Raman spectroscopy as a tool for tracking cyclopropane fatty acids in genetically engineered Saccharomyces cerevisiae

Kamila Kochan a, Huadong Peng b, Eunice S. H. Gwee ac, Ekaterina Izgorodina c, Victoria Haritos *b and Bayden R. Wood *a
aCentre for Biospectroscopy, School of Chemistry, Monash University, Clayton Campus, 3800, Victoria, Australia. E-mail: bayden.wood@monash.edu
bDepartment of Chemical Engineering, Monash University, Clayton Campus, 3800, Victoria, Australia. E-mail: victoria.haritos@monash.edu
cSchool of Chemistry, Monash University, 17 Rainforest Walk, Clayton, VIC 3800, Australia. E-mail: katya.pas@monash.edu

Received 2nd August 2018 , Accepted 2nd September 2018

First published on 11th September 2018


Cyclopropane fatty acids (CFAs) are a group of lipids with unique physical and chemical properties between those of saturated and monounsaturated fatty acids. The distinctive physicochemical characteristics of CFAs (e.g. oxidative stability, self-polymerization at high temperatures, etc.) results from the presence of a cyclopropane ring within their structure making them highly useful in industrial applications. CFAs are present in several species of plants and bacteria and are typically detected with standard lipid profiling techniques, such as gas or liquid chromatography. In this work we investigated several strains of S. cerevisiae, genetically modified to introduce the production of CFAs, in comparison to control strain using confocal Raman spectroscopy (CRS). The aim of our work was to demonstrate the potential of CRS not only to detect changes introduced due to the CFAs presence, but also to track CFAs within the cells. We present for the first time Raman and IR spectra of CFA standard (cis-9,10-methyleneoctadecanoic acid), completed with quantum chemical calculations and band assignment. We identified marker bands of CFA (e.g. 2992, 1222, 942 cm−1) attributed to the vibrations of the cyclopropyl ring. Furthermore, we analysed lipid bodies (LBs) from modified and control yeast using CRS imaging and identified multiple changes in size, number and composition of LBs from engineered strains. We observed a significant reduction in the degree of unsaturation of LBs using the ratio of bands located at 1660 cm−1 (ν(C[double bond, length as m-dash]C)) and 1448 cm−1 (δ(CH2)) in the modified cell lines. In addition, we were able to detect the presence of CFAs in LBs, using the established marker bands. CRS shows tremendous potential as technique to identify CFAs in lipid bodies providing a new way to track lipid production in genetically modified single yeast cells.


Introduction

The emerging interest in the development of new engineering strategies aimed at enhancing lipid production in yeast is dictated by their potential for industrial applications. Amongst those, the most commonly mentioned is biodiesel production.1–3 Biodiesel is an alternative, biodegradable and renewable energy source produced from triacylglycerols (TAGs), obtained from animal, plant or microbial oils.1–3 The latter is considered the most promising TAG source due to the low cost of production, short life cycle of the microorganism, independence from the climate, season, etc.1–3 A realistic alternative fuel should not only produce significant environmental and economic benefits, but also be amenable to large scale production and provide a net energy gain over the energy used for production.2 Consequently, special attention is devoted to microorganisms capable of accumulating significant amounts of lipids.2 Oleaginous microorganisms are a group of microorganisms, including microalgae, fungi, yeast and bacteria, characterised by the content of lipid being in excess of 20% in biomass weight.1–3 However, the amount of lipids accumulated by oleaginous organisms can be substantially higher.1–3 In fact, some microbes (such as Rhodosporidium sp., Rhodotorula sp. and Lipomyces sp. ) have enormous capabilities of accumulating intracellular lipids, with some producing 70% of their total biomass dry weight.2,4,5 The high yield of TAG production by oleaginous microorganisms is very beneficial for industrial applications (including, but not limited to, biodiesel production) and can be achieved via genetic engineering. This has resulted in growing interest in various engineering strategies aiming at enhancing TAG production and accumulation in microorganisms. The detailed lipid content and fatty acid profile differs between species but most produce TAGs formed from several saturated and unsaturated fatty acid chains.6

Cyclopropane fatty acids (CFAs) are a subgroup of fatty acids, containing a cyclopropane ring within their structure. They occur infrequently in plants (such as gymnosperms, Malvales, Litchi and other Sapindales)7,8 and bacteria (such as Escherichia coli, Streptococcus sp. and Salmonella sp.).9,10 CFAs are derived from corresponding unsaturated fatty acids (UFAs) and are formed through cyclopropanation, catalysed by cyclopropane synthases (CPSs).7,9,10 In bacteria, CFAs occur in the cell membranes, with the vast majority in the cis form.10 However, the presence of the trans form was also detected in the cell envelope of Mycobacterium tuberculosis, where it was involved in the regulation of Mycobacterium virulence.11 In general, CFAs are considered to enhance the chemical and physical stability of membranes.9 It was suggested that the CFAs reduce the fluidity of membranes, therefore reducing their permeability.9,12 The presence of CFAs was also demonstrated to increase the tightness of packing within lipid bilayers.13,14 In addition, lipid bilayers containing CFAs compared to lipid bilayers containing UFAs, exhibit a greater chain order.14 The influence of CFAs presence on the properties of the membrane was postulated to result from the cyclopropane moiety acting as a barrier preventing the propagation of motion from one chain to another.15

The presence of the ring within the CFAs structures translates into their unique physical and chemical properties, placed in between those of saturated (SFAs) and monounsaturated fatty acids (MUFAs).7 For instance, hydrogenation of CFAs results in ring opening and formation of methyl-branched fatty acids, exhibiting low temperature properties similar to MUFAs, while at the same time demonstrating the oxidative stability characteristic for SFAs (non-susceptibility for oxidation).7,9 The unique physical and chemical properties of CFAs make them highly desirable for industrial applications (production of lubricants, coatings, polymers, etc.).7 However, currently none of the natural sources of CPSs are suitable for commercial production.8

