WO2012059748A1

WO2012059748A1 - Method, apparatus and software for identifying cells

Info

Publication number: WO2012059748A1
Application number: PCT/GB2011/052122
Authority: WO
Inventors: Chris Denning; Daphne Goh; Ioan Notingher; Flavius Cristian Pascut
Original assignee: The University Of Nottingham
Priority date: 2010-11-04
Filing date: 2011-11-01
Publication date: 2012-05-10
Also published as: GB201018628D0

Abstract

This invention concerns a method and apparatus for identifying a spectral signature of a target cell type that can be used to distinguish the cells of that cell type from other cell types in a mixture. The invention also concerns a method and apparatus for identifying cells of a target cell type from cells of other cell types in a mixture using the spectral signature. The method for identifying a spectral signature comprises obtaining a target cell spectrum for each of a plurality of cells of the target cell type using Raman Spectroscopy, the target cell spectrum resulting from exciting a substantial portion of the cell of the target cell type with a laser and obtaining a non- target cell spectrum for each of a plurality of cells of one or more other cell types using Raman Spectroscopy, the non-target spectrum resulting from exciting a substantial portion of the cell of the other cell type with a laser. The target cell spectra are compared to the non-target cell spectra to identify a spectral signature characteristic of target cell spectra that distinguishes cells of the target type from cells of the one or more other types.

Description

METHOD, APPARATUS AND SOFTWARE FOR IDENTIFYING CELLS

This invention concerns a method, apparatus and software for identifying cells and in particular, but not exclusively, for identifying cardiomyocyte cells. The invention has particular, but not exclusive application to separating one type of cells, such as cardiomyocyte cells, from a mixture of different types of cells.

Background of the Invention Cell therapy and regenerative medicine are widely acknowledged as key medical technologies of the 21 ^st century which promise to provide treatments for many currently incurable diseases. These technologies require high-purity cell populations with specific phenotype, depending on the treatment. Research during the last decade has shown that a large number of cell types can be derived from pluripotent cells. However, the conditions to derive specific cell types remain suboptimal, differentiation protocols are not standardized, generally producing only low yields of the desired differentiated lineages within highly heterogeneous populations that are not suitable for clinical use due to the presence of mainly unwanted cell types. For example for cardiomyocytes, strategies have been developed to enrich the number of cardiomyocytes derived from human embryonic stem cells (hESC) to >90% purity involving complex genetic modification that is not ideal for the clinic. In order to rapidly overcome these obstacles and enable the delivery of validated hESCs for clinical use, further technological advances are required, in particular in manufacturing and quality assessment of these therapeutic products. It is desirable that such technologies are robust, automated, to enable integration with existing manufacturing technologies, and comply with the strict criteria of drug regulatory agencies.

Non-invasive techniques able to monitor biochemical changes in individual cells (eg including in vitro differentiation of pluripotent cells) or to characterise the purity of cell populations, are of particular interest. Such technologies may overcome the limitations of the currently available techniques (eg western blotting, PCR, etc) which are invasive and are not suitable for characterising heterogeneous populations since they require a large number of cells and results represent averages over entire cell populations. Other techniques working on single live cells, such as fluorescence imaging, rely on existence of lineage-specific surface markers, which are expressed only in a limited number of cell types, or require laborious and expensive genetic modification of cells. Chan, J. W. and Leiu, D. K., Huser, T. and Li, R. A. 2009, "Label-free separation of human embryonic stem cells and their cardiac derivatives using Raman spectroscopy", Analytical Chemistry 81 : 1324- 1331 (hereinafter referred to as "Chan"), discusses label-separation of human embryonic stem cells and their cardiac derivatives using Raman Spectroscopy. A laser focus was positioned at the centre of a cell and spectra were acquired for 2 minutes. The reported discrimination accuracy for cardiomyocytes against undifferentiated hECs was 66%, below what is likely to b e required for a commercially useful application.

Summary of Invention

According to a first aspect of the invention there is provided a method of identifying a spectral signature of a target cell type that can be used to distinguish the cells of that cell type from other cell types in a mixture comprising obtaining a target cell spectrum for each of a plurality of cells of the target cell type using Raman Spectroscopy, the target cell spectrum resulting from exciting a substantial portion of the cell of the target cell type with a laser, obtaining a non-target cell spectrum for each of a plurality of cells of one or more other cell types using Raman Spectroscopy, the spectrum resulting from exciting a substantial portion of the cell of the other cell type with a laser and comparing the target cell spectra to the non-target cell spectra to identify a spectral signature characteristic of target cell spectra that distinguishes cells of the target cell type from cells of the one or more other types.

By obtaining spectra from exciting a substantial portion of the cell, it has been found that the signature characteristics identified can be used to distinguish cells of the target cell type from cells of other cell types with discrimination accuracy better than 66%. It is believed that one of the reasons for the increased accuracy is that cells tend to have high intracellular chemical variability, which leads to high variations in the Raman spectra measured at various locations inside the cell. Therefore, in order to accurately identify spectral signatures for a particular cell type is it advantageous to obtain a spectrum that is representative of a substantial portion of the cell. The target and/or non-target spectrum may be obtained by exciting a substantial portion of the cell with a laser having a laser spot (or multiple laser spots) with an area of the order of the footprint of the cell (for example, the laser spot may have a diameter of 30 to 130 microns).

However, preferably, the method may comprise generating multiple spectra associated with different portions of the cell (of the target cell type or other cell type). The multiple spectra may be processed to determine differences between the target cell spectra and the non-target cell spectra that are statistically significant. The multiple spectra may be obtained by scanning (a substantial portion) of each cell with a laser, for example as a rastor scan. A substantial portion of the cell may be more than 50% of the cell, more than 75% of the cell, preferably more than 90%> of the cell and even more preferably more than 99% of the cell.

Processing of the multiple spectra may comprise a principal component analysis of the spectra. The signature characteristics of the target cell spectra that distinguish cells of the target cell type from cells of the one or more other types may be a threshold score for one or more principal components of the principal component analysis.

The laser used for obtaining the spectra may emit light in the near-infra-red (700 to l OOOnm). It has been found that cells may be exposed to such a laser for longer periods and at higher power densities relative to laser light in the visible range (400 to 600nm.)

