Mammalian SRP receptor switches the Sec61 translocase from Sec62 to SRP-dependent translocation

Two distinct pathways deliver secretory proteins to the Sec61 protein translocase in the endoplasmic reticulum membrane. The canonical pathway requires the signal recognition particle (SRP) and its cognate receptor (SR), and targets ribosome-associated proteins to the Sec translocase. The SRP-independent pathway requires the Sec translocase-associated ER membrane protein Sec62 and can be uncoupled from translation. Here we show that SR switches translocons to SRP-dependent translocation by displacing Sec62. This activity localizes to the charged linker region between the longin and GTPase domains of SRα. Using truncation variants, crosslinking and translocation assays reveals two elements with distinct functions as follows: one rearranges the translocon, displacing Sec62 from Sec61. A second promotes ribosome binding and is conserved between all eukaryotes. These specific regions in SRα reprogramme the Sec translocon and facilitate recruitment of ribosome-nascent chain complexes. Overall, our study identifies an important function of SR, which mechanistically links two seemingly independent modes of translocation.

P roteins destined for secretion from the cell typically possess an N-terminal hydrophobic signal sequence, which targets them to the protein-translocating channel at the endoplasmic reticulum (ER) membrane 1 . Once the protein has been translocated across the membrane, the signal sequence is usually cleaved off and the mature protein can then fold and move through the secretory pathway via vesicular trafficking. Several distinct targeting pathways deliver proteins to the proteinconducting channel, which is formed by the heterotrimeric Sec61 complex (Sec61abg) 2 . Perhaps the best-characterized pathway is the canonical signal recognition particle (SRP) pathway, which is conserved in all domains of life 3,4 . SRP is a ribonucleoprotein complex, which binds to ribosomes and can recognize the signal sequence as it emerges from the ribosome exit tunnel 5,6 . Recognition of the signal sequence by the SRP54 component of SRP leads to a transient retardation in translation concomitant with targeting to the ER membrane via an interaction with the cognate SRP receptor (SR) [7][8][9][10] . SR mediates the transfer of the ribosome, together with the nascent chain, to the Sec61 translocase 11 . Translation resumes and the nascent chain is fed directly from the ribosome into the pore of the Sec61 translocon, which is co-aligned with the ribosome exit tunnel 12 . SRP then can be released from its receptor such that both can now participate in further rounds of targeting 13 .
In eukaryotes, SR is composed of two subunits, the 70 kDa peripheral membrane protein SRa (ref. 14), which is anchored to the membrane by the 30 kDa SRb, which has a single N-terminal transmembrane (TM) domain and short luminal domain 15 . Both SRa and SRb are GTPases; SRa has a bacterial homologue FtsY, and both of them share a characteristic GTPase domain (NG domain), which is also found in SRP54 and its bacterial homologue Ffh, as well as in a third member of the SRP GTPases, FlhF 16,17 . The NG domains of SRa and SRP54 interact in a GTP-dependent manner. Complex formation is kinetically accelerated by the presence of the SRP RNA and by the binding of a bona fide signal sequence 18,19 . Studies with the bacterial SRP-SR complex reveal that its subsequent interaction with the translocon (SecYEG) induces molecular rearrangements, which stimulate GTP hydrolysis in both GTPases leading to the release of signal sequence from SRP and its transfer to the translocon 20 .
SRa is bound to SRb by its SRX domain, which possess a longin domain fold 21,22 . The NG and SRX domains are connected by a flexible linker, rich in charged residues, whose function is poorly characterized. FtsY lacks such linker and SRX domains, and instead possesses a natively unfolded A-domain, which is important for membrane binding via two lipid-binding motifs [23][24][25] . SRb, in contrast, is closely related to the ARF and Sar1 family of GTPases 15,26 . However, details of the SRb GTPase cycle are not well understood.
Not all secretory proteins use the SRP targeting pathway; characterized first in the yeast Saccharomyces cerevisiae, an SRP-independent translocation pathway was identified that, instead of SRP, absolutely requires the ER membrane protein Sec62 (ref. 27). Sec62 associates with the same core Sec61 channel used by the SRP pathway, forming a larger Sec translocase, alongside Sec63, Sec71 and Sec72 (ref. 28). Furthermore, there is no obligate coupling of translation and targeting with this pathway, unlike with SRP-dependent targeting. Substrates can be released from the ribosome and maintained in an unfolded conformation by chaperones of the heat-shock protein 70 (Hsp70) family 29,30 . They can then interact with the Sec translocase, which again forms the protein-conducting channel. In this case, the action of the ER Hsp70s (Kar2 and Lhs1), which are recruited to the luminal face of the Sec translocase by Sec63, drive the energetics of the translocation reaction 31,32 .