The lipid composition of cells can be studied using various experimental techniques. Traditionally, the most widely used include gas chromatography (GC) and fluorescence staining. GC enables one to obtain the total lipid profile for whole cells, however, without any information upon their spatial distribution. Fluorescence staining, on the other hand, allows one to visualise LBs, but does not provide information about their composition.16 Both methods, although very useful, are burdened by the lack of ability to provide spatially localised chemical information. This gap was recently fulfilled by microscale imaging via molecular-based techniques such as Raman spectroscopy (RS). RS in a non-invasive and non-destructive technique requiring minimal sample preparation and enabling in situ, online analysis. In recent years imaging by confocal Raman spectroscopy (CRS) has been broadly applied to various cells, including red blood cells,17 white blood cells,18 endothelial cells,19 hepatocytes,20 algae,21 yeast16,22 and many others.23

Although, RS provides the ability to simultaneously obtain information about multiple chemical components (e.g. pigments, proteins, carbohydrates, etc.), it is particularly suitable to study lipids24 because lipids contain multiple non-polar groups (C–C, C[double bond, length as m-dash]C), making them strong Raman scatterers.24 Primarily, RS offers the possibility to identify various groups of lipids (free fatty acids, TAGs, phospholipids, cholesterol, etc.) and discriminate between them.25 This enables the differentiation between esterified and non-esterified forms via the presence of the ester carbonyl band located at ∼1740 cm−1 (ν(C[double bond, length as m-dash]O)), enabling one to determine the presence of TAGs.16,20 Furthermore, it allows the discrimination between UFAs and SFAs25,26 and can be used to determine their relative content (using ratios of UFA-related bands located at ∼3012 cm−1, ∼1656 cm−1, ∼1301 cm−1 and SFA-related bands located at ∼2855 cm−1, ∼1444 cm−1, ∼1266 cm−1).27–29 Raman spectra of various, biologically relevant lipids have been demonstrated in several reviews.25,26,30 However, hitherto there are no experimental or theoretical spectra of CFAs available in the literature.

In our previous work, we demonstrated several single- and multigene engineering approaches enabling the increase in amount of intracellular lipids (TAGs) in Saccharomyces cerevisiae.16,31 In the next stage, we focused on investigating whether those strategies can also be applied for CFAs, which are highly valued in industry. Currently, the detection of CFAs is based on either liquid32 or gas chromatography,33 although recently a 1H NMR approach was shown.34 Here, we demonstrate for the first time the application of CRS imaging for determining the presence of CFAs in several engineered Saccharomyces cerevisiae strains. We first obtained spectra (Raman and IR) of the CFA standard and assigned the bands via theoretical calculations. Subsequently, we performed CRS imaging of single cells from various strains to determine the presence, composition and heterogeneity of LBs. Finally, we investigated the presence of CFAs in the engineered cell lines by comparing Raman spectra of the lipid bodies with the CFA standard and using Partial Least Square Discriminant Analysis (PLS-DA) to predict lipid body composition.

Materials and methods

Yeast cell lines and culture condition

Five engineered strains of S. cerevisiae (CP1, CP4, CP5, CP6, CP7) in comparison to a control strain (CON) were compared. Details of all strains are given in Table 1. Cyclopropane fatty acid synthase from E. coli (Ec.CFAS) was regulated by promoter GAL1. The details of other genes including Ald6, SeACSL641P, AtDGAT1, AtClo1 can be found in our previous paper.31 Based on the auxotrophy difference (SC-Leu, SC-His-Ura, SC-His-Leu-Ura), the engineered yeast strains were maintained using synthetic complete (SC) minimal medium. Then, induced by galactose and incubated at 30 °C, 250 rpm in 250 mL flasks until cells were harvested at 72 h for the following analysis. Detailed cell culture conditions have previously been described.16,31
Table 1 Engineered Saccharomyces cerevisiae strain names and introduced genes
Strain Description
a pSP-GM2, pIYC04, pESC-leu2d. Δ indicates endogenous gene was knocked out.
CON BY4741 – empty vectorsa
CP1 BY4741 – Ec.CFAS
CP4 BY4741 – Ec.CFASAtDGAT1
CP5 BY4741 – Ec.CFASAtDGAT1ΔTgl3
CP6 BY4741 – Ec.CFASAtDGAT1ΔTgl3Ald6SeACSL641P
CP7 BY4741 – Ec.CFASAtDGAT1ΔTgl3Ald6SeACSL641PAtclo1


Lipid analysis and quantification by gas chromatography (GC)

Harvested cell pellets were freeze-dried overnight to obtain the dry cell weight (DCW) of each culture. Slightly modified Bligh Dyer extraction procedure35 was used to extract yeast lipids, then dissolved in the chloroform layer, which was loaded on the thin layer chromatography (TLC) plate to separate the phospholipids and triacylglycerols (TAGs).36 Then, scraped TAG and phospholipids silica spots from TLC plates were methylated to form the fatty acid methyl esters (FAME), which were quantified by GC as described previously.37

Sample preparation for Raman spectroscopy

Yeast cells in Phosphate Buffer Solution (PBS) were centrifuged (1000 rpm, 5 min) in order to obtain a pellet. The supernatant was then removed; the pellet was re-suspended in 500 μL of ultrapure water, vortexed for 2 min and centrifuged again. This step was repeated three times to ensure complete removal of any residual PBS. The final pellet was resuspended in 500 μL of ultrapure water and gently mixed. 100 μL of the yeast solution was placed on a Raman grade CaF2 window (Crystran Pty. Ltd, UK) and air-dried. From each cell line, at least three independent samples were prepared. The cyclopropane fatty acid (CFA) standard: cis-9,10-methyleneoctadecanoic acid (CycC19) and fatty acid standard: stearic acid (SA) were purchased from Sigma Aldrich and prepared directly prior to measurement by placing them on a Raman grade CaF2 slide (both in solid state).

