The auxotrophic yeast S. cerevisiae strain CEN.PK.113-5D (MATa ura3-52 HIS3, LEU2 TRP1 MAL2-8^c SUC2; kindly provided by Dr. P. Kötter, University of Frankfurt, Germany; Entian and Kötter, 2007) was used as a host strain in this study. The host strain was transformed with the plasmids for constitutive and equivalent expression of Aβ42 and Aβ40, under control of the GPD1 promoter: p416GPD-Aβ42 and p416GPD-Aβ40, respectively. The detailed construction of these plasmids and strains has been reported previously (Chen and Petranovic, 2015). Both Aβ constructs consist of the Kar2 signal sequence (42 amino acids) in front of Aβ42 or Aβ40 sequence, as described previously (Treusch et al., 2011).

Strains were grown in defined minimal medium as described previously (Chen and Petranovic, 2015), containing per liter: 10 g glucose, 5 g (NH₄)₂SO₄, 3 g KH₂PO₄, 0.5 g MgSO₄.7H₂O, 125 μl antifoam 204 (Sigma-Aldrich, USA), trace metals solution, and vitamins. Trace metals and vitamins solutions were prepared as described previously (Jensen et al., 2014). Batch fermentations were performed at 30°C in 1.2-l bioreactors (DasGip, Germany) with a working volume of 700 ml. Cultures were operated with 800 rpm agitation and 1 vvm gas flow of either pure dried air (aerobic) or dried air mixed with nitrogen gas to obtain a mixture containing 2% oxygen (micro-aerobic). Culture pH was measured with a pH sensor (Mettler Toledo, Switzerland) and maintained at 5.0 by automated addition of 2 M KOH or 2 M H₂SO₄. The CO₂ and O₂ concentrations in the exhaust gas were analyzed real-time with a GA4 gas analyzer (DasGip, Germany).

Biomass and extracellular metabolites were determined as described previously (Zhou et al., 2016). Biomass concentrations were measured as the optical density at 600 nm (OD₆₀₀) and cell dry weight (CDW). CDW was measured by filtering 5 ml of culture samples over a pre-weighted, pre-dried 0.45 μm nitrocellulose filter (Sartorius Stedim, Germany), microwave drying of the filter plus biomass and determining the increase in weight of the dried filter. During the different growth phases, OD/CDW ratios were constant (1.45 ± 0.04 OD/g l⁻¹). Extracellular glucose, ethanol, glycerol and acetate concentrations were analyzed on an Ultimate 3000 HPLC (Dionex, Sunnyvale, USA) equipped with an Aminex HPX-87H column (Bio Rad, USA). The column was eluted at 45°C using 5 mM H₂SO₄ at a flow rate of 0.6 ml min⁻¹.

Ten OD₆₀₀ of cells were harvested by centrifugation at 2,000 × g for 5 min at 4°C, and washed once in ice-cold distilled water. Cell pellets were resuspended in boiling 0.25 M Na₂CO₃ solution and processed as described previously (Parrou and François, 1997). Glycogen and trehalose were converted into glucose with amyloglucosidase (Sigma-Aldrich) and trehalase (Sigma-Aldrich), respectively, in acidic environment (pH 5.2). Glucose levels were determined enzymatically using a Glucose (HK) assay kit (Sigma-Aldrich).

Viability and intracellular ROS were measured by propidium iodide (PI, Thermo Fisher Scientific, USA) and dihydrorhodamine 123 (DHR123, Sigma-Aldrich, USA) staining respectively, as described previously (Chen and Petranovic, 2015). For the staining, 0.5 OD₆₀₀ of cells were taken at different phases and incubated with 0.5 μg ml⁻¹ of PI or 5 μM of DHR123 for 20 min. Cells were analyzed with a Guava flow cytometer (Merck, Germany) using a 488 nm laser for excitation. Fluorescence was detected with a 690/50 filter for PI or a 525/30 nm filter for rhodamine 123. Five thousand cells were analyzed for each sample. Two populations could be distinguished based on fluorescence intensity. Positively stained cells fall within the more fluorescent population (mean fluorescence 5–100 times higher). Results are shown as the fractions stained positively by PI or DHR123.