The pathway a particular yeast precursor will take is largely determined by the signal sequence, with more hydrophobic signal sequences being targeted via SRP and less hydrophobic sequences translocated via the Sec62 pathway 33 . The Sec62 pathway is, however, not restricted to yeast; homologues of Sec62 and Sec63, but not the non-essential Sec71 or 72, are also present in higher eukaryotes and furthermore can complement deletions of their homologues in yeast; suggesting that they perform similar functions [34][35][36] .
Recently, the first substrates for the mammalian Sec62 pathway were identified and they are typically short secretory proteins including insulin, apelin and statherin [37][38][39] , as well as the insect protein preprocecropin, which was known for a long time to translocate independently of SRP 40 . The fact that these precursors are very short makes them poor substrates for SRP, as they are likely released from the ribosome before SRP has had a chance to bind to their signal sequence efficiently. Before binding to the translocon, some of these short secretory proteins can also interact with upstream cytosolic factors, including calmodulin and components of GET-targeting pathway, which typically delivers C-terminally anchored membrane proteins to the ER 39,41,42 .
Both the SRP-and Sec62-dependent pathways converge at the Sec61 translocase, but require different accessory factors to associate with Sec61. SR and ribosomes are required in the SRP pathway, and Sec62 in the SRP-independent Sec62 pathway. Evidence suggests that Sec63 is likely to be involved with both pathways 43,44 . At present, little is known about the dynamics and interconvertability of these complexes.
Here we have explored the interaction of both Sec62 and SR with the mammalian translocon and find that interaction of ribosomes and Sec62 with the core Sec61 translocon is mutually exclusive. Furthermore, we identify a specific role for SR, involving the SRa linker domain, in displacement of Sec62, which provides a mechanism for the interconversion of translocons.

Results
Sec62 interacts with Sec61 in a ribosome-sensitive manner. We first compared the interaction profile of Sec61 with translocon-associated proteins in intact canine pancreatic rough microsomes (RMs), where 480% of translocons are engaged with ribosomes 45 , and microsomes stripped of ribosomes by puromycin-high salt treatment (PKRMs). The compound bismaleimidohexane (BMH), a cysteine-reactive homobifunctional reagent was used to monitor crosslinking from the single endogenous cysteine in Sec61b (Fig. 1a). As reported previously 46 , the Sec61b cross-link profile was strikingly different for these two preparations. In particular, crosslinking between Sec61b and Sec62 was very weak in the RM, but dramatically increased in the PKRM.
We repeated this experiment with membranes stripped of ribosomes with EDTA and high salt (EKRM) and monitored crosslinking of Sec62 by western blot. As observed with PKRMs, there was a dramatic enhancement in Sec62-Sec61b crosslinking with EKRMs compared with RMs (Fig. 1b). In fact more than 50% of Sec62 could be cross-linked to Sec61b with EKRMs (Fig. 1b). To test whether the loss of crosslinking between these two components was specifically due to the removal of ribosomes from the membrane, purified canine salt-washed ribosomes were allowed to rebind to the EKRMs and the cross-link assay was repeated (Fig. 1b). This led to a dose-dependent reduction in crosslinking between Sec62 and Sec61b. Hence, as reported previously 34 , the Sec62-Sec61b interaction is strongly dependent on the absence of ribosomes.
Binding of SR to Sec61 positions SRa close to Sec61b. To explore the interaction of SRP receptor with the translocon, we made use of a recombinant form of the SRP receptor, which lacks the N-terminal luminal and TM domains of SRb (Fig. 2a), and which is functionally active in co-translational translocation 47 . In canine microsomes, SR is considerably sub-stoichiometric to Sec61 complex 45 , hence any SR-binding sites are unlikely to be saturated. Again, monitoring crosslinking from the single cysteine in Sec61b, we assessed the effect of adding recombinant SR on the environment of the Sec61 complex (Fig. 2b, lane 7). In the case of RMs, SRa/bDN had little effect on the Sec61b crosslinking profile. In contrast, with EKRMs we observed a rather dramatic alteration in the cross-link profile on addition of SR. In particular, we observed a strong increase in a B80 kDa cross-link species, in good agreement with the expected size of an SRa Â Sec61b adduct (Fig. 2b). This 80 kDa band was also observed with RM but to a much lesser extent. To assess whether the 80 kDa cross-link species observed in the presence of SRa/bDN was indeed an SRa Â Sec61b cross-link, we performed crosslinking and then reisolated recombinant SRa via the histidine (His)-tag using Ni-NTA resin, following membrane solubilisation (Fig. 2c). The 80 kDa cross-link species and the weaker band above this bound to the Ni-NTA resin, unlike Sec61b and all the other Sec61b cross-link products. The presence of the double band likely arises due to crosslinking of the single cysteine of Sec61b to one or the other of two different cysteines within SRa resulting in products with different electrophoretic mobility.
