(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Yeast derlin Dfm1 employs a chaperone-like function to resolve misfolded membrane protein stress [1]

['Rachel Kandel', 'Division Of Biological Sciences', 'The Section Of Cell', 'Developmental Biology', 'University Of California San Diego', 'La Jolla', 'California', 'United States Of America', 'Jasmine Jung', 'Della Syau']

Date: 2023-01 

Protein aggregates are a common feature of diseased and aged cells. Membrane proteins comprise a quarter of the proteome, and yet, it is not well understood how aggregation of membrane proteins is regulated and what effects these aggregates can have on cellular health. We have determined in yeast that the derlin Dfm1 has a chaperone-like activity that influences misfolded membrane protein aggregation. We establish that this function of Dfm1 does not require recruitment of the ATPase Cdc48 and it is distinct from Dfm1’s previously identified function in dislocating misfolded membrane proteins from the endoplasmic reticulum (ER) to the cytosol for degradation. Additionally, we assess the cellular impacts of misfolded membrane proteins in the absence of Dfm1 and determine that misfolded membrane proteins are toxic to cells in the absence of Dfm1 and cause disruptions to proteasomal and ubiquitin homeostasis.

Funding: This work was supported by the National Institutes of Health (1R35GM133565 to SEN;R00GM135515 to SD), the National Science Foundation (2047391 to SEN), and the Pew Charitable Trusts (34089 to SEN). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

As a complement to our work on the function of Dfm1 in relieving misfolded membrane protein toxicity, we also sought to determine how misfolded membrane proteins cause toxicity. We determine that misfolded membrane proteins, but not other types of misfolded proteins, impact proteasome and ubiquitin homeostasis. We also identified several proteins that promote cellular health upon misfolded membrane protein by resolving the proteasome and ubiquitin stress that misfolded membrane proteins trigger. Intriguingly, we also find that not all membrane protein aggregates are toxic. The combination required for toxicity appears to be both (1) aggregated misfolded membrane proteins; and (2) ubiquitinated misfolded membrane proteins. Either of these features alone is not sufficient for toxicity.

In the present study, we determine that Dfm1 prevents membrane protein toxicity because of a previously unidentified chaperone-like function that is independent of Cdc48 recruitment. This function is distinct from Dfm1’s role in protein retrotranslocation, while also relying on many of the same functions deployed by Dfm1 to promote retrotranslocation. We further determined that human homologs of Dfm1 have also retained this ability. This study is the first to demonstrate chaperone-like activity for any rhomboid protein. Many rhomboid proteins use similar functions to Dfm1 to promote retrotranslocation, and the rhomboid protease RHBDL4 has recently been characterized as acting on aggregation-prone substrates [ 19 , 28 , 35 – 37 ]. It will be an interesting and important further line of inquiry to determine if a chaperone-like activity is a common ability of rhomboid proteins, both for the pseudoproteases and proteases.

(A) WT, dfm1Δ, and hrd1Δ cells containing either GAL pr -HMG2-GFP or EV were compared for growth by dilution assay. Each strain was spotted 5-fold dilutions on glucose or galactose-containing plates to drive HMG2-GFP overexpression, and plates were incubated at 30°C. (B) Dilution assay as described in (A) except using WT, dfm1Δ, and doa10Δ cells containing either GAL pr -STE6*-GFP or EV. (C) Dilution assay as described in (A) except using WT, dfm1Δ, and hrd1Δ cells containing either GAL pr -PDR5*-HA or EV. (D) Dilution assay as described in (A) except using WT, dfm1Δ, and cdc48-2 cells. (E) Dilution assay as described in (B) except using WT, dfm1Δ, and cdc48-2 cells. (F) Dilution assay as described in (C) except using WT, dfm1Δ, and cdc48-2 cells. (G) Dilution assay as described in (A) except using WT or dfm1Δ cells expressing human CFTR, CFTRΔF508, or A1PiZ and plated only on glucose-containing plates. All dilution growth assays were performed in 3 biological and 2 technical replicates (N = 3). CFTR, cystic fibrosis transmembrane receptor; ERAD, endoplasmic reticulum-associated degradation; EV, empty vector.

We have previously observed that in dfm1Δ cells, when a misfolded membrane protein is strongly expressed, the cells show a severe growth defect [ 33 ]. This is seen specifically in the absence of Dfm1, and this growth defect is not observed in the absence of other ERAD components, indicating a specific function for Dfm1 in sensing and/or adapting cells to misfolded membrane protein stress ( Fig 1 ) [ 33 ]. This is in line with a previous study linking Dfm1 to ER homeostasis [ 34 ].

Dfm1 is a member of the derlin subclass of rhomboid proteins. Rhomboid proteins are a widely conserved family of proteins, found in all domains of life [ 20 – 23 ]. There are 2 major categories of rhomboid proteins: active rhomboid proteases and inactive rhomboid pseudoproteases. While the inactive rhomboid pseudoproteases lack a catalytic site, they have been implicated in a wide variety of biological processes, including protein quality control, protein trafficking, and cell signaling [ 24 – 28 ]. Derlin proteins, including Dfm1, are rhomboid pseudoproteases that are critical for ERAD of a wide variety of substrates, both in yeast and mammalian cells [ 8 , 29 – 32 ].

In yeast, ER membrane substrates can be targeted by the DOA (degradation of alpha2) pathway or the HRD pathway (hydroxymethyl glutaryl-coenzyme A reductase degradation), utilizing the E3 ligases Doa10 and Hrd1, respectively. Additionally, the yeast derlin Dfm1 is specifically required for the retrotranslocation of misfolded membrane substrates, in both the HRD and DOA pathways [ 8 ]. Dfm1 facilitates retrotranslocation of membrane proteins through several mechanisms including (1) recognition and binding to misfolded membrane proteins; (2) thinning the lipid bilayer to reduce the thermodynamic barrier to extraction; and (3) recruiting the ATPase Cdc48 to the ER [ 8 , 19 ].