Five OD₆₀₀ of cells were harvested, washed twice in distilled water and resuspended in PBS before measurement. Oxygen consumption was measured at 30°C in a 1 ml temperature controlled closed chamber with a Clark oxygen electrode (Gilson) as described previously (Albers et al., 2007). One milliliter of PBS was added to the chamber and incubated until baseline (100% dissolved O₂) was stable. One OD₆₀₀ of cells were added to the chamber to measure the endogenous oxygen consumption rate. When the rate was stable, 25 mM of glucose was added to the chamber to obtain the initial oxygen consumption rate with glucose. Sodium thiosulfate (NaS₂O₃) was used for the zero-point calibration of the sensors. Oxygen consumption rates were determined from the slope of a plot of oxygen concentration vs. time and expressed as mM/OD/h.

One OD₆₀₀ of cells were harvested during PD. Mitochondria were stained with 1 μg ml⁻¹ Rhodamine 123 (Rh123, Sigma-Aldrich, USA) for 30 min and washed once in PBS. One OD₆₀₀ of cells was harvested during EX for lipid droplets measurement. Non-linear microscopy was applied to detect lipid droplets and mitochondria signals as described previously (Mertz, 2004). Two laser beams, 817 (5 mW) and 1,064 nm (6 mW) of synchronized pico-second pulse trains were spatially and temporally overlapped to be coupled into an inverted Nikon microscope (Eclipse TE2000-E microscope). CARS and TPEF signals were generated in live yeast cells to monitor lipid droplets and Rh123 stained mitochondria, respectively. The laser beams were focused on the samples using a 40x objective (Nikon Plan Fluor N.A. 1.3, working distance 0.21 mm). CARS and TPEF signals were collected in forward and epi-direction, respectively, with a high NA lens with a 661/20 and 514/30 or 609/45 filter into a single photon counting (SPC) photomultiplier tube (HPM-100-40, Becker & Hickl GmbH). Z-stacks were acquired for each sample with 0.2 μm step size for a maximum of 10 μm sample thickness. Each single acquisition in the z-stack was taken in 1 sec with a resolution of 256 pixels for a 34.4 μm field of view. The images were opened in ImageJ and then processed in Imaris (Bitplane software for image processing). The lipid quantification for control and Aβ42 expressing cells was done by ImageJ and GIMP (GNU Image Manipulation Program). Statistics was performed for 10 images of each strain. The quantities were determined from the CARS channel in which lipids were detected. The channel was turned into a binary color format (0 and 1) in GIMP first for lipids and their area in each slice (Li, i = slice number) was calculated using ImageJ. Subsequently cell areas in each slice (Ci, i = slice number) were also calculated. The ratio of the summation of lipid area / summation of cell area gave the value of lipid content in a cell as shown in Equation 1. The distance between slices (Δz) was canceled out since it was constant for all cells.