Raman spectroscopy of yeast cells

Raman spectroscopy measurements were performed using WITec confocal CRM alpha 300 Raman microscope, equipped with a CCD detector (cooled to −60 °C), 600 grooves per mm grating and an air-cooled solid-state laser operating at 532 nm, coupled to the microscope with an optical fibre with a diameter of 50 μm. All data was collected using a dry Olympus MPLAN (100×/0.90 NA) objective. Prior to data collection, the monochromator of the spectrometer was calibrated using Raman scattering line produced by a silicon plate (520.5 cm−1). The size of mapped area was adjusted individually, depending on the cell size, with sampling step of 0.3 μm. The laser power was 7 mW. All spectra were collected in the spectral range 0–3725 cm−1 and with spectral resolution of 3 cm−1. The integration time for each spectrum was 0.5 s.

Vibrational spectroscopy measurements of the CFA standard

Raman measurements of the CFA and fatty acid standards (CycC19 & SA) were performed using the system described in the previous section. RS spectrum of each standard was obtained by averaging three individual spectra, each collected in the spectral range of 0–3725 cm−1, with spectral resolution of 3 cm−1, integration time of 1 s and 100 accumulations. The Attenuated Total Reflectance-Fourier Transform Infrared (ATR-FTIR) spectra of CFA standard were recorded using a Bruker Alpha FTIR (Ettlingen, Germany) spectrometer with a globar source, KBr beam splitter, a deuterated triglycine sulfate (DTGS) detector and an Attenuated Total Reflection (ATR) sampling device containing a single bounce diamond internal reflection element (IRE). The ATR-FTIR spectrum of the CFA standard was obtained by averaging three individual spectra, each collected in the spectral range of 600–4000 cm−1 with spectral resolution of 8 cm−1 and accumulation of 64 interferograms. The collection of each ATR-FTIR spectrum was preceded by collection of background (128 interferograms).

Data analysis

Directly after collection, all Raman spectra were subjected to cosmic spike removal (CRR)(filter size: 2, dynamic factor: 4). The ATR-FTIR spectrum of CFA was smoothed using Savitzky–Golay algorithm with 9 smoothing points (polynomial fit 2). The quantum chemical calculations were performed using ethyl and butyl groups as side chains, two different functionals (M06-2X and wB97XD) and two different basis sets (cc-pVDZ and aug-cc-pVDZ). The high wavenumber region of the spectra was scaled using scaling factors (ESI, Table 1) because the calculations assume harmonic oscillations. The maps of distribution of components were created by displaying the integrated intensity of selected Raman bands in the ranges: 3050–2800 cm−1 (organic components) and 1460–1420 cm−1 (lipids). To obtain the average LB spectrum for each cell line the following procedure was implemented. For each LB from each measured cell an average LB spectrum was obtained by k-means cluster analysis (KMC) (spectral range: 1800–600 cm−1) and extracted. Subsequently, all spectra of LBs originating from cells of selected cell lines were averaged. For PLS-DA, 630 spectra extracted from LBs of all cell lines were used (nCON = 274, nCFAs = 356). Spectra from LBs from all modified cell lines (CP1, CP4, CP5, CP6, and CP7) were assigned to CFA-containing group. The dataset was randomly divided into calibration and validation subsets, with 66% of spectra kept in the calibration subset. PLS-DA was performed on 2nd derivatives of RS spectra from LBs. Prior to analysis spectra were smoothed, using Savitzky–Golay algorithm with 11 smoothing points (polynomial fit 2) and normalised using the Standard Normal Variate (SNV) method. 2nd derivatives were calculated using 11 smoothing points (Savitzky–Golay algorithm) and mean centred before PLS-DA. The presented PLS-DA model was created using the spectral ranges of 3100–2800 cm−1 and 1800–650 cm−1 and 5 latent variables.

Results and discussion

Vibrational spectroscopic characterisation of CFA standard

First, we collected Raman spectra of a CFA standard: cis-9,10-methyleneoctadecanoic acid and of a fatty acid standard: stearic acid (C18:0). A direct comparison between the average spectra of those compounds is presented in Fig. 1 (1800–200 cm−1) and Fig. 2 (3200–2800 cm−1). The RS spectrum of the CFA standard shows bands typical for vibrations of FA chains along with additional modes, originating from vibrations of the cyclopropane ring. The bands associated with vibrations related to FA chain are marked with dashed lines in Fig. 1 and 2. These include bands located at: 2882 cm−1, 2846 cm−1, 1466 cm−1, 1442 cm−1, 1298 cm−1, etc., originating from stretching and deformation modes of CH2 & CH3 groups.25,26,30 The bands located between 1100–1000 cm−1 (1173, 1127, 1104, 1065 cm−1) are associated with C–C stretching modes and were also reported for saturated FAs.25 Minor shifts in the bands related to FA chain between SA and CFA spectrum can be observed. The relative intensity of various bands also differs between SA and CFA spectra (particularly in the region 1200–1000 cm−1). Despite this, the presence and compatibility of the aforementioned bands in SA and CFA spectra indicates that they originate from vibrations of the FA chain. Vibrations associated with the functional groups within the cyclopropane ring most probably also contribute to those bands, although none of them is specific for the cyclopropane ring, as all were previously reported for chain FAs.25,26,30 However, the CFA spectrum compared to SA spectrum contains multiple additional bands (Fig. 1 and 2, red background), located in the high wavenumber region (3071, 2992 cm−1) and in the fingerprint region (1222 cm−1, 942 cm−1 and several low intensity bands between 830–780 cm−1), absent in Raman spectrum of SA (Fig. 1 and 2), along with spectra of other chain FAs.25
image file: c8an01477a-f1.tif
Fig. 1 (A) Raman spectra (in the range 1800–200 cm−1) of the CFA standard (bottom) and stearic acid (C18:0) (top) together with (B) their 2nd derivatives. Corresponding bands in both spectra, associated with FA chain vibrations, are marked with black dashed lines in (A). Red background in (A) and (B) marks the region where bands are present in the Raman spectrum of CFA, but absent in the Raman spectrum of stearic acid. The precise band positions are given in (B).

image file: c8an01477a-f2.tif
Fig. 2 (A) Raman spectra (in the range 3200–2800 cm−1) of the CFA standard (bottom) and stearic acid (C18:0) (top) together with (B) their 2nd derivatives. Corresponding bands in both spectra, associated with vibrations of FA chain, are marked with black dashed lines in (A). Red background in (A) and (B) marks the region with bands present in the Raman spectrum of CFA, but absent in the Raman spectrum of stearic acid. The precise band positions are given in (B).