The method may comprise checking, either before or after carrying out Raman spectroscopy, that the plurality of target cells are cells of the target cell type. The checking step may be carried out using immuno- fluorescence techniques. In this way, the results of the method are ensured to relate to characteristics of the target cells as a cross-check has been carried out to ensure that the plurality of target cells are in fact target cells. Such a technique is preferred over simply relying on a process producing cells of the target cell type, as such processes may not produce a high yield of the target cells leading to erroneous results. The method may be carried out for a range of target cells and the method further comprise compiling a database of signature characteristics for each cell type. The information in such a database may be used in methods for identifying cells. According to a second aspect of the invention there is provided a data carrier having stored thereon instructions, which, when executed by a processor, cause the processor to carry out the method according to the first aspect of the invention.

According to a third aspect of the invention here is provided a data carrier having stored thereon a database of spectral signatures determined in accordance with the first aspect of the invention.

The data carrier may be any suitable storage device such as a hard disk, RAM, ROM, CD-ROM, memory stick or the like or any suitable signal for conveying data, such as an electromagnetic or electronic signal.

According to a fourth aspect of the invention there is provided apparatus comprising a processor, the processor arranged to carry out the method of the first aspect of the invention. The apparatus may further comprise a device for carrying out Raman spectroscopy and automatically feeding the detected spectrum to the processor.

According to a fifth aspect of the invention there is provided a method of identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells comprising:- obtaining a spectrum for a candidate cell generated using Raman Spectroscopy, and

determining whether the candidate cell is a target cell by determining whether the spectrum has a spectral signature characteristic of a target cell, wherein the spectral signature has been determined by the method of the first aspect of the invention.

The target and/or non-target spectrum may be obtained by exciting a substantial portion of the cell with a laser having a laser spot (or multiple laser spots) with an area of the order of the footprint of the cell (for example, the laser spot may have a diameter of 30 to 130 microns). However, preferably, the method may comprise generating multiple spectra associated with different portions of the cell (of the target cell type or other cell type). The multiple spectra may be processed to determine differences between the target cell spectra and the non-target cell spectra that are statistically significant. The multiple spectra may be obtained by scanning (a substantial portion) of each cell with a laser, for example as a rastor scan. A substantial portion of the cell may be more than 50% of the cell, more than 75% of the cell, more than 90%> of the cell and preferably more than 99% of the cell.

The method may comprise the additional step of separating cells identified as cells of the target cell type from cells identified as cells of other types. In this way, a sample of cells is provided having a greater purity than of cells of the target cell type than the mixture of candidate cells. In another embodiment, the method may be used to evaluate a process for forming cells of the target cell type, the number of cells identified as being of the target cell type or the ratio of cells of the target cell type to cells of other types providing an indication of the effectiveness of the process for forming cells of the target cell type. The spectral signature characteristic of a target cell may be a score for a first principal component of a principal component analysis falling within an expected range of values. For example, the range of values may be a score above or below a threshold value. According to a sixth aspect of the invention there is provided apparatus for identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells comprising:- memory having stored therein a spectral signature characteristic of a target cell, wherein the spectral signature has been determined by the method of the first aspect of the invention;

a processor arranged to obtain a spectrum for a candidate cell generated using Raman Spectroscopy, and

determine whether the candidate cell is a target cell by determining whether the spectrum has the spectral signature characteristic of a target cell. The apparatus may further comprise a separator, such as optical tweezers, for separating cells identified as cells of the target cell type from cell of other types.

The calls may be stem cells. The target cell type may be cardiomyocyte cells.

According to a seventh aspect of the invention there is provided a method of identifying cardiomyocyte cells form other cells in a mixture of candidate cells comprising:- obtaining a spectrum for a candidate cell using Raman Spectroscopy, determining whether a value of the spectrum at one or more wave-numbers selected from the group consisting of:

i) 825 to 950 cm ¹

ii) 1035 to 1 170 cm ¹; and

iii) 1320 to 1420 cm ¹

is within a range of expected values for cardiomyocyte cells, and

identifying the candidate cell as a cell of the target cell type if the value is determined as being within the range of expected values.

The value may be an area under each peak at the one or more wave-numbers.

According to an eighth aspect of the invention there is provided a method of identifying cardiomyocyte cells from other cells in a mixture of candidate cells comprising:- obtaining a spectrum for a candidate cell using Raman Spectroscopy, determining from characteristics of the spectrum a score for a first principle component of a principle component analysis carried out on spectra of cardiomyocyte cells,

determining whether the score is within a range of expected values for cardiomyocyte cells, and

identifying the candidate cell as a cell of the target cell type if the score is determined as being within the range of expected values.

It has been found that use of such spectral signatures of the Raman spectrum of the cardiomyocyte cell can result in better discrimination of the cardiomyocyte cells from the other cells than prior art techniques. In particular, it has been found that looking at other spectral signatures of the Raman spectrum of the cardiomyocyte cell, such as the second, third, fourth, etc principal components, does not significantly improve the discrimination of cardiomyocyte cells from the other cells and, in fact, may reduce the accuracy of the discrimination.

Accordingly, the seventh aspect of the invention may comprise determining whether a value of the spectrum at only one or more wave-numbers selected from the group consisting of:

iv) 825 to 950 cm ¹

v) 1035 to 1 170 cm ¹; and

vi) 1320 to 1420 cm ¹

is within the range of expected values for cardiomyocyte cells.

The wavenumbers may be 854cm^{" 1} and/or 938cm^{" 1} and/or 1084cm^{" 1} and/or 1 123 cm^{" 1} and/or 1340cm^{" 1} and/or 1450 cm^" . The value may be a Raman peak area for a peak centred approximately at these wavenumbers.

The value may be the Raman peak area for a peak centred at 854 cm^{" 1}. This Raman peak may be assigned to CH₂ rocking vibrations in tyrosine and C-O-H ring vibrations in carbohydrates, this peak having a high intensity in myosin, actin and glycogen, which are highly abundant in parts of cardiomyocyte cells. The range of expected values may be a boundary set at the intersection between a probability distribution for the Raman peak area at 854 cm^{" 1} for non-cardiomyocyte cells and a probability distribution for the Raman peak area at 854 cm^{" 1} for cardiomyocyte cells. It has been found that using this characteristic to discriminate between cardiomyocyte cells and the other cells may provide 84.3% sensitivity and 100% specificity for identifying cardiomyocyte cells.