Samples for microarray analysis were taken from duplicate cultures during EX, PD, SP1, and SP2, respectively. Samples of cells were frozen rapidly in liquid nitrogen to prevent mRNA turnover (Piper et al., 2002). Total RNA was extracted using the RNeasy Mini Kit (QIAGEN, Germany) with a FastPrep homogenizer (MP Biomedicals, USA) to disrupt cells. Quality of total RNA was assessed by an Agilent 2100 bioanalyzer (Agilent Technologies, USA). Further RNA preparation and hybridization of biotin-conjugated aRNA fragments to Yeast Genome 2.0 Arrays (Affymetrix GeneChip, USA) was performed by the Bioinformatics and Expression Analysis core facility (BEA) of the Karolinska Institute, Sweden. Microarray data are deposited at the Genome Expression Omnibus website (GEO, http://www.ncbi.nlm.nih.gov/geo/) with series number GSE94793. Raw RNA data (CEL files) were preprocessed by Bioconductor and R version 3.2.3. The PIANO package was used for gene ontology (GO) terms and gene set analysis (GSA; Väremo et al., 2013). Only gene sets significantly enriched by distinctly up or down-regulated genes (p < 0.05) were considered in this study. The KEGG pathway and Gene Ontology functional categories enrichment analysis were used to investigate the transcriptional regulation of lipogenesis by using online software tool David (Database for Annotation, Visualization, and Integrated Discovery, http://david.abcc.ncifcrf.gov; Huang et al., 2008). Adjusted P < 0.05 and fold changes <0.81 or >1.2 were used as thresholds to identify significantly differentially expressed genes. Pheatmap package (cran.r-project.org/web/packages/pheatmap/index.html) was used to generate clustered heatmaps of gene sets.

qPCR was performed as previously described (Liu et al., 2016). cDNA was synthesized from 1 μg of total RNA using the QuantiTect Reverse Transcription Kit (QIAGEN, Germany). Two microliters of synthesized cDNA were used as the template for qPCR with the DyNAmo Flash SYBR Green qPCR kit (Thermo Fisher Scientific, USA). Housekeeping gene ACT1 was used as a reference gene to normalize RNA levels. Used primer sets are listed in Table S4.

Lipid extraction and separation were performed as described previously (Khoomrung et al., 2013). Ten micrograms of freeze-dried cells were mixed with 7 ml of chloroform:methanol (2:1, v/v) solution in extraction tubes containing 50 μg of cholesterol as internal standard. Each tube was vigorously vortexed and placed in microwave reaction vessel (12 × 3cm I.D., 0.5 cm thickness, Milestone Stard D, Italy) which was heated to 60°C within 6 min and kept at 60°C for 10 min. After the sample was cooled down to room temperature, 1.7 ml of NaCI was added and the tube was vortexed vigorously. Thereafter, the sample was centrifuged at 1,912 g for 10 min and the organic phase was transferred into a clean extraction tube. Organic solvent was evaporated and residues dissolved in 200 μl of chloroform:methanol (2:1, v/v) for HPLC-CAD analysis (Corona, USA). Lipid peaks were identified based on the spectrum of a standard mix (SE, TAG, CH, ES, PA, CL, PE, PC, PS, and PI). Quantification of lipids was performed using serial dilutions of the standard mix from concentrations of 10–1,000 μg ml⁻¹. The average log₁₀ of peak area was plotted against log₁₀ of concentration. Correlation (r²) was determined for all standard curves by linear regression.

Significance of differences observed in physiological parameters between strains were determined using two-tailed, student t-tests. Cultures were independent replicates and, as parameters were determined using identical procedures, student t-tests were performed as two-sample with equal variance. Unless specified explicitly, three independent replicate cultures of the Aβ42 and control strains and two independent replicate cultures of the Aβ40 strain were analyzed under both fully aerated and micro-aerobic conditions. P < 0.05 was considered to indicate significant differences. Values are represented as mean ± SEM.

