We have performed multimodal imaging of live fibroblast cells infected by murine cytomegalovirus (mCMV). The infection process was monitored by imaging the two-photon fluorescence signal from a GFP-expressing strain of mCMV, whilst changes to lipid droplet configuration were observed by CARS imaging. This allowed us to identify three visually distinct stages of infection. Quantitative analysis of lipid droplet number and size distributions were obtained from live cells, which showed significant perturbations across the different stages of infection. The CARS and two-photon images were acquired simultaneously and the experimental design allowed incorporation of an environmental control chamber to maintain cell viability. Photodamage to the live cell population was also assessed.
©2011 Optical Society of America
The physical dimensions of viruses are smaller than the resolution limit of far field microscopy thereby precluding direct observation. As a result, many studies focus on examining the host–pathogen interactions which can be observed with far field microscopy. These studies generally involve labelling cells with fluorescent markers. However, labels can degrade into toxic photochemical reaction products and can also perturb the native biochemical pathways which operate between host and pathogen. This has restricted many previous investigations to working with fixed rather than living cells.
In this paper we demonstrate the application of far field microscopy to the study of host–pathogen interactions in unlabelled live cells. Specifically, we examine the correlation between cytomegalovirus infection and the behaviour of cellular lipid droplets by imaging live cells on a multimodal microscope platform which combines coherent anti-Stokes Raman scattering (CARS) and two-photon fluorescence (TPF). This work extends our previous study  which was based only on fixed cell samples.
The importance of label-free live cell imaging has led to the development of a variety of new methods for far field microscopy. Among them, CARS has emerged as the ideal technique for imaging cellular lipid droplets [2, 3]. The basic functions of lipid droplets as a store for energy reserves, hydrophobic compounds, and membrane building blocks have long been understood . Recently, the more subtle and complex role played by these organelles has been revealed, including the discovery that lipid droplets are appropriated by the hepatitis C virus during the infection cycle [5–8]. CARS is clearly a valuable tool for the study of the link between lipids and viral replication. Used alone however, CARS is insufficient for identifying the infection state of cells, so we combine CARS with TPF to simultaneously image cellular lipid droplets whilst identifying and monitoring viral activity using a genetically modified murine cytomegalovirus (mCMV). The virus causes the host cell to express green fluorescent protein (GFP) upon infection.
The use of the modified virus allows the cellular expression of GFP to be utilised as a definitive marker of infection, to indicate that cellular translation of viral mRNA has begun. The modified genetic code persists in the DNA of viral progeny, therefore GFP expression continues for an extended time period after the initial infection. Furthermore, with this technique, it is possible to use the GFP expression to correlate observed cell changes with the overall timescale of the infection process. This imaging technique avoids the problems of toxicity, changes to cell function and fluorophore fading inherent to fluorescent labels.
In our previous study  we combined CARS and TPF imaging to compare fixed samples of healthy (uninfected) and infected mouse fibroblast cells. The main results were that the virus causes morphological changes to the cell, including expansion of the cell nucleus, and that an apparent rearrangement of lipid droplets occurs at some point during the infection process.
The aim of this study is to understand this dynamic process by relating the observed changes in lipid droplet configuration with the GFP expression, and hence the infection stage. A preliminary study of fixed cell samples is used to systematically identify the stages of viral infection by observing the quantity of GFP expression within the cell. We then image the lipid droplet behaviour in live cells during the infection process by using the fixed cell data as a reference to identify the stage of infection from the level of GFP expression.
2. Experimental Method
2.1. Sample Preparation
The method for sample preparation and infection with the genetically modified virus was described in detail in a previous paper . Briefly, NIH/3T3 fibroblast cells with a passage number in the range 11–20 were cultured in Dulbecco’s modified Eagle’s medium (DMEM), with L-glutamine (200 mol/m3), penicillin/streptomycin (10 000 units/ml) and 10 % calf serum. Standard growth conditions of 37 °C and a 5 % CO2 environment were maintained during an incubation period of 12 hours. Cells were seeded onto glass-bottom Petri dishes (P35Gcol-1.0-14-C, MatTec). Infection with mCMV wildtype mutant C3X was carried out using a genetically modified version of the Smith strain  at a multiplicity of infection of 1, with the growth media containing 3 % calf serum in order to reduce cell growth.
2.2. Preliminary Study
A preliminary study was conducted with fixed cells, prepared as described above, with the exception that the cells were seeded onto glass coverslips and fixed using 1 % formaldehyde in phosphate buffered saline. By using a synchronized transfection method  a range of samples were fixed at known time points between 0 and 96 hours post transfection (hpt). The samples therefore consisted of a large number of cells fixed at the same time (hpt) in the infection cycle.