The value may be the Raman peak area for a peak centred at 938 cm^{" 1}. This peak is assigned to C-C and CH₃ stretching vibrations of certain proteins (rich in Lys, Asp, Leu and Val) and C-O-C glycosidic bond vibrations in carbohydrates. This Raman peak also has a high intensity in myosin, troponin and glycogen. The range of expected values may be a boundary set at the intersection between a probability distribution for the Raman peak area at 938 cm^{" 1} for non-cardiomyocyte cells and a probability distribution for the Raman peak area at 938 cm^{" 1} for cardiomyocyte cells. It has been found that using this characteristic to discriminate between cardiomyocyte cells and the other cells may provide 90.2% sensitivity and 100% specificity for identifying cardiomyocyte cells. The eighth aspect of the invention may comprise determining from characteristics of the spectrum a score for only the first principle component of the principle component analysis carried out on spectra of cardiomyocyte cells.

The realisation that confining consideration to these characteristics of the spectrum of cardiomyocyte cells provides adequate discrimination may allow processing of the Raman spectrum to be simplified without unduly affecting the results of the separation process.

According to a ninth aspect of the invention there is provided a method of identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells comprising:- obtaining a multiple spectra for a candidate cell generated using Raman Spectroscopy, each spectra associated with a different portion of the cell, and

determining whether the candidate cell is a target cell by determining whether the spectra have a spectral signature characteristic of the target cell type.

Description of the Drawings

Embodiments of the invention will now be described, by example only, with reference to the following drawings, in which:-

FIGURE 1 is an analysis of beating frequency of cardiomyocytes prior and after laser irradiation as required for collection of Raman spectra; FIGURE 2A shows bright-field images of typical live cells derived from hESCs;

FIGURE 2B shows Raman spectra of individual live cardiomyocyte (CM) and non-cardiomyocyte (non-CM); FIGURE 2C shows Immuno- staining images of the same cells: cell nuclei DAPI and cardiac a-actinin;

FIGURE 3 shows average Raman spectra of cardiomyocytes (CMs), non- cardiomyocytes (non-CMs) and their computed difference spectrum;

FIGURE 4A shows the first four PC loading spectra in the PCA analysis of Raman spectra of cardiomyocytes and non-cardiomyocytes; FIGURE 4B shows the distribution of the scores corresponding to the first principal component PC I for cardiomyocytes and non-cardiomyocytes;

FIGURE 5 shows a comparison between the loading spectrum of PC I and selected chemicals abundant in cardiomyocytes compared to other phenotypes;

FIGURES 6A to 6E show, respectively, bright-field; immuno -staining for cardiac phenotype (a-actinin) and cell nucleus (DAPI); PAS staining for glycogen; Raman spectral image corresponding to the 938 cm- 1 band; all obtained from a typical cardiomyocyte; and selected Raman spectra at positions indicated by the square and star (acquisition time 1 second per spectrum);

FIGURE 7 shows a distribution of the Raman peak area at 854cm^{" 1} for cardiomyocytes and non-cardiomyocytes;

FIGURE 8 shows a distribution of the Raman peak area at 938cm^{" 1} for cardiomyocytes and non-cardiomyocytes;

FIGURE 9 shows apparatus for identifying a spectral signature of a target cell type that can be used to distinguish the cells of that cell type from other cell types in a mixture; and

FIGURE 10 shows apparatus for identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells. Detailed Description of Embodiments

Described below is a technique for using Raman spectroscopy to identifying a spectral signature/marker specific to a target cell type, such as mature phenotypes derived from pluripotent cells, which can be used to distinguish the cells of that cell type from other cell types in a mixture. This identification is carried out on individual live cells maintained under sterile physiological conditions (culture medium and CO₂) and does not require preparation, fixation, labels or any other contrast enhancement. Using the phenotypic specific spectral features allows non-invasive phenotypic identification of individual live cells.

Embodiments are also described that use the spectral signatures/markers identified using this technique to identify cells of a target cell type from cells of other cell types in a mixture of candidate cells. In one embodiment, specific Raman spectral peaks are used to identify cardiomyocytes derived from human embryonic stem cells. In another embodiment, a score for a first principle component of a principle component analysis is used to differentiate between target cell types and other cell types.

Referring to Figure 9, apparatus for identifying a spectral signature/marker specific to a target cell type comprises a Raman spectroscopy system 1 for carrying out Raman spectroscopy on a sample 2 of cells. The system comprises a laser 6 that generates a laser beam having a spot size at the sample 2 of approximately 2 microns. The laser 6 is arranged such that the laser beam can be scanned across each cell of the sample 2 in incremental steps, for 2micron steps. Light from the spot is collected in a spectrometer 3 for detecting the generated Raman spectra. Data relating to the Raman spectra is sent from the spectrometer 3 to a processor 4. Processor 4 is arranged to process the received spectra to identify spectral signatures/markers specific to a target cell type. The processor 4 may be under the control of software stored in memory 5 and the resulting spectral signatures/markers identified by the processor may be stored in a database in memory 5.

In general, the method of identifying a spectral signature/marker of a target cell type comprises obtaining a target cell spectrum for each of a plurality of cells of a sample of target cells using the apparatus. The sample of target cells may have been checked by another technique to determine that the cells are in fact of the target cell type. Such a cross-check may be carried out by any suitable technique, for example using immuno-fluorescence techniques. In this embodiment, each cell of the sample is scanned by the laser to produce multiple spectra for each cell, each spectrum associated with a different portion of the cell. These multiple spectra fed to the processor and the processor averages the spectra to determine a single spectrum for each cell.

A similar method is then used to obtain non-target cell spectra for a sample of cells of other cell types.