We also analysed cross-linked samples by western blot with Sec61b and SRa antiserum in parallel ( Supplementary Fig. 1a). This clearly revealed that the strong 80 kDa species also cross-reacts with SRa antisera.
To test whether endogenous SR, that is, with the SRb TM domain intact, could also be cross-linked to Sec61b, we treated EKRM with BMH and performed denaturing immunoprecipitation with anti-SRa antiserum (Fig. 2d). Again, an 80 kDa Sec61b cross-link species was detected and could be specifically immunoprecipitated with SRa antibodies but not with an unrelated antiserum.
Finally, we tested whether endogenous purified SR and Sec61 complex could also interact. Sec61 complex purified from pancreatic microsomes and reconstituted alone in liposomes and treated with BMH only revealed Sec61b cross-links to Sec61a, Sec61g and homodimers of Sec61b (Fig. 2e). In contrast, proteoliposomes that also contained purified endogenous SR and signal peptidase complex (SPC) revealed an additional albeit weak 80 kDa Sec61b cross-link species in excellent agreement with the cross-link between SRa and Sec61 observed with recombinant SR and in EKRM.
SRP receptor blocks Sec61b Â Sec62 crosslinking. In contrast to the Sec61b Â SRa cross-link species, which increased in abundance when SR was added to EKRMs, most other Sec61bderived cross-linked adducts were reduced, in particular, the cross-link between Sec61b and Sec62 (Fig. 2b). Moreover, this effect could be confirmed by pre-incubating EKRMs with increasing concentrations of SR, which revealed a dose-dependent inhibition in crosslinking between Sec62 and Sec61b (Fig. 3a). When this experiment was repeated using a mutant SRP receptor, which contains only the SRX domain of SRa (SRa 126 /bDN, Fig. 2a) 5 , a much weaker reduction in Sec61b-Sec62 crosslinking was observed. This suggests that either the linker and/or NG regions are important for the inhibitory effect observed (Figs 2b and 3b). Similar results were obtained using the bifunctional reagent bismaleimidoethane, which has a shorter spacer arm between the reactive maleimides than BMH (10 and 16 Å, respectively; Fig. 3b).
As both SRa and SRb are GTPases, we also assessed whether the inhibition of the Sec61-Sec62 interaction was dependent on guanine nucleotides. However, pre-incubation of SR and EKRMs with either GDP or the non-hydrolyzable analogue GppNHp (5 0 -guanylyl imidodiphosphate) had no effect on the SR-induced reduction of Sec61 Â Sec62 crosslinking (Fig. 3c). Consistent with this result, crosslinking between HisSRa and Sec61b was also insensitive to guanine nucleotides ( Supplementary Fig. 1b). nucleotides, led us to investigate whether the SRa linker domain, rather than the GTP-binding NG domain, might be responsible. Comparison of the linker region from different eukaryotes reveals two distinct charged regions, which we term CBR and RBR (Fig. 2a, Supplementary Fig. 2). A series of mutants were generated lacking either just the first one (DCBR), both charged domains (DCBRDRBR) or the entire linker domain (Dlinker), but retaining the SRX and NG domains intact (Fig. 2a).
When the cross-link assay was repeated with these mutants, in contrast to the full-length SR, none of them could reduce the Sec61b Â Sec62 product (Fig. 3d). These results strongly suggest that the linker domain is important for this effect. When the construct lacking the entire linker domain (residues 126-315) was used (SRa Dlinker ), an additional cross-link adduct of 92 kDa was present corresponding precisely to the size of Sec62 plus SRa Dlinker . . Samples were then solubilized with Triton X-100 and high salt and then incubated with Ni-NTA beads to recover hisSRa together with any cross-link products that also contain hisSRa. Samples were then analysed by SDS-PAGE and western blot with Sec61b antibodies. (d) EKRM were treated with BMH (40 mM) before denaturing immuno-precipitation with either anti-SRa or a nonrelated antiserum (n.r. ). An aliquot of the total reaction (5%) along with the immunoprecipitated material was analysed by SDS-PAGE and western blot with Sec61b antibodies. Positions of the cross-link products and the IgG heavy chains (h.c.) are indicated. (e) Proteoliposomes reconstituted with purified Sec61 alone or together with purified endogenous SR and signal peptidase complex (SPC) were treated with BMH (10 mM) and analysed as in a.
To test whether the CBR region is sufficient for this effect, we performed the same experiment with a construct where the entire linker is replaced with just the CBR (Dlinker þ CBR) (Fig. 3d). Indeed this construct was also able to reduce crosslinking between Sec61b and Sec62. Our data show that the first charged region is required for Sec62 displacement.
