|Home | About | Journals | Submit | Contact Us | Français|
Torsin ATPases are the only representatives of the AAA+ ATPase family that reside in the lumen of the endoplasmic reticulum (ER) and nuclear envelope. Two of these, TorsinA and TorsinB, are anchored to the ER membrane by virtue of an N-terminal hydrophobic domain. Here we demonstrate that the imposition of ER stress leads to a proteolytic cleavage event that selectively removes the hydrophobic domain from the AAA+ domain of TorsinA, which retains catalytic activity. Both the pharmacological inhibition profile and the identified cleavage site between two juxtaposed cysteine residues are distinct from those of presently known proteases, suggesting that a hitherto uncharacterized, membrane-associated protease accounts for TorsinA processing. This processing occurs not only in stress-exposed cell lines but also in primary cells from distinct organisms including stimulated B cells, indicating that Torsin conversion in response to physiologically relevant stimuli is an evolutionarily conserved process. By establishing 5-nitroisatin as a cell-permeable inhibitor for Torsin processing, we provide the methodological framework for interfering with Torsin processing in a wide range of primary cells without the need for genetic manipulation.
The human genome encodes four Torsin ATPases (TorsinA, TorsinB, Torsin 2A, and Torsin3A), all of which are members of the ATPases associated with a variety of cellular activities (AAA+)3 family (1, 2). Of those, TorsinA is the best characterized Torsin ATPase due to its association with the congenital movement disorder DYT1 dystonia (3). TorsinA has been implicated in a variety of cellular processes, including protein quality control and cellular stress responses (4,–9), although its precise mechanistic functions remain to be identified (2, 10). Although Torsins are phylogenetically closely related to Clp/HSP100 proteins, they are characterized by a number of atypical features (for a recent review, see Refs. 2 and 10). On the primary structural level, these include differences in the Walker A motif responsible for nucleotide binding, as well as the absence of an otherwise conserved arginine residue required for ATP hydrolysis (10).
TorsinA is directed to the lumen of the endoplasmic reticulum by a cleavable N-terminal signal sequence, which is followed by a hydrophobic domain responsible for membrane anchoring and ER retention (11). TorsinA is partitioned between the ER and the contiguous perinuclear space of the nuclear envelope, and its localization is controlled in part by its association with LAP1 and LULL1, type II transmembrane proteins residing in the nuclear envelope and ER, respectively (12,–14). Another unusual property of Torsins is that they lack significant basal ATPase activity in isolation and are tightly regulated by LAP1 and LULL1, which integrate into the Torsin ring via AAA-like domains to trigger ATP hydrolysis through an active site complementation mechanism (15,–17). Thus, Torsins and their ATPase activating cofactors form a composite, membrane-spanning assembly. This assembly is sensitive to the redox environment of the ER (18) and is perturbed by the DYT1 dystonia-causing mutation, presumably due to a defect in the Torsin-activator interface (16), underscoring the importance of the regulation of this assembly. Although a key role for a nucleotide-proximal cysteine has been established in the context of redox regulation (18, 19), the functional significance of the cysteine cluster positioned near the luminal membrane interface is presently unclear.
In this study, we demonstrate that this cysteine cluster, which is highly conserved in the Torsin family, serves as a cleavage site for an unusual proteolytic activity. This activity cleaves specifically between two juxtaposed cysteine residues in response to perturbation of the oxidative environment in the ER and perinuclear space. The identified cleavage site and its close positioning relative to the membrane collectively represent an unprecedented case of regulatory proteolysis. Together with the identification of a potent and specific protease inhibitor, this study establishes the framework to identify this novel membrane-embedded or -associated proteolytic activity. We, furthermore, demonstrate that this activity is particularly abundant in the ER of stimulated B cells. Here, TorsinA is processed in response to LPS stimulation even in the absence of reducing agents, establishing physiological relevance of TorsinA processing during B cell differentiation.
HEK 293T cells, HeLa cells, and HFF (human foreskin fibroblast) cells from ATCC were routinely cultured at 37 °C in DMEM supplemented with 10% (v/v) FBS in a humidified incubator saturated with 5% (v/v) CO2. TorA mutant TorA ΔC49 with a N-terminal MHC-I HLA-A signal sequence and a C-terminal HA tag was cloned by a standard PCR-based procedure into pcDNA3.1+ vector. For transient expression, plasmids were transfected into HEK 293T cells using LipofectamineTM 2000 (Life Technologies). Experiments were performed 24 h post-transfection as described previously (20), unless otherwise specified.
Wild-type C57BL/6 mice were maintained at our animal facility strictly following the guidelines provided by the Wistar Institute Committee on Animal Care.
