Based on experimental protocols commonly applied in yeast research, the influence of various growth conditions on the lipid composition of the wild-type strain BY4741 was investigated by quantitative shotgun mass spectrometry. The tested parameters were: temperature (15, 24, 30, 37°C), carbon source (glucose, raffinose, glycerol), different growth phases (early and middle logarithmic as well as early and stationary phase) complete (YP) vs. synthetic complete (SC) medium, galactose-induced over-expression of a cytosolic as well as a trans-membrane protein and the presence or absence of the selective drug G418. In addition, the lipidome of haploid (a and α) and diploid cells was determined (see Materials & Methods
for further details). In order to assess the influence of growth conditions on the lipid composition of yeast cells, it was necessary to ensure that factors such as the handling of the cells, lipid extraction and data acquisition by mass spectrometry were not a source of variations of the yeast lipidome. Therefore, in each experimental run, triplicate control cultures grown under “reference conditions” were included (SC glucose at 30°C and harvested in early logarithmic (log) phase (OD600
1); corresponding to samples: “OD 1”, “SCglc”, “T30”, and “a” (for details see Materials & Methods
)). Variance analyses, as well as hierarchical clustering revealed that the lipidomes of the control samples grown on different occasions (days, media batches) exhibit a remarkable similarity, thus confirming the reliability as well as the consistency of our approach (, Table S1
Hierarchical clustering and principal component analysis (PCA) of yeast lipidomes from different growth conditions.
The acquisition of mass spectra of about 80 samples allowed for the absolute quantification of a total of about 160 individual lipid species belonging to 18 lipid classes. The amount of data produced necessitated the development of a suitable approach simplifying and facilitating data analysis and interpretation. To this end, quantitative lipidomic features were derived from the measurements of individual lipid species. Their level of detail ranged from categories through classes down to the species level. For example, quantities of all GP species comprised of two double bonds in their fatty acid moieties were summarized as one lipidomic feature, i.e. GP DB-2. Together with the lipidomic feature GP DB-1 and GP DB-0 (i.e. all GP comprising one or zero double bond, respectively) we obtained a simple expression for the degree of unsaturation of the entire category of GP. We proceeded similarly for the other lipid categories and the lipid classes. Other molecular parameters like FA chain length and hydroxylation were also included in the features. Thus, by way of example, the lipidomic feature PI C-34 summarizes all PI species containing a total of 34 carbon atoms in their hydrocarbon chain.
The absolute quantification of lipid species and the summary of their amounts as lipidomic features provide a quantitative as well as qualitative, functional and biologically relevant description of a lipidome. The calculation of variances of the lipidomic features across the entire sample set or a subset of samples identifies the features that undergo the most pronounced changes. Additionally, the lipidomic features were used for hierarchical clustering of the datasets (see Materials & Methods
). This approach simplified the data analysis and interpretation process.
Influence of different factors on the yeast lipidome - general observations
The quantification of the lipidome of yeast cells grown under the various conditions led to four major observations. First, the carbon source has a strong effect on the overall lipid composition. Lipidomes of cells grown in raffinose and glycerol cluster together, while lipidomes of the glucose-grown cells form a separate cluster (). Second, the progression through the growth phase, i.e. from early logarithmic to stationary phase, causes remarkable changes in the lipidome. This is manifested by the very distant cluster formed by lipidomes of mid-log (OD3.5), early stationary (OD6) and late stationary (ODstat) yeast cultures. Third, there is a pronounced effect of decreasing the growth temperature to 15°C (T15). While the other lipidomes from the growth temperature experiment (T24, T30, T37) group within the “glucose cluster”, the lipidome of cells grown at 15°C appears to be significantly different. Fourth, the galactose-induced overexpression of a cytosolic or a transmembrane protein (within an induction time of up to 3 h) as well as the presence of the selective drug G418 and the different mating types (a, α, diploid) are not causing pronounced changes in the lipid composition of the yeast cells. The lipidomes of galactose-induced cells are part of the “non-glucose cluster”, while the lipidomes of the different mating types cluster tightly with the lipidomes of glucose-grown cells (for a complete dataset, see Table S1
). The nature of the changes caused by different carbon sources, growth phase, and growth temperature is presented in the following sections in more detail.
