Aggresome-Like Formation Promotes Resistance to Proteotoxicity in Cells from Long-Lived Species

Abstract

The capacity of cells to maintain proteostasis declines with age, causing rapid accumulation of damaged proteins and protein aggregates, which plays an important role in age-related disease etiology. While our group and others have identified that proteostasis is enhanced in long-lived species, there are no data on whether this leads to better resistance to proteotoxicity. We compared the sensitivity of cells from long- (naked mole rat [NMR]) and short- (Mouse) lived species to proteotoxicity, by measuring the survival of fibroblasts under polyglutamine (polyQ) toxicity, a well-established model of protein aggregation. Additionally, to evaluate the contribution of proteostatic mechanisms to proteotoxicity resistance, we down-regulated a key protein of each mechanism (autophagy—ATG5; ubiquitin-proteasome—PSMD14; and chaperones—HSP27) in NMR fibroblasts. Furthermore, we analyzed the formation and subcellular localization of inclusions in long- and short-lived species. Here, we show that fibroblasts from long-lived species are more resistant to proteotoxicity than their short-lived counterparts. Surprisingly, this does not occur because the NMR cells have less polyQ82 protein aggregates, but rather they have an enhanced capacity to handle misfolded proteins and form protective perinuclear and aggresome-like inclusions. All three proteostatic mechanisms contribute to this resistance to polyQ toxicity but autophagy has the greatest effect. Overall, our data suggest that the resistance to proteotoxicity observed in long-lived species is not due to a lower level of protein aggregates but rather to enhanced handling of the protein aggregates through the formation of aggresome-like inclusions, a well-recognized protective mechanism against proteotoxicty.

Proteostasis involves multiple protein quality control processes that work together to ensure the health of the cellular proteome (1). A decline in proteostasis with age, a hallmark of aging (2,3), leads to the accumulation of misfolded proteins and protein aggregates, leading to proteotoxicity (4–9).

Using a comparative biology approach, we and others have shown that long-lived species have more robust proteostasis mechanisms, including chaperones, autophagy, and the proteasome (10–14). Similarly, protein fractions from long-lived bats show low levels of protein ubiquitination, and resistance to urea-induced protein unfolding relative to those observed in similar fractions from mice (15). Moreover, studies by Triplett et al. have shown that naked mole rats (NMRs) maintain high levels of autophagy throughout most of their life span (16). Finally, studies using primary skin fibroblasts from a large set of long- and short-lived species show that there is a correlation between species longevity and the levels of insoluble proteins (17), suggesting that long-lived species might have better protection against protein misfolding and less protein aggregation.

Together, these data suggest that long-lived species might have very efficient mechanisms for maintenance of protein homeostasis. However, there are no data confirming that enhanced proteostasis in long-lived species actually protect cells from proteotoxicity. In this study, we directly tested the ability of cells from long-lived species to respond to polyQ-YFP-induced proteotoxicity. This is a well-established and widely used model to study protein aggregation (18,19), that is based on an insertional mutation in genes that generates the expansion of repeats of the polyglutamine tract (polyQ) in their coding regions. PolyQ tracts coded by the cytosine-adenine-guanine (CAG) repeats are found in many proteins. There are usually less than 35–40 CAG repeats; however, an insertional mutation that causes the number of glutamines to exceed the threshold of 35–40 repeats leads to protein misfolding and aggregation (20).

Expansions of polyglutamine (polyQ) tracts in unrelated proteins is known to be a cause of nine neurodegenerative diseases which include Huntington’s disease, dentatorubropallidoluysian atrophy, spinal and bulbar muscular atrophy, and six autosomal dominant forms of spinocerebellar ataxia (SCA1,2,3,6,7 and 17) (21). One of the first studies to use the fluorescently tagged polyQ was to understand the molecular mechanisms of CAG-repeated expansion diseases. These experiments were performed in COS-7 cells (22). This was later used in models such as Caenorhabditis elegans to study the threshold of aggregate formation, the effect of aging and treatments on protein aggregation, and proteostasis mechanisms in general (19,23,24). Hence using this fluorescently labeled polyQ enables to visualize and study the aggregate formation.

To evaluate the hypothesis that enhanced protein homeostasis might be the main contributor to the reduced level of protein aggregates observed in fibroblasts from long-lived species, we transfected skin fibroblasts from two long-lived species (NMR and little brown bat [LBB]), and two phylogenetically related short-lived species (mouse and evening bat) with both polyQ19-YFP, which does not aggregate and polyQ82-YFP, which aggregates and examined cell survival, cellular distribution of polyQ82 inclusions, and levels of protein aggregates (inclusions). We found that while resistance to proteotoxicity is superior in long-lived species, it was surprisingly not due to differences in the levels of protein aggregation. Instead, we observed that cells from long-lived species form aggresome-like inclusions, with a perinuclear cellular location, next to centrosomes. These aggresome-like inclusions were less abundant in the short-lived species. Furthermore, our data indicate that inclusions of polyQ82 start to appear at lower concentrations in long-lived species, compared with short-lived species, suggesting that inclusion formation is a protective response to sequester toxic misfolded proteins and oligomers.