To determine the effects of human Aβ peptides on S. cerevisiae growth characteristics, strains constitutively expressing either human Aβ42 or Aβ40 peptide (hereafter referred to as Aβ42 and Aβ40 strain) or carrying the empty vector (control strain) were grown in well-aerated bioreactors using glucose as carbon-source (Figures 1A,B). Physiological parameters from these cultures are presented in Table 1. The maximal specific growth rate of the Aβ42 strain was 17% lower than that of the control and Aβ40 strains (p < 0.0003), in accordance with our previous report (Chen and Petranovic, 2015). The Aβ42 strain also showed a less pronounced but significantly decreased maximal glucose uptake rate (p = 0.02), which was reflected by 15% reduced biomass yield during EX (p < 0.02). Maximal ethanol production rates were not significantly altered (p > 0.4) during EX. On the contrary, we observed substantially increased production of two fermentation products: glycerol and acetate (Table 1). Glycerol yields were increased in both Aβ strains (p < 0.01), but was highest in the Aβ42 strain (2.7-fold compared to control strain, p < 0.00005). Despite the well-aerated conditions, the Aβ42 strain also produced more acetate than control and Aβ40 strains (1.28- and 1.21-fold compared to control and Aβ40 strains, respectively, p < 0.01). Overall these results suggest an altered redox-cofactor balancing in the heterologous Aβ42 expressing strain (Bakker et al., 2001). During SP, cells rely heavily on the storage carbohydrates glycogen and trehalose as energy sources. No significant differences in the accumulation of glycogen were observed among these strains (Figure S1A). Levels of trehalose were significantly higher in the Aβ42 strain during PD and SP1 (when extracellular carbon source was depleted; Figure S1B), which may be linked to its alternative role in stress-resistance. Trehalose can stabilize proteins in their native folding during heat shock and reduce the aggregation of denatured proteins in yeast cells (Singer and Lindquist, 1998). Reduced viable cell fractions might explain reduced growth and substrate consumption rates in the Aβ42 strain. However, viability analysis showed that fractions of dead cells were low for all strains during different phases (<2.5%). The slightly, yet significantly, elevated fractions of dead cells in Aβ42 cultures (p < 0.007, Figure S2A), did not fully explain the observed differences in fermentation kinetics.

Mitochondria are pivotal to the survival of cells, including neurons, due to their role in energy metabolism. The increased glycerol production in the Aβ42 strain suggested that the surplus of NADH produced in anabolism, could not be completely reoxidized by mitochondrial respiration (Table 1). Despite the differences in glycerol and acetate production, the maximal oxygen uptake rates during the fully respiratory PD phase were surprisingly similar for both Aβ strains, yet lower than control strain (Table 1). These observations suggested that not consumption of oxygen per se, but rather its usage was affected specifically by Aβ42 expression. Together with our previous observations that the respiratory rate of the Aβ42 strain was significantly reduced in ethanol-grown cultures (Chen and Petranovic, 2015), this suggests that the toxicity triggered by Aβ42 might involve reduced mitochondrial functionality. To further test this, we evaluated the respiratory capacity of all strains during different growth phases. During EX no significant differences were observed between strains, which matched with fermentation being the main catabolic pathway at this phase. However, during the subsequent phases (PD and SP1), the respiratory capacity of the Aβ42 strain was significantly decreased compared to both other strains (p < 0.05, Figure 2A). Impaired mitochondrial function can lead to excess reactive oxygen species (ROS) production which contributes to AD pathology (Livnat-Levanon et al., 2014). In the Aβ42 strain, the ROS-positive fractions were significantly increased during phases of high metabolic activity, i.e., EX and PD, compared to the other strains (p < 0.003, Figure S2B). To investigate if these physiological indications of mitochondrial dysfunction coincided with aberrant mitochondrial structures, we stained the mitochondria with Rhodamine 123 and employed TPEF (two-photon excited fluorescence) microscopy to investigate the morphology in fully respiratory growing cells. Control cells exhibited continuous mitochondrial structures (networks), whereas Aβ42 expressing cells displayed a more fragmented structure of mitochondria (Figure 2C and Videos S1, S2). Such fragmented structures of mitochondria were previously observed in aged yeast cells and could enhance the turn-over of damaged mitochondria by mitophagy (Mao and Klionsky, 2013; Breitenbach et al., 2014).

