WO2008056160A1 - Microfluidic device and assay for the quantitation of nucleic acid analytes based on hybridization - Google Patents

Abstract

To facilitate quantitative analysis of analytes in samples, a sample can be applied to one or more immobilised reagents along the direction of a flow path. As analytes interact with immobilised reagents, their concentration in the sample is reduced, and so the signal intensity for a particular analyte decreases along the path. By analysing the change in signal intensity along the path (typically a decrease), the original concentration of the analyte in the sample can be determined.

Description

MICROFLUIDIC DEVICE AND ASSAY FOR THE QUANTITATION OF NUCLEIC ACID ANALYTES BASED ON HYBRIDIZATION

All documents and on-line information cited herein are incorporated by reference in their entirety,

TECHNICAL FIELD

This invention is in the field of analysing data from solid phase assays. BACKGROUND ART

DNA and protein arrays are now routinely used for analysing biological samples. They generally have a solid support, such as a microscope slide, to which multiple DNA molecules and/or proteins are immobilised in multiple sites. A sample is incubated across the array, contacting all sites simultaneously. After incubation, the solid support is investigated to determine which sites are occupied, thereby indicating which analytes were present in the sample.

These experiments are carried out under conditions such that only a fraction, usually a small (and almost invariably unknown) fraction, of the target analytes binds to the array. Furthermore, it is not usually known for any given probe/target combination whether the target or the probe is in excess. These factors make it difficult to determine the absolute concentration of targets in the sample and it is normal practice to design experiments so that a target's concentration is measured relative to another target.

There are two ways in which this is done. Two samples to be compared may be hybridised separately to two arrays. In this case, the hybridisation levels of each target will be normalised with reference to a common standard within each sample. Such standards may be intrinsic standards, such as a mRNA for a housekeeping gene, or added as a 'spike'. Alternatively, the targets in the two samples may be labelled with dyes of different colours and mixed before they are hybridised to the microarray. Both methods have drawbacks.

It is an aim of the invention to permit direct measuring or estimation of the absolute amount of analytes in a sample.

DISCLOSURE OF THE INVENTION

To facilitate quantitative analysis, a sample can be applied to one or more immobilised reagents along the direction of a flow path. As analytes interact with immobilised reagents, their concentration in the sample is reduced, and so the signal intensity for a particular analyte decreases along the path. By analysing the change in signal intensity along the path (typically a decrease), the original concentration of the analyte in the sample can be determined. In contrast to current microarray protocols, therefore, the invention does not simultaneously apply a sample to all immobilised reagents, but sequentially applies the sample to reagents. In this manner, quantitative information can be derived. Thus the invention provides a method for analysing a sample containing one or more analyte(s), comprising steps of: (i) applying the sample to a device having one or more immobilised binding reagent(s), where the sample is applied at a proximal end of a path and moves in a direction to a distal end of the path, such that analytes in the sample can interact with the immobilised binding reagents; (ii) determining an amount of interaction between an analyte and a binding reagent at two or more points on the path; and

(iii) comparing the amounts of interaction at the points and, based on the comparison, inferring the amount of analyte in the sample.

A device can include binding reagents for analysing multiple analytes along a single path. It can also include multiple paths. Thus the methods of the invention can be used for the parallel analysis of multiple analytes in multiple samples.

The invention also provides a method for applying a liquid sample containing a plurality of analytes to a device having a plurality of immobilised binding reagents, wherein the whole volume of the sample is applied to each binding reagent. The sample may be applied along a path, and continuous flow of the sample along the path is preferred. Thus the sample should be applied to the device without interrupting the liquid flow e.g. without introducing gas bubbles into the flow.

The analytes

The invention is suitable for the analysis of various analytes in various types of sample. Typical analytes are nucleic acids, including DNA and RNA (e.g. mRNA), and proteins. The samples may be derived from various sources e.g. from eukaryotes, from prokaryotes, etc.

Typical samples will be lysates of a population of cells or of tissues. In some embodiments, particularly where a high proportion of analyte is captured, the invention is particularly useful for analysing material derived from single cells e.g. in conjunction with the details disclosed in reference 1.

Analytes in a sample may be labelled prior to performing the method e.g. if transcription takes place in the presence of labelled RNA precursors then RNA can be applied to a device and hybridised label can be detected. As an alternative, analytes may be labelled during the method e.g. a sample can be applied to a device, and then captured analytes can be labelled in situ.

The immobilised binding reagents

In methods of the invention, analytes in a sample flow along a path, such that they can interact with one or more binding reagent(s) immobilised on a device.

The immobilised binding reagents can include nucleic acids for hybridisation, antibodies for antigen binding, antigens for antibody binding, lectins for capturing sugars and/or glycoproteins, sugars for capturing proteins, small organic compounds (e.g. <1000 Da), etc. Preferred binding reagents are specific for a chosen target e.g. a nucleic acid sequence for specifically hybridising to a target of interest, an antibody for specifically binding a target antigen of interest. The degree of specificity can vary according to the needs of an individual experiment e.g. in some experiments it may be desirable to capture only nucleic acids that are precisely complementary to an immobilised probe, but in other experiments it may be useful to capture also a target which includes nucleotide mismatch(es), etc.

Analytical reagents (binding reagents) may be immobilised on the device in various ways. For instance, they may be immobilised on the surface of a solid support, as in a conventional DNA microarray, usually via a linker. Thus the path in step (i) will run along the surface of the support. As an alternative, reagents may be immobilised within a porous substrate (e.g. a gel), in which case the path in step (i) will run through the substrate (e.g. through a gel). Other arrangements are also possible.

A preferred binding reagent is a nucleic acid for hybridising to and capturing nucleic acid analytes within a sample. A path will usually include a series of different immobilised nucleic acids. The sequence of the immobilised nucleic acids will be chosen according to the target analytes of interest. More preferably, the analytical components retain specific mRNA transcripts. The immobilised nucleic acids are preferably DNA, are preferably single-stranded, and are preferably oligonucleotides (e.g. shorter than 200 nucleotides, <150nt, <100nt, <50nt, or shorter). Retention of mRNA rather than DNA by such reagents can conveniently be achieved by removing DNA form a sample before analysis.

Other useful binding reagents can capture proteins from a sample. These reagents will typically be immunochemical reagents, such as antibodies, although other specific binding reagents can also be used e.g. receptors for capturing protein ligands and vice versa. Techniques for the specific capture of proteins by immobilising reagents to solid surfaces are well known in the art e.g. from ELISA, and for surface plasmon resonance, protein arrays, antibody arrays, etc. To detect binding in an immunochemical assay format then it is may be necessary to use a second antibody (e.g. in a 'sandwich' assay).

The analytical components in any particular device will generally be chosen based on knowledge of the sample to be analysed by that device, and on the desired results.

A device may include reagents for analysing both nucleic acids and proteins.

Methods for immobilising analytical binding reagents onto surfaces are well known in the art. Methods for attaching nucleic acids to surfaces in a hybridisable format are known from the microarray field e.g. attachment via linkers, to a matrix on the surface, to a gel on the surface, etc. The best-known method is the photolithographic masking method used by Affymetrix for in situ synthesis of nucleotide probes on a glass surface, but electrochemical in situ synthesis methods are also known, as are inkjet deposition methods. Methods for attaching proteins (particularly antibodies) to surfaces are similarly known.

Immobilised nucleic acids can be pre-synthesised and then attached to a surface, or can be synthesised in situ on a surface by delivering precursors to a growing nucleic acid chain. Either of these methods can be used according to the invention.

To avoid the possibility that an analyte in a sample could flow past an immobilised reagent on a path, reagents are preferably arranged such that they occupy the full width of a path (Figure 7). Furthermore, immobilised reagents are preferably arranged such that they are encountered singly along a flow path, rather than simultaneously (Figure 8). Different immobilised binding reagents are preferably arranged in discrete cells or patches, to facilitate data analysis, because including different reagents within the same area means that it may not be clear which of the reagents gives rise to a signal. Patches can take various shapes (e.g. rectangular, square, elliptical, circular, etc.), including irregular shapes.

In some embodiments of the invention ('mono-patch embodiments'), a reagent for analysing a particular analyte will be included only as a single patch. In other embodiments ('multi- patch embodiments'), a reagent for analysing a particular analyte will be included in a series of patches (e.g. 2, 3, 4, 5, 6, 7, 8, 9, 10 or more patches of reagent for the same analyte).

In mono-patch embodiments, a signal gradient will be generated along the path's intersection with that patch. The shape of the gradient can be used to infer the analyte concentration in the original sample, as described in more detail below.

In multi-patch embodiments, a gradient will be generated on each patch, but a patch-to- patch gradient will also be generated. One or both of these gradients can be used to infer the analyte concentration in the original sample, as described in more detail below. In these multi-patch embodiments, it is preferred that all of the patches for a single reagent should have substantially the same shape, size and reagent density. Having substantially identical patches in this way makes it easier to convert the gradient information into quantitative data.