The processor 4 then compares the target cell spectra to the non-target cell spectra to identify a spectral signature/marker characteristic of target cell spectra that distinguishes cells of the target type from cells of the one or more other types. An example of the use of Raman micro-spectroscopy (RMS) to detect and image molecular markers specific to cardiomyocytes (CMs) derived from human embryonic stem cells (hESCs) will now be described. However, it will be understood that similar techniques can be used to identify spectral signatures/markers for other types of cells.

Example

The technique of the invention is non-invasive and, therefore, can be used for discrimination of individual live CMs within highly heterogeneous cell populations as obtained following differentiation of hESCs. Principal component analysis (PCA) of the Raman spectra was used to build a classification model for the cells as CMs. Retrospective immuno-staining imaging was used as a gold-standard for phenotypic identification of each cells. The discrimination accuracy of CMs from other phenotypes was >97% specificity and >96 % sensitivity, calculated using cross- validation algorithms (target 100% specificity). Comparison between Raman spectral images corresponding to selected Raman bands identified by the PCA model and immuno-staining of the same cells allowed assignment of the Raman spectral markers. It was concluded that glycogen and myofibril proteins are responsible for the discrimination of CMs, glycogen being the main contributor. The study demonstrates that RMS can be used for non-invasive phenotypic identification of hESCs progeny. Such label-free optical techniques can be used in the separation of high-purity cell populations with mature phenotypes or enable repeated measurements to monitor time- dependent molecular changes in live hESCs during differentiation in-vitro.

Individual live cells were analysed by RMS and retrospective phenotypic identification of all cells was carried out by the gold-standard method of immunofluorescence imaging integrated on the Raman microscope. The effect of the laser irradiation on cells was evaluated by performing a beating frequency analysis on individual CMs. The Raman spectra of the cells were analysed using multivariate statistical methods with the aim of developing high-accuracy models for discrimination of CMs from other unwanted cell types. The spectral markers identified in these models were then correlated with cellular biochemical changes related to differentiation to the cardiac phenotype. The distribution of the molecular spectral markers in individual live cells was mapped using selected Raman bands as identified by the multivariate models.

MATERIALS AND METHODS Materials and General Cell Culture All tissue culture reagents were purchased from Invitrogen (Paisley, UK) and chemicals from SigmaAldrich (Poole, UK) unless otherwise stated. Mouse embryo fibroblasts and hESC cultures were maintained at 37°C, 5% CO₂ in a humidified atmosphere. Medium was changed daily for hESC culture and every 3-4 days during differentiation. Purified chemicals used for comparison with Raman spectra of cells were purchased from SigmaAldrich: myosin heavy chain from rabbit muscle (M7659- 1 MG), Troponin T from human heart (T0175-50UG), glycogen from Mytilus edulis (G1767- 1 VL). All chemicals were used without further purification.

Human Embryonic Stem Cell Culture and Differentiation

The hESC line HUES7 was cultured in feeder-free conditions in conditioned medium on Matrigel-coated flask and cultured using trypsin passaging between passages 17- 35. Differentiation was by forced aggregation of defined numbers of hESCs. To obtain single beating CMs, beating clusters were manually dissected, washed in phosphate- buffered saline (PBS), and then incubated for 30 minutes at room temperature in buffer 1 ( 120mMNaCl, 5.4mM KC1, 5mM MgS04, 5mM sodium pyruvate, 20mM taurine, l OmM HEPES, 20mM glucose, pH 6.9), for 45 minutes at 37°C in buffer 2 (120mM NaCl, 5.4mM KC1, 5mM MgS0₄, 5mM sodium pyruvate, 20mM taurine, l OmM HEPES, 0.3 mM CaCl₂, 20mM glucose, 1 mg/ml collagenase B, pH 6.9) and for 1 hour at room temperature in buffer 3 (85mM KC1, 5mM MgSO i, 5mM sodium pyruvate, 20mM taurine, ImM EGTA, 5mM creatine, 30mM K₂ HP0₄, 20mM glucose, 1 mg/ml Na₂ ATP, pH 7.2). Finally, cell clusters were dissociated by repeated pipetting through a PI 000 tip, and the liberated cells were seeded in D-FBS in the purpose-built cell-chambers for Raman measurements.

Immuno-staining of cells

Immediately following Raman measurement, samples were fixed with 4% paraformaldehyde and permeabilized with 0.1 % Triton-Xl OO, then incubated with mouse monoclonal anti- -actinin (1 : 800; Sigma) for 1 hour at room temperature. Cell nuclei were stained with 4',6-diamidino-2-phenylindole (DAPI, 100 ng/ml) at 1 : 1000 dilution in PBS for 5 minutes at room temperature.

For glycogen staining, a Periodic Acid-Schiff (PAS) Kit was used (Sigma-395B). After the cells were stained with -actinin, the samples were rinsed with de-ionised water and oxidized in Periodic Acid Solution for 5 minutes. The samples were rinsed 3 times with de-ionised water and treated with Schiff's reagent for 15 minutes. This was followed by 10 minutes de-ionised water wash and the samples were counterstained with hematoxylin for 90 seconds. The samples were rinsed again with de-ionised water for evaluation under light microscope.

Analysis of beating frequency for individual cardiomyocytes

Beating frequency analysis was carried out on individual beating cardiomyocytes maintained on the Raman microscope under similar conditions as for Raman measurements. Videos recorded under white-light illumination were analysed using routines developed in MATLAB (The MathWorks, Natick, MA). First, all video frames were converted into grayscale and only pixels that contained the cell were selected for analysis (typically -400 pixels). To increase the accuracy of the analysis method, the contrast between the lightest and darkest regions of a cell was increased by digitising the intensity shifts at each pixel by using a threshold value. The beating frequency at each pixel of a cell was obtained by calculation the Fourier transform of the timedependent intensity shift at each individual pixel. The beating frequency of the cell was then calculated as the mean of the frequency values obtained at all pixels.

Assuming that beating is homogeneous within the cell, the digitising threshold was selected as the value which provided the smallest standard deviation between frequencies measured at different pixels inside the cell. The standard deviation was then used as a measure of the error for the beating frequency.