The affected mitochondrial functionality is likely to have a more severe impact on cells under oxygen-limited conditions. Oxygen limitation is common due to increased oxygen-requirements during PD in less-controlled cultivations, such as shake-flasks, in which oxygen-transfer rates are lower (Anderlei and Büchs, 2001). We also noticed this effect by the differences in viability of the same Aβ42 strain reported here (Figure S1) and previously in shake-flasks (Chen and Petranovic, 2015). In standard bioreactor cultures, the dissolved oxygen levels were maintained above 30% of saturation at all time, but to test the combinatorial effect of oxygen limitation and Aβ peptides expression on cell physiology, oxygen-limited cultivations were also performed. To this end, the oxygen concentration in the inflowing gas was reduced to 2%, leading to dissolved oxygen levels below 1% of saturation. The reduced dissolved oxygen levels only reduced the maximal growth rate of the Aβ42 strain on glucose significantly (8% reduction, p < 0.05) and did not significantly affect growth of control and Aβ40 strains (Figure 2B). Similarly, total biomass yields were only significantly reduced in the Aβ42 strain (13% reduction, p = 0.05; Table S1). Under well-aerated, standard conditions, we observed differences in maximal oxygen uptake rates between Aβ42 and control strains (Table 1). When the oxygen supply was severely limited, oxygen uptake rates were also restricted and differences between strains were absent (Table S1). However, despite these equal oxygen consumption rates, the Aβ42 strain produced less carbon dioxide as shown by the almost one-third reduction in respiratory quotient (defined as the ratio of carbon dioxide produced/oxygen consumed) compared to control strain (p < 0.05, Table S1). This indicates an altered usage of the oxygen consumed by the Aβ42 strain, likely due to less efficient respiration. In addition to these effects on physiology, oxygen limitation also resulted in a doubling of the fraction of dead cells in all strains (<6%; Figure S2A). The ROS-positive fractions were similar compared to well-aerated conditions for all strains during EX and PD, but lower during SP, likely due to reduced oxygen availability (Figure S2B).

Oxygen limitation may thus contribute to enhanced cell death, however the drastic effect of Aβ42 expression on CLS observed in shake-flask cultures (Chen and Petranovic, 2015) was not observed in oxygen-limited bioreactor cultures. Only 5.25 ± 0.44 and 2.39 ± 0.64% of cells were identified as dead in Aβ42 and control cultures, respectively, during SP2, i.e., 2 days after extracellular carbon source exhaustion (Figure S2A), in contrast to 37.15 ± 1.21 and 14.06 ± 1.82%, respectively, in shake-flask cultures (Chen and Petranovic, 2015). A putative explanation could be the different glucose concentrations in the medium used for bioreactors and shake-flask cultivations, i.e., 10 g L⁻¹ instead of 20 g L⁻¹, respectively. Shake flask cultures grown in medium containing 10 g L⁻¹ of glucose which is identical to the medium used in bioreactor cultures, revealed indeed a strong effect on viability. The higher initial glucose concentration resulted in a strongly reduced CLS, i.e., the fractions of dead cells were 5–7 fold higher during SP2 for the different strains (Figures S3A,B). Both initial glucose concentrations are generally not considered to induce calorie restriction related hormesis effects, but strongly affect other culture parameters in stationary phase. When using 10 g L⁻¹ of glucose the pH dropped to only 4.5, instead of to 3.1, when 20 g L⁻¹ of glucose was used (Figures S3C,D). Final cell concentrations correlated with initial glucose concentrations (Figures S3E,F). The acidic environment poses a strong stress on cells and can result in reduced CLS. An additional confounding factor were higher acetate levels in shake-flask cultures, which were highest for the Aβ42 strain on 20 g L⁻¹ glucose (Figure S4). Altogether oxygen limitation, low pH-values and high acetate concentrations may contribute to the strongly reduced viability and CLS of all strains and especially of the Aβ42 strain in less-controlled shake-flasks environment. These effects are smaller in bioreactor cultures, because oxygen is not limited and a decrease in pH is prevented by online-monitoring and automated titration of base.

