The T-cell leukemia related rpl10-R98S mutant traps the 60S export adapter Nmd3 in the ribosomal P site in yeast

Mutations in the ribosomal protein Rpl10 (uL16) can be drivers of T-cell acute lymphoblastic leukemia (T-ALL). We previously showed that these T-ALL mutations disrupt late cytoplasmic maturation of the 60S ribosomal subunit, blocking the release of the trans-acting factors Nmd3 and Tif6 in S. cerevisiae. Consequently, these mutant ribosomes do not efficiently pass the cytoplasmic quality control checkpoint and are blocked from engaging in translation. Here, we characterize suppressing mutations of the T-ALL-related rpl10-R98S mutant that bypass this block and show that the molecular defect of rpl10-R98S is a failure to release Nmd3 from the P site. Suppressing mutations were identified in Nmd3 and Tif6 that disrupted interactions between Nmd3 and the ribosome, or between Nmd3 and Tif6. Using an in vitro system with purified components, we found that Nmd3 inhibited Sdo1-stimulated Efl1 activity on mutant rpl10-R98S but not wild-type 60S subunits. Importantly, this inhibition was overcome in vitro by mutations in Nmd3 that suppressed rpl10-R98S in vivo. These results strongly support a model that Nmd3 must be dislodged from the P site to allow Sdo1 activation of Efl1, and define a failure in the removal of Nmd3 as the molecular defect of the T-ALL-associated rpl10-R98S mutation.


Introduction
The eukaryotic ribosome is composed of 4 ribosomal RNA molecules and roughly 80 ribosomal proteins that come together to form the large (60S) and small (40S) subunits that make up mature 80S ribosomes. Eukaryotic ribosome biogenesis is an intricate assembly pathway that requires the assistance of hundreds of trans-acting factors [1][2][3][4][5][6], and although ribosomes catalyze protein synthesis in the cytoplasm, this process takes place predominantly in the nucleus. Each nascent subunit must therefore be exported through the nuclear pore complex, a process that is dependent on the export receptor Crm1 [7][8][9]. In 60S export, Crm1 is recruited via the export adapter Nmd3, which contains a leucine-rich nuclear export signal (NES) [10][11][12]. In yeast, additional factors assist in 60S export, including the non-canonical export receptor Arx1 [13,14], the mRNA export receptor Mex67-Mtr2 [15], Ecm1 [16], and Bud20 [17]. However, only Nmd3 appears to be universally conserved in eukaryotes as a dedicated 60S export factor [9][10][11][12]. Nascent 60S subunits entering the cytoplasm are bound by additional trans-acting factors, including the anti-association factor Tif6 (eIF6 in mammalian cells) [18]. Trans-acting factors bound to subunits in the cytoplasm block important functional regions and are thought to prevent immature subunits from engaging prematurely with the translational machinery (reviewed in [19]). For example, Tif6 binds to the inter-subunit face of the 60S subunit and sterically block 40S joining [20,21]. In addition, our lab and others have recently shown that Nmd3 occupies the A, P, and E sites of the tRNA channel where it would prevent association of ligands for these sites [22,23]. Newly exported subunits also lack several critical ribosomal proteins, including Rpl10 [6,24]. [Note that Rpl10 is also named uL16 in the new nomenclature [25], however, we use Rpl10 here for clarity with genetic mutations]. As such, pre-60S particles enter the cytoplasm in an inactive state and must undergo cytoplasmic maturation, a process of shedding trans-acting factors and assembling the final ribosomal proteins [26,27].
60S cytoplasmic maturation events take place in a highly ordered pathway [28], and the final two known steps include the release of Tif6 by Efl1 and Sdo1 [18,29,30] and the release of Nmd3, dependent on the GTPase Lsg1 [24,31]. After these steps, 60S subunits are competent to participate in translation. Genetic analyses place the release of Nmd3 downstream of and dependent on the prior release of Tif6 [28,32,33]. However, recent structures of pre-60S particles call that order of events into question [34]. A critical event that takes place upstream of Tif6 and Nmd3 release in cytoplasmic maturation is the addition of ribosomal protein Rpl10 [24,35], which binds in a cleft between the central protuberance and the P stalk of the 60S subunit ( Fig 1A). An internal loop of Rpl10 (referred to here as the P site loop) extends toward the catalytic center of the ribosome where it completes the peptidyl-transferase center and makes contact with the acceptor stem of P site ligands, including tRNA, during translation [36]. Extensive genetic analyses have demonstrated the importance of Rpl10 in 60S cytoplasmic maturation. Specifically, the P site loop of Rpl10 is required for the release of both Tif6 and Nmd3 [32,37].
While the mechanism of Nmd3 release is not understood, Tif6 release has been shown to depend on Sdo1 binding in the ribosomal P site where it cooperates with the GTPase Efl1, a paralog of the translation elongation factor EF2, to evict Tif6 [34]. Although Sdo1 has been shown to stimulate Efl1 GTPase activity on 60S subunits [33], structural studies suggest that Efl1 may release Tif6 by competing with it for binding, independent of GTPase activity [34]. We and others have proposed that the release of Tif6 by Efl1 and Sdo1 acts as a quasifunctional "test drive" of nascent 60S subunits in which translation factor-mimics assess the functionality of newly made subunits. Subunits that fail the test drive (including Rpl10 mutants) retain Tif6 and Nmd3, and are prevented from engaging in translation [32,34].
Deficiencies in the human Sdo1 protein, the Shwachman-Bodian-Diamond syndrome protein (SBDS), are associated with Shwachman-Diamond syndrome (SDS), a congenital disorder characterized by bone marrow failure and exocrine pancreatic insufficiency [38]. Interestingly, SDS is also associated with an elevated risk of myelodysplastic syndrome and acute myeloid leukemia development [39]. In addition, recurring mutations in RPL10 are associated with pediatric T-cell acute lymphoblastic leukemia (T-ALL) in humans, with the most commonly identified mutation being rpl10-R98S [40]. These mutations in T-ALL, like those in SDS, block the release of Tif6 and Nmd3. Together, these findings suggest that defects in late 60S cytoplasmic maturation may contribute to carcinogenesis.
