Interaction of the cotranslational Hsp70 Ssb with ribosomal proteins and rRNA depends on its lid domain

Cotranslational chaperones assist in de novo folding of nascent polypeptides in all organisms. In yeast, the heterodimeric ribosome-associated complex (RAC) forms a unique chaperone triad with the Hsp70 homologue Ssb. We report the X-ray structure of full length Ssb in the ATP-bound open conformation at 2.6 Å resolution and identify a positively charged region in the α-helical lid domain (SBDα), which is present in all members of the Ssb-subfamily of Hsp70s. Mutational analysis demonstrates that this region is strictly required for ribosome binding. Crosslinking shows that Ssb binds close to the tunnel exit via contacts with both, ribosomal proteins and rRNA, and that specific contacts can be correlated with switching between the open (ATP-bound) and closed (ADP-bound) conformation. Taken together, our data reveal how Ssb dynamics on the ribosome allows for the efficient interaction with nascent chains upon RAC-mediated activation of ATP hydrolysis.

A lready during translation at the ribosome, the emerging polypeptide chain is subject to enzymes for covalent modification, targeting factors for localization and chaperones for de novo protein folding [1][2][3] . In eukaryotes, the nascent polypeptide chain is modified by methionine amino-peptidases and N-acetyl transferases 2,4,5 . For targeting and insertion of membrane or secretory proteins into the endoplasmic reticulum membrane, the nascent chain is recognized and guided by the signal recognition particle (SRP) (refs 1,6-8). To prevent aggregation and assist in folding of newly synthesized proteins, Hsp70 chaperones interact with nascent chains. Members of the Hsp70 family are highly conserved among all kingdoms of life and are known to bind short, largely hydrophobic sequences exposed by unfolded proteins 9 .
Hsp70 proteins share a number of structural and functional characteristics. They consist of a highly conserved N-terminal ATPase domain, also termed nucleotide binding domain (NBD), which is connected via a linker to the substrate binding domain (SBD). The SBD is divided into a b sandwich subdomain (SBDb), an a-helical subdomain (SBDa or lid domain) and a less-conserved C-terminal domain of variable length 9 . The Hsp70-protein substrate interaction depends on ATP binding and on allosteric regulation between the NBD and the SBD. The ATP-bound state is characterized by a fast exchange rate of substrate (low affinity state), while in the ADP-bound state exchange is much slower (high affinity state) 10 . During the Hsp70 cycle, the chaperone switches between the ATP-bound state (open conformation) and the ADP-bound state (closed conformation) by major conformational rearrangements involving mainly the SBDa (refs [10][11][12][13]. Two types of cochaperones regulate the switch between the two states: Hsp40s stimulate ATP hydrolysis, and nucleotide exchange factors accelerate the exchange of ADP with ATP (refs 9,14). Due to the high conservation, the allosteric mechanism recently established for DnaK likely applies to all canonical members of the Hsp70 family 15 .
Fungi from the large division of Ascomycota possess an evolutionary conserved subfamily of Hsp70s, termed the Ssb-type Hsp70s, which is named after two nearly identical homologues Ssb1 and Ssb2 (collectively Ssb) from Saccharomyces cerevisiae 16 . Ssb is distinguished from other Hsp70 homologues as it interacts directly with the ribosome by a mechanism that is currently unknown 3,17 . When bound to the ribosome, Ssb interacts with a large variety of nascent polypeptides and assists their cotranslational folding 18 . Because the ratio of ribosomes and Ssb in a yeast cell is near-balanced 19 each ribosome could cooperate with one Ssb molecule throughout protein synthesis. However, only about 50% of Ssb is ribosome-associated at steady state, while the remainder is free in the cytosol 3 . This cytosolic pool of Ssb is involved in additional functions, not directly connected to cotranslational protein folding 3,20,21 .
Ribosome-bound Ssb cooperates with a specific cochaperone termed the ribosome-associated complex (RAC), which is composed of the Hsp70 protein Ssz1 and the Hsp40 protein Zuo1 (refs 22-24). Deletion of either ZUO1, SSZ1 or SSB1/SSB2 results in similar growth defects including slow growth, cold sensitivity and hypersensitivity towards paromomycin. The combined deletion of any of the above genes does not result in additive effects, suggesting that Ssb, Zuo1 and Ssz1 function together in the same pathway 3,25,26 . Of note the Ssb chaperone system is not induced at high temperature, but rather SSB1 and SSB2 are down-regulated together with ribosomal proteins on heat shock 3,27 . Rather than being heat shock proteins Ssb and RAC belong to the group of chaperones linked to protein synthesis 3,28 . The ribosome interaction of RAC has been characterized biochemically and structurally 22,[29][30][31] . RAC binds close to the tunnel exit with contacts to Rpl22 and Rpl31 (refs 29-31), and several expansion segments of ribosomal RNA (rRNA), and stretches across the surface of the ribosome to the 40S ribosomal subunit 31 . Although the precise position of, for example the NBD of Ssz1 or the J-domain of Zuo1 are not resolved in the cryo-EM structures, the crosslink of Zuo1 to Rpl31 supports the idea that the J-domain is positioned close to the tunnel exit and possibly to Rpl31 (ref. 29). As the J-domain of Zuo1 activates the Ssb ATPase and switches it to the high affinity substrate binding state 32 , Ssb and the J-domain of Zuo1 should occupy neighbouring binding sites at the ribosome.