ERAD is a well-conserved process from yeast to mammals. ERAD of membrane proteins requires 4 universal steps: (1) substrate recognition [ 3 – 6 ]; (2) substrate ubiquitination [ 7 ]; (3) retrotranslocation of substrate from the ER to the cytosol [ 8 – 13 ]; and (4) degradation by the cytosolic proteasome [ 2 , 14 , 15 ]. A hexameric cytosolic ATPase, Cdc48 in yeast and p97 in mammals, is required for retrotranslocation of all ERAD substrates [ 16 – 18 ]. In the context of this paper, substrate refers to a protein that is targeted by the ERAD pathway.

While misfolded proteins are recognized as a source of cellular stress, the mechanisms by which cells prevent this stress and how this stress impacts cells is not fully understood. Eukaryotic cells are equipped with protein quality control pathways for preventing the accumulation of aggregation-prone misfolded proteins. The endoplasmic reticulum (ER) is responsible for folding both secretory and membrane proteins and is well equipped with quality control pathways for refolding or eliminating misfolded proteins. One of the major pathways of protein quality control at the ER is ER-associated degradation (ERAD) [ 1 ]. ERAD utilizes the ubiquitin proteasome system (UPS) to selectively target and degrade misfolded or unassembled proteins at the ER [ 2 ].

Results

Absence of Dfm1 and expression of integral misfolded membrane proteins causes growth stress Previous research from the Neal lab has revealed that accumulation of a misfolded membrane protein in the absence of Dfm1 causes a severe growth defect in the substrate-toxicity assay [33]. In the substrate-toxicity assay, yeast strains with a misfolded protein under the control of a galactose inducible promoter are plated in a spot assay onto selection plates with either 2% galactose or 2% dextrose as a carbon source (Fig 1) [38]. This allows for comparison of growth of yeast strains with different genetic perturbations with expression of misfolded substrates. This growth defect can be seen with strong expression of 3 misfolded membrane proteins in dfm1Δ cells: Hmg2, Pdr5*, and Ste6* (Fig 1A–1C). We have previously shown that this growth defect is specific to misfolded membrane proteins at the ER, as expression of a luminal ERAD substrate, CPY*, in dfm1Δ cells elicits no growth defect [33]. Interestingly, this growth defect is not seen when membrane proteins accumulate in the absence of other ERAD components, such as the E3 ligases Hrd1 and Doa10 (Fig 1A–1C). In the case of dfm1Δ, hrd1Δ, and doa10Δ cells, misfolded membrane proteins accumulate at the ER due to defects in ERAD, but only in the case of dfm1Δ cells, is a growth defect observed with misfolded membrane protein expression. Altogether, we surmise that this growth defect triggered by the absence of Dfm1 along with expression of misfolded membrane protein is due to cellular stress caused by misfolded membrane protein toxicity. By utilizing the substrate-toxicity assay, we observed a growth defect in dfm1Δ cells and normal growth in hrd1Δ and doa10Δ cells upon expression of ERAD membrane substrates (Fig 1A–1C). The cell biological difference among these ERAD knockout strains is that membrane substrates are ubiquitinated in dfm1Δ cells but not ubiquitinated in hrd1Δ and doa10Δ cells, due to the absence of the ER E3 ligases, as determined through western blot for ubiquitin (Fig 7B). One possibility is that the growth stress is not specific to dfm1Δ cells and is solely dependent on the accumulation of ubiquitinated membrane substrates. To rule out this possibility, we utilized a temperature-sensitive Cdc48 allele strain, cdc48-2, which, like dfm1Δ cells, results in accumulation of ubiquitinated ERAD membrane substrates [8]. While we used cdc48-2 cells at the permissive temperature of 30°C, ERAD is still compromised for Hmg2, as previously reported, and we validated using a cycloheximide chase that Pdr5* and Ste6* degradation is also impaired at 30°C in cdc48-2 cells (S1A and S1B Fig) [8,39]. The substrate-toxicity assay was employed on cdc48-2 strains expressing membrane substrates Hmg2, Pdr5*, and Ste6* (Fig 1D–1F). These strains showed a growth defect while growing on galactose plates due to inherent slow growth of cdc48-2 strains, but this was not worsened by expression of misfolded integral membrane proteins, despite the E3 ligases that ubiquitinate these proteins still being present. These results indicate that Dfm1 plays a specific role in the alleviation of misfolded membrane protein stress.

Disease-associated membrane proteins cause growth stress Since a wide variety of misfolded membrane proteins elicit growth stress in dfm1Δ cells, we hypothesized that growth stress would also be observed with expression of clinically relevant human misfolded membrane proteins. We tested expression of WT cystic fibrosis transmembrane receptor (CFTR), CFTRΔF508, the most common disease-causing variant of CFTR, and the Z variant of alpha-1 proteinase inhibitor (A1PiZ), a protein variant that results in alpha-1 antitrypsin deficiency (AATD). All these substrates are targeted by ERAD when expressed in yeast, with CFTR and CFTRΔF508 being targeted by machinery for membrane proteins and A1PiZ being a soluble misfolded protein in the ER lumen [40,41]. When these proteins were expressed in dfm1Δ cells, both CFTR and CFTRΔF508 resulted in a growth defect, while none was observed with expression on A1PiZ (Fig 1G). Expression of any of the proteins in WT yeast cells resulted in no growth defect (Fig 1G). It was not wholly surprising that WT CFTR also elicited growth stress in dfm1Δ cells. Previous studies have shown that while virtually all CFTRΔF508 is targeted to ERAD, about 80% of WT CFTR is degraded via ERAD in yeast and mammals [40,42,43].

