Tetracyclines activate mitoribosome quality control and reduce ER stress to promote cell survival

Mutations in mitochondrial genes result in increased sensitivity to nutrient deprivation, such that glucose deprivation can induce cell death. Specifically, cells cultured in equimolar concentrations of galactose in place of glucose become dependent on mitochondrial metabolism, resulting in cells harboring mitochondrial mutations being particularly sensitive to cell death (Soustek et al, 2018; Balsa et al, 2019, 2020; Perry et al, 2021). We previously showed that inhibition of mitochondrial translation by tetracyclines rescues mitochondrial disease models from cell death by a currently unknown mechanism (Perry et al, 2021). Here, utilizing mitochondrial disease complex I (A3796G ND1) mutant cybrid cells, we interrogate mitoribosome quality control pathways as potential mediators of tetracycline‐induced rescue under glucose deprivation.

Recent cryogenic electron microscopy studies in phosphodiesterase 12 (PDE12) mutant HEK293 cells indicate that mitoribosome elongational stalling induced by aberrant tRNA accommodation promotes mitoribosome splitting and recruitment of quality control factors (Desai et al, 2020). Doxycycline inhibits bacterial protein synthesis by binding to the bacterial SSU rRNA in the decoding region, stalling translation by disrupting the accommodation of tRNA (Wilson, 2014). This mechanism is similar to defective tRNA processing in PDE12 mutant cells (Desai et al, 2020). With this in mind, we sought to investigate mitoribosome elongational stalling and quality control factors as initial molecular signals in promoting doxycycline‐induced survival in ND1 mitochondrial mutant cells. ND1 mutations are commonly found in MD patients (Gorman et al, 2016), and we have previously used this cell model to validate genes and compounds, including tetracyclines, that rescue cell death under glucose‐deprivation conditions (Soustek et al, 2018; Balsa et al, 2019, 2020; Perry et al, 2021).

The mitochondrial translation cycle includes distinct stages of initiation, elongation, and termination, where conserved classes of proteins regulate progression between each phase (Kummer & Ban, 2021). To investigate if the general inhibition of mitochondrial translation was sufficient to promote survival in ND1 mutants, we compared the effects of mitoribosome‐targeting doxycycline with a distinct inhibitor, N‐formyl methionine mimetic actinonin, that stalls mitochondrial translation by inhibiting peptide deformylase (Richter et al, 2013, 2015) (PDF, Fig 1A), the enzyme responsible for the removal of N‐formyl groups on nascent peptides as they emerge from the ribosome exit tunnel prior to folding (Richter et al, 2013, 2015). It has also been suggested that actinonin inhibits mitochondrial translation indirectly through inhibition of the mitochondrial AAA‐protease (Richter et al, 2015), illustrating another non‐mitoribosome targeting mechanism. Nonetheless, the mechanism of inhibition of mitochondrial translation by actinonin is distinct to that of tetracyclines. Treatment with actinonin was unable to promote survival in ND1 cells at a wide range of concentrations (Fig 1A), suggesting that inhibition of mitochondrial translation is not necessarily sufficient for the promotion of MD cell survival during glucose starvation. We have also previously identified (Perry et al, 2021) that other mitoribosome‐targeting inhibitors pentamidine (Sun & Zhang, 2008) and retapamulin (Wilson, 2014; Meydan et al, 2019) promoted survival similar to doxycycline. Together, these results suggest that general suppression of mitochondrial protein synthesis is not sufficient to promote survival, rather, a signal must be initiated through direct inhibition of the mitoribosome. In an unbiased approach to identify factors involved in doxycycline rescue, we utilized the Dependency Map (DepMap) proteomics datasets, where repurposing screens have been performed to investigate the utility of current FDA approved small molecules, including doxycycline, for cancer treatment (Corsello et al, 2020) (Fig 1B). Here, positive correlation between the protein abundance and sensitivity to doxycycline identified proteins involved in cell death. Hence, we sought to identify proteins for which high expression was associated with reduced sensitivity to doxycycline, which we reasoned might be involved in promoting doxycycline‐mediated survival in models of MD (Perry et al, 2021). Gene‐ontology enrichment analysis identified high (negative) correlation with numerous mitochondrial protein subclasses including “mitochondrion” and “transit peptide,” and more specifically the mitoribosome including “translational elongation” and “large mitoribosome subunit” (Fig 1B). Interestingly, these studies identified survival correlation with mitoribosome quality control protein MALSU1, in line with potential elongational stalling and splitting mechanism described above. To note, MTRES1 and mtRF‐R did not correlate with doxycycline treatment. MRPL14, an accessory subunit of the mtLSU that MALSU1 binds during splitting (Brown et al, 2017; Desai et al, 2020), was also identified in this screen, further substantiating the importance of this binding module in surviving cancer cells and potentially in mitochondrial mutant cells in the context of doxycycline mechanisms of action (Fig 1B).

To interrogate the role of the mitoribosome quality control pathway in facilitating tetracycline‐induced survival in mitochondrial ND1 mutant cells, we utilized CRISPR Cas9 gene editing to knock‐out (KO) respective proteins involved in this pathway. Elongational stalling results first in splitting of the large and small subunit where MALSU1, along with LOR8F8 and NDUFAB1 (MALSU1 module), binds and stabilizes the large subunit. Rescue factors MTRES1 and mtRF‐R then bind and stabilize the nascent peptide and tRNA. We first knocked out MTRES1, mtRF‐R, and MALSU1 to identify which, if any, proteins may be involved in doxycycline survival (Fig 1C and D). These studies illustrated that doxycycline was still able to rescue ND1 cell death in the absence of MTRES1 or mtRF‐R (Fig 1C), but survival was significantly reduced in the absence of MALSU1, where the recruitment of the entire MALSU1 module, including NDUFAB1 and LOR8F8, is expected to be compromised (Fig 1D). It was also noted that basal survival rates of ND1 cells in galactose were drastically reduced in MTRES1 and mtRF‐R KO conditions (Fig 1C), indicating that these cells rely on mitoribosome quality control under nutrient stress, but doxycycline specifically requires the MALSU1 module to promote survival (Fig 1D). We further investigated whether this mechanism can be applied to the general tetracycline class of antibiotics by testing additional analogs, including the parental tetracycline, along with synthetic analogs minocycline and 7002, a compound we previously illustrated to inhibit mitochondrial translation and promote survival in mitochondrial mutants (Perry et al, 2021) (Fig 1E). These studies illustrate that tetracycline antibiotics universally depend on MALSU1 to promote survival in ND1 cybrid cells (Fig 1E). Ectopic overexpression of MALSU1 in complex I mutant cells independent of tetracyclines was unable to promote survival of ND1 mutant cells under glucose deprivation, indicating that MALSU1 is required but not sufficient to promote survival. It also suggests that the survival signal requires the tetracycline‐inhibited mitoribosome for MALSU1 to protect against cell death (Fig EV1A–C).

