The third layer is functionalised with rabbit IgG and the other three only contain BSA. The sample, 1 μΐ of anti-rabbit IgG Alexa Fluor 488, is injected into the channels. B) Fluorescence images

245 of each layer in the stack. The 3D microarray is disassembled and rinsed before imaging each layer separately. As expected, the anti-rabbit IgG bound to the third layer in a checkerboard pattern.

Fluorescent rings appear around the spots on the first layer of the 3D microarray stack and to a lesser extent on the subsequent layers. Reflection images indicate that this binding is on the wax edge surrounding the exposed nitrocellulose. The same ring-effect is observed in a sandwich assay when the

250 labelled-antibody is added after disassembling the stack

Figure 5: Fluorescence images of individual spots from a sandwich assay experiment. The two rows represent two repeats. In both cases, the second layer in the stack was functionalized with anti-mouse IgG (pictured) and the other three blocked with BSA (not shown). In the first column (A and C), 0.3 μg/mL of mouse IgG was injected into the channels. The second column (C and D) is the negative

255 control, containing only BSA. After running the 3D assay, the stack is disassembled and the layers are incubated in anti-mouse IgG Alexa Fluor 488. This fluorescent antibody binds to the border between the wax and the nitrocellulose, most likely because of the dye. The scale bar is 200 μπι. The second row highlights variations in alignrnent between experiments. While the corner-pin alignment is sufficient for the sample to pass through a stack with four layers of nitrocellulose, some samples bind Figure 6: A multiplexed array of affinity columns. A) The four layers in the stack were functionalised with: BSA, mouse IgG, rabbit IgG, and BSA. The top view shows the injection pattern of the directly- labelled antigens. We prepared three different samples: anti-mouse IgG Alexa Fluor 488 (binding to layer 2), anti-rabbit IgG Alexa Fluor 488 (binding to layer 3) and a combined sample (binding to both 265 layers 2 and 3). Samples containing anti-mouse IgG were injected to form and 'M' and samples

containing anti-rabbit IgG to form an 'R'. After injecting the samples, the stack is disassembled and the layers are rinsed and imaged separately. B) Fluorescence images of each layer in the stack. C) The bar graph shows the average intensity from the three samples and the empty channels for each layer. The error bars are the standard deviation between spots containing the same sample.

270 Figure 7: Limit of detection for a sandwich assay detecting mouse IgG. A) The fluorescence image is from the second layer in the stack, which was functionalised with anti-mouse IgG. Eight

concentrations of mouse IgG (ranging from 1781 pM to 10 pM, spiked into BSA) were injected sequentially three times. The stack was then disassembled, rinsed and incubated in fluorescent anti- mouse IgG. The wax background was removed from the image to visualise the low concentration

275 samples. The scale bar = 500 μπι. B) The average dose response curve from three independent

experiments. The black dotted line was used to determine the limit of detection and was calculated from the average intensity of 0 pM of mouse IgG added to 3x its standard deviation. Each point is the average signal-to-background of nine spots, taken from three different experiments. The error bars are the standard deviation between experiments.

280 Figure 8: Schematic of the Device

Figure 9: Images of another embodiment of the device: A) Materials used to make the microarray stack. B) A stack of paper arrays, aligned and held together with a temporary adhesive, pressed between a Kimwipe™ and a PDMS seal and supported by a microscope slide. C) The assembled device including an inlet extension to support larger volumes (e.g for rinsing). D) Side view of the 285 assembled device.

Figure 10. Concentration series of fluorescent IgG passively adsorbed to patterned cellulose-based filter paper. A) A typical fluorescence image of one slice from the stack (slice #1). B) Concentration series for three stacked slices. The antibody in this example is non-specifically adsorbed to the fibres, which causes the intensity to drop as the antibody passes through the layers. Each point represents the 290 average of five spots. Error bars are the standard deviation.

References

1. M. Hartmann, J. Roeraade, D. Stoll, M. Templin and T. Joos, Analytical and Bioanalytical 295 Chemistry, 2009, 393, 1407-1416.

2. N. Schneiderhan-Marra, A. Kirn, A. Doettinger, M. Templin, G. Sauer, H. Deissler and T. O. Joos, Cancer Genomics & Proteomics, 2005, 2, 37-42.

3. H. J. Lee, A. W. Wark and R. M. Corn, Analyst, 2008, 133, 975-983.

4. R. Wilson, Expert Review of Proteomics, 2013, 10, 135-149. 300 5. R. J. Caiazzo, Jr., A. J. Maher, M. P. Drummond, C. I. Lander, O. W. Tassinari, B. P. Nelson and B. C. S. Liu, Proteomics Clinical Applications, 2009, 3, 138-147.

6. V. Espina, A. I. Mehta, M. E. Winters, V. Calvert, J. Wulfkuhle, E. F. Petricoin and L. A. Liotta, Proteomics, 2003, 3, 2091-2100.

7. H. Sun, G. Y. J. Chen and S. Q. Yao, Chemistry & Biology, 2013, 20, 685-699.

305 8. J. LaBaer and N. Ramachandran, Current Opinion in Chemical Biology, 2005, 9, 14-19.