Dfm1 has a dual role in ER protein stress and ERAD retrotranslocation Previous work from the Hampton lab establishing a role for Dfm1 in misfolded membrane protein retrotranslocation also identified several motifs of Dfm1 that are essential for its retrotranslocation function [8]. Additionally, by employing an unbiased genetic screen, our lab recently identified 5 residues of Dfm1 that are required for retrotranslocation [19]. Here, we tested whether these residues, critical for Dfm1’s retrotranslocation function, are required for alleviating the growth stress in dfm1Δ cells expressing Hmg2. Fig 2A shows a schematic of Dfm1, with the regions of the protein important for retrotranslocation function highlighted, and a list of specific motifs and residues that are mutated listed in a table [8,19]. Dfm1 contains 2 motifs that are well conserved among the rhomboid superfamily, the WR motif in Loop 1 and the GxxxG (Gx3G) motif in transmembrane domain (TMD) 6 [8,27]. Both of these motifs are required for Dfm1-mediated retrotranslocation [8,44]. We first tested the requirement of the conserved rhomboid motif mutants by expressing Hmg2 with WR mutants (WA and AR) and Gx3G mutants (Ax3G and Gx3A) and observed no restoration in growth (Fig 2B). Our previous work determined that Loop 1 mutants (F58S, L64V, and K67E) obliterated Dfm1’s ability to bind misfolded membrane substrates, and TMD 2 mutants (Q101R and F107S) reduce the lipid thinning ability of Dfm1, a function that aids in Dfm1’s retrotranslocation function [44]. Accordingly, we utilized these mutants in our growth assay and did not observe a rescue of the growth defect (Fig 2C). We have previously shown that alteration of the 5 signature residues of the Dfm1 SHP box to alanine (Dfm1-5Ashp) ablates its ability to recruit Cdc48 (Fig 2D). We also established, Dfm1’s Cdc48 recruitment function is required for Dfm1’s retrotranslocation function, whereas the Dfm1-5Ashp mutant impairs its retrotranslocation function [8]. Notably, in contrast to the other mutants tested, Dfm1-5Ashp was still able to alleviate the growth defect like WT Dfm1 (Fig 2E). These results suggest that Dfm1’s substrate engagement and lipid thinning function is required for alleviating membrane substrate-induced stress, whereas Dfm1’s Cdc48 recruitment function is dispensable for alleviating the growth stress. We validated that expression of WT Dfm1 and all Dfm1 mutants were comparable using western blot (S1C Fig). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Dfm1 retrotranslocation defective mutants show differing abilities to restore growth. (A) Depiction of Dfm1, which highlights L1, TM2, TM6, and its SHP box domain. The table indicates the Dfm1 region, amino acid mutation, and the corresponding function that is specifically impaired. All mutants have been previously identified as being required for retrotranslocation and when mutated did not restore growth in dfm1Δ cells expressing an integral membrane protein (GAL pr -HMG2-GFP). (B) dfm1Δ cells with an add-back of either WT DFM1-HA, EV, DFM1-WA-HA, DFM1-AR-HA, DFM1-Ax3G-HA, or DFM1-Gx3A-HA containing either GAL pr -HMG2-GFP or EV were compared for growth by dilution assay. Each strain was spotted 5-fold dilutions on glucose or galactose-containing plates to drive Hmg2-GFP overexpression, and plates were incubated at 30°C. (C) Dilution assay as described in (B) except using an add-back of either WT Dfm1-HA, EV, Dfm1-F57S-HA, Dfm1-L64V-HA, Dfm1-K67E-HA, Dfm1-Q101R-HA, or Dfm1-F107S-HA. (D) Depiction of Dfm1 and Dfm1-5Ashp. Dfm1 is an ER-localized membrane proteins with 6 TMDs. Both versions of Dfm1 have a cytoplasmic shp box, but the 5Ashp mutant is unable to recruit the cytosolic ATPase Cdc48. (E) Dilution assay as described in (B) except using add-back of either EV, WT DFM1-HA, or DFM1-5Ashp-HA mutant. (F) Dilution assay as described in (B) except with add-back of human Derlin-1-Myc or Derlin-2-Myc. All dilution growth assays were performed in 3 biological replicates and 2 technical replicates (N = 3). ER, endoplasmic reticulum; EV, empty vector; TMD, transmembrane domain. https://doi.org/10.1371/journal.pbio.3001950.g002

Human derlins relieve growth stress Dfm1 is a rhomboid pseudoprotease and a member of the derlin subclass of rhomboid proteins. The human genome encodes 3 derlins, Derlin-1, Derlin-2, and Derlin-3. Yeast Dfm1 is the closest homolog of the mammalian derlins [27]. All 3 are ER-localized proteins that are implicated in ERAD and adaptation to ER stress [29,30,45–48]. We expressed human Derlin-1 and Derlin-2 in dfm1Δ+Hmg2 cells. We opted to only test Derlin-1 and Derlin-2, as they are more structurally similar to Dfm1 and also contain cytoplasmic SHP box motifs [19]. Both human derlins were able to rescue growth in these cells in the substrate-toxicity assay (Fig 2F). This was surprising, as we had previously found that mammalian derlins cannot complement the retrotranslocation function of Dfm1 in yeast cells for self-ubiquitinating substrate (SUS)-GFP, a similar ERAD substrate to Hmg2 [19].

