Single cell assessment of yeast metabolic engineering for enhanced lipid production using Raman and AFM-IR imaging
Biodiesel is a valuable renewable fuel made from derivatized fatty acids produced in plants, animals, and oleaginous microbes. Of the latter, yeasts are of special interest due to their wide use in biotechnology, ability to synthesize fatty acids and store large amounts of triacylglycerols while utilizing non-food carbon sources. While yeast efficiently produce lipids, genetic modification and indeed, lipid pathway metabolic engineering, is usually required for cost-effective production. Traditionally, gas chromatography (GC) is used to measure fatty acid production and to track the success of a metabolic engineering strategy in a microbial culture; here we have employed vibrational spectroscopy approaches at population and single cell level of engineered yeast while simultaneously investigating metabolite levels in subcellular structures.
Firstly, a strong correlation (r2 > 0.99) was established between Fourier transform infrared (FTIR) lipid in intact cells and GC analysis of fatty acid methyl esters in the differently engineered strains. Confocal Raman spectroscopy of individual cells carrying genetic modifications to enhance fatty acid synthesis and lipid accumulation revealed changes to the lipid body (LB), the storage organelle for lipids in yeast, with their number increasing markedly (up to tenfold higher); LB size was almost double in the strain that also expressed a LB stabilizing gene but considerable variation was also noted between cells. Raman spectroscopy revealed a clear trend toward reduced unsaturated fatty acid content in lipids of cells carrying more complex metabolic engineering. Atomic force microscopy-infrared spectroscopy (AFM-IR) analysis of individual cells indicated large differences in subcellular constituents between strains: cells of the most highly engineered strain had elevated lipid and much reduced carbohydrate in their cytoplasm compared with unmodified cells.
Vibrational spectroscopy analysis allowed the simultaneous measurement of strain variability in metabolite production and impact on cellular structures as a result of different gene introductions or knockouts, within a lipid metabolic engineering strategy and these inform the next steps in comprehensive lipid engineering. Additionally, single cell spectroscopic analysis measures heterogeneity in metabolite production across microbial cultures under genetic modification, an emerging issue for efficient biotechnological production.
KeywordsLipid bodies Spectroscopy Metabolic engineering Heterogeneity Triacylglycerol Fatty acids
atomic force microscopy–infrared spectroscopy
caleosin from Arabidopsis thaliana
attenuated total reflection-Fourier transform infrared spectroscopy
confocal Raman spectroscopy
dry cell weight
fatty acid methyl ester
gas chromatography with flame ionization detector
phosphate buffer solution
partial least square
saturated fatty acid
standard normal variate
triacylglycerol lipase 3
unsaturated fatty acid
Biodiesel is a versatile renewable fuel composed of fatty acid methyl esters produced from animal and plant lipids-triacylglycerols (TAGs) [1, 2]. Microbial oils also have advantages as a source of TAG: microorganisms can be grown on non-food sugars, are minimally affected by seasons, generally robust and have short life cycles . Many types of oleaginous eukaryotic species accumulate TAG in lipid bodies (LB), intracellular organelles considered primarily as storage vesicles for neutral lipids, in amounts up to 70% of total DCW biomass [1, 3, 4]. However, for cost-effective and high yielding production of TAG in oleaginous organisms such as yeast, metabolic pathway engineering and maximizing yield across the cultures is necessary [1, 3, 5, 6].
The metabolic engineering strategies for improvement of lipid production include up and down regulation of genes involved in the main steps of yeast lipid production: fatty acid (FA) biosynthesis, lipid accumulation and sequestration [1, 7]. Increased fatty acid production can be achieved by higher expression of aldehyde dehydrogenase (ALD), acetyl-CoA synthetase (ACS)  and acetyl-CoA carboxylase (ACC) [9, 10] and FA accumulation in TAG is catalyzed by diacylglycerol acyltransferase (DGAT)  and other acyltransferases. Heterologous expression of highly active acyltransferases such as the DGAT1 from Arabidopsis thaliana has been shown to increase lipid yield in the yeast Saccharomyces cerevisiae  and similarly, LB stabilization proteins such as caleosin (AtClo1)  have been shown to support LB formation. Down regulation of lipid mobilization enzymes such as the major TAG lipase, Tgl3, in S. cerevisiae, can also increase lipid content of cells .