To confirm the contribution of the ring vibrations to the aforementioned bands and their assignment, we performed Density Functional Theory (DFT) calculations of two model systems, 3-(2-ethylcyclopropyl)propanoic acid and 5-(2-butylcyclopropyl)pentanoic acid using M06-2X38 and ωB97XD39 functionals combined with cc-pVDZ and aug-cc-pVDZ basis sets.40 We used two different lengths of a side chain to further study the effect of the side chain on the cyclopropyl ring vibrations. The predicted Raman spectra are presented in Fig. 3 and their detailed band assignment is presented in ESI (see Table S1). The model system with the ethyl chain is given in Fig. 3A and B, whereas the model with the butyl chain in Fig. 3C and D. The fingerprint region for both models (normalized to one) is presented in Fig. 3B for ethyl and 3D for butyl. Initially, it appears that the presence of the ethyl groups resulted in spectra exhibiting more bands than in the case of the butyl groups. However, a similar number of bands are present in both spectra. The main difference between the spectra came from the relative intensity of some bands. Since the butyl group is longer than the ethyl one by 2 carbon atoms, Raman spectra of the butyl model showed higher intensity of bands originating from vibrations of the side chains compared to those originating from the cyclopropyl ring. The predicted spectra above 2800 cm−1 were scaled to reproduce the 2847.27 cm−1 band that is assigned experimentally to νs(–CH2). The scaling parameters are included in the ESI. The rest of the spectra were left unscaled. Analysis of the predicted spectra for the two DFT functionals and four different basis sets revealed small variations in the bands. It was observed that the complexity of vibrations increased with the length of the alkyl chain, giving rise to an increase in coupled vibrations and a decrease in pure vibrations. The longer butyl alkyl chain appeared to also influence intensities of the predicted bands. Upon comparison, both ethyl and butyl side chains predicted similar assignment of the fingerprint region, which confirms that the alkyl chain length has a minimal effect on the predicted Raman spectra. Furthermore, in the text we only discuss results obtained with ωB97XD/aug-cc-pVDZ for the butyl model system. Experimentally, there are 3 distinct peaks assigned to bands arising from the cyclopropyl ring – 942 cm−1, 1222 cm−1 and 2992 cm−1 and this assignment agrees well with the predicted spectra. 942 cm−1 (calculated to be at 958 cm−1) corresponds to α(–C–C–C– (ring)), whereas the bands at 1222 cm−1 (calculated to 1226 cm−1) can be confidently assigned to a combination of stretching and bending vibrations of the ring (labelled as νs(–C–C–C– (ring)) + τ(–CH2(ring)) + α(H–C–C–H (ring) in Table S1). The band at 2992 cm−1 (calculated to be 2958 cm−1 after scaling) was assigned to a combination of stretching vibrations such as νs(–CH2(ring)) and ν(–CH(ring)). To this end, the calculated spectra, regardless of level of theory chosen in this study and length of alkyl side chain, were able to reliably predict and assign characteristic vibrations of the cyclopropyl ring, confirming the identified CFA marker bands above (with the main ones located at 1222 and 2992 cm−1).


image file: c8an01477a-f3.tif
Fig. 3 A comparison of the experimental Raman spectrum of CFA standard with calculated Raman spectra using varying length of side chains: (A, B) ethyl and (C, D) butyl groups. For each length of side chains spectra were calculated using combinations of functionals: M06-2X, wB97XD and basic sets: cc-pVDZ, aug-cc-pVDZ, as annotated on the figure. Spectra are presented in spectral ranges: (A, C) 3100–2800 cm−1 and (B, D) 1800–400 cm−1. The scaling factors for each calculated spectrum are given in ESI, Table S1. All spectra in (A, C) were normalized to 1 in the range 3100–2800 cm−1. All spectra in (B, D) were normalised to 1 in the range 1800–400 cm−1. Grey background highlights the spectral regions of: (A, C) 3010–2980 cm−1, (B, D) 1260–1200 cm−1 and 960–920 cm−1. Black stars mark the band located at 2992 cm−1 in the experimental spectrum.

Raman spectral characteristics of the CFA standard (experimental spectrum as well as theoretical calculations) was additionally supplemented with results from a complementary technique – ATR-FTIR. The ATR-FTIR spectrum of a CFA standard, theoretical calculations using the same approach as for RS spectra and a table with band assignments are presented in ESI (Fig. S1, S2 and Table S2).

Multigene modification of yeast leads to significant increase in number of LBs with heterogeneous size

In the next step, we collected Raman maps of single cells (n = 126) from all studied strains (n = 6). Raman mapping enabled us to quickly visualise the cell area, by integrating the area under the bands in the region 3050–2800 cm−1 (corresponding to symmetric and asymmetric stretching of C–H groups from all organic components) along with the presence of LBs (by integrating the area under the lipid-originating band e.g. at ∼1444 cm−1). Representative results obtained for three cells from each strain are presented in Fig. 4 (cell area together with LBs). The control cells rarely demonstrated LBs and if so – they were usually rather small (<1 μm) (Fig. 4: 1b). A similar situation was observed for the CP1 strain, with only a slight increase in the number of cells containing single LBs only. Of note is that the CP1 strain was modified only to induce the production of CFAs, however, without any modifications aimed at increasing the overall lipid yield. For the CP4 and CP5 strains an increase in the prevalence of LBs can be observed. Even though they remain rather small (majority with a diameter ≤1 μm), they began to appear in multiple numbers in one cell (Fig. 4: 8b, 10b). A clear and significant increase in the prevalence of LBs is visible for CP6 and CP7 (Fig. 4: 13–18). Overall, LBs of CP6 and CP7 strains are larger than in other cell lines (e.g.Fig. 4: 13b and 16b), however, at the same time they demonstrate a substantial heterogeneity in size (e.g.Fig. 4: 15b), even within a single cell.
image file: c8an01477a-f4.tif
Fig. 4 Maps of distribution of selected components for three representative cells from each cell line obtained by integration the area under the Raman bands in the range: (a) 3050–2800 cm−1 (representing the area of the cell) and (b) 1460–1420 cm−1 (representing the lipids and visualising the LBs). Each row contains cells from different cell line (CON, CP1, CP4, CP5, CP6, CP7). Each number (1–18) corresponds to a different cell. For each cell a map of distribution of organic components (visualising the cell area and labelled ‘a’) and of lipids (visualising LBs and labelled ‘b’) is presented. The colour scale for each map was adjusted individually to visualise LBs and is presented below (values given in CCDs).

