CANX and CALR downregulate ebolavirus GP1,2 expression at steady state
We identified PDIA3 as an EBOV-GP1,2 interacting protein in the ER by liquid chromatography tandem mass spectrometry (LC-MS/MS), resulting in discovery of EBOV-GP1,2 degradation via reticulophagy24. In the same study, we also identified three ER chaperones, CALR, CANX, and Heat Shock Protein Family A (Hsp70) Member 5 (HSPA5), that interacted with EBOV-GP1,2. To understand how these chaperones affect EBOV-GP1,2 expression, FLAG-tagged GP1,2 was expressed with HA-tagged PDIA3, HSPA5, CALR, or CANX in HEK293T cells, and their steady state expression was analyzed by western blotting (WB). GP1,2 expression was strongly reduced by PDIA3, CALR, and CANX, whereas HSPA5 did not have any effect (Fig. 1A). We previously reported that PDIA3 also decreases expression of GP1,2 with a MLD-deletion (GP∆MLD), but not sGP or ssGP24. When CALR and CANX were expressed with these different GP forms, CANX decreased GP∆MLD expression, whereas CALR did not (Fig. 1B, lane 3, 4, 11, 12), and neither decreased sGP and ssGP expression (Fig. 1B, lanes 5–8, 13–16). When GP interactions with CALR and CANX were tested by co-immunoprecipitation (IP), GP1,2, GP∆MLD, sGP, and ssGP all interacted with both CALR and CANX (Fig. 1C). These results demonstrate that CALR and CANX selectively downregulate GP1,2 expression, although they interact with all different GPs.
To validate these results, we tested the endogenous CALR and CANX activity on GP1,2 expression. When GP1,2 was expressed in HEK293T wild-type (WT), and PDIA3-, CALR-, or CANX-knockout (KO) cells, GP1,2 levels were much higher in all three KO cell lines than in the WT cell line, confirming the CALR and CANX inhibitory activity (Fig. 1D). CALR and CANX activity were further confirmed by their dose-dependent inhibition of GP1,2 expression in WT and their KO cell lines (Fig. 1E). Next, we tested CALR and CANX activity against GP1,2 proteins from the other ebolaviruses, including Sudan ebolavirus (SBOV), Bundibugyo ebolavirus (BDBV), Taï Forest ebolavirus (TAFV), and Reston ebolavirus (RESTV). Expression of these different GP1,2 proteins was strongly reduced by both ectopic (Fig. 1F) and endogenous (Fig. 1G) CALR and CANX. Notably, when GP1,2 from another filovirus, MARV, was tested, its expression was strongly increased by both ectopic and endogenous CALR and CANX (Fig. 1H). When CALR and CANX activity were further tested in several other cell lines, including A549, HeLa, Hep G2, mouse primary macrophages (MΦ), and human monocytic THP1 cell-derived macrophages (THP1-MΦ), they reduced GP1,2 expression to a similar level as in HEK293T cells (Fig. 1I). Collectively, these results demonstrate that CALR and CANX downregulate ebolavirus GP1,2 steady state expression in a cell type-independent manner.