Splenocytes were obtained from mice by mashing the spleens through cell strainers followed by RBC lysis (155 mm NH4Cl, 10 mm Tris-HCl, 0.1 mm EDTA) (Sigma). Naive B cells were purified from mouse spleens by negative selection using CD43 (Ly48) MicroBeads, according to the manufacturer's instructions (Miltenyi Biotech). Mouse B cells were cultured in the RPMI 1640 media (Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS), 2 mm l-glutamine, 100 units/ml of penicillin G sodium, 100 μg/ml of streptomycin sulfate, 1 mm sodium pyruvate, 0.1 mm non-essential amino acids, and 0.1 mm β-mercaptoethanol.
Mouse spleens and peripheral lymph nodes were mashed through 40-μm cell strainers; splenocytes and lymph node cells were collected after RBC lysis. Mouse livers were cut into small pieces, washed with PBS, snap frozen, and ground with a mortar and pestle in liquid nitrogen. Cells and tissue homogenates were lysed in RIPA buffer (10 mm Tris-HCl, pH 7.4, 150 mm NaCl, 1% (w/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 1 mm EDTA) supplemented with complete protease inhibitors (Roche).
Generation of anti-TorA antibodies was described previously (15). Antibodies to mouse light chain and heavy chain (SouthernBiotech), p97 (Fitzgerald), PDI (Abcam), and actin (Sigma) were obtained commercially. LPS (Sigma), CpG-1826 oligodeoxynucleotides (TIB-Molbiol), dithiothreitol (DTT) (Sigma), Tris(2-carboxyethyl)phosphine hydrochloride (TCEP) (Pierce), NEM (Sigma), PMSF (Sigma), N,N,N′,N′-tetrakis(2-pyridylmethyl)ethane-1,2-diamine (TPEN) (Sigma), 5-nitroisatin (Sigma), thapsigargin (Enzo Life Sciences), tunicamycin (Enzo Life Sciences), brefeldin A (Cell Signaling), Endo H (New England Biolabs), and PNGase F (New England Biolabs) were purchased commercially. Subtilase cytotoxin (SubAB) was provided by Dr. James C. Paton and Dr. Adrienne W. Paton at the University of Adelaide. We developed and synthesized the IRE-1 RNase inhibitor, B-I09 (21).
To induce unfolded protein response, 5 μg/ml of tunicamycin, 5 μm thapsigargin and various concentrations of DTT were added to HeLa cell cultures and incubated at 37 °C for various time periods. Total RNA was extracted by RNeasyTM Plus mini kit (Qiagen). The cDNA of XBP-1 (X-box binding protein 1) was amplified by RT-PCR using a first strand cDNA synthesis kit (Roche). XBP-1 splicing was analyzed by RT-PCR using primers 5′-ACAGCGCTTGGGGATGGATG-3′ and 5′-CCATGGGGAGATGTTCTGGAG-3′, encompassing the missing sequence in the spliced XBP-1. The PCR fragments were separated by 11% PAGE in TBE (Tris borate-EDTA) buffer and visualized by ethidium bromide staining (22).
Human TorsinA and control (GFP) shRNA constructs were purchased from Sigma. The lentivirus was produced according to instructions provided on the Broad Institute website. HeLa cells in a 6-well plate were infected by addition of 1 ml of lentivirus supplemented with 8 μg/ml of Polybrene transfection reagent (Millipore). Antibiotic selection (1 μg/ml puromycin) was started after 24 h.
Twenty-four hours post-transfection, cells were metabolically labeled overnight, lysed, and immunoprecipitated with anti-TorA antibodies as described previously (15). The immunoprecipitates were separated by SDS-PAGE and analyzed by autoradiography. Proteinase protection assay was performed as described previously (23) prior to immunoprecipitation with anti-TorsinA antibodies.
Metabolically labeled HeLa cells were incubated in 300 μl of ice-cold permeabilization buffer (50 mm Tris, 250 mm sucrose, 5 mm MgCl2, pH 7.5) supplemented with 0.025% (w/v) digitonin and 75 mm NaCl on ice for 5 min. Cytosol fraction was squeezed out by centrifugation at 13,000 rpm and 4 °C for 10 min. The cell pellet was resuspended in 300 μl of permeabilization buffer supplemented with 2-fold ATP regeneration system (6 mm phosphoenolpyruvate, 20 units/ml of pyruvate kinase, and 4 mm ATP). 20 μm protease inhibitors from protease inhibitor library (Yale Center for Molecule Discovery) were incubated with 100 μl of resuspended cells on ice for 15 min before adding 2 mm DTT (final concentration). Finally, a 100-μl aliquot of cytosol fraction was added and the mixture was incubated at 37 °C for 3 h. The cells were lysed by 1% SDS and TorA was immunoprecipitated for autoradiography.
