The biosynthesis of phospholipids is linked to the cell cycle in a model eukaryote

The structural challenges faced by eukaryotic cells through the cell cycle are key for understanding cell viability and proliferation. In this study, we tested the hypothesis that the biosynthesis of structural lipids is linked to the cell cycle. If true, this would suggest that the cell’s structure would form part the control of the cell cycle. Lipidomics (31P NMR and MS), proteomics (Western immunoblotting) and transcriptomics (RT-qPCR) techniques were used to profile the lipid fraction and characterise aspects of its metabolism at seven stages of the cell cycle of the model eukaryote, Desmodesmus quadricauda. We found considerable, transient increases in the abundance of phosphatidylethanolamine during the G1 phase (+35%, ethanolamine phosphate cytidylyltransferase increased 2·5×) and phosphatidylglycerol over the G1/pre-replication phase boundary (+100%, phosphatidylglycerol synthase increased 22×). The relative abundance of phosphatidylcholine fell by ~35% during the G1. N-Methyl transferases for the conversion of phosphatidylethanolamine into phosphatidylcholine were not found in the de novo transcriptome profile, though a choline phosphate transferase was found, suggesting that the Kennedy pathway is the principal route for the synthesis of PC. The fatty acid profiles of the four most abundant lipids suggested that these lipids were not generally converted between one another. The relative abundance of both phosphatidylinositol and its synthase remained constant despite an eightfold increase in cell volume. We conclude that the biosynthesis of the three most abundant structural phospholipids is linked to the cell cycle in D. quadricauda.


Introduction
The processes governing control of the cell cycle in eukaryotic organisms have been researched and characterised in considerable depth over the last half-century. This work has shown that checkpoints and the expression and degradation of cyclins and cyclin-dependent kinases are important in controlling progress through the cell cycle in fungi and metazoans (1)(2)(3). Similar mechanisms and homologous proteins were subsequently found in plants (4)(5)(6).
However, successful completion of the cell cycle also presents a number of structural challenges. The plasma and compartment membranes must expand and undergo topological remodelling during the cell cycle, whilst maintain biochemical and barrier functions. The correct components of the daughter cells and their spatial arrangement must be organised. Finally, lipid membranes must be divided so that two or more viable daughter cells are produced. Interruption in the functions of the membrane, mis-timed membrane lysis or incorrect spatial distribution of cell components all represent perilous threats to cell survival that must be avoided for a cell to be viable.
The success of the cell cycle across countless species indicates that the factors that affect membrane behaviour, shape and size are under careful control through repeated cycles of cell division. This is also supported by recent reports about the structural integrity of cells through growth and division. Evidence for checkpoints that couple the structural integrity of the plasma membrane to DNA synthesis in eukaryotic cells is beginning to emerge (7).
Control of division through cell size has also received attention in single-cell organisms such as yeast (8), and prokaryotes (9,10) but also in green algae (11,12). This work is particularly interesting in the light of evidence for the modulation of the composition of structural lipids in prokaryotes through their cell cycle (13,14). The changes in the topology of the cell envelope of prokaryotes and their change in lipid composition (14) are consistent with the wealth of evidence that lipid composition has an important influence on the geometry of the structures formed (15)(16)(17)(18)(19)(20).
This evidence therefore indicates that lipids have a considerable role in determining membrane behaviour and thus cell structure. Taken with the importance of structural integrity for cell viability, this raises questions about how membrane systems are managed through the cell cycle. One suggestion is that the composition of lipid membranes is remodelled to minimise the energetic costs of changing their shape through the cell cycle (21). This led us to the hypothesis that the biosynthesis of structural lipids is linked to the cell cycle.
To test this hypothesis, we elected to use a model organism that has a well-characterised cell cycle, shows the structural challenges through the eukaryotic cell cycle as clearly as possible and comprises typical eukaryotic lipids. We also wanted to be able to collect samples of cells at different stages of the cell cycle without introducing artefacts associated with the drugs needed to synchronise cell cultures. All of these conditions were met by Desmodesmus quadricauda. This organism undergoes multiple fission producing eight daughter cells in one cell cycle ( Fig. 1). Multiple fission is common in green algae, with some species dividing into up to 32 daughter cells in one mitosis (22,23). Cell division in D. quad. is also timed carefully as this photosynthetic species preferentially undergoes cytokinesis when there is insufficient light for the light-dependent part of photosynthesis. Thus, cultures can be synchronised through light and dark periods without the need for drug-based inhibition of DNA synthesis (24,25). D. quad. is also a good model for other eukaryotic cell types as it comprises similar lipids to most eukaryotes (21,26,27).