As an analyte flows over a patch, a proportion of it will be captured. If analytes are not captured across the whole of the cross-section of a flow path (e.g. where capture reagent is immobilised only at the bottom of a channel) then, in multi-patch embodiments, it is preferred that consecutive patches of reagent for capturing the same analyte should be sufficiently separated along the path such that, during flow between them, analyte can diffuse in solution to be redistributed to give a substantially constant concentration throughout the depth of the solution (i.e. across the path's cross-section). At the start of each patch, therefore, the analyte will be evenly distributed. The separation distance between consecutive patches of the same reagent will depend on (or, alternatively, dictate) factors such as the flow rate along the path, viscosity of the solution, channel height, etc. As described in more detail below (Table II), essentially complete redistribution can be achieved with a separation of <1mm (e.g. <500μm, <400μm, or shorter) under typical flow conditions. Separation of consecutive patches of the same reagent can be achieved in various ways e.g. by having inert gaps between patches, or by including between them patches that capture a different analyte. For instance, capture reagents for three analytes A', B' and C could be arranged A-A-A-B-B-B-C-C-C, where each patch of A is separated to allow redistribution of A' in between, or they could be arranged ABCABCABC, where the presence of patches B and C between each patch of A allows analyte A' to redistribute. The distance between patches of reagent for capturing the same analyte in a multi-patch device will usually be at least 10^ynm, where V is selected from 0, 1 , 2, 3, 4, 5, 6, 7, etc.

In mono-patch embodiments, or in multi-patch embodiments where sequential patches for a first analyte are separated by one or more patches for other analyte(s), adjacent patches may abut or may overlap, but it is preferred that adjacent patches are separated by a gap. Overlapping patches are not preferred.

Patches preferably have a rectangular or square shape. For a patch with width W and area A, the length of such a patch will thus be A ÷ W. For abutting patches of the same size, the centre-to-centre separation will be A ÷ W, and to allow a gap between patches then the centre-to-centre separation will more typically at least 1.5(A÷W) or 2(A÷W). Some embodiments of the invention may also work for round, oval, or irregular patch shapes.

A device preferably contains at least 10^w different immobilised analytical binding reagents along a path, wherein N is selected from 0, 1 , 2, 3, 4, 5, 6, 7, 8 or more. Immobilisation of at least 10⁶ different oligonucleotides onto a single surface is well known in the field of microarrays. In mono-patch embodiments, the 10^w different reagents will typically be arranged in 10^w different patches; in multi-patch embodiments, the 10^w different reagents will be arranged in at least 2x10^N different patches.

Each patch of immobilised reagent preferably has an area of less than 10^x m², where X is selected from -6, -7, -8, -9, -10, -11 , -12, etc. Microarrays with patch sizes in the order of 10μm x 10μm (Ae. 10^"10 m²) are readily prepared using current technology. Patches with a small area improve the sensitivity of detection. When materials bind to the immobilised analysis reagent they are confined to a small area, increasing signal to noise ratio.

The length of a patch along the direction of flow will typically be at least 1f/ nm, where L is selected from 1 , 2, 3, 4, 5, 6, 7, etc. The length of a patch in a channel will usually be greater than the height of the channel. As the length of a patch increases for a fixed density, it becomes possible to capture more analyte. If a patch is long enough then it is possible to capture substantially all of its target analyte from a sample, and then quantification of its amount can readily be achieved. By using a flow path, it is possible to detect when signal drops to zero (i.e. there has been full capture for practical purposes).

The amount of reagent in a patch (or series of patches) should ideally be in excess of the analyte i.e. there should be enough reagent to capture substantially all of the analyte in a manner such that a gradient can be formed along the patch. If the reagent is not in excess then it will be saturated and the method will reveal that the analyte in question was present at a concentration above a particular value, but full quantitation will not be possible. In such circumstances, a method can be repeated e.g. with a different device having a higher concentration of immobilised binding reagent, with a diluted sample, a faster flow rate, etc. In multi-patch embodiments, it may suffice for the total amount of a capture reagent in all patches to be in excess, but it may be preferred to have an excess of capture reagent in each individual patch.

A typical set of parameters may be a channel of width 1mm, height 25μm, and length 60mm. Patches of capture reagents span the whole width of the channel. For some applications, the patch length may be 1mm with a capture reagent density of about 10¹⁰ molecules per cm². The flow rate may be expressed as a volumetric flow rate and be about 7.5 nl/min, which corresponds to a linear flow rate of 5μm/s. The temperature and viscosity of the buffer are dependent on the desired hybridisation conditions, which may be room temperature. Such a set of parameters may be p re-determined based on model calculations (e.g., model #1 below). In addition, or alternatively, the set may be refined or selected from an initial set of experimental data. Usually, these factors will be selected such that: (i) in a mono-patch embodiment, a patch will capture at least 50% (e.g. >60%, >70%, >80%, >90%, >95%, >98%, >99%, etc.) of its analyte during an experiment; (ii) in a multi-patch embodiment, each patch will in sequence capture between 20% and 80% of the (remaining) analyte during an experiment.

In some embodiments of the invention, binding with immobilised reagents occurs in a competitive manner i.e. an analyte displaces a substance that is already bound to an immobilised reagent.

Devices Devices used in methods of the invention include immobilised binding reagent(s) arranged such that they can interact with a sample moving along a path in or on the device.

Characteristics of the immobilised reagents are disclosed above. The device itself can take various forms. For instance, it may be a standard DNA or protein array, on which one or more flow paths can be defined e.g. by overlaying a suitable component. The choice of materials for manufacture of the device is influenced by a number of design considerations, and suitable materials can readily be selected by the skilled person based on the requirements of a particular device. For example, the material(s) should be amenable to microfabrication, stable to the reagents used during analysis, and compatible with the methods used for observing and measuring cells and molecules. Materials impermeable to the reagents used during analysis will generally be used. For some applications, it will be necessary to attach reagents covalently to the surface of a material. For some applications it will be desirable to use a hard material; other applications may need a flexible material. Where fluorescence is used for detection then. the material should be transparent to the excitation and emission wavelengths, and also have low intrinsic fluorescence at these wavelengths. Where electroosmosis is used to move material about the device then the material should be charged during use, or should be able to carry charge. For example, the skilled person can choose to give a positive or negative charge to silicon, glass and PDMS surfaces by derivatising them with appropriate silanising reagents. Materials that can propagate an illuminating evanescent wave (by total internal reflection) may be preferred for use with certain detection techniques. Suitable materials and fabrication methods are well known. Hard materials such as silicon and glass, for which microfabrication methods have been in use for many years, can be used. Recent developments in 'soft lithography', which exploits the potential to mould devices in polymers, such as polydimethylsiloxanes (PDMS), have enabled convenient methods for fabrication of microfluidic devices at the cellular scale. Thus devices can be made from a variety of materials, including but not limited to silicon oxides, polymers, ceramics, metals, etc. and mixtures thereof. Specific materials that can be used include, but are not limited to: glass e.g. microscope slides; polyethylene; PDMS; polypropylene; and silicon. PDMS is a particularly useful material, and the devices can be conveniently made by using casting, injection molding or UV-patterning and curing. Where devices include multiple components then these may be made from different materials (e.g. glass and PDMS) or from the same material.

Preferred devices comprise a plurality of channels and a plurality of immobilised binding reagents. The channels may be straight and substantially parallel to each other. The binding reagents may be immobilised in straight lines that run substantially orthogonal to the channels. Various techniques can be used to immobilise reagents such as nucleic acids in a series of parallel straight stripes.

Where the device includes a plurality of channels, these will generally be arranged next to each other within a single plane. It is possible to stack planes of channels, such that the channels will be arranged three-dimensionally, but ease of manufacture (particularly applying reagents to the insides of analysis channels) and result gathering (particularly reading analysis data within channels) means that substantially planar arrangements of channels are preferred. The overall device, however, may extend beyond the plane of the channels.

The flow path

Methods of the invention move analytes in a sample along a path, from its proximal end to its distal end. Along the path, analytes in the sample will encounter immobilised binding reagents, and so can be captured by the device. Capture along a path contrasts with a typical DNA array experiment, in which analytes in a sample are allowed to interact with the whole array surface simultaneously, without any directional flow.

A path will usually be defined by walls of a channel in the device. A sample will be applied to the proximal end of the channel and will move to its distal end (i.e. from a channel's entrance to its exit). A device may include multiple channels, thereby defining multiple flow paths and permitting the analysis of multiple samples in parallel. A device with multiple channels on the surface of an array is disclosed in reference 1.

A channel will preferably have a roof, thus restricting the distance that an analyte can move away from the device's immobilised reagents. Thus the path may be defined by a channel that is closed, except in the direction of flow. A liquid introduced into the channel will be able to flow along the path's axis, but cannot flow upwards or sideways out of the path.