HEK 293T cells transfected with HA-tagged TorA or TorA ΔC49 on a 10-cm plate were lysed in 1 ml of phase separation buffer (50 mm Tris, pH 7.5, 2% (w/v) Triton X-114 (Sigma), 200 mm NaCl, and 5 mm MgCl2) supplemented with complete protease inhibitors (Roche) on ice for 30 min. Phase partition was performed as described (14) with two washing steps (24). The aqueous and hydrophobic phases were brought to the same volume and Triton X-114 concentration prior to immunoblotting.
Ten 15-cm plates of HEK 293T cells transiently expressing HA-tagged TorA were incubated at 37 °C for 3 h in the absence or presence of 5 mm DTT. Cells were lysed with 1% SDS in PBS and diluted 10-fold with immunoprecipitation buffer (50 mm Tris, pH 7.5, 75 mm NaCl, 5 mm MgCl2, and 0.5% Nonidet P-40) prior to immunoprecipitation with 200 μl of HA affinity beads. The immunoprecipitates were separated by 12% SDS-PAGE, electroblotted onto a PVDF membrane (Bio-Rad), and stained with Coomassie staining. The protein band corresponding to the processed TorA were cut out and sequenced by Edman degradation on a Procise Protein Sequencing System (Applied Biosystems) by Alphalyse.
TorsinA (TorA) has been implicated in cellular stress response pathways including ER stress and oxidative stress (7, 18, 19, 25, 26). We therefore set out to investigate whether TorA levels are modulated in response to conditions of exacerbated protein misfolding. To this end, we metabolically labeled HeLa cells with [35S]cysteine/methionine overnight and then exposed cells to 5 mm DTT, 5 μg/ml of tunicamycin, 5 μm thapsigargin, or heat shock (42 °C) in the continued presence of [35S]cysteine/methionine. Cells were lysed in SDS after 3 h of treatment and endogenous TorA was immunoprecipitated using an affinity-purified TorA antibodies. The immunoprecipitates were resolved via SDS-PAGE followed by autoradiography (Fig. 1A, upper panel). Endogenous TorA is efficiently immunoprecipitated from cells as evidenced by the band of the expected electrophoretic mobility (~37 kDa, Fig. 1A, arrow). When cells are treated with the reducing agent DTT, an additional band of approximately ~35 kDa is also observed (Fig. 1A, third lane, arrowhead). This additional band was not observed when cells were treated with tunicamycin, thapsigargin, or exposed to heat shock (Fig. 1A, fourth to sixth lanes).
To confirm that our treatments adequately induced an unfolded protein response resulting from protein misfolding, we monitored XBP-1 splicing via RT-PCR to validate the induction of the unfolded protein response. As expected (27), XBP-1 splicing as a readout of IRE-1 activity was potently induced by the reducing agent DTT, the calcium ionophore thapsigargin, as well as the N-glycosylation inhibitor tunicamycin, all of which are established inducers of the unfolded protein response (Fig. 1A, lower panel). Thus, the appearance of the second TorsinA species is selectively induced by a reducing agent, and not by unrelated pharmacological agents that induce protein misfolding or heat shock.
To investigate whether the lower molecular mass species induced by DTT stress is in fact a derivative of TorA, we treated cells that stably express either a shRNA against GFP (shGFP) as a control or against TorA (shTorA) with DTT, and immunoprecipitated TorA. Both TorA and the DTT-inducible protein are present in the shGFP cells (Fig. 1B, second and third lanes, arrow and arrowhead), but are absent from shTorA cells (Fig. 1B, fifth and sixth lanes). We conclude that DTT stress does indeed induce the formation of a TorA species of a ~35 kDa, which we will refer to as “TorAp.”
To generalize our findings, we next investigated whether TorAp could be induced by DTT in primary cells. We metabolically labeled HeLa cells and HFFs with [35S]cysteine/methionine overnight, and then treated the cells with increasing concentrations of DTT for different periods of time. The cells were harvested for immunoprecipitation using the TorA antibodies, and the immunoprecipitates were subjected to SDS-PAGE followed by autoradiography (Fig. 1C). TorAp is produced in both HeLa cells and HFFs in response to DTT stress (Fig. 1C, arrowhead), confirming that this phenomenon applies to primary cells as well. Of note, we observed both a time and a concentration dependence for the formation of the TorAp species, as judged by a densitometric quantification (Fig. 1D).
Having demonstrated that TorAp can be detected in several cell types (Fig. 1), we sought to determine the nature of TorAp. We considered two possibilities: first, TorAp may be specified by an alternative, DTT-induced transcript resulting either from alternative splicing or utilization of an alternative transcriptional start site. Second, TorAp may represent a post-translationally modified form of TorA.