CANX and CALR decrease EBOV entry
To understand the functional impacts of CANX and CALR activity on EBOV-GP1,2 expression, we used viral assays to test how EBOV entry is affected by these two ER chaperones. Initially, we used HIV-1 pseudoviruses that expressed EBOV-GP1,2 to measure EBOV entry, which is a standard assay for studying GP1,2 activity24, 25, 26. Envelope glycoprotein (Env)-deficient HIV-1 luciferase reporter viruses were produced in the presence of EBOV-GP1,2 from HEK293T WT, and PDIA3-, CALR-, or CANX-overexpressing, and PDIA3-, CALR-, or CANX-KO cells. An equal amount of these pseudoviruses was collected to infect HEK293T cells, and viral entry was determined by measuring intracellular luciferase activity. Like PDIA3, ectopic CANX and CALR expression strongly decreased, whereas their KOs strongly increased EBOV entry (Fig. 2A). Next, we tested whether their inhibitory activity could be recaptured by using EBOV replication and transcription-competent virus-like particles (trVLPs), which model the entire viral replication cycle27. Like PDIA3, both CALR and CANX consistently suppressed EBOV replication more than 10-fold in three consecutive passages (p0, p1, p2) in HEK293T cells (Fig. 2B). Furthermore, we tested how they affect GP1,2 levels in virions. EBOV virus-like particles (VLPs) were produced from CALR- or CANX-overexpressing, and CALR- or CANX-KO cells after expressing EBOV-GP1,2 with its matrix protein VP40. Both ectopic and endogenous CALR and CANX reduced GP1,2 levels in these VLPs, although the endogenous protein activity was slightly weaker than the ectopic protein activity (Fig. 2C, lower panels). Finally, we tested how CALR and CANX affect expression of glycoproteins from 6 other enveloped viruses, including influenza A virus (IAV)-hemagglutinin (HA), vesicular stomatitis virus (VSV)-glycoprotein (G), HIV-1 Env, Middle East respiratory syndrome coronavirus (MERS-CoV) spike protein (S), severe acute respiratory syndrome coronavirus (SARS-CoV) S, and SARS-CoV-2 S proteins. Neither ectopic nor endogenous CALR and CANX had any noticeable effect (Fig. 2D, Fig. 2E), except that ectopic CALR increased VSV-G and HIV-1 Env expression (Fig. 2D, lane 5 to 8). These results demonstrate that CALR and CANX decrease EBOV entry by reducing its GP1,2 levels in virions, which is consistent with their downregulation of GP1,2 in viral producer cells.
Interrelationship between CANX, CALR, and PDIA3 in EBOV-GP1,2 downregulation
CANX and CALR interact with PDIA3 and promote glycoprotein folding in the ER28, 29. To understand their interrelationship in EBOV-GP1,2 downregulation, we created single PDIA3-, CALR-, or CANX-KO, and dual PDIA3/CANX- or PDIA3/CALR-KO cell lines from HEK293T cells (Fig. 3A). Next, EBOV-GP1,2 was expressed with HSPA5, PDIA3, CALR, and CANX in these cells, and levels of GP1,2 expression were compared. HSPA5 barely affected GP1,2 expression in any of these cells (Fig. 3B, lanes 3, 8, 13, 18, 23). In contrast, PDIA3 (Fig. 3B, lanes 2, 7, 12, 17, 22) and CANX (Fig. 3B, lanes 5, 10, 15, 20, 25) strongly decreased GP1,2 expression in all these cells. However, CALR decreased GP1,2 expression much less effectively in the single PDIA3- or CANX-KO, and dual PDIA3/CANX- or PDIA3/CALR-KO cells, although an effective decrease was detected in the single CALR-KO cells (Fig. 3B, lanes 4, 9, 14, 19, 24). These results demonstrate that CALR is dependent on CANX and PDIA3, while CANX is independent of CALR and PDIA3 in downregulating GP1,2 expression. To further understand the role of PDIs in this downregulation, cells were treated with Tizoxanide (TIZ), which specifically inhibits the PDIA3 disulfide reductase activity30. Notably, GP1,2 downregulation by PDIA3, CALR, and CANX was all inhibited by TIZ in a dose-dependent manner by all three proteins (Fig. 3C). Thus, although the CANX activity was independent of PDIA3, it still requires a disulfide reductase activity to downregulate GP1,2.
CANX and CALR target EBOV-GP1,2 to the autolysosomes for degradation via ERAD machinery
To understand how CANX/CALR downregulates EBOV-GP1,2 expression, we blocked protein degradation pathways with their respective inhibitors. First, we blocked ERAD with three inhibitors, including kifunesine (Kif), eeyarestatin I (EerI), and lactacystin (Lac). Kif inhibits the class I α-mannosidases that initiate ERAD; EerI inhibits valosin-containing protein (VCP)/p97, a member of the AAA + ATPase family that is required for retro-transportation of ERAD client proteins from the ER to the cytoplasm; and Lac inhibits proteasomes. GP1,2 downregulation by CALR/CANX was blocked by Kif and EerI but not Lac (Fig. 4A), indicating that class I α-mannosidases and VCP/p97, but not proteasomes, are required. Second, we blocked the autolysosome activity with bafilomycin A1 (BafA1), concanamycin A (ConA), and NH4Cl, and the proteasome activity again with MG132 and Lac. CALR and CANX activities were both blocked by autolysosome inhibitors, but not proteasome inhibitors (Fig. 4B). Third, we blocked autophagosome synthesis by 3-methyladenine (3-MA), LY 294002 (LY), and wortmannin (Wort). CALR and CANX activities were both blocked by these inhibitors (Fig. 4C).