Overall, our data show that cells from long-lived species are more resistant to proteotoxicity, and that this resistance is due to enhanced handling of the protein aggregates through the formation of aggresome-like inclusions, a protective mechanism against proteotoxicty (25–28).

Experimental Procedures

Cell Culture

Skin fibroblasts from short-lived species and long-lived species belonging to the same phylogenetic clade and having similar body size were used. For rodents, we used laboratory mice (Mus musculus, 35 g and maximal reported life span [MLS] ~4 years), versus NMRs (Heterocephalus glaber, 30 g and MLS ~30 years) and for bats, we used evening bat (EB) (Nycticeius humeralis, 11 g and MLS ~6 years) versus little brown bat (LBB) (Myotis lucifugus 8 g and MLS~34 years). Mouse skin fibroblasts were from C57BL/6 mice, while fibroblasts from NMR were from the Comparative Biology of Aging Core from the Nathan Shock Center at San Antonio. Bat cells (from young wild-caught animals) were obtained from the Comparative Biology of Aging Core from the Nathan Shock Center at University of Alabama at Birmingham. Briefly, primary skin-derived fibroblast cell lines derived from young animals (no breeders; males and females), were prepared by enzymatic digestion of skin, and cultured in low-glucose Dulbecco’s modified Eagle’s medium and 10% cosmic calf serum (Hyclone Laboratories, Logan, UT) in a 37°C (17,29) incubator with a gas phase of 21% O₂, 5% CO₂. NMR and mouse fibroblasts were cultured at 35°C (30). Temperature was chosen based on the previous literature in the field (17,29,30). For NMR fibroblast cultures, we used Biocoat collagen I-coated tissue culture dishes (Advance Biometrix, San Diego, CA) (11). All cells were used at passages below 10 and all comparisons were made between similar passage cells.

Plasmids and Cell Transfection

pEYFP-N1 Q19-YFP, and pEYFP-N1 Q82-YFP plasmids were kind gifts from Dr. Richard Morimoto. The plasmids pEYFP-N1-Q19 and pEYFP-N1-Q82 contain the polyglutamine-encoding sequences, cloned as a cassette into the EcoRI site of pEYFP-N1 (CLONTECH) resulting in the expression of yellow fluorescent protein (YFP) (23,31). Thermo Fisher Scientific’s Neon Electroporation System was used for transfection. Fibroblasts were collected by trypsinization at 70% confluency. Cells were centrifuged to remove the media, washed with DPBS, and the cell pellet was collected in resuspension buffer (R buffer). This cell suspension was mixed with the plasmid of interest and was electroporated. Cells were seeded in the respective cell culture dishes with preincubated media immediately after electroporation.

Immunofluorescence

Cells were seeded on collagen-coated Zeiss no. 1.5 cover slips in six-well plates, washed with DPBS twice, and fixed with either methanol-free 4% formaldehyde or ice-cold methanol. Cells were incubated with primary antibody overnight at 4°C. This was followed by incubation with the respective secondary antibodies for 1 hour at room temperature. Cover slips were mounted on microscope slides using Prolong Gold Antifade Reagent. A Zeiss LSM 780 confocal microscope was used for imaging. Primary antibodies used for immunofluorescence were γ-tubulin (c-20) rabbit pAb (Santa Cruz-7396-R), Vimentin(D21H3) XP rabbit mAb (Cell Signaling Technology-5741), Lamin A/C (4C11) mouse mAb (Cell Signaling Technology-4777), p62/SQSTM1(D6M5X) rabbit mAb (rodent specific, Cell Signaling Technology 23214), p62/SQSTM1 rabbit polyclonal antibody (Thermo Fisher Scientific, PA5-20839), HSP 27 (rodent preferred, Cell Signaling Technology 2442), and HDAC6 rabbit pAb (Abclonal A11429). Anti-rabbit IgG (H+L), F(ab’)2 Fragment (Alexa Fluor 647 Conjugate) from Cell Signaling Technology (4414), and Anti-mouse IgG (H+L), F(ab’)2 Fragment (Alexa Fluor 647 Conjugate) from Cell Signaling Technology (4410) were used as secondary antibodies.

