Desiccation causes significant structural changes
Desiccation causes cellular damages even in anhydrobiotes . In budding yeast, studies have shown that desiccation causes changes of the cellular structure, such as cell wall, plasma membrane, nucleus, vacuole, lipid droplets, peroxisomes and mitochondria (for reviews, see [10-12]). Here, transmission electron microscopy (TEM) was employed to examine and compare the ultrastructure of desiccated yeast cells in log- (desiccation-sensitive) and stationary- (desiccation tolerant) phases. First, we examined the desiccation progress of yeast cells in the 15 ml glass tubes (see Method section for details). Cells in the tubes were weighted during the desiccation process for weight loss, which was not observed 3.5 days (~80 hours) after desiccation, indicating that cells were fully desiccated after 3.5 days of desiccation. To check the remaining amount of extracellular water, the 3.5-day-desiccated cells were placed in a 65°C oven for 1 hours, then to a 125°C oven for 15 minutes . A 7% weight loss was observed after oven incubation, suggesting that the desiccated cells contained about 7% extracellular water.
Staining with potassium permanganate, which provides enhanced staining of cellular membranes , revealed a typical ultrastructure of overnight-grown (log-phase) yeast cells without desiccation (Fig. 1A). Stationary cells (3-day culture) without desiccation had more mitochondria than log-phase cells and the cristae of these mitochondria were more prominent (Fig. 1B). Additionally, hydrated stationary-phase cells possess one or a few dark-stained vacuoles and the vacuoles are often associated with the nucleus, forming the nucleus-vacuole junctions , which function as sites for piecemeal microautophagy that degrades portions of the yeast nucleus during starvation  (Fig. 1B). When cells were stored in the desiccated state for 14 days, significant structural changes were observed in both log- and stationary phase cells. In dried desiccation-sensitive log-phase cells, vacuoles diminished in almost all cells. ER membranes became curled, forming small circular structure (whorls), especially near the plasma membrane. These whorls were much smaller that observed in stationary desiccated cells (see next section). The gap between the inner and outer nuclear membrane expanded significantly (Fig. 1C). This phenomenon was also observed in dividing cells during log-phase growth. Circular ER membrane structures were mostly observed in the mother, but not the daughter cell (Fig. 1D). Desiccation also caused dramatic structural changes in desiccation-tolerant stationary cells, including increased ER membrane and reduced vacuole number . These changes are the major focus of this study and are described in details in the following sessions.
Desiccation triggers endoplasmic reticulum (ER) stress and unfolded protein response (UPR)
It has been reported that misfolded and unfolded proteins trigger endoplasmic reticulum (ER) stress, which in turn, induces unfolded protein response (UPR) in yeast . When treated with the ER stressor dithiothreitol, the ER forms multi-layered, ring-shaped ER whorls [30-31]. ER whorls and whorl-like ER structure were also observed in stationary-phase yeast desiccated for 14 days (Figs. 2A, 2B), suggesting desiccation causes accumulation of unfolded/misfolded proteins , which induces ER stress, ultimately triggering UPR.
In budding yeast, UPR is mediated by the transmembrane endoribonuclease Ire1. Activation of Ire1 cleaves the mRNA of HAC1, which binds to a DNA sequence called UPR element to increase the expression of UPR target genes . To further examine whether the ER whorls are related to UPR, we compared the cross sections of wild type, ire1Δ and hac1Δ mutants desiccated for 14 days. About 200 cell cross sections of each strain (from 3 independent experiments, 60-70 cells per experiment) were examined. ER whorls were observed in about 13% of wild type cells, but no ER whorls were observed in ire1Δ or hac1Δ mutant (Figs. 2C, 2D), further suggesting the ER whorls observed in desiccated cells (Fig. 2A, 2B) were induced by UPR during desiccation.