Here, we present a detailed structural and biochemical analysis of the Ssb-ribosome interaction. We determined the structure of full-length Ssb from the thermophilic Ascomycota Chaetomium thermophilum 33,34 in the open ATP-bound conformation and characterized its ribosome binding by crosslinking and mutagenesis experiments. Ssb is positioned on the ribosome by dual interaction with ribosomal proteins and rRNA in close proximity of the tunnel exit. Ssb binding is modulated by RAC and our data allow to derive a model for the positioning and the dynamics of Ssb at the ribosome.

Results
Ssb interacts with ribosomal proteins at the tunnel exit. To better characterize the interaction of Ssb with ribosomes, we first investigated its positioning via crosslinking experiments. To this end, we employed a Dssb1Dssb2 strain expressing the point mutant Ssb1-A577K (termed Ssb1*), which displayed enhanced affinity for aSsb ( Supplementary Fig. 1a). Ssb1* was ribosome associated and fully complemented growth of a Dssb1Dssb2 strain ( Supplementary Fig. 1b,c). Crosslinking in cell extract generated several crosslink products detected by aSsb (Fig. 1a). After separation of the cytosol from ribosomes via a sucrose cushion, a prominent crosslink of B160 kDa was detected in the cytosolic fraction, while additional crosslinks of weaker intensity were observed in the ribosomal fraction (Fig. 1a). The 160 kDa band represents a crosslink between Ssb1 and the cytosolic Sse1, which functions as a cochaperone of the cytosolic pool of Ssb (refs 35-37) ( Fig. 1a; Supplementary Fig. 1d). Many of the weaker crosslinks migrated with a molecular mass compatible with crosslink products between Ssb and ribosomal proteins, most of which possess a molecular mass between 6 and 25 kDa. Consistently, most of the smaller crosslinks were recovered in the ribosomal pellet (Fig. 1a, green asterisks). We next probed the crosslinks obtained in a wild type or in a Dssb1Dssb2 extract with antibodies directed against ribosomal proteins surrounding the tunnel exit (Fig. 1b). While no differences in the crosslinking pattern of Rpl25, Rpl17, Rpl26 or Rpl31 between the Ssb1* and Dssb1Dssb2 strains were observed, Rpl35 (13.9 kDa), Rpl39 (6.3 kDa) and Rpl19 (21.7 kDa) formed crosslink products of the expected molecular mass in the wild type but not in a Dssb1Dssb2 extract (Fig. 1c). The general crosslinking pattern observed in extracts derived from wild type cells was similar, with bands of slightly weaker intensity, than observed in extracts from the Ssb1* strain ( Supplementary Fig. 1e). The crosslink band detected with aRpl35 was fuzzy or, in some experiments, migrated as a doublet of bands (Fig. 1c), which both represented crosslinks between Ssb and Rpl35 ( Supplementary Fig. 1f). From these data, we conclude that Ssb contacts Rpl35, Rpl39 and Rpl19, which are all located in close proximity of the tunnel exit.
Chaetomium thermophilum 33,34 , which is 75% identical to Ssb1 from Saccharomyces cerevisiae. Within the Hsp70 family, domain organization is highly conserved ( Fig. 2a; Supplementary Fig. 2; Ssb residue numbering is given for C. thermophilum unless stated otherwise). In order to crystallize full-length Ssb, we stabilized the ATP-bound, open conformation adapting a cysteine crosslinking strategy that was successfully used for the structure determination of Escherichia coli DnaK (ref. 11). To identify the relevant residues to create the cysteine bridge, we created a homology model of full-length Ssb based on DnaK (refs 11,38) using SWISS-MODEL 39,40 . On the basis of this model, the conserved threonine in the NBD P-loop was replaced by an alanine (T208A) to reduce the intrinsic ATPase activity. Two cysteines (in the NBD and SBDa) had to be introduced to link both domains via a disulfide bond. While the position in the NBD was readily determined (E51), we could not directly identify the relevant position of the second cysteine from the homology model. Therefore seven residues (534-540) in Ssb SBDa were individually mutated. After purification and oxidation, only the Ssb T208A-E51C-D534C mutant crystallized in the presence of ATP ( Supplementary Fig. 3a,b). The crystals belong to the monoclinic space group P 1 2 1 1 with unit cell parameters of a ¼ 66.9 Å, b ¼ 128.2 Å and c ¼ 79.7 Å (Table 1) and contain two molecules in the asymmetric unit. The structure was solved at 2.6 Å resolution by molecular replacement using DnaK as a search model 11 (Table 1). The resulting electron density map was of high quality showing well defined ATP and Mg 2 þ ligands, as well as the engineered C51-C534 disulfide bridge ( Supplementary  Fig. 4). The final model has excellent stereochemistry (Ramachandran allowed 99.5%) and only small parts of the structure are disordered (residues 1-3, 393-394, 561-571 and 613-614), indicated as dotted lines (Fig. 2b).
