A ribosome-associating factor chaperones tail-anchored membrane proteins.

Hundreds of proteins are inserted post-translationally into the endoplasmic reticulum (ER) membrane by a single carboxy-terminal transmembrane domain (TMD). During targeting through the cytosol, the hydrophobic TMD of these tail-anchored (TA) proteins requires constant chaperoning to prevent aggregation or inappropriate interactions. A central component of this targeting system is TRC40, a conserved cytosolic factor that recognizes the TMD of TA proteins and delivers them to the ER for insertion. The mechanism that permits TRC40 to find and capture its TA protein cargos effectively in a highly crowded cytosol is unknown. Here we identify a conserved three-protein complex composed of Bat3, TRC35 and Ubl4A that facilitates TA protein capture by TRC40. This Bat3 complex is recruited to ribosomes synthesizing membrane proteins, interacts with the TMDs of newly released TA proteins, and transfers them to TRC40 for targeting. Depletion of the Bat3 complex allows non-TRC40 factors to compete for TA proteins, explaining their mislocalization in the analogous yeast deletion strains. Thus, the Bat3 complex acts as a TMD-selective chaperone that effectively channels TA proteins to the TRC40 insertion pathway.

The final step in TA protein targeting to the mammalian ER is mediated by TRC40 (also called Asna1)2,4. This factor interacts directly with TA substrates via their TMD, and upon targeting to the ER membrane via putative receptor(s), releases the TA protein in an ATPase-dependent manner for insertion into the lipid bilayer. The yeast homolog of TRC40, termed Get3, plays an analogous function in conjunction with its ER receptors Get1 and Get2 (3), and still poorly understood cytosolic factors Get4 and Get5 (5-10). In both cases, TRC40/Get3 must selectively and efficiently capture TA proteins via recognition of a hydrophobic TMD that is prone to aggregation, inappropriate interactions, or sequestration by other chaperones.

To investigate how TRC40 efficiently captures its TA protein cargo, we reconstituted this event in vitro and analyzed its requirements (Sup. Fig. S1). Sucrose gradient purified ribosome-nascent chains (RNCs) containing a tRNA-tethered TA protein (Sec61β) were released with puromycin in the presence of cytosolic fractions. Capture of the radiolabeled Sec61β by TRC40 was assayed by crosslinking. As expected, Sec61β released into a complete cytosol was captured by TRC40 in a TMD-dependent and energy-stimulated manner. Surprisingly, TRC40 in size fractionated cytosol was unable to capture Sec61β, which instead crosslinked to an unidentified protein we provisionally term p38 (Sup. Fig. S2). TRC40 capture could be restored by adding back one of the missing fractions. Further fractionation confirmed this ‘capture-stimulating’ factor is distinct from TRC40 (Sup. Fig. S3). Thus, efficient substrate capture by TRC40 in a complex environment requires an additional protein factor, the absence of which leads to substrate interactions with non-TRC40 proteins (e.g., p38) despite the availability of TRC40.

We postulated the capture-stimulating factor could be an intermediary that delivers substrates to TRC40. Such a factor should recognize TA substrates in a TMD-dependent manner, be found in fractions that contain ‘capture-stimulating’ activity, and be more easily observed upon TRC40 saturation. Crosslinking analyses of Sec61β overproduced in complete cytosol revealed two interacting partners of ~120 kD and 35 kD that met these criteria (Sup. Fig. S4). Our earlier affinity purification of Sec61β from large-scale translation reactions2 contained an associated product of ~120 kD (in addition to the major TRC40 band) that we identified by mass spectrometry as Bat3 (also known as Scythe/Bag6) (Sup. Fig. S5). Bat3 affinity purified from reticulocyte lysate co-purified with proteins of ~18 kD and 35 kD (Fig. 1A) that we identified by mass spectrometry as Ubl4A and C7ORF20, for which we propose the name TRC35 (Sup. Fig. S5). Antibodies against either Bat3, Ubl4A or TRC35 could co-immunoprecipitate the other two components, and affinity-depletion of any one of these components resulted in substantial depletion of the other two proteins (Sup. Fig. S6). Thus, Bat3 is part of a stable heterotrimeric complex in the cytosol.