2.3. Live Cell Chamber
A portable cell chamber was built to provide a suitable environment for sustaining live cells on the microscope stage. The chamber was designed to attach directly onto the microscope, as shown in Fig. 1. Since the function of the chamber was to sustain a stable population of living cells, access for passaging or perfusion was not required. Therefore, the chamber was designed to be airtight, to maintain a 5 % CO2 atmosphere, and to provide a controlled temperature of 37.0 °C . The top window of the chamber was designed to touch the growth media in order to minimize signal loss due to refraction and to prevent the formation of condensation on the glass. We found that this set-up enabled live cells to survive on the microscope for at least 7 days. This was verified by observing a gradual growth in cell population from the initial 50–70 % to 100 % confluence over the period. A phenol-red indicator confirmed that neutral pH was maintained, suggesting that the CO2 concentration remained constant inside the chamber [12, 13]. As cytomegalovirus completes its replication cycle after 72 hours at most , the longevity of cells in the chamber is more than sufficient for the process being investigated.
2.4. Imaging Instrumentation
A custom-built laser-scanning inverted CARS microscope setup , shown in Fig. 2, was used to record both CARS and TPF images simultaneously. The laser source was a pair of titanium-sapphire lasers (Mira 900, Coherent), which were mode-locked and synchronized via a phase-locked loop system  (Synchrolock, Coherent).
The 2845 cm−1 Raman vibration, which corresponds with the aliphatic C–H vibrations found in lipids , was targeted by tuning the pump laser wavelength to 714 nm and the Stokes laser to 896 nm. This generated a CARS signal at 593 nm, which was reflected out of laser beam path with a long wave-pass dichroic mirror (683dcxr, Chroma) and further filtered with a short wave-pass filter (680/SP-25, Semrock) and a narrow band-pass filter (600/14-25, Semrock). The forward CARS (F-CARS) signal was detected with a photomultiplier tube (PMT) (R3896, Hamamatsu).
The Stokes beam at 896 nm falls within the excitation band of GFP, allowing simultaneous TPF imaging with CARS. Unlike CARS, TPF is non-directionally scattered  and is most efficiently collected in the backward direction. A dichroic mirror behind the objective lens reflected the TPF signal out of the beam path. An identical PMT (R3896) was used to detect the signal. Two filters, a short wave-pass (680/SP-25, Semrock) and a broad band-pass (510/84-25, Semrock), were used to block the pump and Stokes laser wavelengths whilst transmitting most of the GFP emission spectrum. The band-pass filter also blocks the CARS signal at 593 nm. The picosecond pulse duration used in this set-up is ideal for CARS microscopy, but less so for TPF. Nonetheless, it is still capable of producing high-quality TPF images .
Image acquisition was achieved with a customized LabVIEW program, which controls the beam scanning mirrors and records the CARS and TPF PMT signals. The laser power at the sample was adjusted with neutral density filters (not shown in the figure). Bright field microscopy was used for all sample alignment and focusing in order to minimize the exposure of the sample to the laser beam.
3. Results and Discussion
3.1. Multimodal microscopy: simultaneous CARS and TPF
We used CARS to target the C–H Raman band at 2845 cm−1. Bright spots were observed in the cytosol, which were identified as the resonant CARS signal from lipid droplets . The cellular and nuclear membranes were also visible in CARS images due to the nonresonant background, revealing the cell morphology and the outline of the nucleus as a dark elliptical region.
Simultaneous and sequential CARS and TPF images are shown in Fig. 3. Sequential TPF images (Fig. 3(C)) were collected by scanning the Stokes laser beam only, the wavelength of which falls within the two-photon excitation band of GFP. It is evident that the TPF images generated simultaneously with CARS (Fig. 3(B)) are comparable to the images generated sequentially (Fig. 3(C)), with both TPF images revealing similar features. The bright high-contrast spots in the simultaneous image (Fig. 3(B)) could in principle arise from either CARS or GFP. To confirm that cross-channel leakage was not significant, pairs of images were taken with the phase of the pump and Stokes beams shifted by approximately 180° using the phase-locked loop system (Synchrolock). When the beams are out of phase, pump and Stokes pulses do not arrive at the sample simultaneously, so frequency mixing cannot occur and no CARS signal is generated. This test allows ready discrimination of the TPF signal from the CARS and confirmed that the filters prevented cross-channel leakage. The bright spots in TPF images recorded simultaneously with CARS (Fig. 3(B)) were therefore unambiguously identified as sites of GFP accumulation.