Another factor affecting the productivity of microbially produced lipids is heterogeneity in individual cell performance across a culture . Some of the variance is due to the stochastic nature of biological processes within gene regulation, transcription and translation, but response can be further distorted by variable responses to environmental conditions  including genetic engineering. While heterogeneity in cellular processes among isogenic microbial cells in a population has been known for some time, recent developments in visualization technologies have allowed measurement and monitoring of metabolite production both of populations and single cells within.
The formation and growth of LBs in cells are most commonly analysed by fluorescence imaging or histochemical staining ; these enable the visualization of LBs in multiple cells at once, but without providing details of their chemical composition. Vibrational spectroscopy-based techniques such as infrared (IR) and Raman spectroscopy (RS) are used in combination with digital imaging to enable rare insights into single cells of microalgae and yeast [17, 18, 19, 20]. In particular, confocal Raman spectroscopy (CRS) has revealed intracellular structures such as LBs, due to its high spatial resolution falling in the range of few hundred nm depending on the excitation wavelength  and chemical composition information of selected intracellular structures [18, 19, 22]. The spatial resolution of imaging with the use of conventional IR spectroscopy is restricted by wavelength diffraction spatial resolution limit (~ 5.5 µm), making it less suitable for the study of intracellular structures of microorganisms. Recently, however, a novel technique has been developed, based on a combination of IR spectroscopy and Atomic Force Microscopy (AFM–IR). In AFM–IR, IR absorption spectrum is acquired indirectly, by measuring the resulting thermal expansion of the sample. This approach overcomes the limitation of conventional IR-based imaging and achieves a spatial resolution of ~ 100 nm, thus enabling analysis of cellular constituents at the single cell level.
The aims of this research were to determine the effectiveness of metabolic pathway engineering approaches for enhanced lipid content in yeast through an examination of subcellular structures including LBs and to examine the heterogeneity of response in S. cerevisiae yeast cultures via vibrational spectroscopic approaches. We measured the overall changes in lipid content and other metabolites at a population level in whole cells by ATR-FTIR following the introduction of genetic modifications, then undertook detailed studies of subcellular structures and changes in their chemical composition in single cells using CRS and AFM–IR. These approaches provided detailed, single cell information on subcellular structures at high-resolution for the engineered strains and also allowed representative sampling of the population for these traits. The increase in the total lipid content of engineered cells demonstrated in bulk samples reflected either a greatly increased number or greatly increased size of LBs; the increased size was likely due to the influence of the LB stabilizing protein, caleosin, expressed in these cells. In engineered yeast, LB fatty acid composition shifted towards lower content of UFA relative to SFA in the highly lipidic strains and cytoplasmic carbohydrate stores were heavily reduced. Vibrational spectroscopy analysis of yeast cells revealed unprecedented information on the effectiveness and effects of metabolic engineering strategies for higher lipid content that will also guide future approaches to the field.
Yeast cell lines and culture conditions
Saccharomyces cerevisiae strain names and introduced genes
AtDGAT1 Tgl3Δ Ald6 SEACS L641P
AtDGAT1 Tgl3Δ Ald6 SEACS L641P ACC1 S659A, S1157A AtClo1
Saccharomyces cerevisiae were maintained based on their auxotrophy using yeast synthetic complete (SC) minimal medium, containing 6.7 g/L of yeast nitrogen base, 20 g/L glucose as well as a mixture containing appropriate nucleotide bases and amino acids for the various dropout options (SC-Leu, SC-His-Ura, SC-His-Leu-Ura). All strains were stored in 15% glycerol at − 80 °C before being cultured in 5 mL yeast SC minimal medium and incubated at 30 °C, 250 rpm, overnight. The culture OD600 nm was diluted to approximately 0.4 into 50 mL of SC induction medium containing 2% (w/v) galactose and 1% (w/v) raffinose in 250 mL flasks. The flasks were capped with aluminum foil and incubated at 30 °C, 250 rpm until cells were harvested at 72 h for the following analysis.