We cultivated populations of D. quadricauda, synchronised them and extracted the lipid fraction at defined points in the cell cycle. We developed a novel method for preparing cells that is compatible with established procedures for extracting lipids from biological samples (14,28,29). Lipid class abundance was measured using 31 P Nuclear Magnetic Resonance Spectroscopy (NMR) and the fatty acid composition of lipids determined using high resolution mass spectrometry (HRMS). We assembled a de novo transcriptome of D. quad. to identify homologues of genes involved in cell cycle regulation and lipid metabolism. Finally, we used a combination of the reverse-transcription quantitative polymerase chain reaction (RT-qPCR) and Western Blotting to determine the mRNA and protein abundance of putative enzymes involved in lipid biosynthesis.
It was important to test this hypothesis because the physical integrity of cells through the cell cycle is a fundamental part of cell viability, but the control of the physical process is relatively poorly understood.
Understanding how the structure of cells fail is of interest in controlling cell growth, either to hinder it (antibiotics, anti-tumour compounds) or to promote it (tissue regeneration). The processes that exist in evolved systems also have implications for preparations of artificial cells and for nanotechnology.

Results
In order to test the hypothesis that lipid biosynthesis is linked to the eukaryotic cell cycle, we profiled the abundance of structural lipid classes through the cell cycle alongside a formal characterisation of growth and the cell cycle (Fig. 2). The growth characterisation was done through the cell volume and mass of RNA per cell, while that of cell cycle was done through monitoring DNA replication (DNA mass) and nuclear divisions (number of nuclei) per cell. As expected under given growth conditions, the cells increased their volume about eight-fold, which was accompanied by increase in mass of major macromolecules such as RNA (Fig. 2). Concomitant with cell growth the cells entered three sequences of DNA replication sequentially, nuclear and cellular division leading to division into eight-celled daughter coenobia (Fig. 2). We developed novel procedures for handling cultures as established procedures for handling this cell type for proteomics were not suitable for handling the lipid fraction.
The lipids were extracted (n = 6 biological replicates) in a similar manner to previous work (14,28-30). Phosphorus ( 31 P) NMR was then used to measure the relative abundance of lipid classes (28,29).
This showed that the relative abundance of major phospholipids was modulated through the cell cycle (Fig. 3). The lipid fraction was dominated by four lipid classes through the cell cycle: phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PI) and phosphatidylglycerol (PG). They make up over 95% of the phospholipid fraction between them (Fig. 4). During the G1 phase, in which the greater part of cell expansion occurs, the relative abundance of PC, PE and PG changed significantly (Fig. 4), including a significant change after the first commitment point. We therefore analysed the relationship between lipid biosynthesis through particular phases, as well as throughout the whole cell cycle. These are described in order.
Lipid remodelling in G1-The relative abundance of PC falls by a third in the G1 phase (Fig. 4, Table S1). However, during this phase, the cell volume increases by a factor of ~2·5 (Fig. 2), indicating that the overall mass of phospholipid in a cell increased by around the same factor. This suggests that the biosynthesis of PC does not occur at the same rate as cell growth. We therefore investigated the control of the biosynthesis of PC. There are two well-characterised routes for the biosynthesis of PC in eukaryotes. One is the methylation of PE, the other is the transfer of choline phosphate onto a diglyceride. We assembled a de novo transcriptome of D. quadricauda and then we investigated with combination of RT-qPCR, Western Blotting and lipidomics techniques to investigate which pathway was more important.
In the assembled transcriptome, we searched for putative homologues of individual enzymes involved in methylation of PE. However, we were unable to find phosphatidyl-N-methylethanolamine N-methyltransferase (PEMT). Mass spectrometric profiling of lipids shows that there are a number of isoforms of PC that are without equivalent in PE (Fig. S3), suggesting that not all PC can come directly from PE. Furthermore, a large increase in the abundance of PE is not met with an increase of PC, suggesting that PE is not a substrate for biosynthesis of PC.