3.2. Cell Viability
The setup enabled acquisition of high-contrast CARS images using a reasonably low laser power of approximately 35 mW at the sample. Figure 4 shows a stack of images of a single live cell at increments of 0.25 μm, starting from the coverslip and scanning upwards in the z-direction. Bright field images were taken before (Fig. 5(A)) and after (Fig. 5(B)) recording the stack to assess possible photodamage. The image taken after recording the stack shows morphological changes to the cellular membrane, suggesting that some photodamage had occurred. Images of the same cell recorded two hours later, Fig. 6, revealed that the cell had become rounded and detached from the surface of the Petri dish. Nearby cells retained a normal appearance, indicating that the cell’s death was due to photodamage.
The accumulated laser exposure experienced by a cell during stack acquisition is far greater than required for imaging a single slice. We tested cell viability against a range of laser powers in order to determine the optimum parameters for 2D imaging. We found that acquiring a 512×512 pixel image and integrating over 4 scans provided the optimum balance between noise, image contrast and acquisition time. The minimum power achieved at these parameters was approximately 28 mW at the sample. Increasing the laser power increased the image contrast, but also increased the likelihood of photodamage.
Viability was tested with different cells over a range of laser powers in order to determine whether higher laser powers caused immediate damage to the cell morphology. It was found that powers up to 120 mW could be used without causing morphological changes in the cell on the timescale of the imaging. However the accumulated photodamage typically caused a delayed change in cell morphology after imaging. As a result, long time-course experiments, which involve imaging a single cell over many hours cannot be carried out at present. Instead, we are able to image many cells within the same live cell population over a time period of interest. This gives an overall view of the state of the cell population at the specific time point, and thus allows the change in overall cell population to be assessed over the time-course of infection.
3.3. Preliminary Study: Identification of Infection Stages
Samples of infected cells were fixed at 3 hour intervals and visually classified into three distinct groups: group 1, cells show low or nil concentrations of GFP, along with no significant changes to lipid droplet configurations compared to uninfected cells (0–9 hpt); group 2, cells show higher concentrations of GFP and significant changes to lipid droplet configuration (9–24 hpt); group 3, cells show major morphological changes and lipid droplet configuration changes, plus very high concentrations of GFP observed in the cells (> 24 hpt). We refer to these groups respectively as the early, intermediate and late stages of infection.
All cell samples in the early infection stage (0–9 hpt) displayed similar features to those exhibited by the cell in Fig. 7. During this stage, CARS images showed both uniform size and spatial distributions of lipid droplets throughout the cytosol. TPF images had zero or limited contrast, indicating either no GFP, or the presence of a low concentration of GFP in the cytosol. There was no TPF signal from the nucleus, as GFP is expressed in the cytoplasm. The images also show areas of small accumulations of GFP in the cytosol, which could indicate the sites of GFP expression or suggest vesicle-mediated GFP transport.
Cell samples infected for 9–24 hpt, the intermediate stage, typically exhibited the features shown in Fig. 8. The cells in this group displayed some lipid droplet clustering, where nonuniform spatial distributions of lipid droplets and some relatively large lipid droplets were identifiable. In addition, a general increase in cell nucleus area was evident. This is consistent with our earlier study , although in that investigation we were unable to associate nuclear expansion with a definite infection stage. The larger overall TPF signal indicated higher concentrations of GFP in the cytosol, thus confirming a progression of the infection.
Samples containing cells infected for > 24 hpt, Fig. 9, we designated as the late stage of infection. Cells in this infection stage imaged using CARS show a dramatic reduction in the number of lipid droplets. The corresponding TPF images indicated high concentrations of GFP in both cytosol and nucleus, showing that GFP is imported into the cell nucleus at this late infection stage. Additionally, many images display fragmented cell nuclei, which suggests the cells are undergoing apoptosis.
Since GFP is expressed as a result of viral protein translation, we expect the fluorescence intensity to increase as the infection progresses. Figure 10 shows the measure of intensity of intracellular fluorescence per cell throughout the infection process. The graph shows a significant increase in fluorescence intensity as the infection progressed, where an average intensity of 17.3, 160.4 and 1025.9 (normalised units) was measured for each group of infected cells. Pre-infection TPF images were also acquired where an average intensity of 0.09 was measured, indicating that there was no significant autofluorescence background in TPF images.
3.4. Imaging of Live Infected Cells
Although our preliminary study allowed us to group the infection process into three visually distinct stages, it is unsuitable for quantitative analysis of the observed changes in lipid droplet configuration because the formaldehyde fixing method causes rapid degradation of lipid droplets . Therefore, the use of live cell samples is imperative for further analysis of lipid droplet characteristics during the infection process. Different cells in live samples had often progressed to different stages of the infection process, therefore within one image, each cell may be at a different stage of infection, as evident in Fig. 13(B). Using the data from the preliminary study we were able to assign the live cells to a definite stage (early, intermediate or late) of the infection process.