To confirm that EBOV-GP1,2 is targeted by macroautophagy/autophagy, we blocked the chaperone-mediated autophagy (CMA), which requires LAMP2A (lysosomal associated membrane protein 2A) as a receptor31. When LAMP2A was silenced by small interfering RNAs (siRNAs), CALR and CANX activities were not affected (Fig. 4D). Formation of autophagosomes is dependent on a number of ATG proteins, which induce lipidation of soluble MAP1LC3/LC3-I to become LC3-II on phagophore membranes32. When LC3-II synthesis was blocked in HEK293T cells by knocking out either ATG3 or ATG5, CALR and CANX activities were both blocked (Fig. 4E). We further knocked out a soluble autophagy receptor SQSTM1/p62 in HEK293T, HeLa, and A549 cells, and another autophagy receptor, histone deacetylase 6 (HDAC6) in A549 cells. CANX and CALR activities were both blocked in all three SQSTM1-KO cell lines (Fig. 4F), and none of them were blocked in the HDAC6-KO cell line (Fig. 4G). Thus, EBOV-GP1,2 is recruited to autophagosomes in a SQSTM1-dependent, but HDAC6-independent manner, in the presence of CANX and CALR.
To confirm that EBOV-GP1,2 is degraded in autolysosomes, we measured the GP1,2 turnover rate in the presence and absence of CANX and CALR after blocking cellular translation by cycloheximide (CHX). GP1,2 had a half-life of approximately 10 h, which was decreased to less than 2 h by CANX and CALR (Fig. 4H). This decrease was restored by treatment with BafA1 (Fig. 4H). We then used confocal microscopy to determine GP1,2 localization in autolysosomes using LAMP1 as a marker. GP1,2 and LAMP1 co-localization was clearly detected in the presence of CANX and CALR, which was further enhanced by treatments with BafA1 or NH4Cl (Fig. 4I). Collectively, these results demonstrate that CANX and CALR target EBOV-GP1,2 to the autolysosomes for degradation via ERAD machinery.
RNF26 decreases EBOV-GP1,2 expression in an E3 ubiquitin ligase activity-independent manner
Our results showed that SQSTM1/p62 is required for EBOV-GP1,2 degradation via autophagy. To be an active autophagy receptor, SQSTM1/p62 needs to be polyubiquitinated33. In fact, SQSTM1/p62 is polyubiquitinated by RNF26, which also tethers endosomes to the ER34. In addition, RNF26 recruits the ubiquitin conjugating enzyme E2 J1 (UBE2J1) that plays an important role in ERAD35, 36. Thus, we investigated the role of RNF26 in EBOV-GP1,2 polyubiquitination and degradation.
Initially, we investigated whether RNF26 targets EBOV-GP1,2. To confirm the ER localization of RNF26, RNF26 with a C-terminal mCherry tag was expressed in HeLa cells with the ER marker CALR or the Golgi marker TGOLN2 (Trans-Golgi network integral membrane protein 2), both of which with a C-terminal GFP tag. RNF26 co-localized with CALR, but not TGOLN2, when detected by confocal microscopy (Fig. 5A). To detect the RNF26 and EBOV-GP1,2 interaction, EBOV-GP1,2 with a C-terminal GFP tag was expressed with RNF26-mCherry in HeLa cells, and their co-localization was determined by confocal microscopy. EBOV-GP1,2 alone was detected on the plasma membrane, but was found in the cytoplasm in the presence of RNF26 (Fig. 5B). Importantly, strong EBOV-GP1,2 and RNF26 co-localization was clearly detected, indicating that their interaction occurred in the ER. In addition, ectopic RNF26 strongly decreased the expression GP1,2 from EBOV, RSTV, SBOV, and TAFV in HEK293T cells (Fig. 5C). Thus, RNF26 targets ebolavirus GP1,2 and decreases its expression.