Fig. 1

Identification of a TMD-interacting protein complex

a, Cytosolic proteins bound and eluted from anti-Bat3 or anti-GFP (control) affinity columns are shown. HC and LC are IgG heavy and light chain. b, Sec61β (WT), a deletion construct lacking its TMD (ΔTMD), or the insertion-deficient 3R mutant were translated in reticulocyte lysate, affinity purified on an anti-Sec61β column, and immunoblotted for the indicated products. Total lysate was included for comparison. An autoradiograph of the blot revealed equal recovery of the three translated substrates. c, Crosslinking products of Sec61β (from pooled sucrose gradient fractions 6-9 in Sup. Fig. S4) were immunoprecipitated under denaturing or native conditions. Non-immune serum (N.I.S) was included as a control. d, Versions of Sec61β containing the TMD from the indicated proteins were analyzed for interaction with Bat3 and TRC40 by in vitro translation, crosslinking, and immunoprecipitation. An aliquot of the total translation reaction is shown for each substrate (‘subst.’), as well as the immunoprecipitation products of the crosslinking reactions. e, The TMD of Sec61β was mutated to change its hydrophobicity as indicated. f, Each construct was analyzed for its interactions with Bat3 complex and TRC40 as in panel b. An aliquot of the total translation product was analyzed by autoradiography to visualize the substrates.

The TMD-dependent interaction of the Bat3 complex with TA proteins was verified by immunoblotting of Sec61β affinity purified from in vitro translation reactions (Fig. 1B). Furthermore, the TMD-dependent ~120 kD and 35 kD crosslinks to Sec61β (Sup. Fig. S4) could be specifically immunoprecipitated under denaturing conditions with anti-Bat3 and anti-TRC35 antibodies, respectively, or under non-denaturing conditions by anti-Ubl4A antibodies (Fig. 1C). Analysis of additional TA protein TMDs by crosslinking showed that both TRC40 and the Bat3 complex interacted with the ER-targeted VAMP2 and Sed5 TMDs, while neither interacted with the TMDs from spontaneously inserting cytochrome b5 (Cb5) or mitochondrially-targeted Fis1 or F1L (Fig. 1D). Differences in TMD hydrophobicity (either length or maximal hydrophobicity) among these substrates appeared to be a key determinant, since altering Sec61β TMD hydrophobicity to match that of Cb5, Fis1, or F1L markedly reduced interaction with both Bat3 complex and TRC40 (Fig. 1E,F; Sup. Fig. S7). Thus, the Bat3 complex interacts directly with the TMDs of several ER-targeted TA proteins with similar substrate-specificity as TRC40.

We surmised that the Bat3 complex was responsible for the stimulatory activity needed by TRC40 for efficient substrate capture. In cytosolic fractions immunodepleted (by ~95%) of Bat3 (which also depletes Ubl4A and TRC35 without affecting TRC40 levels; Fig. 2A and Sup. Fig. S6), Sec61β capture by TRC40 is diminished, with a concomitant increase in capture by p38 (Fig. 2B); a similar effect is seen after immunodepletion of TRC40, where p38 was the primary soluble interaction partner. Replenishment of Bat3 depleted extracts to the original level with affinity purified Bat3 complex [prepared using an anti-TRC35 antibody (Sup. Fig. S8)] fully restored TRC40 capture activity (Fig. 2B). The effect of Bat3 depletion on TRC40 substrate capture was not simply a consequence of the RNC-release assay since translation of full-length TA substrate in Bat3-depleted reticulocyte lysate also showed diminished TRC40 interactions with increased p38 interaction (Sup. Fig. S9). Thus, maximally efficient capture by TRC40 of TA proteins upon their release from the ribosome requires the Bat3 complex. These findings are consistent with recent observations suggesting that depletion of Bat3 can impair insertion of a TRC40-dependent TA protein11.

Fig. 2

The Bat3 complex mediates substrate capture by TRC40

a, Translation extracts were passed over anti-GFP (control), anti-Bat3, or anti-TRC40 affinity resins and different amounts of each depleted lysate analyzed by immunoblotting. b, Substrate capture assay using either total cytosol, or cytosol immunodepleted (‘Δ’) with the indicated affinity resins. In this assay (see Sup. Fig. S1), radiolabeled Sec61β RNCs are released with puromycin and capture by TRC40 is assessed by crosslinking. The portion of gel showing the TRC40-Sec61β crosslink is shown. Failure of TRC40 to capture substrate typically results in capture by a 38 kD protein (p38). The ‘addback’ lanes are Bat3-depleted cytosol replenished with affinity purified Bat3 complex (prepared using an anti-TRC35 resin; Sup. Fig. S8) at three concentrations spanning that present in the original cytosol. The ‘mock’ addback sample was prepared in parallel, but employed an irrelevant affinity resin (anti-GFP) in lieu of TRC35 affinity resin (Sup. Fig. S8). A reaction lacking crosslinker (XL) is shown in the first lane. Aliquots of each reaction (prior to crosslinking) were also analyzed by immunoblot against Bat3 and TRC40 to document their relative amounts in the reactions.