High Content Imager

The Image X press Micro wide field high content imaging (HCI) system, a fully-automated machine, was used to capture images of live cells. Live cells can be imaged over multiple days, as this imager comes with an incubator and an environmental control system to set the temperature (35°C or 37°C), CO₂ supply (5%), and humidity. Transfected cells were seeded in Greiner glass 96-well plates and time-lapse video microscopy was performed using the HCI system. The acquired images were analyzed using Fiji-ImageJ software. Mean cell fluorescence was calculated as in Bersuker et al. (2013) (32). Live cell images in which cells start to form polyQ82-YFP inclusions were selected. The outline of the cell was delineated using Fiji-Image J software and total cell fluorescence was calculated, which was divided by cell area to derive mean cell fluorescence at which inclusions started to form (10 hours after transfection). The number (percentage) of inclusions was calculated by counting the number of cells with inclusions divided by the total number of cells (around 500 cells per species) 36 hours after transfection.

Cell Survival

Using time-lapse cell imaging, we measured cell survival by following individual transfected cells. We considered a cell to be dead when it underwent rounding, shrinking, and/or disintegration. We followed cultured fibroblasts for 72 hours and the deaths were annotated over that time. The data were plotted as a survival curve using GraphPad Prism software and analyzed by log-rank test (Mantel-Cox test) where each curve was compared to its respective scrambled siRNA control.

Location of Inclusions

Inclusions were classified based on the location with respect to the nucleus: perinuclear—inclusions located within 2 µm of the nuclear membrane, frequently forming dents in the nuclear membrane versus far—inclusions which are located more than 2 µm from the nuclear membrane. Similarly, inclusions were classified based on their location with respect to the centrosome: peri-centrosomal—located within 2 µm of centrosome, and far—located greater than 2 µm from centrosome (25).

ATG5, PSMD14, and HSP27 SiRNA Knock Down

siRNA targeting HSP27, PSMD14, and ATG5 were purchased from Thermo Fisher. Scrambled siRNA was used as control. Lipofectamine RNAiMAX transfection reagent (Thermo Fisher) was used to deliver siRNA (27.3 pmol siRNA with 6 µL RNA imax). Forty-eight hours after siRNA transfection, cells were transfected overnight with polyQ19 or polyQ82. Cells were monitored by HCI as described previously. The silencing efficacy was determined by western blotting using antibodies against HSP27 (Cell Signaling Technology 2442), PSMD14 (Thermo Fisher 38–0200), and ATG5 (Thermo Fisher PA5-35201), respectively.

Statistics

The survival curves were analyzed by GraphPad Prism software using log-rank (Mantel-Cox) test. Hazard ratios were calculated using GraphPad Prism software for two survival curves, which is a measure of how often an event, in this case death of the cell, happens in one group compared with how often it happens in another group over time. Here, we report hazard ratios as significant if the lower limit of the 95% confidence interval of the hazard ratio is greater than 1. We also reported median cell survival, the time at which 50% of the cells are alive. For all other comparisons, a parametric Student-unpaired t test was used in GraphPad Prism, and similar standard deviations were assumed for compared populations. All these tests were two-tailed, and a 95% confidence level was used to identify significant difference. The data represent three independent experiments with cells obtained from three different animals, and each experiment was done in duplicate.

Results

NMR Fibroblasts Are More Resistant to polyQ-Induced Proteotoxicity Than Those of Mouse

To determine whether cells from long-lived species are more resistant to polyQ82 proteotoxicity, we measured cell survival by live cell imaging (Figure 1A and B). We observed no significant differences in cell survival between mouse and NMR fibroblasts transfected with the nonaggregative control, polyQ19. In the presence of the aggregative polyQ82, mouse fibroblasts displayed a significant decrease in survival (p < .0001; Figure 1A), while no significant differences between polyQ19 and polyQ82 were observed in NMR cells (p = .1469; Figure 1B). The hazard ratio for survival of mouse cells was 7.2, indicating that mouse cells transfected with polyQ82 were 7.2 times more likely to die than control (mouse cells transfected with polyQ19), whereas for NMR the hazard ratio was 1.9. Furthermore, our data show that median survival time of mouse cells transfected with polyQ82-YFP was 20 hours, while for NMR, it was 70 hours. The percentage survival time (relative to polyQ19) at 72 hours after transfection was 11% and 56% for mouse and NMR cells, respectively (Figure 1C). These data demonstrate that NMR fibroblasts are more resistant to proteotoxicity induced by polyQ82-YFP than are mouse fibroblasts.

Figure 1.