Having observed that MALSU1 is required to promote survival in mitochondrial mutant cells by tetracycline, we investigated the requirement for other mitoribosome quality control factors. Mitochondrial initiation factor 3 (MTIF3) has been previously implicated in binding the mitoribosome small subunit during biogenesis (Khawaja et al, 2020; Itoh et al, 2022) and during translation to access polycistronic transcripts (Remes et al, 2023). Specifically, MTIF3 sterically blocks the premature association of large‐ and small subunits (Khawaja et al, 2020; Itoh et al, 2022), analogous to the role of MALSU1 at the large subunit (Brown et al, 2017; Desai et al, 2020). Further, MTIF3 has ribosome dissociative activity (Luo et al, 2019), qualifying it as a potential mediator of mitoribosome splitting following tetracycline treatment and downstream cell survival promoting effects. CRISPR‐Cas9 depletion of MTIF3 in ND1 cells resulted in a reduced ability of doxycycline to promote cell survival (Fig 2A and B), implicating this protein in potentiating signaling at the tetracycline‐inhibited mitoribosome, albeit to a lesser extent than MALSU1 (Fig 2B).

To further define MALSU1 as a key player in survival, we dissected the role of the MALSU1 module by knocking‐out LOR8F8. LOR8F8, a LYRM‐chaperone, binds to and recruits NDUFAB1 to MALSU1, constituting the mature MALSU1 module capable of binding the split mitoribosome (Brown et al, 2017). Similar to MALSU1 and MTIF3, LOR8F8‐KO‐sensitized cells to glucose deprivation (Fig 2B). However, doxycycline was able to efficiently rescue LOR8F8 deficient cells to a much greater extent than MALSU1 and MTIF3 KO cells (Fig 2B). The ability of doxycycline to promote survival in the LOR8F8‐KO cells indicates that MALSU1, but not other MALSU1‐module proteins, is necessary to promote survival.

We next sought to investigate the regulation of MALSU1 at the mitochondria and mitoribosome in response to tetracyclines. To study the recruitment of different quality control factors to the purified mitoribosome, we utilized HEK293 suspension cells (Expi293F), a model previously used to study mitoribosome elongational stalling and quality control (Desai et al, 2020). ND1 mutant cybrid cells grown in monolayer culture did not provide sufficient mitochondria to purify appreciable quantities of the mitoribosome, so Expi293F suspension cells were used as an alternate system to study changes in the molecular composition and architecture of the doxycycline‐inhibited mitoribosome. To first understand the kinetics of MALSU1 recruitment, we performed a time‐course experiment, isolating mitochondria from Expi293F cells at a series of different time points following doxycycline treatment (Fig 2C). Here, we found that levels of MALSU1 within mitochondria were highest at 12 h following doxycycline treatment (Fig 2C). Using these timepoints as a guide, we isolated mitoribosomes from Expi293F cells at early (1 h) and later (12 h) time points to assess MALSU1‐mitoribosome recruitment. We found that MALSU1 levels were detectable at the mitoribosome after 1 h, with much larger increases observed after 12 h of doxycycline treatment (Fig 2D). We also evaluated the regulation of MTIF3 in these experiments, observing that doxycycline did not significantly alter the amount of this protein present at the mitochondria or mitoribosome (Fig 2D), highlighting MALSU1 as a key mediator of signaling in response to elongational stalling. The necessity of MTIF3 for doxycycline‐induced survival, however, implicates this protein in potentiating the survival signal initiated by tetracycline treatment. Similar to Expi293F cells, the abundance of MALSU1 in mitochondrial fractions of ND1 mutant cells increased following doxycycline treatment, with accumulation starting at 1 h and maximizing from 3 to 24 h (Fig 2E). We also observed a notable increase in mitochondrial accumulation of MTIF3, along with mitoribosome proteins MRPL4 and MRPS16, potentially reflecting a stimulation of mitochondrial quality control or biogenesis in response to doxycycline treatment in the context of mitochondrial mutations and glucose starvation (Fig 2E). Interestingly, when compared to matched glucose controls, we observed that MALSU1 and MRPs are decreased upon glucose starvation and are stabilized by doxycycline, most notably at the 24‐h time point (Fig 2F). We observed concomitant increases in MALSU1 protein in the whole cell extract at 24 h without substantial changes in transcript levels (Figs 2F and EV1D–G), indicating likely increased mitochondrial stability of MALSU1 at the tetracycline‐inhibited mitoribosome. These results highlight a potential stabilization of MALSU1 at the mitochondria in response to tetracyclines at early timepoints.

To investigate the mechanisms whereby MALSU1 is required for ND1 mutant cell survival under glucose starvation, we sought to further investigate doxycycline‐dependent regulation of MALSU1 recruitment to the mitoribosome. Rapid isolation of the mitoribosome, initially from Expi293F cells, was achieved through successive sucrose gradient centrifugations to obtain mitochondria, which were then solubilized in dodecyl‐β‐D‐maltoside (β‐DDM) and cardiolipin. This efficiently extracted mitoribosomes from the inner mitochondrial membrane for fractionation of the mtSSU, mtLSU, and monosomes (Desai et al, 2020).