Crystal structure of ATP-bound Ssb. Ssb is a typical member of the Hsp70 family and the crystal structure shows the canonical domain architecture (NBD, SBDb and SBDa) of Hsp70 chaperones (Fig. 2a,b). The Ssb nucleotide binding domain (NBD, blue) and the substrate binding domain (SBD, purple and brown) are connected by the conserved inter-domain linker DLLLLDV (green) responsible for allosteric communication 15,41 . The

Rpl26
Rpl25 Figure 1 | Ssb contacts ribosomal proteins at the tunnel exit. (a) Ssb is crosslinked to proteins of the large ribosomal subunit. Isolated ribosomes of the Ssb1* strain were incubated with ( þ ) or without ( À ) the crosslinker BS 3 . The crosslinked sample (Total) was separated into a cytosolic supernatant (cyt) and a ribosomal pellet (ribo) via centrifugation through a low-salt sucrose cushion and was analysed via immunoblotting with aSsb. Shown is a short exposure of the upper part of the blot (4116 kDa), which contains the strong crosslink between Ssb1* and Sse1 (Ssb1*-Sse1) and a long exposure of the lower part of the blot (o116 kDa), which contains the weaker crosslinks between Ssb1* and ribosomal proteins (green asterisks). Rpl17 served as a marker for the ribosomal fraction. (b) Schematic representation of selected ribosomal proteins (Rpl19, Rpl35 and Rpl39 in pink; Rpl31 in blue; Rpl17, Rpl25 and Rpl26 in white) surrounding the tunnel exit (black) of the yeast ribosomal large subunit (grey). (c) Identification of ribosomal proteins crosslinked to Ssb1*. Crosslinking was performed in total cell extract of Ssb1* or Dssb1Dssb2 strains. Aliquots were analysed via immunoblotting using aSsb, aRpl35, aRpl39, aRpl25, aRpl19, aRpl17, aRpl26 and aRpl31 as indicated. Crosslink products between Ssb1* and ribosomal proteins (Ssb1*-XL) are indicated with red asterisks.
actin-like fold 42 and comprises two lobes (I and II), further divided into four subdomains (IA, IB, IIA and IIB) (Fig. 2b). ATP binds between the two lobes. Aspartate 13 coordinates the Mg 2 þ ion through two water molecules, and Mg 2 þ in turn coordinates both the ATP b and g phosphates ( Supplementary Figs 4a and 5a). Superposition of the NBDs (including the active site) of Ssb with DnaK shows an r.m.s.d. below 1 Å (for 358 Ca atoms) underlining the high structural conservation ( Supplementary Fig. 5b). As in the ATP-bound open conformation of DnaK, the b-sheet subdomain of the SBD (Ssb-SBDb, residues 404-513) contacts the NBD subdomains IA, IIA and IB (Fig. 2b). It shows the typical, distorted b-sandwich fold formed by two layers of b-sheets built by three and five strands ( Supplementary Fig. 5c,d). In Hsp70s the residues involved in substrate binding are conserved and the loop between the first and second b-strand of the SBDb (L 1-2 ) is important for substrate binding as it contributes to the shape of the binding cleft ( Supplementary Fig. 2). When we compare Ssb-SBDb with DnaK-SBDb two important differences are observed: methionine 410 (threonine 403 in DnaK) seems to restrict the entrance to the substrate binding cleft, and glutamate 411 (methionine 404 in DnaK) may be engaged in polar contacts with the substrate. Methionine 404 in DnaK forms van der Waals   contacts with the central leucine of the substrate peptide 43 ( Supplementary Fig. 5c,d). These adaptations in Ssb-SBDb are conserved in all members of the Ssb subfamily (Supplementary Fig. 2) but their influence on Ssb substrate specificity has not been addressed so far. In Ssb, the SBDb is followed by SBDa (residues 517-613, helical lid domain) that comprises four a-helices (A-D), with aB to aD forming a three-helix bundle. Our structure shows a high degree of flexibility for this helical bundle as indicated by high B-factors ( Supplementary Fig. 6). In cytosolic Hsp70 proteins SBDa comprises five a-helices. In Ssb, the helix aC of Ssb-SBDa contains a nuclear export signal (NES) (ref. 44) (residues 575 IEQALSEAM 583 ) of which the hydrophobic side chains are engaged in van der Waals contacts forming the hydrophobic core of the bundle.