Mutations in RPL10 are thought to block late 60S maturation by specifically impairing the Efl1-and Sdo1-dependent release of Tif6. A previous mutational analysis of the P site loop of . B) 10-fold serial dilutions of the glucose repressible RPL10 strain AJY3373 (P GAL -RPL10) harboring WT RPL10, rpl10-S104D, or rpl10-R98S vector, and either empty vector or the indicated TIF6 or EFL1 plasmids. Cells were spotted onto glucose-containing selective media to repress genomic RPL10. C) Tif6-GFP and Tif6-V192F-GFP localization monitored by fluorescence microscopy in WT RPL10, rpl10-S104D, and rpl10-R98S cells. AJY2766 (P GAL -RPL10, TIF6-GFP) and AJY3941 (P GAL -RPL10, TIF6-V192F-GFP) expressing WT RPL10, rpl10-S104D, or rpl10-R98S from plasmids were grown in the presence of glucose to repress genomic RPL10. DIC, differential interference contrast. Scale bar, 5μm. https://doi.org/10.1371/journal.pgen.1006894.g001 Rpl10-R98S traps Nmd3 in the ribosomal P site Rpl10 in yeast identified a class of mutants, typified by rpl10-S104D, that arrest ribosome biogenesis by failing to promote the Efl1-dependent release of Tif6 [32]. This arrest appears similar to that observed in sdo1 mutants as wells as in T-ALL-associated rpl10-R98S mutants. The RPL10 mutations also appear similar to disease-related mutations in SBDS in their impact on the binding affinity of Sdo1 to 60S subunits in vivo [34] and to mature ribosomes in vitro [41,42]. While reduced binding of Sdo1 to 60S subunits could explain the defect in Tif6 release, an understanding of how this leads to a block in Nmd3 release is unknown. Moreover, how Tif6 release is coordinated with Nmd3 release is not well understood.
In this work, we explore the molecular defect of the T-ALL related rpl10-R98S mutant in ribosome assembly using Saccharomyces cerevisiae. Despite phenotypic similarity, the rpl10-R98S mutant blocks 60S biogenesis by a mechanism that is distinct from the P loop rpl10-S104D mutant. We show that the molecular defect of the T-ALL associated rpl10-R98S mutation is a failure to release Nmd3 from the P site, thereby preventing the release of both Nmd3 and Tif6.

Results
The T-ALL related rpl10-R98S mutation blocks 60S biogenesis by a mechanism distinct from other rpl10 mutations Mutations in RPL10 were identified as driver-mutations in pediatric T-cell acute lymphoblastic leukemia (T-ALL) [40]. These mutations occurred almost exclusively at residue arginine 98 (R98), with the exception of one patient harboring the Q123P mutation, which lies adjacent to R98 at the base of the P site loop of the Rpl10 (Fig 1A). We previously showed that these T-ALL mutants exhibited a large subunit biogenesis defect in which Tif6 and Nmd3 release were blocked [40]. This phenotype appeared similar to the defect observed in a class of Rpl10 P site loop mutants, typified by the rpl10-S104D mutant, identified in an earlier mutational analysis of RPL10 [32]. We therefore anticipated that the molecular defect of the T-ALL related rpl10-R98S mutation was the same as that caused by rpl10-S104D and would be suppressed by the same mutations. We previously showed that rpl10-S104D can be suppressed by the dominant TIF6 mutant, TIF6-V192F, which belongs to a class of mutations that map to the 60Sbinding surface of Tif6 [32]. Such dominant mutations in TIF6 also suppress the growth defects due to loss of Efl1 [18,29] or Sdo1 [30] by weakening Tif6 affinity for the 60S subunit, thereby uncoupling the release of Tif6 from the requirement for Efl1 and Sdo1. We also identified mutations in EFL1 itself that suppressed rpl10-S104D. These mutations are thought to predispose Efl1 to undergo a conformational change that normally requires proper signaling through Sdo1 binding in the P site [32]. Thus, a probable explanation for the effect of rpl10-S104D on the release of Tif6 is that the Rpl10 mutant ribosome is defective for Sdo1 binding. Indeed, we recently found that both rpl10-S104D and rpl10-R98S ribosomes had a weakened affinity for Sdo1 compared to wild-type ribosomes in vitro [41,42], while others have also demonstrated that the T-ALL associated RPL10 alleles impair Sdo1 binding to the 60S subunit in vivo [34].
Unexpectedly, we found that although TIF6-V192F and EFL1 mutants could suppress the rpl10-S104D growth defect, they could not suppress the rpl10-R98S defect (Fig 1B). This result was surprising because both rpl10 mutants are defective for Tif6 release, and the lower affinity of Tif6-V192F for 60S should promote its release. We considered two possibilities to explain the difference between the two rpl10 mutations: 1) Tif6-V192F remains trapped on rpl10-R98S pre-60S subunits or 2) Tif6-V192F is released from rpl10-R98S subunits, but its release does not suppress the full biogenesis defect. To determine whether Tif6-V192F was released, we monitored the localization of Tif6-GFP versus Tif6-V192F-GFP in rpl10-R98S cells, and in rpl10-S104D cells as a control. The nuclear localization of Tif6 in wild-type cells reflects its ability to be released from pre-60S subunits in the cytoplasm and subsequently recycled to the nucleus [18,28,30,43]. As previously reported [32,40], wild-type Tif6-GFP was mislocalized to the cytoplasm in both rpl10 mutants, indicating a failure in Tif6 release from the 60S subunit ( Fig 1C, left panels). Also in agreement with previous observations, we found that the steady state localization of Tif6-V192F-GFP was nuclear in rpl10-S104D cells, demonstrating that it recycles back to the nucleus [32]. Unexpectedly, we found that Tif6-V192F-GFP was also localized to the nucleus in rpl10-R98S cells (Fig 1C, row 3). These results indicate that TIF6-V192F can bypass the block in its release in both rpl10 mutants. However, while its release could suppress rpl10-S104D, the release of Tif6-V192F was not sufficient to suppress the rpl10-R98S growth defect.
