BACKGROUND A common requirement in many analytical situations is to detect binding between species. For example, many immunoassays are based on detection of binding between an analyte and a specific antibody or fragment thereof. The assay might be conducted in a variety of formats where an analyte competes with a labelled analogue for a binding site or where the analyte binds jointly to two different recognition molecules (in a so-called'sandwich assay') for example. Hybridisation assays, where the binding between complementary sequences of nucleic acids is detected, are also common.

Binding assays can be conducted in so-called'homogeneous'formats where no physical separation steps are used to separate bound and unbound species, or alternatively can involve separation. The latter approach allows potentially interfering materials to be washed away from the bound sample before measurement, and therefore minimises spurious background signals, but is less suited to automated high volume testing than simpler homogeneous procedures.

Luminescence methods are very widely used for high-sensitivity assays, but suffer from the problem that many sources of interfering background are often present in typical samples. For example, serum samples often have highly fluorescent bilirubin present, blood samples are strongly coloured and can quench emission by energy transfer or by absorption of exciting and/or emitted light. Thus, it is a challenging task to use luminescence methods to detect very low levels of analytes without some separation or washing step to minimise background.

One approach which has been taken to conduct a binding assay to detect a luminescent substance without the need to separate unbound material is to make use of selective excitation of surface-bound species. There are two common means to achieve selective excitation at surfaces. One approach is to make use of the evanescent wave that exists at an interface between media of high and low refractive index when light is totally internally reflected at the interface. The second approach is to utilise coupling of excitation energy into energy states of electrons in a metal film at the interface (surface plasmon/polariton states), and transfer of this excitation to surface bound species. It is convenient in the first instance to describe evanescent wave processes and their application to surface-binding assays.

When light incident from a medium of high refractive index n2 to a second medium of lower refractive index nl the light is totally internally reflected at the interface if it is incident at an angle equal to or greater than the critical angle (f) where sin (f) = nl/n2.

Although no energy passes across the interface in the case of total internal reflection, an electric field (the'evanescent wave') that decays exponentially with distance exists in the rarer medium. The characteristic penetration depth (d), which is the distance from the interface at which the electric field amplitude has decayed to 1/e of its surface value, is proportional to wavelength of illumination (1) and is given by d= (l /nl)/2p [sin2f (nl/n2) 2]'i2 For example, if the interface is between fused silica and water, of approximate refractive index 1.46 and 1.34 respectively, and the incident light is of wavelength 500nm and at 70 degrees to the normal, d is of the order of 150nm. The penetration depth increases as the critical angle is approached.

Although there is no energy flow across the interface as described, an absorbing species at or near the interface can accept energy from the evanescent wave. This has the effect of exciting the acceptor and attenuating the reflected component of the exciting light. If the acceptor is luminescent it can emit light that can be detected.

Although emitted light can be detected at right angles to the surface, much of the emission couples back into the higher index medium and therefore a detector is often placed on the optical axis of the exciting radiation with an appropriate optical filter to block exciting light.

Evanescent wave excitation has been applied to the construction of sensors, to immunoassays and to fluorescence microscopy. The principles and applications of evanescent wave excitation are described in'Fibre Optic Chemical Sensors and Biosensors' (Wolfbeis, O. ed CRC Press, Boston (1991) for example, and have been discussed in patent applications such as US 5745231 by Groger et al and US 3939350 by Kronick et al.

Examples of common detection geometries for evanescent wave excitation are shown in Figure 1. Figure la shows in outline the geometry used for on-axis detection. A light source (L), filtered if necessary by an optical filter (fui) to isolate required excitation wavelengths, illuminates a waveguide (W) at an angle such that total internal reflection is achieved. In the medium (M) which is of lower refractive index than (W) an evanescent wave excites bound luminescent label (LA), the luminescence from which is filtered by optical filter (F2) to block the exciting radiation, collected by the optional lens (LE) and detected by (P) which is a photosensitive device such as a photomultiplier or semiconductor diode. An alternative on-axis geometry (not shown) uses an optical fibre as a waveguide. This is commonly used for sensor applications, detecting fluorescence coupled into the fibre, usually by means of a dichroic beam splitter which transmits the exciting radiation but reflects luminescence into a detector.