We also monitored membrane binding of the different SR linker mutants using EKRM in a floatation assay ( Supplementary  Fig. 3). Constructs with CBR (that is, full-length SR and Dlinker þ CBR) showed the strongest membrane binding, consistent with the role of this domain in interacting with the translocon. Although the other mutants showed lower membrane binding, all of them could still interact with the membrane. In the case of the Dlinker construct, this is in good agreement with the observed cross-link between SRa Dlinker and Sec62. This indicates that the inability of mutants lacking the CBR domain to perturb the Sec62-Sec61 interaction is not simply due to a loss of membrane or translocon interaction.

SRa linker domain is important for ribosome binding of SR.
The linker domain is also known to be required for ribosome binding of mammalian SR 47,48 . We made use of the linker mutants to assess whether this property of SR might also depend on a specific region of the linker domain. Using a sedimentation assay with salt-washed canine pancreas ribosomes, binding of full-length SR to ribosomes could be observed as reported previously (Fig. 4a) 47,48 . Removal of both charged regions (DCBRDRBR) or the entire linker (Dlinker) abolished ribosome binding; however, deletion of only the N-terminal region (DCBR) led to increased ribosome association. Deletion of just the RBR domain strongly reduced ribosome binding. In contrast, a b c GppNHp While the ability of SR to bind ribosomes appears linked to the RBR region of the linker domain, this contrasts to the effect of SR on Sec61b Â Sec62 crosslinking, which is closely linked to the CBR region (Fig. 3d). Hence the linker domain appears to be involved in both binding the ribosome and destabilizing the Sec62-Sec61 interaction. However, these two functions require distinct regions of the linker domain.
SR ribosome binding is evolutionarily conserved. Having observed that the linker domain harbours two distinct charged regions, we used multiple sequence alignments to further dissect these elements (Supplementary Fig. 2a). The CBR, which is involved in Sec62 displacement, is more specific to animals, while the RBR is evolutionarily well-conserved between fungi, plants and animals. As the latter was implicated in ribosome binding, the role of this region was further assessed using a fungal SR. Analogous linker domain mutations were generated in SRa from the thermophilic fungus Chaetomium thermophilum (Supplementary Fig. 2b) and their ability to bind to Chaetomium 80S ribosomes was tested (Fig. 4b). Compared with human SRa, full-length CtSRa is more stable in the absence of SRb and allowed us to monitor ribosome binding independent of SRb. Moreover, we recently showed that while CtSRa/SRbDN can bind ribosomes, both CtSRbDN alone and the minimal CtSRa 138 / SRbDN complex are unable to bind ribosomes 26 . Full-length CtSRa alone readily bound to the ribosomes in a sedimentation assay (Fig. 4b), as well as to canine ribosomes ( Supplementary   Fig. 4b). Deletion of the non-conserved linker region (DCBR) had very little effect on ribosome binding. In contrast, deletion of the conserved RBR region as well almost completely abolished ribosome binding (DCBRDRBR). Replacing the entire linker with just the RBR domain completely restored ribosome binding (Dlinker þ RBR). Hence the role of the RBR is conserved between yeast and mammals.
Unlike the ribosome-binding activity of SRa, CtSRa was unable to destabilize the interaction between Sec61b and Sec62 ( Supplementary Fig. 4c), suggesting that this interaction is specific to higher eukaryotes. Interestingly, the CtSRa NG domain alone could also be cross-linked to Sec62, as was observed with the human SRa Dlinker /bDN construct ( Supplementary Fig. 4c), indicating that the NG-translocon interaction is conserved.

SR inhibits translocation of Sec62-dependent substrates.
Having observed that human SR can perturb the association between mammalian Sec62 and Sec61, the functional effect of SR on the translocation of Sec62-dependent substrates was tested. Insect preprocecropin A has long been known to translocate in an SRP-independent manner and was recently shown to instead require Sec62 (refs 37,49). Similarly, the small mammalian secretory proteins apelin and statherin also translocate in a Sec62-dependent manner 49 . Apelin, statherin and preprocecropin A derivatives, possessing a N-glycosylation opsin tag to monitor ER translocation, were synthesised in vitro in rabbit reticulocyte lysate in the presence of labelled methionine. To exclude the possibility of any co-translational translocation, the translation reaction was treated with puromycin before the addition of RMs in  the presence or absence of SRa/bDN. Precursor translocation was then monitored via N-linked glycosylation. All three substrates showed a marked decrease in translocation in the presence of SRa/ bDN (Fig. 5a). In contrast, the C terminus of the tail-anchored membrane protein cytochrome B5-opsin construct, which can insert into the ER membrane spontaneously 50 , was N-glycosylated equally efficiently in the presence or the absence of SR, indicating that the loss of N-linked glycosylation of the short secretory proteins is not due to an indirect effect of SR on oligosaccharyltransferase activity. To further investigate the specificity of this perturbation, we compared the inhibitory effect of the different SR linker deletion constructs on preprocecropin translocation. We performed the analysis in a co-translational reaction so that we could monitor effects on the SRP-dependent translocation of preprolactin in parallel (Fig. 5b,c). Control reactions performed in the absence of membranes or in the presence of membranes followed by treatment with Endo H confirmed the identity of the translocated species ( Supplementary Fig. 5).