Although the biochemical functions of Bat3, Ubl4A, and TRC35 are poorly understood, the latter two proteins have recognizable homologues in yeast (Mdy2/TMA24/Get5 and Yor164C/Get4, respectively) that have been analysed by recent biochemical, structural, and genetic studies5-7,12,13. Get4 and Get5 form a stable complex that can interact with Get3, the yeast TRC40 homolog5,7-10,12,13. Deletion strains of Get4 and Get5 appear to phenocopy Get1, Get2, or Get3 deletions5-7, and all five genes cluster together in synthetic genetic interaction arrays5,7. These observations have implicated Get4 and Get5 in TA insertion at a step prior to Get3, but their functions have been unclear.

Given that Bat3 complex facilitates TRC40-substrate interactions, we hypothesized that Bat3 complex might capture TA substrates at the ribosome and transfer them to TRC40. Such a model requires the Bat3 complex to be at or near the ribosome. Indeed, affinity-purified RNCs containing a TA protein were markedly enriched in the Bat3 complex relative to empty ribosomes (Fig. 3A,B). Quantification showed that Bat3 complex and SRP occupy ~27% and ~7%, respectively, of TMD-containing RNCs (Fig. 3C). For comparison, SRP occupancy on a genuine SRP substrate isolated by the same method was ~30% (ref. 14; Sup. Fig. S10). Notably, Bat3 complex occupancies of <5% were observed for empty ribosomes and RNCs containing a mutant TMD, illustrating substrate selectivity for Bat3 complex recruitment. We confirmed that our affinity purified RNC preparations were tRNA-associated, as judged by their sensitivity to alkaline hydrolysis and selective precipitation with cetyl-trimethyl-ammonium-bromide (CTAB) (Sup. Fig. S11). Selective enrichment of Bat3 complex on TA protein RNCs was also seen using a magnetic bead ‘pull-up’ strategy that minimizes non-specific recovery of aggregates (Sup. Fig. S12). Thus, as suggested for SRP14, the Bat3 complex appears to be recruited to ribosomes when a functional TMD is inside the ribosome.

Fig. 3

Bat3 complex captures substrates on ribosomes for transfer to TRC40

a, Ribosomes were purified under native conditions from reticulocyte lysate and analyzed for TRC40, Bat3, Ubl4A, and SRP54 by immunoblotting. Approximately 30-50% of SRP, 2-5% of Bat3 complex, and undetectable (<1%) amounts of TRC40 are ribosome-bound. b, Affinity and size purified RNCs (from 60 ul translation reactions) of Sec61β, a TMD-lacking version (ΔTMD), and the 3R mutant were analyzed by immunoblotting for the indicated antigens. For comparison, 0.5 ul translation lysate was analyzed. L9 is a ribosomal protein. Autoradiography confirmed equal recovery of each substrate. c, The amounts of ribosomes, Ubl4A, and SRP54 were quantified in total lysate, purified empty ribosomes, or purified RNC preparations. Shown are the amounts of SRP54 and Ubl4A, normalized to 100 ribosomes, averaged from multiple independent purifications (n=3 for empty ribosomes, n=7 for Sec61β-RNCs, and n=6 from β(3R)-RNCs). d, Translations of Sec61β in TRC40-depleted lysates lacking or replenished with recombinant zebrafish TRC40 were rapidly diluted, crosslinked, and separated by centrifugation into soluble and ribosome fractions. Sec61β, Bat3 crosslinks (immunopreciptiated with anti-Bat3), and TRC40 crosslinks are shown in the total (T), soluble (S), and ribosome (P) fractions. e, Cytosolic fractions from HT1080 cells lacking or stably over-expressing HA-tagged human TRC40 were bound to anti-HA resin and selectively eluted with TEV protease (which cleaves between the HA tag and TRC40). The eluted products, along with starting lysates, were analyzed by immunoblotting for Bat3, TRC40, or SGTA (a control protein). Note that endogenous TRC40 is present, but not visible at this exposure of the blot. Identical results were obtained when elution of the affinity column was with HA peptide instead of TEV protease (data not shown).