The Ssb-SBDa is required for ribosome binding. Further analysis of the Ssb structure in combination with a detailed sequence alignment revealed the presence of a positive patch located in SBDa. This positive patch is mainly formed by conserved Arg and Lys residues (K597, K598, K604, R605 and K609) located in the helices aC and aD (Fig. 2c, left and middle panel; Fig. 3a). These residues are conserved in members of the Ssb subfamily ( To test whether these Ssb specific, positively charged residues are involved in Ssb-ribosome interaction, we created a series of mutants, in which the relevant positively charged residues had been replaced pairwise and performed quantitative ribosome binding assays (Fig. 3). The first pair K568/R569 (Sc K567/R568) is in a disordered loop connecting helices aB and aC (Ssb L BC ), the second and third pair K597/K598 (Sc R596/K597) and K604/R605 (Sc K603/R604) are located in helix aD (Ssb D1 and Ssb D2, respectively). We created reverse charge double mutants in yeast Ssb1, resulting in three variants: Ssb L BC (K567E/R568E), Ssb D1 (R596D/K597D) and Ssb D2 (K603D/R604D) (Fig. 3a). As the Ssb antibody is directed against the very C-terminus of Ssb, mutations within this region strongly affect antibody recognition ( Supplementary Fig. 7a), therefore an N-terminally myc-tagged Ssb1 (mycSsb1) was employed in these experiments ( Supplementary Fig. 7a). All three mutants (L BC , D1, and D2) fully complemented growth defects of a Dssb1Dssb2 strain (Fig. 3b). However, ribosome-binding of the three mutants was reduced to less than 50% of the mycSsb1 control (Fig. 3c,d;  Supplementary Fig. 7b). Due to possible synergistic effects of these mutations, double (L BC -D1, L BC -D2) and triple (L BC -D1-D2) mutants were tested. While double mutants fully complemented growth defects of the Dssb1Dssb2 strain, the triple mutant displayed slightly reduced growth at 20°C and at 30°C in the presence of 50 mg ml À 1 paromomycin (Fig. 3b). When the paromomycin concentration was raised to 500 mg ml À 1 , a concentration at which even growth of the wild type strain was reduced, the triple L BC -D1-D2 mutant displayed a severe growth defect (Fig. 3b). However, all three mutants were severely defective with respect to ribosome-binding (Fig. 3c,d). These findings indicated that the growth defect of the L BC -D1-D2 mutant was not due to a defect in ribosome binding, but was due to a negative effect caused probably by the accumulation of negative charges within the very C-terminal region of Ssb1. This is also supported by the observation that, consistent with previous data 16 , a series of Ssb1 C-terminal truncation variants, lacking the last three (DC3), eight (DC8) or twenty-three (DC23) residues fully supported growth of the Dssb1Dssb2 strain (Fig. 3b), but displayed severe ribosome-binding deficiency (Fig. 3c,d;  Supplementary Fig. 7b) as did internal deletions (DNES) (Fig. 3b-d; Supplementary Fig. 7b). In order to assay the effect of these variants on the integrity of the helical bundle, we analysed them by CD spectroscopy. As expected from the crystal structure, the DNES variant showed reduced helicity compared with the reverse charge mutants and the WT ( Supplementary  Fig. 3c,d). These effects could be explained by destabilization of the helical bundle due to disruption of the hydrophobic core. Taken together, our data indicate that the loop L BC and helix aD as part of the three-helix bundle form an interaction platform that allows Ssb to efficiently bind to the ribosome. The interaction was strictly dependent on conserved, positively charged residues clustering into a three-dimensional epitope.
Ssb-ribosome interaction is modulated by RAC. Ssb1 interacts with the ribosome by contacting Rpl35 and Rpl39 in close proximity to the tunnel exit. The Ssb cochaperone RAC (refs 3,14) contacts Rpl31 and stimulates ATP hydrolysis in Ssb (Fig. 1b, blue) [29][30][31][32] . Therefore, we wanted to test whether RAC influences Ssb binding to the ribosome. We employed a strain lacking RAC (Dzuo1Dssz1) or a strain where wild type Zuo1 was replaced by the Zuo1-H128Q variant (RAC-H128Q), which does not stimulate the ATPase activity of Ssb due to a mutation in the conserved HPD motif in the J-domain 32 . With respect to ribosome binding, RAC-H128Q behaves similar to wild type RAC, which fully binds to ribosomes under low-salt conditions and is released from ribosomes under high-salt conditions (Fig. 4a). However, ribosome binding of wild type Ssb1 was significantly reduced in a Dzuo1Dssz1 or in a RAC-H128Q strain (Fig. 4a,b; Supplementary Fig. 8) indicating that ribosome-association of Ssb was hampered when RAC was absent or non-functional as a cochaperone. We next probed the interaction of Ssb with Rpl35 and Rpl39 in the Dzuo1Dssz1 and RAC-H128Q strains employing the strong crosslinks of Ssb1 to Rpl35 and Rpl39 in the wild type  (Figs 1c and 4c). The crosslink between Ssb1 and Rpl35 was reduced to less than 10% and the crosslink between Ssb1 and Rpl39 was reduced to less than 5% when RAC was absent or non-functional (Fig. 4c-e; Supplementary Fig. 8

Figure 4 | Ssb-ribosome interaction is modulated by RAC and involves rRNA expansion segments.