Fig. 5 shows the results of analysis of all collected Raman maps. The percentage of cells with LBs (Fig. 5A) demonstrates the prevalence of LBs in each cell line, whereas the ratio of the number of LBs to the number of cells with LBs for each cell line (Fig. 5B) shows the tendency to occur as individual or several LBs per cell. The value equal to 1 indicates the presence of only single LB per cell, whereas the value above 1 shows the presence of multiple LBs per cell. The CP1 strain exhibits similar characteristic to the control, with only slightly increased prevalence of LBs (Fig. 5A). In both cases, LBs always occur individually per cell (Fig. 5B). In addition, for all those strains (CP4, CP5, CP6 and CP7) the ratio of the number of LBs to the number of cells containing LBs exceeds 1, indicating that for each strain it was possible to identify cells with multiple LBs (Fig. 5B). This ratio is higher for the strains CP4, CP5, CP6 than CP7, as in those strains at least half of all measured cells contained min. 2 LBs. Particularly for CP6, LBs often were heterogeneous in size. In case of the CP7, the number of cells with ≥2 LBs was smaller than in CP4, CP5, and CP6, although such cells were still observed. Most of the LBs in CP7 were large (diameter >1 μm), often with a single LB filling most of a cell. Altogether, these results demonstrate that a single gene modification, aimed at inducing the production of CFAs (CP1 strain) did not influence the overall LB formation. At the same time, the multi-gene-based engineering strategies (CP6, CP7 strains) were shown to remain effective in enhancing lipid production, even when combined with modification aimed at introduction of CFA production. The excessive lipids were stored in the form of LBs, however, demonstrated significant heterogeneity in their size not only within the cell line, but even within individual cells.


image file: c8an01477a-f5.tif
Fig. 5 (A) Bar chart showing the percentage of all measured via CRS cells that showed the presence of LBs for each cell line. (B) Bar chart showing the ratio of the number of LBs to the number of cells demonstrating the presence of LBs for each cell line (on the basis of CRS measurements). Ratio equal to 1 indicates that for all cells containing a LB, only single LBs per cell were present. Ratio above 1 indicates the presence of multiple LBs in single cells.

All engineered strains exhibit LBs with altered UFA/SFA ratio

The average spectra of LBs (together with their standard deviation) from all modified cell lines in comparison to control are presented in Fig. 6 (RS spectrum of LBs with marked all band positions is presented in ESI, Fig. S3). In all cases, spectra show a typical lipid profile, indicating the presence of TAGs, containing both UFA and SFA chains. The high lipid contribution to the spectra of LBs from all cell lines is also clearly visible via the high wavenumber region (3100–2800 cm−1), particularly through the presence of a band at 2855 cm−1 (ν(CH2)) (ESI, Fig. S4). The presence of TAGs is confirmed by the band, located at 1747 cm−1 and assigned to ester carbonyl stretching.25 The SFAs are primarily manifested in the spectrum by the presence of bands at ∼1448 cm−1 (δ(CH2)) and ∼1308 cm−1 (τ(CH)) (Fig. 6) along with the aforementioned band at 2855 cm−1 (ESI, Fig. S4). The main evidence of the UFA presence is the band located at 1660 cm−1 originating from C[double bond, length as m-dash]C stretching (Fig. 6, marked with a star). The second typical UFA band in the fingerprint region is located around 1270 cm−1 (δ([double bond, length as m-dash]C–H)) (Fig. 6, marked with an arrow). This band is a part of a characteristic doublet (∼1308 cm−1 and ∼1270 cm−1) observed for mixtures of UFA and SFA.25,26,30 Although, this doublet might be used as a measure of degree of unsaturation for lipid mixtures,41 in case of biological materials its usefulness is limited because both bands are often not well resolved and strongly affected by baseline changes and the presence of compounds such as haem (inducing a photothermal raise of background).42 Even though haem itself is not located within LBs, its presence around them may affect the average spectra of LBs. This is particularly the case for small LBs (diameter ≤ 1 μm) because of spatial resolution depth probed by CRS. The haem from outside of small LBs may still contribute to their spectra. Some contribution from haem can be seen e.g. in the average RS spectrum of LBs from control (Fig. 6 and Fig S3) through the presence of bands located at 757 cm−1 (ν(pyr breathing); ν(CαCβ)), 1127 cm−1 (ν(C–C)) and 1588 cm−1 (ν(CαCm)asym).
image file: c8an01477a-f6.tif
Fig. 6 Average Raman spectra (in the range 1800–600 cm−1) of LBs from each cell line (black line, in each panel) together with standard deviation (grey background, in each panel). Each spectrum was obtained by averaging all average spectra of LBs originating from the selected cell line (more details are given in the ‘Materials and methods’ section). The black star marks the position of the band originating from ν(C[double bond, length as m-dash]C), located at 1660 cm−1 and the black arrow marks the band located at 1270 cm−1 ((δ([double bond, length as m-dash]C–H)). All spectra were normalized to the band located at 1448 cm−1 (δ(CH2) marker band for SFA). Blue dashed line marks the height of the band in control. The red dashed line marks the position of band in all modified lines.