The rpl10-R98S mutant is suppressed by mutations in NMD3 and in TIF6 The results described above suggest that while the rpl10-S104D mutant is specifically defective in the Efl1-dependent release of Tif6, the rpl10-R98S defect is more complex. To gain insight into the mechanism by which the rpl10-R98S mutant stalls 60S biogenesis, we identified extragenic suppressors of the rpl10-R98S growth defect. We isolated multiple independentlyderived spontaneous suppressors of rpl10-R98S that appeared readily as fast-growing colonies among the severely growth-impaired rpl10-R98S background. Because Nmd3 release, in addition to Tif6 release, is impaired in rpl10-R98S cells, we considered the possibility that mutations in Nmd3 that promote its release might suppress the rpl10-R98S growth defect. The idea that suppressing mutations could be identified in NMD3 was also supported by our earlier observation that ectopic expression of WT NMD3 partially suppressed the rpl10-R98S growth defect [40] (Fig 2A). Thus, we began by sequencing the NMD3 genomic locus in each of the suppressors. Remarkably, of 18 spontaneous suppressors analyzed, 16 contained mutations in NMD3 (Table 1) (the NMD3-Y379D mutant was described previously [42]). To confirm that the NMD3 mutations were responsible for rpl10-R98S suppression, we expressed the mutants ectopically and found that all NMD3 alleles tested were dominant suppressors of rpl10-R98S (Fig 2A). Because the remaining two spontaneous suppressors contained wild-type NMD3, we performed whole genome sequencing on one isolate to identify mutations elsewhere in the genome that could be responsible for rpl10-R98S suppression. SNP analysis revealed a mutation in TIF6, resulting in a Gly to Val change at amino acid position 189 (TIF6-G189V). Sanger sequencing of the TIF6 locus from the final remaining suppressor identified the TIF6-T185A allele. To identify additional mutations in TIF6 that could suppress rpl10-R98S, we performed random PCR mutagenesis of TIF6 and identified additional suppressing mutations at amino acid positions T185 and G189 as well as mutations at P163 and T211 ( Fig 2B and Table 2). Thus, despite the fact that we had initially discounted TIF6 as a target of suppressing mutations because TIF6-V192F did not suppress rpl10-R98S, other alleles of TIF6 are suppressors. Unlike the case for NMD3, ectopic expression of WT TIF6 did not suppress rpl10-R98S and, in fact, slightly inhibited growth (Fig 2B). On the other hand, Tif6-V192F, which has weakened affinity for 60S subunits, neither inhibited nor suppressed the growth of rpl10-R98S cells, suggesting that the mechanism by which the TIF6 suppressing mutations acted was not through weakening the interaction between Tif6 and the ribosome.

rpl10-R98S suppressors restore Tif6 and Nmd3 release
Suppression of the rpl10-R98S growth defect implies that the NMD3 and TIF6 alleles identified above restore ribosome biogenesis by allowing the release of Tif6 and Nmd3 from mutant 60S subunits. We examined ribosomal subunit levels from rpl10-R98S cells harboring representative NMD3 and TIF6 mutants using sucrose density gradient sedimentation. This analysis confirmed that the biogenesis defect of rpl10-R98S cells was suppressed by either NMD3-Y379D or TIF6-G189V, evidenced by the apparent wild-type 60S:40S ratio and the disappearance of halfmer polysomes ( Fig 2C). We also checked the localization of both Tif6 and Nmd3 in rpl10-R98S mutant cells suppressed with either NMD3-Y379D or TIF6-G189V. To monitor Tif6 localization, a C-terminal GFP tag was integrated into the TIF6 genomic locus in wild-type RPL10 cells, rpl10-R98S cells, and rpl10-R98S cells harboring the suppressing NMD3-Y379D or TIF6-G189V alleles in the genome. Again, we saw that Tif6 was mislocalized to the cytoplasm in rpl10-R98S cells ( Fig 2D, compare row 2 to row 1), consistent with a failure in its release from pre-60S subunits. However, the localization of Tif6-GFP was restored in rpl10-R98S cells containing NMD3-Y379D ( Fig 2D, row 3). In addition, although wild-type Tif6-GFP was mislocalized to the cytoplasm in rpl10-R98S cells, Tif6-G189V-GFP localization was restored to the nucleus ( Fig 2D, row 4). Because Nmd3 is distributed throughout the cytoplasm during steady-state conditions [44], we monitored the localization of Nmd3 in strains containing the leptomycin B (LMB)-sensitive crm1-T539C mutation [45]. Treatment with LMB in this mutant background blocks Crm1-dependent nuclear export, and the subsequent nuclear accumulation of Nmd3 in wild-type RPL10 cells reflects its ability to be released from pre-60S subunits in the cytoplasm and shuttle back to the nucleus [10,11] (Fig 2E, row 1). As previously reported, Nmd3 could shuttle to the nucleus in wild-type RPL10 cells, but was distributed throughout the cytoplasm in rpl10-R98S cells ( Fig 2E, row 2) [40]. However, Nmd3-Y379D could shuttle to the nucleus in rpl10-R98S cells, and wild-type Nmd3 shuttling was restored in rpl10-R98S cells expressing Tif6-G189V ( Fig 2E, rows 3-4). These results microscopy in cells containing genomic TIF6-GFP and either WT RPL10 (AJY3937) or rpl10-R98S (AJY3938) as the sole copy. Suppressed rpl10-R98S cells contained either NMD3-Y379D (AJY3939) or TIF6-G189V-GFP (AJY3940) in the genome. Scale bar, 5μm. E) Nmd3 localization monitored by indirect immunofluorescence, following treatment with LMB, in P GAL -RPL10 cells containing an LMB-sensitive crm1 allele (AJY1958) and either wild-type RPL10 or rpl10-R98S plasmid. Suppressed rpl10-R98S cells contained either NMD3-Y379D (AJY3909) or TIF6-G189V (AJY3943) in the genome. Scale bar, 5μm.
demonstrate that the NMD3-Y379D and TIF6-G189V suppressors simultaneously restore the localization of both Tif6 and Nmd3. The ability of mutant Nmd3 to restore the release and nuclear shuttling of Tif6 was a surprising result because previous genetic analyses suggested that the release of Nmd3 was downstream of the release of Tif6 [28,32]. The results shown here are the first indication that a block in Nmd3 release could inhibit Tif6 release.