As with the post-translational assay, full-length SR led to a significant inhibition of preprocecropin translocation. In contrast, the DCBR, DCBRDRBR and Dlinker constructs all failed to block translocation. The construct where the linker is replaced by the CBR domain alone (Dlinker þ CBR) also inhibited translocation, although not quite as strongly as full-length SR. These results are, therefore, in close agreement with the ability of the different linker mutants to effect the Sec61-Sec62 interaction.
Full-length SR was also able to reduce translocation of preprolactin, although the effect was much weaker than observed for preprocecropin. This is consistent with previous studies using endogenous SR and Sec61 complex reconstituted into proteoliposomes, which also showed that increasing levels of SR relative to Sec61 reduced translocation of the efficiency of preprolactin 45 .
Importantly, none of the linker mutants were able to inhibit preprolactin translocation. The lack of effect of the Dlinker þ CBR construct, which did block preprocecropin translocation, therefore indicates distinct underlying mechanisms of inhibition for the two pathways. Neither the full-length SR nor any of the linker mutants had any effect on cytochrome B5 integration.

Discussion
Here we have shown that the interaction of Sec62 and Sec61 as revealed by crosslinking is highly sensitive to the presence and the absence of ribosomes. In contrast to the core Sec61 heterotrimer, the heptameric yeast Sec complex has been shown to be unable to bind to ribosomes 51 . Consistent with this, low resolution cryoelectron microscopy structures of the Sec complex reveal additional density above the cytoplasmic face of the core Sec61 heterotrimer 52 , which would likely occupy a similar position as the ribosome 53 and hence occlude binding. This is in good agreement with the observation that crosslinking between mammalian Sec62 and Sec61b is strongly induced by the removal of ribosomes from RMs and is inhibited by their readdition. Furthermore, cross-links between Sec62 and Sec61b were shown previously to be exclusively absent from the ribosome-associated membrane protein fraction following detergent solubilisation 34,35 . ARTICLE The fact that more than 50% of Sec62 can be cross-linked to Sec61, following removal of ribosomes, suggests that the interaction of these components is strongly favoured if ribosomes are absent. Furthermore, if the free Sec complex is unable to interact with ribosomes 51 , this raises the question as to how ribosomes can subsequently rebind to such complexes. Here we find that SR is able to destabilize the interaction between Sec62 and Sec61, which we detect as a loss of crosslinking between Sec62 and Sec61b and a block in the translocation of Sec62dependent substrates.
It is well established that SR catalyses the transfer of incoming ribosome-nascent chain complexes from SRP to the Sec61 translocon 11 . Hence, the ability of SR to interact with Sec61 and displace or reposition Sec62, immediately suggests a mechanism by which the Sec complex can be reorganized to permit the subsequent docking of the ribosome-nascent chain complex (Fig. 6). Furthermore, the SR-Sec61 intermediate would preferentially recruit SRP-bound ribosome-nascent chain complexes, potentially providing an additional step where targeting fidelity can be enhanced. Moreover, this would also rationalize the previous observation that SRP-bound ribosomenascent chains can access a pool of Sec61 to which free ribosomes are unable to bind 54 .
A recent study showed that arresting translocation of a cotranslational substrate using a small, tightly folded domain at the luminal side of the translocon, led to the stable recruitment of Sec62 (ref. 55). Hence, although dispensable for co-translational translocation of most substrates 37-39 , a subset of nascent chains may require Sec62 to facilitate later steps of their the translocation. Such nascent chains may possess particular features, which initiate reorganization of the ribosometranslocon complex and thereby expose a stable binding site for Sec62.
Our experiments revealed that crosslinking of SRa to Sec61b was much stronger with PKRMs as compared with RMs. There is still a low amount of SRa to Sec61b crosslinking in RM, rationalised by the fact that B80-90% of translocons are engaged with ribosomes 45 , hence a small population of translocons are still available to bind SR.
SR also had little effect on the overall cross-link profile of Sec61b with RMs as compared with PKRMs. This suggests that once the ribosome is fully engaged with the translocon, access of SR is likely to be strongly hindered. This is rationalized by cryoelectron microscopy structures of the ribosome-Sec61 (ref. 53) and ribosome-SRP-SR 6 complexes, which predict a strong potential overlap of density at the ribosomal exit tunnel.