A comparison between RS spectra of LBs from all studied strains shows a clear difference between control (CON) and all modified strains. The relative intensity of the UFA-related bands (1270, 1660, and 3009 cm−1) to the SFA-related bands (1308, 1448, and 2855 cm−1) is significantly higher for the control strain.

This is particularly visible using the set of bands at 1660 and 1448 cm−1. Fig. 6 presents spectra normalized to the lateral, with the blue dashed line indicating the height of the band at 1660 cm−1 in LBs from control strain and the red dashed line – height of this band in all modified strains. Interestingly, the UFA-related band from stretching of C[double bond, length as m-dash]C (1660 cm−1) shows similar height in all modified cell lines, with only minor variations (within the SD). Higher UFA chains content in the LBs of the control strain manifest itself in spectra also via the band at 1270 cm−1. Although it shows low intensity, it is still noticeable for the LBs from CON strain (Fig. 6, black arrow). For all modified strains however, even though some increase in intensity of Raman signal in this region can be observed, it remains clearly smaller than in LBs from CON line and with no distinct peak. A similar situation is observed for the band at 3009 cm−1 (ESI, Fig. S4) with the band being clearly more pronounced for LBs of CON strain compared to LBs from the modified strain. These results remain in good correlation with the FA profile obtained via GC (ESI, Table S3), clearly indicating a substantially higher UFA/SFA ratio for control compared to all modified strains (similar between each other). It is important to underline that the lipid profile obtained by GC reflects the overall lipid composition (including e.g. membrane lipids). Therefore, for cell lines with high LB content the GC results will reflect primarily the LBs composition and very likely correspond to it with high accuracy. However, for cell lines with low content of LBs (such as CON), the composition of LBs may have a limited impact on the overall lipid composition. Although the overall lipid composition from GC for such cell lines (low content of LBs) may still be indicative of LBs composition (e.g. point towards higher UFA/SFA ratio, as in this case), it does not necessarily represent it exactly. Therefore, the exact lipid composition estimated via GC for cell lines with low LB content does not necessarily correspond exactly to the composition of their LBs. However, GC results, accurate with the respect to whole cells and indicative for LBs, combined with high spatial resolution CRS aimed at LBs explicitly, together provide a strong proof of a change in the composition of LBs resulting from the introduced modifications. In all modified strains, despite the influence of the introduced modifications on total lipid content along with the number and size of LBs results in a significant reduction in the ratio of UFA to SFA chains. Interestingly, the ratio of UFA/SFA chains remained similar in LBs for all modified strains, independently of the engineering strategies (single, multigene) and their efficiency towards increasing the total lipid content. This change in the ratio of UFA/SFA chains might therefore be attributed to the introduction of the cyclopropane fatty acid synthase from E. coli, as this modification was shared for all lines. Furthermore, this modification aimed at inducing the production of CFAs, deriving from UFA through cyclopropanation.

RS enables the detection CFAs in LBs

GC results confirmed the presence of CFAs in all modified strains (ESI, Table S3). However, as mentioned before they reflect the overall composition of cells, without providing any specific information about the localisation of chosen lipids. The location of CFAs within the yeast cell is important, as for industrial applications the desired position for CFAs are LBs. However, at the same time CFAs are known to be associated mainly with membranes.9–14 In order to confirm that the CFAs produced by engineered cell lines are in LBs and are not only within the membranes, we isolated and separated the TAGs and phospholipids (PLs) and subsequently determined the composition of both fractions using GC. The results (ESI, Table S4) confirmed the presence of CFAs in both fractions, indicating that at least a part of produced CFAs is associated with TAGs (and therefore – with high probability present in LBs). We further investigated whether it is possible to detect the presence of CFAs in the LBs of modified cell lines using CRS. For this purpose we used a set of 630 spectra (nCFAs = 356, nCON = 274) extracted from LBs of all cell lines and analysed them using Partial Least Square Discriminant Analysis (PLS-DA). As the CFAs were detected in TAG fractions of all modified cell lines (ESI, Table S4), LBs from all engineered cell lines were assigned as containing CFAs. The results of the PLS-DA are presented in Fig. 7.
image file: c8an01477a-f7.tif
Fig. 7 Results of PLS-DA on a set of 630 Raman spectra (nCFAs = 356, nCON = 274) extracted from LBs of S. cerevisiae cells, genetically engineered to produce CFAs (marked with red diamonds and red lines) as well as S. cerevisiae control cell line (marked with black triangles and black lines). (A) Predicted class membership based on RS spectra and PLS-DA model for validation set together with regression vectors for (B) control and (C) CFAs-containing LBs with most prominent bands indicated. The stars in (C) mark bands discriminative for the CFA-containing group (in reference to control), but not associated with the differences in UFA/SFA ratio between both groups. PLS-DA was performed on 2nd derivatives of spectra after pre-processing (described in the ‘Materials and methods’ section) using the spectral ranges: 3100–2800 and 1800–650 cm−1. The average 2nd derivatives of Raman spectra of both groups (control vs. CFAs) are presented in (D). Black arrows mark the bands associated with the differences in UFA/SFA ratio (2855 and 1660 cm−1). The spectral region between 1280 and 1200 cm−1 was magnified to visualise the differences in this region.

The PLS-DA model was able to discriminate between CON and CFA-containing LBs on the basis of their RS spectra with 98% sensitivity and 100% specificity (Fig. 7A). The regression vectors (Fig. 7B and C) show a series of discriminative bands. However, from the regression vectors it is obvious that the major differentiating feature between the CON and CFA groups was not the presence/absence of CFAs, but rather the differences in UFA/SFA ratio. The regression vector for the control group shows bands characteristic for UFAs (3012, 1656, 1272 cm−1), whereas the regression vector for CFA-containing group demonstrates bands related to SFAs (2855, 1444, 1302 cm−1). This result is not unexpected, since the differences in the UFA/SFA ratio is known to be well reflected in Raman spectra and have a large impact on them.19,25,28 Furthermore, the differences in UFA/SFA ratio between LBs from control and modified cell lines were visible already in the average spectra (Fig. 6) and indicated by the overall lipid profile obtained through GC (ESI, Tables S3 and S4).