Expi293F cells were treated with doxycycline for 48 h to induce mitoribosome stalling, after which mitoribosomes were isolated for analysis. Doxycycline‐treated mitochondria, and more drastically the mitoribosomes, exhibited increased MALSU1 abundance (Fig 3A). These results are consistent with recruitment of MALSU1 to the mitoribosome at earlier timepoints (Fig 2D) and may reflect increases in its mitochondrial import and/or stability in response to doxycycline. To investigate whether MALSU1 was enriched at the large subunit in doxycycline‐treated cultures, we sucrose‐gradient‐fractionated mitoribosomes to separate and analyze the location of MALSU1. Here, we found that MALSU1 was enriched at the mtLSU upon doxycycline treatment (Fig 3B). Additionally, we observed a shift in the sedimentation of the MALSU1‐bound mtLSU with doxycycline treatment, indicating a potential change in the size and/or molecular shape of the large subunit in this condition when compared to controls (Fig 3B). To validate these effects in ND1 cybrid cells, crude mitochondria were isolated, solubilized with β‐DDM and cardiolipin, and fractionated by sucrose gradients in a similar fashion as described above. Mitochondria from CRISPR‐Cas9 MALSU1‐depleted ND1 cells were analyzed in parallel to investigate the necessity of MALSU1 in shifting the sedimentation pattern of the doxycycline treated large subunit. As observed in Expi293F suspension cells, doxycycline treatment in ND1 cells resulted in an increase in mitochondrial MALSU1 (Fig 3C) that was enriched at the large subunit (Fig 3D and E). The shift in large subunit sedimentation was again observed with doxycycline treatment (Fig 3D), mimicking Expi293F cultures (Fig 3B). The sedimentation pattern of the large mitoribosome subunit was still altered in MALSU1‐KO cells (Fig 3F), indicating that doxycycline may split the large subunit independent of MALSU1, although MALSU1 binding to the large subunit is required for survival (Fig 1D and E).

To further investigate the subcellular localization of MALSU1 in response to doxycycline, we fractionated ND1 (Fig 3G) and Expi293F (Fig 3H) cells to identify if MALSU1 is enriched in specific compartments in response to mitoribosome stalling. These experiments showed that MALSU1 is specifically (Fig 3G and H) and dose‐dependently (Fig 3I) enriched at the mitochondria of doxycycline‐treated cells, without changes in mitochondrial abundance of the matrix protein LRPPRC or the outer‐mitochondrial membrane protein TOM70. It has been previously reported that MALSU1 has a short half‐life (Morgenstern et al, 2021) and that its expression is lost in rho0 cells (Wanschers et al, 2012), highlighting a tightly regulated and mitochondria‐specific localization. Increased whole cell levels of MALSU1 upon doxycycline treatment at earlier timepoints (e.g., ND1 24 h and Expi293F 1 and 12 h, Figs 2D–F and EV1G) may reflect an accumulation of MALSU1 at the mitochondria due to its stabilization at the mitoribosome. MALSU1 expression is apparently unchanged at the whole cell level when compared to increases in mitochondrial fractions at later time points (48 h). In these conditions, this may suggest remodeling of the mitochondrial proteome in doxycycline treated cells, that when normalized to mitochondrial protein content reflects higher MALSU1 levels. This is consistent with decreases in mitochondrial markers TOM70 and cytochrome C (Fig 3G and H). These studies support the notion that MALSU1 binding to the mitoribosome large subunit drives tetracycline‐induced survival responses and shows a direct association between the extent of mitoribosome inhibition and MALSU1 recruitment.

Taken together, mitoribosome stalling caused by doxycycline results in mitoribosome splitting where doxycycline acts as a “splitting factor” and MALSU1 binds and stabilizes the large subunit to provide a pro‐survival signal. Based on its partial requirement in doxycycline‐induced cell survival (Fig 2A and B) and its reported ribosome dissociating activity (Luo et al, 2019), it is possible that MTIF3 cooperates in tetracycline‐induced ribosome splitting to promote MALSU1 binding.

We have previously reported that pharmacological inhibition of ER stress sensor/transducer IRE1α promotes survival in mitochondrial complex I disease cells (Soustek et al, 2018), implicating ER stress in promoting death of cells with defective mitochondrial respiration. To determine whether doxycycline‐induced survival involves modulation of this pathway, we tested if doxycycline treatment reversed the activation of X‐box binding protein 1 short fragment (XBP1s), an ER‐stress‐induced transcription factor produced through translation of an alternatively spliced mRNA generated through the endoribonuclease activity of IRE1α (Huang et al, 2019; Carreras‐Sureda et al, 2022). A potent transcription factor, XBP1s, regulates the expression of a wide variety of ER‐homeostatic genes. Western blot analysis indicated that XBP1s is strongly and specifically activated in ND1 cells, but not WT cells, after 48 and 72 h under glucose restriction conditions (Fig 4A). Doxycycline treatment efficiently reduced XBP1s activation at 48 h and completely eliminated it at 72 h, consistent with improved cell survival and reduced PARP1 cleavage (Fig 4A). Interestingly, CMT3, a tetracycline analog that does not exhibit mitoribosome translation inhibition or rescue mitochondrial mutants (Perry et al, 2021), did not reverse the activation of XBP1s (Fig 4A). These results indicate that inhibition of mitochondrial translation by doxycycline results in the pro‐survival resolution of ER stress that is exacerbated in complex I mutants.

Unfolded proteins in the ER bind to IRE1α on its luminal domain, promoting dimer/oligomerization and activation of the cytosolic kinase and RNAase domains, transducing ER stress (Huang et al, 2019). To further investigate doxycycline effects on IRE1α activation, we analyzed its oligomerization state using blue‐native polyacrylamide gel electrophoresis (BN‐PAGE, Fig 4B) as previously described (Sundaram et al, 2018). In whole cell extracts, we did not observe substantial changes in the oligomerization state with doxycycline treatment (Fig EV2A), inconsistent with results on XBP1s activation (Fig 4A). It has been previously reported that IRE1α can be enriched at mitochondrial ER contact sites (Takeda et al, 2019), a potential subcellular location wherein the oligomerization state is regulated. Consistent with this, we found IRE1α to be enriched in crude mitochondrial (Mito/ER) samples (Fig 3G and H). Hence, due to IRE1α enrichment with mitochondria, along with doxycycline mode of action at the mitoribosome, we isolated crude mitochondria containing ER fractions from ND1 cells and performed BN‐PAGE to analyze the IRE1α oligomerization status in the ER concentrated at the mitochondria. Here, we observed a marked dose‐dependent decrease in oligomerization state of IRE1α with doxycycline treatment without changes in total IRE1α levels or associated mitochondrial markers (Fig 4B). These results indicate that ND1 cell‐survival decisions through IRE1α are to some extent controlled at the mitochondria‐associated ER, and that a signal may be initiated at the mitochondrial ribosome upon elongational stalling with doxycycline treatment. These effects were also observed for additional markers of ER stress, including the ER lumen UPR chaperone BiP and downstream p38 MAPK phosphorylation, which were both reduced in a similar fashion to XBP1s (Fig 4B). In fact, we have previously illustrated that pharmacological inhibition of p38 (SB203580) rescues survival in ND1 mutant cells concomitant with metabolic changes similar to that of doxycycline treatment implicating MAPK signaling in the survival mechanism (Perry et al, 2021). Treatment of ND1 cells with IRE1α inhibitor 4μ8C reverses the galactose‐induced phosphorylation of p38 (Fig EV2B) similar to doxycycline or chemically distinct and active pentacycline‐based mitoribosome inhibitor 7002, but not the inactive CMT3 (Fig EV2C). This illustrates that MAPK activation is downstream of IRE1α and can be controlled by mitoribosome inhibition under these conditions.