Next, we investigated how RNF26 targets EBOV-GP1,2. RNF26 has 433 amino acids (aa) that comprise five transmembrane (TM) domains in the N-terminal half region and a RING-finger in the C-terminal region (Fig. 5D). We created five RNF26 deletion mutants to express its different regions that cover 76–433, 241–433, 1-385, 76–385, and 241–385 aa, and four point-mutation mutants C395S, C399S, C401S, and 3C/3S that have the critical Cys residues in the RING-finger replaced with Ser (Fig. 5D). When RNF26 WT and these mutants were expressed with EBOV-GP1,2 in HEK293T cells, all these RNF26 proteins were expressed at a similar level (Fig. 5E). However, unlike the others, mutants 241–433 and 241–385 did not decrease GP1,2 expression (Fig. 5E, lanes 4, 7). Because mutants C395S, C399S, C401S, 3C/3S, and 1-385 were still active, we conclude that the RING-finger is not required for this RNF26 activity. In addition, because mutant 76–385 was still active, whereas mutant 241–385 was not, we conclude that the 76–240 region that contains TM3, TM4, and TM5 domains is required for the RNF26 activity.
Next, we investigated how RNF26 interacts with EBOV-GP1,2 by co-IP. EBOV-GP1,2 was expressed with RNF26 WT and its mutants 76–433, 241–433, 1-385, 76–385, and 241–385, and RNF26 proteins were pulled down to detect EBOV-GP1,2. RNF26 WT and its mutants 76–433, 1-385, and 76–385 pulled down EBOV-GP1,2, whereas mutants 241–433 and 241–385 did not (Fig. 5F). These results confirm the RNF26 interaction with EBOV-GP1,2 that was suggested from the previous confocal microscopy data. In addition, they demonstrate that EBOV-GP1,2 interacts with RNF26 via its TM3, TM4, and TM5 domains.
Finally, we tracked the RNF26 and EBOV-GP1,2 interaction in live cells by bimolecular fluorescence complementation (BiFC). A basic yellow fluorescent protein Venus was divided into N-terminal (VN) and C-terminal (VC) fragments. HA-tagged VN and FLAG-tagged VC were fused to the C-terminus of RNF26 or EBOV-GP1,2, respectively. Expression of RNF26-VN or GP-VC alone in cells did not produce any green fluorescence, while co-expression did produce green fluorescence (Fig. 5G). This green fluorescence colocalized with the red fluorescence produced from RNF26 after staining with fluorescent anti-HA, confirming the specificity of these BiFC signals. These BiFC signals also colocalized with CALR that has a C-terminal blue fluorescent protein (BFP) tag (Fig. 5G). Collectively, these results confirm RNF26 downregulation of and its interaction with EBOV-GP1,
RNF26 supports CALR, but not CANX or PDIA3 to downregulate EBOV-GP1,2
To understand whether RNF26 is required for CALR, CANX, and PDIA3 activities, we knocked out RNF26 in HEK293T cells and generated three RNF26-KO cell lines 1-E4, 3-B8, and 3-F10 (Fig. 6A). When EBOV-GP1,2 and RSTV-GP1,2 were expressed in WT and 1-E4 cells, levels of their GP1,2 expression were much higher in KO than WT cells (Fig. 6B). Thus, endogenous RNF26 also decreases GP1,2 expression. Next, we detected PDIA3, CALR, and CANX activities in 1-E4 cells, using WT cells as a control. PDIA3 and CANX downregulated EBOV-GP1,2 expression in KO cells, whereas CALR did not (Fig. 6C). We then set up a co-IP assay to test whether RNF26 interacts with PDIA3, CALR, and CANX using GFP as a control. We detected RNF26 interaction with CALR and CNAX, but not PDIA3 and GFP (Fig. 6D). Lastly, we used co-IP to compare EBOV-GP1,2 polyubiquitination after expressing Ub in HEK293T cells in the presence of PDIA3, CALR, CANX, RNF26, and RNF26 catalytically inactive mutant 3C/3S. We detected polyubiquitinated GP1,2 products at above 170 kDa, which were greatly increased by PDIA3, CALR, and CANX, but not RNF26, or its mutant 3C/3S (Fig. 6E, IP, bottom panel). Collectively, our results demonstrate that RNF26 is not responsible for EBOV-GP1,2 polyubiquitination and selectively supports the CALR, but not CANX or PDIA3 activity.