Desiccation causes expansion of endoplasmic reticulum
ER stress induces expansion of ER membrane and this expansion requires UPR signaling . When 3-day stationary cells were desiccated for 14 days, a significant increase of ER membrane was observed in yeast cells with the ER membrane appearing to be disorganized and tangled around the nuclear membrane (Fig. 3A). This expanded and disorganized membrane was not observed in 3-day stationary cells prior to desiccation (Fig. 1B), nor in desiccated log-phase cells (Fig. 1C). These expanded membrane was different from the ER whorls, which were more organized and usually not associated with nuclear membrane. In desiccated stationary-phase cells, the ER membrane close to plasma membrane was often observed broken into small pieces, forming circular structures of 30-50 nm in diameter (Fig. 3B). One possibility is that the circular membrane structures are cross-sections of the ER whorls, or whorl-like structure observed in Figures 2A and 2B. These circular structure was rarely observed in ire1Δ or hac1Δ mutant.
Much less ER membrane accumulation, including both the disorganized membrane near nucleus and the circular membrane near plasma membrane, was observed in stationary-phase ire1Δ or hac1Δ mutant (Figs. 2C, 2D). We measured 6 cross sections from each strain (wild type, ire1Δ or hac1Δ) that contain the disorganized membrane near nucleus. It showed an approximately 40% less membrane accumulation in ire1Δ or hac1Δ mutant than in wild type. This reduced membrane expansion may be caused by blocking of UPR signaling in ire1Δ or hac1Δ mutant or be due to the fact that both ire1Δ and hac1Δ are involved in in phospholipid biosynthesis .
The nuclear and plasma membrane also underwent dramatic changes during desiccation in stationary-phase cells. Outward folding of the nuclear membrane was observed (Fig. 3C). This was not observed in desiccated log-phase cells and thus may be an adaptive response mounted by stationary-phase cells to cope with dehydration-induced cell shrinkage.
Another prominent feature of the 14-day desiccated stationary cells is the rupture of the nuclear envelop. The nuclear envelop opening was always associated with disorganized ER membrane (Figs. 3A, 3B). Similar to the wild type, broken nuclear membrane was also observed in desiccated stationary-phase hac1Δ and ire1Δ mutants (Figs. 2C, 2D). Examination of about 200 cells from TEM cross sections of each strain/growth condition (from 3 independent experiments, 60-70 cells per experiment) showed that 6-8% of cells displayed nuclear opening in desiccated wild type stationary cells or hac1Δ and ire1Δ cells, but not in desiccated log-phase cells (Figs. 1C, 1D).
Unfolded protein response does not affect the desiccation tolerance
We next checked the survival rates of yeast after 14 days of desiccation. The survival rate of the stationary wild type yeast was about 60%, but only 10% for log-phase cells. Interestingly, the survival rates of both ire1Δ and hac1Δ were slightly lower that the wild type, but the difference was not statistically significant (p > 0.05) (Fig. 4), suggesting that inactivation of UPR does not affect desiccation tolerance.
Vacuoles diminish during desiccation
The yeast vacuole is the largest organelle in a yeast cell and has similar functions to the lysosome in higher eukaryotes or the plant vacuole. In yeast the vacuole serves as the primary site for protein degradation and recycling, especially during starvation. Using TEM, we observed that hydrated 3-day stationary cells usually have one or a few dark-stained vacuoles (Fig. 1B). Examination of about 200 cells from TEM cross sections (from 3 independent experiments) showed that fourteen days after desiccation, over 90% of the stationary cells had no vacuoles (Figs. 2A-2B, 3). When observed, the vacuoles were lightly stained (Fig. 6A). Vacuoles were also absent in most of the log-phase desiccated cells (Figs. 1C, 1D).