Dfm1 solubilizes misfolded membrane protein aggregates independent of Cdc48 recruitment The above studies show that Dfm1 residues critical for retrotranslocation—through substrate binding and its lipid thinning function—are also important for alleviating membrane substrate-induced stress. Conversely, add-back of a Dfm1 Shp box mutant (Dfm1-5Ashp) (Fig 2D), which does not recruit Cdc48 and cannot retrotranslocate proteins, is able to restore growth in the substrate-toxicity assay. We surmise that Dfm1’s actions—independent of its Cdc48 recruitment function—may be directly acting on misfolded membrane substrates to prevent growth stress. One possibility is that Dfm1 may directly act on misfolded membrane substrates by functioning as a chaperone-like protein to prevent misfolded membrane protein toxicity. We hypothesize that Dfm1 acts as either a holdase, preventing the aggregation of misfolded membrane substrates, or as a disaggregase, separating proteins in existing protein aggregates. To address this hypothesis, we employed a detergent solubility assay in dfm1Δ+Hmg2 cells with add-back of WT DFM1 or DFM1 mutants. ER microsomes were isolated and incubated in 1% dodecyl maltoside (DDM) and subjected to centrifugation to separate aggregated substrate (pellet fraction) from non-aggregated substrate (soluble fraction). As shown in Fig 3A, nearly all Hmg2-GFP in dfm1Δ cells was pelleted (aggregated). Conversely, with Dfm1 and Dfm1-5Ashp add-back cells, nearly all Hmg2-GFP was solubilized (non-aggregated). This striking all-or-nothing phenotype of Hmg2 aggregation demonstrates an important role for Dfm1 in influencing membrane protein aggregation. As a control for these studies, we examined Hmg2 in both WT, hrd1Δ cells, and hrd1Δ+Hrd1 cells. Nearly all protein was soluble in all 3 strains (Fig 4A). We also tested a properly folded ER membrane protein, Sec61-GFP in dfm1Δ cells. In contrast to Hmg2-GFP, majority of Sec61-GFP was in the detergent-solubilized supernatant fraction and there was no change in Sec61-GFP detergent solubility with Dfm1 or Dfm1-5Ashp add-back in dfm1Δ cells (Fig 4B). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Dfm1 reduces misfolded membrane protein toxicity through a chaperone-like activity. (A) Western blot of aggregated (pelleted) versus non-aggregated (soluble) membrane proteins at the ER. Lysates from dfm1Δ cells containing HMG2-GFP, with either add-back of WT DFM1-HA, EV, DFM1-5Ashp-HA, DFM1-AR-HA, and DFM1-AxxxG-HA were blotted using anti-GFP to detect Hmg2. Top: Total fraction. Middle: ER pelleted fraction. Bottom: ER DDM solublilized fraction. (B) Western blot of aggregated versus non-aggregated membrane proteins at the ER as in (A) but with add-back of either WT DFM1-HA, EV, DFM1-F58S-HA, DFM1-L64V-HA, DFM1-K67E-HA, DFM1-Q101R-HA, and DFM1-F107S-HA. (C) DDM solubilized Hmg2-GFP binding to Dfm1-HA was analyzed by Co-IP, using anti-GFP to detect Hmg2 and anti-HA to detect Dfm1-HA. As negative control, cells not expressing Hmg2-GFP were used. Sec61 was analyzed as another negative control for nonspecific binding using anti-Sec61 (3 biological replicates, N = 3). (D) Graphic depicting integrated model of Dfm1’s function in misfolded membrane protein stress. Top: Misfolded membrane proteins in the absence of Dfm1 forming aggregates within the ER membrane. Bottom: Cells with WT Dfm1 or 5Ashp-Dfm1 promoting non-aggregated misfolded membrane proteins and preventing cellular toxicity. Data Information: All detergent solubility assays were performed with 3 biological replicates (N = 3). DDM, dodecyl maltoside; ER, endoplasmic reticulum; EV, empty vector. https://doi.org/10.1371/journal.pbio.3001950.g003 PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Dfm1 specifically influences solubility of misfolded membrane proteins. (A) Western blot of aggregated versus non-aggregated membrane proteins at the ER. Lysates from WT, hrd1Δ, or hrd1Δ+HRD1 cells containing HMG2-GFP were blotted using anti-GFP to detect Hmg2. T is total protein, P is pellet (ER aggregated) fraction, and S is soluble (ER non-aggregated) fraction. (B) Western blot of aggregated versus non-aggregated membrane proteins at the ER as in (A) except with dfm1Δ cells containing SEC61-GFP with add-back of EV, WT DFM1-HA, or DFM1-5Ashp-HA. Anti-GFP was used to detect SEC61-GFP. (C) Western blot of aggregated versus non-aggregated membrane proteins at the ER as in (A) except with dfm1Δ cells containing PDR5*-HA with add-back of WT DFM1-HA or EV. Anti-HA was used to detect PDR5*-HA. (D) Western blot of aggregated versus non-aggregated membrane proteins at the ER as in (A) except with dfm1Δ cells containing STE6*-GFP with add-back of WT DFM1-HA or EV. Anti-GFP was used to detect STE6*-GFP. (E) Western blot of aggregated versus non-aggregated membrane proteins at the ER as in (A) except with dfm1Δ cells containing CPY*-GFP with add-back of EV or WT DFM1-HA. Anti-GFP was used to detect CPY*-GFP. (F) Western blot of aggregated versus non-aggregated membrane proteins at the ER as in (A) except with dfm1Δ cells containing HMG2-GFP with add-back of EV, DERLIN-1-Myc, and DERLIN-2-Myc. Anti-Myc was used to detect DERLIN-1-Myc and DERLIN-2-Myc. Data information: All detergent solubility assays were performed with 3 biological replicates (N = 3). ER, endoplasmic reticulum; EV, empty vector. https://doi.org/10.1371/journal.pbio.3001950.g004 It appears Dfm1—independent of its Cdc48 recruitment function—functions as a chaperone-like protein to influence the aggregation of misfolded membrane proteins. We next explored additional Dfm1 residues that are required for solubilizing membrane substrates. Accordingly, mutants in the conserved rhomboid motifs (AR and Ax3G) were employed in the detergent solubility assay. DFM1-AR and DFM1-Ax3G add-back resulted in aggregated HMG2 (Fig 3A). Similarly, retrotranslocation defective Dfm1 mutants in Loop 1 (F58S, L64V, and K67E) and TMD 2 mutants (Q101R and F107S) in the detergent solubility assay were not capable of solubilizing Hmg2 (Fig 3B). This all-or-nothing effect that Dfm1’s presence has on aggregation led us to determine whether Dfm1 binds to Hmg2 even after solubilization with DDM. Indeed, using co-immunoprecipitation, we found that Dfm1 physically interacts with solubilized Hmg2 (Fig 3C). Altogether, Dfm1 is critical in influencing the aggregation state of its ERAD membrane substrate (Fig 3D). Although the ability to recruit Cdc48 is vital for Dfm1’s retrotranslocation function, it is not required for this newly established chaperone-like function. The chaperone-like function of Dfm1 is generalizable to other misfolded membrane proteins but not non-membrane misfolded proteins. We tested other membrane ERAD membrane substrates targeted by the HRD (Pdr5*) or DOA (Ste6*) pathways, and they were both completely solubilized in the presence of Dfm1 in the detergent solubility assay (Fig 4C and 4D). In contrast, the misfolded ER luminal protein, CPY*, was completely solubilized regardless of the presence or absence of Dfm1 (Fig 4E). Notably, we also observed that Derlin-1 and Derlin-2 were able to prevent aggregation of Hmg2 in dfm1Δ cells, indicating that other derlin proteins have a conserved chaperone-like function (Fig 4F). We also investigated whether Hmg2-GFP appeared in puncta in dfm1Δ cells through confocal microscopy. We found no significant difference between the percentage of GFP in puncta or the number of puncta with add-back of WT DFM1 or any of the DFM1 mutants in dfm1Δ cells (S1D–S1F Fig). This is in line with the view of some in the field that toxic aggregates are generally below the visible detection limit for confocal microscopy and that puncta identified through microscopy tend to be representative of sequestrosomes, a cellular adaptation to the accumulation of aggregation prone proteins [49].