EBOV-GP1,2 is polyubiquitinated with K27-linked Ub chains
To understand how EBOV-GP1,2 becomes polyubiquitinated, we determined the Ub chain linkage on EBOV-GP1,2. Seven lysine residues within Ub can be utilized for ubiquitination (K6, K11, K27, K29, K33, K48, K63) and linkage type directs the modified proteins to different cellular fates. Initially, we generated seven Ub mutants, K6R, K11R, K27R, K29R, K33R, K48R, and K63R, where each of its seven lysine residues were individually mutated to arginine. We also generated a Ub mutant, 7K/R, in which all seven lysine residues were mutated to arginine. EBOV-GP1,2 was expressed with each of these Ub mutants and PDIA3, CALR, or CANX in HEK293T cells, and EBOV-GP1,2 polyubiquitination was analyzed again by co-IP. As we already observed, these three proteins promoted EBOV-GP1,2 polyubiquitination in the presence of Ub; and importantly, similar levels of polyubiquitination were detected in the presence of mutants K6R, K11R, K29R, K33R, K48R, and K63R, but not K27R and 7K/R (Fig. 7A, IP, lower panels). These results suggested that Ub K27 plays an important role in EBOV-GP1,2 polyubiquitination.
To confirm the important role of K27 in EBOV-GP1,2 polyubiquitination, we created another seven Ub mutants, K6, K11, K27, K29, K33, K48, and K63, that only express each of its seven lysine residues individually. When these mutants were used to analyze EBOV-GP1,2 polyubiquitination, PDIA3, CALR, and CANX promoted EBOV-GP1,2 polyubiquitination only in the presence of Ub and its mutant that only retains K27, but not mutants that retain K6, K11, K29, K33, K48, or K63 (Fig. 7B, IP, low panels). These results confirm the important role of the Ub K27 residue in EBOV-GP1,2 polyubiquitination, suggesting that EBOV-GP1,2 is polyubiquitinated via K27-linked Ub chains.
RNF185 is responsible for EBOV-GP1,2 polyubiquitination and degradation
K27-linked ubiquitination plays an important role in the innate immune response and T cell signaling37. Many E3 Ub ligases have been reported to catalyze K27-linked ubiquitination, including RNF18538, membrane-associated RING-CH-type 8 (MARCH8)39, and TRIM2540. To identify the critical E3 ligase in EBOV-GP1,2 degradation, EBOV-GP1,2 and Ub were expressed with RNF185, RNF26, TRIM25, or MARCH8, and their catalytically inactive mutants 3C/3A, 3C/3S, 2E/2A, or W114A in HEK293T cells. GP1,2 proteins were pulled down and GP1,2 polyubiquitination was detected as we did previously. Among these four E3 ligases, RNF185, TRIM25, and MARCH8 strongly promoted EBOV-GP1,2 polyubiquitination, whereas their mutants did not, and, as we already observed, neither did RNF26 and its mutant (Fig. 8A, IP, lower panel). In addition. TRIM25 and MARCH8 did not decrease GP1,2 expression (Fig. 8B, lanes 7, 8). Nonetheless, MARCH8 inhibited GP1,2 proteolytic cleavage, as we previously reported 26. Notably, RNF185 decreased GP1,2 expression as did PDIA3, CALR, CANX, and RNF28 (Fig. 8B), whereas its mutant 3C/3A did not (Fig. 8A, Input, top panel, lanes 2, 3). Furthermore, GP1,2 downregulation by RNF185 was blocked by BafA1, ConA, 3-MA, Ly, and Wort, but not by Lac and MG132 (Fig. 8C, lanes 1–9). In contrast, GP1,2 downregulation by RNF26 was only slightly blocked by BafA1, ConA, and 3-MA, but not any other inhibitors (Fig. 8C, lanes 10–18).