GC measurements of total fatty acid content
Harvested cells were pelleted by centrifugation at 3000 rpm for 5 min, frozen at − 80 °C for ~ 1 h and subsequently freeze-dried overnight using a FreeZone® 4.5 Liter Freeze Dry Systems (Labconco Corporation, USA) to obtain the dry cell weight (DCW) of each culture. Dry cells (~ 20 mg) were treated with 2 mL methanol/hydrochloric acid/chloroform (10:1:1, v/v/v) and heated at 90 °C for 1 h in sealed test tubes to convert lipids to fatty acid methyl ester (FAME). FAME was washed with 0.9% NaCl solution (1 mL) and extracted with hexane after mixing. FAME samples (1 µL) were analyzed by Agilent 7890A gas chromatography with flame ionization detection (GC-FID) as described previously .
Lipid body visualization using confocal fluorescence microscope
Imaging of lipid droplets after Nile red (Sigma-Aldrich, USA) staining of unfixed stationary phase yeast cells was undertaken 72 h after induction of gene expression [11, 24]. 1 mL of harvested cells were transferred into a 1.5 mL reaction tube and washed twice with 1 mL sterile 50 mM Tris–HCl (pH 7.5). 1 µL of Nile Red stock solution was added into the cell suspension to obtain the final concentration of 1 μg/mL, gently mixed, incubated for 20 min at room temperature and centrifuged at 1000g for 2 min. 1 μL of dense cell suspension was mounted on a standard microscope slide and imaged by a Leica Microsystems SP5 confocal microscope coupled with HCX PL APO 63×/1.4 OIL CS oil-immersion objective in Monash Micro Imaging, Monash University and the data collected by Leica LAS X (Leica Microsystems, Inc.) microscope control software.
Sample preparation for vibrational spectroscopy-based techniques
Yeast cells in phosphate buffered saline (PBS) were collected by centrifugation (1000 rpm, 5 min) and the pellet resuspended in 500 µL of ultrapure water, gently mixed and again centrifuged. This step was repeated 3 times to ensure removal of any residual PBS. The final pellet was resuspended in 500 µL of ultrapure water. For each yeast strain, 100 µL of the suspension was placed on each of two CaF2 slides and air-dried to obtain a dispersed layer of single cells. From each set of two samples, one was subjected to Raman measurements and the other was mounted on a flat magnetic stainless-steel substrate and designated for AFM–IR measurements. The remaining suspension (300 µL) was centrifuged and the pellet was placed directly on the Attenuated Total Reflection (ATR) crystal.
ATR-FTIR data were recorded using a Bruker Alpha FTIR (Ettlingen, Germany) spectrometer with an ATR sampling device containing a single bounce diamond internal reflection element and equipped with a globar source, KBr beam splitter and a deuterated triglycine sulfate detector. Spectra were recorded at a resolution of 6 cm−1 in the spectral range of 4000–900 cm−1. For each strain, 3 biological replicates were studied and for each of these, 3 technical replicates were recorded (nsingle_strain = 9, ntotal = 45). Background spectra were collected directly prior to each measurement (64 scans). After recording the background, 0.5 µL of yeast pellet was placed on the crystal and air-dried for approximately 10 min. Each spectrum was recorded using 64 co-added interferograms.