However, despite of the differences in UFA/SFA ratio being the dominant differentiating feature between both groups, the regression vector for the CFA-containing LBs exhibits some bands not associated with the changes in UFA/SFA relative content (Fig. 7C, black stars). In particular, the bands at 2994 and 1223 cm−1 can be observed. These bands are not present in Raman spectra of TAGs and FA constructed from saturated and/or unsaturated chains.25

Both bands are however present in the Raman spectrum of CFAs standard (Fig. 1 and 2). A direct comparison of Raman spectra of stearic acid (C18:0) and CFA standard (Fig. 3) as well as quantum chemical calculations clearly indicated that they originate from the cyclopropane ring vibrations. The possibility of these bands resulting from protein contribution to the spectrum is highly unlikely. As discussed in the previous section the CFA-containing group (CP1, CP4, CP5, CP6, CP7) consisted in the majority of spectra from larger LBs than the CON group (Fig. 4). The potential contribution of non-lipid material from outside of the LB (such as proteins) is a result of possible to achieve depth and spatial resolution. A larger protein contribution can be expected for smaller lipid bodies, with a size below the depth and spatial resolution. Therefore, in the discussed results the possible contribution of proteins would be expected to be much higher for the control group. Altogether, this suggests that the bands at 2994 and 1223 cm−1, observed in the regression vector of the CFA-containing group, are in fact markers of the CFAs presence in the LBs and demonstrates the ability of RS to detect and track CFAs in LBs.

Conclusion

Cyclopropane fatty acids are a group of lipids with an unusual structure, containing a cyclopropane ring. This translates directly to their unique physicochemical properties, which makes them highly desirable for industrial applications. CFAs are produced by several species of plants and bacteria. Currently however none of the natural sources of CFAs is sustainable for commercial use. At the same time, due to the untypical properties and potential applications, there is an emerging interested in studies focused on CFAs, as well as on sensitive methods for their detection.

Here, we demonstrated an approach, based on confocal Raman imaging, for determination of the CFA presence within S. cerevisiae strains, genetically engineered to induce the CFA production. The study of yeast cells was preceded by detailed spectral characterisation of CFA standard. We present for the first time the experimental Raman spectrum of a CFA standard (cis-9,10-methyleneoctadecanoic acid) along with quantum chemical calculations accompanied by detailed band assignments. Using the calculated spectra and through direct comparison with spectra of FA chains, we were able to identify bands associated with vibrations of cyclopropane ring (including primarily 2994 and 1222 as well as 3071 and 942 cm−1), constituting marker bands for this group of lipids. We supplemented this work with experimental spectra, quantum chemical calculations and band assignments for the CFA standard in IR, to provide its full vibrational characteristics.

Furthermore, we investigated several engineered cell lines in comparison to control. We were able to determine the impact of induced genetic modifications on the number, size and heterogeneity of produced LBs. Moreover, CRS demonstrated a significant decrease in the unsaturation of LBs (through the ratio of bands at 1660 and 1448 cm−1) in all modified cell lines compared to controls. The unsaturation level of LBs remained constant between the modified cell lines, regardless of the effectiveness of the engineering strategies towards increasing total lipid content. This indicates that the decrease of unsaturation in the modified cell lines is associated with the introduction of CFA production, as all engineered lines shared this modification. Finally, we were able to detect the presence of the CFAs within the LBs of modified cell lines, using CRS combined with PLS-DA. Even though the discrimination between LBs from control and engineered cell lines was dominated by spectral signatures related to differences in unsaturation, the regression vector of LBs from modified cell lines showed also the identified previously CFA marker bands (2992, 1222 cm−1). This demonstrated the potential of CRS for CFA spatially localised detection.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

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. We thank Mr Finlay Shanks for instrumental support. The authors would also like to acknowledge the allocation of computational resources through the Monash eResearch Centre and the National Computational Infrastructure in Canberra.

