GB2450356A - Method of Determining the Genotype of a Polymorphism using labelled nucleotides - Google Patents

Abstract

A method of determining the frequency of a genotype of a polymorphism in a population of nucleic acids comprises annealing a primer 16 to a template DNA 10 at a position adjacent a known or suspected polymorphic base. The primer 16 is preferably linked to the surface of a microarray slide 18. Labelled chain-terminating nucleotide analogues corresponding to different alleles of the polymorphic base are introduced to extend the primer 16. The resulting labelled primer-template complexes are then visualised, for example, using total internal reflection fluorescence microscopy. The differently labelled molecules are counted in order to determine the frequency of each genotype in the population.

Description

Method of Determining the Genotype of a Polymorphism This invention

relates to a method of determining the genotype of a polymorphism, and in particular to a method of determining the proportion of nucleic acid molecules in a population having a predetermined genotype. In preferred methods, the polymorphism is a single nucleotide polymorphism.

It is estimated that I in every 1,000 bases in the human genome exhibits a single nucleotide polymorphism (SNP); i.e., within the human population there is a variability in the base present at that position (Syvanen, A.C. (2001) Nature Reviews Genetics 2, 930-942). Such variations may be present in the coding sequence of a gene, in which case they may, for example, be associated with a disease, or the susceptibility to a disease; or they may affect drug metabolism. Similar effects may result when SNPs are present in the regulatory regions of genes. Detecting such polymorphisms is important in diagnostics, pharmacogenomics and disease risk assessment. In addition, the many SNPs located in non-coding regions of the genome may be used in linkage disequilibrium studies to identif' genes involved in particular phenotypes, including those responsible for disease.

There exists a wide variety of techniques for the detection of SNPs (Syvanen, A.C. (1999) Hum. Mutat 13, 1-10). Given the vast numbers of polymorphisms that are present in the human genome, and the numbers of samples that need to be characterised, there is an increasing requirement for high-throughput genotyping techniques. Furthermore, there is also a need for high throughput methodologies to screen for SNPs that are related to a particular phenotype.

One family of SNP detection techniques is so-called "minisequencing." Minisequencing is analogous to a conventional sequencing reaction, in that a DNA polymerase is used to extend an oligonucleotide primer hybrid ised to a template DNA. However, unlike conventional sequencing reactions, in which many hundreds or thousands of new bases may be incorporated, in minisequencing typically only one or a few bases are added. Here, the primer defines the location of the base whose genotype is to be determined. The primer is annealed immediately adjacent the polymorphic base on its 3' side. The genotype of the polymorphic base is determined by the base, if any, which is incorporated at the 3' end of the primer. "Readout" techniques used with minisequencing include conventional fluorescence, fluorescence polarisation, mass spectrometry, ELISA and chemiluminescence. A wide variety of assay formats have been developed in order to increase the throughput of the methods, for example through multiplexing approaches.

These include a variety of array-based formats. The technology of minisequencing is described in US 2003/008253 1.

Another approach to genotyping is based on hybridisation of oligonucleotide probes specific to one or the other allele to the target DNA. Perfectly matched target DNA-probe hybrids have a higher melting temperature than hybrids with a single base mismatch. It is possible to use this principle to discriminate between alleles. Typically a pair of probe oligonucleotides is used, one for each allele of interest. A disadvantage of this approach is that the pairs and reaction conditions must be empirically fine-tuned for each particular polymorphism to give accurate genotyping results (Syvanen, A.C. (2001) Nature Reviews Genetics 2, 93 0-942).

An extension of the hybridisation principle is to use the allele-specific oligonucleotides as primers in a polymerase chain reaction (PCR). This is a popular method because it exploits hardware and reagents commonly available in many laboratories. However, the amplification and detection steps potentially introduce additional uncertainties into the measurement. Furthermore, the probes must again be optimised for each SNP.

A variety of other methods are used in genotyping. These include the use of ligation in which the DNA ligase enzyme discriminates between perfect matches and single base mismatches; cleavage using "invader" probes; and the use of restriction endonucleases.

What all of these techniques have in common is that the readout mechanism (be it fluorescence, fluorescence polarisation, or mass spectrometry, for example) relies on a bulk detection methodology, with its attendant uncertainties. Many of the detection methods also are somewhat indirect, involving additional processes, which can increase uncertainty, time to perform and cost of assays. Such steps may include amplification by PCR, electrophoresis and bulk fluorescence detection.

WO 02/072892 discloses a technique for single molecule sequencing. This method involves visualising the incorporation of labelled riucleotides into immobilised polynucleotide template molecules in a time-resolved manner. This method could not be multiplexed for genotyping as the reactions would proceed in parallel at each template or primer location of an array. The reactions could not be observed in parallel, which would result in loss of information.

WO 00/36 152 discloses a method of genotyping a nucleic acid in a single molecule configuration by detecting release of labelled phosphates. This technique relies on detection of a transient event, i.e. release of a fluorescently labelled phosphate group. In order to observe this event, sequential images would need to be obtained. This, again, inhibits multiplexing.

A major problem with current techniques for determining the genotype of a polymorphism is that they are not well suited to the quantitative analysis of complex mixtures. In some applications, this is not critical. For example, where a patient's genotype at a particular site is being assessed, the only possibilities are 100% of one allele (homozygous); 100% of the other allele (homozygous); or 50% of each allele (heterozygous). Therefore, even if the uncertainties in the measurements of the two alleles are quite large (e.g. 20%) the three possibilities can readily be discriminated.

However, in some cases it is of interest to determine the frequencies of polymorphisms more precisely. For example, where one is looking for an association between a SNP and a phenotype, one may compare pooled samples of DNA from many individuals. The frequencies of alleles are then compared with the occurrence of the phenotype(s) of interest to identify any correlations. Such an investigation may be done for many different SNPs to identify those that are of interest. This type of experiment is often referred to as statistical genotyping.

Another example might be the investigation of mutants of a virus. For example, where a single base mutation in a viral genome is associated with a drug resistance phenotype, it might be desirable to monitor the frequency of such an allele in a patient as treatment progresses.

In both cases, it is desirable to measure the frequency of a particular allele in a complex mixture. The precision and accuracy of the information required are significantly higher than where one is simply determining a genotype in an individual patient.

There are a number of reasons why such quantitative measurement is difficult to achieve.

Some of these problems are outlined below.