Because such channels are closed, except along their length, access to their internal surfaces cannot easily be performed, which has an impact on the methods used for attachment of binding reagents to the device. Devices can be assembled from a base member and a lid member (e.g. Figure 19). The base member may include channels with an open upper side, allowing access to the channel's internal surfaces. After reagents have been immobilised onto the base member the lid member is attached to close the channels' upper side. The lid and base members will join to seal the channel, to prevent materials leaking between channels. In an alternative arrangement, reagents are applied to the lid rather than to the base. Where the lid covers the channel, it may be flat. In an alternative arrangement, the lid may itself include part of the channel.

A channel will preferably have a height h less than its width w. With reference to Figure 9, the w/h ratio will usually be at least 2 (e.g. 3, 4, 5, 6, 7, 8, 9, 10, 12, 14, 16, 18, 20, 25, 30, 35, 40, 50, 60, 70, 80, 90, 100 or more). Where patches of reagent are immobilised in a channel, the channel's height will usually be less than the length of the patches.

Channels will typically have a substantially constant cross-sectional area, and preferably a substantially constant cross-section shape. Variations in cross-sectional area lead to variations in flow rate through the channel, which is not usually desirable. A rectangular cross-section area is preferred, as explained in more detail below. A flow path's width is preferably less than 1cm e.g. <2mm, <1mm, <500μm, <400μm, <300μm, <250μm, <200μm, <150μm, <100μm, <50μm, etc. Channel widths of this order are used in available microfluidic devices.

A path may be substantially straight, or it may be curved. For instance, devices with multiple straight channels are disclosed in reference 1 , and serpentine channels are disclosed in reference 2. In a preferred device, a device includes a plurality of channels and a plurality of immobilised binding reagents, where the channels are straight and are substantially parallel to each other, and the binding reagents are immobilised in straight lines that run substantially orthogonal to the channels (see also reference 3). This arrangement ensures that the patches occupy the full width of the channels, and that all channels contain the same series of analysis reagents.

A serpentine path (having multiple connected parallel straight portions) can be conveniently arranged over a single straight line of immobilised reagent such that material flowing along the path crosses the line multiple times, but at different points on the line. Each intersection of the path and the line creates a separate patch for a multi-patch embodiment. If the path is arranged over multiple straight lines of immobilised reagents then the material can interact with each line multiple times. Figure 20 shows two devices having serpentine path arrangements. Figure 21 then shows one of these serpentine paths flowing over multiple parallel lines of immobilised reagent, with the straight portions of the serpentine path running in the Y direction and the reagent lines in the X direction.

A path may be branched. Each branch can be arranged to receive substantially the same materials as the others, or different cell contents can be directed down different branches e.g. mRNA down one branch and DNA down another, or positively charged proteins down one branch and negatively charged proteins down another. Sub-paths may re-join; if they do not, then a path can have more than one distal end. A branched device is discussed in more detail below.

Once a sample's contents have reached a path's distal end, it is preferred that they should not then flow back along the channel in the reverse direction, as this will interfere with gradients that were established during the initial flow. If the methods used to interpret binding patterns (see below) are sufficiently sophisticated, however, reverse flow of this type can be accommodated, as can other complex flow patterns. For simplicity of signal analysis, however, a simple flow from proximal to distal is preferred. This single pass flow contrasts with reference 2, where material was passed along a channel multiple times in both directions, with the aim of capturing analytes uniformly along the channel. In some circumstances, however, reciprocal flow can be used to simplify signal analysis. For instance, if a sample flows in a multi-patch device over at least three separate patches, each comprising a binding reagent for the same analyte, reciprocal flow over the middle patch(es) can be used advantageously. If flow over each patch along the flow path results in capture of an essentially constant proportion of an analyte (e.g. 5% of the total when passing over each patch, such that 0.95^N of the initial total remains after flowing over N patches and/or after flowing over a patch N times) then repeated reciprocal flow over a patch can be used to capture a larger amount than would be captured if a single flow direction were used. For example, in a situation where there are three patches, with repeated flow over the second patch for an odd number of passes (3, 5, 7, 9, etc.), the difference in signal between the first and third patches will be increased and can be used to facilitate quantitative analysis. Such flow patterns can be used to ensure that a greater proportion of total analyte in a sample is captured. Furthermore, efficient capture of analyte generally requires a slow flow rate, maximising the contact time between analyte and capture reagent, but slow flow rates can be difficult to control. By using reciprocal flow the same total contact time can be achieved but with flow rates that are faster and thus easier to control. More details are given below. Methods of the invention require analytes in a sample to move along a path. This movement can be achieved in various ways. For instance, the whole sample can be moved along the path, taking its analytes with it, or analytes in the sample may themselves be moved without moving the bulk of the sample. Various techniques can be used to move a cell's contents along the channel e.g. based on pumping, suction, capillary flow (including flow driven by micro-posts e.g. as in reference 4), electrokinesis, etc. A sample may move along the path as a plug, which may be preceded and/or followed by liquid and/or gas. For instance, a fixed volume of a sample may be introduced and then pushed along a flow path by either a bubble of air or by a buffer solution.

Where a sample moves along a path, its flow will preferably be laminar, not turbulent. Thus flow will usually have Reynolds number below 500. Laminar flow is the norm within microfluidic devices.

Where analytes move independent of the bulk sample, electrophoresis can be used, in which charged analytes move within an electric field.

Electrokinetic movement of analytes (e.g. by electroosmosis or by electrophoresis) requires a potential to be applied along a channel, with the polarity dictating the direction of movement. Electrokinetic movement in microfabricated devices is reviewed in reference 5. Figure 26 illustrates an electrokinetic arrangement.

Electroosmosis is a process by which fluid flows along a charged channel when a potential is applied along it. If a channel's surface is positively charged (e.g. along one or more walls) then, when a potential is applied across the channel, fluid within the channel can move by electroosmosis towards the anode. Movement of bulk fluid can bring about movement of analytes within the fluid. Electroosmosis and electrophoresis can be experienced at the same time. For instance, a negatively charged mRNA molecule will move towards an anode by electrophoresis. If a channel's walls are positively charged then fluid movement within the channel will also be towards the anode, and so the mRNA will move towards the anode by both electroosmosis and electrophoresis. If a channel's walls are negatively charged, however, the mRNA will experience electroosmotic flow towards the cathode, which will oppose the electrophoretic flow. The net effect of the opposing electroosmotic and electrophoretic flows on mRNA movement will depend on factors such as the magnitude of the electric field, the charge on the channel's walls, the solvent being used (e.g. depending on viscosity), the temperature (again, viscosity can change), ionic strength, presence of surfactants, pH, etc. These factors can be varied during design of the device (e.g. choice of materials, etc.) and/or during use (e.g. choice of temperature, electric field, etc.) in order to achieve the desired movement of particular components. Alteration of pH during use is a preferred way of controlling movement. Electroosmotic movement of material through the device is preferred. Movement of mRNA along a path is advantageously achieved by having (a) a negative potential at the proximal end relative to the distal end (proximal cathode, distal anode) and (b) a positive charge on the channel wall(s). Charged walls can be achieved by using a positively charged material for their manufacture. Where electroosmosis is used to move material along a channel then at least one wall of the channel will have an appropriately charged surface during use. The polarity and magnitude of the charge can be selected depending on the direction and rate of movement desired in any particular analysis. Polarity can depend on both the underlying material used to make the channel, on any surface-attached material (e.g. immobilised nucleic acids) and on any surface modifications. If a positively-charged material is used for one wall of a channel, and DNA and/or RNA is immobilised in discrete locations on the opposite wall, localised zeta potential variations will result in contraction and expansion of the bulkflow streamline distribution, giving rise in turn to differential transverse mass transfer rates and hence mixing. A similar effect can also be obtained using embedded electrodes, as demonstrated in references 6 and 7, by using pulses of electrical potential across a channel between the electrodes. Mixing effects and turbulent flow may be desirable in some assays but may be undesirable in others. The skilled person can choose these conditions according to their needs, and suitable conditions can be determined empirically.

As nucleic acids are charged molecules then they can cause changes in electroosmotic properties when immobilised on channel walls. If required then, in such situations, uncharged analogs of nucleic acids can be used instead e.g. PNA. Electrokinetic movement can be controlled precisely, and movement speed and direction can be varied simply by varying the potential as required. Electrokinetic movement can also be stopped, which can be used e.g. to allow introduction of reagents by mechanical means (e.g. injection, pumping, etc.). The flow rate of material through a channel should be controlled e.g. to obtain the desired capture in any particular experiment. If flow is too rapid, for example, then analytes will be swept along without having chance to come into useful contact with an immobilised binding reagent. For oligonucleotide patches with an area of 10μm x 10μm on the lower surface of a channel with a height and width of 10μm, a flow rate of 12.5 pl/sec is adequate to capture approximately 80% of a 250nt RNA target in the patch. A useful flow rate for any given combination of device, analyte and capture reagent can be determined empirically.

Determining analyte/reagent interactions

After analytes and binding reagents have interacted, the invention does not require special methods for detecting the interactions. On the contrary, methods currently used to detect signals on devices such as microarrays can be used with the invention, provided that the method can detect quantitative changes in signal intensity along the path. The detection method used will, of course, depend on the nature of the molecular targets and on any label that may be used.

