ER-misfolded proteins become sequestered with mitochondria and impair mitochondrial function

Proteostasis is a challenge for cellular organisms, as all known protein synthesis machineries are error-prone. Here we show by cell fractionation and microscopy studies that misfolded proteins formed in the endoplasmic reticulum can become associated with and partly transported into mitochondria, resulting in impaired mitochondrial function. Blocking the endoplasmic reticulum-mitochondria encounter structure (ERMES), but not the mitochondrial sorting and assembly machinery (SAM) or the mitochondrial surveillance pathway components Msp1 and Vms1, abrogated mitochondrial sequestration of ER-misfolded proteins. We term this mitochondria-associated proteostatic mechanism for ER-misfolded proteins ERAMS (ER-associated mitochondrial sequestration). We testify to the relevance of this pathway by using mutant α-1-antitrypsin as an example of a human disease-related misfolded ER protein, and we hypothesize that ERAMS plays a role in pathological features such as mitochondrial dysfunction.

P rotein homeostasis depends on a delicate balance between maintaining protein conformation, refolding misfolded proteins and degrading damaged proteins. This process is most finely controlled in the endoplasmic reticulum (ER). The ER is a main hub of protein production and serves as the synthesis site for secretory and membrane proteins. More than one-third of a cell's proteome is typically synthesized in the ER 1 . Proteins enter the ER unfolded and are processed by a complex protein folding machinery consisting of foldases, oxidoreductases, chaperones, lectins and oligosaccharide processing enzymes to facilitate proper folding 2 . In general, proteins only leave the ER after they have reached their native states. However, protein folding in the ER is relatively slow and inefficient and requires the continuous removal of polypeptides that fail to attain their native structure 2,3 . To offset the harmful consequences of misfolded protein accumulation, ER-retained misfolded proteins are most commonly destroyed. The primary mechanism of disposal is referred to as ER-associated degradation (ERAD). During ERAD, terminally misfolded proteins are delivered to the proteasome, which resides in the cytoplasm. Destruction of ERAD substrates thus requires delivery from the ER by retrotranslocation into the cytoplasm, where they become subsequently processed and degraded by the ubiquitin-proteasome system 4,5 .
ER and mitochondria form physical contact sites, resulting in electron-dense structures that bridge these two organelles 6,7 . These membrane contact sites are also referred to as mitochondria-associated membranes (MAMs) 8 . More recently, a protein complex termed ERMES (ER-mitochondria encounter structure) has been identified that tethers ER and mitochondria and that facilitates the exchange of Ca ++ and lipids between these two organelles 9 . ER-mitochondria tethers have been implicated in lipid homeostasis, the formation of autophagosomes, and apoptotic signalling [10][11][12][13] , contributing to the view that ERmitochondria connections are hubs for integrating interorganelle crosstalk. Here, we addressed the question of whether further to these functions, mitochondria play a role in ER proteostasis.

Results
ER-Fluc is targeted to ER, processed, and misfolded. To assess and follow the trafficking of ER-localized misfolded proteins we used a modified misfolding-prone firefly luciferase reporter (Fluc) 14 . We targeted Fluc to the ER with the help of the prolactin signal peptide (PLN) placed upstream, and a KDEL ER-retention sequence placed downstream of the luciferase coding sequence (ER-Fluc construct). As a control, we used a similar construct without the PLN ERtargeting and the KDEL ER-retention signals, so that Fluc remains in the cytoplasm (cyto-Fluc construct). To assess the subcellular localization of the ER-Fluc constructs, HEK293 cells were transiently transfected with ER-Fluc and cyto-Fluc constructs and the cytosolic and ER fractions were isolated and analyzed by Western blot. Compared to cyto-Fluc which was found predominantly in the cytosolic fraction, ER-Fluc was significantly enriched in the ER fraction (Fig. 1a). Detection of Fluc by Western blot demonstrated two different molecular weight forms of ER-Fluc and we hypothesized that the higher molecular weight form reflects ER-mediated Nglycosylation. To corroborate the identity of the higher molecular weight form of ER-Fluc we used an endoglycosidase H (Endo H) treatment to deglycosylate the protein. Subsequent high-resolution SDS-gel analysis confirmed the identity of the higher molecular weight form as N-glycosylated Fluc and the lower molecular weight form as unglycosylated Fluc, both with cleavage of the signal peptide, demonstrating that ER-Fluc had passed into the ER (Fig. 1b). Apparently, while cleavage of the signal peptide is complete, glycosylation of the Fluc reporter in the ER is partial, most likely as a result of its non-ER nature. Further evidence for expression of the ER-Fluc reporter in the ER was provided by confocal microscopy, which revealed co-localization of ER-Fluc with the ER-marker protein calreticulin. Transfection with the HA-tagged ER-resident protein β-galactosyltransferase (Gal-Trans) served as a positive control (Fig. 1c).
As a possible indicator of protein misfolding we studied the activity of Fluc as enzymatic activity is reduced when Fluc is misfolded. The enzymatic activity of ER-Fluc was dramatically reduced compared to cyto-Fluc (Fig. 1d). This loss of activity cannot be explained by N-glycosylation of the ER-Fluc reporter or by the PLN-signal interfering with luciferase's enzymatic activity, as a significant part of ER-Fluc is cleaved and not glycosylated (see Fig. 1b). We hypothesize that the inactivity of the ER-Fluc most likely reflects incorrect folding inside the endoplasmic reticulum.