Misfolded membrane proteins do not activate the unfolded protein response The canonical ER stress pathway triggered by the accumulation of misfolded proteins is the unfolded protein response (UPR) [50]. The UPR is known to be induced by the accumulation of misfolded soluble proteins within the ER lumen. To test if misfolded membrane protein accumulation at the ER activates UPR, we used a fluorescence-based flow cytometry assay. In this assay, yeast cells encoding both a galactose inducible misfolded protein or empty vector (EV) and UPR reporter 4xUPRE-GFP were treated with or without 0.2% galactose and 2 μg/mL of the ER stress inducing drug tunicamycin or DMSO as the vehicle control. GFP expression was measured by flow cytometry for 5 hours following galactose treatment. We found that GFP expression did not increase by 5 hours post-galactose addition in dfm1Δ cells compared to pdr5Δ cells expressing any of the following substrates tested: Hmg2, Ste6*, or EV (Fig 5A–5D, 5G and 5H). Dfm1Δ cells are able to activate UPR, as addition of tunicamycin to these cells allowed them to activate the UPR at similar levels as pdr5Δ cells (Fig 5A–5H). As expected, expression of the ER luminal substrate CPY* activated the UPR in dfm1Δ cells and pdr5Δ cells (Fig 5E and 5F). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 5. Misfolded membrane protein stress in dfm1Δ cells does not activate the UPR. (A) UPR activation for indicated strains with overexpression of a misfolded integral membrane protein. pdr5Δ cells containing GAL pr -Hmg2-6Myc and 4xUPRE-GFP (a reporter that expresses GFP with activation of the UPR) were measured for GFP expression using flow cytometry every hour for 5 hours starting at the point of galactose induction and tunicamycin or equivalent volume of DMSO was added at the 1-hour time point. Figure depicts the GFP fluorescence in A.U. for indicated conditions 5 hours post-galactose addition. In figure legend, “Gal” indicates addition of 0.2% galactose to cultures and “Raf” indicates addition of 0.2% raffinose to culture, and “Tm” indicates presence (+) or absence (-) of 2 μg/mL tunicamycin. (B) Flow cytometry-based UPR activation assay as described in (A) except using dfm1Δ cells. (C, E, and G) Flow cytometry-based UPR activation assay as described in (A) except using cells containing GAL pr -Ste6*-GFP, GAL pr -CPY*-HA, or EV, respectively. (D, F, and H) Flow cytometry-based UPR activation assay as described in (B) except using cells containing GAL pr -Ste6*-GFP, GAL pr -CPY*-HA, or EV, respectively. Data information: All data are measured mean ± SEM; N = 3 biological replicates. The data underlying this figure can be found in Table A–H in S1 Data (Sheet 1). A.U., arbitrary unit; EV, empty vector; UPR, unfolded protein response. https://doi.org/10.1371/journal.pbio.3001950.g005 Our findings from flow cytometry experiments were further corroborated by measuring Hac1 splicing via polymerase chain reaction (PCR) (S2A and S2B Fig). When the UPR is active, the mRNA of the transcription factor Hac1 is spliced to create a transcript 252bp shorter than the full-length transcript [51]. Samples with a band for both spliced and unspliced Hac1 indicated UPR activation, while a single band of the unspliced variant indicated no UPR activation. The results from these experiments were in agreement with the flow cytometry-based assay; we found no HAC1 splicing with misfolded membrane protein overexpression in dfm1Δ cells (S2A and S2B Fig).

Accumulation of misfolded membrane proteins up-regulates proteasome components After determining the UPR is not activated in dfm1Δ cells expressing Hmg2, we next sought to determine the transcriptional changes that occur with misfolded membrane protein stress. To address this question, we utilized RNA sequencing (RNA-seq). We prepared and sequenced cDNA libraries from mRNA extracted from pdr5Δ cells, hrd1Δpdr5Δ cells, and dfm1Δ pdr5Δ cells containing galactose inducible Hmg2 or EV, 2 hours post-galactose treatment. These yeast strains were generated from a yeast knockout collection with the BY4742 strain background, and pdr5Δ cells are commonly used as the wild-type background for the knockout collection. We validated using the substrate-toxicity assay that dfm1Δ pdr5Δ+Hmg2 strains in this background also display a growth defect (S3A Fig). We used principal component analysis (PCA) to determine genes that were up-regulated and down-regulated most in dfm1Δ cells expressing Hmg2 versus the control strains; WT+EV, WT+Hmg2, hrd1Δ+EV, hrd1Δ+Hmg2, and dfm1Δ+EV (S1 Data). Principal component 1 (PC1) value of all replicate strains except for dfm1Δ+Hmg2 cells clustered closer to each other than they did to either replicate of the dfm1Δ+Hmg2 cells, indicating that these strains were transcriptionally distinct from the others sequenced (S3B Fig). Additionally, the dfm1Δ+Hmg2 replicates were fairly distinct from each other, so while there were genes up-regulated in both replicates, there were also variable transcriptional changes (S3B Fig). This variability is likely representative of biological variability in these strains rather than experimental variability as it was only observed between these replicates and not replicates of the other strains tested. Up-regulated (+PC1 values) and down-regulated (-PC1 values) genes in dfm1Δ+Hmg2 cells were used for gene ontology (GO) analysis. The most overrepresented group of up-regulated genes were those classified as being involved in “Proteasomal Ubiquitin-Independent Protein Catabolic Processes,” “Regulation of Endopeptidase Activity,” and “Proteasome Regulatory Particle Assembly” (S3D Fig). Several proteasome subunits were represented in this list of up-regulated genes. The most overrepresented group of down-regulated genes in this dataset were those classified as being involved in “rRNA Export from Nucleus,” “rRNA Transport,” and “Translational Termination” (S3E Fig). Because a down-regulation of the mRNA for genes encoding ribosomal proteins is a general feature of stressed yeast cells [52], we focused on the up-regulation of proteasome components. Plotting the PC1 and PC2 values for dfm1Δ+Hmg2 cells for the highest PC1 value genes, we observed a large overlap between genes in this dataset and those that are targets of the transcription factor Rpn4 (S3C Fig, highlighted in red).