Taken together, this suggests that PE is not generally used to make PC in this organism and that the two lipids are not directly linked metabolically. This strongly suggested that PC was produced endogenously through transfer of choline rather than methylation of PE.
The increase in the relative abundance of PE during the G1 phase (40%) is accompanied by a factor of eight increase in the abundance of mRNA and factor of 2 increase in the abundance of the protein Ethanolamine Phosphate cytidylyl Transferase 1 (EPT1), the enzyme that synthesises PE from ethanolamine phosphate and diglycerides (Fig.   5, S4). This showed that the synthesis of PE is based upon transfer of ethanolamine onto a diglyceride rather than modification of PC or decarboxylation of phosphatidylserine (PS). This is therefore consistent with the results of the biosynthesis of PC that show that PC and PE are produced separately and not interconverted.

Lipid remodelling over the pS/S boundary
There is a rapid increase in the relative abundance of PG and a reduction in that of PE towards the end of the pS phase and into the replication phase (S). This suggested that PE is biosynthesised rapidly in G1 before the relative abundance of PG overtakes it. We therefore tested the hypothesis that biosynthesis of PG relied upon PE.
A comparison of the fatty acid profile of PE and PG suggests that they are not closely related (Fig. S3). The fatty acid residue (FAR) profile of PG is more consistent with prokaryotic lipid metabolism (more saturated, shorter FARs), where PE is more eukaryotic (dominated by FA(16:4) and unsaturated C18 FARs), Fig. S4. This is consistent with our understanding that the bulk of the PG fraction resides in chloroplasts (a prokaryotic compartment) and the bulk of PE is typically in the plasma membrane and eukaryotic compartments. Lastly, enzymes involved in converting PE to PG have not yet been reported. This led us to determine how the synthesis of PG is controlled with respect to the cell cycle.
Western blotting revealed that the abundance of PG synthase 1 (PGS1) increases by a factor of 22 from the beginning of the pS phase, and starts to fall back zero by the start of the cell division, preceded by chloroplast division (Fig. 5 and S4), indicating that PG is produced by this enzyme from cytidyl diphosphate-diglyceride (CDP-DAG) and glycerol-3-phosphate. These data are consistent with de novo synthesis of PG from phosphatidic acid in chloroplasts and not from PE, indicating a control of the biosynthesis at pS/S boundary that is not tied to other lipids.

Lipid remodelling in the G2, M and G3
The abundance of scarcer lipids was modulated through Mitosis and G3 (Table S1). Small amounts of phosphatidic acid (PA, FAR profile in Table S2), PS ( Table S3) and evidence of plasmalogens of PC and PE were found. The abundance PA is 0·5-1·0% from G1 to the end of G2, but increases to around 3% during mitosis. PC-plasmalogen may also increase in abundance during the same period. As the cell increases in volume only from ~710 µm 3 to ~830 µm 3 ( Fig. 2) during G3 and Mitosis, these shifts are not expected to be consistent with major structural changes in either eukaryotic or prokaryotic compartments. PE-plasmalogen and PS remain roughly constant at ~1% throughout the cell cycle (Table S1). Although such species have been detected before (31,32), it is not clear what their role is in the cell cycle, if any.

Lipid remodelling across the whole cell cycle
The relative abundance of PI is maintained at ~5% throughout the cell cycle (Fig. 4), indicating that rate of its synthesis correlates closely with the increase in the volume of the cell. However, the population of PI molecules must increase by a factor of eight through the cell cycle. RT-qPCR indicated that the expression of PI synthase (PIS) is constant through the cell cycle (Fig. 5). This implied a constant supply of PI relative to other phospholipids and suggests its production is separate from that of PG, PE and PC. Fatty acid profiling of isoforms of PI and PC showed that PI's profile is not consistent with PC (Fig. S1), nor with the much less abundant PIPs (Fig. S2).