Naked mole rat (NMR) fibroblasts were more resistant to proteotoxicity induced by polyQ82 as compared to mouse fibroblasts. Fibroblasts of mouse and NMR transfected with polyQ19-YFP (triangle; solid line) and polyQ82-YFP (circle; dash line) were followed using the live cell time-lapse imaging. Cell death was identified by collapsing of the cell and disintegration. Cell survival plotted as survival curve for mouse (A) and NMR (B) fibroblasts (cells were analyzed for 72 hours after transfection). (C) Percentage of cell survival relative to polyQ19 was measured at 72 hours after transfection in mouse (black bars) and NMR (gray bars) cells. The data represent three independent experiments (cells obtained from three different animals) and done in duplicate, with an N of 19–32 cells per experiment, and analyzed by log-rank t test (mantel-cox test). PolyQ19 and polyQ82 survival curves in mouse cells were significantly different with a p < .0001 and with a hazard ratio of 7.2. The asterisks (****) denotes a statistically significant difference at p ≤ .0001. No significant difference in NMR cells (p = .1469) with a hazard ratio of 1.9.

Knocking Down Autophagy, Ubiquitin-Proteasome, or HSP27 Affected NMR Cell Survival

Previous data in our laboratory showed that NMR fibroblasts show enhancement in autophagy, heat shock chaperones, and ubiquitin proteasome (11). To test the role of each of these mechanisms in the resistance to polyQ82 proteotoxicity, we used siRNA against key proteins for each mechanism: ATG5 (autophagy), PSMD14 (Ub-proteasome), and HSP27 chaperone. PolyQ82-YFP-induced proteotoxicity was analyzed by live cell imaging as previously described. As shown in Figure 2A, there were no significant differences between survival of control and cells transfected with scrambled siRNAs. However, we found that knocking down each of these mechanisms affected cell survival, even in cells transfected with polyQ19 as compared to scrambled siRNAs (Figure 2B and D). The most detrimental effect was observed with siRNA targeting autophagy (ATG5 gene, Figure 2B), where both polyQ19 and polyQ82 show 100% cell death at 54 hours and 27 hours after transfection, respectively. A significant increase in cell toxicity was observed when we measured the hazard ratio for survival in siRNA cells transfected with polyQ82 compared to polyQ19 (hazard ratio of 1.75; p < .05; curves with triangles in Figure 2B and Supplementary Figure 1A). The inhibition of the proteasome with siRNA against PSMD14 led to 63% and 100% cell death at 72 hours after transfection with polyQ19 and polyQ82, respectively. Whereas siRNA against HSP27 shows a 46.6% and 87.3% cell death 72 hours after transfection with polyQ19 or polyQ82, respectively. Likewise, PSMD14 and HSP27 siRNA cells transfected with polyQ82 showed an increase in cell toxicity (hazard ratio of 1.85 and 2, respectively) as compared to siRNA cells transfected with polyQ19 (p < .05; curves with triangles in Figure 2C and D and Supplementary Figure 1A). All the three siRNA, ATG5, PSMD14, HSP27, treatments also have a greater impact on cells transfected with polyQ82 in relation to their respective scrambled siRNA controls (hazard ratios of 0.277, 0.435, and 0.547, respectively, p < .05) than survival of cells transfected with polyQ19 compared to their respective siRNA controls (hazard ratios of 0.0991, 0.239, and 0.29; Figure 2). However, when we knocked down HSF1 by 90%–100%, we did not observe a negative effect on resistance to proteotoxicity in NMR cells (Supplementary Figure 3). HSF1 is the major transcription factor that regulates the expression of heat shock chaperones, including the HSP27; so, these results might be due to a compensatory effect in NMR cells. Similar findings were observed in mouse cells, where knocking down ATG5 or HSP27 increased cell death compared to controls (scrambled siRNA), but these effects were not as dramatic as in NMR cells (Supplementary Figure 2). Overall, the data show that all three pathways contribute to the increased resistance to polyQ toxicity, with autophagy having the greatest effect.

Figure 2.

Knocking down of protein quality control processes affected naked mole rat (NMR) cell survival. Knockdown of autophagy, proteasome, and HSP27 was done using siRNA targeting ATG5, PSMD14, and HSP27, respectively. (A) Survival curves of NMR cells transfected with polyQ19 and polyQ82 and cells transfected with scrambled siRNA followed by polyQ19 and polyQ82. All survival curves were compared to the corresponding scrambled siRNA. (B) Effect of ATG5 knocking down in NMR cell survival transfected with polyQ19 (p ≤ .0001; hazard ratio of 10.1) and polyQ82 (p ≤ .0001; hazard ratio of 10.6) with respect to polyQ19 scrambled siRNA-transfected cells (solid line). (C) Effect of PSMD14 knocking down in NMR cell survival transfected with polyQ19 (p ≤ .001; hazard ratio of 4.2) and polyQ82 (p ≤ .001; hazard ratio of 9.2) with respect to polyQ19 scrambled siRNA-transfected cells (solid line). (D) Effect of HSP27 knocking down in NMR cell survival transfected with polyQ19 (p ≤ .01; hazard ratio of 3.40) and polyQ82 (p ≤ .05; hazard ratio of 7.1) with respect to polyQ19 scrambled siRNA-transfected cells (solid line). Data represent one experiment with an N = 25–47 cells and analyzed by log-rank t test (Mantel-Cox test). The asterisks (*,**,***,****) denotes a statistically significant difference at p ≤ .05, p ≤ .01, p ≤ .001, and p ≤ .00001, respectively. NS stands for no significant difference.