The transcription factor Rpn4 is involved in misfolded membrane protein stress Rpn4 is a transcription factor that up-regulates genes with a proteasome-associated control element (PACE) in their promoters [53]. From our RNA-seq data, there was a remarkably high overlap between the genes that were observed to be up-regulated in dfm1Δ cells expressing Hmg2 and those that are known Rpn4 targets [53]. We reasoned that Rpn4 may be involved in adapting cells to misfolded membrane protein stress and predicted rpn4Δ cells should phenocopy dfm1Δ cells by exhibiting a growth defect induced by ERAD membrane substrates. Using the substrate-toxicity assay, we found expression of misfolded membrane proteins in rpn4Δ cells resulted in a growth defect equivalent to that seen in dfm1Δ cells (Figs 6A and S4A), indicating that Rpn4 is also required for alleviating misfolded membrane protein stress. As with dfm1Δ cells, this effect was specific to membrane protein expression, as expression of CPY* in rpn4Δ cells did not result in a growth defect (S4B Fig). This is in line with previous research demonstrating Rpn4 is activated in response to misfolded membrane protein accumulation and that misfolded membrane protein expression can result in proteasome impairment, even in WT cells [54,55]. Finally, we tested a transcription factor that can regulate Rpn4 and has many overlapping transcriptional targets with Rpn4, Pdr1 [56], and did not observe any growth defect in pdr1Δ+Hmg2 cells (S4C Fig). As a readout for Rpn4 activity, we measured the abundance of a GFP-tagged version of the proteasome component Pre6-GFP in dfm1Δ+Hmg2 cells 0- and 5-hours after galactose induction through flow cytometry. Pre6 is a component of the 20S core of the proteasome that can be transcriptionally up-regulated by Rpn4, and Pre6-GFP has been used by others as a marker for the proteasome [57,58]. In comparison to WT control strains and other substrates tested, dfm1Δ+Hmg2 had a significant increase in Pre6-GFP after 5 hours (Fig 6B and 6C). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 6. Misfolded membrane protein toxicity results in proteasome impairment. (A) WT, dfm1Δ, and rpn4Δ cells containing either GAL pr -HMG2-GFP or EV were compared for growth by dilution assay. Each strain was spotted 5-fold dilutions on glucose or galactose-containing plates to drive Hmg2-GFP overexpression, and plates were incubated at 30°C. Three biological replicates and 2 technical replicates (N = 3). (B) PRE6-GFP levels as measured by flow cytometry at 0 versus 5 hours post-galactose induction in WT cells containing either EV, GAL pr -CPY*-HA, or GAL pr -HMG2-GFP. (C) Pre6-GFP levels as in (B) except in dfm1Δ cells. (D) Quantification of CFUs formed on appropriate selection plates from proteasome sensitivity inhibition assay. pdr5Δ cells containing SUS-GFP or EV in log phase were treated with 25 uM of proteasome inhibitor MG132 or equivalent volume of DMSO for 8 hours and samples were diluted 1:500 and 50 uL of each sample was plated. (E) Proteasome sensitivity assay as in (D) except using hrd1Δpdr5Δ cells. (F) Proteasome sensitivity assay as in (D) except using dfm1Δhrd1Δpdr5Δ cells. Data information: For (B) and (C), all data are mean ± SEM, with 7 biological replicates (N = 7). For (D), (E), and (F), all data are mean ± SEM, 3 biological replicates and 2 technical replicates (N = 3); statistical significance is displayed as two-tailed unpaired t test, *P < 0.05, ns, not significant. The data underlying this figure can be found in Table I–K in S1 Data (Sheet 2). CFU, colony-forming unit; EV, empty vector. https://doi.org/10.1371/journal.pbio.3001950.g006

Misfolded membrane protein stress in dfm1Δ cells leads to proteasome impairment Because Rpn4 appears to be active in membrane protein-stressed dfm1Δ cells, we hypothesized that proteasome function is impacted in dfm1Δ cells expressing an integral membrane protein. We tested this using an MG132 sensitivity assay developed by the Michaelis lab [54]. MG132 is a drug that reversibly inhibits proteasome function [59]. For this assay, cells in liquid culture were treated with MG132, plated, and counted the number of colony-forming units (CFUs) resulting from each strain. Due to the risk of the retrotranslocation defect being suppressed in dfm1Δ cells with constitutive expression of a misfolded membrane protein, and thus possibly artificially increasing the number of CFUs resulting from treatment of dfm1Δ cells with MG132, we opted to instead test dfm1Δ hrd1Δpdr5Δ cells. These cells are unable to suppress the retrotranslocation defect of dfm1Δ cells, due to the absence of Hrd1, which has been characterized to function as an alternative retrotranslocon for membrane substrates when Dfm1 is absent [33]. We utilized the engineered misfolded membrane protein SUS-GFP as the substrate for these experiments. SUS-GFP contains the RING domain of Hrd1 and catalyzes its own ubiquitination, thus still causing the stress that is elicited by ubiquitinated misfolded membrane proteins in dfm1Δ cells [60]. We predicted that cells with compromised proteasome function would be more sensitive to MG132 treatment, resulting in fewer CFUs. Strikingly, no CFUs resulted from MG132 treated dfm1Δhrd1Δpdr5Δ cells constitutively expressing SUS-GFP (Fig 6F). All others strains and treatments tested did not show as dramatic of a change in the number of CFUs, either with MG132 or DMSO treatment (Fig 6D–6F). These results demonstrate that proteasome function is impacted in dfm1Δ cells with misfolded membrane protein accumulation.