RNF185 is another E3 Ub ligase on the ER membrane that plays an important role in ERAD41. Its 192 aa comprise one N-terminal RING-finger and two C-terminal TM domains (Fig. 8D). To understand how EBOV-GP1,2 is targeted, we created seven RNF185 deletion mutants including ∆2–18, ∆19–38, ∆39–80, ∆81–132, ∆133–155, ∆171–192, and ∆133–192, that lack the indicated regions (Fig. 8D). In addition, we created four point-mutation mutants, C39A, C42A, C79A, and 3C/3A that had the critical Cys residues in the RING-finger replaced with Ala (Fig. 8D). When RNF185 WT and these mutants were expressed with EBOV-GP1,2 in HEK293T cells, all these RNF26 proteins were detected by WB (Fig. 8E). Mutants ∆39–80, ∆171–192, ∆133–192, and 3C/3A did not decrease GP1,2 expression (Fig. 8E). These results demonstrate that the RING-finger and TM2 domain, and importantly, the E3 Ub ligase activity are all required for RNF185 downregulation of GP1,2 expression.
Next, we tested the RNF185 interaction with EBOV-GP1,2 by co-IP. EBOV-GP1,2 was expressed with GFP, RNF185 WT, and its mutants ∆2–18, ∆19–38, ∆39–80, ∆81–132, ∆133–155, ∆171–192, or ∆133–192 in HEK293T cells. GFP and RNF185 proteins were pulled down and their interactions with EBOV-GP1,2 were determined. RNF185 WT and its mutant ∆133–155 pulled down EBOV-GP1,2 with a similar efficiency; mutants ∆2–18, ∆19–38, and ∆81–132 pulled down EBOV-GP1,2 with much less efficiency; GFP and mutants ∆171–192 and ∆133–192 did not pull down any EBOV-GP1,2 (Fig. 8F). These results demonstrate that EBOV-GP1,2 interacts with RNF185 via its TM2 domain. In addition, the 2–38 and 81–132 region that are amino terminal to the RING-finger or TM1 domain are also required for RNF185 interaction with EBOV-GP1,2, likely by contributing to RNF185 structure integrity. To further validate these results, we tested whether these deletions affect RNF185 polyubiquitination activity toward EBOV-GP1,2. Mutants ∆39–80, ∆81–132, ∆171–192, and ∆133–192 all showed a deficiency in promoting EBOV-GP1,2 polyubiquitination, as did mutant 3C/3A when compared to the WT protein (Fig. 8G), consistent with our previous results.
RNF185 polyubiquitinates EBOV-GP1,2 on K673 in its cytoplasmic tail
EBOV-GP1,2 has a very short cytoplasmic tail (CT) that only has four residues 673KFVF676. To determine whether K673 is targeted for polyubiquitination, we created two EBOV-GP1,2 mutants by replacing K673 with Ala (K673A) or deleting all these four residues (∆CT). We found that RNF185, TRIM25, and MARCH8 could no longer polyubiquitinate mutants K673A and ∆CT (Fig. 9A, IP, lower panel). Thus, K673 is the polyubiquitination site on EBOV-GP1,2 for all these three E3 Ub ligases. Consistently, we also found that PDIA3, CALR, and CANX did not promote the polyubiquitination of mutants K673A and ∆CT (Fig. 9B, IP, lower panel). Finally, we determined whether these two mutants are still sensitive to those E3 ligases and ER proteins. We found that they were both resistant to MARCH8, TRIM25, RNF185, PDIA3, CALR, and CANX, although they were still as sensitive to RNF26 as was WT GP1,2 (Fig. 9C). To further confirm the role of RNF185 in EBOV-GP1,2 downregulation, we used siRNAs to knock down RNF185 (Fig. 9D). Notably, RNF185-knockdown strongly increased GP1,2 expression (Fig. 9E). Importantly, it also disrupted GP1,2 downregulation by PDIA3, CALR, and CANX (Fig. 9F). Collectively, these results demonstrate that CANX-CALR cycle co-opts RNF185 to polyubiquitinate EBOV-GP1,2 on K673, resulting in EBOV-GP1,2 degradation via reticulophagy.