NMR Cells Have a Perinuclear Cellular Location of polyQ82 Inclusions

To determine whether this resistance to proteotoxicity in NMR cells was due to a lower level of protein aggregation, we measured the levels of protein aggregates in NMR and mouse cells. Our data show that 36 hours after transfection, NMR cells have similar (if not higher) levels of polyQ82 protein aggregates than mouse cells (Figure 3A). To establish whether there was any difference in the concentration at which the polyQ82-YFP starts forming inclusions, we used a method previously described by Kopito’s lab (32). As shown in Figure 3B, the outline of the cell is marked using an imaging software (Fiji) at the frame where inclusions start to appear, and mean cell fluorescence is calculated by dividing total cell fluorescence by total area of the cell. Our data show that NMR fibroblasts form inclusions at 65% less mean YFP fluorescence than mouse fibroblasts, suggesting that NMR cells start forming polyQ82 aggregates at lower concentrations of polyQ82 protein than mouse cells.

Figure 3.

PolyQ82 inclusions in naked mole rat (NMR) fibroblasts show a perinuclear cellular location. (A) The percent of inclusions was calculated at 36 hours after transfection. (B) Mean cell fluorescence was calculated in the frame where inclusions start to appear. The quantitation of the percent of mean cell fluorescence is shown on the right. The data represent mean ± SD of three independent experiments. Indirect immunofluorescence was used to identify location of inclusions respective to nuclear membrane and centrosomes. (C) Location of inclusions polyQ82-YFP inclusions (GFP) were identified with respect to nucleus (DAPI) and nuclear membrane [protein Lamin A/C], and overlay of three channels is also shown in both top (mouse) and bottom (NMR) panels. Quantitation of percentage of cells with perinuclear inclusions is shown in the bar graph to the right (x-axis: inclusion’s cellular location based on proximity of inclusions to the nucleus: perinuclear and far; y-axis: percentage of cells). (D) Location of inclusions polyQ82-YFP inclusions (GFP) were identified with respect to nucleus (DAPI) and centrosomes (γ-tubulin), a marker for aggresomes. Overlay of three channels in both top (mouse) and bottom (NMR) panels. Quantitation of cells with inclusions near to centrosomes is shown in the bar graph. The data was analyzed by unpaired t test. N = 19–33 cells were imaged in each experiment. The asterisks (**,***) denote a statistically significant difference between mouse and NMR fibroblasts at p ≤ .01 and p ≤ .001, respectively. White bar = 10 µm.

Interestingly, in all these experiments, we observed a clear difference between mouse and NMR fibroblasts in terms of cellular location of the polyQ82-YFP inclusions. Specifically, we observed that inclusions in NMR fibroblasts were often located in the perinuclear region (forming dents in the nuclear membrane), whereas in mouse, inclusions were generally located more randomly and near the periphery of the cell. To establish whether the inclusions are perinuclear or intranuclear, we used indirect immunofluorescence by targeting lamin A/C proteins to stain the nuclear membrane. These experiments clearly show that in mouse fibroblasts most inclusions were randomly located, with only 27% of them being perinuclear, whereas in NMR fibroblasts, inclusions were predominantly perinuclear (80%) or close to the nucleus (Figure 3C).

This perinuclear location suggests that these aggregates are forming aggresomes, a protective type of aggregate that primarily localizes in the perinuclear region, in close association with centrosomes and surrounded by the intermediate filament vimentin. Aggresomes are eventually cleared by degradation (25,33). Therefore, we explored whether polyQ82 aggregates in NMR were near to centrosomes using an anti γ-tubulin antibody (a marker for centrosomes). Figure 3D shows that 48 hours after transfection with polyQ82-YFP, NMR cells have a significantly higher number (over 65%) of inclusions close to the centrosome (red dots), as compared to mouse fibroblasts (~35%). Furthermore, in mouse cells polyQ82 aggregates did not colocalize with vimentin, while in NMR cells, we were unable to observe any vimentin signal due to incompatibility of the antibodies (Supplementary Figure 6).

