Label-free in vivo analysis of intracellular lipid droplets in the oleaginous microalga Monoraphidium neglectum by coherent Raman scattering microscopy

Oleaginous photosynthetic microalgae hold great promise as non-food feedstocks for the sustainable production of bio-commodities. The algal lipid quality can be analysed by Raman micro-spectroscopy, and the lipid content can be imaged in vivo in a label-free and non-destructive manner by coherent anti-Stokes Raman scattering (CARS) microscopy. In this study, both techniques were applied to the oleaginous microalga Monoraphidium neglectum, a biotechnologically promising microalga resistant to commonly applied lipid staining techniques. The lipid-specific CARS signal was successfully separated from the interfering two-photon excited fluorescence of chlorophyll and for the first time, lipid droplet formation during nitrogen starvation could directly be analysed. We found that the neutral lipid content deduced from CARS image analysis strongly correlated with the neutral lipid content measured gravimetrically and furthermore, that the relative degree of unsaturation of fatty acids stored in lipid droplets remained similar. Interestingly, the lipid profile during cellular adaption to nitrogen starvation showed a two-phase characteristic with initially fatty acid recycling and subsequent de novo lipid synthesis. This works demonstrates the potential of quantitative CARS microscopy as a label-free lipid analysis technique for any microalgal species, which is highly relevant for future biotechnological applications and to elucidate the process of microalgal lipid accumulation.


Supplementary Figures
Supplementary Figure S1: Application of the Nile Red reagent for neutral lipid droplet staining. Cells from Monoraphidium neglectum and Chlamydomonas reinhardtii CC1690 (cell-wall containing strain as staining control) were transferred to nitrogen-free medium, incubated for five days to induce lipid accumulation and then subjected to the Nile Red staining procedure. The data points represent average relative fluorescence values per cell (n=4); error bars indicate SEM. In contrast to C. reinhardtii (grey lines), lipid droplet formation in M. neglectum during the -N treatment did not result in an increase of the Nile Red fluorescence (black lines). Blue dotted lines highlight Raman peaks which are attributable to carotenoids. Note that the strong signal is not due to high carotenoid abundance, but rather the resonant excitation of delocalized πelectron systems in the pigments, leading to a strong amplification of the Raman signal. Green dotted lines indicate Raman bands which belong to lipids. The resonances at 865 cm -1 and 1085 cm -1 belong to the choline and phosphate group of lipids, respectively, indicating the presence of polar lipids in lipid droplets (LDs). Thus, it is likely that the TAG rich core of LDs is surrounded by a phospholipid layer. The signal at 1750 cm -1 originates from the C=O alkyl-ester vibration, indicating the presence of DAGs and TAGs inside the LD. The signal at 1302 cm -1 and 1442 cm -1 is representative of the CH 2 shear vibration of alkyl groups, whereas 1264 cm -1 and 1660 cm -1 indicates C=C stretching vibration of alkyl groups. The CH 2 signal is indicative of total lipids, since CH 2 groups belong to both saturated and unsaturated fatty acids, whereas the C=C double bond appears only in unsaturated fatty acids. The intensities of the peaks at 1264 cm -1 and 1302 cm -1 strongly depend on the structure and conformation of the fatty acid inside the lipid molecule and are as such not suited as indicators of the quantity of the respective lipid type. Therefore, the peaks at 1660 cm -1 and 1442 cm -1 were utilized to determine the amount of desaturation among the total lipids from an individual LD, which is given by the ratio of I 1660/1442 . Background substraction was performed by the Lieberalgorithm [1].

Cell growth parameters
To monitor cell growth, the parameters optical density (at 750 nm), cell count (Beckmann Coulter Counter) and dry biomass weight (25 mL culture washed once with distilled water and dried overnight at 105 °C) were recorded in technical triplicates per biological replicate. Technical replicates were averaged prior data analysis. Cell weight (as pg cell -1 , for subsequent determination of the cellular neutral lipid content as shown in Supplementary Fig. S5) was obtained by the following formula: −1 ] * 1000 * 10 12

Lipid extractions and chromatography
Lipid extractions and chromatographic analyses were performed in one (day 0 and 1) or two (day 2, 4 and 8) technical replicates per biological replicate from 10 -30 mg lyophilized biomass. After homogenization (3 × 45 s at 6,500 rpm using a Precellys 24, Peqlab, Erlangen, Germany), total lipids were extracted and separated into polar and neutral lipids as described previously [4]. Net total, neutral and polar lipid amounts were determined gravimetrically. The cellular neutral lipid content was obtained by: Nile Red staining procedure (Supplementary Fig. S1) In addition to this, images were also acquired by illuminating the sample with the modulated 1064 nm beam only. Still, a very weak signal of the chlorophyll fluorescence could be observed in the lipid channel especially at low lipid concentrations (e.g. at day 0). This image (referred to as background) was subtracted from the CARS image generating the final image, which contains only lipid signals. For additional sample information delivered by the chlorophyll fluorescence, a 600 nm long-pass filter is implemented into the signal beam path, which reflects wavelengths shorter than 600 nm. A second PMT detects the signal and generates similar to the CARS channel a fluorescence image displaying the exact position of the chloroplast in a simultaneous acquisition manner.