When fluorescence labelling is used, there are a number of different steps that can introduce bias into the measurement. Most of these result from the bulk nature of the measurement, which means that all the biases are rolled up into one number. For example, two different fluorescent dyes may be used to label two different alleles. In such cases the dyes may be incorporated into the DNA with different efficiencies, they may have different absorption coefficients for the light used to excite them, they may have different fluorescence quantum yields, the light sources may have different intensities, the filters and optics used may have different transmission efficiencies and/or the detector may have different sensitivities. Thus the relative measured intensities for the two dyes will not be directly related to the relative frequencies of the alleles. All of these factors could in principle be measured; however, in practice there are significant uncertainties and sources of variability, which are difficult to eliminate. The approach that is usually taken is to include internal standards in each assay, which can be used to normalise the results.

Where an amplification-based method is used, it may be assumed that the amplification efficiency of the two alleles is equal. This may not always be the case, which could result in a bias in the measurement, particularly where quantitative results are required. The amplification process itself may introduce additional uncertainties. For example, the yield of the reaction may be affected by contaminants in the sample.

Another issue that affects bulk measurements is that a single quantity (e.g. fluorescence) is measured, typically at two emission wavelengths. Such a measurement always includes a level of background signal. Again, using fluorescence as an example, sources of background could include the detector itself, contamination of the fluorescence with excitation light (due to the limitations of the optical system used), and/or unwanted fluorescence from materials (or their contaminants) used in the method. The fluorescence is measured as a single quantity and thus it is difficult to determine the contribution of the background signal. Furthermore, such a signal will contain a contribution of random noise from a variety of sources.

Another key parameter, which limits the utility of many prior art genotyping techniques, is the throughput of reactions. Related to this is the cost per reaction, which is strongly influenced by the time taken to perform each reaction, including all preparative and processing steps. It is also affected by the cost of the reagents and equipment required for the assay. Depending on the design of the assay, a method may be efficient for analysing a large number of samples for a few SNPs, or alternatively it may be more suited to analysing a small number of samples for many SNPs.

Multiplexing refers to the ability to perform more than one genotyping assay in the same reaction vessel. Existing genotyping techniques vary widely in their suitability for multiplexing.

Many methods require some sort of purification or processing step before detection. These types of manipulations add to the time required to perform the assay, and often make a methodology harder to automate.

Existing genotyping methods vary widely in the amount of material required to perform an assay. In some cases, sample quantity is not a concern; however in others it may be critical. For example, where many SNP genotypes are required from one sample; or where a sample is exceptionally difficult to obtain, there will be real advantages to having a technique with maximal sensitivity.

According to an aspect of the present invention there is provided a method of determining the frequency of an allele of a polymorphism in a population of nucleic acid molecules including: providing a target nucleic acid, the target nucleic acid including a known or suspected polymorphism; providing at least two distinguishably labelled nucleotides, the first labelled nucleotide corresponding to a first genotype of the known or suspected polymorphism, the second labelled nucleotide corresponding to a second genotype of the known or suspected polymorphism; synthesising an oligonucleotide complementary to a region of the target nucleic acid, the region including the known or suspected polymorphism; wherein the complementary oligonucleotide includes the first labelled nucleotide corresponding to the first genotype of the known or suspected polymorphism, or wherein the complementary oligonucleotide includes the second labelled nucleotide corresponding to the known or suspected polymorphism; detecting individual labelled nucleic acid molecules using optical microscopy; and counting the number of oligonucleotides including each labelled nucleotide, thereby to determine the number of nucleic acid molecules in the population having each genotype.

This method enables the frequency of different alleles within a population to be accurately determined. As the number of labelled oligonucleotides (which corresponds to the number of nucleic acid molecules having a particular genotype) is determined at the level of the individual molecules, inaccuracies that result from bulk calculations are avoided.

Furthermore, as this method does not require amplification of the nucleic acid, potential mistakes that may be incorporated during amplification are avoided.

The polymorphism is preferably a single nucleotide polymorphism.

In a preferred embodiment, synthesising the oligonucleotide includes extending a primer that has been annealed to the target nucleic acid. This method of synthesising an oligonucleotide helps to ensure that the oligonucleotide synthesised is truly complementary to the region of the target nucleic acid including the polymorphism.

Preferably, the primer is linked to a surface. This helps to increase throughput of the reaction. It is also enables many different primers to be used within the same reaction.

In the preferred embodiment, a plurality of different primers is linked to the surface, preferably in an array format. This allows several different polymorphic alleles to be quantified in a single reaction.

In the preferred embodiment, the primer is designed to anneal to the target nucleic acid at a position immediately 3' of the polymorphism. This means that the first base to be added to the primer will be complementary to the polymorphic base in the target nucleic acid.

Preferably, the primer is extended by a single base. This enables chain-terminating nucleotides to be added to the primer and facilitates incorporation of a single labelled nucleotide into the synthesised oligonucleotide. This helps to ensure that the method is quantitative.

It is preferred that extending the primer includes mixing the target nucleic acid with a mixture of chain-terminating nucleotides, wherein at least one of the chain-terminating nucleotides corresponding to one of the alleles of the polymorphism is labelled. This is a convenient method for incorporating a single labelled base into the oligonucleotide complementary to the region of the target nucleic acid including the polymorphism.

Preferably the mixture of chain-terminating nucleotides includes one type of nucleotide corresponding to one allele of the polymorphism labelled with one type of label and another type of nucleotide, corresponding to another allele of the polymorphism, being labelled with a different type of label. This allows two different alleles of the polymorphism to be quantified.

Preferably, nucleotides that do not correspond to an allele of the polymorphism in the nucleic acid are included in the mixture, but are not labelled. This may help to ensure the enzyme used is specific enough.

In the preferred embodiment, the optical microscopy is Total Internal Reflection Fluorescence Microscopy. In other embodiments the optical microscopy may be scanning confocal microscopy, epifluorescence microscopy, near field optical scanning probe microscopy, fluorescence correlation spectroscopy, Raman microscopy, or variants thereof.

The nucleotides may be fluorescently labelled, in which case the labels may include fluorescent dyes and/or fluorescent nanoparticles. In another embodiment the nucleotides may be labelled with Raman-active labels, in which case the labels may include Raman-active molecules and/or Raman-active nanoparticles.

Preferred embodiments of the present invention are now described, by way of example only, with reference to the accompanying drawrngs, in which: Figure 1 is a schematic illustration of a preferred method; Figure 2 is a schematic illustration of a preferred method; Figure 3 is a schematic illustration of differential labelling of alleles of an SN?; Figure 4 is a schematic showing the principle of array-based genotyping; Figure 5 illustrates the sequences of two "genotypes" of a model system; Figure 6 is a schematic illustration of an apparatus for use in the preferred method; and Figure 7 shows the results of an experiment using the method A preferred embodiment of the present invention relies on single molecule fluorescence detection. In outline, an exemplary method involves the following steps: i) The target (or template) DNA containing a known or suspected SNP is prepared. This may include a denaturation step to separate the two strands of the DNA.

ii) A probe (typically an oligonucleotide), complementary to the sequence immediately adjacent the base-pair where the SNP is located, is hybridised to the target DNA. This may occur free in solution; however, in a preferred method, the oligonucleotide is attached to the surface of an array slide.

iii) A reaction is performed to extend the probe molecule by a single base. This reaction is catalysed by an enzyme such as a DNA polymerase. The additional incorporated base is complementary to the base at the polymorphic position. The incorporated base is typically a chain-terminating nucleotide, such as a dideoxynucleotide triphosphate (ddNTP).

Nucleotides containing each base that may be complementary to the base of interest are used in the extension reaction. Each base is preferably labelled with a different fluorescent dye, enabling the base incorporated to be detected and the genotype of the template molecule to be thereby determined.

iv) The population of fluorescently labelled molecules is then visualised by an appropriate method, such as Total International Reflection Fluorescence Microscopy (TIRFM). There are a number of different approaches to single molecule fluorescence imaging, including techniques such as confocal microscopy and near-field scanning optical microscopy. Any suitable detection method may be used, although a preferred method is TIRFM. The method used should, preferably, be some form of wide-field imaging technique, which allows for the relatively rapid acquisition of data.

For TIRFM, or another technique, to be able to detect molecules, and further to obtain genotype information from them, the following features are preferably incorporated: I) The molecules should be labelled with a suitable fluorescent dye or other optical label ii) The specific dye should be related to the genotype at the chosen location iii) The labelling should be specific to the polymorphism of interest iv) The molecule should be immobilised on a surface Some visualisation techniques, (such as Fluorescence Correlation Spectroscopy), can be carried out without the molecules being immobilised (i.e. the molecules may be detected in solution). However, surface attachment has an additional advantage in that it allows array-based methods to be used.

A separate image is acquired of the fluorescence from each dye (using, for example, a series of filters).

v) The images are then analysed to count the number of molecules labelled with each base, and thereby derive the frequency of each genotype. This step may include corrections for any remaining errors due to e.g. differing labelling efficiencies, or false positives due to misincorporations.

By this method, quantitative determination of the frequency of different alleles of a polymorphism in a complex sample by counting individual molecules is possible. Each of these stages is described in more detail below. Throughout this application references to a "nucleotide" should be held to include references to a suitable "nucleotide analogue" as appropriate.

Template preparation and annealing The template DNA should be of sufficient purity for the method to succeed and should not contain contaminants that could, for example, inhibit the enzyme, or fluorescent contaminants, which could interfere with the detection process. if the DNA is from a genomic source it is likely that the size of the DNA fragments will be large. In such a case, it may be necessary to shear the DNA or otherwise break it up into shorter fragments.

This reduces the viscosity of the solution and should also improve the kinetics of the annealing reaction.

In Figures 1 and 2, template DNA 10 is shown. Each molecule includes one of two possible alleles of a polymorphic base 12, 14 represented in Figure 1 by a triangle and by a square respectively.

The primer 16 is designed to anneal 11 immediately downstream of the polymorphic base 12, 14 such that the most 3' nucleotide of the primer 16 hybridises to the base immediately 3' of the polymorphic base 12, 14. A simple approach to annealing lithe primer 16 and template 10 is to mix them together and heat them up sufficiently to denature the double-stranded template DNA 10. The temperature at which this occurs can also be influenced by the composition of the buffer, principally its ionic strength. The mixed solution is then cooled slowly to allow hybridisation to occur. The conditions (temperature and salt ii concentration) at the endpoint of the reaction should be chosen to minimise the formation of non-specific hybrids. They should also be compatible with subsequent steps of the protocol. The skilled person would appreciate this and know how to determine the best conditions.

Preferably, the primers 16 are immobilised at an end through a linker to the surface of a microarray slide 18 (for example, fabricated from quartz or glass). Such primers 16 may be synthesised in situ, or deposited on the surface using a spotting robot. In this case the annealing reaction II, and subsequent steps, will be performed on the slide 18 (Lindroos, K. et al. (2001) NucI. Acids Res. 29, E69-9). In an alternative method, the reactions are performed in solution and the reaction products subsequently immobilised on a slide 18 for visualisation. The primers 16 should be deposited at sufficiently low density such that incorporated fluorescence can be resolved at the level of the individual molecule. The spacing between molecules may be approximately 300nm, but depends on the wavelength of light used and the numerical aperture of the objective lens used.

Primer extension Primer extension 13 involves adding a single base 20 to the 3' end of the primer 16. The identity of the incorporated base is determined by the base immediately adjacent to the 3' end of the primer 16. This should be the position of the base 12, 14 in the template DNA whose genotype is to be determined. The reaction mixture includes the following components: U A buffer and salts appropriate to the reaction U The annealed primer-template complexes U Four modified (chain-terminating) deoxynucleotides 20. Typically ddNTPs are used, but other chain terminating substrates may be suitable. At least two of the four are labelled. The labelled deoxynucleotides 20 are those that are to be discriminated by the assay (i.e. one labelled base corresponding to one allele, the other labelled base corresponding to a second allele). Any suitable labels could be used, for example Cy3 and Cy5. The two labels preferably emit fluorescence at different wavelengths or have other distinguishable characteristics (such as different fluorescence lifetime). The other bases present help to ensure that fidelity of the labelling reaction has been maintained.

A DNA polymerase enzyme 20 to catalyse the primer extension.

The reaction is incubated under the optimal conditions for the reaction, which can be determined by the skilled person.

After the reaction 13 is complete, the slides 18 are washed to eliminate any unbound ddNTPs 20. Figure 3 shows the differential labelling of alleles of an SNP obtained.

Incorporation of an adenine nucleotide (4-pointed star) indicates presence of a thymine at the polymorphic site. Incorporation of a guanine nucleotide (6-pointed star) indicates presence of a cytosine at the polymorphic site. Therefore, molecules with different alleles can be distinguished by the type of florescence they emit.

Visualisation The fluorescently labelled primer-template complexes, attached to the surface of a slide 18, are then visualised 15. An appropriate technique is TIRFM (but other microscopy techniques may be suitable). This technique has been described in some detail elsewhere (Axelrod, D. (1989) Methods Cell Biol. 30, 245-270; Knight, A. et al. (2005) Eur. Biophys.

J. 35, 89). However, any technique giving a sufficiently high signal to noise ratio that single molecules can be detected is suitable.

In TIRFM, a high signal-to-noise ratio is achieved by using an evanescent wave to excite fluorescence in molecules attached to a surface. Evanescent waves penetrate a small distance (of the order of a wavelength of the light used to produce them) into a medium; and they decay exponentially away from the surface.

The excitation light (typically laser radiation) is incident on an interface between the glass slide and a less-dense medium (such as water or air). The angle of incidence must be greater than an angle O (the critical angle), when total internal reflection will occur. Under this condition, the incident beam is, as the name suggests, completely reflected from the interface. However, at the point of reflection, an evanescent wave penetrates a short distance into the second medium. This wave decays exponentially away from the interface, and therefore can be used selectively to excite fluorescence at the surface. This is useful for single molecule fluorescence, as it greatly enhances the signal to noise ratio.

To excite probes with spectrally distinct emission wavelengths, it is preferred that distinct excitation wavelengths are used, in other words, multiple lasers. Where the laser wavelength for one probe does not coincide with the transmission band of the emission filter for the other probe, multiple lasers may be used simultaneously. However, it may be advantageous to use the lasers alternately. This can be facilitated by the use of shutters.

The beams can be made coaxial by the use of dichroic mirrors, which transmit one wavelength and reflect another. Arrangements to fine-tune the angle of incidence and alignment of the beam will also be required. So that all the rays in the beam are incident at approximately the same angle, the beam should be reasonably well collimated. However, a long focal-length lens is useful to focus the beam and make a smaller, more intense spot. If the focal length is sufficiently long, the range of angles in the beam will be sufficiently small (i.e. it will be approximately collimated).

In a typical experiment, the molecules are visualised in a buffer solution. The solution is enclosed by a microscope cover slip, separated from the slide surface by 10 tm thick PTFE spacers. The fluorescence from the labelled molecules is captured by an oil-immersion microscope objective. This type of objective has a very high numerical aperture. This means that it collects light from the maximum possible solid angle, and also means that it has the highest possible resolution. However, such lenses also have a narrow depth of focus, which means that focussing becomes critical. They also have very short working distances and will suffer from spherical aberration when imaging through more than a few micrometres of water, meaning that thin specimens should be used. Typical magnifications that may be used are lOOx or 63x.

After collection by the objective lens, the light is passed through a filter optimised for transmission of the wavelengths of interest. Interference filters are preferred for this application, as they combine the maximum transmission in the band of interest with the maximum blocking of other wavelengths, and a very sharp transition between the two.

This helps to avoid residual excitation light from the laser, which is a major source of background. (Most fluorescent dyes have very small Stokes shifts, which means that the excitation and emission wavelengths may be very close together.) This also helps to eliminate other sources of stray light and background. A filter wheel (which may be motorised) orsimilar arrangement is preferred to exchange filters rapidly, so that the two (or more) fluorescent labels can be separately imaged. Alternatively, a "dual view" arrangement can be used allowing multiple dyes to be imaged simultaneously.

Most fluorescent dyes are prone to some form of photobleaching. This is a process whereby the excited state of the molecule undergoes an irreversible change into a non-fluorescent form. This process results in a loss of signal. If the photostabilities of the dyes in use differ, this could also lead to a bias in the measurement. Therefore it is preferred to minimise photobleaching, and to acquire data rapidly before this becomes a concern.

Photobleaching can be minimised by using the lowest illumination intensity needed to give a sufficient intensity of fluorescence. Bleaching is also typically an oxygen-dependent process, and can be greatly reduced by degassing buffer solutions and including an "oxygen scavenger" system and reducing agents in the observation buffer. Another method of reducing bleaching is to select the most photostable dyes available. Images may also be acquired so rapidly with sensitive cameras that bleaching is minimised.

A host of other photophysical and photochemical processes may modulate the intensity of fluorescence in a reversible fashion, leading to phenomena such as "blinking" of fluorophores. Provided that the characteristics of the dyes in question are well understood, such phenomena may be eliminated, or circumvented by, for example, time averaging or integration of fluorescence data, or manipulation of the experimental conditions.

The characteristics of the evanescent wave include the penetration depth, intensity and polarisation characteristics. These can be modified to suit experimental requirements by modifying a number of parameters of the input beam, including the angle of incidence, intensity, wavelength and polarisation, and also the refractive indices of the slide and the solution.

Optical microscopy techniques are limited in resolution by diffraction processes. A given imaging system has a characteristic known as the "point spread function", which is the shape of the distribution of the light in the image of a point source (and a single molecule is very close to being such a source). Two molecules which are so close that these distributions significantly overlap cannot be resolved or distinguished from each other.

However, where it is only necessary to count molecules this may still be determined by noting the greater intensity. It is preferred, however, that the assay be designed so that the spatial separation of the molecules is sufficient that they can be resolved and therefore counted unambiguously (e.g. by depositing the oligonucleotides at a low density on the surface). This is also aided by ensuring that the effective pixel size of the detector is somewhat less than the size of the point spread function (e.g. by using a high magnification).

The signal from the molecules is relatively small, so that a detector that is very sensitive, but with a low noise level, should be used. Such detectors include silicon intensified target and intensified CCD cameras. A modem detector that is well suited to this application is the electron-multiplying CCD. Greater sensitivity reduces exposure times and/or illumination intensities. Reducing noise has a similar effect, and makes it easier to unambiguously identify the molecules in the images obtained.

The use of an imaging detector allows the molecules to be spatially resolved, permitting them to be quantified by the unambiguous method of counting. It also allows data to be acquired from many molecules at once.

Image processing and data analysis Once images are obtained, they are processed to identify and count the single molecule "spots" in each image. Typically two images are obtained, one for each fluorescent label (i.e. one for each allele), and the ratio of the numbers obtained gives the relative frequencies of the alleles.

Figure 4 illustrates the principle of array-based genotyping. The upper part of Figure 4 shows schematically how frequencies of SNP alleles may be derived by counting single molecules. Here we are determining the frequency of the T allele (black spots) in the population (black+white spots). In the first example, there are 10 black spots and 10 white spots, giving a frequency of 50%. The dashed lines indicate the perimeter of the printed microarray spot. Similarly, counting the black and white spots in the other images yields the frequency of the T allele.

The lower panel shows how printing many oligonucleotide primer spots in an array enables many parallel genotypes to be determined in a sample. In reality, such a microarray could have as many as tens of thousands of spots.

The precise image processing method used for the counting may vary. Steps could include corrections for uneven illumination, and noise reduction by averaging or median filtering in space or time. One strategy is to segment the image into spots and background by thresholding at a certain intensity level. The threshold level may be fixed or selected adaptively, based on the characteristics of the image. After thresholding, further processing may be used to clean up the thresholded image, for example by removing isolated pixels (so-called binary morphological operations). Objects of the correct size to be single molecule images are then counted. Alternatively, various algorithms based on cross correlation may be used to identif' features in the image corresponding to the expected point spread function of a single molecule.

In determining the frequencies of the alleles, a certain number of molecules should be counted to obtain statistically significant results. Where one allele is much less abundant than the other, more images should be acquired to enable sufficient molecules to be counted.

The frequency of a given allele is given byfA = flA/(flA+flB), wherefA is the frequency of allele A; flA is the number of molecules of allele A detected; nn is similarly the number of molecules of allele B detected; and only these two alleles of the polymorphism are present.

There are a number of advantages to the above described method.

Since it is a single-step reaction, a washout step could be used before visualisation. This would permit the use of higher concentrations of label than would be possible in a dynamic experiment.

The relative frequencies of alleles can be assessed simply by counting fluorescent spots.

Data analysis is much simpler because video data is not used. Only still images are needed Since the method involves counting individual labelled nucleic acid molecules, the above-mentioned problems relating to measurement of a bulk quantity (fluorescent intensity) are circumvented. Moreover, when individual molecules are analysed, it is much easier to distinguish signal fluorescence from background fluorescence.

As the molecules are observed as a population of individuals rather than as an amorphous bulk typically consisting of many millions of molecules, more detailed information about the population of molecules can be obtained.

The above-described method allows multiplexing, i.e. the ability to perform more than one genotyping assay in the same reaction vessel. This can be achieved in several ways, for example, by using probes labelled with fluorescent markers having different wavelengths, or probes being arrayed on a microscope slide.

The use of single molecule detection technology provides more quantitative results and the ability to perform parallel analysis of complex mixtures. lf combined with microarray technology, where a plurality of different primers is deposited on a surface of a slide, high-throughput screening of multiple SNPs in parallel can be achieved.

The TIRF method is very robust and reliable and is therefore preferred.

The skilled person would appreciate that many modifications could be made to the above described method.

In some embodiments it may be desirable to add more than a single base to the annealed primer. This may be in particular where the polymorphism involves more than a single nucleotide.

The reaction products of the labelling step could be immobilised on the slide after the reaction. Whilst this would not permit such high throughput, might be more convenient in certain cases.

The complex could, in principle, be immobilised on the surface through the polymerase, the primer, or the template, as desired.

The reaction steps could be performed in a fluidic or microfluidic device, to permit automation of the reaction process. A suitably designed device might be compatible with simultaneous visualisation.

The reaction could be monitored in real time. if the concentration of the labelled ddNTPs is sufficiently low that background fluorescence is not a problem, then single nucleotide addition steps could be observed. The labelled free ddNTPs diffuse too rapidly to form an image. On incorporation, a spot appears where the labelled ddNTP is immobilised. This spot might subsequently disappear due to photobleaching. It is proposed that a sequence of images are acquired and analysed for incorporation events. This variation will have several advantages. Because the data are acquired dynamically, an image could be selected where the spots reached an optimum density for quantitation. This would allow results to be obtained over a wider range of concentrations.

An automated positioning system could be used to image each array spot in turn.

The visualisation could be performed in air. Air lenses typically have a lower numerical aperture (equivalent to the solid angle over which light is collected), which means that they would offer lower sensitivity and resolution. However, an advantage is greater simplicity of visualisation and more rapid exchange of slides. In this case, the slides could be washed and dried before visualisation.

An image-processing algorithm could be used to count the molecules in each image. Such algorithms could include an adaptive threshold ing step, followed by feature identification.

Alternatively, some form of cross-correlation with a template image could be used.

The oligonucleotides could be arrayed on the surface of a microscope cover slip. They could then be visualised by a variant of TIRF known as objective-type TIRF, where the exciting laser beam is introduced through the objective lens.

Many different types of enzyme could be used to perform the labelling reaction, in which the incorporated base is complementary to the base of interest in the template strand.

In a small proportion of reactions, the enzyme may incorporate the wrong base. Careful control of the reaction conditions is used to minimise this, as it directly translates into a measurement error. One way to reduce this problem is to use enzymes that have a "proof-reading" capability. This relies on an exonuclease activity, which removes misincorporated bases; a method for applying this to minisequencing reactions is described in WO 2004/003228.

In the most common implementation, the method would be used with two labelled ddNTPs. In many cases, the discrimination between two alternative alleles is sufficient.

However, in some scenarios there might be more than two alleles of a particular SNP.

Also, in array applications, it may be desirable to characterise SNPs with different pairs of bases on the same array. In principle, this is simply a matter of using an additional labelled base, and possibly an additional laser and set of filters.

The nucleotides need not be fluorescently labelled. Other suitable types of label include Raman-active molecules and Raman-active nanoparticles. If fluorescently labelled it could be via fluorescent dye or fluorescent nanoparticles, for example.

As indicated above, several different visualisation methods could be used, for example, scanning confocal microscopy, epifluorescence microscopy, near field optical scanning probe microscopy, fluorescence correlation spectroscopy, Raman microscopy, or variants thereof.

Optical waveguides could be used as an alternative to TIRF. These devices consist of a high refractive index material, within which light propagates. As with TIRF, an evanescent wave is generated at the interface with the medium of lower refractive index.

A different optical arrangement is used to couple the exciting laser illumination into the waveguide, possibly incorporating fibre optics. Such an arrangement may be physically more compact and better suited to end-user instrumentation.

A phenomenon related to TIRF is Surface Plasmon Resonance. This is a phenomenon that occurs when the surface of the slide is coated with a thin layer of a metal such as gold, silver, or aluminium (Axelrod (1989)). When the beam is incident at the critical angle, a greatly increased intensity is obtained, whereas at other angles very low levels of light penetrate. This could potentially be used to increase the signal to background ratio.

Conventional epifluorescence illumination can be used to image single molecules, and this might be an alternative detection method. However, very critical work must be taken to eliminate sources of background. See Funatsu eta!. (1995) Nature 374, 555-559.

Confocal microscopy is a technique that restricts fluorescence excitation to a very small volume using a laser beam focussed to a diffraction-limited spot. This could, be used for single molecule detection as an alternative to TIRF.

Two photon microscopies are an extension of confocal microscopy where non-linear optical effects are exploited to further restrict the excitation volume. This could also be used instead of TIRF.

The above-described technique could have many different uses. The main application is envisaged to be the characterisation of frequencies of SNP alleles in pooled samples from populations of individuals. The goal of such experiments would typically be to identify correlations between the frequencies of particular alleles and certain phenotypes.

However, the technique is more widely applicable in that it can be used in any scenario where it is desirable to measure the frequencies of single-base differences in mixed samples of DNA. Such samples do not necessarily arise from pooling, but may be naturally occurring.

An example of naturally occurring mixed samples could be the analysis of viral DNA from a patient. Typically mutations will occur in the viral genome during the progress of an infection. Certain specific mutations may be known to be associated with a particular phenotype such as drug resistance; for example, in HIV. Monitoring the frequencies of such mutations could be used in the management of treatment; for example, if a mutation known to cause resistance to one drug began to increase in frequency, the patient could be switched to a different drug before the whole viral population became resistant. The method described would be ideal for this type of analysis, particularly as many different mutations could be monitored simultaneously through the array-based approach.

A further example could be the monitoring of human or animal populations for the occurrence of pandemic strains of avian influenza. Samples of influenza virus from many individuals could be pooled to monitor the frequency of particular mutations in the population.

Other organisms could be profiled in a similar way, where it is desired to measure the frequencies of certain mutations in particular populations.

Another situation where mixed samples may arise is in forensics. Crime scene samples may contain DNA from several individuals, including the perpetrator(s), victims(s), and innocent individuals. DNA fingerprinting is performed using short tandem repeats rather than SNPs, however some SNPs might contain useful information about a perpetrator where there is no match in the database, for example genetically determined aspects of appearance such as hair or eye colour.

One of the advantages of this method is that it allows multiplexing, or scaling-up, of single molecule detection techniques, which enables a significant increase in throughput. In other words many different samples can be analysed in a single reaction.

For genotyping, the use of chain terminating nucleotides ensures that the fluorescence detected is from the position of interest, and not from some other position "downstream".

The requirement for collecting sequences of images in the prior art complicates data acquisition, as compared to the present method, which only needs static images.

To explain the multiplexing issue: o To achieve sufficient magnification to enable single-molecule imaging, high magnification objectives are used. This limits the field of view to approximately the size of a single printed microarray spot.

o Therefore, in the present method, the spots in the array are visualised sequentially, through the use of a motorised stage. This means that each spot is visualised at a different time.

a An advantage of the proposed method, then, is that it is a static method -the image only reads out the end-point of the labelling reaction. Because we have used a primer and a chain terminating nucleotide, we can be confident that the labelling is specific to the site of interest.

EXAMPLE

Introduction

A model system based on plasmid DNA molecules and primers that anneal near the junction between the polylinker region and the plasmid backbone is now described. This model system illustrates the principle of using single molecule detection for genotyping of SNPs. The base adjacent the polylinker is either an A or G dependent on the polylinker orientation.

Model System The model system consisted of the plasmids pBluescript 11 KS+ and SK+ (Stratagene) (Alting-Mees and Short (1989) Nucleic Acids Research 17, 9494) and a primer complementary to a region adjacent to the multiple cloning site. Figure 5 illustrates the sequences at the boundary between the plasmid backbone (left) and the multiple cloning site (right). The sequence of the primer is indicated by a single underline. The first base of the multiple cloning site is the model polymorphism, and is indicated by a double underscore. The remaining multiple cloning site sequence is indicated with a dashed underline. The primer anneals to the complementary sequence (lower strand). Immediately 5' to the primer-binding site is either a I (KS, polymerase incorporates an A) or a C (SK, polymerase incorporates a G).

Depending on the orientation of the multiple cloning site, the base immediately 3' of the primer on the complementary strand is either a C or a I. Therefore the complementary base that will be added by the polymerase is either a Got an A. Since the nucleotide analogues used are chain terminators, the subsequent bases are not significant for the outcome of the method. By combining the SK and KS plasmids in different ratios, different allele frequencies can be simulated.

Methods and Materials Materials Template DNA Plasmid DNA was purified by using the Fastplasmid Mini purification kit (Eppendorf, Hamburg, Germany). Concentration and purity were determined by absorbance spectroscopy.

Primers The sequence of the primer used is given in Figure 5. Primers were synthesised and purified by MWG (Ebersberg, Germany).

Labelling Reagents Labelling ("minisequencing") reactions were performed using a Perkin-Elmer AcycloPrimeTM-FP SNP detection kit using a G/A terminator mix. This mix contains RI 10-acyGTP, TAMRA-acyATP, and two unlabelled terminators (where "acy" indicates the "acyclo" nucleotide analogues that are used as terminators). The kit also includes a mutated thermostable polymerase (AcycloPolTM) and a reaction buffer.

Methods Minisequencing Reactions Reactions were performed as recommended by the manufacturer, except where otherwise stated. Each reaction mix was made up as follows: Component Volume (gil) AcycloPolTM 0.05 lOx Reaction buffer 2 AcycloTerminatorTM mix G/A SNP Primer (10mM) 0.5 Water 6.45 Template DNA (50 ng) 10 Total 20 Reactions were performed in Thermo-Fast 96 wells plates (Abgene, Epsom, UK). Plates were sealed with adhesive plate seals (Abgene, Epsom, UK) and the minisequencing reactions were run on a GeneAmp PCR system 2700 (Applied Biosystems, Foster City, USA). The programme used was as follows: Temperature ( C) Duration (s) Cycles 120 1 15 30 4 120 After the completion of the reactions, the samples were stored al -20 C until required.

Prepa ration of Samples for Visualisation To minimise photobleaching during observation, an "oxygen scavenger" solution was prepared. At working concentration, this consisted of 0.02 mg/mI catalase, 0.1 mg/mT glucose oxidase, 20 mM dithiothreitol (DTT) and 3 mg/mI glucose in I x Tris-EDTA (TE) buffer pH 8. Sample reactions were diluted typically by a factor of:10 in TE followed by a further 1:10 dilution into the oxygen scavenger solution.

To immobilise the molecules for visualisation purposes, polylysine-coated microscope slides were used (Menzel Glaser, Braunschweig, Germany). Visualisation chambers were constructed from a slide, a 22 x 22 mm cover slip, and spacers made from strips of 10 im thick PTFE film (Goodfellow). After the addition of a 10 tl aliquot of sample, the cell was sealed using clear domestic nail varnish. After 5-10 minutes for the nail varnish to dry and DNA molecules to bind to the surface, the slide was placed on the microscope for visualisation.

Instrumentation The system used for TIRF imaging of the labelled DNA molecules is shown in 6. This is a schematic of the apparatus that was used in the experiments described. Fluorescence is excited using one of three lasers 60, the beams of which are combined using a series of dichroic mirrors 62 and a polarisation rotator 90. The beam is focussed to a spot at the viewing position using a lens 66. The angle of incidence of the beam is controlled using the translation stage 68 a rotation stage 70 and a pentaprism 72. A cubic quartz prism 74 is used to couple the laser beam into the microscope slide 76. Total internal reflection occurs at the interface between the slide 76 and the sample medium 78 and generates an evanescent wave. Any fluorescence is captured using a high numerical aperture I OOx 1.4 NA oil immersion objective lens 80 and the emission wavelength selected using an interference filter 82 before being imaged by an electron-multiplication CCD camera. For clarity, many beam-steering mirrors and other components have been omitted. (Other components shown include safety interlock shutters 64 and excitation selection shutters 84; mirrors 88, and cover slip 86).

The system is based on an Olympus 1X71 inverted microscope, equipped with an Andor iXon EMCCD camera. A filter wheel on the emission port of the microscope contained interference filters (Semrock, Rochester, NY, USA) to select the wavelength range appropriate to the fluorescent probe. Shutters were used to select the appropriate laser, as indicated in the table below. Image sequences of 100 or 200 frames were collected with an exposure time of I frame per second. For genotyping experiments, the identical area of the specimen was imaged with both laser/filter combinations to enable both dyes to be counted.

Dye Base Excitation Laser Emission Filter Wavelength (nm) Band Centre (nm) Band Width (nm) RhO ci 488 536 40 TAMRA A 532 593 40 Data analysis Data acquisition and analysis were performed with software developed using MATLAB and the Image Processing and Signal Processing Toolboxes (The Math Works, Natick, MA, USA). A number of algorithms are possible to detect and count the molecules present in the images. For demonstration purposes, we selected a relatively simple image-processing algorithm, as follows: 1. An average image was calculated for each image sequence.

2. A background image was calculated from the mean image, using a morphological opening operation with a "rolling ball" structuring element with a diameter larger than the expected size of a single molecule image.

3. The background image was subtracted from the mean image.

4. The image was thresholded to identify objects (groups of connected pixels).

5. The objects were filtered to remove objects that were too small or too large. Also, objects that overlapped in the RI 10 and TAMRA fluorescence images were removed.

6. The number of remaining objects in each image was counted.

For display purposes, the background-subtracted images from Step 3 above were contrast-enhanced and combined as a false colour image. The binary images obtained in Step 5 were merged in a similar way.

Results Figure 7 shows the results obtained. The experiment shown used a mix of equal quantities of the SK and KS forms of the plasmid. The images are 41 im (256 pixels) square and represent an average of 100 frames with each filter. The exposure time per frame was I s.

a) TAMRA fluorescence image and b) RI 10 fluorescence image. Note the roughly equal numbers of spots in each image; typically, each spot corresponds to a single molecule.

Some of the larger spots may be aggregates or contaminating particles and are excluded from the analysis, as are spots that occur in both channels. The contrast in each image has been enhanced for clarity.

c) and d) Results of feature analysis of the images in a) and b). The images were thresholded and objects consisting of groups of connected pixels were analysed. Objects which contained too few, or too many, pixels to be consistent with the point spread function were excluded, as were any objects which shared pixels with an object on the other channel. The remaining objects were then counted. In this case, 41 RHO-labelled and 43 TAMRA-labcl led objects were identified.

From the results of analysing an equimolar mixture of the two templates (SK and KS, corresponding to the G and A genotypes respectively) 41 RI 10-labelled (G allele) and 43 TAMRA-labelled (A allele) objects were counted. Therefore the relative frequency of the 1 allele is estimated as 49%.

Discussion As expected, approximately equal numbers of spots of each colour were visible in the and identified by the image analysis. The number of molecules counted would be expected to follow a Poisson distribution, where the standard deviation is equal to the root of the number of objects. Therefore, by increasing the number of objects counted, the confidence in the value obtained can be increased. This could be achieved by collecting larger images, or collecting more images from other fields of view; alternatively, the surface density of molecules could be increased.

Where rare alleles are to be quantitated, more molecules should be counted to give the same level of confidence in the frequency.

The above-described exemplary method makes a useful model system for the minisequencing reaction. In these experiments, the reactions were performed in solution and the reaction products then immobilised on the surface of a microscope slide for visual isation. This demonstrates the principle of using single molecule imaging to analyse the products of a genotyping reaction. In practice the preferred approach would be to perform the reactions on a slide with an array of immobilised primers, as described elsewhere in this application. In addition to permitting multiplexing, such an approach would also enable washing of the slides to reduce non-specific binding.

Claims (24)