(a) Ribosome-binding of Ssb is reduced when RAC is absent, non-functional, or if ATP hydrolysis is prevented by the Ssb1-K73A mutation. Total cell extract (tot) was separated into a cytosolic fraction (cyt) and a ribosomal pellet (ribo) under low-salt (LS) or high-salt (HS) conditions. Immunoblots were decorated with aSsb, aZuo1, aSse1 (cytosolic marker) or aRpl35 (ribosomal marker). (b) The ribosome-bound fraction of Ssb in extracts derived from RAC-H128Q, Dzuo1Dssz1 or Ssb1-K73A strains is reduced. Fig. 8), error bars represent the s.d. (c,d) The contact between Ssb1 and Rpl35 or Rpl39 is strongly reduced if RAC is absent, non-functional, or if ATP hydrolysis is prevented by the Ssb1-K73A mutation. Crosslinking was performed in cell extracts of strains expressing Ssb1* (Supplementary Table 1) and (c) wild type RAC, RAC-H128Q, or carrying the Dzuo1Dssz1 mutation or (d) carrying the Ssb1*-K73A mutation. Immunoblots were decorated with antibodies directed against Rpl35 or Rpl39. Crosslink products between Ssb1* and ribosomal proteins (Ssb1*-XL) are indicated with red asterisks. (e) Comparison of crosslinking efficiencies between Ssb1* and Rpl35 or Rpl39 in cell extracts derived from Ssb1* strains expressing wild type RAC, RAC-H128Q, carrying the Dzuo1Dssz1 mutation, or expressing Ssb1*-K73A. The band intensities of crosslink products between Ssb1* and Rpl35 (Ssb1*-Rpl35) or Rpl39 (Ssb1*-Rpl39) were determined in at least 3 independent experiments. The intensity of Ssb1*-Rpl35 and Ssb1*-Rpl39 crosslinks in the Ssb1* strain was set to 100%. Error bars represent the s.d. does not complement growth defects of a Dssb1Dssb2 strain 23 . Ssb1-K73A also showed moderately reduced binding to ribosomes, similar to the ribosome-binding defect observed in the absence of functional RAC (Fig. 4a,b). Importantly, however, crosslinking of Ssb1-K73A to Rpl35 and Rpl39 was severely reduced (Fig. 4d,e). Thus, the combined data strongly support a model in that in the closed conformation Ssb was in close contact to Rpl35 and Rpl39, while this was not the case in the open conformation.

Quantification of the ribosome-bound fraction of Ssb under low-salt conditions is based at least on 3 independent experiments (Supplementary
Ssb contacts specific expansion segments of rRNA. As the interaction of Ssb with nontranslating ribosomes is salt sensitive 3,18 (Fig. 4a) and we detected a strict requirement of a positive surface patch for ribosome binding, we anticipated that ribosome binding of Ssb might involve interaction with ribosomal RNA. To experimentally test this hypothesis, we used the CRAC (ultraviolet crosslinking and analysis of complementary DNA (cDNA)) methodology 45 , which was previously used successfully to identify RNA-protein interactions in ribosomal subunits [45][46][47] . We found that Ssb1 directly contacts rRNA elements on the 60S subunit. Ssb1 was efficiently crosslinked to the eukaryote specific expansion segments ES24 and ES41 close to the tunnel exit and to ES39 more distant from the tunnel exit (Fig. 4f,g). These data support the contribution of the positively charged patch on Ssb to the interaction with rRNA.

Discussion
Ssb interacts with nascent polypeptide chains when they emerge from the ribosomal tunnel exit 3,18,26 and so it was long anticipated that Ssb must interact with the ribosome close to the polypeptide tunnel exit. However, the exact localization, molecular details of this interaction, and the interplay with its cochaperone RAC remained unclear.
The crystal structure in the open ATP-bound conformation shows that Ssb represents a canonical member of the Hsp70 family. Overall, the structure is highly similar to the E. coli Hsp70 homologue DnaK in its open conformation 11,48 . The structure was readily obtained using a crosslinking approach developed for DnaK (ref. 11) underlining the high conservation within the Hsp70 family. The nucleotide and substrate binding domains are connected by a conserved, hydrophobic linker region. The substrate binding site is more narrow compared with DnaK due to specific changes in SBDb, probably reflecting different substrate specificity. The Ssb SBDa comprises four helices and displays high flexibility. All elements required for ATP binding, for substrate binding and for allosteric regulation are present, suggesting that Ssb follows the same molecular mechanism of allostery as DnaK (ref. 15).