Analysis of various TMDs for their capacity to recruit Bat3 complex and SRP to the ribosome from inside the tunnel showed a general correlation with hydrophobicity, with bona fide ER-targeted TMDs (from Sec61β, VAMP2, and Sed5) showing the highest recruitment (Sup. Fig. S10). However, a strict concordance seems unlikely since a moderately hydrophobic Sec61β mutant (3A3G) that fails to interact with Bat3 complex post-translationally (Fig. 1F) nonetheless mediated its recruitment from inside the ribosome (Sup. Fig. S10). Interestingly, Bat3 complex did not crosslink to ribosome-tethered substrate (Sup. Fig. S13), but only upon release with puromycin or translational termination. This is opposite to SRP, which contacts substrates on the ribosome, but not after release. Thus, ribosomes synthesizing TMD-containing proteins can recruit both SRP and Bat3 complex at a stage before TMD emergence into the cytosol. Both of these complexes stay on the ribosome during further translation (Sup. Fig. S10), but upon substrate release, only the Bat3 complex remains associated in a TMD-dependent manner. Because changes in SRP recruitment do not influence Bat3 recovery on RNCs, the two factors may not compete with each other. However, this remains to be determined, as the binding site for Bat3 complex is not known.

Selective recruitment of the Bat3 complex to ribosomes before TA protein release suggested this may be the site of initial substrate capture. We therefore looked for a substrate-Bat3-ribosome intermediate upon TRC40 depletion. Sec61β translated in a TRC40-depleted lysate was ~70% ribosome-associated, where it could be crosslinked to Bat3 (Fig. 3D). When recombinant TRC40 was included in the depleted translation extract, the proportion of Sec61β in the ribosome fraction was decreased to ~30%, crosslinking to Bat3 was lost, and Sec61β in the soluble fraction was crosslinked to TRC40. Non-ribosomal Sec61β-Bat3 complexes can also transfer substrate to TRC40, since incubation of this fraction with TRC40-containing fractions resulted in a decrease in Bat3 crosslinks and concomitant increase in TRC40 crosslinks (data not shown). Thus, Bat3 complex is recruited to the ribosome, where it can interact with TA substrates upon their translational termination. This putative Bat3-substrate intermediate, whether on the ribosome or free in solution, is then converted into the productive Sec61β-TRC40 targeting complex. Consistent with this model, Bat3 and TRC40 can interact as determined by their co-immunoprecipitation under detergent free conditions (Fig. 3E).

Bat3 recruitment to RNCs is restricted to the period between synthesis of the TMD and release from the ribosome (Fig. 4A), a time window that seems exceedingly brief. To explain this conundrum, we measured the recruitment window by performing a time course of TA protein synthesis combined with selective precipitation of tRNA-associated polypeptides with CTAB. Nascent chains available for Bat3 complex recruitment would be nearly full-length, but still contain a covalent tRNA. After validating that such tRNA-associated full-length species can be selectively recovered (Sup. Fig. S14), we quantified the amount of this population during a time course of Sec61β synthesis. Five minutes into this time course (~2 min after full-length Sec61β chains are first detected), a remarkable 50% of full-length Sec61β was precipitated by CTAB (Fig. 4B). This proportion diminished over time as translational termination ensued and completed Sec61β chains accumulated. Comparison of experimental data to theoretical expectations (Sup. Fig. S15) estimate a t1/2 for Sec61β termination of ~ 1 min.

Fig. 4

Translation termination is delayed for a TA protein

a, Schematic of Sec61β synthesis illustrating that nascent chains competent for recruitment of Bat3 complex would be near full-length and contain a covalently associated tRNA (red star). Such recruitment-competent polypeptides should migrate on gels close to the full-length size, but be precipitated by CTAB, which selectively precipitates tRNA-associated proteins. b, Time course of Sec61β and Sec61β(ΔTMD) (both tagged at the C-terminus with the same 12-residue epitope tag) synthesized at 25°C in reticulocyte lysate. At each time point, samples were analyzed directly or after CTAB precipitation. The proportion of total full-length polypeptide that was CTAB precipitated is plotted. The dashed lines indicate theoretical expectations (Sup. Fig. S15) for a t1/2 of translational termination of 15 and 60 sec. c, Model for shuttling of TA proteins from the ribosome to TRC40 via a pre-targeting intermediate involving recruitment of the Bat3 complex to the ribosome.