To understand the mechanisms behind the phenotypic changes in Aβ42 and Aβ40 strains, gene expression levels during EX, PD, SP1, and SP2 were quantified using microarrays. The global expression pattern was characterized using principal component analysis (PCA). PCA resulted in strong grouping of biological replicates, indicative of a high degree of reproducibility. The first and second PCA components clearly separated all of samples from different growth phases by their gene expression profiles (Figure 3A and Figure S5). To further examine the extent of changes in each phase, we performed pair-wise comparisons between the strains at each time-point. The expression of 472, 394, 391, and 280 genes was significantly different (adjusted p < 0.001) during EX, PD, SP1, and SP2, respectively, between the Aβ42 strain and the control strain (Figure S6A). Although the physiology of the Aβ40 strain was not notably different from the control strain, expression of 89, 486, and 1,118 genes was significantly different (adjusted p < 0.001) during EX, PD, and SP1, respectively. No significantly changed genes were identified later in SP2 between Aβ40 and control strains (Figure S6B). Of note, the overlap in differentially expressed genes was <20% between Aβ42 and Aβ40 strains during all phases, suggesting that different biological processes were influenced (Figure S6C).

To gain more insight in biological processes affected by Aβ42 and Aβ40 expression, gene set analysis (GSA) was performed on the significantly differentially expressed genes. Significant enrichment of 126 and 129 gene sets was identified (p < 0.005) among genes differentially expressed in Aβ42 and Aβ40 strains, respectively compared to the control strain. The complete list of gene sets can be found in Figures S7, S8. The most significantly enriched gene sets are shown in Figure 3B for both Aβ strains. Gene sets associated with protein processing (post-translational modifications, transport, and degradation), such as, “endoplasmic reticulum (ER),” “protein glycosylation,” “vacuole” were enriched with up-regulated genes during EX, PD, and SP1 in the Aβ42 strain, but only during EX in the Aβ40 strain. Gene sets related to protein synthesis such as, “ribosome,” “translation,” “mitochondrial translation,” and “cellular amino acid biosynthetic process,” were enriched with down-regulated genes in the Aβ42 strain during EX and PD. Remarkably, these genes were up-regulated in the Aβ42 strain compared to the control strain during SP. The expression profiles of these protein synthesis related genes were studied throughout the different culture phases. This revealed that for genes, whose expression was higher in the Aβ42 strain than control strain in SP, this was due to a less pronounced decrease in expression (data not shown). These protein synthesis-related gene sets were enriched with up-regulated genes in the Aβ40 strain especially in PD. This opposite regulation of these genes clearly indicated different responses to Aβ42 and Aβ40 expression. For illustration, the genes involved in amino acid biosynthesis pathways of which expression levels were significantly changed can be found in Figure S9 and Table S2. In addition to these processes directly involved in protein synthesis and processing, transcription related gene sets such as, “nucleic acid binding,” “protein import into nucleus,” “ribonucleoprotein complex,” and “mRNA transport” were enriched with down-regulated genes in both Aβ strains.