To discern between these formal possibilities, we metabolically labeled cells with [35S]cysteine/methionine overnight and treated the cells with cycloheximide to block de novo translation prior to DTT treatment. After 30 min of cycloheximide treatment, we stressed cells with DTT for 1.5 and 3 h before immunoprecipitating endogenous TorA (see diagram in Fig. 2A). At 1.5 h after addition of DTT, we robustly detected TorAp (Fig. 2A) even in the presence of a cycloheximide concentration of 100 μg/ml that exceeds the dosage required to completely block de novo protein synthesis. Similarly, TorAp is readily observed after 3 h of DTT treatment in a parallel control experiment from which cycloheximide was omitted (Fig. 2A). We attribute the noticeable decrease in the total signal at the 3-h time point in the presence of both DTT and cycloheximide to the cumulative cellular toxicity elicited by those compounds. Regardless, we can already conclude from the first (1.5 h) time point that TorAp is not a novel isoform made from the TorA gene, but instead is derived from a pre-existing TorA pool.
Arguably, the most likely explanations for the smaller molecular mass of TorAp relative to TorA are (i) deglycosylation due to exposure to a cytosolic N-glycanase or (ii) proteolytic processing. To characterize the putative precursor-product relationship, we first asked whether TorA, which is glycosylated at positions Asn143 and Asn158 (28), becomes deglycosylated in response to DTT treatment, which could be indicative of a previously suggested TorsinA dislocation from the ER to the cytosol (29). To this end, we immunoprecipitated TorA from [35S]cysteine/methionine-labeled cells and treated the immunoprecipitates with Endo H or PNGase F. Endo H removes high mannose N-glycans, whereas PNGase F removes complex N-glycans and fully deglycosylates N-glycoproteins. If TorAp represents a deglycosylated state of TorA, then treating DTT-stressed samples with Endo H or PNGase F should collapse all TorA protein present to a single species with the same electrophoretic mobility as TorAp. After immunoprecipitation followed by SDS-PAGE and autoradiography, TorA can be observed as a single major band at ~37 kDa under normal cellular conditions (Fig. 2B, lane 3, arrow, TorA). TorA immunoprecipitates treated with Endo H collapse to a single lower molecular mass species (Fig. 2B, lane 4), indicating that TorA is efficiently retained in the ER and does not escape to the Golgi where it would acquire Endo H resistance, consistent with previous observations (11). TorA immunoprecipitates treated with PNGase F similarly collapse to a single, fully deglycosylated TorA (TorA-CHO) species (Fig. 2B, lane 5, arrow, TorA-CHO). We then performed the same experiment, but in DTT-stressed cells. Both TorA and TorAp are readily observed upon DTT treatment (Fig. 2B, lane 7, arrow, TorA and TorAp). Treating the immunoprecipitates with Endo H causes the two TorA bands of TorA and TorAp to shift to two lower molecular mass bands: one band corresponds to the TorA Endo H-digested species seen in the control experiment without DTT, and one additional lower band represents Endo H-digested TorAp (Fig. 2B, lane 8). Treating the immunoprecipitates with PNGase F also causes TorA and TorAp to shift to two lower molecular mass bands of TorA-CHO and TorAp-CHO (Fig. 2B, lane 9, arrow, TorA-CHO and TorAp-CHO).
Based on the observations that (i) TorAp remained sensitive to Endo H and PNGase F treatment and because (ii) TorA did not collapse to a size equivalent to that of TorAp after PNGase F treatment, we can rule out that TorAp merely represents a deglycosylated variant (also see Fig. 6B). It is thus unlikely that TorAp gained access to the cytosol, where it would be readily attacked by N-glycanase to yield a deglycosylated form (30).
To confirm the suggested topology of TorAp, we next performed a protease protection assay. We homogenized [35S]cysteine/methionine-labeled HeLa cells in a hypotonic buffer in the absence of detergent and performed a protease protection assay using the ER luminal protein PDI as a control. As expected, TorA is protected from proteinase K digestion (Fig. 2C, upper, lane 4), confirming that it is a luminal ER protein similar to PDI (Fig. 2C, lower, lane 2), but can become digested upon membrane solubilization with Nonidet P-40 (Fig. 2C, upper, lane 5). The observable net loss of the signal in the presence of Proteinase K is due to a temporary disruption of the ER during the lysis process and the concomitant leakage of a small percentage of luminal components, rendering them sensitive to the protease. We then tested whether this was also true for TorAp. Indeed, TorAp is likewise protected from proteinase K digestion (Fig. 2C, upper, lane 8), but can become digested upon membrane solubilization (Fig. 2C, upper, lane 9), similar to the ER luminal protein PDI, which was monitored as a control by subjecting the supernatants obtained from the primary anti-TorsinA IP to a secondary IP with anti-PDI antibodies (Fig. 2C, lower panel). Taken together, these data rule out that TorAp represents a deglycosylated cytosolic or non-glycosylated form of TorA, suggesting that a cleavage event is responsible for the lower molecular mass of TorAp.