Mitochondrial localization of misfolded ER-Fluc. We studied the cellular localization of ER-Fluc using cell fractionation experiments. Surprisingly, Western blot analysis demonstrated that a substantial part of ER-Fluc is present in purified mitochondrial fractions (Fig. 1e). Transfections with mitochondria-targeted firefly luciferase (MTS-Fluc) and myc-tagged ER-resident protein Gal-Trans were used as positive and negative controls, respectively (Fig. 1e). Evidence for expression of mitochondria-targeted firefly luciferase (MTS-Fluc) in mitochondria was provided by immunofluorescence studies, which revealed co-localization with the mitochondrial marker protein TOM20 ( Supplementary Fig. 2). Using the mitochondrial matrix protein COX1 for normalization, we estimated the relative amount of transfected reporters enriched in the mitochondrial fraction. Compared to the native ER-resident protein Gal-Trans, a significant amount of misfolded ER-Fluc was recovered in the mitochondrial fraction, for quantification see Fig. 1f and Supplementary Fig. 3.
The detection of the two major forms of ER-Fluc in the mitochondrial fraction, representing the ER-associated modifications of cleavage of the signal peptide and N-glycosylation, indicates that the ER-Fluc present in the mitochondrial fraction has passed through the ER. To further assess the association of misfolded ER-Fluc with mitochondria, we treated purified mitochondrial fractions with proteinase K to digest ER-Fluc attached to the outside of mitochondria. In addition, mitochondrial samples were subjected to proteinase K treatment following permeabilization of the outer mitochondrial membrane by hypotonic shock (20 mM HEPES). The ER-Fluc associated with the mitochondrial fraction was partially digested upon treatment with proteinase K. However, a significant part of the ER-Fluc stayed intact after treatment with proteinase K alone or following permeabilization of the outer membrane by hypotonic shock and was degraded only after treatment with proteinase K in combination with Triton X-100, dissolving both outer and inner mitochondrial membrane ( Supplementary Fig. 4A, for quantification by densitometry analysis see Supplementary  Fig. 4B). As control we used cells expressing mitochondriatargeted Fluc. MTS-Fluc in the mitochondrial fractions was protected from treatment with proteinase K alone or in combination with hypotonic shock and was degraded only after a combined treatment with proteinase K and Triton X-100 ( Supplementary  Fig. 4A, B). As controls for mitochondrial integrity and intramitochondrial localization we used proteins TOM70, TIM23, and COX1 for outer membrane, intermembrane space, and mitochondrial matrix, respectively.
Microscopy studies in ER-Fluc expressing cells. The results of the cell fractionation experiments suggested that misfolded ER-Fluc is associated with mitochondria and becomes partially imported.
However, cell fractionation experiments inherently are compromised by some pollution of the mitochondrial fraction with ER proteins. In addition, we cannot exclude that small amounts of protein aggregates accumulating in ER membrane inclusions may fragment away from the main ER network and potentially co-fractionate with mitochondria. To address these limitations, we decided to use a previously described split-GFP mito-localization system combined with confocal and electron microscopy and FACS analysis 15 . Split-GFP has the advantage of low off rates allowing for augmentation of even small signals. Briefly, this system consists of two vectors; one carrying the first ten β-strands of GFP fused to mitochondria-targeted mCherry (MTS-mCherry-GFP 1-10 ) and one carrying the eleventh β-strand of GFP (GFP 11 ) fused to a reporter protein. Manders' correlation coefficient was used to demonstrate co-localization of the mitochondria-targeted mCherry-GFP 1-10 with TOM20, a mitochondrial protein of the outer membrane ( Supplementary Fig. 5). Mitochondria-localized GFP fluorescence is observed only when the GFP 11 -fused reporter associates with mitochondria and complements the mitochondrial-targeted mCherry-GFP [1][2][3][4][5][6][7][8][9][10] . In our experiments we used the ER-localized wild-type luciferase (ER-FlucWT-HA-GFP 11 ), a double mutant aggregation-prone luciferase variant (FlucDM) targeted to ER (ER-FlucDM-HA-GFP 11 ), the non-misfolding ER-resident protein Gal-Trans, and MTS-HA-GFP 11 (as a positive control) 14 . Staining with antibodies against calreticulin was used to demonstrate ER localization of the ER-Fluc constructs and of ER-Gal-Trans (Fig. 1c). Alongside each reporter, cells were transiently co-transfected with MTS-mCherry-GFP 1-10 and the resulting mitochondria-associated GFP fluorescence was measured by FACS and visualized by confocal microscopy. We further included cotransfection with MTS-Fluc-HA-GFP 11 together with MTS-mCherry-GFP 1-10 in this analysis to fully represent the reporters used in the cell fractionation experiments, i.e., ER-Fluc, MTS-Fluc, and ER-Gal-Trans. In addition to subcellular localization, expression levels of the reporter protein were monitored by staining with anti-HA antibodies.