Detection may occur in situ within a device, or may occur in a disassembled device. For instance, in a device having a channel member and a lid member (see above), with capture reagents immobilised on the lid member, the lid can be removed after analytes have passed through the channels, and the lid can be analysed separately e.g. using the reagents, techniques, devices and software already used to analyse microarrays.

For analytes such as RNA and protein, further biochemical processing may be needed in order to introduce detectable labels after a target analyte has interacted with an immobilised binding reagent. Fluorescent labels are preferred for use with the invention.

Fluorescence in the channels can be excited using an evanescent wave. These waves extend out of the surface of a material by ~^ΛA of the wavelength of the illuminating light i.e. they will extend outwards by ~150-350nm, which is more than enough to extend illumination throughout a patch of immobilised oligonucleotides. Other sources of light for excitation can also be used e.g. lasers, lamps, LEDs, etc.

Proteins can be detected by one of several known methods that exploit antibodies. For example, a protein that has been captured by an immobilised antibody can be detected by applying a second labelled antibody specific for a different epitope from the first antibody, to form a 'sandwich' complex. For RNA analytes, detection can be achieved by incorporating fluorescent nucleotides into a complementary strand using an enzyme such as reverse transcriptase. mRNA analytes may be captured by immobilised nucleic acid probes and then cDNA can be synthesised in situ by using the immobilised probe as a primer. The reverse transcription reaction preferably incorporates labelled nucleotides into the cDNA in order to facilitate detection of the hybridisation [8]. This can be achieved by the use of dNTPs with suitable fluorophores attached. Unlike a sequencing reaction, it is not necessary to use different coloured fluorophores for different nucleotides, because individual nucleotides do not need to be distinguished. Similarly, there is no need to label every nucleotide, and so 1 , 2, 3 or 4 of dATP, dCTP, dGTP and dTTP may be labelled, and a mixture of labelled and unlabelled dNTPs can be used. Incorporation of a large number of fluorophores into the cDNA (e.g. in at least 5% of incorporated dNTPs, such as >10%, >20%, >30%, >40%, >50%, >75%, or more) means that the cDNA can readily be detected by any of the familiar means of fluorescence detection, thereby revealing a positive signal even for a single hybridisation event. Thus even low-abundance mRNAs can be detected. High levels of label incorporation can, however, carry a risk of self-quenching.

An alternative way of detecting mRNA molecules is to capture them via their 5' ends (e.g. using a capture reagent that is immobilised at its 3' end), leaving their 3' poly-A tails exposed away from the capture surface. Labelled oligo-dT can then be applied to the captured mRNA molecules. The oligo-dT molecules will be captured (tandem hybridisation) by the poly-A tails and then can be detected. The tandem hybridisation step may be performed at a lower temperature (e.g. at room temperature) than the mRNA capture to reduce the risk that the capture reagent and mRNA will melt. By using oligo-dT of an appropriate length it is possible to achieve essentially uniform labelling of each mRNA (same number of dye molecules for each transcript), thereby facilitating quantification. A device for implementing this labelling procedure may have a separate inlet channel, or an on-device reservoir, for supplying oligo-dT after capture.

Rather than incorporate fluorophores directly, it is also possible to incorporate a specific functional group to which fluorophores can later be coupled ('post-labeling') e.g. after steps such as reverse transcription, washing, etc. Intercalating dyes can also be used as labels.

The amount of interaction between an analyte and a binding reagent will be determined at two or more (e.g. 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , etc.) points on the path, thereby allowing a signal gradient to be defined (see below).

In a mono-patch embodiment, the amount of interaction will be determined at multiple points within a single patch. These points will usually be defined arbitrarily e.g. as pixels or lines of pixels on an electronic representation of the patch. In a multi-patch embodiment, the amount of interaction will be determined in multiple patches, and it may optionally be determined at multiple points within each of those patches. In a multi-patch embodiment, however, it will normally suffice to detect the total amount of signal for each patch rather than detect gradients within each of those patches. Further details are given below.

The amount of interaction may be assessed in various ways. For instance, the total amount of fluorescence signal may be assessed, etc. Depending on the parameter that is measured, subsequent steps in the method may require a way of converting the parameter to an absolute analyte concentration e.g. to convert a fluorescence signal into the actual number of analyte molecules retained by a binding reagent. Methods for conversions of this type are disclosed in, for instance, references 9, 12 & 13, as described in more detail in the examples below. These methods may involve the use of high resolution scanners, such as current apparatuses that can identify single fluorophores with a pixel resolution of ~150 nm. For example, references 10 to 13 describe a single molecule reader (commercially available as the 'CytoScout' from Upper Austrian Research GmbH) in which a CCD detector is synchronized with the movement of a sample scanning stage, enabling continuous data acquisition at a pixel size of 129nm.

Inferring amounts of analyte

After measuring the amount of signal at multiple points along a path, the amounts of signal at different points are compared and then used to infer the amount of analyte in the sample.

In mono-patch embodiments, multiple datapoints will be measured in a single patch, to generate a signal curve along the path. In multi-patch embodiments, it may be sufficient to provide a single datapoint for each patch, to generate a patch-to-patch curve. In both cases, the number of points to be measured will depend on the method used to analyse that curve and the desired accuracy and/or on the number of patches of a particular reagent in the device. In some embodiments, for instance, the curve may include a number of signal points, and the shape of that curve can be related to the amount of analyte. In other embodiments, datapoints may be measured until signal declines to zero (or to background) and the point at which zero is reached can be related to the amount of analyte. If a device can capture all of a target analyte then quantification of its amount can be achieved directly e.g. by integrating the total signal for that analyte, or by direct counting. The inferred amount of analyte may be relative or qualitative. For instance, signal from an analyte in a first sample on a first path may be compared to signal from the same analyte in a second sample on a second path, and the method of the invention can reveal whether the level of the analyte in the first sample is lower than, equal to or greater than in the second sample. This comparison of levels may be quantitative, even though absolute amounts in each sample are not determined (e.g. the second sample contained twice as much analyte as the first sample, without revealing the actual concentration). The first and second paths may be on the same device, such that signals are compared on the same device, or they may be on separate devices, to allow comparison of samples between different experiments on different devices. In another form of qualitative or relative analysis, if an analyte saturates a binding reagent then its amount may be inferred to be greater than the amount which will saturate that binding reagent. In preferred embodiments, however, the inferred amount of analyte is absolute or quantitative. In these embodiments, observed signal data will generally be fitted to a mathematical model of expected signal.

For quantitative analysis of signal on a single patch (e.g. in mono-patch embodiments, or in single patches of multi-patch embodiments), a gradient of signal can be assessed using a model that describes how known quantities of an analyte interact with patches of analyte in the device. For instance, model #1 described in the examples below may be used. Other models of the interactions between nucleic acid analytes and immobilised probes are also available and can be used with the invention e.g. the model described in reference 14. The model used may depend on the characteristics of a particular experiment e.g. depending on flow rate, path dimensions, path shape, density of immobilised binding reagents, etc.

For quantitative analysis in multi-patch embodiments, gradients in each patch may be measured and analysed as described for mono-patch embodiments. Where analytes in a sample can redistribute between consecutive patches, however, a simpler analysis measures the total amount of signal in each sequential patch. The total amount of captured analyte in consecutive patches can be used to infer the total amount of analyte in the original sample. For instance, if the amount of signal drops proportionally by 80% from patch-to-patch then that ratio can be used to infer the total amount of analyte. Further details of such analyses are described in more detail in the examples.

Where total amounts of signal are compared patch-to-patch in multi-patch embodiments then, for simplicity of analysis, it is preferred that each patch for a single analyte should have substantially the same capture characteristics e.g. shape, size, reagent density, etc. If patches are not identical then analysis is more complicated e.g. it may be necessary to measure signal from the same area of each patch and to adjust calculations accordingly.

If analyte flows along a path in a single direction then the pattern of signal retained by immobilised binding reagents will be simpler than situations where analytes flow in forward and reverse directions i.e. it is easier to fit the signals to a mathematical model. As flow paths increase in complexity then the inference process also becomes more complex. Thus it is preferred with the invention for an analyte to flow along a path in a single direction.

Products In a single sample, such as a population of mRNA molecules, some analytes may be more abundant than others. Devices may be adapted to facilitate their parallel analysis. For instance, a device may have a flow channel comprising a proximal portion and a plurality of distal portions, with a branch between them, such that a liquid flowing from the proximal portion is split by the branch to flow into the plurality of distal portions. The flow channel will have the characteristics described above (immobilised binding reagents, efc.). In some embodiments of a branched device, binding reagents for low abundance analytes may be located before the branch, such that the whole sample can be analysed. In contrast, binding reagents for high abundance analytes can be located after the branch. The branch splits the sample into sub-samples, and thus reduces the amount of analyte that can interact with downstream binding reagents, but this reduction is not important for high abundance analytes. Figure 16 illustrates such a device, with a 3-way branch and multi-patch binding reagents for four analytes A to D.