If TorA is indeed cleaved by a protease to produce TorAp, we should be able to inhibit TorAp formation with a suitable protease inhibitor. We first tested whether generic enzyme inhibitors such as NEM, EDTA, TPEN, and PMSF could inhibit TorAp production, which would allow us to tentatively assign the proteolytic activity to the family of cysteine proteases, metalloproteases, or serine proteases, respectively. To enrich for TorA and TorAp and to facilitate the uptake of inhibitors, we transfected HEK 293T cells with TorA-HA and squeezed out the cytosol by digitonin permeabilization followed by centrifugation. Given that the ER membrane is more permeable to small molecules than other cellular membranes (31), we anticipate that the majority of inhibitors should gain access to the ER lumen following this procedure, although we cannot exclude the formal possibility that some inhibitors cannot gain access to ER-resident proteases, which would result in a false-negative observation. Because those cell-impermeable inhibitors would be of little utility in a cellular setting, we consider this a minor limitation.
The resulting permeabilized cells were resuspended in buffer and incubated with 5 mm NEM, EDTA, TPEN, or PMSF for 15 min, followed by readdition of the cytosol and −/+ 2 mm DTT for 3 h. The cells were then lysed in SDS, treated with Endo H to ease electrophoretic analysis, and blotted against the HA tag. We observed that NEM efficiently inhibited production of TorAp (Fig. 3A, TorAp-HA - CHO), whereas EDTA, TPEN, and PMSF had no effect. We next extended our analysis to a protease inhibitor library of over fifty compounds (Fig. 3B, supplemental Table S1). We similarly permeabilized [35S]cysteine/methionine-labeled HeLa cells with digitonin to squeeze out the cytosol, resuspended the permeabilized cells in buffer, and incubated the suspension with 20 μm of each protease inhibitor for 15 min before readdition of the cytosol and 2 mm DTT for 3 h. We lysed the cells with SDS and immunoprecipitated endogenous TorA. The immunoprecipitates were resolved by SDS-PAGE followed by autoradiography. We observed a near-complete inhibition of TorA proteolytic processing by protease inhibitor number 29, which corresponds to the compound 5-nitroisatin (Fig. 3B, arrowhead, see C for the chemical structure). To assess the affinity of this compound, we repeated the experiment with varying concentrations of 5-nitroisatin, and observed potent inhibition of TorAp production with 30 μm 5-nitroisatin (Fig. 3D). A densitometrical analysis of TorAp production as a function of 5-nitroisatin concentration allowed us to crudely estimate that the compound has an IC50 value of approximately ~25 μm, qualifying 5-nitroisatin as a potent lead compound.
Having established that TorA processing can be inhibited by cysteine protease inhibitors, such as NEM (Fig. 3A) and 5-nitroisatin (Fig. 3, B–D), we next set out to map the proteolytic cleavage site. Because we observe DTT-induced processing of C terminally HA-tagged TorA and can detect the cleavage product by blotting against the HA tag (Fig. 3A), the cleavage site most likely occurs closer to the N terminus of the TorsinA. To rule out autocatalytic or internal processing, e.g. via protein splicing (32), we reconstituted a TorA-LULL1 complex of authentic full-length components in proteoliposomes, using previously established methodology (15). No processing was observed upon exposure to DTT (Fig. 4A), indicating that a trans-acting proteolytic activity is responsible for TorA processing.
Based on the electrophoretic mobility of TorAp, we estimated that proteolysis must occur between the N-terminal hydrophobic domain and the domain boundary of the AAA+ domain (Fig. 4B). Given that this region is cysteine-rich (cf. Fig. 5A), a property that is detrimental for mass spectrometry-based approaches, we employed Edman degradation to identify the putative cleavage site. To this end, we transiently transfected 293Ts with TorA-HA and treated cells with 5 mm DTT for 3 h. Cells were lysed in SDS and TorA-HA was immunoprecipitated with HA affinity beads. Immunoprecipitates were resolved via SDS-PAGE, transferred to a PVDF membrane, and stained with Coomassie Blue to visualize TorA and TorAp. The TorAp band was excised and analyzed via Edman degradation. Due to a suboptimal signal-to-noise ratio, multiple amino acid assignments could be made for a number of amino acid positions (Table 1). However, given that we have rigorously established that the DTT-inducible product is a derivative of TorA (Figs. 11–3), we can, nevertheless, define the starting position based on the Edman degradation results (Table 1, cf. Fig. 4B). After comparing the N-terminal sequencing results with the amino acid sequence of TorA, only one possible combination of residues can give rise to the sequencing results obtained. Those residues are “-GQK-SLS-E,” where dashes correspond to residues that give blank or low intensity signals due to unstable phenylthiohydantoin-derivatives, such as cysteine and arginine (33). Therefore, the N-terminal sequencing data are consistent with the TorAp starting with amino acids “CGQKRSLSRE” located at positions 50–59, indicating that the proteolytic cleavage site occurs between TorA amino acids 49 and 50 (Fig. 4B). Of note, residues 49 and 50 are two strictly conserved adjacent cysteine residues located within the N-terminal hydrophobic domain of TorA (cf. Fig. 5A)(11).