The positive controls MTS-HA-GFP 11 and MTS-Fluc-HA-GFP 11 localized exclusively to mitochondria and cells co-expressing MTS-HA-GFP 11 or MTS-Fluc-HA-GFP 11 together with MTS-mCherry-GFP 1-10 showed a prominent split-GFP signal (Fig. 2b, for quantification Fig. 2c, d). The misfolding ER reporters ER-Fluc-WT-HA-GFP 11 and ER-FlucDM-HA-GFP 11 showed a split-GFP signal of varying intensity in cells co-expressing MTS-mCherry-GFP 1-10 . In contrast, the non-misfolding ER reporter ER-Gal-Trans-HA-GFP 11 showed only a minor, if any, split-GFP signal (Fig. 2a, b, for quantification Fig. 2c, d). These data were further corroborated using Pearson's correlation coefficient for co-localization analysis of split-GFP and MTS-mCherry fluorescent signals (Fig. 2e). To control for the transfection and expression efficiency of the reporters, we used antibodies specific to the HA-tag, and quantified the percentage of cells showing co-localization of split-GFP with MTS-mCherry signals relative to the total number of cells expressing both HA-tag and mCherry. This revealed that for the positive controls MTS-HA-GFP 11 and MTS-Fluc-HA-GFP 11 more than 90% of cells coexpressing MTS-HA-GFP 11 or MTS-Fluc-HA-GFP 11 together with MTS-mCherry-GFP 1-10 showed a split-GFP signal. In cells coexpressing one of the misfolding ER-Fluc-HA-GFP 11 reporters together with MTS-mCherry-GFP 1-10 we observed 15-20% split-GFP positive cells. In contrast, very little, if any, cells expressing ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-021-02873-w split-GFP were found among cells co-expressing the non-misfolding ER-Gal-Trans-HA-GFP 11 reporter together with MTS-mCherry-GFP 1-10 ( Fig. 2f, for illustration see low magnification images Supplementary Fig. 6). Together, the data from the split-GFP experiments support the conclusions from the cell fractionations that, in contrast to the non-misfolded ER reporter protein ER-Gal-Trans, a significant fraction of the misfolded ER reporter protein ER-Fluc becomes associated with mitochondria.
To further demonstrate the mitochondrial localization of the split-GFP signal and to exclude aberrant mislocalization of MTS-mCherry-GFP 1-10 , we used staining with the mitochondrial marker TOM20. The mitochondrial split-GFP signal of the positive controls, MTS-HA-GFP 11 and MTS-Fluc-HA-GFP 11 , colocalized with TOM20 and mCherry fluorescence, as did the split-GFP signal of ER-FlucWT-GFP 11 and ER-FlucDM-GFP 11 (Fig. 2b). Compared to ER-FlucWT, the split-GFP fluorescence was not further increased by the DM mutations as assessed by fluorescence microscopy quantification and FACS analysis (Fig. 2c, d). This observation indicates that misfolding of the FlucWT protein inside the ER is not further aggravated by the additional DM mutations.
We finally used a correlative light and electron microscopy approach in cells co-transfected with ER-FlucWT-GFP 11 and MTS-mCherry-GFP 1-10 to demonstrate the localization of split-GFP and mCherry within mitochondria 16 . Both fluorescent signals (mCherry, split-GFP) co-localized and were found within mitochondria as revealed by images of ultrathin sections (110 nm) acquired from correlative light and electron microscopy (Fig. 2g). These data largely exclude the possibility that the colocalization of split-GFP and mitochondrial-targeted mCherry observed by confocal microscopy in cells co-transfected with ER-FlucWT-GFP 11 and MTS-mCherry-GFP 1-10 is a mere result of the association of ER-Fluc with the mitochondrial surface. Rather, these data demonstrate that the co-localization of the split-GFP signal with mCherry is at least in part a result of localization of ER-Fluc within mitochondria.
Expression of ER-Fluc impairs mitochondrial function. We next investigated whether the accumulation of misfolded ER-Fluc (ER-FlucWT, ER-FlucDM) in mitochondria affects mitochondrial function compared to transfection with the non-misfolding ER-reporter ER-Gal-Trans. For most of these assays, treatment with the ionophore CCCP served as positive control. We first assessed total mitochondrial content using staining with Mitotracker Deep Red. Decreased Mitotracker DR staining indicated a reduced mitochondrial mass in the ER-Fluc transfected cells (Fig. 3a). We also observed that ROS formation, as assessed by FACS analysis of cells stained with MitoSOX Red, was significantly elevated (Fig. 3b). Mitochondrial membrane potential was determined using the red fluorescence indicator DilC1(5) and found to be decreased in cells expressing ER-Fluc (Fig. 3e). Further, ATP content was significantly lower in cells expressing ER-FlucWT or ER-FlucDM (Fig. 3c). We next shifted the cellular ATP synthesis from glycolysis to oxidative phosphorylation by replacing glucose with galactose in the growth media 17 . Replacing glucose with galactose augmented the decrease in ATP content in both ER-Fluc transfected cells and in wild-type cells treated with CCCP ( Fig. 3d, Supplementary Fig. 7). Alongside these changes in mitochondrial mass, ROS production, membrane potential, and ATP content we found significantly increased mitophagy in ER-Fluc transfected cells. Mitophagy was assessed using cotransfection with pKeimaRed vector followed by FACS analysis (Fig. 3f). Elevated levels of ROS and mitophagy were further accompanied by increased apoptosis, indicated by increased Caspase 3 activity (Fig. 3g). In contrast, in cells expressing the non-misfolding ER-Gal-Trans reporter none of the markers used to assess mitochondrial function (mitotracker, mitoSOX, membrane potential, ATP content, ATP content in the presence of galactose, mitophagy, Caspase 3, for original data underlying the graphs in Fig. 3 see Supplementary Fig. 15A-E) were significantly different from the wild-type HEK293 controls.