Misfolded membrane protein stress does not cause proteasome sequestration One hypothesis that we explored to understand the mechanism by which proteasomes are impaired with misfolded membrane protein stress was direct sequestration of proteasomes at the ER. Using dfm1Δ cells expressing EV or Hmg2, we used western blotting to detect ER recruitment of Pre6, a proteasome component (Fig 6G). Proteasome recruitment was similar between both strains. We also tested aggregation versus solubility of Pre6 at the ER in both strains and this was also not affected in either strain (Fig 6G). These results indicate an indirect mechanism for proteasome impairment in membrane protein-stressed cells.

Growth defect in dfm1Δ cells is ubiquitination dependent The observation that a growth defect triggered by misfolded membrane proteins is only seen in the absence of Dfm1, and not in cells lacking either of the ER E3 ligases Hrd1 and Doa10, led us to hypothesize that this growth defect is dependent upon ubiquitination of the misfolded membrane proteins. The substrate-toxicity assay results using cdc48-2 cells indicate that the growth defect is not solely due to defective ERAD or the accumulation of ubiquitinated misfolded membrane proteins. Nonetheless, we still explored the possibility that misfolded membrane protein-induced toxicity is dependent on substrate ubiquitination. We examined whether growth defects were seen in either dfm1Δhrd1Δ or dfm1Δdoa10Δ cells expressing either Hmg2 (an Hrd1 target) or Ste6* (a Doa10 target), respectively (Fig 7A). These results showed no growth defect in the double mutants for which the membrane protein expressed was not ubiquitinated by the absent E3 ligase: dfm1Δhrd1Δ cells expressing Hmg2 and dfm1Δdoa10Δ cells expressing Ste6* (Fig 7A). We validated that substrates were indeed not ubiquitinated in E3 ligase knockouts via western blot for ubiquitin (Fig 7B). In contrast, a growth defect was observed in the double mutants for which the absent E3 ligase did not participate in ubiquitination of the expressed membrane protein: dfm1Δhrd1Δ expressing Ste6* and dfm1Δdoa10Δ expressing Hmg2. This indicates that growth stress in dfm1Δ cells is dependent upon ubiquitination of the accumulated misfolded membrane protein. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 7. Ubiquitin stress contributes to misfolded membrane protein toxicity. (A) WT, dfm1Δ, dfm1Δhrd1Δ, and dfm1Δdoa10Δ cells containing either GAL pr -Hmg2-GFP, GAL pr -STE6*-GFP, or EV were compared for growth by dilution assay. Each strain was spotted 5-fold dilutions on glucose or galactose-containing plates to drive Hmg2-GFP overexpression, and plates were incubated at 30°C. (B) Indicated strains expressing either Hmg2-GFP or Ste6*-GFP were grown to log-phase, lysed, and microsomes were collected and immunoprecipitated with α-GFP conjugated to agarose beads. Samples were then subjected to SDS-PAGE and immunoblotted by α-Ubiquitin and α-GFP. Three biological replicates (N = 3). (C) Dilution assay as described in (A) except using WT and dfm1Δ cells containing either GAL pr -Hmg2-GFP, GAL pr -Hmg2-K6R-GFP, GAL pr -Hmg2-K357R-GFP, GAL pr -Hmg2- (K6R and K357R)-GFP, or EV. (D) WT and dfm1Δ cells containing either CUP1 pr -Ub or EV and GAL pr -HMG2-GFP or EV were compared for growth by dilution assay. Each strain was spotted 5-fold dilutions on glucose or galactose-containing plates to drive Hmg2-GFP overexpression, and plates were incubated at 30°C. Galactose plates containing 50 μM Cu2+ were used to allow expression of Ub driven by the CUP1 promoter. (E) Western blot of monomeric ubiquitin in WT, dfm1Δ, and hrd1Δ expressing HMG2-GFP. Anti-ubiquitin was used to blot for ubiquitin and anti-PGK1 was used to blot for PGK1 as a loading control. (F) Quantification of western blots from (E). Each strain was normalized to PGK1 and the monomeric ubiquitin quantification of WT+HMG2-GFP was used to normalize all strains. (G) Dilution assay as described in (A) dfm1Δ, ubp9Δ, ubp14Δ, and doa4Δ cells. (H) Dilution assay as described in (A) except using WT, dfm1Δ, and ubp6Δ cells containing either GAL pr -HMG2-GFP or EV. (H) Western blot of aggregated versus soluble membrane proteins at the ER. Lysates from dfm1Δ cells containing HMG2-GFP or HMG2-K6R-GFP with EV or DFM1-HA were blotted using anti-GFP to detect Hmg2. T is total fraction, P is pellet (ER aggregated) fraction, and S is soluble (ER non-aggregated) fraction. Data information: All dilution growth assays were performed in 3 biological replicates and 2 technical replicates (N = 3). For (F), all data are mean ± SEM, 3 biological replicates (N = 3); statistical significance is displayed as two-tailed unpaired t test, *P < 0.05, ns, not significant. Detergent solubility assay in (H) was performed with 3 biological replicates (N = 3). The data underlying this figure can be found in Table L in S1 Data (Sheet 3). ER, endoplasmic reticulum; EV, empty vector. https://doi.org/10.1371/journal.pbio.3001950.g007 As an alternative approach to determine if membrane proteins must be ubiquitinated to cause toxicity in the absence of Dfm1, we tested the expression of well-characterized, stabilized Hmg2 mutants. These mutants, Hmg2 (K6R), Hmg2 (K357R), and Hmg2 (K6R, K357R), were previously identified by the Hampton lab in a genetic screen for stabilized Hmg2 mutants [61]. Both K➔R stabilized mutations disrupt Hmg2 ubiquitination, and these sites are hypothesized to be Hmg2 ubiquitination sites. While the Hampton lab has shown that ubiquitination levels of both substrates are nearly undetectable with western blot, they also showed that the K6R mutant is not further stabilized in an ERAD deficient background, while the K357R mutant is slightly more stable in an ERAD deficient background than in a WT background [61]. We propose that because of this slight level of degradation in the K357R mutant, some fraction of this mutant must be ubiquitinated and targeted to the Hrd1 ERAD pathway. Our model predicts that toxicity of misfolded membrane proteins is ubiquitination dependent. Thus, we would expect that the fully stabilized Hmg2-K6R with negligible ubiquitination should not elicit a growth defect, whereas Hmg2-K357R, which is a poor ERAD substrate but still can still be targeted for degradation, should elicit a growth defect in dfm1Δ cells. Indeed, we observed no growth defect in dfm1Δ cells expressing the K6R mutant, while the K357R mutant still showed a growth defect. Moreover, the growth defect is still observed in the double mutant Hmg2- (K6R, K357R), which phenocopies Hmg2-K6R, in that it is completely stabilized, consistent with the model that growth stress in the absence of Dfm1 is dependent on the accumulation of ubiquitinated membrane substrates (Fig 7C).