Two classes of TIF6 mutations affect Tif6 release by distinct mechanisms
Tif6 binds to the 60S subunit primarily through Rpl23/uL14 [21], and mutations such as Tif6-V192F, map to this interface and weaken the binding of Tif6 to the subunit [18,29,30]. Unlike Tif6-V192F, the four residues in Tif6 that are mutated in the rpl10-R98S suppressors cluster together on a surface of Tif6 that is close to, but distinct from, the Tif6-60S interface ( Fig 3A). The position of the rpl10-R98S suppressing mutations in the Tif6 structure suggests that they act by a different mechanism. Intriguingly, the rpl10-R98S suppressing mutations in TIF6 cluster in a region that interacts directly with Nmd3 [22] (see below). Because this distinct class of TIF6 alleles (and NMD3 alleles) could restore Tif6 release in rpl10-R98S cells, we wondered whether their mechanism(s) of suppression was specific to rpl10-R98S, or if they could suppress other mutants that block Tif6 release. As discussed earlier, rpl10-S104D mutants are also defective for Tif6 release. We transformed rpl10-S104D cells with either WT or mutant TIF6 or NMD3 vectors. While NMD3-Y379D and TIF6-G189V dramatically improved rpl10-R98S cell growth, neither mutant could suppress the rpl10-S104D growth defect ( Fig 3B). Depletion of Efl1 also prevents Tif6 release [18,29]. We transformed a glucoserepressible EFL1 strain (P GAL -EFL1) with either WT or mutant TIF6 or NMD3 vectors and plated onto glucose-containing medium. Again, we found that while TIF6-V192F could suppress Efl1 depletion, neither NMD3-Y379D nor TIF6-G189V suppressed the loss of Efl1 (Fig 3C). These results suggest that the mechanism(s) by which NMD3-Y379D and TIF6-G189V restored Tif6 release in rpl10-R98S cells is not a general Tif6-promoting mechanism, but specifically bypasses the rpl10-R98S defect.

Suppressors of rpl10-R98S alter the Nmd3-60S interaction
We recently resolved the structure of Nmd3 on the 60S subunit [22], providing important insights into how Rpl10, Nmd3, and Tif6 interact. Nmd3 consists of three domains: an eIF5Alike domain associates with the L1 stalk and occupies the ribosomal E site, an eL22-like domain occupies the P site and displaces the Rpl10 P site loop from its observed position in mature ribosomes (Fig 4A), while the N-terminal domain extends from the P site toward the SRL and contacts Tif6, providing a direct link between Nmd3 and Tif6 (Fig 4A and 4B). That we could resolve the N-terminal domain of Nmd3 only in the presence of Tif6 suggests that Tif6 stabilizes Nmd3 through their direct interaction. Importantly, all TIF6 alleles that suppress rpl10-R98S can be mapped to this Tif6-Nmd3 interface (Fig 4B). In addition, several of the rpl10-R98S suppressing mutations in NMD3 also occur in the zinc-binding N-terminal domain that contacts Tif6 (Table 1). Although the resolution of the N-terminal 39 residues of Nmd3 was not sufficient to unambiguously trace chains, it is likely that the mutated residues within this domain of Nmd3 affect its interaction with Tif6. In addition to mutations in the N-terminal domain of Nmd3, the majority of mutations that suppressed rpl10-R98S occurred in the eL22-and eIF5A-like domains of Nmd3, and mapped to the Nmd3-60S interface ( Table 1, Fig 4C). We therefore considered the possibility that these mutations in Nmd3 might alter its interaction with the 60S subunit. Although residue Y379, (mutated in the NMD3-Y379D suppressor) does not make a direct interaction with the ribosome, it is located between N378 and N380, which make hydrogen bonds to eL42 and 25S rRNA, respectively (Fig 4D). The substitution of aspartate for tyrosine at position 379 likely destabilizes these interactions, leading to suppression of rpl10-R98S. To test this idea, we made alanine substitutions in N378 and N380 to directly disrupt their interactions with the Tif6 residue V192 (green) is located at the interface between Tif6 and Rpl23, while Tif6 residues mutated in rpl10-R98S suppressors (magenta) are clustered in a nearby region distinct from the interface with the 60S subunit. (Assembled from PBD file 5ANB) B) 10-fold serial dilutions of AJY3373 (P GAL -RPL10) containing WT RPL10, rpl10-R98S, or rpl10-S104D vectors, and either empty vector or the indicated alleles of NMD3 on centromeric vectors or TIF6 alleles on high copy vectors. Cells were spotted onto glucose-containing selective media to repress genomic RPL10. C) Serial dilutions of the glucose repressible EFL1 strain AJY2981 (P GAL -EFL1) containing empty vector or the indicated TIF6 or NMD3 plasmids. Cells were spotted onto glucose-containing selective media to repress Efl1. https://doi.org/10.1371/journal.pgen.1006894.g003 Rpl10-R98S traps Nmd3 in the ribosomal P site interface showing hydrogen bonds between Nmd3 residues N378 and N390 with eL42 and 25S rRNA, respectively. E) 10-fold serial dilutions of AJY3373 (P GAL -RPL10) expressing WT RPL10 or rpl10-R98S, and either WT NMD3 or the indicated NMD3 mutant from vectors, as indicated. Cells were spotted onto glucose-containing selective media. F-G) Sucrose gradient sedimentation of Nmd3. Extracts were fractionated by sucrose gradient sedimentation and the position of Nmd3 in gradients was monitored by Western blotting using anti-ribosome and asked whether this mutant, NMD3-N378A, N390A, could suppress the rpl10-R98S growth defect. Indeed, the NMD3-N378A, N390A mutant suppressed the rpl10-R98S growth defect to a similar degree as NMD3-Y379D (Fig 4E).
To further test if the Nmd3 mutants identified as rpl10-R98S suppressors altered the interaction between Nmd3 and the 60S subunit, we compared the co-sedimentation of wild-type and mutant Nmd3 with 60S subunits. We tested the eIF5A-like domain mutant, NMD3-Y379D, and an N-terminal domain mutant, NMD3-C35G, both of which could complement in the absence of wild-type NMD3 (S1 Fig). We found that while wild-type Nmd3 sedimented predominantly in the 60S-containing fractions and was absent from the top of the gradient (indicating that very little free protein was present in cells), both Nmd3-Y379D and Nmd3-C35G mutant proteins were enriched in the free protein fractions at the top of the gradients (Fig 4F). A larger population of free protein relative to 60S-bound Nmd3 could result from a weakened affinity of Nmd3 for the 60S subunit, or from an increased rate of Nmd3 release. While sedimentation cannot distinguish between these possibilities, each scenario suggests an altered interaction between Nmd3 and the 60S subunit.