Remarkably, Sec61β-ΔTMD (lacking the TMD) showed substantially less tRNA-associated polypeptide at each time point (Fig. 4B) corresponding to a termination t1/2 of ~ 15 sec. Importantly, the last 12 codons (containing an epitope tag) of both constructs are identical, arguing against differences in residues close to the peptidyl transferase centre as the basis for differences in termination rate. The translation elongation rate (~1.5-1.8 sec per residue) and the ~10-fold slower termination rate (relative to elongation) we measured for our control substrate (Sec61β-ΔTMD) are consistent with earlier estimates using this same in vitro system15. However, the unexpectedly long termination rate of Sec61β suggests that the TMD slows this reaction. The magnitude of termination delay seen for Sec61β (~1 min) is comparable to the elongation ‘arrest’ mediated by SRP measured under similar conditions in this same system16. This provides a plausible explanation for how Bat3 complex could have sufficient time to be recruited to RNCs containing a TMD inside the tunnel. The sequence parameters that modulate termination remain to be established, but is unlikely to be simple hydrophobicity given earlier studies showing similar termination delays by relatively hydrophilic sequences17.

Our findings explain how the hydrophobic TMDs of TA proteins are safely shielded from inappropriate interactions and aggregation during their delivery to the ER membrane (Fig. 4C). Our working model posits that the TMD is shielded throughout targeting, first by the ribosome, then by the Bat3 complex, and finally by TRC40. The sequential handoffs of substrate between these complexes are likely to be tightly regulated to prevent exposure of the TMD to aqueous solution. Perhaps the Bat3 complex regulates the nucleotide cycle and/or conformation of TRC40 to facilitate its loading with substrate. Based on recent crystal structures of Get3 (18-22), efficient substrate loading likely requires a nucleotide-dependent transition from an ‘open’ to ‘closed’ conformation18,19 that is perhaps aided by the Bat3 complex. In the absence of this activity, capture by TRC40 would be slower, explaining why it cannot operate efficiently in the presence of competing cytosolic factors, but does manage in a purified system (data not shown).

A role for the Bat3 complex in TA protein targeting may therefore not be absolutely essential23, but would increase fidelity and efficiency. Such a function would seem to be highly conserved, since yeast have a homologous complex (Get4 and Get5) that shows genetic, physical, and functional interactions with Get35-10,12,13, and whose absence can lead to aggregation and partial mislocalization of TA proteins5-7. Thus, the Bat3 complex appears to represent a conserved TMD-selective chaperone that acts at the ribosome. Where on the ribosome the Bat3 complex binds, how it is recruited when a TMD is inside the ribosomal tunnel, and how its function is coordinated with several other ribosome-associating factors including SRP24, NAC25, and RAC26, remain important questions for future studies27.

Methods Summary

Reagents and standard methods

Constructs, antibodies, and recombinant proteins derived from earlier studies2,18 are fully described in Supplementary Methods. Antibodies to Bat3 and Ubl4A were produced in rabbits immunized with recombinant Bat3 fragment (residues 1-250) or full-length Ubl4A, respectively. TRC35 antibodies were against a synthetic peptide (NH2-CGSPIELD-COOH) conjugated to KLH. In vitro transcription, translation in reticulocyte lysate, sucrose gradient fractionation, chemical crosslinking, and immunoprecipitation was as before2. RNCs were generated by transcription and translation of PCR products lacking a stop codon and sucrose gradient purified for use in the capture assay. CTAB precipitation of tRNA-associated polypeptides has been described28.

RNC release assay

[35S]-labeled RNCs were incubated with the indicated additions for 15 min at 32°C. Standard assay conditions were 100 mM KAc, 20 mM Hepes, pH 7.4, 7 mM MgAc2. Energy regenerating system consisted of 1 mM ATP, 1 mM GTP, 10 mM Creatine phosphate, 40 ug/ml creatine kinase. The base cytosolic fraction was prepared from reticulocyte lysate that was bound and eluted from DEAE-sepharose to remove hemoglobin and concentrated to half the original volume. Immunodepletions from this lysate employed immobilized antibodies. Following RNC release, crosslinking was with 250 uM bismaleimidohexane (BMH, a homo-bifunctional cysteine-reactive crosslinker) on ice for 30 min and quenched with 25 mM 2-mercaptoethanol.

Bat3 complex purification

Reticulocyte lysate was bound to phenyl-sepharose, washed in 50 mM Hepes pH 7.4, and eluted with 1% Triton X-100, 50 mM Hepes, pH 7.4. The eluate was bound and acid-eluted from anti-Bat3 affinity resin to identify interacting proteins by mass spectrometry. Functional Bat3 complex was isolated from the phenyl sepharose fraction using anti-TRC35 resin, eluted with immunizing peptide, and concentrated by binding and elution from S-sepharose.

RNC purification

Affinity purification of size-fractionated RNCs was essentially as described24, but with an antibody against the N-terminus of Sec61β and elution with immunizing peptide. The final RNCs were sedimented to remove any spontaneously released polypeptides.