This suggested first that PI and PC are not generally interconverted and second that there are many isoforms of PI that have non-signalling roles, something also observed in HeLa cells (33). PI appears only to be present in the membranes of eukaryotic compartments (34), suggesting to be exclusively eukaryotic in origin. The FA profile of PI in D. quadricauda (Table S1) is characterised by polyunsaturated and longer chain configurations. These results are consistent with PI as a lipid that is required in small amounts throughout the cell cycle. Through HRMS alone, we also observed the presence of a number of known phytolipids and phytosterols throughout the cell cycle that have not been observed in this species before (Tables S4-6). These species were not visible by 31 P NMR so were not quantified and thus could not be linked to the cell cycle.

Discussion
This study was motivated by the hypothesis that the biosynthesis of structural lipids is linked to the cell cycle in eukaryotes. Testing this hypothesis provided evidence that the biosynthesis of the three most abundant phospholipids (PC, PE and PG) in the model used are modulated through the cell cycle and linked to it. PC, PE and PG all dominate the overall lipid fraction at different points. This is in contrast to PI that does not change in relative abundance through the cell cycle. Evidence from MS, Western blots and RT-qPCT suggests that PC, PE and PG, and the next most abundant, PI, are not interconverted between one another and are thus metabolically independent after the CDP-DAG is assembled.
The evidence for independent synthesis of the four most abundant lipids and that three of them dominate the organism's lipid fraction at different stages of the cell cycle is consistent with cell-cycle-based control of lipid biosynthesis through different pathways that are switched on and off appropriately, and locally. It also characterises those stages of the cell cycle as having a particular focus in their lipid metabolism. Furthermore, the change in abundance of EPT1 and PSG1, and the increase in abundance of PE and PG that follow, suggests that these are the principal synthetases of these lipids in D. quad. The abundance of the enzymes increased rapidly just before the abundance of the appropriate lipid does, and then most of the enzyme is lost in a way that limited synthesis of those lipids, up to 90% in the case of EPT1. This indicated that the final step in the biosynthesis of those two crucial lipids occurs through only one enzyme each, and that the expression of both enzymes is linked to the cell cycle.
Significant changes in PC, PE and PG-lipids that are both abundant and well known to have structural rolesinvites questions about the physical role of these molecular species. Studies of lyotropic phase behaviour have established that PG and PC are bilayer-forming lipids and that the anionic PG may have an important charge-based effect (35). Unlike PC and PG however, PE is a non-bilayer forming lipid that typically favours the assembly of curved lipid mesophases (15). Recent work in biological systems indicates that PE is part of the control of fluidity of membranes in vivo (36) and has an essential role in cytokinesis in at least one eukaryotic cell type (37). This suggests that an increase in the abundance of PE in the G1 phase represents a shift in membrane properties through this phase. PE's propensity for forming inverse mesophases (15,38) suggests that more curved membranes are required as the cells enter from the pS to S phase. During this period, preparations for DNA synthesis are made but the cells do not grow as rapidly (22), suggesting re-organisation is more important than expansion at this point. This is consistent with the theory that cells remodel their lipid composition in order to lower the energy of the succeeding phase (21), as the internal structural needs of the cell change at this point. The precise role and distribution of PE is required to understand this more deeply and could be used to answer the importance of the site of biosynthesis in driving membrane reorganisation.
These results are also of interest in the light of evidence that the rate as well as the timing of lipid biosynthesis differs between lipid classes. For example, it is not clear from this work that the biosynthesis of PI has a peak, either through the abundance of the lipid or the expression of its synthase (PIS). It is possible that it is produced continuously, correlating with cell size. However, it is not clear from our data whether there is a mechanism that links PI synthesis to cell size or whether inhibiting PI synthesis would disrupt progress through the cell cycle through structural means. This is important because PI is also beginning to be recognised as a non-bilayer lipid.
Studies of the lyotropic phase behaviour of PI have shown that it has a concentration-and time-dependent effect on the geometry of lipid systems, induces considerable inverse curvature at lower hydrations (17) and introduces defects into lipid bilayers (39). This is similar to PE that is also characterised by inverse curvature (15). This evidence hints that a local abundance of PI may be able to reduce the energetic cost of membrane fission.

Experiments of the distribution of lipids in Chinese Hamster Ovary cells have shown that PE and PI-derived PIP2
must be present the cleavage furrow of CHO cells in order for cytokinesis to take place (37,40,41). This suggests that lipids associated with inverse curvature such as PI and PE have a role in membrane scission.