Raman micro-spectroscopy measurements Raman spectra were recorded using a home-built Raman micro-spectrometer optimized for live-cell studies. The instrument consisted of an inverted microscope with water-immersion objective (60x/NA 0.90) (Olympus, Essex, UK), a 785 nm -170 mW laser (before objective) (Toptica Photonics, Munich, Germany), a spectrometer equipped with a 830 lines/mm grating and cooled deep-depletion back-illuminated CCD detector (Andor Technologies, Belfast, UK) and a high-precision automated step-motor stage (Prior, Cambridge, UK). The spectrometer was calibrated prior to each experiment using a standard tylenol sample and the spectral resolution was -1.5cm^{" 1} in the 600- 1800cm^{_ 1} region. Purpose designed titanium cell-chambers were built, which incorporated MgF2 coverslips (0.17 mm thick) at the bottom to enable acquisition of Raman spectra of the cells. The Raman microscope was equipped with an environmental enclosure (Solent, Segensworth, UK), which combined with the inverted optical configuration, allowed the cells to be maintained under sterile physiological conditions (culture medium, 37°C temperature, 5% C02 ). Raman spectra of cardiomyocytes (50 cells) and non- cardiomyocytes (40 cells) were recorded during several months using more than 20 cell culture flasks, including cells at various passages.

For each cell population, both CMs and non-CMs were measured. The Raman spectrum of each individual cell represented the average of a total of 625 spectra measured at different positions inside the cell by raster-scanning the cell through the laser focus in 2 μιη steps (equivalent to a grid of 25 by 25 points). The acquisition time at each position was 1 second. After the acquisition of Raman spectra was completed, the position coordinates of each cell was recorded, the cells were fixed and prepared for immuno-staining (cardiac phenotype and cell nucleus). The cardiac phenotype of the cells was established using a wide-field fluorescence staining system integrated on the Raman microscope. The retro-positioning of each cell was achieved by using the cell coordinates (accuracy ~5 μιη), allowing identification of each individual cell into two groups, CM or non-CM. The exact phenotype of the non-CMs was found impractical to establish by immuno-staining method.

Data analysis and processing

Data pre-processing consisted of removal of spectra containing cosmic rays, background subtraction and normalization. As 625 Raman spectra were acquired for each individual cell, the small fraction of individual spectra containing cosmic rays (typically less than 1 %) were eliminated without affecting the outcome of the data analysis. The average of the Raman spectra measured at points outside of the cell (automatically identified using a Principal Component Analysis routine) represented the background spectrum (contributions from the culture medium, MgF₂ coverslip and microscope objective). The Raman spectrum representative of each cell was obtained by algebraic subtraction of the background spectrum from the average of the Raman spectra at all positions inside the cell. All Raman spectra were then normalised using the standard normal variance method.

The Raman spectra of cells were analysed by Principal Component Analysis (PCA) using functions in MATLAB. Raman spectral images corresponding to selected Raman bands were obtained by calculating the area under the spectral bands after subtraction of estimated local linear baselines and representing the integrated intensity value at each measurement position in the cell.

RESULTS

Effect of laser irradiation on cardiomyocytes: analysis of beating frequency Biomolecules are sensitive to light and exposure of cells to lasers can cause localised heating or induce toxic photochemical reactions. Although lasers in the visible range (400-600 nm) allow rapid measurement of high high-quality Raman spectra of fixed cells, such lasers can induce cell damage even at low laser powers and short exposure time.

In contrast, when cells are irradiated by lasers in the near-infrared region (700- 1000 nm), cells can remain viable even after longer exposures at much high laser power densities.

Previous reports on hESCs and hESCs-derived cardiomyocytes mentioned that, visually, laser exposure at 70mW for 5 minutes (785 nm laser) did not seem to affect the beating of the hESC-CM clusters and trypan blue viability tests indicated that cells remained viable. However, the experiments used five-fold higher laser power and five-fold longer acquisition time, as required to sample the entire volume of the cell rather than single-point measurements. Therefore, this longer laser exposure required a new evaluation of the laser effect on cells to ensure that the proposed RMS technique is non-invasive. As no evident morphological changes indicating necrosis or apoptosis were observed after laser irradiation (Fig. 1 , A and B), more subtle cellular changes were evaluated by measuring the beating frequency of individual CMs before and 15 minutes after the Raman measurements. The beating frequency before laser irradiation was found to b e in the 0.75- 1.75 Hz range (Fig. 1 C), in agreement with recent reports for cardiomyocytes. The mean standard deviation values are -24%, indicated that beating frequency can vary considerably within individual CMs. Fig. 1 C shows that the mean change in beating frequency of hESCs-derived CMs following measurement of Raman spectra was only 18%, lower than the standard deviation. Therefore, these results demonstrate that, apart from not inducing cell death, laser irradiation as required for acquisition of Raman spectra for phenotypic identification of CMs did not affect the cells to induce significant changes in their beating.

Molecular markers for phenotypic identification of cardiomyocytes Fig. 2 presents bright-field, immuno-fluorescence images and Raman spectra of typical individual live cells derived from differentiated hESCs. The immuno- staining image for cardiac phenotype (a-actinin) and cell nuclei (DAPI) for the same cells highlight the high heterogeneneity of the cell populations derived from hESCs. The percentage of CMs in the cell population was typically less than 10%, therefore the retrospective phenotypic identification using the gold-standard immuno -staining assay for each individual cell was essential. The RMS study disclosed in Chan aimed to discriminate hESC-derived CMs from undifferentiated hESCs without a gold-standard method for retrospective phenotypic identification of CMs and reported a modest accuracy of 66%.

Fig. 2B shows that the sampling method used in this study led to high signal-to-noise Raman spectra which are representative of the entire cell analysed. Several spectroscopic features can be recognized in the Raman spectra of both CMs and non- CMs, corresponding to molecular vibrations of cellular biochemicals (nucleic acids, protein, lipids and carbohydrates). To determine spectral differences, the average Raman spectra of CMs (50 cells) and non-cardiomyocytes (40 cells) derived from hESCs are showed in Fig. 3. The computed standard deviation spectra were als o included (grey lines) to show the intercellular variance within each group of cells. The computed difference spectrum (CMs minus non-CMs) is also presented in Fig. 3, including the combined standard deviation spectrum.

Fig. 3 shows that spectral differences between CMs and non-CMs can be identified in several spectral regions. The first region is 825-950cm^{_ 1} and is associated to molecular vibrations of proteins, carbohydrates and lipids. The second region is 1035- 1 170 cm^{" 1} associated mainly to vibrations of carbohydrates, lipids and nucleic acids. The third region is the highly convoluted region between 1320- 1420 cm^{" 1}, where contributions are dominated by proteins, carbohydrates and nucleic acids. A more detailed discussion on the molecular assignments of these bands is discussed below.

As the spectral differences between CMs and non-CMs include multiple Raman bands, a multivariate statistical model has been developed for discrimination of CMs. Principal component analysis (PCA) decomposes multivariate data sets in uncorrelated principal components in such a way that the first few components will capture most of the variation present in the dataset. Fig. 4A shows that out of 91 principal components (PCs) needed to identically reconstruct the original data set, only the first four PCs had a high signal-to-noise ratio and captured 86% of the variance between the Raman spectra of all cells, including CMs and non-CMs. Inspecting individual principal components and their contribution to the total variance led to the conclusion that using only the first principal component (PC I ), accounting for 65% of variance, is adequate for class discrimination purposes. It is also important to note that the main spectral bands in PCI are also found in the computed difference spectrum between average Raman spectra of CMs and non-CMs. The inclusion of PC2, PC3 or PC4 (capturing 12.36%, 7.33%) and 2.15%> of variance) did not improve the discrimination between CMs and non-CMs.

The probability distribution of the PC I scores are presented in Fig. 4B, showing a clear distinction between CMs and non-CMs. By selecting the boundary between the two classes at the intersection of the two fitted probability functions (Fig. 4B), the classification accuracy of the PCA model was 96%> sensitivity and 100% specificity. In this case, the specificity is the ability of Raman spectroscopy to identify and exclude all non-CM cells while the sensitivity parameter represents the ability to correctly identify CMs. Although PCA has been successfully used to determine the Raman spectral bands able to discriminate CMs, the values for specificity and sensitivity give no indication on the ability of these spectral markers for phenotypic identification of new cells. To determine the true accuracy for phenotypic identification of CMs, cross-validation (CV) was used to determine the sensitivity and specificity parameters for a certain target sensitivity or specificity. The different CV methods can also be labelled with the percentages corresponding to the dataset splitting, where the first figure refers to the fraction of the dataset used for building the classification model and the second figure represents the fraction of dataset on which the model is applied. In leave-one- out CV (LOOCV), all spectra except one are used to build a model and then to classify the left out spectrum. This method is repeated so that each spectrum is predicted once. The PCA model has the flexibility to allow setting the classification boundary between CMs and non-CMs to emulate a realistic scenario, where we can target highly specific regime of target 100%) specificity as required in regenerative medicine. This type of modelling is required to ensure that no unwanted cells are predicted as CM, thus leading to a lower purity cell population.

However, this regime may lead to more CMs being misclassified as non-CMs, thus decreasing the number of cells available for treatment. The cross-validation specificity and sensitivity for 70/30%, 80/20% and LOOCV calculated for target 100% specificity are showed in the following table.

For the LOOCV, all possible permutations were used to calculate the sensitivity and specificity errors, since the possible permutation was relatively small. However, for the 70/30%) and 80/20%), the number of possible permutations was high, therefore only 5000 randomly chosen combinations were used. CV results in the table show that the Raman spectral bands determined by PC I provide >97 % specificity and >96 % sensitivity. The table also shows that the values for the predicted sensitivity and specificity are independent on the cross-validation method, demonstrating the robustness of the Raman spectral markers. It is worth noting that by adjusting the classifications boundary, the prediction specificity can be further improved up to 100%), however at the expense of lower sensitivity. Imaging and assignment of the Raman spectral markers

CMs are specialized to perform specific functions and consequently express specific biochemicals to produce a large number of myofibrils. Beating CMs also require a higher amount of energy compared to other cell types; therefore they store larger amounts of glycogen. In addition, to the high classification and discrimination ability, assignment of the Raman bands identified by the PCA model can also provide insight into the chemical differences between CMs and non-CMs. Comparison between the loading spectrum of PCI and the Raman spectra of few selected biochemicals known to be more abundant in CMs (myosin, troponin, glycogen) are shown in Fig. 5. There is a striking similarity between the Raman spectrum of glycogen and PCI , suggesting that glycogen is one of the main contributors to the spectral signatures/markers that discriminate between CMs and non-CMs. Several Raman bands corresponding to glycogen can be identified: C-O-H ring vibrations at 860 cm^{" 1}, C-O-C glycosidic bonds vibrations at 938 cm^{" 1}, C-C and C-0 stretchings at 1084 cm^{" 1} and 1 123 cm^{" 1}, while C-H deformation vibrations contribute at 1340 cm^{" 1}, 1381 cm^{" 1} and 1450 cm^{" 1}. The presence of glycogen inside the cardiomyocytes has been well documented, as these cells have long been used as a valid model of glycogen metabolism.

Fig. 5 also shows that myosin and troponin also contribute to PC I , as several bands of high intensity in the Raman spectra of the purified chemicals can also be identified in the PC I loading: 853 cm^{" 1} assigned to C-C and CH3 stretching vibrations in lysine, aspartic acid, leucine, valine, and 936 cm^{" 1} associated with C-C stretching of protein backbones. One of the main features of Raman micro-spectroscopy is the ability to collect Raman spectra from micrometric regions of live cells. This feature can be used in this study to help the molecular assignment for the Raman spectral markers of CMs by observing the spatial distribution of these biomolecules within individual cells and comparing with immuno- staining images. Fig. 6 compares the Raman spectral map of the 938 cm^" ¹

Raman band for a typical CM (Fig. 6D) with the immuno-staining images for myofibrils (a-actinin, Fig. 6B) and glycogen (PAS, Fig. 6C) of the same cell. Fig. 6D indicates that the molecules contributing most to the identification of CMs are localized in the centre of this cell, at the same location where the PAS staining also shows a significantly higher concentration of glycogen. Fig. 6E shows a typical Raman spectrum measured at a position with high abundance of both glycogen and myofibrils (top spectrum, square). The Raman spectrum shows a very close similarity with PCI and pure glycogen spectra in Fig. 5. On the other hand, the immuno-staining for a-actinin (Fig. 6 B) suggests that myofibrils are distributed over the entire cells, with an accumulation at the central region. The bottom Raman spectrum in Fig. 6E (star) measured at a position rich in myofibrils but with a lower glycogen concentration, indicated a spectrum similar to myosin and troponin, with a strong band at 938 cm^{" 1} but no band at 860 cm^{" 1}. Therefore, the closer similarity between the Raman map and glycogen staining combined with the band assignment identification suggests that glycogen is the main contributor to the Raman spectral signatures/markers of hESC-derived cardiomyocytes, while proteins found in myofibrils have a lower contribution.

Fig. 6 also highlights another important point: the high intracellular chemical variability of the CMs, which led to high variations in the Raman spectra measured at various locations inside the cells. For example, although the two spectra presented in Fig. 6E correspond to two positions only ~12μιη apart inside the same cell, their spectral features are completely different. As typical cells have diameters of -50- 60μιη and laser spot is ~2μιη, it is important that the cells are raster scanned through the laser spot (2 μιη step size) and all spectra averaged in order to obtain a Raman spectrum representative to each individual cell. Therefore, measuring only single spectra at single locations inside cells by Raman micro-spectrometers with lasers focused to micrometer sizes may require shorter acquisition times but are not be appropriate for these cells due to high spectral variability.

Although this example focuses on hESCs-derived cardiomyocytes, the results demonstrate the potential of RMS for non- invasive phenotypic identification of hESCs progeny.

The spectral signatures/markers identified using these techniques can be used for separation and purification of cell populations as required for medical applications such as cell-therapy or regenerative medicine. In addition, non-invasive monitoring of time-dependent molecular changes in live hESCs during differentiation may also have huge beneficial impact on refinement and standardisation of differentiation protocols to induce the efficient differentiation of pluripotent stem cells towards desired phenotypes. Figure 10 illustrates one embodiment of apparatus according to the invention suitable for this purpose. The apparatus comprises a Raman spectroscopy system 101 for carrying out Raman spectroscopy on a sample 102 of cells. The system comprises a laser 106 that generates a laser beam having a spot size at the sample of approximately 2 microns. The laser 106 is arranged such that the laser beam can be scanned across cells of the sample in incremental steps, such as 2micron steps. Light from the spot is collected in a spectrometer 103 for detecting the generated Raman spectra. Data relating to the Raman spectra is sent from the spectrometer 103 to a processor 104. Processor 104 is arranged to process the received spectra to identify cells of a target cell type through the presence of spectral signatures/markers in the received spectra, as described in more detail below. The processor 104 may be under the control of software stored in memory 105. The processor 104 is also connected with a separator, such as an optical tweezers system 107, the system 107 responsive to signals from the processor 104 indicating whether a candidate cell is a cell of the target cell type. In use, a sample 102 of candidate cells are passed through the Raman Spectroscopy system 101 and the each cell of the sample is scanned to produce multiple spectra for each cell. These spectra are passed to the processor 104 and the processor carries out an analysis, as described below, to determine whether the spectra for each cell comprises spectral signatures/markers indicative of a target cell type. Data on the cells identified as being cells of the target cell type is fed to the separator 107 and the separator 107 separates the cells identified as being cells of the target cell type from the other cells. This process may be carried out as a batch process or as a continuous (or at least substantially continuous) process. Various spectral analysis methods can be used to for identification of cells of the target cell type within highly heterogeneous cell populations of different cell types, as those derived from hESC. Below we give examples of univariate (use only one peak) or multivariate (several peaks or entire spectrum) methods for cardiomyocytes. Univariate Analysis

The processor 104 averages the multiple spectra for each cell to obtain a single average spectrum for that cell and compares the area of the Raman peak at 854 cm^{" 1} in the average spectrum to the peak area expected for cardiomyocytes. Figure 7 shows the probability distribution functions peak area at 854 cm^{" 1} for cardiomyocytes and non-cardiomyocytes. A boundary is set at the intersection of the Gaussian fits of the probability distribution functions and cells having a Raman peak area at 854 cm^{" 1} for the average spectrum above the boundary are identified as being cardiomyocyte cells and cells having a Raman peak area at 854 cm^{" 1} for the average spectrum below the boundary are identified as being non-cardiomyocyte cells. The univariate model using this Raman peak provides 84.3% sensitivity and 100%) specificity for identification of cardiomyocytes.

This prediction accuracy can be increased if classification is carried out using the 938cm^{" 1} Raman peak. The probability distribution of this peak area for cardiomyocytes and non-cardiomyocytes is shown in Figure 8 with the boundary at the intersection shown by the dotted line. This univariate model provides 90.2% sensitivity and 100%) specificity for identification of cardiomyocytes. Multivariate Analysis

In multivariate analysis methods, multiple Raman peaks are used at the same time for classification, therefore the accuracy is usually superior to univariate methods. In this embodiment, Principal Component Analysis is used. However, it will be understood that other suitable variable reduction methods could be used.

Figure 4A shows the first 4 loading spectra responsible for 64.78%), 12.36%), 7.33% and 2.15%) of variance between Raman spectra of all cells. The loading spectra contain peaks corresponding to the main biochemicals as identified in the difference spectra in Figure 3. However, as discussed above, it is only necessary to consider PCI in order to achieve sufficient discrimination between cardiomyocytes and non-cardiomyocytes. Therefore, only the scores corresponding to PC I are used for phenotypic classification. The probability distribution of the scores corresponding to first principal component PC I is shown in Figure 4B. A boundary is set at the intersection of the Gaussian fits of the probability distribution functions for the scores and cells having a PC I score above the boundary are identified as being cardiomyocyte cells and cells having a PC I score below the boundary are identified as being non-cardiomyocyte cells. The classification accuracy is 96%> sensitivity and 100% specificity.

It will be understood that various modifications and improvements can be made to the above described embodiments without departing from the scope of the claims as appended hereto.