MALSU1, over other mitoribosome quality control proteins, is required for doxycycline‐induced survival (Fig 1D and E). To investigate if MALSU1 is required for doxycycline‐induced inactivation of IRE1α, we compared IRE1α oligomerization and XBP1s activation between doxycycline treated control and MALSU1‐KO mutant cells. In control and MALSU1 KO cells, we observe an increase in IRE1α oligomerization at mitochondrial contacts, along with XBP1s and BiP activation with galactose treatment (Fig 4C). However, unlike in control cells, IRE1α oligomerization, XBP1s cleavage, and BiP activation are not repressed in MALSU1 KO cells with doxycycline treatment, consistent with inability of doxycycline to inhibit phosphorylation of p38 MAPK (Fig EV2D) and promote survival of MALSU1 KO cells (Fig 4C). These studies show that MALSU1 must be present for doxycycline to modulate IRE1α‐mediated UPR activity in ND1 mitochondrial mutant cells.

NDUFS4^−/− mice are a widely employed preclinical model of mitochondrial disease, largely characterized by neuromuscular decline (Kruse et al, 2008). Previously, we used this model to investigate the utility of doxycycline in vivo, where we found doxycycline supplemented mice exhibited improved survival and fitness when compared to controls (Perry et al, 2021). Specifically, doxycycline delayed the development of neurological defects in NDUFS4 KO brains. Thus, we investigated the status of IRE1α in the brains of NDUFS4 KO mice (Fig 4D). We found that NDUFS4 KO brains exhibited marked increased expression of IRE1α protein levels when compared to their control and doxycycline‐treated counterparts (Fig 4D). These results indicate that chronic treatment of doxycycline may attenuate the UPR in brains with mitochondrial dysfunction, suppressing IRE1α. Taken together, doxycycline‐induced survival signaling at elongating mitoribosmes requires MALSU1 and initiates cross‐talk between mitoribosome quality control and ER‐stress effectors, highlighting an important survival mechanism in mitochondrial disease mutants under nutrient stress.

Canonical regulation of IRE1α involves the interplay between inhibitory binding of BiP to monomeric IRE1α on the luminal domain and the activating recruitment of BiP to unfolded proteins in the ER lumen, releasing BiP from IRE1α to chaperone and maintain protein homeostasis (Huang et al, 2019). This release allows for dimerization of IRE1α, transducing initial pro‐survival UPR. If ER proteostasis is unable to be re‐established, unfolded proteins bind to the IRE1α luminal domain directly, causing oligomerization activating cell death responses (Huang et al, 2019). MARCH5, a mitochondrial outer membrane ubiquitin ligase, has recently been implicated in attenuating the oligomerization of IRE1α under prolonged ER stress (Takeda et al, 2019). With marked decreases in IRE1α oligomerization upon mitochondrial translation inhibition with doxycycline (Fig 4B and C), we sought to investigate if this regulation was mediated through MARCH5. CRISPR Cas9 KO of MARCH5‐sensitized ND1 cells to galactose‐induced ER stress with heightened levels of XBP1s and P‐p38, but this response could still be reversed by doxycycline treatment (Fig EV3A). Importantly, doxycycline still rescued MARCH5‐depleted ND1 cells from galactose‐induced cell death (Fig EV3B), suggesting that MARCH5 is not involved in the cross‐talk between the doxycycline inhibited mitochondrial ribosome and IRE1α activation at the ER.

Doxycycline treatment results in marked MALSU1‐dependent decreases in ER stress as evidenced by reduced IRE1α oligomerization and activation and downstream reductions in XBP1s, BiP, and phosphorylation of p38 MAPK (Figs 4A–C and EV2D). To further investigate tetracyclines effects on the UPR, we analyzed the activation of additional UPR sensors PERK and ATF6. Similar to IRE1α, PERK, and ATF6 are ER membrane proteins with luminal sensing and cytosolic signal‐transducing domains (Sundaram et al, 2018; Wiseman et al, 2022). Activation occurs following the accumulation of unfolded proteins in the lumen that out‐compete protein folding chaperones (Sundaram et al, 2018; Wiseman et al, 2022). ATF6 is proteolytically processed to translocate first to the golgi, and then to the nucleus where the cleaved product acts as a transcription factor, promoting the expression of ER‐stress responsive genes (e.g., BiP/HSPA5) (Wiseman et al, 2022). PERK is activated upon oligomerization and trans‐autophosphorylation, once active it phosphorylates eif2α, resulting in the specific translation of the transcription factor ATF4 and the subsequent induction of UPR genes (e.g., DDIT3, PPP1R15A) (Madhavan et al, 2022). Hence, we investigated the activity of all three UPR sensors upon mitochondrial mutation and glucose deprivation, how this activity is regulated by doxycycline, and whether the effects are MALSU1‐dependent (Fig 4E). Here, we observe that all three UPR sensors are strongly activated upon glucose deprivation in ND1 cells, as evidenced by increased IRE1α phosphorylation, ATF6 cleavage, and PERK phosphorylation, all of which could be suppressed by doxycycline (Fig 4E). Consistent with cell survival (Fig 1D and E) and IRE1α oligomerization (Fig 4C) data, MALSU1 expression was required for the doxycycline‐dependent resolution of UPR activation for all arms of the UPR, and these results can be expanded to downstream UPR effector XBP1s, phosphorylation of p38 MAPK, and luminal chaperone BiP (Fig 4E). Activation of each respective arm of the UPR can be linked to the expression of specific ER‐stress‐responsive genes (Wiseman et al, 2022) and can be monitored using qPCR. Consistent with the observed activation of IRE1α, ATF6, and PERK on the protein level, transcripts linked to each specific arm were found to be activated upon nutrient stress, and doxycycline attenuated this activation (Fig 4F). These results show that glucose deprivation results in activation of all arms of the ER UPR, and that doxycycline resolves this global ER stress in a manner dependent on the mitoribosome quality control protein MALSU1.