To confirm the deduced cleavage site, we cloned a C terminally HA-tagged TorA construct in which the first 49 amino acids are deleted (TorA Δ49-HA) to represent the cleavage product. Because TorA Δ49-HA is deprived of its authentic signal sequence, we installed an N-terminal MHC-1 HLA-A signal sequence to ensure proper ER targeting as well as defined signal peptidase-mediated processing (Fig. 4C). We transfected TorA-HA and TorA Δ49-HA into 293T cells, treated cells with DTT, harvested the protein lysates, and treated the lysates with Endo H. We subjected the lysates to SDS-PAGE and blotted against anti-HA to detect TorA-HA and TorA Δ49-HA. If TorA Δ49-HA is the DTT-induced cleavage product, then it should have the same electrophoretic mobility as TorAp-HA. As judged by SDS-PAGE and immunoblotting, we found that this is indeed the case (Fig. 4D). Along the same line, TorA Δ49-HA should also no longer be susceptible to DTT-induced proteolytic cleavage if it indeed represents the cleavage product. When cells transfected with TorA Δ49-HA are stressed with DTT, no additional lower molecular mass band is observed (Fig. 4D). We thus conclude that TorA is proteolytically cleaved between residues Cys49 and Cys50 to produce TorA Δ49.
It was previously suggested that the N-terminal hydrophobic domain of TorA forms an amphipathic α helix rendering TorA a monotopic membrane protein (11). Because TorA is cleaved within its N-terminal membrane domain between Cys49 and Cys50, effectively removing the majority of hydrophobic domain, it seems reasonable to propose that cleavage could mobilize TorA from the membrane. To address this question, we performed a Triton X-114-based phase separation experiment from 293T cells transfected with TorA-HA or TorA Δ49-HA and indeed observed that TorA Δ49-HA partitions into the aqueous phase, whereas TorA-HA exclusively partitions to the hydrophobic phase (Fig. 4E).
These data support the idea that TorA is proteolytically cleaved within its hydrophobic domain, liberating it at least in part from the ER membrane. The observation that some of the processed TorsinA remains membrane-associated could be due to its binding to remaining full-length TorsinA or cofactor, although we note that even PDI, an established soluble, ER resident protein, partitioned into the hydrophobic phase to a minor extent.
TorA contains six strictly conserved cysteines, three of which are located within the N-terminal membrane domain (Fig. 5A). The Caenorhabditis elegans homolog OOC-5 has previously been shown to contain two disulfide bonds, one located near the C terminus and one near the N terminus, the latter of which could not be unambiguously mapped (19). Because human TorA is cleaved between Cys49 and Cys50 to create TorAp, we investigated whether the N-terminal cysteines of TorA are important for proteolytic processing. We transfected 293Ts with TorA-HA wild-type or constructs containing Cys44, Cys49, or Cys50 mutated to serines (TorA C44S-HA, C49S-HA, or C50S-HA). Cells were either not treated or treated with DTT, and their lysates were harvested for subsequent immunoblotting analysis. TorA-HA wild-type is proteolytically processed into TorAp when cells are treated with DTT, as was previously demonstrated (Fig. 5B). TorA-HA C44S, C49S, and C50S mutants are also proteolytically cleaved in response to DTT treatment (Fig. 5B). However, proteolytic cleavage is also evident for each of the cysteine mutants even in the absence of DTT (Fig. 5B, third, fifth, and seventh lanes). This is most apparent for TorA-HA C49S (Fig. 5B, fifth lane). Because TorA cysteine mutants show basal levels of cleavage even without DTT stress, maintaining the structural integrity of the N-terminal cysteine cluster of TorA is likely critical for preventing uncontrolled proteolytic processing.