For detection off-axis the geometry of Figure lb is used where the labelling letters have the same significance as in Figure la. The lens (LEa), filter (F2a) and detector (Pa) illustrate the alternative position if luminescence is to be detected in the rarer medium. It is to be understood that these units will be in air, even if the surface of the waveguide is covered with aqueous assay medium.

Although the evanescent wave excitation offers advantages there are also some potentially serious problems with this approach which limit applicability. For example, * If off-axis (e. g. right-angle) geometry (Figure lb) is used to detect the emitted luminescence this is relatively inefficient because most of the luminescence couples back into the waveguide.

* Exciting light might leak from the waveguide, coupled by surface imperfections, dust or scratches for example. This light can excite luminescence from the bulk medium, rather than from a surface-associated layer. Thus, background will be excited efficiently by light leaking from the waveguide. Such luminescence will be detected efficiently if the right-angle geometry of Figure 1 b is used.

* If in-line geometry (Figure la) is used to detect emitted luminescence the optical filters used to block exciting light must be very efficient since the reflected exciting light is very much more intense than any luminescence.

* Exciting light striking the blocking filter commonly excites some luminescence from the filter. The level of such unwanted emission will often limit the detection sensitivity that can be achieved.

* Exciting light might give rise to luminescence from the waveguide, or Raman scattering might be significant at high excitation levels. If ultra-violet excitation is used, luminescence of the waveguide is potentially a serious problem for very high sensitivity detection. In practical measurement formats it might be necessary to use plastic (e. g. polystyrene or acrylic) waveguides for economy and ease of manufacture. Plastics are often luminescent as a result of traces of plasticiser or unpolymerised materials within the matrix.

* Laser excitation is very convenient since it is highly monochromatic and directional and often polarised and this facilitates optical rejection of scattered light. However most lasers are expensive and bulky.

* If lifetime resolution is used with pulsed or modulated excitation and long-lived luminescent labels in an effort to minimise background, the long-lived emission often seen from solids such as waveguide and filters can limit sensitivity.

These and other potential problems limit the widespread application of evanescent wave excitation to luminescence sensing.

One recent approach which minimises some of the above problems is described by Gryczynski et al in'Two-Photon Excitation by the Evanescent Wave from Total Internal Reflection' (Analytical Biochemistry, Vol. 247 No. 1 pages 69-76 (1997)).

This paper describes evanescent wave excitation of the fluorescent calcium-binding dye Indo-1 using 770nm radiation from a Ti-sapphire laser. The two-photon approach limits the effective distance from the interface at which labels can be excited efficiently because of the square-law dependence of excitation on the electric field strength in the evanescent field.

Two-photon excitation in the far red or near infra-red spectral regions avoids the problem of excitation of background by light leaking from the waveguide into the bulk medium. The square-law dependence on light intensity prevents significant excitation by low-level illumination. It is also relatively straightforward to discriminate against scattered long wavelength light using efficient optical filters to isolate luminescence of interest. This facilitates the use of on-axis detection, where low levels of fluorescence must be seen against a very large background of exciting light. On-axis detection is usually efficient when fluorescence is excited by an evanescent wave, and is almost always used where the waveguide is an optical fibre. For two-photon excitation however the exciting light intensity can be so high that two-photon excitation of luminescent species in the optical filters might be a problem. This is minimised by placement of filters in regions where the exciting beam is expanded to minimise the local power density.

Although the two-photon technique described is useful, it does not avoid the problem of excitation of unwanted background luminescence by the evanescent wave. Materials in the analyte might well have substantial two-photon excitation cross sections, similar to that of the label for example. In addition, there is the possibility of excitation of luminescence within the waveguide itself by multiphoton processes. This is particularly likely where a polymer waveguide is used which might have impurities, traces of monomer or plasticisers present. Very intense laser pulses can also stimulate inelastic scattering processes which might give rise to interfering signals from the waveguide. One of the main disadvantages from a practical viewpoint is the need for a laser source for efficient excitation in most cases. For example, the cited reference used a femtosecond pulsed laser, which is extremely expensive and physically large.

These problems arise from the nature of the excitation process, which proceeds via so- called'virtual'states of very short lifetime. This means that the instantaneous peak power required for efficient excitation is very high so that a powerful fast-pulsed laser is required. It is possible to excite two-photon transitions with a continuous laser beam, but this is normally inefficient and requires extremely efficient filtering to reject the exciting radiation.

An alternative means to excite luminescence at a surface has been used, based on generation of surface plasmons and related surface phenomena in thin metal films at an interface. When light of p-polarisation is incident from a dielectric medium onto a thin metal film at the interface with a second medium of lower refractive index, under appropriate circumstances energy from the light beam can be coupled into quantised oscillations of the electron plasma within the metal. This in turn gives rise to an evanescent field penetrating into the rarer medium. In order to couple energy efficiently into the surface states the incident radiation must strike the interface at an appropriate angle and the film must be of an appropriate thickness. A periodic grating can be used instead of a continuous film to similar effect. The application of surface plasmon resonance to surface binding assays has been described in a number of publications and patents, for example US 5485277 by Foster.

A luminescent molecule placed in the evanescent field region can accept energy from the surface states and become electronically excited. If the molecule is spaced a few nanometres from the film, so that the luminescence is not quenched by energy coupling back into the metal, emission can be stimulated efficiently by surface plasmon effects.

The application of surface plasmon enhancement for coherent two-photon excitation of fluorescence is discussed in'Two-Photon-Excited Fluorescence Enhanced by a Surface Plasmon'by H. Kano and S. Kawata (Optics Letters Vol. 21 No. 22.

November 1996, pages 1848-1850).

The effective electric field in the evanescent wave near the interface can be markedly enhanced relative to that in the higher refractive index medium. For example, the cited reference describes calculations for a silver film of thickness 4.8nm at the interface between glass of refractive index 1.5 and air (refractive index 1) excited by a light source of wavelength 800nm. At the resonance angle, the electric field in the near surface layer is enhanced 300-fold relative to that in the glass. Relative to excitation with an evanescent wave at an uncoated surface, the surface-plasmon- enhanced excitation was found to be 90-fold more efficient for two-photon excitation of a fluorescent dye.

The cited reference describes two-photon excitation mediated by a virtual state. A process such as this requires high light intensity for efficient operation, and therefore requires expensive and bulky laser excitation. As with the evanescent wave excitation method cited previously, there is little selectivity against excitation of background, since unwanted luminescent species in the medium might be excited as efficiently as the label species of interest. Thus, although the surface plasmon effect is of considerable interest as a potential means to enhance the sensitivity of luminescence detection at surfaces, it does not solve the problems of a practical robust sensor device.

It is a purpose of the present invention to minimise and potentially to avoid the limitations of evanescent wave sensing and of sensors based on excitation of luminescence via excited surface states such as plasmons.

SUMMARY OF INVENTION According to the present invention there is provided a method of conducting a luminescence assay to detect the surface association of a luminescent species wherein said luminescent species is provided as a label that is bound to or capable of being bound to the surface of a waveguide that is a medium of higher refractive index than its surroundings or is a medium coated with a thin metal film or grating within which electronically-excited surface states such as surface plasmons can be stimulated by appropriate illumination the luminescent species and the waveguide are contacted together with a medium possibly containing a species able to influence their binding the waveguide is illuminated with a source of light so as to give rise either to an evanescent wave penetrating from the surface or to excitation of said surface states and luminescence is detected in an emission region characteristic of the luminescent label and distinct from the exciting radiation said method being characterised in that the luminescent species is stimulated by absorption of photons of the illuminating radiation, giving rise to luminescence of shorter wavelength than the said radiation either 1) by further excitation by the illuminating radiation of metastable excited states populated by previous exposure of ground state species to a source of excitation energy or 2) by pooling of excitation energy from two or more species excited by single photon absorption of the illuminating radiation, either via further excitation of one or more of the said species or through stepwise transfer of excitation energy to one or more further species which emits said luminescence Further features of the invention are defined in the appended claims.