Nmd3-C35G is predicted to disrupt the stabilizing interaction between Nmd3 and Tif6. To further test the significance of this interaction, we asked if the rpl10-R98S suppressing TIF6 mutations had an impact on Nmd3 sedimentation. We compared Nmd3 sedimentation in cells expressing either wild-type or mutant TIF6 as the sole copy. We found that both mutants tested (TIF6-G189V and TIF6-P163L) could complement in the absence of wild-type TIF6 (S2 Fig). However, while the majority of Nmd3 was found in the 60S fraction in wild-type TIF6 cells, Nmd3 was distributed in 60S fractions as well as lighter fractions in TIF6-G189V and TIF6-P163L cells (Fig 4G). This altered Nmd3 sedimentation observed in TIF6 mutants suggests that, like rpl10-R98S suppressing mutations in NMD3, mutations in TIF6 alter the interaction of Nmd3 with the 60S subunit.

Nmd3 inhibits Efl1 activation on rpl10-R98S 60S subunits in vitro
In our recent Nmd3-60S structural work, we observed Nmd3 engaged with the L1 stalk in multiple L1 stalk positions, ranging from partially open to fully closed [22]. In the fully closed position, the eL22-like domain of Nmd3 occupies the ribosomal P site in a position that is incompatible with Sdo1 binding in the P site. Nmd3 binding in the P site should therefore prevent the Sdo1-dependent activation of Efl1. We tested this idea in vitro with purified components. It has been shown previously that Sdo1 activates the GTPase activity of Efl1 in the presence of 60S subunits [33]. We used this assay to ask if Nmd3 would inhibit Sdo1-dependent Efl1 activity. We found that although Nmd3 binds to 60S subunits and activates Lsg1 [22], the presence of Nmd3 did not inhibit the Sdo1-dependent activation of Efl1 (Fig 5A). We were unable to test the effect of Tif6 stabilization of Nmd3 because in our hands the addition of Tif6 completely inhibits Sdo1-dependent Efl1 GTPase activity (S3 Fig). The rpl10-R98S mutant blocks both Tif6 and Nmd3 release in vivo, and genetic suppression of this T-ALL associated mutant showed that mutations in Nmd3 restored Tif6 release. In addition, rpl10-R98S suppressing mutations alter the interaction of Nmd3 with 60S subunits, possibly destabilizing Nmd3 binding in the P site. Based on these observations, we considered that the Tif6 release defect in rpl10-R98S mutant ribosomes might be attributed to stabilization Nmd3 antibody. Anti-Rpl8 was used to monitor the position of 60S subunits. F) Extracts were prepared from AJY3249 (P GAL -NMD3) cells expressing WT (pAJ409) or mutant NMD3 (pAJ2805 or pAJ3609) plasmid as the sole copy. G) Extracts were prepared from AJY1700 (tif6Δ) cells expressing either wild-type (pAJ2846) or mutant TIF6 (pAJ2833 or pAJ3401) as the sole copy.
https://doi.org/10.1371/journal.pgen.1006894.g004 of Nmd3 in the P site, thereby blocking Sdo1 binding. We reasoned that the suppressing mutations identified in NMD3 could destabilize Nmd3 from the fully engaged state in the P site. We found that whereas the addition of Nmd3 to wild-type subunits had no effect on Sdo1-dependent activation of Efl1, the addition of Nmd3 to rpl10-R98S subunits inhibited Sdo1-dependent activation of Efl1 GTPase (Fig 5B, compare lanes 3 and 4 to lanes 7 and 8). In contrast, the addition of Nmd3-Y379D, which suppresses the rpl10-R98S mutant in vivo, restored Sdo1-dependent activation of Efl1 GTPase in the presence of rpl10-R98S subunits (Fig 5B, lane 9). Thus, a mutation that alters the interaction of Nmd3 with the subunit overcomes the defect in Efl1 activation. These results strongly suggest that the molecular defect of the T-ALL-associated rpl10-R98S mutation is stabilization of Nmd3 in the P site where it blocks productive binding of Sdo1.
We considered the possibility that the Nmd3-Y379D mutant failed to inhibit Efl1 activity due to a weakened affinity for 60S subunits. To test this, we used an in vitro assay for Nmd3-dependent Lsg1 activation. Because we previously showed that Nmd3 binding to 60S subunits was required for Lsg1 GTPase activation [22], we used Lsg1 activation as a proxy to test Nmd3 binding. We compared Lsg1 activation in the presence of either wild-type 60S subunits or rpl10-R98S 60S subunits and either wild-type Nmd3 or Nmd3-Y379D. We found that wild-type Nmd3 stimulated Lsg1 GTPase on both wild-type and rpl10-R98S 60S subunits (Fig 5C, lanes 4 and 8, Fig 6D). Interestingly, rpl10-R98S subunits supported slightly higher stimulation of Lsg1 GTPase activity compared to wild-type subunits. A similar increase in Lsg1 activity was previously observed in the presence of Tif6, which stabilizes Nmd3 in the fully engaged conformation where the eL22-like domain occupies the P site [22]. Thus, the increase in Lsg1 activation with rpl10-R98S subunits provides additional evidence that rpl10-R98S subunits trap Nmd3 in the fully engaged conformation. In contrast to wild-type Nmd3, Nmd3-Y379D supported only a low level of Lsg1 activation in the presence of wildtype 60S (Fig 5C, compare lanes 4 and 5). If Nmd3-Y379D simply had a reduced affinity for subunits, we would expect the reduced Lsg1 activity to be overcome at higher concentrations of Nmd3-Y379D, where more Nmd3 protein would be bound. However, Lsg1 activity was not recovered by increasing the concentration of Nmd3 (Fig 5D). Because the eIF5A domain of Nmd3 can bind to the L1 stalk independently of the eL22 domain binding in the P site, increasing concentrations of Nmd3 may saturate binding to the ribosome via the L1 stalk without increasing binding in the P site, explaining the lower activation of Lsg1 by Nmd3-Y379D on wild-type ribosomes. Interestingly, in the presence of rpl10-R98S subunits, Nmd3-Y379D stimulated Lsg1 to a greater extent than that observed for wild-type Nmd3 with wild-type 60S subunits (Fig 5C and 5D). This suggests that rpl10-R98S subunits might stabilize both wildtype and mutant Nmd3 in the P site, but Nmd3-Y379D can be repositioned in the presence of Sdo1. Together, these results argue strongly that the T-ALL-associated mutation rpl10-R98S traps Nmd3 in the P site. Mutations in Nmd3 that suppress rpl10-R98S weaken the interaction of Nmd3 with the P site, restoring the ability of Nmd3 to be retracted from the P site in the presence of Sdo1.