In other embodiments of a branched device, different distal portions have different dimensions, such as different widths and/or heights. For instance, the widths of two distal portions may differ from each other by 20% or more. These dimensions are varied in order to ensure different flow rates along each distal portion. If the different distal portions have identical sets of immobilised binding reagents then, after the branch, a sample will encounter the same reagents but under different conditions. As explained herein, the binding characteristics to a patch of binding reagent can vary according to flow rate, and so the sample in each branch will interact differently with the binding reagents. Thus, in a single experiment on a single device, it is possible to select an optimum flow rate for a particular pair of analyte and binding reagent. Different binding reagents with very different properties (such as diffusion coefficients) can thus be explored simultaneously. Figure 17 illustrates such a device.

The invention provides a computer programmed to analyse an image of a device having one or more immobilised binding reagent(s), to which a sample was applied at a proximal end of a path and moved in a direction to a distal end of the path, such that analytes in the sample interacted with the immobilised binding reagents, wherein the image is analysed to determine an amount of interaction between an analyte and a binding reagent at two or more points on the path, and wherein the amounts of interaction at the points are compared to provide an inferred amount of analyte in the sample.

The invention provides a device comprising a substrate, a sample entrance, and a sample exit, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, (ii) the substrate has three or more different binding reagents attached thereto and arranged such that a sample flowing along the flow path can encounter two or more separate patches of each of said three different binding reagents. Thus, after flowing along the flow path on this device, an analyte in a sample will have encountered at least three different binding reagents at least two times each. By comparing the interaction at the at least two patches then, as described above, the amount of analyte in the sample can be inferred. This device permits this inference to be made for at least three different analytes. Other features of this device can be as described elsewhere herein. The invention also provides a device comprising a substrate, a sample entrance and a sample exit, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, (ii) the substrate includes a first and a second patch of immobilised binding reagent, wherein both of the patches can capture the same analyte, (iii) the first and second patches are arranged such that the separation between them, measured along the flow path, is greater than (e.g. at least 1.5x, 2x, 3x, 4x, 5x or more) the shorter linear dimension, measured along the direction of the flow path, of them. Other features of this device can be as described elsewhere herein. In contrast to the arrangement in reference 2, the separation between patches allows analyte to diffuse in solution to be redistributed, as discussed above. The invention also provides a system comprising a device and an image capture apparatus, wherein: (A) the device comprises a substrate, a sample entrance, and a sample exit, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, and (ii) the substrate has at least one binding reagent attached thereto, arranged in separate patches such that a sample flowing along the flow path can encounter two or more separate patches; and (B) the image capture apparatus can capture an image of the flow path and can measure more than one (e.g. 2, 4, 8, 16, 256, 2048, 32768 or more) intensity level of any signal on the path. Other features of the device can be as described elsewhere herein. The image capture apparatus will usually be able to detect fluorescence on the flow path.

Reciprocal flow

As mentioned above, if a sample flows in a multi-patch device over at least three separate patches, each comprising a binding reagent for the same analyte, reciprocal flow over the middle patch(es) can be used advantageously.

Thus the invention provides a method for analysing a sample that contains one or more analyte(s), wherein the method comprises steps of: (i) applying the sample to a flow path that has a proximal end and a distal end, and situated between the proximal and distal ends has at least a first, a second and a third patch of immobilised binding reagent, wherein each of the three patches can capture the analyte; (ii) moving the analyte along the flow path in a direction from the proximal end towards the distal end such that it can interact with the first patch; (iii) moving the analyte along the flow path in a direction from the proximal end towards the distal end such that it can interact with the second patch; (iv) moving the analyte along the flow path in a direction from the distal end towards the proximal end such that it can again interact with the second patch; (v) optionally repeating steps (iii) and (iv) to permit further interaction with the second patch; and (vi) moving the analyte along the flow path in a direction from the proximal end towards the distal end such that it can interact with the third patch. The amounts of interaction on the three patches can be used for analysing the sample, as described elsewhere herein.

The first, second and third patches should be sufficiently separated from each other along the path that, as described above, the analyte can diffuse in solution while flowing between them to be evenly distributed when the next patch is encountered or re-encountered. Where reciprocal flow is used, however, the distance between the patches should also ensure that the leading edge of analyte in a sample does not interact with the third patch before its trailing edge has passed the second patch.

Figure 25 illustrates a way of implementing this method, as explained in more detail below. Unlike reference 2, reciprocal flow is not applied to the first or third patches. The number of times that step (v) is performed (i.e. the number of direction changes and thus interactions with the second patch) can be varied according to requirements. For instance, if full capture of analyte is desired then the flow rate, channel dimensions, temperature, efe., can be used to calculate how many passes will be required, and this calculation can then be verified. In other embodiments, a fixed number of passes may be used and this information can be used as part of data analysis.

The flow rates through the flow path at the first and third patches will usually be substantially identical. The flow rate at the second path may be slower or faster than that, although substantially the same flow rate may be used here as well.

The flow path can include more than three patches for capturing a particular analyte i.e. there can be more than one patch between the first and third patches.

The invention also provides a device comprising a substrate, a sample entrance, a sample exit, and a pump, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, (ii) the substrate includes a first, a second and a third patch of immobilised binding reagent, wherein each of the three patches can capture the same analyte, (iii) the three patches are arranged such that the separation between each of them, measured along the flow path, is greater than (e.g. at least 1.5x, 2x, 3x, 4x, 5x or more) the shortest linear dimension, measured along the direction of the flow path, of them, and (iv) the pump can move a liquid along the flow path in both a direction from the entrance to the exit and also from the exit to the entrance. Other features of this device can be as described elsewhere herein. General

The term "comprising" encompasses "including" as well as "consisting" e.g. a composition "comprising" X may consist exclusively of X or may include something additional e.g. X + Y.

The term "about" in relation to a numerical value x means, for example, x+10%. Where necessary, the term "about" can be omitted.

The word "substantially" does not exclude "completely" e.g. a composition which is "substantially free" from Y may be completely free from Y. Where necessary, the word "substantially" may be omitted from the definition of the invention.

The use of terms such as "diameter" and "circumference" in relation to an element does not necessarily imply that the element is circular (or, in a three-dimensional context, spherical).

Where patches of an immobilised reagent "can capture the same analyte" then they may contain the same reagent as each other, or they may contain reagents that differ from each other but can still capture the same analyte. For instance, two different monoclonal antibodies may capture the same reagent. The term "antibody" includes any of the various natural and artificial antibodies and antibody-derived proteins which are available, and their derivatives, e.g. including without limitation polyclonal antibodies, monoclonal antibodies, chimeric antibodies, humanized antibodies, human antibodies, single-domain antibodies, whole antibodies, antibody fragments such as F(ab')₂ and F(ab) fragments, Fv fragments (non-covalent heterodimers), single-chain antibodies such as single chain Fv molecules (scFv), minibodies, oligobodies, dimeric or trimeric antibody fragments or constructs, etc. The term "antibody" does not imply any particular origin, and includes antibodies obtained through non-conventional processes, such as phage display. Antibodies of the invention can be of any isotype (e.g. IgA, IgG, IgM i.e. an α, Y or μ heavy chain) and may have a K or a λ light chain. BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 illustrates hybridisation profiles on two patches along a multi-patch flow path.

Figure 2 shows a device used with the invention. Sample is passing along a surface channel in a piece of PDMS. The PDMS is sealed against a microscope slide so that the channel runs perpendicular to three stripes of immobilised binding reagent on the slide's surface. Thus analyte in the sample can interact with the reagent at the points where the channel and the stripes intersect. An image capture apparatus is also shown.

Figures 3 and 4 show graphs relating to 'model #1 '. Figure 3 shows p(r) as a function of r, and Figure 4 shows H(r) as a function of r.

Figures 5 and 6 show high resolution scans of hybridised signal on an array. In Figure 6, the x-axis is distance in μm and the y-axis is signal intensity profile. Figure 7 illustrates the preferred patch width relative to the channel width.

Figure 8 illustrates the preferred arrangement of neighbouring patches along a channel.

Figure 9 illustrates dimensions of channels (height h and width w).

Figure 10 shows how a model is corrected to account for situations where a whole sample does not flow over a patch.

Figure 11 to 14 shows data obtained in flow experiments. Figure 12 shows an example of hybridisation on three patches of 1mm length each in a multi-patch configuration, where for purposes of the data display the distance between the patches has been reduced from 4.5mm to several 10μm. Figure 15 shows data from Figure 11 fitted to 'model #1'. Figures 16 and 17 illustrate devices including branched flow paths. In Figure 17, liquid can flow from inlet port (1) to outlet ports (2,3,4) via three distal branches (5,6,7).

Figure 18 illustrates the hybridisation profiles seen in a device, varying only the flow rate. The arrow shows the direction of the flow path.

Figure 19 shows assembly and use of a 2-channel device. Lines of oligonucleotides (3) are attached to a glass support (2). A piece of PDMS including two surface channels (1) is attached to the glass support. Separate samples containing mRNAs are passed through the two channels, and the PDMS is then removed. Reverse transcription occurs, extending immobilised mRNAs. Signals are seen, and their gradients are analysed as described.