Structural changes during the desiccation process
Next, we evaluated the ultrastructural changes during desiccation by examining stationary-phase yeast cells dried for 1, 2, 5 or 10 days. No significant difference was observed between cells dried for 1 or 2 days (Fig. 5A), in which nucleus was usually intact, similar to pre-desiccated cells, except that the mitochondrial cristae became less obvious. Vacuoles were mostly dark-stained, but light-stained spots were often observed inside the vacuoles (Fig. 5B), suggesting a possible initiation point for vacuole degradation. More light-stained vacuoles were observed in 5-day desiccated cells. Observation of about 200 cell cross sections from 3 independent experiments showed that engulfment of lipid droplet occurred in about 9-10% of either 2-day or 5-day-dried cells (Fig. 5C) and vacuoles were absent in more than 50% of 10-day desiccated cells and some lipid droplets were surrounded by ER membrane (Fig. 5D).
Endoplasmic reticulum (ER) and lipid droplets (LDs) dynamic in desiccated yeast cells
Lipid droplets (LDs) are ER–derived neutral lipid storage organelles . They are universally conserved in both prokaryotes and eukaryotes and their biogenesis primarily occurs from ER, where newly synthesized neutral lipids emerge from and are surrounded by a phospholipid monolayer . During desiccation, LDs were often observed to be associated with normal (Fig. 6A), disorganized (Fig. 6B), or circular ER membrane (Fig. 6C). It was observed that some of the circular structures had no membranes (arrow in Fig. 6C), suggesting that the ER membrane might be degraded and the free fatty acids were stored in LDs. If this is the case, we speculate that the overall LDs would increase during desiccation. To test this possibility, we checked the LD levels every 3 days for 15 days during desiccation. To our surprise, the LD level significantly decreased, rather than increased during the desiccation process (Fig. 6D, 7B). We reasoned that LDs might be consumed during desiccation to provide energy. This notion is supported by the TEM observation that LDs were engulfed during early stage of the desiccation by vacuoles (Fig. 5C), which degrade LDs by a process that resembles microautophagy [37-38].
Defects of lipid droplet synthesis reduce desiccation tolerance
To further test the possible role of LDs in desiccation tolerance, we examined yeast strains that are defective in LD formation. The yeast Pah1 is a homologue of mammalian lipin 1 protein, which plays important roles in glycerolipid biosynthesis. It regulates lipid droplet formation and nuclear/ER membrane growth . Deletion of PAH1 causes significant decrease of number of LDs . BODIPY staining revealed that the number of LDs in the 3-day stationary cells of pah1Δ was significantly less than that in wild type either prior to or after 14 days of desiccation (Fig. 7A, B). No significant change of LDs in pah1Δ mutant was observed after 14 days of desiccation. The survival rate of the pah1Δ mutant was significantly lower than the wild type (Fig. 7C). TEM observation revealed that after 14 days of desiccation, the LDs in pah1Δ cells lack clear boundaries and are usually smaller than those in wild type cells. (Fig. 7D). Measurement of 50 LD diameters from TEM cross sections showed that the average LD size/diameter of pah1Δ was about 1/3 of the wild type.
Pah1 is regulated by the Nem1-Spo7 phosphatase complex. Deletion of NEM1 or SPO7 partially inhibits the Pah1 activity and causes a phonotype similar to pah1Δ . When the nem1Δ and spo7Δ deletion mutants in stationary phase were dried for 2 weeks, we found that their survival rates were much lower than in wild type cells, but higher than that of the pah1Δ (Fig. 7C). BODIPY staining showed that the number of LDs in nem1Δ or spo1Δ was also in between the wild type and pah1Δ (Fig. 7A). TEM observation of the 14-day dried nem1Δ or spo1Δ cells showed a structure similar to pah1Δ, except that LDs were larger than that in pah1Δ (Fig. 7D-E). In all three mutants, much less ER membrane accumulation was observed.
Next, we used the antifungal agent cerulenin, which inhibits the biosynthesis of fatty acids and steroids, to reduce LD formation . In the presence of 4 µg/ml cerulenin, the number of LDs was significantly reduced (Figs. 7A, 7B). After 14 days drying, the survival rate of the cerulenin treated cells was significantly lower than the non-treated control (p < 0.01), similar to that of the pah1Δ mutant (Fig. 7B), further suggesting that accumulation of LDs is important to desiccation tolerance.