HSP27 Colocalizes with polyQ82 Inclusions in NMR Fibroblasts but not in Mouse Fibroblasts

To confirm the data on differences in aggresome formation, we used immunofluorescence and confocal imaging to investigate if proteins that form part of the aggresome signaling pathway colocalize with polyQ82 aggregates. We first analyzed Hsp27, HDAC6, and p62 because they participate in each of the state of the aggresome pathway: initiation, transport, and degradation, respectively. Interestingly, we found that HSP27 colocalized with inclusions only in NMR but not in mouse fibroblasts (Figure 4A). P62 colocalized with polyQ82 inclusions in both mouse and NMR fibroblasts (Figure 4B), while HDAC6 did not colocalize with polyQ82 inclusions in either mouse or NMR fibroblasts. However, we observed that HDAC6 signal was attenuated in NMR cells suggesting an antibody incompatibility. To test further whether these proteins colocalize with polyQ82, we run a pull-down experiment using a polyclonal polyQ antibody. Our data show that indeed P62, HDAC6, and Hsp27 coimmunoprecipitated with polyQ in both mouse and NMR cells, although HDAC6 and HSP27 coimmunoprecipitates significantly more in NMR cells than in mouse cells (Supplementary Figure 4). These data support our finding that protein of the aggresome pathway colocalizes with polyQ82 aggregates and there is a difference in how mouse and NMR cells handle polyQ82 aggregates.

Figure 4.

PolyQ82YFP inclusions colocalized with HSP27 in naked mole rat (NMR) fibroblasts. Immunofluorescence using DAPI, polyQ82-YFP (GFP), HSP27 (A), HDAC6 (B), and p62 (C), followed by the overlay of three channels in both mouse (top) and NMR (bottom) panels. Data represent independent two different experiments with two different cell lines. Zeiss LSM 780 confocal microscope was used for imaging. 40× water objective. White bar = 10 µm.

Improved Resistance to Proteotoxic Stress Was Apparent in Other Long-Lived Species

To test whether this difference in proteotoxicity was also observed in other short- and long-lived species, we studied fibroblasts from the long-lived little brown bat (LBB; MLS = 34 years) and compared them to fibroblasts from the short-lived EB (MLS = 6 years). Our results in Figure 5 show that there was no difference in overall cell survival in response to polyQ82-YFP between long- and short-lived bat fibroblasts (LBB vs EB). The overall survival curves of EB and LBB have similar hazard ratios (2.2 and 2.0, respectively, Figure 5A and B). However, while fibroblasts from the short-lived EB have a median survival of 22 hours, median survival for the long-lived LBB fibroblasts is 38 hours, indicating that LBB cells transfected with polyQ82 have a better median cell survival than EB. We also measured the mean cell fluorescence at which inclusions start forming (as described earlier) and we observed that LBB fibroblasts formed inclusions at two times less mean cell fluorescence than fibroblasts from EB (Figure 5C). We also analyzed the location of the polyQ82 inclusions with respect to the nucleus and centrosome (Figure 5D). Because the lamin A/C antibody did not work in EB cells, location of inclusions with respect to the nucleus was established using 4’,6-diamidino-2-phenylindole (DAPI) staining. Our data show that 90.4% of LBB fibroblast inclusions were perinuclear, whereas in EB fibroblasts, only ~60% of the inclusions were located perinuclearly (Figure 5D). In addition, we also noticed differences in terms of proximity of inclusions with respect to the centrosome, where 64% of LBB fibroblasts had inclusions close to or colocalizing with the centrosome, whereas for EB fibroblasts, this was only 36% (Figure 5E).

Figure 5.

Improved survival to proteotoxic stress was apparent in other long-lived species. Cell survival curves of skin fibroblasts from short-lived (evening bat [EB]) (A) and long-lived (little brown bat [LBB]) (B) bats were transfected with polyQ19-YFP (solid lines) and polyQ82-YFP (dashed lines). (C) The quantitation of the percent of mean cell fluorescence is shown in the graph below. PolyQ82 cellular localization in LBB cells using protein lamin A/C (D) and γ-tubulin (E). Quantitation of perinuclear and centrosomes localization is shown below each figure. Black bars represent EB cells and grey bars represent LBB cells. Survival curves were analyzed according to log-rank (Mantel-Cox) test. The hazard ratios for both EB and LBB cells is 2.10. Median survival for EB = 22.5 hours, and for LBB = 37.5 hours. N = 13–32 cells per cell line were analyzed. For mean YFP fluorescence, data shown is the mean ± SD and analyzed by unpaired t test. The data represent three independent experiments where different EB and LBB cell lines were used. N = 19–32 cells per cell line were analyzed. Data asterisk (**,***) denotes a statistically significant difference between EB and LBB fibroblasts at p ≤ .01 and p ≤ .001, respectively. White bar = 10 µm.