Raman measurements were collected using WITec confocal CRM alpha 300 Raman microscope (WITec, Ulm, Germany). The spectrometer was equipped with an air-cooled solid-state laser operating at 532 nm, a CCD detector, cooled to − 60 °C and 600 grooves/mm grating. The laser was coupled to the microscope via an optical fiber with a diameter of 50 µm. For data collection, a dry Olympus MPLAN (100×/0.90NA) objective was used. The monochromator of the spectrometer was calibrated using Raman scattering line produced by a silicon plate (520.5 cm−1). For each strain, 3 biological replicates were studied and for each of these replicates, 6 individual cells were mapped (nsingle_strain = 18, ntotal = 90). The size of mapped area was adapted individually, depending on the size of the cell. Data were collected in the spectral range of 3705–0 cm−1, with the spectral resolution of 3 cm−1. The integration time for a single spectrum was 0.1–0.3 s. Laser power was adjusted individually for each sample, not exceeding the range of 5–7 mW. Raman measurements and initial data analysis were performed using WITec software (WITec Plus, Ulm Germany). Raman images were constructed by integration of selected marker bands without any preprocessing. Cluster analysis was carried out after cosmic spike removal (CRR) and background subtraction (polynomial fit, 2nd order). The Raman data were analyzed with k-means Clustering (KMC) using the Manhattan distance and Ward’s algorithm.
AFM–IR measurements were performed with a NanoIR2 system (Anasys Instruments Inc., Santa Barbara, USA). The IR source was an optical parametric oscillator (OPO) laser, producing a 10 ns pulse at a 1 kHz repetition rate. A silicon cantilever (AppNano, Mountain View, CA 94043, USA) was used with nominal radius of 10 nm and a nominal spring constant of 0.5 N/m. The system was purged with N2 to control humidity. For each strain, 3 biological replicates were studied, with 6 single cells investigated for each biological replicate (nsingle_strain = 18, ntotal = 90). All single spectra were collected in the range of 1800–900 cm−1 with a spectral resolution of 8 cm−1 and IR maps at fixed wavenumber values, to investigate the distribution of selected components (specific wavenumber values are given in Results). Simultaneously to each IR map, the AFM height and deflection images were acquired. The maps were subsequently combined in MatLab (Mathworks, Natick, USA), PLS_toolbox (Eigenvector research, Manson, USA) and analyzed using k-means clustering to identify the presence and location of LBs. Following this analysis, single spectra were recorded from cytoplasm and LBs. All presented single spectra were normalized using the Standard Normal Variate (SNV) method and smoothed using the Savitzky-Golay algorithm with 13 smoothing points.
Absorbance data obtained for the engineered and control yeast strains were assessed for statistical significance by one way analysis of variance (ANOVA) at p < 0.01, α = 0.05; where significant differences were indicated, Student’s t test was applied the post hoc to data for engineered strains compared to control.
Results and discussion
Total fatty acid content: correlation between GC-FAME and ATR-FTIR spectroscopy
The first region (1800–1500 cm−1) includes the band at 1745 cm−1 (Fig. 1d, black arrow), attributed to stretching of C=O groups of lipids, whereas the second one (2800–3050 cm−1) contains a variety of bands originating from stretching of CH2 and CH3 groups in lipid chains . Increase in the intensity of the band at 1745 cm−1 as well as the bands in the high wavenumber region between different cell lines is clearly visible (Fig. 1d–g) and follows the same trend as an increase in the total fatty acid content assessed from GC (Fig. 1c). Calculation of the ratio of those bands to the amide I band (Fig. 1a, b) enabled the visualization of changes in the lipid to protein ratio  in the studied cell lines, as it represents the total fatty acid content in the dried mass of cells. A PLS regression model, built on the basis of 2nd derivatives of ATR-FTIR spectra in the high wavenumber region, enabled the prediction of total fatty acid content (as a percentage of dry cell weight). The correlation between predicted and assessed from GC total fatty acid content for the validation set, together with the parameters of the PLS regression model, is shown in Fig. 1h. This approach provides a fast and straightforward measure of the effectiveness of lipid metabolic engineering strategy. The 2nd derivatives of ATR-FTIR spectra revealed significant variation in the band at 1636 cm−1 (Fig. 1f) assigned to deformation vibration of water, most likely due to intracellular water . No correlation was observed between this and any of the lipid-related bands.
Our initial analysis of metabolically engineered cells undertaken on populations of whole cells addressed the basic question of lipid production between strains and demonstrated the power and rapid analysis of vibrational spectroscopy similar to that previously reported for naturally occurring yeasts by Ami et al. . However, to enable us to address questions relating to subcellular structures and their chemical components and in regard to cell-to-cell variability within metabolic engineered cells required spectroscopic approaches with greater spatial resolution power.