Complete deletion of the linker domain led to the formation of a novel cross-link between SRa and Sec62. This is also observed with the CtSRaNG domain, indicating that this interaction from the NG domain appears conserved. Human and Chaetomium NG domains both contain multiple cysteines but only one of these is conserved and is located in helix 3 of the N-domain, hence this is a strong candidate to be involved in crosslinking to Sec62. Escherichia coli FtsY, which lacks the linker domain, is known to interact with cytoplasmic loop 5 of SecY via its NG domain. Therefore, in the absence of the linker, SRa may also bind to the Sec61 complex in a similar manner, positioning the NG domain close to Sec62. If the linker is still present, but unable to displace Sec62, as with full-length CtSRa, this may interfere with the binding of NG domain.
As well as interacting with the translocon, SRa also binds to the ribosome 47,48 . Our study characterizes a long stretch of positively charged residues in the linker region of eukaryotic SRa of previously unknown function. We identify two distinct elements within the linker and show that ribosome binding requires the second element (RBR) that is conserved between fungi and mammals. Notably, a large number of factors that act on the nascent polypeptide use the tunnel exit as a hub for interaction. Although these factors are unrelated in structural and functional terms, they seem to use a common mechanism for ribosome interaction involving a stretch of positively charged residues 56,57 . The RBR identified in this study within the linker region of SRa probably acts by the same mechanism.
The first element (CBR) being unique to higher eukaryotes and rich in lysine residues is necessary and sufficient for the displacement of Sec62 by SRa. Sec62 contains several positively charged regions also rich in lysine residues. Hence the linker of SRa and Sec62 are likely to compete for Sec translocon binding via electrostatic interactions. Indeed, Sec63 has a very negatively charged C terminus 34,35 and so might well contribute to the mutually exclusive binding site. In this respect, CBR and Sec62 might also compete for binding to Sec63.
Interestingly, mammalian Sec62 differs from its yeast homologues by the presence of an additional positively charged region in the N-terminal region 58 (Supplementary Fig. 6). This might represent an extra level of regulation in higher eukaryotes, and explain why the additional positive-charged (CBR) region within the linker domain of human SR is required to displace Sec62 from Sec61.
Taken together, we have identified two distinct interaction motifs in SR and show that one of them serves a function that is distinct from the 'canonical' eukaryotic RBR in regulating association of Sec62 with Sec61 in the ER membrane. Cells under different physiological conditions can express different levels of SR 59 . In combination with our findings, this may provide a mechanism to differentially modulate the flux of substrates through two distinct ER translocation pathways, allowing cells to efficiently respond to changing profiles of secretory substrates with distinct properties under different physiological conditions.

Methods
Purification of mammalian SR and deletion variants. For purification of mammalian SR, a bi-cistronic construct containing human SRa and murine SRbDTM was cloned in the pET16b vector (Novagen). Linker deletion variants of SR were created using the same construct. Expression was performed in E. coli BL21 (DE3) cells with overnight induction with 1 mM isopropyl-b-D-thiogalactoside at 16°C. The cells were lysed using a sonicator followed by micro fluidizer in SR buffer (25 mM HEPES pH (7.5), 300 mM KOAc, and 5 mM Mg(OAc) 2 ) and cleared by centrifugation at 63,000 g for 30 min using a JA25.50 (Beckman) rotor. The resulting supernatant was purified first by Ni-NTA affinity chromatography. The eluate was applied to a Q-sepharose column and the flow-through, then bound to SP-sepharose resin and eluted with a 0.2-1 M KOAc salt gradient in SR buffer. Finally, the SR was purified by size exclusion chromatography on a Superdex 200 column 47 . The final concentration of salt was reduced to 150 mM before snap freezing in liquid nitrogen.
Purification of CtSR and linker deletion variants. CtSRa and the deletion variants and CtSRb were cloned as hexa-histidine-tagged proteins via the pETHis vector 26,60 . Untagged CtSRbDTM was cloned into the pET24a vector (Novagen). CtSR His-tagged CtSRa and untagged CtSRbDTM were expressed individually using the auto-induction method in E. coli BL21 (DE3) cells at 24°C overnight 61 . Pellets of cells expressing CtSRa and CtSRbDTM were lysed together in the lysis buffer (20 mM Tris pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 20 mM imidazole and 0.02% (v/v) Nonidet-P40 ) using a microfluidizer. After the lysis, the cell debris was removed by centrifugation at 67,000 g using JA25.50 (Beckman) rotor. The supernatant was collected and incubated at 4°C for 1 h on a rotating wheel. The CtSR complex was then purified as described for mammalian SR 47 . His-tagged CtSRa, CtSRa deletion variants and CtSRbDTM were expressed and purified similarly.