Ubiquitin homeostasis is disrupted with misfolded membrane protein accumulation There is increasing evidence that suggests ubiquitin homeostasis and maintenance of the free ubiquitin pool is critical for cellular survival under normal and stress conditions [62–65]. Because we observed that growth defect in dfm1Δ cells is dependent on ubiquitination of membrane substrates, we hypothesized that ubiquitin conjugation to accumulating membrane proteins reduces the availability of free ubiquitin, impacting cell viability. One hypothesis that would explain substrate ubiquitination dependency of the growth defect in dfm1Δ cells is that the pool of monomeric ubiquitin is depleted by accumulation of misfolded membrane proteins. If this hypothesis is correct, exogenous ubiquitin should rescue the growth defect seen from substrate-induced stress in dfm1Δ cells. To that end, dfm1Δ+Hmg2 cells harboring a plasmid containing ubiquitin under the control of the copper inducible promoter CUP1 [66] were tested in the substrate-toxicity assay. These cells were plated on 2% galactose and 50 μM copper to induce expression of Hmg2 and ubiquitin in dfm1Δ cells, respectively. Notably, supplementation of ubiquitin restored the growth defect (Fig 7D). We blotted for monomeric ubiquitin to determine whether this pool is depleted in dfm1Δ+Hmg2 cells and found that it was reduced compared to WT and hrd1Δ strains with Hmg2 (Fig 7E and 7F). In contrast, dfm1Δ without overexpressed Hmg2 do not show a decrease in monomeric ubiquitin (S5A and S5B Fig).

Deubiquitinases prevent or resolve misfolded membrane protein stress The deubiquitinase (DUB) Ubp6 is a peripheral subunit of the proteasome and recycles ubiquitin from substrates prior to proteasome degradation [65]. Accordingly, ubp6Δ cells were employed in the substrate-toxicity assay to determine whether this protein is involved in alleviating misfolded membrane protein stress by replenishing the free ubiquitin pool. By utilizing the substrate-toxicity assay, we found Hmg2 or Ste6* expression causes a growth defect in ubp6Δ cells (Figs 7F and S5C). Like dfm1Δ and rpn4Δ cells, this growth defect was specific to misfolded membrane proteins and was not observed with CPY* (S6D Fig). To confirm whether this effect was specific to Ubp6, we also tested DUB Doa4, another regulator of free ubiquitin, in the substrate-toxicity assay. Unexpectedly, we found that doa4Δ cells phenocopy ubp6Δ cells with Hmg2 expression (S6A Fig). From this observation, we tested a collection of DUB KOs in the substrate-toxicity assay. Of the 14 yeast DUBs tested (out of 22 DUBs total), we observed a growth defect with both ubp9Δ and ubp14Δ cells (Figs 7H and S6B). Interestingly, Ubp6, Doa4, and Ubp14 have all previously been implicated in ubiquitin homeostasis and, to date, no research has been conducted into the specific role of Ubp9 [67]. Interestingly, when a misfolded cytosolic substrate, ΔssCPY*, is expressed in ubp6Δ, doa4Δ, ubp9Δ, and ubp14Δ cells, there is no growth defect observed (S6C Fig).

Absence of deubiquitinases and RPN4 in combination with DFM1 do not exacerbate toxicity We tested double knockouts of dfm1Δrpn4Δ, dfm1Δubp6Δ, and rpn4Δubp6Δ cells expressing Hmg2 in the substrate-toxicity assay to determine whether these genetic backgrounds display the same or different growth defect than either of the single knockouts. Expression of either Hmg2 or Ste6* in either dfm1Δrpn4Δ orF020dfm1Δubp6Δ cells resulted in a growth defect that phenocopied that observed in any of the single knockouts (S7A and S7B Fig), whereas expression of CPY* showed no growth defect (S7C Fig). In contrast, rpn4Δubp6Δ cells showed a growth defect in the absence of substrates, whereas rpn4Δ and ubp6Δ displayed normal growth. Moreover, rpn4Δubp6Δ cells along with expression of Hmg2 or Ste6* resulted in synthetic lethality (S7A–S7C Fig). This indicates that there is an exacerbation of stress in rpn4Δubp6Δ background, whereas there is no increase in toxicity when RPN4 or UBP6 are knocked out in combination with DFM1. It is likely that there are several parallel pathways contributing to preventing stress from misfolded membrane proteins and resolving this stress, and Dfm1 appears to be one of the major mediators of misfolded membrane stress prevention. We also tested expression of previously described Hmg2 mutants K6R and K357R in rpn4Δ and ubp6Δ cells (S7D Fig). As with dfm1Δ cells expressing these mutants, expression of Hmg2-K6R does not cause toxicity while Hmg2-K357R does cause toxicity in both rpn4Δ and ubp6Δ. Thus, ubiquitination of misfolded membrane proteins influences toxicity in dfm1Δ, rpn4Δ, and ubp6Δ cells.

Misfolded protein aggregation toxicity requires protein ubiquitination We originally hypothesize that ubiquitination of membrane proteins was promoting those proteins to become aggregated. To test this hypothesis, we measured aggregation versus solubility of Hmg2-K6R in dfm1Δ cells in the detergent solubility assay (Fig 7I). Surprisingly, Hmg2-K6R phenocopied Hmg2 in dfm1Δ cells, with virtually all of the protein being in the aggregated fraction. This demonstrates ubiquitin does not influence aggregation and not all misfolded membrane protein aggregates are toxic.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001950

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/