Discussion
Mutations in the ribosomal protein Rpl10 (uL16) can be drivers of T-ALL [40]. We previously showed that these T-ALL mutations, including rpl10-R98S, disrupt late steps in cytoplasmic In the fully closed L1 stalk position, the eIF5A domain of Nmd3 is engaged with the E site, while the eL22-like domain occupies the P site. This position is stabilized by the interaction between the N-terminus of Nmd3 with Tif6. We propose that, following Rpl10 loading, the linkage between Nmd3 and Tif6 is broken, destabilizing the N-terminus of Nmd3 and allowing Nmd3 retraction from the P site and Sdo1 binding.
https://doi.org/10.1371/journal.pgen.1006894.g006 maturation of the 60S subunit, preventing the release of Nmd3 and Tif6 [40]. However, in that work we did not identify the molecular mechanism that led to the block in Tif6 and Nmd3 release. Because genetic analysis had placed the release of Nmd3 after the release of Tif6 [28], we expected that mutations that enhanced the release of Tif6 would suppress rpl10-R98S. Indeed, we had shown that one such mutant, TIF6-V192F, suppressed some mutations within the P site loop of Rpl10 (rpl10-S104D). Surprisingly, we found here that TIF6-V192F could not suppress rpl10-R98S. Instead, we found that mutations that enhanced the release of Nmd3 could suppress rpl10-R98S. This was difficult to reconcile with our earlier genetic analysis. However, recent atomic structures of Nmd3 on the 60S subunit revealed that Nmd3 spans the joining face of the 60S subunit, interacting with Rpl1 (uL1) of the L1 stalk, stretching through the E and P sites, and interacting directly with Tif6 [22,23]. Notably, the domain of Nmd3 in the P site would sterically block Sdo1 binding and prevent Tif6 release. Based on this structure, we proposed that Nmd3 interaction with Tif6 must be broken to permit the retraction of Nmd3 from the P site by the opening of the L1 stalk [22]. This retraction would allow Sdo1 to bind in the P site to trigger the release of Tif6. Thus, a model emerged that the defect of rpl10-R98S may be a failure in retraction of Nmd3 from the P site (Fig 6).
The eL22-like domain of Nmd3 contains an extended loop that projects into the mouth of the polypeptide exit tunnel. Because the P site loop of Rpl10 and the loop of Nmd3 occupy a similar position, it was recently suggested that the loading of Rpl10 could potentially influence Nmd3 release [23]. Arginine 98 is at the base of the Rpl10 loop and interacts with the sugar phosphate backbone of residues 1126 and 1127 of 25S rRNA. Loss of this interaction is expected to destabilize the P site loop and alter the interaction between Nmd3 and Rpl10. Although this could, in principle, affect the loading of Rpl10, our results indicate that the R98S mutation prevents the retraction of Nmd3 from the P site rather than the loading of Rpl10. This model of rpl10-R98S function is strongly supported by the genetic and biochemical results reported here.
We identified mutations in NMD3 and TIF6 that were extragenic suppressors of rpl10-R98S. These mutations mapped to the interface between Nmd3 and the 60S subunit and between Nmd3 and Tif6 and altered the interaction of Nmd3 with the 60S subunit. In addition, mutations in Nmd3 engineered to specifically disrupt hydrogen bonding with eL42 and rRNA also suppressed rpl10-R98S. Thus, mutations that destabilized Nmd3 on the pre-60S subunit suppressed rpl10-R98S. From a genetic perspective, it may seem counterintuitive that mutations that weaken the affinity of Nmd3 for the ribosome can act as dominant suppressors. However, their dominant nature can be readily explained. For example, wild-type Nmd3 will become trapped on pre-60S subunits in the cytosol, preventing entry of those 60S subunits into the translating pool but also sequester wild-type Nmd3 on the stalled subunits. Mutant Nmd3 that can be released will recycle to the nucleus unimpeded and support ribosome production.
We were also able to recapitulate the genetic effects of rpl10-R98S and suppression by mutations in Nmd3 with a reconstituted system of purified components. For this work, we used the activation of Efl1 by Sdo1 as a proxy for productive Sdo1 binding. Whereas wild-type Nmd3 did not compete with Sdo1 for binding in the P site on wild-type ribosomes, Nmd3 inhibited Sdo1 binding in the presence of rpl10-R98S ribosomes. Furthermore, this inhibition was reversed by an Nmd3 mutant, Nmd3-Y379D, that suppresses rpl10-R98S mutant cells in vivo. These results identify the defect of rpl10-R98S-containing ribosomes as a failure to release Nmd3 from the P site, and show that suppressing mutations bypass the entrapment of Nmd3 by destabilizing interactions of Nmd3 with the 60S subunit.
We proposed previously that Nmd3 is retracted from the P site by opening of the L1 stalk [22]. But whether Nmd3 release is coincident with retraction of the stalk or is a subsequent step remains an unanswered question. The persistence of Nmd3 binding to the L1 stalk could explain the observation that lsg1 mutants, or the depletion of Lsg1, prevent recycling of Nmd3 but not Tif6 [22,31], leading to the conclusion that Nmd3 is released after Tif6. Although our results cannot definitively distinguish between these two models, several observations lead us to favor the latter. First, single particle reconstruction of the Nmd3-60S complex revealed Nmd3 bound to the L1-stalk in multiple conformations from partially open to fully closed, suggesting that Nmd3 interaction with Rpl1 is maintained independently of other interactions on the subunit [22]. And second, Nmd3-Y379D, which bypassed the inhibition of Efl1 activation on rpl10-R98S subunits, could efficiently stimulate Lsg1 GTPase on rpl10-R98S subunits. However, in the presence of wild-type 60S subunits, Nmd3-Y379D failed to activate Lsg1 to the same degree as did wild-type Nmd3, regardless of the concentration of protein. If Nmd3-Y379D simply had lower affinity for 60S, we would expect the defect in Lsg1 activation to be overcome at higher concentrations of Nmd3-Y379D. That we did not observe this is consistent with the interpretation that Nmd3-Y379D binding to the L1 stalk could be saturated, but its lower affinity for the P site prevented maximal activation of Lsg1. Thus, the Y379D mutation preferentially affects Nmd3 binding in the P site relative to overall subunit binding.