To examine a possible mitochondrial localization of misfolded ATZ, we cloned WT-AT and ATZ fused to a myc-tag in an expression vector and transiently transfected HEK293 cells with these constructs. Treatment of total cell lysates with endoglycosidase H followed by Western blot analysis showed that both WT-AT and ATZ are almost completely glycosylated (Fig. 4a), testifying to ER processing. Visualization of the WT-AT and ATZ reporters by confocal microscopy demonstrated ER localization of both WT-AT and ATZ (Fig. 4b). Fractionation experiments followed by Western blot analysis showed the presence of WT-AT and ATZ proteins in ER fractions, as well as in the growth media, where they were secreted. As expected, mutant ATZ is secreted with less efficiency compared to WT-AT. Most notably, mutant ATZ-but not WT-AT-extensively accumulated in the mitochondrial fraction (Fig. 4c), for quantification see Fig. 4d and Supplementary Fig. 8. Treatment of mitochondrial fractions isolated from ATZ expressing cells with proteinase K either alone or in combination with hypotonic shock (20 mM HEPES) revealed that the majority of misfolded ATZ is subject to proteinase K digestion, with a limited amount of ATZ being protected both from treatment with proteinase K and a combined hypotonic shock/proteinase K treatment ( Supplementary Fig. 9). To visualize mitochondrial localization, cells were co-transfected with GFP 11 tagged WT-AT or ATZ together with MTS-mCherry-GFP [1][2][3][4][5][6][7][8][9][10] . This demonstrated significant mitochondrial GFP fluorescence for cells expressing ATZ, but only a weak, if any, split-GFP signal for cells expressing WT-AT (Fig. 4e). These data were corroborated by calculation of the Pearson's correlation coefficient which was used to quantify the co-localization of split-GFP and MTS-mCherry signals for WT-AT and ATZ compared to the positive controls MTS-HA-GFP 11 and MTS-Fluc-GFP 11 (Fig. 4f).
We wished to corroborate the confocal microscopy results obtained with the split-GFP system by direct tagging of ATZ with a fluorescent protein. Transfection of HEK293 with an ATZ reporter labelled with BFP showed co-localization of BFP with both mitochondrial reporter MTS-mCherry and mitochondrial marker TOM20 (Supplementary Fig. 10A). Further studies combining the BFP-tag with the split-GFP system using a double BFP-and HA-tagged ATZ GFP 11 reporter and MTS-mCherry-GFP 1-10 together with staining of the mitochondrial marker TOM20 demonstrated that while the majority of BFP-labelled ATZ was localized outside mitochondria, i.e., in the ER, a significant part of BFP-labelled ATZ showed a split-GFP signal, which co-localized with TOM20 and MTS-mCherry. Similar results were seen when staining the HA-tag of ATZ GFP 11 with HA-antibodies ( Supplementary Fig. 10C). For more precise localization of transfected ATZ, we combined immunolabelling with electron microscopy (EM) in ultrathin sections of HEK cells transfected with ATZ-myc vector. EM-immunolabelling using anti-myc antibodies revealed the presence of ATZ inside mitochondria, in addition to the expected majority of ATZ being associated with ER (Fig. 4g). Mitochondrial localization of misfolded ER-Fluc in yeast Saccharomyces cerevisiae requires ERMES. To study whether mitochondrial sequestering of misfolded ER proteins is evolutionary conserved, we used split-GFP constructs adapted to the yeast system. To target Fluc to the ER, the signal peptide of the native yeast ER protein KAR2 was placed upstream and the HDEL ER-retention sequence downstream of luciferase resulting in yeast yER-FlucDM-GFP. Co-transformation of yER-FlucDM-GFP with yElo3-mCherry, Elo3 is a yeast ER-resident protein, revealed localization of yER-FlucDM-GFP in the ER (Supplementary Fig. 11) 19 . Compared to the homogeneous yElo3-mCherry signal, the yER-Fluc signal additionally revealed irregular blobs indicative of protein aggregates. For the split-GFP assay, performed to assess localization in mitochondria, the yER-FlucDM-GFP 11 fusion protein was used as a reporter and yMTS-GFP 11 and yElo3-GFP 11 were used as positive and negative controls respectively. The reporters were co-transformed along with yMTS-mCherry-GFP 1-10 . Mitochondria-associated GFP fluorescence was visualized and quantified using confocal microscopy. Yeast cells transformed with yMTS-mCherry-GFP 1-10 were used as a control for autofluorescence. Confocal microscopy demonstrated absence of a mitochondrial split-GFP signal for cells expressing the GFP 11 -tagged ER-resident protein yElo3-GFP 11 , but a prominent mitochondrial split-GFP signal for yER-FlucDM-GFP 11 , indicating trafficking of the misfolded ER reporter to mitochondria (Fig. 5a, d and e).
Mitochondria have specialized surveillance systems, which target faulty proteins designed for mitochondrial import. MitoCPR (Compromised Protein Import Response) removes stalled precursor proteins accumulated at the mitochondrial outer membrane, while VMS1 is part of a mitochondrial quality pathway that acts on mitochondrial polypeptides stalled on ribosomes at the mitochondrial surface 20,21 . Neither deletion of mitoCPR components MSP1 and TOM70 or of VMS1 in yeast co-expressing yER-FlucDM-GFP 11 and yMTS-mCherry-GFP 1-10 , measurably affected the intensity of the split-GFP signal or its colocalization with the mitochondrial mCherry marker (Fig. 5b, f). Together this suggests that classical quality control pathways acting on faulty proteins designed for mitochondrial import do not affect trafficking of yER-FlucDM to mitochondria.