The invention therefore relates to methods to excite and detect upconverted luminescence produced by either of the mechanisms (1) or (2) above. Mechanism (1) relates primarily to labels such as energy-storage phosphors. These phosphors are commonly alkaline earth sulphides doped with ions such as cerium, europium and samarium. Excitation of the label occurs by absorption of visible or UV light leading to a population of energetic electrons trapped at sites within the particle matrix. The electrons can be released from the traps by illumination, typically with infra-red light, and recombine with'holes'giving visible luminescence. Trapped electrons can be stored for many hours or more in a typical storage phosphor in absence of infra-red stimulation. Such phosphors are usually unstable to moisture and require some type of coating or encapsulation if they are to be used in an aqueous medium. Phosphors of this type have the advantage that they emit visible light with relatively high efficiency on stimulation with infra-red light of low intensity.

Mechanism (2) relates primarily to upconverting glasses and crystals doped with ions such as erbium, thulium, holmium terbium and dysprosium. Ions such as these can be excited in a stepwise fashion by infra-red light of wavelengths matching the transitions of the ions concerned. These phosphors are commonly sensitised by including ytterbium ions in the matrix as a primary absorber, able to be excited at around 960nm and to transfer the excitation energy in a stepwise fashion to the luminescent lanthanide ions. The emission from an upconverting agent of this type is a non-linear function of excitation intensity as a result of the multiphoton nature of the excitation process.

Consequently the invention does not relate to excitation mediated by two-photon or multiphoton excitation by simultaneous absorption of two or more photons of low energy mediated by so-called'virtual'excited states. Two-photon and multiphoton excitation by the latter mechanism requires that the illumination has a very high power density if excitation is to be reasonably efficient. Such multiphoton excitation can only be achieved efficiently at present using very expensive laser sources, and has the further disadvantage that the excitation is intrinsically able to excite background as well as emission from the target luminescent label.

The method of the invention is applicable particularly to immumoassay and hybridisation assays where surface binding is to be detected. Typically a waveguide (which might for example be an optical fibre from which the surface cladding has been partially removed) is coated with a probe species that is able to bind an analyte of interest. Such coating might be by physical adsorption (e. g. if the probe is an antibody or other protein or a DNA sequence) or by covalent coupling. In the latter case the waveguide might require surface activation to provide suitable binding groups. Glass surfaces are often activated by coupling silanes such as 3- mercaptopropyltriethoxysilane or 3-aminopropyltriethoxysilane to the hydroxylic surface. Surfaces coated with gold or silver, which are used for surface-plasmon mediated excitation, can be activated by binding thiols such as mercaptoethanol, cysteine or 3-mercaptopropyltriethoxysilane. Binding of probe molecules to the functionalised surfaces follows traditional routes that are familiar from surface chemistry. For example, carboxylic acid groups on proteins can be linked to surface- bound amines or organic hydroxyl groups using water-soluble carbodiimides.

In use the waveguide bearing the probe molecules along with a label species is exposed to a solution thought to contain the analyte. The analyte might be one that causes the label to bind to the surface, or alternatively might displace a bound label. After a suitable incubation period the waveguide is illuminated with long-wavelength light and luminescence characteristic of the label is detected either within the waveguide (e. g. if the waveguide is an optical fibre) or from the surface of the waveguide.

DESCRIPTION OF DRAWINGS Fig 1 a illustrates in-axis detection of evanescent wave excitation; and Fig lb illustrates off-axis detection of evanescent wave excitation.

DESCRIPTION OF PREFERRED EMBODIMENTS Examples of labels able to be excited from metastable levels are the so-called electron-trapping or'storage'phosphors. These materials are commonly used to detect infra-red radiation by visible luminescence, and typically are inorganic substances such as strontium and calcium sulphides doped with other ions such as samarium. The electron-trapping phosphor is a semiconductor that can be excited by prior exposure to visible or ultra-violet light, giving rise to very long-lived metastable excited states where electrons are trapped within local potential energy minima below the conduction band of the semiconductor. Exposure to infra-red light of appropriate wavelength can excite the electrons releasing them from these'traps'and visible luminescence is emitted as the electrons fall back to the ground state energy band.

Although the excitation is in the infra-red region of the spectrum, the emission is stimulated by absorption of single photons and does not require energy pooling or multi-photon excitation once the semiconductor has been'charged'by exposure to short wavelength light to populate the energy'traps'.