The increase in abundance of PG and evidence that it arises from prokaryotic rather than eukaryotic lipid biosynthesis is consistent with an expansion in chloroplast size at that point in the cell cycle. This is consistent with a link between lipid biosynthesis and chloroplast as the structure making up the vast majority of cell volume.
The role of PG in photosynthesis is well-established (42). PG's lyotropic behaviour under physiological conditions is dominated by bilayer (membrane-like) systems (19). It is also anionic, suggesting that this lipid's synthesis is associated with an increase in membrane area and a change in electrostatic interactions. Despite being a bulk lipid, evidence that PG has a role in regulating protein orientation in chloroplasts is emerging (43). PGS1 has also been found in chloroplasts in the alga Chlamydomonas reinhardtii (44) and in higher plants (45,46). This suggests that PG has a role in chloroplast membranes in all plants, adding to the view of this lipid as a ubiquitous one (47).
Recent work on fission yeast has begun to show that a gene involved managing lipid metabolism is involved the controlling the cell cycle (8). The gene mga2 regulates lipid homeostasis (48) and lipid synthesis (49) and when deleted, leads to cells that are unable to correct size deviations within individual cell cycles (8). This is consistent with the 'sizer' hypothesis about the control of cell division in eukaryotes (9,10) and may also apply to D. quad. as there is a clear size component to the preparation of a cell for multiple fission. This study shows that individual enzymes involved in lipid metabolism are linked to the cell cycle, however this work suggests that more general genes governing cell size are also involved. This is interesting because it can be used to inform the interpretation of cell-cycle-based lipid biosynthesis in other studies.
This includes general shifts in lipid abundance during cell elongation (50) but also specific ones such as the increase in the abundance of PA in the leaves of Arabidopsis thaliana during dark periods (51). The latter may be consistent with our observations of an increase in the abundance of PA during mitosis. Jueppener et al. reported a qualitative study of lipids through the cell cycle of Chlamydomonas reinhardtii, finding evidence for shifts in the lipid profile through that organism's cell cycle and highlighting the cyclical nature of lipid metabolism in that species (52).
Interestingly, Chlamydomonas reinhardtii appears to make considerable use of uncharged glyceride lipids such as mono-galactosyl diglycerides (MGDGs) and di-galactosyl diglycerides (DGDG) (53). In the present study, we found galactosyl-glycerides and others in D. quad. that have not been reported in this species before (Tables S4-6).
The evidence for the biosynthesis of lipids at particular points of the cell cycle raises questions about the extent of the link between the two. One question is whether there is a particular link between the availability of fatty acids and the synthesis of phospholipids. Evidence that particular FAs can favour the synthesis of phospholipids is beginning to emerge (54), suggesting a link between de novo fatty acid synthesis and progression through the cell cycle. Indeed, studies in yeast have provided evidence for a lipase that releases FAs from TGs that is linked to progress through the cell cycle (55).
An understanding of polyunsaturated fatty acid synthesis in algae may also be useful for harnessing them for industrial triglyceride synthesis (56,57). This is attractive as a sustainable source of essential fatty acids such as docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA). Evidence for lipid metabolism being linked to progress through the cell cycle implies that fatty acid biosynthesis is not directed solely towards triglyceride metabolism. Thus, a characterisation of lipid metabolism through the cell cycle can be used to inform the preparation of industrial cultures in which the balance between progress through the cell cycle and accumulation of triglycerides is struck.

Conclusions
This study was motivated by the hypothesis that the biosynthesis of structural lipids is linked to the cell cycle.
Lipid profiling showed that the lipid fraction is remodelled several times through the cell cycle in this organism. A combination of mass spectrometry, proteomics and transcriptomics indicate that the most abundant phospholipids are not directly connected to one another metabolically, but are connected to progress through the cell cycle. This has implications not only for our general understanding of the cell cycle, but also our understanding of the physical aspects of cell division. It may also be useful for informing the design of artificial cells and the use of algae for industrial production of triglycerides and characterisation of the migration of biomass in food chains through nutrients such as DHA. Continuous re-modelling of the lipid fraction through the course of the cell cycle implies that lipid metabolism is as important as that of proteins and nucleic acids for the success of this process.