ER and mitochondria form physical contact sites that facilitate the cross-talk between these two organelles. In yeast, a protein complex termed endoplasmic reticulum-mitochondria encounter structure (ERMES) has been identified which tethers ER and mitochondria 9 . Deletion of MDM10, an integral mitochondrial membrane protein and structural component of ERMES, lead to the previously shown alteration in mitochondrial morphology with the characteristic tubular mitochondrial network being replaced by condensed mitochondria (Fig. 5b) 22,23 . Staining with MitoTracker Green confirmed these condensed structures as mitochondria ( Supplementary Fig. 12). Notably, deletion of MDM10 in cells expressing yER-FlucDM-GFP 11 together with yMTS-mCherry-GFP 1-10 completely abrogated the split-GFP signal associated with mitochondria. In addition to MDM10, deletion of further core components of the ERMES structure, i.e., MMM1 and MDM34, in cells expressing yER-FlucDM-GFP 11 together with yMTS-mCherry-GFP 1-10 , besides resulting in a condensed mitochondrial structure, abrogated the split-GFP signal associated with mitochondria (Fig. 5b, f). Western blots were used to corroborate the expression of yER-FlucDM-GFP 11 and yMTS-mChery-GFP 1-10 in the yeast deletion mutants (Supplementary Fig. 13). None of the genetic deletions affected the split-GFP signal of the positive control yMTS-GFP 11 cotransformed with yMTS-mCherry-GFP 1-10 .
The artificial ER-mitochondria tether ChiMERA 9 has been used previously to compensate for the tethering defects associated with ERMES mutants. We expressed the ChiMERA tether in MDM10 mutants and controlled expression by Western blot (Supplementary Fig. 14). In contrast to complementation with MDM10 (see below), MDM10 deletion mutants expressing yER-FlucDM-GFP 11 and yMTS-mCherry-GFP 1-10 and transformed with the ER-mitochondria tether ChiMERA showed continued absence of the mitochondrial split-GFP signal together with condensed mitochondria (Fig. 5c, f). Besides recapitulating the observation that ChiMERA cannot rescue mitochondrial morphology in MDM10 mutants 9 , this finding indicates that expression of the ChiMERA tether in MDM10 mutants is insufficient for translocation of misfolded ER-Fluc into mitochondria. ERMES mutants frequently lose their phenotype over time 24 due to appearance of suppressor alleles in the endosomal vascular protein sorting 13 (VPS13), which suppress the phenotypic consequences of ERMES deficiency 25 . We wondered whether this suppression may extend to the mitochondrial split-GFP signal absent in MDM10 mutants expressing yER-FlucDM-GFP 11 together with yMTS-mCherry-GFP 1-10 . Indeed, over time we observed adapted MDM10 mutants, which showed restoration of normal mitochondrial morphology together with rescue of the split-GFP signal (see Fig. 5b and f). Sequencing of the VPS13 gene in two adapted MDM10 mutants revealed that adaptation was associated with dominant VPS13 mutations known to restore mitochondrial morphology in ΔMDM10 mutants, i.e., D716Y and G820R 25,26 .
In addition to ERMES, MDM10 is part of the mitochondrial sorting and assembly machinery (SAM). Point mutants of MDM10 have been used to separate the function of MDM10 in the biogenesis of outer membrane proteins and protein assembly (SAM) from its function in morphology and phospholipid homeostasis (ERMES) 22 . We employed these point mutants to address the role of MDM10 in mitochondrial localization of misfolded ER-Fluc. Expression levels of SAM mutant MDM10 and ERMES mutant MDM10 were controlled by Western blot (Supplementary Fig. 14). Complementation of MDM10 deletion mutants expressing yER-FlucDM-GFP 11 and yMTS-mCherry-GFP 1-10 with wild-type MDM10 or dysfunctional SAM mutant MDM10 resulted in recovery of the mitochondrial split-GFP signal along with restored tubular mitochondrial morphology. In contrast, cells complemented with dysfunctional ERMES mutant MDM10 showed continued absence of the mitochondrial split-GFP signal together with condensed mitochondria (Fig. 5c, f). These results suggest that interaction between ER and mitochondria through ERMES is essential for mitochondrial sequestration of misfolded ER-Fluc.

Discussion
In general, the formation and organization of protein aggregates is not random. Recently, misfolded cytosolic proteins have been reported to accumulate at ER-mitochondria contact regions, where they eventually become transported into mitochondria 14,15,[27][28][29] . The findings we describe here add an additional layer to the recently proposed role for mitochondria in proteostasis. Using two different reporters, ER-Fluc and mutant α-1-antitrypsin, and a combination of cell fractionation, biochemistry, confocal microscopy, FACS, and electron microscopy studies, we here demonstrated that misfolded aberrant proteins formed in the ER can become associated with mitochondria and partially imported into the mitochondrial matrix, resulting in impaired mitochondrial function. Demonstration of this trafficking pathway for misfolded ER proteins both in mammalian cells and in yeast, while suggesting that mitochondrial sequestration of aberrant ER proteins is evolutionary ancient, also allowed us to in part address mechanistic questions.
Inactivation of mitochondrial surveillance pathways which act on import of faulty mitochondrial proteins by genetic deletion of MSP1, TOM70, and VMS1 did not affect trafficking of the ER proteins into mitochondria. However, we found that MDM10, MMM1 and MDM34, all components of ERMESmediated ER-mitochondria contacts, are essential for this pathway. Multiple mitochondrial functions have been linked to ERMES, including biogenesis of outer membrane proteins, lipid homeostasis, membrane dynamics and maintenance of mitochondrial morphology 22,30 . In addition to being a part of ERMES, MDM10 is also present in SAM complexes. The MDM10-SAM interaction is required for biogenesis of the mitochondrial outer membrane transporters TOM22 and TOM40 22,23 . Using mutants of MDM10, which separate the MDM10-ERMES function from the MDM10-SAM function, we show that the mitochondrial sequestration of misfolded ER-Fluc is critically dependent on the ERMES function of MDM10. Our data further show that this function of ERMES can be bypassed by dominant mutations in the conserved eukaryotic endosomal protein VPS13.