Figure 20 shows two masks for etching serpentine channels into PDMS materials. Figure 21 then shows how these channels (running vertically) can be arranged over stripes (horizontal) of immobilised oligonucleotides on a DNA microarray.

Figures 22 to 24 and 27 to 28 show results of hybridisation of a mRNA analyte to immobilised DNA capture probes, showing fluorescence on an array's surface and also the fluorescence intensity profile along a straight line through the capture probes (fluorescence units against pixels). In Figures 22B, 22D, 24B and 24C the x-axis omits the gaps between adjuvant patches of capture probes.

Figure 25 illustrates a reciprocal flow embodiment.

Figure 26 illustrates an electrokinetic flow arrangement. The horizontal analysis channel runs over parallel capture stripes (vertical). An inlet and outlet channel feed into the analysis channel. Sample is introduced into the inlet channel and a plug forms in the analysis channel, the volume of which is determined by the displacement distance between inlet and outlet channels. A potential difference is applied across the analysis channel, causing the sample plug to move to the right. A bias is also applied to the inlet channel, causing the sample in this channel to move away from the analysis channel, thus avoiding contamination. The sample plug is electrokinetically transported across the stripes. MODES FOR CARRYING OUT THE INVENTION

Device preparation

DNA arrays were prepared on glass microscope slides by electrochemical synthesis. Probes were specific for capturing mRNA encoding murine HPRT, and were immobilised with 10⁹-10¹⁰ probes per mm². Targets were generated by in vitro transcription of the HPRT gene with incorporation of Cy3-Iabelled nucleotides, and were present at about 30fmol/μl. Typical arrays used for initial testing had 3 identical stripes running widthways, each 1 mm wide. A parallel straight channel was embossed on the surface of a flat piece of PDMS. The height and width of the channel were varied, but its length extended along most of the PDMS. At each end of the channel, a perpendicular hole extended through the PDMS, allowing liquids to enter/exit at a desired flow rate. The PDMS was arranged on top of the microscope slide, with the channels perpendicular to the array's stripes (Figure 2), with a microscope objective above the assembled device.

Flow and capture In experiment (A), the channel was 1 mm wide and 50μm high. The sample was passed along the channel at 1μm/s. Afterwards, the density of immobilised label was determined using an Axon GenePix 4000B scanner (Figure 11). More than 99.9% of the signal was captured on the first patch.

In experiment (B), the channel height was reduced to 25μm, and the linear flow rate was changed to 5μm/s. Results for all three patches are shown in Figure 12. Unlike the first experiment, signal was captured in the second patch; Figure 13 shows signal for just the first patch, Figure 14 shows the second patch.

Thus a device can be used with different capture efficiencies merely by varying the channel height and/or the flow rate. For example, substantially complete capture of the analyte on a single patch is possible.

In experiment (C), four parallel stripes of oligonucleotide probe, 1 mm width, were attached to Nexterion™ H slides. The probes were specific for the HPRT gene. PDMS with a 1mm wide straight channel, 50μm high, was applied to the slide to create a flow path running perpendicular to the stripes. The intersection between stripe and channel thus created a patch 1 mm x 1mm. The product of HPRT in vitro transcription (10OpM of transcript, Cy3- labelled) was passed along the channel at 0.005μl/min as a 2mm plug (~0.12μl_ volume). Figure 22A shows the fluorescence signal on the four patches, with flow having been from left to right. Decreasing signal is visible on the first patch, with no signal on the other patches. Figure 22B shows fluorescence intensity measured across the four patches. Figures 22C and 22D show results from similar experiments. In experiment (D), similar slides to (C) were used, but the PDMS included three parallel channels. Two channels were compared in which sample flowed through at a flow rate of either 0.05μL/min or 1Ox faster (0.5μl_/min). Fluorescence results are shown in Figure 24A. The faster flow rate visible reduces the amount of signal captured on the first patch. This visual impression is confirmed by the fluorescence intensities shown in Figure 24B (0.05μL/min) compared to Figure 24C (0.5μL/min).

In experiment (E), similar slides to (C) were used, but with two different negative control stripes arranged before the four HPRT stripes. Sample was applied at 48⁰C. Flow control at this higher temperature was less precise than at room temperature, though. Figure 27A shows a fluorescence image of the slide after hybridisation and Figure 27B shows a fluorescence intensity profile measured along the flow path. Results from similar experiments are shown in Figures 28 and 29.

In experiment (F), three parallel stripes of HPRT oligonucleotide probe, 1 mm width, were attached to Nexterion™ H slides. Two different negative control stripes were also attached, running parallel to the HPRT stripes. A serpentine channel in PDMS (1mm wide, 50μm high) was applied to the slide such that a flow path (Figure 23A) runs over the three HPRT stripes then the two negative control stripes, then passes in the opposite direction over the two negative control stripes then the three HPRT stripes, and then back again in the original direction on a third pass. The transcript analyte was applied through the PDMS channel, with a flow rate of 0.5μl_/min in the first pass, 0.01 μl_/min in the second pass, and 0.5μl_/min in the third pass. The results of hybridisation are shown in Figure 23B. Figure 23C shows the fluorescence signal profile along the first pass (the third HPRT patch is absent). Figure 23D shows the profile for the second pass. Figure 23E shows the third pass. Signal decay is clearly seen in each of the first six HPRT patches.

Reciprocal flow

Figure 25A shows a device that is used for reciprocal flow. 39 stripes of capture oligonucleotides are immobilised onto a glass microscope slide. These stripes are shown vertically in Figure 25. A serpentine flow path etched into PDMS is laid over these stripes. The flow path has three parallel interaction channels, arranged so that the flow path intersects each oligonucleotide stripe three times. These three intersections form the first, second and third patches defined above.

Between the three interaction channels the flow path is convoluted so as to increase the length of the path between the three channels. These regions permit analyte to redistribute itself when flowing between each of the three patches, but their increased length also means that the trailing edge of an analyte-containing sample plug can pass over the 39 stripes in the first interaction channel before the plug's leading edge encounters the second interaction channel. Figure 25B illustrates a plug of sample passing from the flow path's proximal end (top of Fig. 25) towards its distal end (bottom of Fig. 25) and passing into the first interaction channel. In Figure 25C the plug has now passed into the second interaction channel and is flowing over the stripes. In Figure 25D the plug has passed through the second interaction channel, and the flow direction is now reversed so that the plug flows back through the second interaction channel (Figure 25E). This cycle is repeated a desired number of times (an even number of direction changes), until finally the flow direction is not reversed after the plug leaves the second interaction channel and thus the material flows into the third channel (Figure 25F). The amount of analyte captured in the first interaction channel will be proportional to the initial analyte concentration. Because of analyte redistribution in the convoluted regions, the amount captured in each pass through the second interaction channel (both directions) will be proportional to the concentration at the start of that pass. Passage through the third channel will pick up (i) again, an amount proportional to the concentration leaving the final pass over the second channel; (ii) zero, if substantially all analyte, for practical purposes, was already captured; or (iii) the residual amount of analyte that was not already captured. The amount of captured analyte can thus be used for calculating analyte concentrations in the original sample.

Analysing mono-patch data A mathematical model of analyte capture by immobilised reagents is described below ('model #1'). The model considers a one-dimensional random walk between two barriers at 0 and b, where a particle starts at initial position u on the x-axis (0 < u < b an integer) at t = 0 and with equal probability moves one unit to the left or right at times t = 1 , 2, ... . The barrier at 0 is absorbing and the one at b is reflecting so that: (i) when the particle reaches the barrier at 0 it is absorbed and the process terminates; (ii) when at any integral time τ (τ > b - ιι), the particle is at the barrier at b, with equal probability it remains there at the next step (τ+1) or moves one unit to the left.

Letting g(t \ u) be the probability that the particle reaches the barrier at 0 for the first time at time t starting from initial position u at t = 0, reference 15 shows that:

Jt ,_ . sϊn(b — u + I ) ci_η - sin (b — u) <x_η gi t U}= - > COS^{1 x} αη SirUΥf_j-7 — . , . : ;

^₁ ' ' (Jb + J )eostib + I )(X_n - I) cos bα_n ' where α_η(η = 1, 2, . . . , b) are the b distinct roots of: sin (b + 1)α - sin bα .

With a channel having a rectangular cross-section with height h m, an immobilised DNA probe of length / m, and a uniform analyte flow rate of v m/s (piston flow) with a diffusion coefficient D m²/s, the hybridisation profile to a single patch can be modelled. Assuming that the target/probe on and off rates are infinite and zero, respectively, so that all target which reaches a probe hybridizes immediately and permanently, the fraction of analyte that hybridizes is then determined primarily by vertical diffusion between the channel ceiling and floor (immobilised probe).