Cells in the early stage of infection provided a low overall intensity in TPF images. GFP expression is initiated by cellular translation of viral mRNA, therefore the initial appearance of GFP within the cell can be attributed to the onset of viral replication. CARS images of live cells acquired during the early infection stage, such as the image shown in Fig. 11, clearly show regularly-sized lipid droplets uniformly distributed throughout the cytosol, consistent with the preliminary study (Fig. 7).
CARS images acquired from cells in the intermediate stage of infection, such as the image shown in Fig. 12, showed non-uniform spatial distribution of lipid droplets throughout the cytosol. In addition, some relatively large lipid droplets were seen, suggesting a clustering effect. This behaviour is consistent with observations in the preliminary study (Fig. 8). Furthermore, some images of live cell infection, including Fig. 12, suggested that lipid droplets migrated towards neighbouring cells. Live cell TPF images revealed GFP in a number of cell nuclei. This feature was not observed in fixed cell images at the same infection stage, where GFP was only detected in the nucleus during the late stages of infection. These live cell images suggest that nuclear import of GFP occurs during the same infection stage as lipid droplet aggregation, indicating a possible link between the two processes. The import of GFP into the nucleus was an unexpected feature. Since GFP is synthesized in the cytosol, and nucleocytoplasmic transport is highly controlled , large molecules such as GFP must be actively transported in order to enter the nucleus. This suggests that the behaviour may not be a cellular function initiated by the virus but rather a side-effect of infection caused by some mechanism occurring within the cell during the infection cycle .
CARS images of cells in the late stages of infection, Fig. 13, show a significant reduction in the number of lipid droplets. Where lipid droplets were still visible, much larger droplet sizes were typical. The high intensity in TPF images provided evidence of fragmented cells and apoptosis.
To confirm that the observed reduction in lipid droplet number is linked to the infection process, a sample of 69 cells was examined in more detail. The cells were first grouped into the three infection stages-early, intermediate and late-based on the level of GFP expression identified by the preliminary study. For each infection stage group, the number of lipid droplets per cell was counted. The histogram in Fig. 14 shows the probability that a cell has a given number of lipid droplets during the early, intermediate and late stages of infection. A clear decrease in lipid droplet number throughout the infection process is apparent. The mean number of lipid droplets are 63, 29 and 14, for the early, intermediate and late infection stages respectively, where the early infection probability distribution is between 37 → 106 lipid droplets, the intermediate infection probability distribution is between 12 → 54 lipid droplets, and the late infection probability distribution show that no cell contains more than 21 lipid droplets from the data set collected.
In addition to a reduction in the number of lipid droplets, there appeared to be an increase in the size of the remaining droplets. In order to quantitatively determine the change in size due to clustering, the areas of lipid droplets were measured in a large sample of cells using ImageJ . 545, 897 and 295 lipid droplet areas were measured for the early, intermediate and late infection stages respectively. A histogram showing the probability of measuring lipid droplets with a certain size during each infection stage is shown in Fig. 15. In the early infection stage, 86 % of the cells’ lipid droplets were less than 1 μm2 in size. This confirmed the observation in CARS images that lipid droplets were broadly similar in size during this infection stage. The graph also shows high probability of measuring lipid droplets with sizes less than 1 μm2 for the intermediate and late infection stages, both 77 %. The difference between the lipid droplet sizes throughout the infection stages are most evident when looking at lipid droplets with area greater than 6 μm2. In the intermediate infection stage, there was an increased probability of finding lipid droplets larger than 6 μm2, with the probability further increasing in the late infection stage. No lipid droplets larger than 5.8 μm2 were measured in cells in the early infection stage, showing that as the infection progresses, the probability of finding larger lipid droplets increases. This behaviour is highlighted by the inset in Fig. 15, where the histogram shows the probability distribution in more detail for the lipid droplets larger than 2 μm2. These results confirm the observation of progressive lipid droplet clustering during infection.
We have shown that a multimodal CARS and TPF microscope system allows the observation and measurement of changes in live fibroblast cells that occur during infection by cytomegalovirus. Imaging with TPF allowed us to identify three distinct stages of the infection process; early, intermediate and late infection stage, whilst simultaneous CARS imaging was used to observe the changes in cell morphology, and to quantify the perturbations to cellular lipid droplets, enabling us to correlate these findings against the defined infection stages. This enabled us to identify several persistent features of cytomegalovirus infection including a change in cell morphology, a decrease in the number of cellular lipid droplets and an overall increase in the size of the remaining droplets, thus suggesting a clustering of lipid droplets. The increase in the size was significant, with some lipid droplets of area greater than 6 μm2 emerging in the later stages of infection. The biological significance of these observed processes requires further investigation.