CARS acquisition settings
After gentle centrifugation to concentrate the culture slightly, the cells were placed on ice until CARS microscopy was performed (within one hour). 10 µL cell suspension was loaded onto a microscopic slide and incubated in the dark for 10 minutes During this incubation period, the non-motile M.
neglectum cells settled at the bottom and due to the fast scanning times (the average acquisition time for a cell of 60 µm² (Supplementary Figure S4) was 88 ms (2D scanning mode) and 375 ms (3D scanning mode), respectively), an additional immobilization step was not required. The lateral and axial resolution of the CARS system was approximated to 400 nm and 650 nm, respectively.
For CARS microscopy the focal intensities were set to 11.6 mW for the Stokes (1064 nm) and 23.3 mW for the pump beam (816.7 nm). Here, the typical pixel dwell time was 31.25 µs with 1024 px x In order to quantify intracellular lipids in the entire volume of the cells a 3D data set for one biological replicate per day with a step size of 500 nm in the axial (z) direction was acquired utilizing a piezo-stage (S/N 222009, Märzhäuser Wetzlar, Germany). A full 3D scan contains approx. 45 images covering the full volume of all cells (diameter 5 -10 µm) within the field of view. In this step a smaller field view (50 µm x 50 µm) was scanned to optimize contrast and sensitivity in combination with a suitable acquisition time (approx. 30 min). These images contain 512 x 512 pixels with a pixel dwell time of 62.5 µs.

CARS image analysis
Each image was recorded three times, initially with both lasers on, then, to assess the fluorescence background, with either the pump beam or the Stokes beam blocked, respectively. The Stokes image was subtracted from the initial CARS image by the 'Image Calculator' function in ImageJ (version 1.46a). For 3D scans, only two images were recorded, the first with both lasers on and the second with the Pump beam blocked. Again, the Stokes image was subtracted from the first. assuming LDs as spherical objects. The cell was approximated as an ellipse and its area given by = 2 2 , where d x is the diameter of the long axis of the cell and d y the diameter of the short axis of the cell. To determine the relative LD area, three steps were applied: first, the area of a cell was determined. Second, the areas of all LDs visible in this cell were summed up. Third, this sum of LD areas was divided by the cell area to yield the relative LD area. This procedure was repeated for 30 cells per day and replicate (90 per day).
The volume of LDs was estimated from the 3D scans. The shape of a LD was estimated as a sphere, and its volume, given by = 4 3 ( 2 ) 3 , was calculated from the diameter of the z-section in which the LD was largest. For the relative LD volume analysis, all LDs in a single cell were quantified as well as all cells per 3D reconstruction (n = 9 per day, except day 1 where n = 8). To calculate the volume of a cell, its shape was approximated by an ellipsoid. The diameter of the height of the cell (d z ) was estimated from the number of z-sections in which the cell size remained constant, and the x and y diameters (d x and d y ) were determined in the z-section in which the cell was largest. The volume of this ellipsoid was calculated according to = 4 3 2 2 2 . The sum of volumes of LDs was divided by the respective cell volume to derive the relative LD volume.

Statistical analysis
Data were collected and subsequently visualized as well as tested for significance (two-sided Wilcoxon Rank Sum test assuming a previous day having e.g. less LDs per cell section than the current one) with a Python script incorporating R (version 2.14.1) by the package 'rpy2' (version 2.2.5).
Significant changes of the gravimetric neutral lipid amounts were calculated in Excel with a two-sided t-test with variance set to heteroscedasticity; prior to this, technical replicates were averaged. For the correlation analysis of the 3D relative LD volume ( Supplementary Fig. S5), average values and standard deviations were obtained by creating -for each day -three lists (NL, CW, RV) with 2.5 * 10 6 data points; those were obtained by resampling from normal distributions described by average and standard deviation of the (i) neutral lipid content (as % dry weight, list NL), (ii) cell weight (as pg cell -1 , list CW) and (iii) relative LD volume per cell (as % of cell volume, list RV). Every data point in list NL was then multiplied by its counterpart in list CW (yielding cellular neutral lipid content as pg cell -1 ), and average and standard deviation were calculated for this new list, yielding the average cellular neutral lipid content and its standard deviation; this was repeated for RV and CW.

Raman spectroscopy based on spontaneous Raman scattering
Raman spectroscopy can be performed on living cells; any additional fixation is not required. Due to the fact that the measurements were not carried out in situ, cells were fixed with 4 % paraformaldehyde (PFA) in 1x phosphate-buffered saline (PBS) for 10 -15 min and stored at 4 °C in the dark till the transportation.
Per biological replicate, 5 cells (a total of 15 cells per day) were analyzed by spontaneous Raman spectroscopy with a custom-built Raman micro-spectroscopy setup. 10 µL cell suspension in water (10 7 cells mL -1 ) were put onto a quartz cover slip (made of quartz in order to minimize fluorescence from usual borosilicate cover slips). The clusters of individual lipid droplets were placed in the focus of a 100x oil immersion objective lens (UPLFNL, NA = 1.3, Olympus, USA) and subsequently investigated. As an excitation source a laser diode (Innovative Photonic Solutions, USA) with an output wavelength of 785 nm was used generating a focal intensity of 25 mW. The Raman spectra were detected by a spectrograph (Acton 2300i, Princeton Instruments, New Jersey, USA) and an aircooled CCD camera (Newton DU 920P BR-DD, Andor, Belfast, UK). The exposure time was 20s. Each presented spectrum is the accumulation of three spectra which were obtained at the same sample