The involvement of mitochondria in the pathogenesis of ATdeficiency with mutant gain-of-function α1-AT has been a mystery, despite detailed studies of liver pathology in the PiZ mouse model demonstrating significant mitochondrial damage and mitochondrial injury, as well as caspase-3-induced apoptosis in situ 31 . Our investigations suggest that aberrantly folded ATZ in the ER becomes associated with mitochondria. While our fractionation experiments cannot separate mitochondria-associated ATZ from ATZ possibly accumulating in inclusions of ER membrane or turned over by ER-phagy, our imaging experiments using confocal microscopy and a combined light and electron microscopy approach demonstrate that at least part of misfolded ATZ becomes associated with the mitochondrial matrix.
Our findings have clear implications for further understanding the pathogenesis of protein misfolding disorders. There is compelling evidence for two apparently unrelated phenomena in neurodegenerative diseases (NDDs): ER stress and unfolded protein response versus mitochondrial dysfunction [32][33][34][35] . A number of pathogenic disease-specific rogue proteins, e.g. Aβ in Alzheimer's disease (AD), α-synuclein in Parkinson's disease, either pass through the ER or have been shown to accumulate as oligomers or aggregates inside the ER [36][37][38] . ER stress may affect mitochondrial function through interorganelle signalling [39][40][41] . However, the herein described role for mitochondria in sequestering ER-misfolded proteins hints to a possible direct structural link connecting protein misfolding in ER and mitochondrial dysfunction. NDDs have recently been proposed to be disorders of ER-mitochondria connectivity 42 . E.g., some features of AD, particularly those manifesting early in the disease process, are apparently unrelated to Aβ plaque and tangle formation, but imply functions related to mitochondria-associated membranes (MAMs) 43 . The missing mechanistic link in this hypothesis was the cause for perturbations in ER-mitochondrial apposition. Our results, which demonstrate that ER-misfolded proteins are sequestered by mitochondria and that this sequestration requires the ERMES function of ER-mitochondrial contact sites, provide a possible answer to this question.
While we realize the many unknowns and the mechanistic details which remain to be elucidated in this mitochondrial sequestration of glycosylated ER-misfolded proteins, we note that it is not without precedence. Most notably, previous reports largely unrecognized by the field have described the mitochondrial import of glycoproteins previously processed in the endoplasmic reticulum 44,45 . In summary, our data provide evidence of a novel degradation pathway for ER-misfolded proteins which become sequestered with mitochondria, highlighting a novel mechanism of cellular proteostasis. We term this mitochondria-associated proteostatic mechanism for ER-misfolded proteins ERAMS, for ER-associated mitochondrial sequestration.

Materials and methods
Luciferase assay. Enzymatic activity of Fluc was determined using the Luciferase Assay System (Promega) as recently described 14 . In brief, HEK293 cells were lysed, the different fractions (total, cytosol, ER, and mitochondria) obtained and their protein content measured using Micro BCA Protein Assay Kit (Thermo Scientific). Equal protein amount was incubated with Passive Lysis Buffer (total volume 20 μl/ well in 96 well plate) for 15 min at RT, mixed with 100 μl of Luciferase Assay Reagent and bioluminescence was measured using plate reader Cytation 5 (Bio-Tek). The specific luciferase activity was calculated by relating the enzymatic activity of each fraction to the corresponding band intensity quantification by Western blot.
Cell culture and preparation of organelle-enriched fractions. HEK293 cells cultivated in DMEM supplemented with 10% FBS were harvested, resuspended in MTE-P buffer (270 mM D-mannitol, 10 mM Tris base, 0.1 mM EDTA, pH = 7.4, supplemented with 1 mM PMSF), and homogenized using a Dounce homogenizer. Homogenates were centrifuged at 1400 × g for 10 min at 4°C to remove cellular debris. An aliquot of supernatant was kept as total cellular protein fraction, the remaining supernatant was centrifuged at 15,000 × g for 10 min at 4°C, resulting in supernatant A and pellet B.
For preparation of ER-enriched fractions, supernatant A was loaded on a sterile 14 × 89-mm polyallomer tube (Beckman) containing a sucrose gradient (3 ml 1.3 M sucrose, 3 ml 1.5 M sucrose, 2 ml 2.0 M sucrose; 10 mM Tris base, 0.1 mM EDTA, pH = 7.6) and centrifuged at 152,000 × g for 70 min at 4°C, acceleration/ deceleration = 1. The cytosolic fraction was taken from the topmost sucrose layer, and the interface containing the ER was collected using a 20-G needle and 1 ml syringe. The ER containing fraction was transferred to a 5.0 ml, 13 × 51 mm polyallomer tube (Beckman) and supplemented with additional ice-cold MTE-P to a final volume of 4 ml. Tubes were covered with parafilm, mixed by inversion until the suspension was homogeneous and ultracentrifuged at 126,000 × g for 45 min at 4°C. Supernatant was discarded, the pellet was resuspended in ice-cold MTE-P and used as ER-enriched fraction.