Electron trapping phosphors are appropriate to the present invention as labels, though the materials commonly used are chemically reactive and easily hydrolysed on contact with aqueous media. Consequently, the labels require coating or encapsulation, for example in a polymer matrix, to maintain compatibility with the aqueous assay medium. Such coating or encapsulation also offers the opportunity to provide reactive groups to bind recognition ligands such as antibodies, lectins, oligonucleotides and nucleic acids, carbohydrates, biotin etc., or to allow appropriate ligands to be bound to the particles by physical adsorption.

Electron trapping phosphors have the disadvantage that the materials emit a low-level background immediately after'charging'by exposure to UV or visible radiation. It is usually necessary to wait for a period of time after such exposure before the materials are used to detect infra-red radiation to allow this spontaneous (thermally-activated) background to decay to an acceptable level.

An alternative upconversion material that is very well suited to the present invention is an energy-pooling medium. Luminescent materials that can be excited by this means have been known for many years, and are described for example by Soules and Hoffman in the article'Luminescent materials (Phosphors)'in the Encyclopedia of Chemical Technology Vol. 14 (1981). Typically these materials use a sensitising species such as ytterbium ion to absorb infra-red radiation and an emissive ion that can accept energy from excited sensitisers in a multi-step process through intermediate metastable excited states. Commonly ions such as erbium and thulium are used as luminescent acceptor species. Applications of such materials as labels is described by Zarling et al in US patent 5,736,410 (1998). In WO-A-98/43072 (PCT/GB 98/00 769) we describe applications of these and other upconversion media inter alia as components of homogeneous assays based on transfer of energy to other species in close proximity. The present invention provides efficient means to excite labels for applications as described in our WO case and the subject matter thereof is included herein by reference.

Labels as described are able to be excited efficiently by relatively low levels of far red and near infra-red light, which excite no detectable background either from waveguides or from any material normally present in an assay medium. A light source such as a very low-cost semiconductor light-emitting diode as is used in a TV remote control for example is able to excite detectable luminescence from upconversion phosphors. For some applications it is convenient to use a laser diode for excitation, but these are small devices using far less power and costing much less than alternative lasers.

Excitation using any source can be continuous, or can be pulsed and/or modulated if required. In the latter case time-resolved detection techniques or, where excitation is modulated, related frequency-domain techniques well known in the art can be used to enhance detection sensitivity. Emission from some upconverting labels can be chosen to have a decay time in a particular range, so that lifetime resolution can be used further to reject background. There will usually be little need for such methods to reject background luminescence generated by the stimulating radiation in the present invention. However, in some circumstances the sample might exhibit significant background from other causes. For example the sample might show long-lived phosphorescence or chemiluminescence. In the example of labels based on electron- trapping phosphors, there is often a weak thermally-activated background emission from the bulk sample, and this can be rejected efficiently if the exciting radiation is pulsed or modulated in a defined manner with lock-in detection.

Where the label of interest has a characteristic decay time, lifetime-resolved detection can be convenient to minimise electrical'pickup'or cross-talk between the driver for the excitation source and the detector, for example. Equally, driving the excitation source with a modulated or pulsed waveform allows lock-in detection which narrows the detection bandwidth and thereby provides enhanced rejection of'noise'and other unwanted signals such as stray light.

It is straightforward to detect luminescence from surface-bound labels either on the optical axis or at an angle to the waveguide. In the former case, which is often appropriate to evanescent wave excitation, it is usually a simple matter to filter out the exciting long-wavelength radiation. Detectors such as photomultipliers often have very low sensitivity in the far red or near infra-red and this minimises problems of traces of exciting light that might leak through optical filters.

On-axis detection is often used for evanescent wave excitation of conventional fluorescent labels by single-photon absorption, because the greater part of the emission couples back into the waveguide. However this need not be such an important effect in the case of the present invention. The excitation process in the case of labels such as the energy-pooling upconversion media allows delocalisation of energy. The emitting species that accepts the excitation need not be aligned favourably to couple energy back into the waveguide. In addition, the labels are not free single molecules, but are in a particulate matrix. This might assist in transmitting emission into the aqueous medium, depending on the shape and refractive index of the particle. Off-axis detection is also appropriate for excitation schemes where surface plasmons are used to enhance excitation, since it is usually undesirable to allow excited states of the label to couple energy back into the surface film.