The effect of temperature on the lipidome was tested by growing cells at 15, 24, 30 and 37°C. As expected, the degree of unsaturation of the GP showed a gradual trend: it was the highest at 15°C and the lowest at 37°C (). This change was mainly caused by a shift from mono- to di-unsaturated GP species. A similar gradual trend was observed for the length of the hydrocarbon chains in GP. While at 15°C GP species with a total sum length of hydrocarbon chains of 32 carbon atoms were most abundant, at 37°C species with 34 carbon atoms became predominant (), meaning that the FA chains of the GP species became longer with increasing temperature. On the level of GP classes, the most striking effects were the opposing trends of PI and PE changes. While PE levels decreased with increasing temperature, PI became more abundant (). Interestingly, changes in PC abundance did not show a clear temperature-dependent trend. However, changes in its unsaturation index followed the general trend of the other GP, i.e. increase in mono-unsaturated species at the expense of di-unsaturated PCs ().
Lipidomics of yeast grown in different temperatures.
Prominent changes also occurred within the sphingolipids. At 15°C, M(IP)2C was the most abundant SP (), but with increasing temperature its levels decreased, and IPC became dominant. MIPC showed a comparable trend. At the species level, a temperature-dependent increase in the sum length of the hydrocarbon chain moieties (from 44 to 46) was observed (as well as a minor increase in hydroxylation) (). Additionally, ergosterol levels also increased gradually, although subtly, with increasing temperature. Furthermore, it is interesting to note that ceramide levels increased significantly at 37°C, while TAG levels increased at 15°C ().
The lipidomes of yeast cells from different growth phases were quantified. Samples were withdrawn in early (OD1), middle logarithmic (OD3.5) and early stationary phase (OD6) as well as from stationary phase after 24 h of continuous growth (ODstat; , inset). During progression through the growth phases, a strong increase in TAG levels was observed. Interestingly, the accumulation of TAG was already more than 2-fold in the mid-log phase when compared with the early log phase and increased further in the later growth phases (). Concurrently with the increase of TAG, the PI level dropped by more than 75% from early to mid-log phase, while the PE level dropped by 50% after mid-log phase. PI levels recovered slightly in later growth phases, while PE levels stabilized. The PC level remained unchanged throughout the course of growth. However, the unsaturation of PC was strongly decreased already in mid-log phase and there was a pronounced increase in its hydrocarbon chain length in stationary phase (Table S1
). On average, the length of GP increased with progression through the growth phases (shift from C32 to C34 and also a small but significant increase of C36), while the unsaturation index remained unchanged ().
Lipidomics of yeast in different growth phases.
With respect to the complex sphingolipids, IPC is the most abundant SP in early log phase. However, the IPC level greatly decreased in mid-log and further in later growth phases. Concurrently with the drop in IPC, MIPC becomes the most abundant sphingolipid, while the M(IP)2C level remained constant throughout the experiment (). Moreover, while progressing through the growth phases, an increase in the length of the hydrocarbon chain moieties (from 44 to 46) was observed. Additionally, the hydroxylation of the SP subtly increased (). These changes were similar to those observed in the temperature experiment.
We investigated the influence of the carbon source in the growth medium on the yeast lipidome by growing the cells in complete medium (YP, see Materials & Methods
) supplemented with glucose, raffinose (YPglc and YPraf, resp.; fermentable carbon sources) or glycerol (YPgly, non-fermentable carbon source).
Hierarchical clustering revealed that the lipidomes of glycerol- and raffinose-grown cells are similar to each other, but differ from the lipidome of glucose-grown cells (). Interestingly, a significant increase in cardiolipin could be observed in cells grown in YPraf and YPgly (). Furthermore, PS was strongly elevated in YPglc, while PA was increased in YPraf and also in YPgly. The abundances of PC and PI also changed depending on the carbon source, while the PE levels were not affected. In general, GP became more unsaturated (to a large extent due to an increase in di-unsaturated PS and PA species, ) and had longer hydrocarbon chains (increase of C34 species at the expense of C32, ) in raffinose- and glycerol-containing media.