Microsome crosslinking in the presence of SR. RMs, EKRMs and PKRMs were prepared from canine pancreas 62,63 . Pancreatic tissue was cut into small pieces, passed through a tissue press and then homogenized in RM buffer (20 mM HEPES-KOH pH 7.6, 50 mM KOAc, 2 mM Mg(OAc) 2 , 250 mM sucrose and 2 mM DTT). The resulting extract was then centrifuged sequentially at 1,000 g and 10,000 g to yield a post-mitochondrial supernatant, which was overlayed onto step gradients of 1.5, 1.75 and 2 M sucrose (in RM buffer) and centrifuged overnight at 235,000 g at 4°C using a Ti45 rotor (Beckman). Membranes banding at the 1.75-2 M cushion interface were collected, diluted with RM buffer without sucrose (RM À ) and pelleted before resuspension in RM buffer at a concentration of 4 eq ml À 1 to yield RMs. EKRMs were prepared by washing RMs with high salt (650 mM KOAc) before resuspension in 20 mM HEPES-KOH pH 7.6, 20 mM EDTA, 650 mM KOAc, 2 mM DTT and 2 M sucrose. Membranes were held on ice for 30 min and then overlayed with sucrose cushions of 1.5, 1 and 0.25 M in the same buffer and centrifuged overnight in the SW40 rotor (Beckman) at 38,000 r.p.m. Membranes floating at the 0.25-1 M interface were collected, diluted with RM( À ) buffer and collected by centrifugation before resuspending in RM buffer to give EKRM. PKRM were prepared in an identical manner, except salt-washed membranes were resuspended in 20 mM HEPES-KOH pH 7.6, 500 mM KOAc, 5 mM Mg(OAc) 2 , 2 mM puromycin, 2 mM GTP and 2 M sucrose (prewarmed to 25°C), and incubated at 25°C for 20 min before floatation.
To purify translocon components, RMs were pre-extracted with 0.1% (w/v) digitonin and then solubilised with 3% (w/v) digitonin in the presence of 500 mM KOAc. The resulting extract was then centrifuged to yield a ribosome pellet and supernatant. To purify SPC, the supernatant was applied to a concanavilin A column. Bound proteins were eluted with a-methylmannoside and then SPC purified on an SP-sepharose ion exchange column 45 . SR was purified from the flow-through of the concanavalin A column using an anti-SRa immunoaffinity column 45,64 . Sec61 was purified from the ribosome pellet fraction by ion-exchange chromatography with Q and SP-sepharose, following release from the ribosome by puromycin treatment 45,64 .
Ten equivalents of EKRMs or PKRMs were incubated with 0.7-10 mM of mammalian SR and linker deletion variants in a 20 ml reaction in the assay buffer (25 mM HEPES-KOH pH 7.5, 120 mM KOAc, 2 mM Mg(OAc) 2 , 250 mM sucrose) at 25°C for 10-15 min. BMH was added to the final concentration of 10 mM and the reaction mixture was further incubated at 25°C for 10 min. The crosslinking reaction was stopped by the addition of 1 mM DTT and incubating the sample on ice for 15 min. The whole sample was then mixed with SDS sample buffer and proteins and crosslinking adducts were analysed by SDS-PAGE and western blotting with either Sec61b (1:3,000), SRa (1:1,000) or Sec62 (1:1,000) antibodies. Complete blot images for cropped panels are shown in Supplementary Fig. 7.
For immunoprecipitation of endogenous cross-linked SRa, EKRM (100 eq) were treated with 40 mM BMH for 10 min at 30°C. After quenching with 10 mM DTT, membranes were reisolated by centrifugation and then resuspended in 70 ml of 20 mM Tris-HCl pH 8.0. 2% (w/v) SDS and 2 mM EDTA and heated for 10 min at 70°C. The denatured membranes were diluted with 700 ml of ice-cold immunoprecipitation (IP) buffer A (20 mM Tris-HCl pH 8.0, 0.4% (v/v) Nonidet-P40, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF (phenylmethylsulfonyl fluoride)) and then 2 ml of anti-SRa or a non-related rabbit antiserum were added. After rolling for 1 h at 4°C, immune complexes were recovered with protein A-sepharose (Genscript) and then washed two times with IP buffer A, two times with 20 mM Tris-HCl pH 8.0, 0.2% (v/v) Triton X-100, 500 mM NaCl, 1 mM EDTA and once with 10 mM Tris-HCl pH 8.0, before elution with SDS-PAGE sample buffer at 70°C for 10 min. Samples were then analysed by SDS-PAGE and western blot with Sec61b antibodies (1:1,000) and detection with protein A-horseradish peroxidase (Sigma) to reduce cross-reaction with IgG heavy chains.
Preparation of Canine and Chaetomium 80S ribosomes. To prepare canine ribosomes 47 , dog RM were resuspended in 50 mM HEPES-KOH pH 7.6, 600 mM KOAc, 12 mM Mg(OAc) 2 , 2 mM DTT, with protease inhibitor cocktail and treated with 0.5 mM puromycin for 15 min at 25°C. The reaction was then overlayed onto a 1.5 M sucrose cushion made up in the same buffer and centrifuged at 444,000 g for 1 h 20 min at 15°C in the MLA-80 rotor (Beckmann). The resulting ribosomal pellet was then rinsed and resuspended in 25 mM HEPES-KOH pH 7.6, 120 mM KOAc, 2 mM Mg(OAc) 2 , 1 mM DTT.