Thus far, our results imply a fairly simple model for suppression of rpl10-R98S-weakened interaction of Nmd3 for the subunit overcomes the retention of Nmd3 in the P site. This can be accomplished by mutations in Nmd3 or Tif6, all of which appear to reduce the affinity of Nmd3 for the ribosome. However, we would also expect mutations that weaken the affinity of Tif6 for the ribosome, such as TIF6-V192F, to suppress rpl10-R98S. That they do not suppress suggests additional subtleties to the mechanism of Nmd3 release. One interpretation is that bypass of rpl10-R98S requires stable association of Tif6 with the subunit. Although Efl1 does not appear to be required for the release of Nmd3 from wild-type subunits, it may assist in dislodging Nmd3 from the P site in rpl10-R98S mutant subunits.
The molecular defect of rpl10-R98S revealed in this work in yeast is likely relevant to human ribosomes in T-ALL patient cells. This conclusion is based on the high degree of conservation of Nmd3 and Rpl10-both human proteins can replace their yeast counterpart [12,46] and the amino acid sequence of Rpl10/uL16 is highly conserved across eukaryotes, with amino acid 98 invariantly an arginine. In addition, rpl10-R98S patients express only mutant RPL10 in the mutated cells, and rpl10-R98S inhibited ribosome biogenesis in mammalian cells, as we have observed in yeast [40]. However, how this defect in ribosome biogenesis ultimately promotes T-ALL is not understood.
We previously demonstrated that the rpl10-R98S mutation had a severe impact on yeast ribosomes at both the structural and functional level, and proposed that rpl10-R98S promotes defective translational fidelity [42]. Within the context of T-ALL, we suspect that rpl10-R98S is a mutation arising early in disease pathogenesis. Whereas translation fidelity defects might support disease development by shifting the spectrum of expressed proteins toward a more oncogenic profile, the ribosome biogenesis defect and resulting lack of functional ribosomes is probably not beneficial for pre-leukemic cells, likely imposing high pressure on these cells to acquire additional mutations that suppress the biogenesis defect. While suppression would reverse the proliferation defect, it may come at the risk of allowing defective ribosomes to be used in translation. So far, the nature of suppressor mutations acquired in rpl10-R98S positive T-ALLs is unclear. Mutations in NMD3 and eIF6 (human Tif6) have not been detected in human rpl10-R98S positive T-ALL cases [42,47], suggesting that alternative suppressor mutations occur in the human patients. Thus, a current challenge is to identify and characterize rpl10-R98S suppressing mutations in T-ALL positive cells.

Strains and plasmids
Yeast strains and plasmids used in this work are listed in Tables A and B of S1 Text. Oligos used in this work are listed in Table C of S1 Text. All cells were grown at 30˚C in rich media (yeast extract and peptone) or appropriate synthetic drop-out medium with 2% glucose or 1% galactose as the carbon source. Strains AJY2781 and AJY2784 were made by introducing plasmid pAJ2522 or pAJ2726 into the rpl10 deletion strain AJY1437 [32] by plasmid shuffle. Spontaneous suppressing mutations were identified in the AJY2784 background, including NMD3-Y379D (AJY2846) and TIF6-G189V (AJY2848). AJY2781, AJY2784, AJY2846, and AJY2848 were used to generate AJY3937, AJY3938, AJY3939, and AJY3940, respectively: pDEGQ2 was introduced into each strain, and the TIF6-GFP::His3MX cassette (amplified from TIF6-GFP cells (Open Biosystems) using AJO454 and AJO1384) was integrated into the genome by homologous recombination. AJY3909 was made by crossing AJY2846 and AJY1837 [31] and AJY3943 was made by crossing AJY2848 with AJY1958. Strain AJY1958 was made by crossing relevant mutants. To make AJY3941, the TIF6 locus was amplified in two parts from TIF6-GFP strain (Open Biosystems) genomic DNA, using mutagenic oligos to introduce the V192F mutation. The 5' half of TIF6-GFP was amplified using AJO453 and AJO933, and the remaining half of was amplified with AJO932 and AJO454, such that the two PCR products overlapped. PCR products derived from these reactions were used in a final reaction to make the full length TIF6-V192F-GFP PCR product, which was integrated into the genome of AJY3901(NatMX::P GAL -RPL10 Δtif6::KanMX) by homologous recombination. AJY3249 was made by amplifying the His3MX-PGAL1-3HA cassette [48] with homology to the NMD3 locus using AJO2031 and AJO2032, integrating into BY4743 (Open Biosystems), followed by sporulation and dissection.
To make pAJ2805, NMD3-Y379D was gap-rescued from the genome of AJY2846 using pAJ409 digested with SnaBI and HpaI. NMD3-Y379D was amplified from pAJ2805 using oligos AJO360 and AJO2035, digested with NdeI and SpeI restriction enzymes, and ligated into the same sites in pAJ2392. pAJ3609 was generated by inverse PCR of pAJ409 using oligos AJO2558 and AJO2554. Similarly, pAJ3581 was made by inverse PCR of pAJ409 using AJO2704 and AJO2705. pAJ729 was made by amplifying the TIF6 locus from genomic DNA using oligos AJO453 and AJO454. PCR product was digested with SalI and HindIII and inserted into the same sites in pRS426. pAJ2846 was made by digesting TIF6 sequence from pAJ2665 [32] with BamHI and XhoI and ligating into the same sites in pRS415. pAJ2828 was made by inverse PCR using oligos AJO1820 and AJO1821 with pAJ2665 as template. To make pAJ2833, the TIF6-G189V sequence was digested from pAJ2828 using SstI and XhoI enzymes, and ligated into the same sites in pRS415. pAJ3023 was made by digesting the TIF6 allele from pAJ2828 using XhoI and SacI enzymes, and ligating into the same sites in pRS426. pAJ3024 was generated by digesting TIF6-V192F from pAJ2240 [32] using XhoI and BamHI enzymes, and ligating into the same sites in pRS426. pAJ2839 and pAJ2840 were made by digesting RPL10 from pAJ2522/pAJ2726 using SacI and BamHI, and ligating into the same sites in pRS313. To make pAJ2982, genomic DNA was amplified using oligos AJO1413 and AJO1414, digested with NdeI and XhoI restriction enzymes, and inserted into the same sites in pET21a. The kemptide tag was added by inverse PCR using AJO1762 and AJO1763. To make pAJ3114, an internal 8HIS tag was integrated between EFL1 amino acids 458 and 459 in a pET21d vector containing EFL1 with its natural C-terminus.