Intrigued by the ability of tetracyclines to suppress ER stress in mitochondrial mutants, we evaluated the contribution of different ER protein homeostasis pathways. ER stress and homeostasis are largely regulated by (i) ER‐associated protein degradation (ERAD), (ii) ER protein secretion, and (iii) ER protein translocation or insertion (Wiseman et al, 2022). To test if tetracyclines are resolving ER stress by activating ERAD, we utilized specific inhibitor of p97‐induced protein degradation eeyarestatin I (ESI, Fig 5A–C) (Wang et al, 2008, 2010). It was noted in these experiments that mitochondrial mutant cells were more dependent on ERAD for survival under glucose deprivation when compared to WT, as the potency of ESI was increased under galactose conditions in ND1 mutant cells (Fig 5A) to a greater extent than in WT cells (Fig 5B). Galactose‐induced phosphorylation of p38 MAPK and XBP1 splicing were increased in ND1 when compared to WT cells, exacerbated by ESI, and suppressed by IRE1α inhibition, illustrating that ERAD modulates MAPK and XBP1s UPR signaling (Fig EV2B). However, we observed that treatment with ESI did not reverse the ability of doxycycline to promote survival in mitochondrial mutant cells under glucose deprivation (Fig 5C), suggesting that doxycycline does not promote survival by enhancing ERAD.

To assess the effects of tetracyclines on ER‐protein secretion, we utilized the secretion inhibitor brefeldin A (BFA), which interferes with ER‐Golgi protein transport (Fujiwara et al, 1988; Orci et al, 1991). It was noted that WT and ND1 cells are less sensitive to BFA under galactose conditions, illustrating a potential decrease in secretory pathway when glucose is limiting (Fig 5D and E). Nonetheless, ND1 cells treated with BFA at IC₅₀ value concentrations can still be rescued by doxycycline (Fig 5F). Consistent with these results, doxycycline can reverse the activation of the UPR in ND1 cells treated with ESI or BFA as evidenced by XBP1s and BiP protein levels, as well as IRE1α and p38 phosphorylation status (Fig 5G). Importantly, galactose‐induced XBP1s and basal expression levels of BiP are increased in ESI and BFA treated cells cultured in galactose, consistent with inhibition of ERAD and secretion, respectively (Fig 5G).

To assess the effects of tetracyclines on ER protein translocation, we probed tetracycline‐treated mitochondrial mutant cells and isolated microsomes (ER) for the accumulation of cathepsin‐C, a glycosylated and secreted protein that can be used as a marker for translocation in ER fractions (Ruiz‐Canada et al, 2009). Here, we found that microsomes derived from ND1 cells exhibit a substantial increase in cathepsin‐C when compared to those derived from WT cells, and doxycycline suppressed this accumulation (Fig 5H and I). Expression or stability of cathepsin‐C was not altered in whole cell extracts as indicated by cycloheximide chase experiments, and expression levels were comparable between WT and ND1 cells (Fig 5H and I) suggesting that ER protein loading is specifically increased upon mitochondrial mutation, consistent with increased BiP in the ER from ND1 cells that is reduced by tetracyclines (Fig 5I). Due to the necessity of MALSU1 in promoting survival of doxycycline‐treated cells, we investigated the effect of doxycycline on ER protein loading in mitochondrial mutant cells lacking MALSU1. We found that tetracyclines were unable to modulate cathepsin‐C microsome levels in MALSU1‐deficient cells (Fig 5J), further implicating this mitoribosome quality control response in the regulation of ER‐protein loading induced UPR. Interestingly, we noted a decrease in numerous ER markers in MALSU1‐depleted microsomes including IRE1α, BiP, cathepsin‐C, and calnexin, with a relative increase in SRP14 when compared to microsomes from control cells (Fig 5J). Altered microsome expression levels of ER proteins may potentially indicate a maladaptive response upon MALSU1 depletion that results in aberrant adaptive UPR function, in line with heightened p38 MAPK phosphorylation in MALSU1‐deficient cells (Figs 4E and EV2D). Nonetheless, in the absence of MALSU1, basal microsome levels of cathepsin‐C which are insensitive to doxycycline treatment are associated with increased sensitivity to glucose starvation (Fig 1D and E).

Endoplasmic reticulum‐protein translocation is initially regulated by the signal recognition particle (SRP), which binds to the ribosome nascent chain complex to promote co‐translational translocation into the ER lumen (Akopian et al, 2013; Voorhees & Hegde, 2015). We observed a marked decrease in SRP subunit SRP14 in doxycycline‐treated microsomes from ND1 cells (Fig 5I) that was not observed in the absence of MALSU1 (Fig 5J) further supporting a decrease in ER‐protein translocation in doxycycline‐treated cultures. Additionally, SRP14 is relatively more abundant in the microsomes of MALSU1‐depleted cells compared to NTC cells when normalized to other ER markers, which may suggest a maladaptive increase in associated ribosomes. Interestingly, doxycycline's ability to reduce ER stress is independent of global translation rates, evidenced by unchanged puromycin incorporation in galactose treated ND1 cells (Fig 6A). Increased protein insertion into the ER of mitochondrial mutant cells may thus explain the increased dependence of cells with ND1 mutation on ERAD (Fig 5A) and the enhanced sensitivity to ER stress as presented in Fig 4A. Further, we examined known mitochondrial stress response factor DELE1 in mediating doxycycline‐induced survival. Here, we show that ATF4 is not activated in cells lacking DELE1 when treated with doxycycline or other mitochondrial toxins including oligomycin (Fig 6B), and doxycycline can rescue cells independent of DELE1 and at concentrations that do not induce ATF4 (1 μM, Fig 6C). This data further supports the prospect of a novel signaling mechanism originating at the mitoribosome that communicates to the ER to promote survival. Thus, we present a model whereby inhibition of mitoribosome elongation with tetracyclines activates mitoribosome quality control and resolves ER stress and IRE1α signaling, potentially through modulation of the rate at which newly synthesized proteins are inserted into the ER (Fig 6D).