A semiconductor laser is convenient as an excitation source for the present invention since its output can be coupled easily into a waveguide at an appropriate angle to excite an evanescent wave and with an appropriate polarisation for surface plasmon excitation. However, the invention can also be used with a semiconductor LED. As discussed by J. Wilson and J. F. B. Hawkes in'Optoelectronics: An Introduction' (Prentice-Hall International, ISBN 0-13-638353 X (1983)), semiconductor LEDs based on materials of high refractive index such as gallium arsenide are very efficient internally in generating photons. However, very little light is coupled out of the chip unless an index matching medium is used to couple the light into the rarer medium.

For a plane waveguide of GaAs at a wavelength where the medium has a refractive index of 3.6 within which light is generated isotropically these authors calculate only 1.3% of the radiation will escape into air. Although this effect is a disadvantage in normal use of LEDs, for the present invention it is potentially beneficial because the devices are inherently very suited to evanescent wave applications. Thus, an infra-red LED equipped with a high-index protective plane layer can be used directly as a waveguide to excite the labels of the invention. A simple way to provide this is to cut or grind the transparent encapsulation of the LED to give a plane surface for use as the waveguide. Ideally, a transparent protective material with a refractive index as close as possible to that of the semiconductor would be used for best efficiency, though this is not essential. The surface can be coated with a thin layer of metal such as silver or gold if surface enhancement is required. Although much of the radiation will be unable to excite surface states in the metal there will be a component of appropriate angle and polarisation to achieve this. Surface plasmon excitation allows efficient measurement of luminescence emitted into the rarer medium, and this might be advantageous where a detector is used that cannot readily be closely-coupled to the waveguide. However, as mentioned previously, the use of the labels of the invention does not automatically give rise to efficient coupling of luminescence back into a waveguide, so that off-axis measurement might also be appropriate to standard evanescent wave excitation. Where a semiconductor detector is integrated into the waveguide it is likely that the detector will respond to the emission of the LED source unless an optical filter can also be integrated. In this case it is likely to be beneficial to use lifetime-resolution of luminescence with fast pulsed and/or modulated excitation to avoid this problem. Many of the labels appropriate to the invention have lifetimes of at least microseconds, so that efficient lifetime resolution is straightforward.

The detector of the present invention can be any photosensitive electronic device including, but not limited to photodiodes and photomultipliers, intensified detectors, semiconductor sensors such as CCD devices etc. The detector can optionally be an imaging detector. For example a CCD array or intensified camera can be used to make measurements of samples in parallel viewing either a group of individual waveguides or a group of individual samples in defined regions on the surface of a single waveguide. It is often advantageous if the detector can be pulsed or modulated in sensitivity for applications where excitation is pulsed or modulated as described earlier. This can be achieved readily for photomultipliers by pulsing or modulating the photocathode or dynode voltages for example. Image intensifiers can also be gated and modulated in this way, as described in our earlier patent EP-A-0 519 930 relating to lifetime-resolved detection which is incorporated herein by reference.

Alternatively, the detector can be equipped with a fast optical gate such as a ferroelectric liquid crystal light valve, or the output from the detector can be modulated or gated in appropriate circumstances to facilitate lock-in detection or similar detection schemes. CCD cameras can be operated as phase-sensitive imagers for lifetime-resolved measurements as disclosed in our earlier paper'New Approaches to Lifetime-Resolved Luminescence Imaging', C. G. Morgan, A. C. Mitchell, J. G.

Murray and E. J. Wall, I. Fluorescence, 7 (1), 65-73,1997, and such cameras are available commercially from Photonic Research Systems Ltd in the UK.

For high sensitivity detection, photon-counting is often used. Suitable detectors include photomultipliers and avalanche photodiodes. the latter are physically very small and can be integrated into a waveguide geometry if necessary. Avalanche photodiodes often have rather high dark count rates, and it is convenient, though not essential, to use them in conjunction with a pulsed or modulated source with appropriate lock-in detection or related signal processing means.