Lipidomics of yeast grown on different carbon sources.
The carbon source also had an impact on the sphingolipids. In glucose medium, IPC was the most abundant SP class. However, in YPraf as well as in YPgly M(IP)2C was predominant (). Moreover, the hydroxylation state of the SP was elevated in glycerol-grown cells.
Additionally, it was interesting to note that differences in the lipid composition could be observed between cells grown in complete (YP) and synthetic complete (SC) medium, irrespective of the carbon source (Table S1
Flexibility of the yeast lipidome
During the course of the experiments we observed that some lipidomic features (e.g. lipid classes or unsaturation indices) varied to a larger extent and more often than others, depending on growth conditions. This observation prompted us to derive general rules about the potential variability of the lipids. To describe the variability of lipidomic features, here we introduce the term flexibility. We define flexibility as the index of dispersion (variance-to-mean ratio, VMR) of a given lipidomic feature across the entire dataset, i.e. changes of its abundance in different conditions. A VMR<1 indicates that data are under-dispersed or, in other words, the abundance of a lipidomic feature deviates from the mean only to a small extent. Thus, a VMR<1 indicates low flexibility of a given feature. On the other hand, a VMR>1 indicates that the data are over-dispersed and have a clumped distribution, that is, the abundance of a lipidomic feature can deviate strongly from the mean. Therefore, the VMR is a useful measure of the capability of a lipidomic feature to undergo quantitative changes (i.e. of its flexibility).
The analysis of the flexibility of the lipid categories and classes revealed that GL is the most flexible lipid category, exhibiting the highest VMR (). This is mainly due to the high flexibility of TAG. Notably, the least flexible lipid categories are the SP and the sterols (Erg), while the GP flexibility is intermediate.
Flexibility of the lipid categories and classes.
Interestingly, when comparing the flexibility of the complex SP classes IPC, MIPC and M(IP)2C with the flexibility of the SP category as such, it turned out that the classes constituting the category are more flexible than the category itself. Thus, the total average amount of complex SP was stable, while the relative abundance of the SP classes within the category was highly variable.
The most flexible GP class is PS (). This is largely due to the strong increase of the PS level in cells grown in YPglc (). The next most flexible classes are PI, PE, PC, and PA. CL is the least flexible GP class, since its abundance changed to a small extent only when glycerol or raffinose was used as carbon source.
Internal variability of lipid classes
Each lipid class may also be characterized by an internal variability. The internal variability describes the extent of change within a lipidomic feature (e.g. the average variation of the length, unsaturation index, or hydroxylation within a class). Therefore, the internal variability is a measure for the extent of changes that occur in the species profile of a given lipid class.
For the description of the internal variability of the individual GP classes, the VMR of the length and unsaturation features of each class were averaged in order to obtain an estimation of the extent of the qualitative changes within the class. For example, it was observed that PC exhibits much stronger changes in its species profile than PI (Table S1
). Aiming at the expression of this observation by a single numerical value (the internal variability index), the VMR of the hydrocarbon chain length and the unsaturation indices within the respective classes (i.e. PI and PC) were calculated and averaged for each class. It turned out, that PC shows a higher internal variability than PI (), confirming the observation in the species profiles. Therefore, the internal variability appears to be an appropriate measure for the qualitative changes that occur within a given lipid class or category.
Internal variability of lipid classes.
Further analysis of the internal variability indices revealed that PS is the most variable GP class (). Again, this is mainly due to the strong changes in PS in yeast grown in YPglc. Apart from that, PA and PC appear to have large internal variability, while PE and PI are rather invariable, with PI being the least variable GP class (except for CL, which was omitted from this analysis, since in the majority of experiments only a few species could be detected). Interestingly, except for PI, the internal variability of all other GP classes is largely determined by the unsaturation index (). Furthermore, the GP classes are internally more variable than the GL classes TAG and DAG.