ND1 and WT cybrid cells previously utilized by us (Perry et al, 2021) were generated by R. Vogel and J. Smeitink (Radboud University Medical Centre) using patient fibroblast mitochondria carrying either WT or the A3796G (ND1) mutation and were routinely validated for identity and contamination. Cybrid cells were cultured at 37°C and 5% CO₂ in DMEM high glucose (25 mM, Gibco) medium supplemented with sodium pyruvate (1 mM), fetal bovine serum (10%), and penicillin–streptomycin (100 U/ml). For glucose deprivation experiments, DMEM no‐glucose (Gibco, 11966‐025) was supplemented with equimolar (25 mM) galactose, sodium pyruvate (1 mM), fetal bovine serum (10%), and penicillin–streptomycin (100 U/ml). Expi293F cells (Gibco) were cultured in suspension in Expi293 expression medium (A14351‐01) without antibiotics using the manufactures protocol. Cells were maintained at a density of 0.5–5 × 10⁶ cells/ml, and cell viability (trypan blue) was assessed upon subculture to maintain viability > 95%.

Reagents used in this study include: Doxycycline hyclate (Sigma), tetracycline (Santa Cruz Biotechnology), minocycline (Santa Cruz Biotechnology), 7002, CMT‐3 (MedChem Express), actinonin (Sigma), eeyarestatin I (Sigma), 4μ8C (Sigma), n‐dodecyl β‐D‐maltoside (β‐DDM, Anatrace), cardiolipin (18:1, Avanti Polar Lipids), and β,γ‐methyleneguanosine 5′‐triphosphate sodium salt (GMPPCP, Sigma).

Antibodies used in this study include: C12orf65 (mtRF‐R, abcam, ab122448, 1:1,000), C6orf203 (MTRES1, abcam, ab151066, 1:1,000), C7orf30 (MALSU1, Proteintech, 22838‐1‐AP, 1:1,000), MRPL4 (Proteintech, 27484‐1‐AP, 1:1,000), MRPS16 (Proteintech, 16735‐1‐AP, 1:1,000), LRPPRC (Proteintech, 21175‐1‐AP, 1:1,000), TOM70 (Santa Cruz, sc390545, 1:1,000), IRE1α (Cell Signaling, 14C10, 1:1,000), β‐tubulin, Histone‐H3 (abcam, ab52946, 1:1,000), VDAC (Santa Cruz, sc390996, 1:1,000), XBP1s (Cell Signaling, D2C1F, 1:1,000), PARP (Cell Signaling, 46D11, 1:1,000), BiP (Cell Signaling, C50B12, 1:1,000), P‐p38 (T180/Y182, Cell Signaling, D3F9, 1:1,000), p38 (total, Cell Signaling, 1:1,000), pan – OXPHOS (ATP5A1, UQCRC2, MTCO1, SDHB, NDUFB8, abcam, STN‐19467, 1:500), Cathepsin‐C (R&D Systems, AF1071, 1:1,000), SRP14 (Proteintech, 11528‐1‐AP, 1:1,000), Calnexin (Cell Signaling, C5C9, 1:1,000), MARCH5 (abcam, ab185054, 1:1,000), HRI (EIF2AK1, abcam, 1:1,000), p‐eif2α (S51, Cell Signaling, 119A11, 1:1,000), eif2α (total, Cell Signaling, 1:1,000), ATF4 (Cell Signaling, D4B8, 1:1,000), MTIF3 (Proteintech, 14219‐1‐AP, 1:1,000), PERK (Cell Signaling, D11A8, 1:1,000), ATF6 (Cell Signaling, D4ZBV, 1:1,000), and Puromycin (Kerafast, 3RH11, 1:1,000).

Survival assays of mitochondrial mutant ND1 cybrid cells were carried out as previously described by us (Perry et al, 2021). Cells were seeded at 1.0 × 10⁵ cells/well in 6‐well plates in glucose‐deprivation media. Doxycycline, actinonin, or eeyarestatin I (ESI) were added at indicated concentrations at the time of seeding. Cells were then incubated for 96 h at which time the surviving cells were counted via trypan‐blue exclusion method. ESI dose–response curves were generated using SRB by seeding ND1 or WT cells with or without glucose at varying concentrations. Cells were then incubated for 48 h after which media was aspirated and analyzed for cell survival via SRB as indicated above. Dose–response curves were generated, and IC₅₀ values were calculated using GraphPad prism software and illustrate the concentration at which 50% of cells survive when compared to nontreated control cells in each respective media condition (glucose or galactose).

CRISPR‐Cas9 gene editing experiments were performed using the GeCKO (Shalem et al, 2014) protocol. sgRNA guides were cloned into lentiCRISPR‐V2 plasmids containing the puromycin resistance vector (Addgene, 98290), transformed into Stbl3 competent E. coli cells, and purified using standard methods. HEK293T cells were then seeded in 6‐well dishes (6 × 10⁵ cells/well) and reverse transfected using Polyfect (Qiagen, 301105) following the manufacturers protocol. Transfection was carried out using 880 ng of lentiCRISPR‐V2 with respective target‐ or nontargeting control sgRNA, 600 ng psPAX2 (Addgene, 12260), and 300 ng pMD2 (Addgene, 12259). Following 24 h transfection, media was replaced with DMEM and virus was generated for an additional 48 h. Lentiviral medium was then harvested, filtered (0.45 μm), and added with polybrene (conc, Cat. No.) onto ND1 cells that had been seeded and adhered to 6 well dishes (3 × 10⁵ cells/well) overnight. Twenty‐four hours after transduction, cells were split into 10 cm dishes and puromycin (0.5 μg/ml) was added and cells were selected for 7–10 days before used for experiments. The following sgRNA sequences were used: mtRF‐R: 5′‐ACCTTTACAACGATGCCTGA‐3′. MTRES1: 5′‐ATGATGTTGTCCTGAAGACG‐3′. MALSU1: sg1 5′‐CAGGCCGCGGACAAAGTTTG‐3′; sg2 5′‐CCCACTAATGTGGCGCAGGG‐3′; sg3 5′‐GCGGAGGGGACGGTCAACGA‐3′. DELE1: 5′‐TGATAATAAAGGACCGCCTG‐3′. MARCH5: 5′‐CCAGGCCTGTCTACAACGCT‐3′. LOR8F8: 5′‐TACCCGCTTAACTGCCAGCG‐3′. MTIF3: sg1 5′‐GCAGAGTATCAGCTCATGAC‐3′ sg2 5′‐GCAATAGGGGACAACTGTGC‐3′ sg3 5′‐AGGCGAACCCCAAAACTGGT‐3′.