References

  1. M.-H. Liang and J.-G. Jiang, Prog. Lipid Res., 2013, 52, 395–408 CrossRef CAS PubMed.
  2. X. Meng, J. Yang, X. Xu, L. Zhang, Q. Nie and M. Xian, Renewable Energy, 2009, 34, 1–5 CrossRef CAS.
  3. H. U. Cho and J. M. Park, Bioresour. Technol., 2018, 256, 502–508 CrossRef CAS PubMed.
  4. Y. Li, Z. Zhao and F. Bai, Enzyme Microb. Technol., 2007, 41, 312–317 CrossRef CAS.
  5. Z. Zhu, S. Zhang, H. Liu, H. Shen, X. Lin, F. Yang, Y. J. Zhou, G. Jin, M. Ye, H. Zou and Z. K. Zhao, Nat. Commun., 2012, 3, 1112 CrossRef PubMed.
  6. F. Thevenieau and J.-M. Nicaud, OCL, 2013, 20, D603 CrossRef.
  7. X.-H. Yu, R. Rawat and J. Shanklin, BMC Plant Biol., 2011, 11, 97–97 CrossRef CAS PubMed.
  8. X.-H. Yu, R. R. Prakash, M. Sweet and J. Shanklin, Plant Physiol., 2014, 164, 455–465 CrossRef CAS PubMed.
  9. D. W. Grogan and J. E. Cronan, Microbiol. Mol. Biol. Rev., 1997, 61, 429–441 CAS.
  10. D. Poger and A. E. Mark, J. Phys. Chem. B, 2015, 119, 5487–5495 CrossRef CAS PubMed.
  11. M. S. Glickman, S. M. Cahill and W. R. Jacobs Jr., J. Biol. Chem., 2001, 276, 2228–2233 CrossRef CAS PubMed.
  12. C. Ying-Ying and J. E. Cronan, Mol. Microbiol., 1999, 33, 249–259 CrossRef.
  13. N. Loffhagen, C. Härtig, W. Geyer, M. Voyevoda and H. Harms, Eng. Life Sci., 2007, 7, 67–74 CrossRef CAS.
  14. R. A. Moss, T. Fujita, Y. Okumura, Z. Hua, R. Mendelsohn and L. Senak, Langmuir, 1992, 8, 1731–1735 CrossRef CAS.
  15. E. J. Dufourc, I. C. P. Smith and H. C. Jarrell, Biochemistry, 1984, 23, 2300–2309 CrossRef CAS.
  16. K. Kochan, H. Peng, B. R. Wood and V. S. Haritos, Biotechnol. Biofuels, 2018, 11, 106 CrossRef PubMed.
  17. B. R. Wood, A. Hermelink, P. Lasch, K. R. Bambery, G. T. Webster, M. A. Khiavi, B. M. Cooke, S. Deed, D. Naumann and D. McNaughton, Analyst, 2009, 134, 1119–1125 RSC.
  18. T. Ichimura, L.-d. Chiu, K. Fujita, H. Machiyama, T. Yamaguchi, T. M. Watanabe and H. Fujita, Sci. Rep., 2016, 6, 37562 CrossRef CAS PubMed.
  19. K. Majzner, A. Kaczor, N. Kachamakova-Trojanowska, A. Fedorowicz, S. Chlopicki and M. Baranska, 3D confocal Raman imaging of endothelial cells and vascular wall: Perspectives in analytical spectroscopy of biomedical research, 2012 Search PubMed.
  20. K. Kochan, E. Kus, A. Filipek, K. Szafranska, S. Chlopicki and M. Baranska, Analyst, 2017, 142, 1308–1319 RSC.
  21. D. Perez-Guaita, K. Kochan, M. Martin, D. W. Andrew, P. Heraud, J. S. Richards and B. R. Wood, Vib. Spectrosc., 2017, 91, 46–58 CrossRef CAS.
  22. A. F. Palonpon, M. Sodeoka and K. Fujita, Curr. Opin. Chem. Biol., 2013, 17, 708–715 CrossRef CAS PubMed.
  23. A. S. Stender, K. Marchuk, C. Liu, S. Sander, M. W. Meyer, E. A. Smith, B. Neupane, G. Wang, J. Li, J.-X. Cheng, B. Huang and N. Fang, Chem. Rev., 2013, 113, 2469–2527 CrossRef CAS PubMed.
  24. B. Vincent, Lipid Technol., 2010, 22, 36–38 CrossRef.
  25. K. Czamara, K. Majzner, M. Z. Pacia, K. Kochan, A. Kaczor and M. Baranska, J. Raman Spectrosc., 2015, 46, 4–20 CrossRef CAS.
  26. H. Wu, J. V. Volponi, A. E. Oliver, A. N. Parikh, B. A. Simmons and S. Singh, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 3809–3814 CrossRef CAS PubMed.
  27. O. Samek, A. Jonas, Z. Pilat, P. Zemanek, L. Nedbal, J. Triska, P. Kotas and M. Trtilek, Sensors, 2010, 10, 8635–8651 CrossRef CAS PubMed.
  28. K. Kochan, E. Maslak, C. Krafft, R. Kostogrys, S. Chlopicki and M. Baranska, J. Biophotonics, 2015, 8, 597–609 CrossRef CAS PubMed.
  29. U. Munchberg, L. Wagner, C. Rohrer, K. Voigt, P. Rosch, G. Jahreis and J. Popp, Anal. Bioanal. Chem., 2015, 407, 3303–3311 CrossRef PubMed.
  30. J. D. Gelder, K. D. Gussem, P. Vandenabeele and L. Moens, J. Raman Spectrosc., 2007, 38, 1133–1147 CrossRef.
  31. H. Peng, L. He and V. S. Haritos, J. Ind. Microbiol. Biotechnol., 2018, 1–11 Search PubMed.
  32. J. C. Obert, D. Hughes, W. R. Sorenson, M. McCann and W. P. Ridley, J. Agric. Food Chem., 2007, 55, 2062–2067 CrossRef CAS PubMed.
  33. A. Caligiani, M. Nocetti, V. Lolli, A. Marseglia and G. Palla, J. Agric. Food Chem., 2016, 64, 4158–4164 CrossRef CAS PubMed.
  34. V. Lolli, A. Marseglia, G. Palla, E. Zanardi and A. Caligiani, J. Anal. Methods Chem., 2018, 2018, 8034042 Search PubMed.
  35. E. G. Bligh and W. J. Dyer, Can. J. Biochem. Physiol., 1959, 37, 911–917 CrossRef CAS PubMed.
  36. O. L. Knittelfelder and S. D. Kohlwein, Cold Spring Harb. Protoc., 2017, 5, 408–411 Search PubMed.
  37. H. Peng, L. Moghaddam, A. Brinin, B. Williams, S. Mundree and V. S. Haritos, Biotechnol. Appl. Biochem., 2018, 65, 138–144 CrossRef CAS PubMed.
  38. Y. Zhao and D. G. Truhlar, Theor. Chem. Acc., 2008, 120, 215–241 Search PubMed.
  39. J.-D. Chai and M. Head-Gordon, Phys. Chem. Chem. Phys., 2008, 10, 6615–6620 RSC.
  40. D. E. Woon and T. H. Dunning Jr., J. Chem. Phys., 1994, 100, 2975–2988 CrossRef CAS.
  41. K. Majzner, K. Kochan, N. Kachamakova-Trojanowska, E. Maslak, S. Chlopicki and M. Baranska, Anal. Chem., 2014, 86, 6666–6674 CrossRef CAS PubMed.
  42. K. Kochan, K. M. Marzec, E. Maslak, S. Chlopicki and M. Baranska, Analyst, 2015, 140, 2074–2079 RSC.

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c8an01477a

This journal is © The Royal Society of Chemistry 2019