1. I. A method of determining the frequency of an allele of a polymorphism in a population of nucleic acid molecules including: providing a target nucleic acid, the target nucleic acid including a known or suspected polymorphism; providing at least two distinguishably labelled nucleotides, the first labelled nucleotide corresponding to a first genotype of the known or suspected polymorphism, the second labelled nucleotide corresponding to a second genotype of the known or suspected polymorphism; synthesising an oligonucleotide complimentary to a region of the target nucleic acid, the region including the known or suspected polymorphism; wherein the complimentary oligonucleotide includes the first labelled nucleotide corresponding to the first genotype of the known or suspected polymorphism, or wherein the complimentary oligonucleotide includes thc second labelled nucleotide corresponding to the known or suspected polymorphism; detecting individual labelled nucleic acid molecules using optical microscopy, and counting the number of oligonucleotides including each labelled nucleotide, thereby to determine the number of nucleic acid molecules in the population having each genotype.
2. 2. A method as claimed in claim 1, wherein the polymorphism is a single nucleotide polymorphism.
3. 3. A method as claimed in claim I or 2, wherein synthesising the oligonucleotide includes extending a primer that has been annealed to the target nucleic acid.
4. 4. A method as claimed in claim 1, 2 or 3, wherein the primer is linked to a surface.
5. 5. A method as claimed in claim 4, wherein a plurality of different primers is linked to the surface.
6. 6 A method as claimed in claim 5, wherein the plurality of different primers is linked to the surface in an array format.
7. 7. A method as claimed in any preceding claim, wherein the primer is designed to anneal to the target nucleic acid at a position immediately 3' of the polymorphism.
8. 8 A method as claimed in claim 7, wherein the primer is extended by a single base.
9. 9 A method as claimed in any preceding claim wherein extending the primer includes mixing the target nucleic acid with a mixture of chain-terminating nucleotides, wherein at least one of the chain-terminating nucleotides corresponding to one of the alleles of the polymorphism is labelled.
10. 10. A method as claimed in claim 9, wherein the mixture of chain-terminating nucleotides includes one type of nucleotide corresponding to one allele of the polymorphism labelled with one type of label and another type of nucleotide, corresponding to another allele of the polymorphism, being labelled with a different type of label.
11. 11. A method as claimed in any preceding claim wherein nucleotides that do not correspond to an allele of the polymorphism in the nucleic acid are included in the mixture, but are not labelled.
12. 12. A method as claimed in any preceding claim, wherein the optical microscopy is Total Internal Reflection Fluorescence Microscopy or a variant thereof.
13. 13. A method as claimed in any of claims Ito II, wherein the optical microscopy is scanning confocal microscopy or a variant thereof.
14. 14. A method as claimed in any of claims Ito I 1, wherein the optical microscopy is epifluorescence microscopy of a variant thereof.
15. 15. A method as claimed in any of claims Ito II, wherein the optical microscopy is near field optical scanning probe microscopy or a variant thereof.
16. 16. A method as claimed in any of claims Ito 11, wherein the optical microscopy is fluorescence correlation spectroscopy or a variant thereof.
17. Il. A method as claimed in any of claims Ito ii, wherein the optical microscopy is Raman microscopy or a variant thereof.
18. 18. A method as claimed in any preceding claim, wherein the nucleotides are fluorescently labelled.
19. 19. A method as claimed in claim 18, wherein the labels include fluorescent dyes.
20. 20 A method as claimed in claim 18 or 19, wherein the labels include fluorescent nanoparticles.
21. 21 A method as claimed in any of claims Ito 17, wherein the nucleotides are labelled with Raman-active labels.
22. 22. A method as claimed in claim 21, wherein the labels include Raman-aetive molecules.
23. 23. A method as claimed in claim 21 or 22, wherein the labels include Raman-active nanoparticles.
24. 24. A method substantially as herein before described, with reference to, and as illustrated in, the accompanying drawings.

    24. A method substantially as herein before described, with reference to, and as illustrated in, the accompanying drawings.

    amei..4ments to the claims have been filed as follows

    I. A method of determining the frequency of an allele of a polymorphism in a population of nucleic acid molecules including: providing a target nucleic acid, the target nucleic acid including a known or suspected polymorphism; providing at least two distinguishably labelled nucleotides, the first labelled nucleotide corresponding to a first genotype of the known or suspected polymorphism, the second labelled nucleotide corresponding to a second genotype of the known or suspected polymorphism; synthesising an oligonucleotide complementary to a region of the target nucleic acid, the region including the known or suspected polymorphism; wherein the complementary oligonucleotide includes the first labelled nucleotide corresponding to the first genotype of the known or suspected polymorphism, or wherein the complementary ohgonucleotide includes the second labelled nucleotide corresponding to the second genotype of the known or suspected polymorphism; visualising individual labelled nucleic acid molecules by obtaining a still image of the labelled nucleic acid molecules using optical microscopy; and counting the number of oligonucleotides in the image labelled with each nucleotide, thereby to determine the number of nucleic acid molecules in the population having each genotype.

    2. A method as claimed in claim 1, wherein the polymorphism is a single nucleotide polymorphism. p

    3. A method as claimed in claim 1 or 2, wherein synthesising the oligonucleotide includes extending a primer that has been annealed to the target nucleic acid. S...

    4. A method as claimed in claim 1,2 or 3, wherein the primer is linked to a surface.

    * *. 5. A method as claimed in claim 4, wherein a plurality of different primers is linked to :. the surface. **I. p P

    6. A method as claimed in claim 5, wherein the plurality of different primers is linked to the surface in an array format.

    7. A method as claimed in any preceding claim, wherein the primer is designed to anneal to the target nucleic acid at a position immediately 3' of the polymorphism.

    8 A method as claimed in claim 7, wherein the primer is extended by a single base.

    9. A method as claimed in any preceding claim wherein extending the primer includes mixing the target nucleic acid with a mixture of chain-terminating nucleotides, wherein at least one of' the chain-terminating nucleotides corresponding to one of the alleles of the polymorphism is labelled.

    10. A method as claimed in claim 9, wherein the mixture of chain-terminating nucleotides includes one type of nucleotide corresponding to one allele of the polymorphism labelled with one type of label and another type of nucleotide, corresponding to another allele of the polymorphism, being labelled with a different type of label.

    II. A method as claimed in any preceding claim wherein nucleotides that do not correspond to an allele of the polymorphism in the nucleic acid are included in the mixture, but are not labelled.

    12. A method as claimed in any preceding claim, wherein the optical microscopy is Total Internal Reflection Fluorescence Microscopy or a variant thereof.

    13. A method as claimed in any of claims Ito II, wherein the optical microscopy is scanning confocal microscopy or a variant thereof.

    14 A method as claimed in any of claims Ito II, wherein the optical microscopy is epifluorescence microscopy of a variant thereof.

    15. A method as claimed in any of claims Ito 11, wherein the optical microscopy is near field optical scanning probe microscopy or a variant thereof.

    16 A method as claimed in any of claims Ito II, wherein the optical microscopy is fluorescence correlation spectroscopy or a variant thereof.

    17. A method as claimed in any of claims Ito 11, wherein the optical microscopy is Raman microscopy or a variant thereof.

    18. A method as claimed in any preceding claim, wherein the nucleotides are fluorescently labelled.

    19. A method as claimed in claim 18, wherein the labels include fluorescent dyes.

    20. A method as claimed in claim 18 or 19, wherein the labels include fluorescent nanoparticles.

    21. A method as claimed in any of claims I to 17, wherein the nucleotides are labelled with Raman-active labels.

    22. A method as claimed in claim 21, wherein the labels include Raman-active molecules.

    23. A method as claimed in claim 21 or 22, wherein the labels include Raman-active nanoparticles.