Proteomic correlation of doxycycline was scored and ranked by P‐value (DepMap, PRISM repurposing screen; Corsello et al, 2020), where proteins enriched in cells surviving under doxycycline treatment were identified (P < 0.0001). Within this statistically significant subset of proteins, positive correlates were further analyzed for gene‐ontology (GO) enrichment against the entire proteome using DAVID functional annotation bioinformatics (https://david.ncifcrf.gov/) (Huang da et al, 2009). GO‐terms and their respective calculated fold enrichment were ranked and plotted. High fold enrichment represented a high number of proteins within a given GO‐term that were enriched in doxycycline‐treated cells.

Subcellular fractionations were carried out using differential centrifugation as previously described (Wieckowski et al, 2009). ND1 or WT cybrid cells were seeded in 2 × 15 cm dishes in glucose media and allowed to reach 80% confluence. Cells were then rinsed with PBS, and galactose media was then added with doxycycline at indicated concentrations and were incubated for 48 h. Cells were then harvested by scraping and pelleted at 300 × g for 5 min at 4°C. Cells were resuspended in PBS, pelleted, and then resuspended in 5 ml of cold isolation buffer (IB, 225 mM mannitol, 75 mM sucrose, 0.1 mM EGTA, 30 mM HEPES‐HCl pH 7.4). Cells were than dounce‐homogenized (~60 up‐down strokes) on ice and centrifuged at 1,000 × g at 4°C for 5 min. Supernatant containing mitochondria, ER (microsomes), lysosomes, and cytosol was then decanted into an additional 5 ml of IB and centrifuged at 10,000 × g at 4°C for 10 min to obtain crude mitochondria (mitochondria/ER) in the pellet. Supernatant containing cytosol, microsomes, and lysosomes was decanted and kept on ice, and mitochondria/ER fractions were resuspended in mitochondria resuspension buffer (MRB, 250 mM mannitol, 5 mM HEPES pH 7.4, and 0.5 mM EGTA). MRB was centrifuged at 10,000 × g at 4°C for 10 min, supernatant was aspirated, and crude mitochondria were snap frozen in liquid nitrogen and stored at −80°C. Lysosomes were then separated from microsomes and cytosol by centrifugation (20,000 × g at 4°C for 30 min), pelleted, and snap frozen. Microsomes in the supernatant were isolated by ultracentrifugation (100,000 × g for 60 min at 4°C) and resuspended in IB with protease inhibitors and snap frozen in liquid nitrogen and kept at −80°C. An aliquot of the resulting final supernatant containing cytosol was also snap frozen in liquid nitrogen and kept at −80°C.

SDS‐ and Native‐PAGE were carried out using standard protocols. For SDS‐PAGE, cells were resuspended in RIPA buffer (Cell Signaling, 9806) with protease and phosphatase inhibitors at 4°C, and protein content was determined using standard BCA colorimetric protocols (Pierce 23228). Microsomes in IB and crude mitochondria in MRB (from Subcellular fractionations) were directly quantified and solubilized with Laemmli SDS‐sample buffer (Bio‐Rad, 1610747). Ten to twenty microgram of protein from respective samples were run on 4–12% SDS‐PAGE gels (Invitrogen, NP0336BOX). Proteins were then transferred to PVDF membranes and probed with specific antibodies. Native‐PAGE analysis was carried out as previously described (Sundaram et al, 2018) with some modifications. Mitochondria/ER (100 μg) were solubilized in 50 μl of Native Page Buffer (Invitrogen, 2360376) supplemented with 1.5% digitonin and protease inhibitors. Cells were then incubated at 4°C for 10 min to complete lysis, and centrifuged at 18,500 × g for 20 min (4°C) to remove insoluble material. Supernatant was then carefully removed and resuspended in 0.5% G‐250 (Invitrogen, BN20041). Samples (10 μg) were then added to each well, and protein complexes were separated on 3–12% NativePAGE gels (Thermo Fisher Scientific, BN1003BOX). Proteins were transferred to PVDF membranes and probed with specific antibodies. Buffers used for NativePAGE include: NativePAGE Running Buffer (20×; Thermo Fisher Scientific, BN2001) and NativePAGE Cathode Buffer Additive (20×; Thermo Fisher Scientific, BN2002).