We appreciate the provision of C3X-GFP virus from Mathieu Blanc of the Division of Pathway Medicine, University of Edinburgh.
References and links
1. I. Robinson, M. A. Ochsenkühn, C. J. Campbell, G. Giraud, W. J. Hossack, J. Arlt, and J. Crain, “Intracellular imaging of host-pathogen interactions using combined CARS and two-photon fluorescence microscopies,” J. Biophoton. 3, 138–146 (2010). [CrossRef]
2. J. Cheng, “Coherent anti-Stokes Raman scattering microscopy,” Appl. Spectrosc. 61, 197A–208A (2007). [CrossRef]
3. C. L. Evans and X. S. Xie, “Coherent anti-Stokes Raman scattering microscopy: chemical imaging for biology and medicine,” Annu. Rev. Anal. Chem. 1, 883–909 (2008). [CrossRef]
6. R. K. Lyn, D. C Kennedy, A. Stolow, A. Ridsdale, and J. P. Pezacki, “Dynamics of lipid droplets induced by the hepatitis C virus core protein,” Biochem. Biophys. Res. Commun. 399, 518–524 (2010). [CrossRef] [PubMed]
7. E. Herker, C. Harris, C. Hernandez, A. Carpentier, K. Kaehlcke, A. R. Rosenberg, R. V. Farese, and M. Ott, “Efficient hepatitis C virus particle formation requires diacylglycerol acyltransferase-1,” Nat. Med. 16, 1295–1298 (2010). [CrossRef] [PubMed]
11. G. Sluder, J. J. Nordberg, F. J. Miller, and E. H. Hinchcliffe, “A sealed preparation for long-term observations of cultured cells,” in Live Cell Imaging: A Laboratory Manual, R. D. Goldman and D. L. Spector eds. (Cold Spring Harbor Laboratory Press, 2005), pp. 345–349.
12. D. J. Stevenson, D. J. Carnegie, B. Agate, F. Gunn-Moore, and K. Dholakia, “Long-term cell culture on a microscope stage: the carrel flask revisited.” Microsc. Anal. 22, 9–11 (2008).
13. M. E. Dailey, D. C. Focht, A. Khodjakov, C. L. Rieder, K. R. Spring, N. S. Claxton, S. G. Olenych, J. D. Griffin, and M. W. Davidson, “Maintaining live cells on the microscope stage,” http://www.microscopyu.com/articles/livecellimaging/livecellmaintenance.html.
14. E. S. Mocarski, T. Shenk, and R. F. Pass, “Cytomegaloviruses,” in Fields Virology, D. M. Knipe and P. M. Howley eds. (Lippincott Williams and Wilkins, 2006) pp. 2702–2772.
15. A. Yu, X. Ye, D. Ionascu, W. Cao, and P.M. Champion, “Two-colour pump-probe laser spectroscopy instrument with picosecond time-resolved electronic delay and extended scan range,” Rev. Sci. Instrum. 76114301 (2005). [CrossRef]
16. X. Nan, J. X. Cheng, and X. S. Xie, “Vibrational imaging of lipid droplets in live fibroblast cells with coherent anti-Stokes Raman scattering microscopy,” J. Lipid. Res. 44, 2202–2208 (2003). [CrossRef] [PubMed]
17. W. Denk, D. W. Piston, and W. W. Webb, “Multi-photon molecular excitation in laser scanning microscopy,” in Handbook of Biological Confocal MicroscopyJ. B. Pawley, (Springer, 2006) pp. 535–549. [CrossRef]
18. J. Bewersdorf and S. W. Hell, “Picosecond pulsed two-photon imaging with repetition rates of 200 and 400 MHz,” J. Microsc. 191, 28–38 (1998). [CrossRef]
19. M. A. Hayat, “Chemical fixation,” in Principles and Techniques of Electron Microscopy Biological Applications (Cambridge University Press, 2000), pp. 43–44.
21. W. Muranyi, J. Haas, M. Wagner, G. Krohne, and U. H. Koszinowski, “Cytomegalovirus recruitment of cellular kinases to dissolve the nuclear lamina,” Science 297, 854–857 (2002). [CrossRef] [PubMed]
22. W. S. Rasband, “ImageJ,” U. S. National Institutes of Health, http://imagej.nih.gov/ij/.