For purification of Chaetomium ribosomes 56 , C. thermophilum cells were grown in a rotary shaker at 90 r.p.m. at 55°C for 3 days. Cells were then harvested with a vacuum filter and ground to a fine powder in a mortar in the presence of liquid nitrogen. The powdered mycelium was then resuspended in 20 mM HEPES-KOH (pH 7.5), 100 mM potassium acetate, 125 mM sucrose, 7.5 mM Mg(OAc) 2 , 1 mM DTT and 0.5 mM PMSF and centrifuged at 28,000 g in SS-34 rotor (Sorvall), for 15 min to remove the insoluble material. To pellet ribosomes, the supernatant was overlaid on a high-salt sucrose cushion (500 mM potassium acetate, 1.5 M sucrose) prepared in the same lysis buffer and centrifuged at 300,000 g in a Ti70 rotor (Beckman) for 18 h. The translucent ribosomal pellet of ribosomes was then resuspended in 20 mM HEPES-KOH (pH 7.5), 120 mM potassium acetate, 5 mM magnesium acetate, 1 mM DTT and 0.5 mM PMSF.
Ribosome co-sedimentation assay. For the co-sedimentation assay, 25 pmol of ribosomes was incubated with 100 pmol of each protein in the assay buffer containing (25 mM HEPES pH 7.5, 250 mM KOAc, 2 mM Mg(OAc) 2 , 0.05% (v/v) Triton X-100 and 1 mM DTT). The reaction mixture was incubated at 25°C for 20 min and then loaded onto a 0.5 M sucrose cushion prepared in the assay buffer. Ribosomes were pelleted by ultracentrifugation at 267,000 g at 4°C for 1 h using a TLA100.3 rotor (Beckman). The proteins in the pellet fraction and in the supernatant (following precipitation with trichloroacetic acid) were analysed using SDS-PAGE and staining with Coomassie Brilliant Blue.
In vitro translocation assays. Templates for preprolactin 65 , opsin-tagged preprocecropin A, apelin, statherin 39 and cytochrome B5 66 were generated by PCR 39  For post-translational translocation assays, translation reactions (25 ml) were carried out using nuclease-treated rabbit reticulocyte lysate (Promega). Translations were performed in the presence of [ 35 S] methionine (0.769 MBq, 43.48 TBq mmol À 1 ; Perkin Elmer). Amino acids minus methionine (Promega) were added to 30 mM. 1 mg of in vitro-transcribed RNA was then added and the sample was incubated for 15 min at 30°C. Puromycin was added to 1 mM following the translation and incubated at 30°C for a further 5 min to ensure effective release of the polypeptide from the ribosome. Recombinant SR was added to 10 mM as well as 10 eq of RM and the sample was then incubated for 20 min at 30°C. Membranes were recovered by centrifugation through an 80 ml high-salt cushion (0.75 M sucrose, 0.5 M KOAc, 5 mM Mg(OAc) 2 , 50 mM HEPES-KOH, pH 7.9) at 100,000 g for 10 min at 4°C in the TLA100 rotor (Beckman). The membrane pellet was resuspended in 20 ml low-salt buffer, 100 mM sucrose, 100 mM KOAc, 5 mM Mg(OAc) 2 , 50 mM HEPES-KOH pH 7.9, 1 mM DTT and treated with 250 mg ml À 1 RNase A at 37°C for 10 mins. The resulting samples were analysed by SDS-PAGE and phosphorimaging using a Typhoon FLA-7000 (GE Healthcare).
For co-translational reactions, RMs were preincubated with 10 mM of the indicated recombinant SR (or an equal volume of SR buffer) on ice for 20 min and were then recovered by spinning at 100,000 g for 20 min and resuspended in RM buffer (250 mM sucrose, 50 mM HEPES-KOH pH 7.5, 50 mM KOAc, 2 mM Mg(OAc) 2 , 1 mM DTT). Translation reactions (25 ml) were carried out as above in the presence of 10 eq of SR/buffer-treated RMs and the samples were incubated for 30 min at 30°C. Translated products were then immunoprecipitated by adding 10 volumes of Triton IP buffer (10 mM Tris-HCl pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% (v/v) Triton X-100) and the appropriate antisera (anti-opsin 67 or anti-pPL (a gift from Sharon Tooze)). The resulting samples were analysed by SDS-PAGE and phosphorimaging using a Typhoon FLA-7000 (GE Healthcare). Data were quantified using Aida (Raytek) and statistical analysis (one-way analysis of variance with Dunnett's post hoc test) was performed using GraphPad (Prism).