Ssb SBDa exposes a positively charged surface region conserved in all members of the Ssb subfamily of Hsp70s, but not in canonical cytosolic Hsp70s such as Ssa. Comparison with the closed conformation of DnaK shows that also in the closed conformation of Ssb this region would be solvent exposed and contacts neither SBDb nor the NBD, indicating that it is available to interact with another partner. To address if the positively charged patch is involved in the interaction of Ssb with ribosomes, we performed crosslinking experiments in combination with ribosome binding assays. A combination of structure based in vivo and in vitro mutational analysis of three positive patches within the SBDa lid domain demonstrated the importance of these residues for ribosome binding. Although these mutations impaired, or when combined, abolished ribosome binding of Ssb almost entirely, cell growth was not significantly affected. This observation suggests that the direct interaction of Ssb with ribosomes is less important than previously anticipated. However, the finding is not too surprising if one considers that Ssb-type Hsp70s are confined to Ascomycota and, for example, mammalian cells do not contain Ssb-type Hsp70 homologues, which directly interact with ribosomes 49 . Likely, when Ssb is not prepositioned on the ribosome, Ssa, the other major cytosolic Hsp70 in yeast, contributes to cotranslational protein folding and thereby compensates for the ribosome binding defects of the Ssb lid domain mutants 28 .   Supplementary Fig. 9). In this state, the substrate (nascent chain, NC; purple) is not tightly bound (low affinity state). (b) On interaction with the cochaperone ribosome-associated complex (RAC), ATP is hydrolyzed and Ssb switches to the closed conformation (ADP-bound, post-hydrolysis state), which involves flipping of SBDa onto the SBDb. Now Ssb can tightly interact with the nascent chain via SBD (high affinity state). The ribosome is shown in grey, the 60 S and 40 S subunits are indicated. Exposed expansion segments of rRNA (ES41, ES24) and ribosomal proteins (Rpl19, Rpl22, Rpl31, Rpl35, Rpl39) are depicted in orange and salmon, respectively. The tunnel exit is highlighted by a black circle (exit). Ssb is shown in blue and domains are labelled (NBD, SBDa, SBDb).
Indeed, Ssa, together with the J-domain chaperone Jjj1 can assist cotranslational protein folding in yeast 50 .
Using CRAC methodology we identified three rRNA expansion segments (ES24, ES39 and ES41) on the 60S ribosomal subunit as direct interaction partners of Ssb. While ES24 and ES41 are located close to the tunnel exit (Fig. 5), ES39 is more distant. In the presence of RAC, ES39 appears shielded while ES41 and ES24 seem available for Ssb binding 30,31 . The crosslink to ES39 could therefore indicate an additional rRNA binding site of Ssb when RAC is absent. Using ribosome-binding experiments we also detected residue-specific contacts between SBDa and the ribosome. Our data indicate that Ssb interacts with Rpl35, Rpl39 and Rpl19 close to the tunnel exit. Using the crosslink data as restraints, we positioned the structure of ATP-bound Ssb on the ribosome 51 . SBDb was docked close to the exit tunnel to allow for the interaction with a nascent chain and SBDa was oriented towards ES41. Choosing this orientation, SBDa is wedged in between Rpl22 and Rpl31, and within crosslinking distance to ES24 (Fig. 5a,  Supplementary Fig. 9). While this orientation of the ATP-bound Ssb in the open conformation correlates with the crosslinks to ES41 and ES24, SBDa is not in close proximity of Rpl35 and Rpl39. We therefore created a model of Ssb in the closed conformation (based on DnaK (ref. 43)) as switching between the open and closed states relocates SBDa by 460 Å (Fig. 5b). When we docked a model of ADP-bound Ssb (closed conformation) via the SBDb to the tunnel exit as before, SBDa is in close proximity of Rpl35 and Rpl39. Thus, docking Ssb to the ribosome in the open or closed conformation faithfully correlates with the observed protein and rRNA crosslinks and indicates that the structural rearrangements underlying allosteric regulation of Hsp70 proteins should also occur in Ssb and give rise to specific crosslinks. This model is further supported by the observation that the crosslinks to Rpl35 and Rpl39 were significantly reduced when Ssb cannot hydrolyze ATP (K73A mutant), as well as when RAC was absent or non-functional. Because RAC is strictly required for efficient ATP hydrolysis in Ssb (ref. 32), these data indicate that crosslinking to Rpl35 and Rpl39 was indeed confined to the closed (ADP-bound) conformation of Ssb. Consistently, it was previously shown that Ssb fails to interact with nascent chains in the absence of functional RAC (ref. 26), as the high affinity substrate binding state of Ssb is not induced. In summary, Ssb interacts with the 60S ribosomal subunit in a dual manner: by specific protein-protein as well as protein-rRNA contacts. Both types of interaction require the Ssb lid domain and are modulated by the cochaperone RAC.
The ribosomal surface around the tunnel exit, which serves as a binding platform for RAC and Ssb, has been described as an universal adaptor site for cotranslational factors including chaperones, enzymes and targeting factors 2,52,53 . It provides two main binding sites: Rpl25/Rpl35 (L23/L29 in E. coli) for SRP, trigger factor, the translocon, YidC/Oxa1 and Map and Rpl31/ Rpl17 (L17/L22 in eubacteria) for RAC, the nascent chain associated complex (NAC), the SRP receptor and peptide deformylase. The crosslinks of Ssb to ES41, which is next to the Rpl31/Rpl17 site and to Rpl35, which is part of the other site, indicate that Ssb bridges both binding sites. Interestingly, during ribosome biogenesis control mechanisms are implemented to ensure that the universal adaptor site is not only correctly assembled, but also protected from premature, unproductive interactions 54 . The ribosomal surface around the tunnel exit directs the highly dynamic interplay of a myriad of factors and we only begin to understand these complex interaction networks.

Methods
Strains and plasmids. Strains and plasmids are listed in Supplementary Table 1  and Supplementary Table 2. The nomenclature for ribosomal proteins used in this study is provided in Supplementary Table 3. DNA cloning and plasmid preparation were performed according to standard methods 55 . All constructs for the expression of mutants and tagged proteins were verified by sequencing and Western blot analysis.
The Chaetomium thermophilum (Ct) SSB coding sequence fused to an N-terminal His 6 tag was PCR amplified from the Ct cDNA and cloned into pET24. The mutations T208A, E51C and D534C were introduced by site-directed mutagenesis as described by the manufacturer (QuikChange Lightning, Agilent Technologies). The coding sequence of the 3-helical bundle of Ssb (residues 536-614) fused to an N-terminal His 6 tag was amplified by PCR from pET24a-His 6 Ssb-E51C-T208A-D534C and cloned into the pET24a vector. Mutations L BC (K568E/R569E), D1 (K597D/K598D), D2 (K604D/R605D), DNES (D575-587) or a combination of them were introduced by site-directed mutagenesis.
For CRAC analysis, the integration of the ProtA-TEV-His 6 tag at the N-terminus of the SSB1 coding sequence into the BY4742 genome was performed as described 59,60 . In brief, a cassette containing the clonNAT resistance gene linked to a 390 nucleotides region upstream of SSB1 coding sequence fused to the ProtA-TEV-His 6 tag sequence was PCR amplified using forward S1 and reverse S4 primers 59,60 and the purified PCR product was transformed into BY4742. The transformants were selected on YPD (1% yeast extract, 2% peptone and 2% dextrose) supplemented with 100 mg ml À 1 clonNAT agar plates.
The oxidation of the disulfide bond (C51-C534) was catalysed by the addition of a CuSO 4 and 1,10-Phenanthroline solution with a final concentration of 0.5 mM and 1.75 mM, respectively, and incubation for 30 min at room temperature. To minimize the formation of inter-molecular disulfide bonds, the concentration of the protein was kept below 10 mM. The oxidized Ssb was further purified by size exclusion chromatography (SEC; S200-26/60, GE Healthcare) in SEC buffer (20 mM Hepes/ NaOH pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , 10 mM KCl, 10% (v/v) glycerol).
Crystallization screens were performed at 291 K by the sitting-drop vapordiffusion method upon mixing equal volumes (0.2 ml) of the Ssb protein solution (10 mg ml À 1 ) and reservoir solution containing 20% (v/v) PEG 3350 and 0.2 M NH 4 H 2 PO 4 . Crystals grew after 3 days.
The three-helix bundle of Ssb and variants were expressed in E. coli BL21(DE3) Rosetta2 cells grown in LB media supplemented with kanamycin and chloramphenicol. Overexpression was induced with isopropyl-1-thio-b-Dgalactopyranoside (IPTG) at an OD 600 of 0.8-1.0 and the culture incubated overnight at 18°C. Cells were harvested and resuspended in lysis buffer (20 mM Hepes/NaOH pH 7.5, 300 mM NaCl, 20 mM MgCl 2 , 20 mM KCl), lysed in a microfluidizer and purified by IMAC: after application of the lysate, the column was washed with 20 column volumes of lysis buffer supplemented with 20 mM imidazole and eluted with 5 column volumes of lysis buffer supplemented with 200 mM imidazole. The eluate was applied to a SEC column S200 26/60 (size exclusion chromatography, GE Healthcare) run in SEC buffer (20 mM Hepes/NaOH pH 7.5, 300 mM NaCl, 5 mM MgCl 2 , 10 mM KCl). Samples were dialyzed overnight at 4°C against CD buffer (10 mM KH 2 PO 4 /K 2 HPO 4 pH 7.5, 150 mM KF).
Data collection and structure determination. Crystals were flash-frozen in liquid nitrogen after cryo-protection by transfer into cryo-solution containing mother liquor and 20% (v/v) glycerol. Diffraction data were measured under cryogenic conditions (100 K; Oxford Cryosystems Cryostream) at the European Synchrotron Radiation Facility (ESRF, Grenoble). Data were processed with XDS (ref. 61 Phenotypic assays. Growth defects were analysed by spotting 10-fold serial dilutions containing the same number of cells onto YPD or YPD þ 25, 50 or 500 mg ml À 1 paromomycin agar plates, and were incubated for 2 days at 30°C or 37°C or 3 days at 20°C, as indicated. CRAC analysis. In vivo CRAC experiments were performed using the BY4742 ProtA-TEV-His 6 -Ssb1 and the parental BY4742 strain as negative control. Yeast cultures were grown to an OD 600 of 2.0 in YPD, harvested and suspended in PBS (phosphate-buffered saline) before ultraviolet-irradiation in a Megatron UV chamber (1.6 J cm À 2 ) for 3 min. The RNA crosslinked to the protein of interest was treated as described 45,47 with the omission of transfer to nitrocellulose. Bands corresponding to the size of the protein of interest and higher were excised directly from the Bis-Tris NuPAGE gel (4-12%, Novex) and were subsequently digested by proteinase K. Extracted RNAs were amplified by RT-PCR. The obtained cDNAs were sequenced using the Illumina HiSeq sequencing platform and the reads were treated and analysed as described 77 with one modification: the reads were mapped to S. cerevisiae genomic reference sequence using Bowtie 2 (v 2.2.5) (ref. 78). Two independent experiments were performed for each sample and one representative experiment is shown.
Crosslinking assays. Protein crosslinking reactions were performed with the homobifunctional, amino-reactive crosslinker BS 3 (bis-(sulfosuccinimidyl)-suberate, spacer length 1.14 nm, Thermo Scientific) either with isolated ribosomes 29 or with total cell extracts prepared from freshly harvested yeast cells, rapidly frozen in liquid nitrogen, and subsequently disrupted with a cryo-mill (MM400, Retsch) 79 . About 500 ml of the cyro-mill cell powder (stored at À 80°C) was then resuspended in 1,200 ml crosslinking buffer (20 mM Hepes/KOH pH 7.4, 80 mM K(H 3 CCOO), 2 mM Mg(H 3 CCOO) 2 , 1 mM PMSF (phenylmethylsulfonyl fluoride), 2 mM DTT), cell debris were removed by centrifugation at 20,000 g for 2 min at 4°C, and aliquots of the supernatant were then incubated in the absence or in the presence of 1.2 mM BS 3 for 20 min at 21°C. Crosslinking reactions were quenched by the addition of glycyl-glycine to a final concentration of 30 mM. Ribosomes and crosslinks to ribosomal proteins were collected by centrifugation at 180,000 g for 35 min. Ribosomal pellets were resuspended in 100 ml ribosomebinding buffer and were precipitated by addition of 5% trichloroacetic acid (TCA). TCA pellets were analysed on 10% Tris-Tricine gels followed by immunoblotting.
Ribosome binding assays. Yeast strains were grown to early log phase on YPD, cycloheximide was added to a final concentration of 100 mg ml À 1 , and subsequently cells were harvested via centrifugation at 4,500 g. Cell pellets were resuspended in ribosome-binding buffer (20 mM Hepes/KOH pH 7.4, 2 mM Mg(H 3 CCOO) 2 , 120 mM K(H 3 CCOO), 100 mg ml À 1 cycloheximide, 2 mM DTT, 1 mM PMSF, protease inhibitor mix: 1.25 mg ml À 1 leupeptin, 0.75 mg ml À 1 antipain, 0.25 mg ml À 1 chymostatin, 0.25 mg ml À 1 elastinal, 5 mg ml À 1 pepstatin A) and disrupted using the glass beads method 80 . After a clearing spin at 20,000 g, each 60 ml of the total glass beads extract (A 260 between 40 and 250) was loaded onto a 90 ml low-salt sucrose cushion (25% (w/v) sucrose, 20 mM Hepes/KOH pH 7.4, 120 mM K(H 3 CCOO), 2 mM Mg(H 3 CCOO) 2 , 2 mM DTT, 1 mM PMSF, protease inhibitor mix) or onto a 90 ml high-salt sucrose cushion (25% (w/v) sucrose, 20 mM Hepes/KOH, pH 7.4, 600 mM K(H 3 CCOO), 2 mM Mg(H 3 CCOO) 2 , 2 mM DTT, 1 mM PMSF, protease inhibitor mix). After centrifugation at 400,000 g at 4°C for 25 min the cytosolic supernatant was collected and the ribosomal pellet was resuspended in 100 ml ribosome-binding buffer. Aliquots of the total glass beads extract, cytosolic supernatant, and resuspended ribosomal pellets were precipitated by addition of 5% TCA. TCA pellets were analysed on 10% Tris-Tricine gels followed by immunoblotting. For quantification purposes, the loading of cytosolic supernatant and ribosomal pellet was adjusted such, that the intensities of both bands on immunoblots were in the linear range of the analysis (Supplementary methods). Quantification was performed using AIDA ImageAnalyzer (Raytest). For statistical analysis at least 3 independent experiments were examined. Circular dichroism spectroscopy. Sample concentration was measured by Bradford assay against a BSA standard (Rotiquant Bradford reagent, Roth) and adjusted to 20 mM. The CD spectroscopy measurements were performed at room temperature in a ultraviolet-spectropolarimeter J750 (Jasco) using 1 mm quartz cuvettes (Hellma).
Immunoblotting. Proteins were separated on 10% Tris-Tricine gels. The following polyclonal antibodies were raised against peptides or purified proteins in rabbit: aSsb