Discussion

The objective of this work was to investigate whether the enhanced proteostasis observed in long-lived species conferred a protection against proteotoxicity arising from aggregation of an exogenous misfolded protein. To test this, we used a molecular approach where cells of both short- and long-lived species were transfected with the same exogenous aggregation-prone protein, polyQ82-YFP (19,24,34,35). PolyQ proteins are broadly used as a protein aggregation model because their aggregation state is sensitive to genetic or pharmacological modifiers (19,36), as well as different stressors and the age of the model organism (37). Although lower levels of insoluble protein fractions in long-lived species (obtained from whole cellular extracts) have been used as an indicator of lower protein aggregation in long-lived species (17), it is not known whether this is due to robust proteostasis processes or because of innate differences in the composition of the cell’s proteome. As a result, it is not known if the improved proteostasis in long-lived species is what leads to resistance to protein aggregation and proteotoxicity.

Our data show that cells from long-lived species (NMR and LBB) were indeed more resistant to proteotoxicity induced by polyQ82 than their short-lived counterparts (mouse and EB). This effect was more robust in NMR fibroblasts than in LBB cells (median survival time was 70 hours versus 38 hours, respectively; Figures 1 and 5). These data are consistent with our previous data where LBB cells have better autophagy, but no difference in heat shock chaperones or proteasome activity as compared to EB fibroblasts, indicating that long- and short-lived bats show less differences in proteostasis mechanisms than observed between NMR and mouse (11). Also, even the short-lived bat is long-lived relative to the mouse; thus, these differences may influence the resistance against proteotoxicity in the cells from these two species of bat. It may also be relevant that these cells were obtained from wild-caught bats; thus, it is also possible that differences in environmental exposure, age, and other factors may influence these preliminary findings, and experimentation under more controlled experimental conditions may be needed to assess these mechanisms.

By using siRNA to block the various proteostasis pathways, we show that all three mechanisms tested, autophagy, proteasome, and chaperones, are important for cell survival even in the absence of a proteotoxicity challenge, since a decline in cell survival was observed even after transfection with polyQ19. Likely, this reflects the need of proteostasis systems in order to withstand the stress of transfection itself. However, the toxic effect was exacerbated by polyQ82. The most dramatic effect on cell survival was observed when macroautophagy was inhibited (ATG5 siRNA), indicating that this mechanism is crucial for cell survival and protection against general stress, as well as proteotoxic stress. While expression of both polyQ19 and polyQ82 induced 100% cell death in ATG5 knockdown cells (Figure 2B), the effect of knocking down the ub-proteasome pathway or HSP27 preferentially increased cell death in cells expressing polyQ82, relative to cells expressing polyQ19 (Figure 2C and D).

These data indicate that these proteostatic mechanisms are crucial not only to protect against the stress of protein aggregation, but also for basal cell survival and to protect against the general stress induced by the transfection and expression of an exogenous protein. Interestingly, our experiments knocking down HSF1 using CRISPR-Cas9 did not show any difference in cell survival (Supplementary Figure 3), suggesting that NMR cells overcome their deficiency with compensatory mechanisms that protected the cell against proteotoxicity. The nature of this protective mechanism is currently unknown.

Contrary to our expectations, our data show that NMR skin fibroblasts did not have any less polyQ82YFP aggregation as compared to mouse fibroblasts. However, we also observed that NMR fibroblasts formed inclusions at a lower mean cell fluorescence (indicative of lower concentrations of polyQ; Figure 3A and B), which may suggest that formation of inclusions could be a protective mechanism to prevent the toxic effects of misfolded monomers and small oligomers as previously described (38–40). Indeed, new evidence from multiple labs has shown that aggregates and inclusions are not always toxic, and that in fact, they could be protective (38,39,41). Specifically, in the field of polyglutamine disease research, it has been shown that large inclusions are cytoprotective while smaller oligomers, misfolded monomers, or microaggregates underlie pathogenesis (40). These data correlate well with our finding of a perinuclear cellular location of polyQ82 aggregates in NMR cells, with polyQ82YFP found to be predominantly located at the perinuclear region (Figure 3C and D).

Our results differ from previous data from Morimoto’s lab, where a decrease in polyQ protein aggregation was observed in the long-lived Age-1 mutants, and knockdown of HSF1 and heat shock proteins in C. elegans resulted in accelerated aggregation of polyQ proteins (19,24,42). These results do not necessarily exclude the protective nature of polyQ inclusions, but suggest that in C. elegans, polyQ monomers are being handled more efficiently, therefore preventing the need to form polyQ inclusions.