As a key component of cellular proteostasis, the ER is responsible for processing of one-third of cellular proteins and harbors an elaborate protein quality control system (PQC) to eliminate misfolded proteins by degradation (Cao and Kaufman, 2012). However, an overload of the PQC machinery results in an ER stress response (ESR) and activates the unfolded protein response (UPR; Scheper and Hoozemans, 2015). In the Aβ42 strain, most genes involved in “response to stress” were significantly higher expressed, including HAC1, the key regulator of UPR. In response to ER stress, HAC1 (HAC1^u) mRNA is spliced to HAC1 (HAC1^s) to initiate synthesis of the active transcription activator Hac1p which induces expression of over 300 of UPR target genes (Travers et al., 2000). The ratio of HAC1^s/HAC1^u was 19, 17, and 13-fold higher in the Aβ42 strain compared to the control strain during EX, PD, and SP1, respectively (Figure S10A). Induction of UPR target genes results in the biosynthesis of chaperones, and other factors involved in the secretory pathway, and ER associated degradation (ERAD) to restore ER homeostasis. The gene sets related to protein processing were strongly triggered upon expression of Aβ42 (Figure 4). Genes encoding “folding” chaperones such as, KAR2, SIL1, SCJ1, JEM1, and disulfide bond formation enzymes PDI1 and ERO1 were indeed all found significantly up-regulated (Figure 4). Higher transcription levels of PDI1 and ERO1 were verified by qPCR (Figures S10B,C). Genes encoding proteins involved in processes throughout the secretory pathway including translocation, glycosylation, GPI (Glycosylphosphatidylinositol) biosynthesis, trafficking between ER and Golgi, ER associated degradation (ERAD), and vacuole were also significantly higher expressed in the Aβ42 strain (Figure 4). Besides ERAD, which is the predominant cellular mechanism to degrade misfolded proteins under ER stress, autophagy can also be activated when the amount of misfolded proteins exceeds the ER capacity (Bernales et al., 2006). In line with this, the Aβ42 strain expressed the autophagy related genes ATG8, ATG22, ATG34, NVJ1, and ATG19 to a higher extent (Figure 4).

In the Aβ40 strain, changes in expression levels of genes involved in protein processing were less pronounced (Figure S11). The ratio of HAC1^s/HAC1^u in the Aβ40 strain was only 5.8-fold higher than control strain in EX. qPCR confirmed that transcript levels of PDI1 and ERO1 were not significantly changed (Figure S10). The distributions of significantly differentially expressed genes between Aβ42 and Aβ40 strains are shown in Figure S12 (p < 0.05). All differentially expressed genes involved in protein processing are listed in Table S3.

It was previously shown that ER stress and activation of UPR pathways play a critical role in lipid metabolism in different model organisms (Hetz, 2012). GSA revealed that the gene sets related to “lipid metabolic process” were overrepresented with up-regulated genes in both Aβ strains during EX. More specifically, gene sets involved in “glycerolipid biosynthesis process,” “phospholipid biosynthetic process,” “glycerophospholipid biosynthetic process,” and “lipid biosynthetic process” were strongly overrepresented among up-regulated genes in the Aβ42 strain during EX, compared to the control strain. Similar results were also found for the Aβ40 strain. However, for the Aβ40 strain, the enrichment was less strong and the amounts of significantly up-regulated genes involved in each processes were, respectively, 30, 34, 31, and 42% lower than for the Aβ42 strain (Figure 5A). The expression level of INO1, the gene encoding an important regulator in lipid metabolism in yeast (Henry et al., 2012), was significantly increased in the Aβ42 strain during EX (data not shown). To see whether the different expression patterns led to alterations in cellular lipid composition, we measured different lipid classes dynamically in all three strains. The major lipid constituents of S. cerevisiae: storage lipids, phospholipids, and sterols were analyzed. Compared to the control strain, the levels of all lipid categories were significantly increased during EX and PD in the Aβ42 strain. The differences were most pronounced during EX. In the Aβ40 strain, only phospholipids showed a slight but significant increase during EX and PD compared to control strain (Figure 5B). To gain insight in the structural storage of lipids we monitored the three-dimensional distribution and amounts of lipids stored at the single-cell level (which usually coalesce into lipid droplets) by CARS (coherent anti-Stokes Raman scattering) microscopy in exponentially growing cells. CARS can probe and image lipid structures by vibrations between carbon and hydrogen bonds and hence no fluorescent labels or tags are needed. The average lipid droplet volume was 1.6-fold larger in the Aβ42 strain (0.078 ± 0.032 μm³) than in the control strain (0.048 ± 0.028 μm³; Figure S13). CARS images showed that the lipid droplets in control cells were numerous and smaller in sizes, whereas they tended to accumulate and form larger drops in Aβ42 expressing cells (Figure 5C and Videos S3, S4).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.