For the channel setup, divide the height h into b equal steps of size Ah. The corresponding diffusion time step is:

(Ah)² h²

At =

2D 2bW

The time for an analyte to flow a distance / over the probe is t = l/v, for which the number of time steps is:

JV = — - = —r-ζ — = 2b r, where r =

At h²v ' ^{' *} /i²v

Target particles are initially distributed uniformly between the channel floor and ceiling, so the probability distribution function p for target absorption is:

1 ^b p{r)= Mm T- y g(2b²r \ u)

H- I

Thus the probability can be expressed as a function of a single variable r, which is a function of channel height, probe length, flow velocity and diffusion coefficient. The function can be evaluated numerically using a sufficiently large value of b (say b > 200) and plotted or tabulated for a range of values of r. A plot is shown in Figure 3. The results of this modelling were in close agreement (typically within 1%) of an earlier more complex simulation model.

The shape of the hybridization probability density function p(r) in Figure 3 is the profile of hybridization intensity along the length of the probe and is constant regardless of channel parameters. The value of r (specified by channel height, probe length, diffusion coefficient and flow velocity) determines what portion of the graph constitutes the hybridization profile. The graph's x-axis can then be interpreted as the x-axis of the channel with distance expressed as:

₇ rh²v I =

D

To determine hybridisation rate, the cumulative distribution function

H(r)= f p(r)dr Jo

gives the hybridization rate for any values of h, I, v and D, and a plot is shown in Figure 4 For example, taking a channel height of 5μm, a probe length of 20μm, a flow velocity of 25μm/s and a diffusion coefficient of 19 μm²/s (250 nucleotides), r = 0.608 and H(0.608)= 0.817, so the hybridization rate is 81.7%. As r is a product of powers of the four parameters h, 1, v and D, the same hybridization rate would be obtained by, for example, (i) halving the 5 height and increasing the velocity by a factor of four; (ii) halving both the probe length and the velocity; (iii) halving the height and reducing the probe length by a factor of four.

Table I shows results (calculated from H(r)) for a target flowing through a channel of width 80μm over a probe of width 80μm and of lengths 10, 20 and 40 μm at a range of channel heights and flow rates, using a uniform velocity profile (piston flow). A diffusion coefficient of 10 19μm²/s was used, corresponding to a target length of 250 nt. The hybridization rate is the percentage of target that hybridizes to the probe. An infinite on rate constant was used for target-probe hybridization, with a zero off rate constant, so that all target which reaches the probe surface hybridizes immediately and remains there. Figure 18 illustrates the hybridisation profiles seen in a device, varying only the flow rate.

15 This model can be used to analyse the empirical hybridisation gradients achieved in flow experiments (A) and (B). By establishing the proportion of analyte that is captured, quantitative data can be inferred.

As described above, the movement of an analyte in the channel can be treated as a random walk, making simplifying assumptions of excess probe and infinite on and zero off 20 rates. The channel hybridization intensity profile for an experiment with flow velocity v, diffusion coefficient D, channel height h and probe length / can then be expressed as a probability density function p(r) of a single variable r, where r = ID / (h²v). A set of experimental hybridization intensities (X|, yi) can thus be fitted by least squares to p(r) by minimizing

25 RSS(A,B) = ∑(By_t -P(Ax₁))² i where the x and y-axes are scaled by factors A and B. The value of r corresponding to the end x_en<j of the probe patch is estimated as r_enci = Ax_end, from which the fraction of target which has been hybridized is estimated as H(r_end).

This fitting procedure is sometimes unstable, due to the scaling of both axes. A more robust 0 procedure is to fit log(y) to log p(r). This can be done directly or, noting that log p(r) is linear for r > 0.2, by fitting to the linear form (defined by m = -2.459668 and c = 0.696036) and minimizing:

RSS(A₃B) = J](IOg(By₁) - (M(Ax₁) + c))²

/|0.2<Λ.Y;<l

Only a subset of the data is used, that which results in a fit in the range 0.2 < r < 1.0, where 5 the upper bound on r is used to exclude excessively noisy data (in log form) caused by lack of signal. The robustness of the estimated r_end value can be assessed by performing a series of fits using different subsets of data.

A comparison of the empirical and model-fitted data for experiment (A) is shown in Figure 15. Thus the model's results are a close match of empirical data. Further models that can be used and/or adapted for calculating analyte concentrations from empirical data may be based on finite difference algorithms [16] or random walks [17].

Analysing multi-patch data

Figure 1 illustrates the signal expected when a sample is passed along one of the channels on a multi-patch device. The y-axis is signal intensity. The x-axis is progress along the path, showing a first patch, a gap, and then a second patch which is the same as the first. As analyte molecules flow along the path and on the first patch, they are progressively removed from solution until they have passed the first patch. Analytes re-equilibrate while flowing along the gap, and then they encounter the second patch and are again removed from solution. The signal intensity (i) on a patch is proportional to the number of molecules hybridised to a patch, which in turn is proportional to the concentration of target analytes in the sample solution to which the patch was exposed:

I *= k'N = kC where N is the number of molecules, C is the concentration of target molecules and k is a constant

Where hybridisation takes place in a microfluidic channel, with the target solution moving in one direction, the number of molecules picked up on a patch will vary with position along the patch in the direction of the flow in some configurations approximately according to:

C_x = C_o e -"^x where C₀ is the bulk concentration and h is a constant depending on flow rate, diffusion coefficient and channel dimensions.

The expression for the intensity thus becomes:

I_x = Jc Q e -^hx If intensity values are integrated over a patch, the expression becomes I_x = k C₀ \ e -^hx dx

Subsequently, if all the patches are of equal length, and since h remains constant (provided the flow rate is constant), the integral is a constant too. Thus the expression simply becomes: I₁ = KQ where the subscript / now refers to a particular patch on the microarray.

Thus the percentage pick up p by any of the patches in the flow path is always the same, even though the target solution becomes increasingly depleted as a result of it. Hence the percentage pickup amount can be determined from the intensities of just two patches along a flow path:

KΓ C —C T - T p = 100— = \00^—^=L = 100^— -^ C C T

If the number of hybridised molecules can be counted directly (e.g. using a single molecule scanner), this number can be combined with the p figure to determine the original concentration of the analyte in a sample solution. If the number cannot be counted, but intensity can instead be related to the number of captured molecules, the p figure and the intensity integrated over a patch can also reveal the original concentration. The signal from any patch can be used, but (i) it is necessary to know whether it is the first, second, third, etc. and (ii) the highest signal intensity relative to background will be in the first patch, and so the first patch is preferred. This is 'model #2'

In an alternative model of the patch-to-patch signal ('model #3'), in a device that provides (i) identical patches along the flow direction of the solution that is to be hybridised, and (ii) identical hybridisation conditions for the patches, the change in the amplitude of either (i) the hybridisation profile as a function of length, or (ii) the integrated intensity value of the hybridisation readout, can be used to find the hybridisation percentage on a given patch with identical properties.

Starting from an initial amount of material Λ/ (e.g. number of molecules, or a given mass, or a constant concentration of solution) that flows over 2 identical patches, the hybridisation percentage p on each of the patches leads to a deposition of pΛ/ material on the first patch, leaving (1-p)-N material for subsequent hybridisation to the second patch. On this second patch, (1-p)-N is deposited. This depletion continues along any further patches, and the ratio of the hybridised amounts (i.e., the intensities of hybridisation readout) is:

I₂ a-<X-pypN

= !-/>

/, a -pN where α is the factor that relates the intensity of the readout to the amount of material that has been hybridised.

In situations where a whole sample does not flow over a patch (as in experiment (B)), because the rear end of a sample does not progress as far as the front end, a correction must be applied. Figure 10 shows the principle of such a correction. The initial concentration is C₀. After some target is picked up by subsequent probe patches, the concentration is smaller, and can be described as c_n=(1-p)ⁿc₀, where p is the pick-up percentage on each patch, and n is the index of the patch running from 1 onwards.

A quantity proportional to the material pickup, m_n, can be measured for each patch, which can be expressed as the product of the concentration of the target just before the patch, the volume that flows over the patch, and the pick-up percentage: m_n=c_n.₁A-l_n where A is the channel cross section, h is the length of the sample that flowed over the first patch, and l_nJr (n-1)Δx is the length of the plug flowing over patch n, while Δx is the pitch of the identical patches. The pick-up percentage can then be calculated from the known or measured quantities of two adjacent patches as:

Using this model with the data for the first and second patches in Experiment (B), the proportion of analyte captured by the first patch was calculated as 83.3+2.4%. The uncertainty in this figure is due to an uncertainty of measuring the precise length of sample that flowed over the first patch; this uncertainty is appropriately propagated through to the final result.

Analyte redistribution between patches in multi-patch devices

After analyte has flowed past a probe patch, the remaining analyte in solution is distributed non-uniformly in the channel's cross-section, with the analyte concentration increasing from floor to ceiling. As the analyte continues to flow along the channel after the patch, however, its distribution gradually becomes uniform due to diffusion. To apply the multi-patch method optimally, there should be enough distance between successive patches for capture of the same analyte to ensure uniform analyte distribution at the start of the second patch. Table Il shows results for a target flowing through a channel of width 80μm over a probe of capture oligonucleotide. The probe patch occupies the full width of the channel (80μm) and has a length of 40μm. Various channel heights and flow rates were modelled, using both parabolic and uniform (piston) velocity profiles. A diffusion coefficient of 19 μm²/s was used, corresponding to a target length of 250nt. An infinite on rate constant was used for analyte-probe hybridization, with a zero off rate constant, so that all target which reaches the probe surface hybridizes immediately and remains there. For the channel floor regions without any capture reagent, the on-rate for capture was zero. The table shows the distance in μm beyond the end of the probe patch for the ratio of target concentration at the channel floor and the ceiling to reach 0.9, 0.95 and 0.99. Quantitation

Using any of these models, the ratio of material that has flowed over a probe patch can be calculated relative to how much has been picked up on that respective probe patch. This is a significant step towards quantifying the amount of material in solution, as opposed to merely quantifying how much material has been picked up by a probe patch.

In order to give a quantitative answer to the question of how much material was in a sample solution before pick-up of target analytes, two questions have to be answered: (1) How much material has been picked up by the patch? (2) What is the ratio of picked-up material to material that has not been picked up? The combination of the two measurements can then yield knowledge of how much material was present prior to hybridisation.

Model #1 can give the answer to question (2) when using a mono-patch geometry, and models #2 and #3 obtain the same answer for a multi-patch geometry while relying on a simpler mathematical model. To answer question (1), however, other methods are required.

One simple method for relating signal intensity to molecular amount is to hybridise a known amount of a labelled target of known molecular weight to a probe, and then measuring hybridisation signal. Methods to achieve this goal are well known. The calibration factor obtained by this method, adjusted for factors such as molecular weight of target, can then be used to convert the integrated intensities to amounts of target.

Single molecule counting can also be used to answer question (1). Examples of single molecule counting have been published (e.g., references 12 & 13). While reference 12 describes only the mathematical process, reference 13 also shows an experimental method how the images for single molecule counting can be obtained. The same apparatus (CytoScout™) has been described in the literature [10].

The counting algorithm for single molecule counting is described in reference 13 as follows: "At low surface densities of N<1000 molecules per spot, more than 95% of all molecules are well separated in the image and were counted via automated algorithms. At higher surface densities, peaks begin to overlap; in such cases, the number of hybridized molecules per spot was inferred from the corresponding integrated fluorescence signal and the single molecule brightness B." Single molecule scanning of microarrays relies on a good discrimination of signal and noise. Figure 4 shows that a single molecule scanner can give a suitably good discrimination, and non-specific binding (such as outside of probe patches) is very low, even for the highly sensitive Cy3 green channel with "contaminating fluorescence" [18].

Also, reference 9 reports that it is possible to produce a microarray assay format that transforms raw data into a defined quantitative unit (i.e., moles) by measuring the amount of array feature present and the cDNA sequence hybridized. Even though this method does not achieve single molecule sensitivity, hybridised target amounts as low as 3.2 x 10^~9 pmol (i.e. ~2000 molecules) have been identified, even using a conventional microarray scanner.

Thus there are several methods in the literature that give absolute answers to the question of how much material has been hybridised to a patch on a microarray. In the prior art, this result on its own was not very meaningful since it was not known using conventional methods, neither a priori nor a posteriori, what fraction of the material in solution has been hybridised. This ratio could be anywhere from a tiny percentage to 100%, depending on the exact hybridisation conditions (temperature, buffer details, duration of hybridisation, ratio of probe to target concentration, etc.). Even on the same microarray, different target molecules generally exhibit different hybridisation rates to the probes (difference in GC content, potential secondary structures, different probe lengths, efc.).

In combination with the methods described above, however, the absolute amount of hybridised material can be combined with the percentage captured by a patch to infer the amount of analyte in the original sample.

It will be understood that the invention has been described by way of example only and modification of detail may be made without departing from the spirit and scope of the invention.

TABLE I

TABLE 11

REFERENCES (the full contents of which are incorporated herein by reference)

[1] WO2006/117541.

[2] Wei et al. (2005) Nucleic Acids Research 33(8):e78.

[3] WO2005/037425.

[4] WO03/103835

[5] Wong et al. (2004) IEEE/ASME Transact Mechatron 9:366-76.

[6] Lee et al. (2004) J. Micromech. Microeng. 14:1390-1398.

[7] Lee et al. (2005) J. Micromech. Microeng. 15:1215-1223.

[8] WO2004/033629.

[9] Rouse et al. (2004) BioTechniques 36:464-470.

[10] Hesse et al. (2004) Anal Chem 76:5960-4.

[11] WO00/25113. See also US-2002/0030811.

[12] Muresan et al., IEEE International Conference on Image Processing, 11-14 Sept 2005. Volume 2:1274 - 1277

[13] Hesse et al. (2006) Genome Research 16:1041-45.

[14] Benn et al. (2006) Analytical Biochem 348:284-93.

[15] Weesakul (1961) Annals of Mathematical Statistics 32:765-9.

[16] Zimmerman et al. (2005) Biomed Microdev 7:99-110.

[17] Pappaert et al. (2006) Biotechniques 41 :609-16.

[18] Martinez et al. (2003) Nucleic Acids Research 31 :e18.

Claims

1. A method for analysing a sample containing one or more nucleic acid analyte(s), comprising steps of:

(i) applying the sample to a device having one or more immobilised binding reagent(s), wherein the binding reagents are nucleic acids, and where the sample is applied at a proximal end of a path and moves in a direction to a distal end of the path, such that analytes in the sample can interact with the immobilised binding reagents;

(ii) determining an amount of interaction between an analyte and a binding reagent at two or more points on the path; and

(iii) comparing the amounts of interaction at the points and, based on the comparison, inferring the amount of analyte in the sample.

2. The method of claim 1 , where the device includes more than one immobilised binding reagent.

3. The method of any preceding claim, wherein the immobilised binding reagent(s) are nucleic acids on the surface of a solid support.

4. The method of any preceding claim, wherein the device is made from a material that is impermeable to the reagents used during the method.

5. The method of claim 3, wherein different immobilised binding reagents are arranged in discrete patches on the solid support.

6. The method of any preceding claim, wherein a reagent for a particular analyte is included once on the path.

7. The method of any one of claims 1 to 4, wherein a reagent for a particular analyte is included more than once on the path.

8. The method of claim 6, wherein repeats of the reagent are sufficiently separated along the path such that, during flow between them, a particular analyte can diffuse in solution to be redistributed to give a substantially constant concentration across the path's cross- section.

9. The method of claim 6 or claim 7, wherein at least two points in step (ii) are located in separate repeats of the reagent.

10. The method of any preceding claim, wherein the path is defined by walls of a channel.

11. The method of any preceding claim, wherein step (ii) uses fluorescence.

12. A device comprising a substrate, a sample entrance, and a sample exit, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, (ii) the substrate has three or more different binding reagents attached thereto and arranged such that a sample flowing along the flow path can encounter two or more separate patches of each of said three different binding reagents.

13. A device comprising a substrate, a sample entrance and a sample exit, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, (ii) the substrate includes a first and a second patch of immobilised binding reagent, wherein both of the patches can capture the same analyte, (iii) the first and second patches are arranged such that the separation between them, measured along the flow path, is greater than the shorter linear dimension, measured along the direction of the flow path, of them.

14. A method for analysing a sample that contains one or more analyte(s), wherein the method comprises steps of: (i) applying the sample to a flow path that has a proximal end and a distal end, and situated between the proximal and distal ends has at least a first, a second and a third patch of immobilised binding reagent, wherein each of the three patches can capture the analyte; (ii) moving the analyte along the flow path in a direction from the proximal end towards the distal end such that it can interact with the first patch; (iii) moving the analyte along the flow path in a direction from the proximal end towards the distal end such that it can interact with the second patch; (iv) moving the analyte along the flow path in a direction from the distal end towards the proximal end such that it can again interact with the second patch; (v) optionally repeating steps (iii) and (iv) to permit further interaction with the second patch; and (vi) moving the analyte along the flow path in a direction from the proximal end towards the distal end such that it can interact with the third patch.

15. A device comprising a substrate, a sample entrance, a sample exit, and a pump, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, (ii) the substrate includes a first, a second and a third patch of immobilised binding reagent, wherein each of the three patches can capture the same analyte, (iii) the three patches are arranged such that the separation between each of them, measured along the flow path, is greater than the shortest linear dimension, measured along the direction of the flow path, of them, and (iv) the pump can move a liquid along the flow path in both a direction from the entrance to the exit and also from the exit to the entrance.

16. A system comprising a device and an image capture apparatus, wherein: (A) the device comprises a substrate, a sample entrance, and a sample exit, wherein (i) the entrance and exit are connected via a flow path such that a sample introduced via the entrance can flow along the flow path towards the exit, and (ii) the substrate has at least one binding reagent attached thereto, arranged in separate patches such that a sample flowing along the flow path can encounter two or more separate patches; and (B) the image capture apparatus can capture an image of the flow path and can measure more than one intensity level of any signal on the path.