Mitoribosomes were isolated following a procedure similar to previously described (Desai et al, 2020). Expi293F cells (750 ml, 6 × 10⁶ cells/ml) treated with doxycycline (1 μM) or vehicle (DMSO) for 48 h were harvested by centrifugation (300 × g, 5 min, 4°C). Cells were resuspended in cold PBS and repelleted. Cells were then resuspended in mitochondrial isolation buffer (MIB, 50 mM HEPES‐KOH pH 7.5, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, and protease inhibitors) at a ratio of 6 ml MIB per 1 g of pellet. Cells were left to swell at 4°C for 10 min, after which were brought to isotonic concentrations of sucrose and mannitol by adding calculated volumes of 4× sucrose mannitol buffer (SM4, 280 mM sucrose, 840 mM mannitol, 50 mM HEPES‐KOH pH 7.5, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, and protease inhibitors). The resulting solution MIB + SM4 (MIBSM) will be used in following steps. Cells were then dounce homogenized (~60 up‐ down‐passes) and cellular debris and nuclei were centrifuged (800 × g, 4°C, 15 min). Supernatant‐1 was filtered and pellet was resuspended in fresh MIBSM for 15 passes. Resulting solution was again centrifuged and supernatant‐2 was combined through a cheesecloth with supernatant‐1. This solution was clarified by centrifugation a (1,000 × g at 4°C for 15 min) and resulting supernatant was collected. Crude mitochondrial pellet was then obtained by centrifugation (10,000 × g at 4°C for 15 min) and was resuspended in MIBSM buffer with DNAase (0.1 U/ml) and left rocking at 4°C for 30 min. Mitochondria were pelleted (10,000 × g at 4°C for 15 min), resuspended in 1 ml of SEM buffer (250 mM sucrose, 20 mM HEPES‐KOH pH 7.5, 1 mM EDTA), and loaded onto a discontinuous sucrose gradient (15–60%) constructed as previously described (Desai et al, 2020). Mitochondria were obtained, snap frozen in liquid nitrogen, and kept at −80°C.

To obtain mitoribosomes, mitochondria were thawed at 4°C and combined with equal parts of lysis buffer (25 mM HEPES‐KOH pH 7.4, 150 mM KCl, 50 mM MgOAc, 1.5% β‐DDM, 0.15 mg cardiolipin, 0.5 mM GMPPCP, 2 mM DTT, and protease inhibitors). GMPPCP was utilized to prevent dissociation of GTPases during the mitoribosome isolation and not included during live cell culture and doxycycline inhibition experiments. Combined solution was mixed and gently homogenized with a dounce homogenizer and was left to rock at 4°C for 30 min. Lysed material was centrifuged at 20,000 × g for 30 min to clarify, and supernatant containing mitoribosomes was carefully decanted. Lysis solution (2 ml) was added on top of a sucrose cushion (1 ml, 1 M sucrose (34%), 20 mM HEPES‐KOH pH 7.4, 100 mM KCl, 20 mM MgOAc, 0.6% β‐DDM, 0.06 mg/ml cardiolipin, 0.25 mM GMPPCP, 2 mM DTT) in SW 55 Ti tubes. Samples were centrifuged (231,550 × g at 4°C for 1 h) to obtain crude mitoribosome pool. Mitoribosomes were reconstituted in 100 μl of resuspension buffer (20 mM HEPES‐KOH pH 7.4, 100 mM KCl, 20 mM MgOAc, 0.3% β‐DDM, 0.03 mg/ml cardiolipin, 0.25 mm GMPPCP, 2 mM DTT) and loaded on top of a 13–30% linear sucrose gradient¹⁵ in SW 55 Ti tubes. Samples were centrifuged (213,626 × g at 4°C for 90 min), and entire volume was fractionated into 13 equal parts. Samples were snap frozen and kept for western blotting analysis. Inputs represent crude mitoribosome pool.

ND1 mitoribosomes were analyzed in a similar fashion, but crude mitochondria (see Subcellular fractionations) were directly lysed and subjected to linear sucrose gradient fractionation without purification of mitoribosome pool. Inputs represent crude mitochondria in these experiments.

Cells were seeded (1.0 × 10⁵ cells/well) in 6 well dishes and treated with the indicated compounds for 48 h under glucose starvation conditions. RNA was then isolated using Trizol (Invitrogen, 15596‐026) and a Zymo‐Spin Direct‐zol RNA Kit (Zymo Research, R2050) following the manufactures instructions. 0.5–1 μg of RNA was then used to reverse transcribe cDNA using a High‐Capacity cDNA Reverse Transcription Kit (Applied Biosystems, 4368813) following. cDNA was then mixed with SYBR Green qPCR master mix (Applied Biosystems, 4309155) and amplified on a CFX 384 Real‐Time system (Bio‐Rad) for analysis. Primers used for this study include Human: MALSU1 F‐CTGCTCGCCCCACTAATG, R‐ACTTGGGACCAGTATGATCTGC. MTIF3. DNAJB9 F‐TCTTAGGTGTGCCAAAATCGG, R‐TGTCAGGGTGGTACTTCATGG. SEC24D F‐GTCAACAAGGTTACGTGGCTA, R‐TAGTGCCCATAATGAGGTGGA. HSPA5 F‐CATCACGCCGTCCTATGTCG, R‐CGTCAAAGACCGTGTTCTCG. DDIT3 F‐GGAAACAGAGTGGTCATTCCC, R‐CTGCTTGAGCCGTTCATTCTC. PPP1R15A F‐ATGATGGCATGTATGGTGAGC, R‐AACCTTGCAGTGTCCTTATCAG.

ND1 or WT cybrid cells were seeded in glucose in 2 × 15 cm² dishes per condition and allowed to reach 90% confluence. Cells were pulsed with cycloheximide (200 μg/ml) for 0, 3, and 6 h after which they were harvested and snap frozen in liquid nitrogen. Whole cells and microsomes were then isolated (see Subcellular fractionations) for western blot analysis.

ND1 cells were seeded in 6 well dishes cultured in glucose, galactose, or galactose supplemented with 1 μM doxycycline and were incubated for 48 h. Cells were then pulsed with puromycin (1 μM) for the indicated time periods and harvested in RIPA buffer for western blotting as described above. Whole‐cell puromycinylated peptides were probed for with an anti‐puromycin antibody.

Animal studies were performed as previously reported by us (Perry et al, 2021). C57BL/6J NDUFS4^−/− or WT mice supplemented with doxycycline or normal chow diets were euthanized at ~p55 and brains were immediately harvested and snap frozen in liquid nitrogen and kept at −80°C. Brain tissue was then pulverized with mortar and pestle until a fine powder was obtained. Tissue was then resuspended in RIPA buffer for western blot analysis.

Animal studies were performed in compliance with the Institutional Animal Care and Use Committee protocol approved by the Beth Israel Deaconess Medical Center Animal Facility.

Replicates presented are derived from independent biological samples. Statistical analysis between two samples was done using student t‐test, and greater than two groups using one‐ or two‐way ANOVA with multiple comparisons where appropriate. Statistical significance was determined with P‐value < 0.05*, < 0.01**, < 0.001***, and < 0.0001**** as noted.