We next addressed the specificity of the cleavage reaction. The first question is whether TorA cleavage is indeed caused by a more reducing environment in the otherwise oxidizing ER, and how we may interpret the inhibitory effect of NEM (Fig. 3A). One interpretation would be that NEM merely inactivates DTT, thus reducing its potency as reducing agent. To eliminate this possibility, we used the potent reducing agent TCEP, which does not react with maleimides. In this case, the inhibition by NEM was still observed (Fig. 5C), ruling out indirect effects and supporting the conclusion that shifting to a more reducing environment in the ER is at least in part responsible for TorA processing.
The second question pertains to the sequence specificity of the cleavage reaction. There are now still two possible interpretations of our observation that the alkylating agent NEM blocks proteolysis. Perhaps the most intuitive one is that that TorA is processed by a cysteine protease, which is inhibited by NEM. An alternative possibility is that the cysteine-rich targeting site is specifically recognized by the protease and that an alkylation of one or several cysteines therein prevents proteolysis. To resolve this issue, we exchanged Cys49 for a tyrosine residue to mimic the bulky product of the alkylation reaction. We found that TorA C49Y was resistant to proteolysis (Fig. 5E). This observation is consistent with the interpretation that alkylation of the cysteine at the protease target site renders it resistant to proteolysis. Combined with the observation that a TorA variant in which all three cysteines in this region are mutated to near-isosteric serines, which are resistant to alkylation is still proteolytically processed in the presence of NEM (Fig. 5D), we consider it unlikely that the protease is a cysteine protease. However, we cannot formally exclude that NEM fails to react with an active site cysteine in a putative cysteine protease for steric reasons.
Having shown that DTT-induced TorA proteolytic processing occurs across several human cell lines and primary cells (Figs. 1 and and33A), we next asked whether this processing occurs under physiologically relevant conditions and in other animal species. To investigate the biological significance of TorA processing, we examined whether TorA is also proteolytically processed in murine B cells, a well established physiological example of unfolded protein response induction and ER expansion enabling a massive production and secretion of immunoglobulins (34,–36).
We first established whether TorA is also proteolytically processed in murine B cells upon addition of DTT. We prepared detergent extracts from B cells that were exposed to a series of conditions eliciting ER stress, and scored for TorA processing via SDS-PAGE in conjunction with immunoblotting. Consistent with our earlier results in human cells (Fig. 1), we observed that TorA is processed selectively in the presence of DTT (Fig. 6A). Notably, the efficacy of conversion from TorA to TorAp is significantly higher than in human cells: TorA is nearly quantitatively processed to TorAp, whereas the conversion in HeLa, 293T, or HFF cells did not exceed 50% even under the most favorable experimental conditions (cf. Figs. 1 and and66A).
We next exploited the quantitative conversion to reinforce our earlier interpretation that TorAp represents an ER-luminal species (cf. Fig. 2B). We exposed B cell-derived detergent extracts from untreated or DTT-stressed cells to Endo H and PNGase F, and analyzed them by SDS-PAGE and immunoblotting. As expected, TorAp remained Endo H and PNGase F sensitive (Fig. 6B). The difference in electrophoretic mobility between untreated and Endo H-treated species is consistent with the loss of both N-glycan modifications, supporting our earlier conclusion that TorAp remains in the ER and does not represent a deglycosylated cytosolic intermediate.
Finally, we asked whether Torsin processing occurs under more physiological conditions of ER expansion even in the absence of DTT. To this end, we isolated detergent extracts from mouse B cells that were either untreated or stimulated by LPS to induce B cell differentiation. To ensure that B cell differentiation occurs with normal kinetics, we monitored the production of the immunoglobulin light (κ) chains as well as the membrane-bound μ (μmem) and secreted μ (μs) heavy chains. As expected, we observed the increased synthesis of the μs heavy chain as well as κ light chain in response to LPS stimulations (Fig. 6C).
Notably, we observed TorA processing starting at 24 h post-addition of LPS (Fig. 6C). Importantly, we observed a time-dependent increase in the levels of TorAp over the course of B cell differentiation, even in the absence of reducing agents. This observation strongly supports the idea that TorA processing is a physiologically relevant phenomenon, as exemplified by B cell differentiation and concomitant ER expansion.
Having established that 5-nitroisatin potently blocks Torsin processing in permeabilized cells, we next wanted to test if primary cells can be effectively manipulated with this compound. Indeed, we found that DTT-induced processing in primary B cells can be blocked by 5-nitroisatin in a dose-dependent fashion (Fig. 7A). Importantly, this inhibition did not require cell permeabilization, and blocked Torsin processing completely at a higher dosage of compound (Fig. 7B). This we consider to be a proof of concept that 5-nitroisatin can be utilized for similar studies in a wide range of cell types.
Notably, 5-nitroisatin treatment leads to a strong increase in the production of immunoglobulin light (κ) chains as well as the membrane-bound heavy chain (μmem) (Fig. 7B). At this point, we cannot exclude the formal possibility of an off-target effect. However, we conclude that these data are consistent with an important role of Torsin processing for B cell biology and represent a phenomenon that warrants further investigation in future studies.