9. S. J. Vella, P. Beattie, R. Cademartiri, A. Laromaine, A. W. Martinez, S. T. Phillips, K. A. Mirica and G. M. Whitesides, Analytical Chemistry, 2012, 84, 2883-2891.

10. P. Hudler, M. Gorsic and R. Komel, Clinical & Experimental Metastasis, 2010, 27, 441 -451.

11. M. Bally, M. Halter, J. Voros and H. M. Grandin, Surface and Interface Analysis, 2006, 38, 310 1442-1458.

12. J.-R. Lee, D. M. Magee, R. S. Gaster, J. LaBaer and S. X. Wang, Expert Review of

Proteomics, 2013, 10, 65-75.

13. L. A. Liotta and E. F. Petricoin, Journal of Clinical Investigation, 2006, 116, 26-30.

14. Y. Lu, W. Shi, J. Qin and B. Lin, Analytical Chemistry, 2010, 82, 329-335.

315 15. Y. Q. Xu and G. Bao, Analytical Chemistry, 2003, 75, 5345-5351.

16. S. Ramachandran, M. Singhal, K. G. McKenzie, J. L. Osborn, A. Arjyal, S. Dongol, S. G. Baker, B. Basnyat, J. Farrar, C. Dolecek, G. J. Domingo, P. Yager and B. Lutz, Diagnostics, 2013, 3, 244-260.

17. J.-Y. Chung and S. M. Hewitt, in Protein Blotting and Detection: Methods and Protocols, eds. 320 B. T. Kurien and R. H. Scofield, 2009, vol. 536, pp. 139-148.

18. J. L. Osborn, B. Lutz, E. Fu, P. Kauffhian, D. Y. Stevens and P. Yager, Lab on a Chip, 2010, 10, 2659-2665.

19. A. W. Martinez, S. T. Phillips, M. J. Butte and G. M. Whitesides, Angewandte Chemie- International Edition, 2007, 46, 1318-1320.

325 20. K. Abe, K. Suzuki and D. Citterio, Analytical Chemistry, 2008, 80, 6928-6934.

21. X. Li, J. Tian, T. Nguyen and W. Shen, Analytical Chemistry, 2008, 80, 9131-9134.

22. Y. Lu, W. Shi, L. Jiang, J. Qin and B. Lin, Electrophoresis, 2009, 30, 1497-1500.

23. E. Carrilho, A. W. Martinez and G. M. Whitesides, Analytical Chemistry, 2009, 81, 7091- 7095.

330 24. D. A. Bruzewicz, M. Reches and G. M. Whitesides, Analytical Chemistry, 2008, 80, 3387- 3392.

25. A. W. Martinez, S. T. Phillips, G. M. Whitesides and E. Carrilho, Analytical Chemistry, 2010, 82, 3-10. 26. A. W. Martinez, S. T. Phillips and G. M. Whitesides, Proceedings of the National Academy of 335 Sciences of the United States of America, 2008, 105, 19606-1961 1.

27. L. Ge, S. Wang, X. Song, S. Ge and J. Yu, Lab on a Chip, 2012, 12, 3150-3158.

28. H. Liu and R. M. Crooks, Journal of the American Chemical Society, 2011 , 133, 17564- 17566.

29. L. Ge, S. Wang, J. Yu, N. Li, S. Ge and M. Yan, Advanced Functional Materials, 2013, 23, 340 3115-3123.

30. J. Yan, M. Yan, L. Ge, J. Yu, S. Ge and J. Huang, Chemical Communications, 2013, 49, 1383- 1385.

31. H. Liu, Y. Xiang, Y. Lu and R. M. Crooks, Angewandte Chemie-International Edition, 2012, 51, 6925-6928.

345 32. N. Noh and S. T. Phillips, Analytical Chemistry, 2010, 82, 4181-4187.

33. G. G. Lewis, M. J. DiTucci and S. T. Phillips, Angewandte Chemie-International Edition, 2012, 51, 12707-12710.

34. Whatman. Inc., Protran Nitrocellulose Data Sheet

http://www.whatman.com/References/Protran%20Nitrocellulose%2 0Membrane%20datasheet Feb09. 350 p_df.

35. H. Hwang, S.-H. Kim, T.-H. Kim, J.-K. Park and Y.-K. Cho, Lab on a Chip, 201 1, 1 1, 3404- 3406.

36. E. Vereshchagina, K. Bourke, L. Meehan, C. Dixit, D. Mc Glade, J. Ducree and Ieee, in 26th IEEE International Conference on Micro Electro Mechanical Systems, 2013, pp. 1049-1052.

355 37. C. M. Sturgeon and E. Diamandis, in Laboratory Medicine Practice Guidelines, American Association for Clinical Chemistry, Washington, DC, 2010, pp. 1-64.

38. T. A. Wagner, C. A. Gravett, S. Healy, V. Soma, J. C. Patterson, M. G. Gravett and C. E. Rubens, Journal of Global Health, 201 1, 1, 210-223.

360