Experimental Procedures
Reagents & Chemicals-Solvents, and fine chemicals were purchased from SigmaAldrich (Gillingham, Dorset, UK) except phosSTOP tablets were purchased from Roche (Welwyn, Hertfordshire, UK; stored at 4°C). Chemicals for the growth medium were purchased from Penta (Chrudim, CZ). Synchronisation was done according to reported procedures (24). Briefly, before the start of the experiment, the cells were synchronized and then grown for one more whole cell cycle. At the beginning of the following light period they were diluted to the initial density (1×10 6 cells mL -1 ). The synchronization itself was carried out by alternating light/dark periods (15 h/10 h), the lengths of which were chosen according to the growth parameters of the cells. The optimum time for turning off the illumination was when the cells started their first protoplast fission. The length of the dark period was chosen to allow all cells of the population to release their daughter cells. Under the conditions described above, cell division started at about the 15th hour of the cell cycle and the cells typically divided into eight daughter cells (Fig. 1, 2).
Preparation of dried cell lysates-The active culture (650 mL) was filtered and the filtrate collected and centrifuged (4000 × g, 5 min). The resulting pellet was resuspended in a mixture of chaeotropes (5 mL; thiourea 1·5 M and guanidinium chloride 6 M), phosSTOP (1 tab/sample, dissolved in 1 mL PBS) and 2-butoxyphenylboronic acid (BPBA, 2 mg/mL final concentration, ethanolic stock solution 100 mg/mL). The suspension (~10 mL) was agitated vigorously with glass beads (1 min, air-tight Falcon tube, 50 mL) before being frozen in liquid nitrogen and then freeze-dried. The freeze-dried material was stored under a nitrogen atmosphere and transported at room temperature.
Isolation of lipid fraction-Freeze-dried cell lysate was powdered (pestle) and rehydrated (PBS, 5 mL) with agitation (1 min) but without sonication. The mixture was frozen (193K) and freeze-dried. The resulting dry, free-flowing powder was resuspended in a mixture of dichloromethane (20 mL) and water (20 mL) and diluted with sufficient methanol to make a stable uniphasic solution (40-45 mL, 500 mL separating funnel). The mixture was then made biphasic by addition of dichloromethane (20 mL). The dichloromethane solution was separated and the aqueous solution washed (dichloromethane, 20 mL). Triethylammonium chloride (TEAC) was added to the remaining aqueous solution (final concentration of 2 mM, 2M stock) and the aqueous solution washed with dichloromethane (2 × 20 mL). The combined organic solutions (~90 mL) were filtered through filter paper and concentrated in vacuo before storage of the resulting lipid film under nitrogen at -20°C.
Solution phase 31 P NMR-Lipid films were dissolved in the CUBO solvent system (28,29) (450 µL, 23-26 mg isolate/sample). Data acquisition was similar to published work (13,14), but using a Bruker 400 MHz Avance III HD spectrometer equipped with a 5 mm BBO S1 (smart) probe operating at 298K. 31 P NMR spectra were acquired at 161·98 MHz using inverse gated proton decoupling, with 2048 scans per sample and a spectral width of 19·99 ppm.
An overall recovery delay of 6·5 s was used which gave full relaxation. Data were processed using line broadening of 2·00 Hz prior to zero filling to 19428 points, Fourier transform and automatic baseline correction. Spectra were processed and analysed using TopSpin 3.2. The dcon function was used to deconvolute spectra in order to determine the integration of each resonance, in a similar manner to previous studies (28,29). The integration of each resonance was divided by the total integration for that spectrum, and assigned according to known shifts (14).
The integrations (as fractions of the total for that spectrum) were used for statistical calculations (n = 6 spectra).
Mass Spectrometry of Lipids-Samples were prepared and measurements taken in a similar manner to published methods (28,59-61). Raw data were processed using software by Kochen et al. (62) with some additional code (14).
The calibrated mass accuracy was 1 (ES+) milli mass units and the resolution was 140,000 for MS1 and 17,500 for MS2 spectra. Analyses were performed using Thermo Xcalibur 3.0.63. Original code for non-standard head groups was written in Matlab R2015b.