However, the support for inclusions being protective comes from increasing evidence using yeast and mammalian cells, indicating that soluble misfolded monomers and intermediary oligomers of the extended polyQ are more toxic than insoluble inclusions (39,43–45). For example, using a different model of protein aggregation, Woerner et al. showed that in HEK293T cells, aggregation of artificial beta sheet proteins or fragments of mutant huntingtin in the perinuclear region is more protective than aggregation in the cytoplasmic region because perinuclear aggregates do not interfere with nucleocytoplasmic protein and RNA transport (46). Similarly, Kaganovich et al. showed that in yeast, misfolded proteins can be sorted between two distinct quality control compartments: the perinuclear “soluble” quality control (JUNQ), where the aggregates can still be degraded; or the peripheral or “insoluble” protein deposits (IPOD), where terminally aggregated proteins are sequestered in the periphery (34). Protective inclusions known as aggresomes are found in mammalian cells. Aggresome formation is a highly regulated process that serves to organize aggregates of misfolded proteins into a single location in the cell (perinuclear) poised to be degraded later (25). Based on our data showing that mouse fibroblasts do not form aggresomes as efficiently as NMR, because polyQ82 inclusions were randomly located primarily in the periphery of the cell, far from centrosomes and polyQ82 inclusions did not colocalize with vimentin; we believe that the difference in cellular location of protein aggregates between NMR and mouse cells plays an important role in cell survival. Because the agressome pathway has been observed in several cell types and using a variety of beta sheet proteins (or other proteins that misfold easily), we speculate that we will observe similar results to those observed with polyQ, if we would use another type of protein aggregates (eg, such as amyloid beta protein) in NMR cells.

Similarly, our data using immunofluorescence and confocal imaging for proteins that participate in the aggresome pathway, that is, p62, HDAC6, and HSP27 (26,28,47) partially support our aggresome hypothesis, because we found a strong colocalization between HSP27 and polyQ82 in the perinuclear region of NMR cells. Cox et al. (2018) also showed that HSP27 inhibits cytotoxicity of alpha synuclein fibrils in a cell culture model (48). In addition, several researchers have shown that polyQ- dependent toxicity correlates with a failure to upregulate or activate HSP27 (49,50). Hence, it is possible that HSP27 might be conferring protection against polyQ-dependent toxicity in NMR fibroblasts. We also measured p62, which is a common component of protein aggregates (51), known to link the recognition of polyubiquitinated protein aggregates to the autophagy machinery (52,53). Thus, it was not entirely surprising that we found p62 colocalization with polyQ82 inclusions in both mouse and NMR fibroblasts. Some reports using knockdown of p62 showed that p62 does not affect formation of aggresomes, indicating that p62 binds to aggresomes after their formation (54). On the other hand, HDAC6 is a protein that plays a role in the formation of aggresomes, by recruiting and transporting polyubiquitinated misfolded proteins and aggregates to the aggresome via the microtubule network (55) Our immunofluorescent data show that polyQ82 inclusions do not colocalize with HDAC6 in either mouse or NMR fibroblasts. However, we observed that the cellular distribution of HDAC6 in NMR cells is different from what is observed in mouse cells: Whereas in mouse cells, HDAC6 immunofluorescence shows a clearly distinct cellular distribution; in NMR cells, HADC6 immunofluorescence was weaker and with a homogenous distribution, questioning the reactivity of the antibody in NMR cells. However, in pull-down experiments using a polyclonal anti-polyQ antibody show that HDAC6 was observed in immunoprecipitated from both mouse and NMR cells, showing a stronger band in NMRs extracts (Supplementary Figure 4). The possible relevance of this observation is currently unknown due to the inconsistency between immunofluorescence and pull-down experiments. Nonetheless, there are other proteins that participate in aggresome formation independently of HDAC6, such as BAG3, a chaperone which uses nonubiquitinated substrates (56). The role of this protein as well as other proteins that participate in the aggresome pathway will be studied in more detail in future experiments.

In summary, our data support the hypothesis that a chronic load of unstable proteins makes functional recycling of unfolded aberrant proteins less feasible, and thus, it is better for the cell to simply sequester these inclusions into ordered aggregates (aggresomes) as the best option for cell survival (33). However, we show that different species differ in their approaches, with longer-lived species (NMR and LBB) favoring the sequestration of aggregates into aggresomes, a mechanism less used by shorter-lived ones (mouse and EB), as shown in our proposed model (Supplementary Figure 7). We also show that a possible mechanism that differentiates these activities is the level and involvement of HSP27, a chaperone involved in the early stages of the aggresome pathway (50,57).

Funding

This work was supported by the National Institute on Aging (R03AG05235 to V.I.P.) of the National Institutes of Health.

Author Contributions

B. S., Z.Y., and V.I.P. designed the research; B.S., Z.Y., I.A., AC.D., J.DM., RT.R., and V.I.P. performed the research; B.S., Z.Y., RT.R., and V.I.P. analyzed the data; and B.S., and V.I.P. wrote the paper.

Conflict of Interest

None reported.