Finally, we wanted to investigate if previously observed, shorter TorA isoforms in organ isolates (37, 38) correspond to TorAp, which would further support the idea that Torsin processing is physiologically relevant and occurs in authentic tissues. To this end, we prepared tissue homogenates from spleen, lymph nodes, and liver, and subjected the resulting detergent extracts to SDS-PAGE and immunoblotting. No processing was observed in spleen or peripheral lymph nodes (Fig. 7C). Consistent with our in vitro stimulation experiment using LPS (cf. Fig. 6C, first lane), the B cell population in the spleens and lymph nodes appears to be mainly unstimulated and undifferentiated in juvenile mice.
Next, we confirmed the previous observation of a second TorA isoform in liver extracts (37, 38). Based on our observation that this isoform co-migrates with TorAp on SDS-PAGE gels (Fig. 7C), we propose that these species are in fact identical.
From a mechanistic perspective, the phenomenon of Torsin processing described here is reminiscent of regulatory proteolysis events leading to the activation of membrane-tethered transcription factors, as in the example of sterol regulation (39) or ATF6 processing in response to ER stress (40). We propose that Torsin processing is similarly mediated by a protease that is membrane-embedded or associated. This proposal is in agreement with the comparatively slow kinetics of Torsin processing (Fig. 1D), because low turnover numbers and incomplete conversion are often observed in intramembrane- or membrane-proximal proteolysis events (41), for example, the processing of SREBP-1 and ATF6 (42, 43).
However, Torsin processing is distinct from these events in that it relies on a conserved cysteine cluster (Fig. 5A) at the membrane interface that acts both as a regulatory switch and as a specific cleavage site. Our data are consistent with a model where at least two of the conserved N-terminal cysteine residues are engaged in an intra-chain disulfide bond, which is reduced in response to reducing agents, exposing a cleavage site that is specifically recognized by a protease that we designate Torsin converting protease (TorCP). Given that the cleavage product never acquires Endo H resistance (Figs. 2B and and66B), it seems reasonable to propose that this activity resides within the ER or the nuclear envelope, rather than the Golgi.
Having firmly established the phenomenon of Torsin processing, it will be critical to define which purpose this process serves. Given that various cell types from distinct animal species are endowed with the capability to mobilize TorsinA from the membrane (Figs. 1D, ,44E, ,66C, and and77C), and that the process can be observed in a physiologically relevant cellular differentiation process as exemplified by B cell maturation (Fig. 6C), we anticipate an important biological role that remains to be identified. It will thus be interesting to test if Torsin conversion is modulated in response to an acute infection. Of note, TorAp has an electrophoretic behavior indistinguishable from the TorA isoforms in liver (Fig. 7C)(37, 38), again suggesting in vivo relevance of our observations.
Because potent inducers of ER stress (with the exception of DTT) do not induce Torsin processing (Figs. 1A and and66A), we consider a major role in counteracting protein misfolding unlikely. Interestingly, several of the Torsins (Tor2A and Tor3A in humans and mice) do not harbor a hydrophobic domain, and TorA variants deprived of their hydrophobic domains retain ATPase activity (16, 17), suggesting that mobilized TorAp could have a function that is distinct from its precursor.
Through the identification of the specific protease inhibitor 5-nitroisatin as a lead compound (Fig. 3, B–D), our study establishes a foundation for functional studies in a variety of experimental settings via facile pharmacological interference. Indeed, 5-nitroisatin can be chemically diversified with ease, and many derivatives are commercially available, facilitating the development of more potent inhibitors or affinity reagents that could be used to identify the protease under question. In the future, the sequence information of the cleavage site that was borne out by this study can inform the design of peptide-based competitive or suicide inhibitors, a strategy that is of proven efficacy in identifying unknown proteolytic activities (44, 45).
C. Z., R. S. H. B., C. C. A. H., and C. S. designed the research; C. Z., R. S. H. B., C. H. T., C. C. A. H., and C. S. performed the research; C. Z., R. S. H. B., C. C. A. H., and C. S. the analyzed data; and C. Z., R. S. H. B., C. C. A. H., and C. S. wrote the paper.
We thank members of the Schlieker laboratory for comments on the manuscript and Janie Merkel (Yale Center for Molecular Discovery) for providing the protease inhibitor library.
*This work was supported, in whole or in part, by National Institutes of Health Grants DP2 OD008624-01 (to C. S.) and (R01CA163910 (to C. C. A. H.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors declare that they have no conflicts of interest with the contents of this article.
This article contains supplemental Table S1.
3The abbreviations used are: