Introduction

Photosynthetic microorganisms including bacteria and algae are essential to aquatic life and the global carbon budget, but the precise ecological roles of many species remain ill defined because of cultivation difficulties which have led scientists to develop methods for detecting uncultured cells in the environment (Amann et al., 1995; Venter et al., 2004; Daniel, 2005). However, some uncultured photosynthetic bacteria and microalgae have been found to play crucial roles in carbon or nitrogen cycling (Weller et al., 1992; Beja et al., 2001; Karl, 2002; Crump et al., 2007; Stepanauskas and Sieracki, 2007; Zehr et al., 2008; Tripp et al., 2010). Recently, culture-independent approaches (for example, in situ hybridisation and 18S rRNA sequencing) have been employed to reveal two new groups of photosynthetic eukaryotes that play key roles in CO2 fixation in the subtropical and tropical northeast Atlantic Ocean. These groups, Euk-A and Euk-B, have not yet been cultivated and, in the case of Euk-B, culture conditions for closely related species have never been defined (Jardillier et al., 2010). It is reasonable to hypothesise that significant populations of uncultivated photosynthetic bacteria and algae remain to be discovered in the oceans. It is essential to monitor the photosynthetic activity of bacteria and algae including those hitherto-uncultivated microorganisms, and to understand their dynamic processes in relation to CO2 fixation and climate change. In addition, uncultured oceanic photosynthetic bacteria and algae represent an untapped biotechnological resource that could lead to discovery of new biomedical drugs and biocatalysts (Bode and Muller, 2005; Fortman and Sherman, 2005; Lorenz and Eck, 2005). Therefore, it would be useful to develop a culture-independent tool to unravel both the ecological function of photosynthetic bacteria and algae in situ, and to reveal their roles in the natural environment in terms of their relative contribution to carbon fixation.

Raman microspectroscopy has been shown to be a useful tool to study microbial cells (Huang et al., 2004, 2007a, 2007b, 2009a, 2009b, 2010; Wagner, 2009; Haider et al., 2010; Hall et al., 2011; Wu et al., 2011). In combination with other biological tools such as stable-isotope probing (SIP) and fluorescence in situ hybridisation, Raman single-cell technology can link bacterial species with their ecological roles without cultivation (Huang et al., 2009a). Raman spectroscopy detects the vibrations of chemical bonds of molecules through the inelastic scattering of incident laser light (Huang et al., 2010). Raman microspectroscopy is able to study single microorganisms and provide intrinsic cellular information (Huang et al., 2004). A typical Raman spectrum contains a rich cellular chemical profile, including information on nucleic acids, protein, carbohydrates, lipids and pigments (for example, carotenoids), which enables the characterisation of different cell types, physiological and phenotypic changes to living cells (Huang et al., 2004, 2007a, 2007b, 2009a, 2010; Wagner, 2009; Haider et al., 2010; Hall et al., 2011). Raman imaging is a hyperspectral technique that can be used to acquire a full Raman spectrum at many hundreds or thousands of positions across a sample (for example, microorganisms on a microscopic slide). The resulting spectra can be interrogated to generate pseudocolour images based on specific peaks of the Raman spectra. As these features relate to chemical entities within the sample, the Raman images can be used to illustrate the distribution of chemical components. Raman imaging would be desirable to ‘visualise’ chemical information of individual cells in a microbial community as it provides single-cell spatial resolution and a temporal dimension to study microbial identity and ecophysiology in complex microbial communities (Huang et al., 2009a). However, the physical process of light scattering that underlies Raman spectroscopy is a rare event, with just one in 106–108 photons incident on the sample undergoing Raman scattering. Typically, the acquisition of a good-quality Raman spectrum from a single cell takes from several minutes to 4 s, making Raman imaging time consuming and hindering high-throughput screening and rapid imaging. Resonance Raman (RR) spectroscopy can enhance the Raman signal up to 106-fold, which significantly reduces the acquisition time of Raman spectra (Robert, 2009). RR occurs when the frequency of incident light matches the electronic transition of the irradiated molecule. Many biological molecules are RR-active compounds and carotenoids are typical analytes for RR spectroscopy (Krebs et al., 2003; Robert 2009). More importantly, nearly all photosynthetic microorganisms contain carotenoids, which are essential elements of light-harvesting complexes and singlet oxygen quenchers (Szalontai et al., 1994; Garcia-Asua et al., 1998; Tracewell et al., 2001, 2005).

In this study, we obtained single-cell RR (SCRR) spectra from carotenoid-containing bacteria and microalgae using short, milliseconds, acquisition times. SIP using 13C bicarbonate was employed to link bacterial species with their CO2 fixation functions. The red shift of SCRR spectra (where Raman bands move to lower Raman shift or wave number position because of 13C labelling on the molecules; Huang et al., 2004) was observed in photosynthetic microorganisms that had incorporated 13C bicarbonate into the cells. The results suggested that carotenoids were good biomarkers to indicate 13CO2 incorporation in photosynthetic microorganisms with a sensitivity as low as 10% 13C content of single cells. Finally, we applied SCRR–SIP to generate Raman images that help in situ and quantitative identification of active 13CO2-fixing microorganisms in culture mixtures and real seawater samples. This shows that SCRR–SIP may have great potential as a rapid in situ technique for imaging and screening of photosynthetic cells in natural microbial communities.

Materials and methods

Chemicals, media, strains and growth conditions

All chemicals used in this study were purchased from Sigma-Aldrich (Dorset, UK) unless otherwise stated. The 13C sodium bicarbonate contains 98 atom % 13C (catalogue number 372382). The microalgae strain AMA was isolated from Arctic supraglacial sites. Synechocystis sp. PCC 6803 and Synechococcus elongatus PCC 7942 were from our laboratory collection. Non-photosynthetic bacterial strains GWO, GWY and GWR with orange, yellow and red pigments, respectively, were isolated from contaminated groundwater in southwest United Kingdom (Huang et al., 2009a). AMA, Synechocystis sp. PCC 6803 and S. elongatus PCC 7942 strains were grown in BG-11 media (Sigma-Aldrich) with various proportions of 12C and 13C sodium bicarbonate as the sole carbon source. To examine the integration of 13C in photosynthetic microorganisms, Synechocystis sp. PCC 6803 and S. elongatus PCC 7942 strains were grown in liquid BG-11 medium supplemented with different percentages of 13C bicarbonate. The final concentration of the total carbon source (12C and 13C bicarbonate) was 5 mM. Different 13C contents were achieved by mixing 1 M filter-sterilised 12C and 13C sodium bicarbonate solutions in appropriate ratios before adding to the BG-11 media. Bacterial strains GWO, GWY and GWR were grown and maintained in R2A agar medium (Oxoid, UK).

Seawater was sampled from the coast at the Dove Marine Laboratory, Cullercoats, England. Photosynthetic organisms in the seawater samples were maintained in glass tanks under open air conditions near (<50 m) to the sampling point for several months. Original sea water was used to maintain the resulting natural photosynthetic microorganisms.

Microcosm setup for cyanobacterial cultures

To eliminate CO2 from the culture vessel, a degasification step was done before the incubation in order to remove dissolved CO2 in the BG-11 medium. This was achieved by boiling the BG-11 medium in a microwave oven for 1 min. Deionised water (10 ml) was added to the BG-11 medium to compensate for the loss of water during the degasification. A 0.45-μm filter attached to a syringe filled with sodium hydroxide was linked to the bottle cap immediately after the degasification and the bottle cap was fastened (Supplementary Figure S1), and the medium was allowed to cool to room temperature. Pre-mixed 12C and 13C sodium bicarbonate solution and 50 μl of cyanobacterial cell suspension (108 per ml) were then added to the BG-11 medium by quickly opening the cap and re-sealing the bottle with a closed cap. Synechocystis sp. PCC 6803 was grown in media with 13C contents of 1.1% (13C natural abundance), 10.8%, 25.3%, 49.6%, 73.8% and 98%. S. elongatus PCC 7942 was grown in media supplemented with 13C contents of 1.1% (natural abundance), 49.6% and 98%. These 13C contents are corrected by taking account of 13C natural abundance and the 98% 13C bicarbonate used in the study. The culture vessel was incubated at 30 °C and 50 μmol photons m−2 s−1 in an orbital shaker and agitated at 150 r.p.m. for 4 days (Innova 44 illuminated rotary incubator, New Brunswick Scientific, Cambridge, UK).

Incubation of seawater samples

A total of 30 ml of the sea water was sampled and 13C-NaHCO3 was added into the seawater samples with a final concentration of 2 mM. The seawater samples were incubated in closed bottles at room temperature (20 °C) for 8 days with natural light. Control seawater samples were incubated with 2 mM 12C-NaHCO3 under the same conditions.

Raman microspectroscopy analysis

Cells were sampled from microcosms and were washed with deionised water before analysing them with the Raman microscope. Each cellular suspension (5–10 μl) was spread on a calcium fluoride (CaF2) slide and allowed to air dry before Raman analysis. The SCRR spectra were acquired using a confocal Raman microscope (LabRAM HR, HORIBA Scientific, London, UK) equipped with an integrated Olympus microscope (model BX41). A × 100 magnifying dry objective (NA=0.90, Olympus, Essex, UK) was used to observe and acquire Raman signals from single cells. The laser beam was targeted on the cell visually using an integrated colour camera and a motorised XYZ stage (0.1 μm step). The Raman scattering was excited with a 532-nm Nd:YAG laser (Torus Laser, Laser Quantum, Manchester, UK). The laser power on a single cell was ∼3.5 mW. The detector was a −70 °C air-cooled charge-coupled device detector (Andor, Belfast, UK). The system was run with a confocal pinhole diameter of 100 μm, enabling a spatial resolution of ∼1 μm to be obtained. Each Raman spectrum was acquired between the range 2172 and 557 cm−1, with 1021 data points and resolution of 1−1.5 cm−1. LabSpec software (HORIBA Scientific) was used to control the Raman system and acquire Raman spectra. Acquisition times for SCRR spectra were 100 ms for single-cell measurements and 1−30 ms for Raman imaging. To each pure or mixed culture samples, 20 single cells were randomly chosen and measured. Pure β-carotene (Sigma Aldrich) was used as a carotenoid control and was measured under the same condition as SCRR.

Raman imaging

SWIFT (HORIBA Scientific) imaging function was coupled into the Raman systems, which allows significant reduction in the time needed for generating Raman images when very short spectral acquisition time is used (milliseconds in this study). In contrast to traditional start–stop mapping methods where the minimum measurement time is typically 500 ms per point, samples are continuously scanned beneath the laser beam, and spectra are acquired ‘on the fly’ as the sample is scanned in the SWIFT mode. SWIFT was employed to generate Raman images of photosynthetic bacterial populations in which the cells had different 13C contents. Three half–half mixed cyanobacterial cultures were imaged by SWIFT: (1) Synechocystis sp. PCC 6803 cells grown in 73.8% 13C bicarbonate BG-11 medium mixed with cells grown in 12C bicarbonate BG-11 medium; (2) Synechocystis sp. PCC 6803 cells grown in 10.8% 13C bicarbonate BG-11 medium mixed with cells grown in 12C bicarbonate BG-11 medium; and (3) Synechocystis sp. PCC 6803 cells grown in 10.8% 13C bicarbonate BG-11 medium mixed with S. elongatus PCC 7942 grown in 12C bicarbonate BG-11 medium. SWIFT was also used to image the seawater samples incubated with 13C-NaHCO3 and 12C-NaHCO3 to visualise the 13C-incorporated photosynthetic cells. Each Raman images was acquired within minutes and the acquisition time of each pixel was 1 ms for mixed cultures, and 10–30 ms for the seawater samples. A diffraction limited spatial resolution of ∼0.7 μm and a mechanical step of 0.5 μm (4 million pixels mm−2) were used to generate the Raman images for cyanobacteria culture mixtures; a mechanism step of 1 μm was used for the seawater sample. Raman images were analysed with LabSpec (HORIBA Scientific), and were generated by classical least squares modelling of hyperspectral data sets. At each image pixel position a linear combination of factors was calculated to estimate the measured spectrum, such that

where Ii is the ith measured spectrum of the data set, Ak is the kth loading input to the calculation, aij is the score of the spectrum Ii with respect to the loading Ak, aijAk is a factor and ɛi is the error between the spectrum (measurement) Ii and the estimation

Loading spectra (Ak) were selected by averaging spectra from manually selected regions of interest within the data set. The images display scores (aik) as a function of x axis and y axis positions. The green or yellow colour in the Raman images indicated the 12C carotenoids in cells, whereas the red colour indicated the 13C carotenoids in cells (characterised by the red shifted RR peaks). In the Raman images of seawater samples, the purple colour was defined as the fluorescence because some organisms with autofluorescence were observed, which appeared as a broad, featureless spectrum without the expected sharper RR peaks.

Raman spectra analysis

All Raman spectra were recorded, smoothed and baseline corrected using LabSpec. Spectra smoothing was performed by the Savitsky–Golay smoothing method. Positions of the SCRR bands were determined using LabSpec and were subsequently imported to SPSS 15 (IBM Corporation, New York, NY, USA) for further analysis. The positions of SCRR Raman bands of Synechocystis sp. PCC 6803 and S. elongatus PCC 7942 were analysed statistically to establish a relationship between the extent of red shift and 13C content in cells. The wave numbers of the Raman bands were plotted against the 13C contents, with the error bars and linear regression lines (SPSS 15). Two-sample t-tests were performed in SPSS between the 12C and 10.8 % 13C PCC6803 samples on three major bands of SCRR. MVSP 3.13g (Kovach Computing Services, Pantraeth, UK) was used to perform principal component analysis for SCRR spectra.

Absorbance analysis of Synechocystis sp. PCC 6803

A total of 5 ml of Synechocystis sp. PCC 6803 was harvested by centrifugation (Eppendorf, Cambridge, UK) at 3000 r.p.m. for 5 min. The cells were then re-suspended in the same volume of phosphate buffer solution. To obtain a light absorbance spectrum of Synechocystis sp. PCC 6803, 200 μl of cells was scanned in the visible range using a microplate reader (Synergy 2, BioTek, Potton, UK). The absorbance spectrum was recorded in triplicate with a spectral resolution of 5 nm in the range of 400 nm to 750 nm.

Mass spectrometry (MS) analysis of 13C content of Synechocystis sp. PCC 6803 proteins

Batches of Synechocystis sp. PCC 6803 cells were grown in BG-11 media with different percentages of 13C-NaHCO3 (1.1%, 10.8%, 49.6% and 98%) as the sole carbon source. The growth conditions were the same as described above. After 4 days of growth, 5 ml of each culture were sampled for single-cell Raman spectroscopy analysis. Meanwhile, 10 ml of each culture were taken and the cells were harvested by centrifugation at 4000 r.p.m. for 10 min. Cells from each Synechocystis sp. PCC 6803 sample were re-suspended in 0.2 ml of SDS sample buffer containing tracker dye and 0.05% (v/v) mercaptoethanol (Sigma, Poole, UK). After incubation at 90 °C for 10 min, 15 μl was applied to a NuPAGE Bis-Tris gel (1.0 mm, 12%; Invitrogen, Paisley, UK) according to the manufacturer's protocol. The gel was stained with colloidal Coomassie Blue (Bio-Rad Laboratories, Hemel Hempstead, UK) and the same prominent protein band (at 50–75 kDa) excised from the four lanes (Supplementary Figure S2). The bands were subjected to in-gel tryptic digestion according to Pandey et al. (2000).

The extracted peptides were separated on an UltiMate 3000 nanoflow liquid chromatography system (Dionex, Camberley, UK) using 5 mm × 300 μm (trapping) and 150 mm × 75 μm (analytical) PepMap C18 reverse-phase columns (Dionex). Linear gradient elution was from 97% solvent A (0.1% formic acid in 3% acetonitrile) to 35% solvent B (0.1% formic acid in 97% acetonitrile) at a flow rate of 300 nl min–1 for 40 min. On-line MS was done on a Maxis UHR-TOF (Bruker Daltonics, Coventry, UK) using profile MS with automatic dependent tandem mass spectrometry (MS/MS) data acquisition.

The nano liquid chromatography–MS/MS data file corresponding to the incorporation experiment at natural isotopic abundance was used as input for database searching (Mascot software supplied by Matrix Science, London, UK) to identify the protein band and its component tryptic peptides. The 13C incorporation ratios in the samples were determined by comparison of the observed isotope distribution pattern of the representative peptide with the theoretical isotope distribution pattern generated using IsoPro v3.1 (http://sites.google.com/site/isoproms).

Results

Cells containing carotenoids display strong RR effects

The absorption spectrum of Synechocystis sp. PCC 6803 in the visible range showed peaks at 440, 620 and 670 nm (Supplementary Figure S3). Carotenoids absorb light in the 400–550 nm range (Gust et al., 1993), and the 532 nm laser for Raman excitation used in this study preferentially excites the carotenoids rather than the chlorophyll pigments.

RR spectra of carotenoid molecules display characteristic Raman bands: v1, v2 and v3 (Figure 1, β-carotene spectrum). The sharp and strong v1 (stretching mode of the C=C bonds at ∼1515 cm−1), v2 (stretching mode of the C-C bonds at ∼1155 cm−1) and v3 (deformation of the methyl groups at ∼1003 cm−1) peaks can be unambiguously assigned to carotenoids (Rimai et al., 1973; Tracewell et al., 2001; Robert, 2009). The position of the v1 band varies according to the length of the conjugated C=C bond chain (Rimai et al., 1973; Carey, 1978; Tracewell et al., 2001; Robert, 2009). The v2 band varies depending on the molecular conformation, and the v3 band is related to carotenoid methyl groups (Rimai et al., 1973; Carey, 1978; Tracewell et al., 2001; Robert, 2009). The RR signals of carotenoids are so strong that with 532 nm laser excitation the SCRR spectra of carotenoid-containing microorganisms predominantly display the v1, v2 and v3 Raman bands, which in Figure 1 arise from the typical carotenoid β-carotene. The positions of the v1 band in particular varied significantly among the various microorganisms studied, with the v2 and v3 bands also affected to a lesser extent (Figure 1). We noted a variation of 17.2, 1.7 and 4.9 cm−1 at the v1, v2 and v3 positions, respectively, which arose from the different conformations and composition of the carotenoid molecules within the cells.

Figure 1
figure 1

SCRR spectra of Synechocystis sp. PCC 6803, microalgae and bacteria with carotenoid pigments. The dominant Raman bands of SCRR spectra were from carotenoids, as shown by comparison with the spectrum of pure β-carotene.

Full size image

The red shift of SCRR spectra due to 13C incorporation in cells

Figure 2 shows the SCRR spectra of S. elongatus PCC 7942 and Synechocystis sp. PCC 6803 grown in media supplemented with different proportions of 13C bicarbonate. The RR bands of both species displayed a red shift, the extent of which was dependent on the level of 13C bicarbonate in the growth medium. For S. elongatus PCC 7942, the ν1, ν2 and ν3 bands of 12C cells shifted from 1522.0, 1157.7 and 1006.0 cm−1 to 1484.9, 1128.2 and 989.5 cm−1, respectively, in 13C-labelled cells (Figure 2). For Synechocystis sp. PCC 6803, the v1, ν2 and ν3 bands of 12C cells shifted from 1516.8, 1155.6 and 1004.3 cm−1 to 1480.9, 1125.3 and 988.6 cm−1, respectively, in 13C-labelled cells (Figure 2). Among the three dominant SCRR bands, the v1 band had the greatest (∼36 cm−1) and ν3 band least (∼16 cm−1) red shift; and the ν2 band had an intermediate ∼30 cm−1 red shift (Table 1) in the range of 1.1–98% 13C cells. This experiment encompasses the full range of labelling possible, considering that the natural abundance of 13C is 1.1% and the 13C content in Sigma-Aldrich 13C sodium bicarbonate is 98%.

Figure 2
figure 2

RR spectra of S. elongatus PCC 7942 and Synechocystis sp. PCC 6803 grown in media containing different percentages of 13C-NaHCO3.

Full size image
Table 1 Raman shift due to 13C incorporation (cm−1)
Full size table

Figure 3 shows that the extent of the red shift of the ν1, ν2 and ν3 Raman bands was proportional to the 13C content of the cells, suggesting that the spectral red shift can be used as a quantitative biomarker for 13C incorporation. Linear regression of data obtained from 20 randomly chosen single cells from three biological replicates (Figure 3) demonstrate that the three Raman bands shift in a linear manner (linear regression R2>0.9). It is notable that the red shifts of the Synechocystis sp. PCC 6803 bands show distinguishable differences between the 12C (containing natural abundance 1.1% 13C) and 10.8% 13C cells (Figure 3). The t-tests of the ν1, ν2 and ν3 bands between the 12C and 10.8% 13C samples indicate that these differences were significant (P-values <0.001). The slopes of the ν1, ν2 and ν3 regression lines of S. elongatus PCC 7942 and Synechocystis sp. PCC 6803 are slightly different (Figure 3). However, for each cell type, the ν1, ν2 and ν3 bands show a stepwise shift to lower energy wave numbers with increasing 13C content in cells (Figure 2 and Supplementary Figure S4), which makes it possible to establish a calibration to estimate the 13C content of a single cell based on its RR spectrum (Figure 3). The principal component analysis was carried out using the whole SCRR spectra and the results show the distinct separations among samples with different 13C contents (Supplementary Figure S5). The largest principal component analysis loadings at axes 1 and 2 were associated with the ν1 and ν2 bands, indicating that these spectral features are the most important factors for the classification of the SCRR spectra into different groups, as shown in Supplementary Figure S5.

Figure 3
figure 3

Linear regression of the ‘red shift’ of the ν1 (a), ν2 (b) and ν3 (c) bands of the 13C-incorporated Synechocystis sp. PCC 6803 single cells and the ν1 (d), ν2 (e) and ν3 (f) bands of the 13C-incorporated S. elongatus PCC 7942 single cells. On each sample, 20 replicates were performed.

Full size image

MS analysis of the 13C incorporation in cells matches SCRR analysis

MS was applied to analyse 13C incorporation in Synechocystis sp. PCC 6803. MS analysis of extracted proteins was used to determine the levels of 13C incorporation in cells. The MS spectra of a representative peptide GGTQGIVGTTAANLNR [M+2H]++ identified in protein Slr1841 of Synechocystis sp. PCC 6803 grown in 1.1%, 10.8%, 49.6% and 98% 13C is shown in Figure 4. The mass shift observed in the MS spectra is due to 13C incorporation in the peptide (Figure 4). The percentage of 13C incorporation was determined by comparison of the observed and theoretical isotope distribution pattern. The results show that the 13C incorporation in the peptide was 1.1%, 13%, 54% and 96%, respectively (Figure 4 and Supplementary Figure S6). The SCRR spectral analysis of 20 single cells from each sample indicated that the corresponding samples contained, respectively, 1.1±0.4%, 9.4±0.5%, 51±0.7% and 95±0.6% of 13C incorporated in the cells of Synechocystis sp. PCC 6803 according to the calibration curves established in Figure 3. The results verify that the 13C contents of single cells measured by SCRR spectroscopy do indeed reflect the 13C-incorporation percentages in the whole cells measured by mass spectrometry.

Figure 4
figure 4

Mass spectra of peptide GGTQGIVGTTAANLNR in protein Slr1841 of Synechocystis sp. PCC 6803 with different 13C incorporation ratios. The mass-charge ratios of the modal ions are labelled.

Full size image

Raman images differentiate cells with varying levels of 13C incorporation

To demonstrate the potential of RR spectroscopy as an in situ ‘biochemical’ imaging technique for photosynthetic microorganisms, two-dimensional Raman imaging was carried out on the three different cyanobacterial cell mixtures described in the Materials and Methods section (Figure 5). The spectra shown in Figure 5 are SCRR spectra taken at the centre of the circled cells with a 1-ms acquisition time. The optical images matched well with the pseudocolour Raman images, as the green colour and the red/yellow colour indicate the locations of 12C- and 13C-incorporated single cells, respectively (Figure 5). The intensity of pseudocolour is proportional to the intensity of the Raman bands. Figure 5a shows a clear difference between 12C- and 73.8% 13C-incorporated cells. The SCRR-based Raman images were so sensitive that they were able to distinguish a subtle difference in the cells with 1.1% and 10.8% 13C incorporation (Figure 5b and c). As shown in Figure 5c, the Raman image of the cells based on RR spectra is consistent with the known morphology of rod-like S. elongatus PCC 7942 cells and spherical Synechocystis sp. PCC 6803 cells.

Figure 5
figure 5

Raman images of mixed cell cultures with different 13C contents: (a) Synechocystis sp. PCC 6803 grown in BG11 medium supplemented with 73.8% 13C-NaHCO3 mixed with the same species grown in BG11 medium supplemented with 12C-NaHCO3. (b) Synechocystis sp. PCC 6803 grown in BG11 medium supplemented with 10.8% 13C-NaHCO3 mixed with the same species grown in BG11 medium supplemented with 12C-NaHCO3. (c) Synechocystis sp. PCC 6803 grown in BG11 medium supplemented with 10.8% 13C-NaHCO3 mixed with S. elongatus PCC 7942 grown in BG11 medium supplemented with 12C-NaHCO3. Two Raman images were generated based on the ν1 and ν2 bands. The 13C-incorporated single cells are displayed in red and the 12C cells green. Some 13C-incorporated cells with lower signal strength are displayed in yellow. The SCRR spectra are coloured to correspond to the 13C-labelled cells and 12C cells indicated by red and green circles, respectively.

Full size image

Figure 6 shows that Raman images can be used to distinguish 13C-labelled microorganisms at the single cell level within a complex microbial community in real seawater samples. The sample incubated with 12C-NaHCO3 provided a baseline for the SCRR spectra used to estimate 13C incorporation of diverse photosynthetic microorganisms (Figure 6a). The Raman images in Figures 6b and c suggest that after 8 days of incubation with 13C-NaHCO3, some species have started to incorporate 13C into their cell components. The Raman imaging results show that cells, tentatively identified as the microalgal Tetraselmis sp. (cells 1b and 2b in Figure 6b and cell 2c in Figure 6c) and the diatom Phaeodactylum tricornutum (cell 4c, a tip of this triradiate-shaped microorganism in Figure 6c) were actively fixing 13C-NaHCO3 whereas one possible photosynthetic species (cell 3b in Figure 6b) was present but incorporated little 13C-NaHCO3 into the cells. Tetraselmis sp. and Phaeodactylum tricornutum in the samples were identified by Zhang (2011). Figure 6c shows that at least four different species demonstrated 13C incorporation: Tetraselmis sp. Phaeodactylum tricornutum and unidentified rod-shaped cells 1c and 3c, which show a red shift of RR bands in comparison with control cells grown in sea water with 12C-NaHCO3. Based on the red shift of the RR bands as shown in Figure 6b (cells 1 and 2) and Figure 6c (cell 2), Tetraselmis sp. is estimated to incorporate ∼20% 13C into cellular compounds after 8 days of incubation.

Figure 6
figure 6figure 6

Raman images of the seawater sample that was grown in original sea water supplemented with 2 mM 12C-NaHCO3 (a) and 2 mM 13C-NaHCO3 (b, c). Two Raman images were generated based on the ν1 and ν2 bands. The 13C-incorporated single cells are displayed in red. The purple colour indicates the fluorescent region of the cells. SCRR spectra corresponding to the marked cells are shown.

Full size image

Discussion

Raman spectra from single cells can be used as intrinsic ‘fingerprint’ without invasive labelling (Huang et al., 2004, 2007a, 2007b, 2009a, 2010; Wagner, 2009; Haider et al., 2010; Hall et al., 2011). However, the broad application of Raman spectroscopy in environmental microbiology is hindered by the naturally weak Raman effect. This study has demonstrated that by using carotenoids as internal biomarkers, SCRR combined with SIP can distinguish and image CO2-fixing cells on a millisecond timescale. This method is also suitable for in situ identification and imaging of CO2-fixing photosynthetic microorganisms in a quantitative manner (Figures 4 and 5).

As a 532 nm laser was chosen to record SCRR spectra from carotenoid-containing cells, the Raman acquisition time for a single cell was reduced to the millisecond level. This rapid Raman detection method paves the way for the development of a Raman cell sorter that could be a vital tool for metabolic analysis of high numbers of environmentally derived uncultured cells.

Cyanobacteria are the largest, most diverse and widely distributed group of photosynthetic prokaryotes (Stanier and Cohenbazire, 1977). In this study, S. elongatus PCC 7942 and Synechocystis sp. PCC 6803 were used as model photosynthetic microorganisms to test the SCRR–SIP methodology and using carbon dioxide fixation as an example of the study of an important eco-physiological property. SCRR–SIP was applied to identify microorganisms actively fixing CO2 in real seawater samples (Figure 6), confirming that this technique could be used to identify and isolate key photosynthetic microorganisms that carry out CO2 fixation in aquatic environments.

In order to establish a reliable quantitative measurement of the 13C content of photosynthetic microorganisms based on the red shift of SCRR, it was essential to grow cyanobacterial strains in conditions that minimise the effect of atmospheric CO2. Preliminary data showed that Synechocystis sp. PCC 6803 appears to prefer CO2 dissolved from the air rather than 13C bicarbonate in the growth medium. To eliminate any potential influence of CO2 from atmosphere, the procedure described in this study was used.

SIP is a method widely used to study carbon and nitrogen flow in ecosystems. Introducing stable isotope compounds into an environmental sample allows researchers to determine which organisms incorporate the substrates of interest. Both mass spectra and vibrational spectra can be used in combination with SIP for work on single cells. In terms of sensitivity, spatial resolution and quantitative detection, mass spectra-based methods such as secondary ion mass spectrometry are superior over those based on vibrational spectra such as Raman spectroscopy. However, SCRR is a noninvasive method and provides molecular profiles or ‘fingerprints’ of cells, whereas secondary ion mass spectrometry is an invasive technology and detects atoms and small ions but not molecules (Wagner, 2009). Nucleic acid-SIP is an effective method for revealing microbial ecological functions in native environments (Radajewski et al., 2000; Manefield et al., 2002; Dumont and Murrell, 2005; Huang et al., 2009a; Chen and Murrell, 2010). It usually employs equilibrium density gradient centrifugation to separate 13C-containing nucleic acids and the analysis of the separated nucleic acids links the function of microorganisms to their taxonomic identities (Radajewski et al., 2000; Manefield et al., 2002; Dumont and Murrell, 2005; Huang et al., 2009a; Chen and Murrell, 2010). Because of the subtle differences in the physical properties between stable isotopes, it would be a challenge to use the centrifugation-based SIP method for the isolation of 13C-incoporated DNA or RNA if 13C corporation was <50%. In this study, we have demonstrated that SCRR–SIP is sufficiently sensitive to distinguish between cells with only 10% 13C incorporation and their cells containing 13C at the natural 1.1% abundance (Figures 3, 4 and Supplementary Figure S4). Applying the SCRR–SIP technique to real environmental samples also demonstrates that it is able to identify photosynthetic cells with only ∼20% 13C content in their cellular compounds (Figure 6). Such a high sensitivity suggests that this technique may be useful in several applications: (1) detection of 13C-incorporated single microbial cells, avoiding the cross-feeding problem in SIP and aiding identification of the species that fix CO2; and (2) observation of the temporal dynamics of the 13C incorporation in different species by taking SCRR spectra or Raman images at different times of the incubation. In Synechocystis, the measured red shifts in the v1 and v2 bands at 0% and 10% 13C incorporation were 7.5 and 2.8 cm−1, respectively, indicating that SCRR–SIP may be able to detect 13C incorporation at levels even <10%. SCRR–SIP is a noninvasive technique, enabling the recovery of cells after measurement, which is especially useful for the study of unculturable microorganisms. For example, the two unknown rod-shaped bacteria identified in Figure 6c appeared to fix CO2 but so far they have not been cultured, and further work is needed to confirm their photosynthetic ability. In certain habitats, a few species of non-photosynthetic obligate chemolithotrophic bacteria that also contain carotenoids have been reported, though these types of bacteria appear to be rare. In marine ecosystems, it is also increasingly clear that non-photosynthetic autotrophic bacteria are widespread in, for example, oxygen minimum zones. Thus, it is also possible that autotrophic non-photosynthetic bacteria exist, and, interestingly, could be detected by this approach if used in future studies. The ν2 Raman bands appeared to vary in a narrower range (usually a few wave numbers) in different species, making it a potential biomarker for 13C incorporation in unidentified photosynthetic microorganisms. However, for samples that consist of various unknown species, the error of detection of 13C incorporation can be raised because of the uncertainty in the positions of the RR bands; therefore affecting the estimation of 13C-incorporation ratios. This indicates that a further investigation of the variation of ν2 Raman bands in different carotenoid-containing cells is required.

To search and screen large numbers of single cells from environmental samples, a rapid Raman imaging technique has great potential. Taking Figure 5a as an example, a 90.5 × 74.7 μm Raman image that consists of 27 000 Raman spectra/pixels was generated in ∼5 min with 1 ms acquisition time per pixel. If we assume a grid of 2 μm × 2 μm contains a single cell, 10 000 single cells can be mapped in ∼30 min. The Raman imaging time can be further reduced by using a more powerful computer, an optimised charge-coupled device detector and a refined spectrum readout strategy. It may also be possible to combine a laser microdissection system with this technique to selectively dissect and collect microorganisms of interest at the single cell level for attempted cultivation, genomic amplification and sequencing. In such a way, Raman imaging may provide a direct screening method for environmental samples without cultivation, and could help to reveal the precise ecological role of the ‘black box’ (unculturable microorganisms) in the microbial world.