High-resolution CRS visualization and chemical analysis reveals subcellular structures and the impact of metabolic engineering
Lipid body characteristics and fatty acid composition of individual engineered yeast cells
Interestingly, LBs in HBY20 were only slightly increased in size (diameter 1.21 ± 0.40 µm) compared to control (diameter 0.87 ± 0.23 µm) (Fig. 5). For HBY31 strain, LBs were observed in all cells, however, in two types of arrangements. Some cells, similarly to HBY20, contained multiple LBs with a diameter of approximately 1 µm, but others contained a single large LB which filled the cell almost entirely (Fig. 5, HBY31 Panel G). This resulted in fewer but overall significantly larger LBs in HBY31 (2.14 ± 1.08 µm) compared to any other strain. HBY31 was the only strain assessed in this study that expressed caleosin (Table 1), a lipid droplet stabilizing protein originating from plants which may account for the larger LBs observed here.
The shift to higher saturated fatty acid content in the most highly engineered strains was not only due to the specificity of the introduced acyltransferase, DGAT1, for saturated fatty acids as the AtDGAT1 gene was expressed in all strains but maybe also due to limited quantities of unsaturated fatty acids available for conversion to triacylglycerol. ∆9-Unsaturated fatty acids are essential components of plasma membranes of yeast  and may be preferentially directed to these locations potentially resulting in more saturated fatty acids directed to triacylglycerol production through the enhanced expression of diacylglycerol acyltransferases.
Comprehensive mapping of cellular constituents by AFM–IR
To obtain a comprehensive measure of biochemical changes in cells resulting from induced genetic modifications, further analysis of the composition of cytoplasm and, in particular, carbohydrate content was required. The ability to simultaneously study the composition of intracellular structures without the need to isolate them is an advantage of high spatial imaging via vibrational spectroscopy. However, as described earlier, the cytoplasm of engineered yeast contains significant and varying amounts of heme in the form of clusters and dispersed within the cytoplasm (Fig. 4e, g, h-Brown). The raised and variable Raman background due to heme content effectively conceals less intense signals such as carbohydrate-related bands excited at lower wavelengths and therefore, IR-based spectroscopy is more useful to investigate carbohydrate components at high spatial resolution. However, as the spatial resolution of conventional IR imaging does not exceed ~ 5 µm, a more sophisticated approach based on AFM–IR imaging was applied in our case. In AFM–IR the phenomenon of IR absorption is not measured directly, but by measuring the thermal expansion of the sample resulting from application. Therefore, the spatial resolution of imaging with the use of AFM–IR is no longer limited to ~ 5 µm, but can be significantly smaller (~ 100 nm), enabling the measurement of spectra representing explicitly the composition of subcellular structures. AFM–IR mapping with selected bands corresponding to proteins, lipids and carbohydrates was performed on 6 replicate cells for each strain and the images were combined and analysed via cluster analysis to confirm presence (or absence) of LBs and determine their precise location within the cell. Subsequently, single spectra (5–10) were collected from areas corresponding to LBs and cytoplasm. A high prevalence of LBs in the HBY31 and HBY20 strains were observed, in agreement with results shown previously by other techniques applied in this study.
AFM–IR reveals cell-to-cell variability in concentrations of key metabolites
Vibrational spectroscopy analysis has previously been shown to be highly informative for measurement and location of target metabolites such as lipids in cultured cells of wild-type microalgae and yeast [17, 19, 32, 33, 34, 35, 36, 37, 38, 39, 40] and in selected studies that include heterologous expression of lipid stabilizing proteins . Our study is the first to our knowledge where the effects of the sequential introduction of a complex metabolic engineering pathway into an organism have been measured using vibrational spectroscopy techniques and have revealed valuable insights into areas for further improvement.