Claims

1. A method of identifying a spectral signature of a target cell type that can be used to distinguish the cells of that cell type from other cell types in a mixture comprising obtaining a target cell spectrum for each of a plurality of cells of the target cell type using Raman Spectroscopy, the target cell spectrum resulting from exciting a substantial portion of the cell of the target cell type with a laser, obtaining a non- target cell spectrum for each of a plurality of cells of one or more other cell types using Raman Spectroscopy, the non-target spectrum resulting from exciting a substantial portion of the cell of the other cell type with a laser, and comparing the target cell spectra to the non-target cell spectra to identify a spectral signature characteristic of target cell spectra that distinguishes cells of the target type from cells of the one or more other types.

2. A method according to claim 1 , wherein the target spectrum and/or non-target spectrum is obtained by exciting a substantial portion of the cell with a laser having a laser spot with an area of the order of the footprint of the cell.

3. A method according to claim 1 , comprising generating multiple target and/or non-target spectra for each cell, each spectrum associated with a different portion of the cell.

4. A method according to claim 3, wherein the multiple spectra are obtained by scanning a laser across the cell.

5. A method according to claim 4, wherein the multiple spectra are processed using principal component analysis to determine principal components and a threshold score for one or more of the principal components that distinguish cells of the target type from cells of the one or more other types.

6. A method according to any one of the preceding claims, wherein the method comprises determining a spectral difference between the target spectra and the non- target spectra and identifying the spectral signature from the spectral difference.

7. A method according to any one of the preceding claims, wherein the laser used for obtaining the spectra may emit light in the near-infra-red.

8. A method according to any one of the proceeding claims comprising checking that the plurality of target cells are cells of the target cell type.

9. A method according to claim 8, wherein the checking step is carried out using immuno-fluorescence techniques.

10. A method according to any one of the preceding claims carried out for a range of target cells, the method further comprise compiling a database of spectral signatures for each cell type.

1 1. A data carrier having stored thereon instructions, which, when executed by a processor, cause the processor to carry out the method according to claim 1.

12. A data carrier having stored thereon a database of spectral signatures determined in accordance with claim 1.

13. Apparatus comprising a processor, the processor arranged to carry out the method of claim 1.

14. Apparatus according to claim 13, further comprising a device for carrying out Raman spectroscopy arranged to automatically feed the detected spectrum to the processor.

15. A method of identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells comprising:- obtaining a spectrum for a candidate cell generated using Raman Spectroscopy, and

determining whether the candidate cell is a target cell by determining whether the spectrum has a spectral signature characteristic of a target cell, wherein the spectral signature has been determined by the method of any one of claims 1 to 10.

16. A method according to claim 15, wherein the spectrum is obtained by exciting a substantial portion of the candidate cell with a laser having a laser spot with an area of the order of the footprint of the candidate cell.

17. A method according to claim 15, comprising generating multiple spectra for the candidate cell, each spectrum associated with a different portion of the candidate cell.

18. A method according to claim 3, wherein the multiple spectra are obtained by scanning a laser across the candidate cell.

19. A method according to any one of claims 15 to 18 comprising the additional step of separating cells identified as cells of the target cell type from cells identified as cells of other types.

20. A method according to any one of claims 15 to 19, wherein the spectral signature characteristic of the target cell is a score for a first principal component of a principal component analysis falling within an expected range of values.

21. Apparatus for identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells comprising:- memory having stored therein a spectral signature characteristic of a target cell, wherein the spectral signature has been determined by the method of any one of claims 1 to 10;

a processor arranged to obtain a spectrum for a candidate cell generated using

Raman Spectroscopy, and

determine whether the candidate cell is a target cell by determining whether the spectrum has the spectral signature characteristic of the target cell type.

22. Apparatus according to claim 21 , further comprising a separator for separating cells identified as cells of the target cell type from cell of other types.

23. A method of identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells comprising:- obtaining a multiple spectra for a candidate cell generated using Raman Spectroscopy, each spectra associated with a different portion of the cell, and

24. A method of identifying cardiomyocyte cells form other cells in a mixture of candidate cells comprising:- obtaining a spectrum for a candidate cell using Raman Spectroscopy, determining whether a value of the spectrum at one or more wave-numbers selected from the group consisting of:

vii) 825 to 950 cm ¹

viii) 1035 to 1 170 cm ¹; and

ix) 1320 to 1420 cm ¹

is within a range of expected values for cardiomyocyte cells, and

25. A method according to claim 23, wherein the value is a Raman peak area for a peak centred at these wavenumbers.

26. A method according to claim 23, wherein the value is the Raman peak area for a peak centred at 854 cm^{" 1}.

27. A method according to claim 23, wherein the value is the Raman peak area for a peak centred at 938 cm^{" 1}.

28. A method of identifying cardiomyocyte cells from other cells in a mixture of candidate cells comprising:- obtaining a spectrum for a candidate cell using Raman Spectroscopy, determining from characteristics of the spectrum a score for a first principle component of a principle component analysis carried out on spectra of cardiomyocyte cells,

determining whether the score is within a range of expected values for cardiomyocyte cells, and identifying the candidate cell as a cell of the target cell type if the score is determined as being within the range of expected values.

29. A method according to claim 27, comprising determining from characteristics of the spectrum a score for only the first principle component of the principle component analysis carried out on spectra of cardiomyocyte cells.

30. A method of identifying a spectral signature of a target cell type that can be used to distinguish the cells of that cell type from other cell types in a mixture substantially as described herein with reference to the accompanying drawings.

31. Apparatus for identifying a characteristic of a target cell type that can be used to distinguish the cells of that cell type from other cell types in a mixture substantially as described herein with reference to the accompanying drawings.

32. A data carrier having stored thereon instructions, which, when executed by a processor, cause the processor to carry out the method substantially as described herein with reference to the accompanying drawings.

33. A method of identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells substantially as described herein with reference to the accompanying drawings.

34. Apparatus for of identifying cells of a target cell type from cells of other cell types in a mixture of candidate cells substantially as described herein with reference to the accompanying drawings.

35. A method of identifying cardiomyocyte cells from other cells in a mixture of candidate cells substantially as described herein with reference to the accompanying drawings.