The lipid signals obtained were relative to the total, i.e. 'semi-quantitative' or relative abundance, with the signal intensity of each lipid expressed relative to the total lipid signal intensity, for each individual, per mille (‰). Raw high-resolution mass-spectrometry data were processed using XCMS (www.bioconductor.org) and Peakpicker v 2.0 (an in-house R script). Lists of known species (by m/z) were used, n = 1740 incl. standards (28,60,61,63). Signals that deviated by more than 5 ppm were ignored, and thus assignments were made on the basis of HRMS only.
Protein extraction-Whole cell protein extracts were prepared as described previously (6,64). Briefly, samples consisting of 2×10 7 cells were harvested and centrifuged, and the pellets washed (SCE buffer [100 mM sodium citrate, 2.7 mM EDTA-Na2, pH 7 (citric acid)], 1 mL). The pellets were frozen in liquid nitrogen (193K) and stored at -70°C. Extracts were subject to both Western immunoblotting and a kinase assay.
Western immunoblotting-Protein extracts were mixed with 5×SDS-PAGE sample buffer (250 mM Tris-HCl (pH 6.8), Kinase assay-The same number of cells from the same volume of culture was used; the cultures were not diluted during experiments. The cleared protein lysates (see above) were used immediately for the assay or were affinity purified by CrCKS1 beads as described (6) and incubated at 4°C for 2 h (64). Histone H1 kinase activity was assayed as previously (68)  Quantitative RT-PCR was performed in a Rotor-Gene RG-3000 (Corbett Science) under the following conditions: initial denaturation, 10 min at 95°C followed by 45 cycles of amplification (20 sec at 95°C, 20 sec at 60°C, 30 sec at 72°C). Each PCR reaction was performed in technical duplicate, differing by less than 5% between each other; the experiments were repeated at least three times with RNA isolated from independent cultures. To ensure that no primer-dimers were present, a melting curve was followed for each PCR. The results were normalized against 18S rRNA [30].
Transcriptomics and transcriptome de novo assembly-RNA from cells prior to CP (pre-CP), at the time when approximately 50 % of them reached the first CP (CP) and when approximately 50 % of the cells divided nuclei into two (M) was isolated the same way as for RT-qPCR. For each sample, three biological replicates were analysed. RNA quality was checked using Agilent Bioanalyzer and only high quality RNA was further processed. Libraries were prepared from 1 µg of total RNA using NEBNext Ultra™ Directional RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA) by poly(A) enrichment and pair-end sequenced using Illumina HiSeq 3000/4000 with read length 2 × 150 nt. Raw reads were adapter trimmed using CLC Genomics Workbench and reads from all conditions were used for de novo RNA assembly using the same software.
Statistical tests-Univariate statistical tests were done using Excel 2013 or 2016. Graphs were produced in OriginLab2018. The integrations of each resonance (as a fraction of the total for that spectrum) were used to calculate the means, standard deviations and student's t-tests reported for the abundance of lipids. The calculations were based on n = 6 biological replicates.

Author Contributions
MV co-designed the study, collected and analysed cell cycle data, advised on species type, co-developed handling procedures, designed experiments, supervised VL and co-wrote the manuscript. VL grew cultures, collected cells, acquired cell cycle data, prepared dried cell pellets and co-developed handling procedures. MČ ran all western blots and collected all qRT-PCR data. MJ acquired and converted all MS data and assisted in forging the collaboration. FR made the NMR instrument available, configured parameters of NMR experiments, collected all NMR data and ensured its quality. KB wrote the grant proposal that funded MV, VL and MČ, prepared and analysed the transcriptome and identified homologues of lipid metabolism genes, designed RT-qPCR and Western blot experiments, and co-wrote the manuscript. ØH wrote the original grant proposal that funded FR, SF and MJ, made equipment available, ensured data quality and. SF conceived the hypothesis, designed the study, extracted lipids and prepared all samples for lipid profiling, developed methods, analysed NMR and MS data and co-wrote the manuscript. All authors commented on the manuscript and approved the final version.

Conflict of Interest:
The authors declare no conflict of interest. 8. Scotchman, E., Kume, K., Navarro, F. J., and Nurse, P. (2021) Identification of mutants with increased variation in cell size at onset of mitosis in fission yeast.