Here, we have used vibrational spectroscopy techniques to measure the impact of metabolic engineering approaches for high lipid production in yeast at both the single cell level and averaged and detailed information about the chemical composition of subcellular structures. The ratio of lipid-related bands (e.g., 1740 cm−1) to amide I was demonstrated to be useful as a quick marker of the total lipid content. An established PLS regression model allowed the successful prediction of total fatty acid content from the ATR-FTIR spectrum, demonstrating the ability of this technique to provide the same information as GC, but within few minutes and without the need for complex sample preparation. Fluorescence imaging enabled us to visualise LBs, confirming their highest prevalence and size in the HBY31 strain. Furthermore, two high spatial imaging techniques based on vibrational spectroscopies were applied (CRS, AFM–IR) and demonstrated substantial changes in intracellular composition with more complex metabolic engineering. CRS imaging provided an insight into LBs composition, revealing a significant decrease in the degree of unsaturation of lipids in the most highly engineered strain—HBY31. In addition, CRS imaging demonstrated significantly increased LB number in HBY20 and HBY31 strains together with a dramatic increase in LB size for HBY31 that expressed the LB stabilizing protein, caleosin. Finally, the AFM–IR imaging demonstrated large changes in the composition of cytoplasm between strains. A decrease in carbohydrate content with concurrent increase in lipid content of cytoplasm was observed, progressing from control through to HBY31 strains. The high concentration of lipid in the cytoplasm of HBY31, in particular, suggests lipid production rate in these engineered yeast is exceeding the rate of lipid sequestration in LBs which could lead to lipotoxicity. Additionally, the shift towards higher saturated fatty acids levels in stored lipids suggests a reduction in the availability of unsaturated fatty acids for TAG formation in the cells. Altogether, our results have demonstrated an increase in the total lipid content resulting from genetic modifications, with the multigene modification approach (Ald6, SEACS L641P , ACC1 S659A, S1157A , AtDGAT1, AtClo1 and Tgl3△) in the HBY31 cell line being most effective. The vibrational spectroscopy approach allowed the simultaneous measurement of intra-strain variability in metabolite production and impact on cellular structures from metabolic engineering.
KK collected ATR, Raman, AFM–IR, performed the data analysis, prepared the figures and drafted the manuscript. HP cultured the yeast cell lines, collected and analysed GC and fluorescence data and contributed to writing and correcting the manuscript. BR contributed to the experimental planning and corrected the manuscript. VH contributed to the design of the experiment, data interpretation and contributed to the writing of the manuscript. All authors read and approved the final manuscript.
We thank Mr. Finlay Shanks for technical support of the instrumentals and are grateful to Ms. Yanqin Xu and Prof. Hongyuan Yang for the gift of the strains BY4741 and BY4741 Tgl3∆.
The authors declare that they have no competing interests.
Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Consent for publication
Ethics approval and consent to participate
We acknowledge the financial support from ARC linkage Project (LE160100185) and Monash University for the graduate and international postgraduate research scholarships awarded to HP. B.R.W. is supported by an Australian Research Council (ARC) Future Fellowship Grant FT120100926.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 11.Peng H, Moghaddam L, Brinin A, Williams B, Mundree S, Haritos VS. Functional assessment of plant and microalgal lipid pathway genes in yeast to enhance microbial industrial oil production. Biotechnol Appl Biochem 2017:1–7. https://doi.org/10.1002/bab.1573.
- 23.Peng; H, He; L, Haritos VS: Impact of yeast lipid pathway engineering and bioprocess strategy on cellular physiology and lipid content. J Ind Microbiol Biotechnol. 2018 (under review).Google Scholar
- 24.Wolinski H, Kohlwein SD. Microscopic and spectroscopic techniques to investigate lipid droplet formation and turnover in yeast. Membrane Trafficking: Second Edition; 2015. p. 289–305.Google Scholar
- 31.Stukey JE, McDonough VM, Martin CE. The OLE1 gene of Saccharomyces cerevisiae encodes the delta 9 fatty acid desaturase and can be functionally replaced by the rat stearoyl-CoA desaturase gene. J Biol Chem. 1990;265(33):20144–9.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.