For preparation of mitochondria-enriched fractions, pellet B was washed twice with 0.5 ml of ice-cold MTE-P, resuspended in 0.8 ml of ice-cold MTE-P buffer and loaded slowly on top of a sucrose gradient (1 ml of 1.7 M sucrose overlaid with 1.6 ml of 1.0 M sucrose in a sterile 5.0 ml, 13 × 51 mm Beckman polyallomer tube). The tube was filled up with an additional 0.8 ml of ice-cold MTE-P buffer and centrifuged for 22 min at 40,000 × g, 4°C (20600 rpm in an SW50.1 rotor). The interface between the 1.7 M and 1.0 M sucrose layers was collected (0.4 ml), mixed with 1.1 ml of ice-cold MTE-P buffer and centrifuged for 10 min at 15,000 × g, 4°C.
The resulting pellet was resuspended in 100 μl of MTE buffer without PMSF and used as mitochondrial fraction.

Construction of mammalian vectors and cell transfection. For a list of vectors used see
Endo H treatment. Endo H glycosidase (Promega) treatment was done according to the manufacturer's protocol. 60 μl of total cellular protein extract were denatured at 95°C for 5 min using denaturing solution, cooled down to room temperature for 5 min, Endo H buffer (Promega) and 2000 units of Endo H glycosidase were added. Samples were incubated for 3 h at 37°C. The resulting samples were resolved by SDS-PAGE and analyzed by Western blot using HRP-conjugated specific anti-Fluc and anti-Myc antibodies (Abcam). The signal intensities of the Fluc and AAT-Myc bands were quantified using ImageQuant.
Proteinase K treatment. Mitochondrial fractions were prepared as described above and OD at 600 nm was measured. Mitochondrial fractions were adjusted to OD600 1.0-1.5 and 10 μl volumes were used per reaction. Mitochondria were permeabilized using hypotonic shock with 20 mM HEPES solution or treatment with 2% Triton X-100. After 30 min incubation at 4°C, proteinase K was added (200 μg/ml final concentration) and incubated for 30 min and 60 min at 4°C. Input control samples were kept with cold MTE buffer w/o detergents and proteinase K. Finally, 10 μg of insulin (Sigma) were added for protein co-precipitation, followed by TCA precipitation for 30 min at 4°C (final concentration −20% TCA). Following TCA precipitation samples were centrifuged at 21,000 × g for 20 min at 4°C, and the resulting pellets were dissolved in 30 μl of 1 M Tris-HCl pH 7.5. Samples were resolved by SDS-PAGE and analyzed by Western blotting. Anti-Fluc and anti-Myc antibodies were used to visualize the target proteins; outer membrane TOM70 antibody, inner membrane TIM23 antibody and matrix COX1 antibody were used to control for mitochondrial integrity and loading.
Mitochondrial mass measurement. For estimation of mitochondrial mass, Mitotracker™ DeepRed FM (M22426, Invitrogen) staining was used according to manufacturer instructions. Cells transfected with vectors pGal-Trans-eGFP, pER-FlucWT-eGFP or pER-FlucDM-eGFP were stained with Mitotracker™ DeepRed 40-48 h after transfection and analyzed by flow cytometry using the BD FACS Canto II and the FlowJo software. GFP signal was used to gate the transfected cells. Treatment of wild-type HEK293 cells with CCCP (10 µM, 12 h) was used as a positive control.
Mitochondrial membrane potential measurement. For measurement of mitochondrial membrane potential, far-red fluorescent indicator DilC1(5) (M34151, ThermoFisher) was used, following manufacturer instructions. In brief, cells transfected with vectors pGal-Trans-eGFP, pER-FlucWT-eGFP or pER-FlucDM-eGFP were stained with DilC1(5) 40-48 h after transfection and analyzed by flow cytometry using the BD FACS Canto II and the FlowJo software. GFP signal was used to gate the transfected cells. Treatment of wild-type HEK293 cells with CCCP (20 µM, 1 h) was used as a positive control.
Mitophagy assay. Mitophagy was assessed using co-transfection with vector pKeimaRed carrying mitochondria-targeted fluorescent protein Keima-Red (Amalgaam MBL AM-V0259M). The fluorescent protein Keima has a pHdependent excitation spectrum that changes depending on pH. A short excitation wavelength (440 nm) is predominant in a neutral environment, whereas a long excitation wavelength (586 nm) is predominant in an acidic environment. pH change upon fusion of mitophagosome with lysosome leads to corresponding shift of excitation spectrum. Mitophagy was monitored by FACS analysis using the BD FACS Fortessa and the FlowJo data analysis software. Keima-Red fluorescent signal was used to gate transfected cells. A violet laser 405 nm and filter set 610/20 was applied to measure the short excitation wavelength fluorescence (SE-fluo), a yellow-green laser 561 nm and the same filter set was applied to measure the long excitation wavelength fluorescence (LE-fluo). "Mitophagy index" proportional to the percentage of mitochondria undergoing mitophagy was calculated as a LE-fluo to SE-fluo+LE-fluo. Treatment with CCCP (10 µM, 16 h) was used as a positive control for mitophagy.
Caspase 3/7 activation assay. Caspase 3/7 activity was measured using the CellEvent™ Caspase-3/7 Green Flow Cytometry Assay Kit (ThermoFisher) according to manufacturer instructions. Non-transfected HEK WT cells and cells transfected with vectors pGal-Trans, pER-FlucWT or pER-FlucDM were cotransfected with pTag-BFP, stained with CellEvent™ Caspase-3/7 assay 40-48 h after transfection and analyzed by flow cytometry using the BD FACS Canto II and the FlowJo software. Tag-BFP signal was used to gate the transfected cells. Cellular ATP production. Intracellular levels of ATP were measured using colorimetric/fluorometric ATP assay kit (ab83355, Abcam). Cells transfected with vectors pGal-Trans-eGFP, pER-FlucWT-eGFP or pER-FlucDM-eGFP were harvested, suspended in specific ATP-assay buffer, lysed by shock-freezing and ATP levels determined according to manufacturer instructions using fluorimeter Cytation 5 (BioTek). The level of transfection monitored by FACS using GFP signal as a marker was around 80%. Treatment with CCCP (20 µM, 2 h) was used as a positive control. To assess mitochondrial OXPHOS-dependent ATP production transfected cells were rinsed with PBS and incubated in DMEM medium supplemented with 2.5 g/L galactose (w/o glucose and FBS) for 4-5 h and extensive aeration before harvesting.
Correlative light and electron microscopy. HEK293 cells transiently cotransfected with MTS-mCherry-GFP 1-10 and ER-FlucDM-HA-GFP 11  HEK293 cells transiently transfected with ATZ-myc vector were processed as described above, with following modifications. Fluorescent heads (PS-Speck, #P7220, ThermoFisher, 1:3 in ethanol absolute) were added as fiducial markers prior to collecting the ultrathin sections on wafers. Following washing, a 5 min preincubation with PBG (PBS with 0.5% bovine serum albumin and 0.2% gelatin type B) was followed by incubation with rabbit anti-myc-tag (Abcam ab9106, 1:100) in PBG for 1 h. After washes with PBG (6 ×1 min and 1 ×5 min), wafers were incubated with a secondary antibody anti-rabbit Alexa 568 (Molecular Probes A21069, 1:100) in PBG for 1 h. For SRRF (Super-resolution radial fluctuations) 500 images per channel were acquired at 15 ms exposure time. These images were further processed with the Plugin Nano-J-SRRF 48 within the open-source platform Fiji 47 . The following parameters were applied: Ring radius: 0.5; Radiality magnification: 7 and Axes in Ring: 7. For electron microscopy, the silicon wafers were processed as described above.
Yeast strains and plasmids. Plasmids containing split-GFP constructs and Elo3-mCherry were ordered from Addgene and the split-GFP constructs were adapted to the current study by subcloning into p403 and p406 vector backbones under control of the TDH3 promoter 15,19 . yER-FlucDM-GFP 11 was targeted to ER by cloning native ER protein KAR2 signal sequence (KAR2SS) upstream FlucDM-GFP 11 , HDEL was added downstream for ER retention. yMTS-mCherry-GFP 1-10 was targeted to mitochondria using mitochondrial targeting sequence of Su9 upstream to mCherry. Artificial ChiMERA tether 9 was modified for the study by replacing GFP with Myc, and cloned into pM methionine centromeric plasmid. For a list of plasmids used in this study see Supplementary Table 2.
Yeast strains used in this study are based on the BY4741 background as listed in Supplementary Table 3. Yeast cells were grown in synthetic complete or drop-out media containing 2% dextrose at 30°C.
All constructs were integrated into wild-type BY4741 genome by LiAc method and selected on corresponding synthetic drop-out agar plates, correct integration was confirmed by colony PCR. MSP1, TOM70, MDM10, MDM34, or MMM1 deletion mutants were created in yeast cells expressing yER-FlucDM-GFP 11 and yMTS-mCherry-GFP 1-10 by replacing the corresponding gene with a leucine cassette, and were grown on plates without Leucine. The Leucine cassette was PCR amplified with homologous overhangs corresponding to the UTR of gene to be deleted for targeted gene deletion. The deletions were confirmed by colony PCR, followed by sequencing. ΔMDM10 strains were complemented with Myc-tagged MDM10 point mutants or artificial ChiMERA tether cloned into methionine centromeric plasmid (pM) and selected on plates without methionine. The point mutants used were MDM10 Y73, 75 A mutant which shows impaired SAM50 function and MDM10 Y296, F298A mutant which exhibits impaired ERMES function 20 . The expression levels of yER-FlucDM-GFP 11 , yMTS-mCherry-GFP 1-10 and complementing mutant MDM10 or ChiMERA tether were corroborated using Western blot.
Yeast microscopy. Live-cell images of yeast were acquired using Leica SP8 confocal microscope with white laser line (WLL) and 63x oil-immersion NA = 1.4 objective. GFP and mCherry were excited at 488 nm and 581 nm respectively. The emission was collected onto HyD detectors in photon counting mode. The emission filter for GFP was set to 500-550 nm and for mCherry to 591-700 nm. Multiframe alternating excitation configuration was used to avoid GFP spillover into mCherry channel. Processing of images was done using ImageJ (NIH) software. For signal quantification, we used embedded ImageJ function Analyze Particles after we segmented our images based on mCherry-mitochondrial signal. GFP signal was calculated only within mitochondrial areas to avoid interference with possible nonspecific cytoplasmic signals. Median fluorescent signal was then averaged in between technical and biological repeats. For quantification of cells exhibiting co-localization, we manually counted the cells where both markers were expressed and collocate when we overlay the GFP and mCherry channels. The ratio of the cells displaying the co-localization divided by the total number of cells was averaged between biological repeats.
Statistics and reproducibility. The number of independent biological replicates (transfection events) used in every experiment is specified in the corresponding figure legend.
Statistical analysis was performed with Prism 5.0 or Excel software. For all assays, except quantification of wide-field fluorescent microscopy, unpaired Student's t-test was used to determine significant difference between analyzed samples. The significance of wide-field fluorescent microscopy quantification was calculated using Mann-Whitney test.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Plasmids generated in this study have been submitted to Addgene. Corresponding ID numbers are present in the Supplementary Tables 1-3