Identification of spontaneous suppressors
Spontaneous suppressors of rpl10-R98S were identified in AJY2784. Individual slow-growing colonies were cultured independently and passaged several times before plating. After streaking for single colonies, large 'fast-growing' colonies were isolated and genomic DNA was prepared. The RPL10 allele was amplified (using AJO264 and AJO268) and sequenced to confirm that cells had not reverted to WT. The NMD3 allele was amplified (using AJO238 and AJO329) and sequenced to identify NMD3 suppressing alleles. The identified NMD3-Y379D suppressor was saved as strain AJY2846. The TIF6-G189V mutant was identified by highthroughput SOLiD sequencing of gDNA from a suppressor isolate, saved as strain AJY2848. Sequencing and SNP analysis was carried out by the Genome Sequencing and Analysis Facility at the University of Texas at Austin.

TIF6 mutagenesis
The open reading frame of TIF6 was randomly mutagenized by PCR using Taq DNA polymerase and oligos AJO453 and AJO454, with wild-type TIF6 plasmid pAJ2846 as template. PCR product was then co-transformed with gapped starting vector, to allow recombination of the mutagenized PCR products back into the starting vector, in AJY3901. Fast growing colonies were selected and TIF6-containing plasmids were extracted and sequenced. TIF6 mutant plasmids were then re-introduced into rpl10-R98S cells to confirm suppression.

Microscopy
For direct fluorescence experiments, cells were grown to saturation in selective medium containing galactose, then diluted back 20-fold in medium containing 2% glucose and grown for an additional hour to repress the expression of genomic RPL10. Images were captured using a Nikon E800 microscope fitted with a 100x Plan Apo objective and a Photometrics CoolSNAP ES camera controlled by NIS-Elements software. For indirect immunofluorescence, cells were grown in selective medium containing galactose before adding glucose to 2% to repress the expression of wild-type genomic RPL10 for 2h. LMB was added to a final concentration of 0.1 μg/ml for 15 min to block Nmd3 shuttling. Cells were fixed with a 1:9 volume of 37% formaldehyde for 40min, then washed with Ksorb buffer (0.1 M potassium phosphate [pH 6.6], 1.2 M sorbitol). Cells were permeabilized in cold methanol followed by washing in acetone. Anti-Nmd3 antibody [49] was diluted 3,000 fold in PBS with 0.1% BSA. Cy3-conjugated donkey anti-rabbit antibody (Jackson ImmunoResearch Laboratories, Inc.) was used at a 300-fold dilution. After antibody application, cells were incubated for 1 min in 1 μg/ml DAPI and mounted in Aqua-Poly/Mount.

Sucrose density gradient sedimentation and polysome profile analysis
For polysome profile analysis, cells were initially grown to saturation in the presence of galactose then diluted back to 3x10 6 cells/ml in glucose-containing media, and growth was continued until cultures reached 1.2x10 7 cells/ml. 100μg/ml cycloheximide (CHX) was added to each culture, followed by incubation for an addition 10min at 30˚C. Cultures were then immediately poured over ice and harvested by centrifugation. To monitor Nmd3 sedimentation, cells were grown only in glucose-containing selective media. Liquid cultures were grown to midlog phase, treated with CHX and harvested as previously described. Cell extracts were prepared at 0-4˚C: Cells were washed with lysis buffer (100mM KCl, 50mM Tris-HCl pH 7.5, 5mM MgCl2, 100μg/ml CHX, 6mM beta-mercaptoethanol (βME), 1mM PMSF, 1μg/ml leupeptin, 1μg/ml pepstatin A), and lysed by vortexing in the presence of glass beads. Extracts were clarified by centrifugation for 10 min at 15,000 g at 4˚C, and 9A 260 units of clarified extract were loaded onto 7-47% sucrose gradients, prepared in lysis buffer, and centrifuged for 2.5 hours at 40,000 rpm (Beckman SW40). Gradients were fractionated (ISCO Model 640) with continuous monitoring at 254nm. To monitor Nmd3 sedimentation, fractions were precipitated with 100% EtOH overnight at -20˚C, then centrifuged for 30min, 15,000 g at 4˚C. Pellets were resuspended in Laemmli buffer and boiled at 99˚C for 3 min. Proteins were separated on 10% SDS-PAGE gels, transferred to a nitrocellulose membrane, and subjected to Western blot analysis using anti-Nmd3 [49] and anti-Rpl8 (from K. Lo) antibodies.

GTPase assays
Specified amounts of Efl1, Sdo1, Lsg1, Nmd3, and 60S, as indicated in the figure legends, were mixed in a 20ul volume in 1x GTPase buffer (20mM HEPES-KOH, pH 7.4, 2mM MgOAc, 50mM KOAc, 1mMDTT) containing 25uM GTP. Reactions were spiked with approximately 1x10 5 cpm of [gamma -32P]-GTP to trace the hydrolysis of gamma phosphate. Reactions were incubated at 30˚C for 10 minutes and stopped by addition of 5ul of 0.5M EDTA. 1ul of each reaction was spotted on a TLC plate (PEI-cellulose, Sigma-Aldrich). Free phosphate was separated from GTP by developing the TLC plate for 10 minutes in 0.8M LiCl, 0.8M CH 3 COOH. Plates were imaged on a phosphoimager screen and signal intensities for free phosphate and GTP were analyzed using ImageJ software. All samples were corrected for non-enzymatic background hydrolysis. Reactions containing Nmd3 were corrected for background hydrolysis by free Nmd3 protein. Percent GTP hydrolysis was calculated as (free phosphate/total phosphate) and picomoles of phosphate release were calculated as Percent GTP hydrolysis Ã [GTP]. Picomole phosphate release values from Nmd3 titration assays were fitted to saturating-specific single site binding curves using Graphpad Prizm software.
MBP-TEV-His6-Nmd3: For WT Nmd3 1 liter of BJ5464 with pAJ1381 was grown to OD 600 of 0.6 in selective medium containing 2% glucose. For Nmd3-Y379D, 25 ml overnight culture of BJ5464 with pAJ2849 grown in selective medium with 2% glucose was diluted to OD 600 = 0.03 in 100ml of same medium and grown for 8-10 hours, then diluted to OD 600 = 0.03 in 500ml of selective medium with 1% raffinose and grown overnight at 30˚C. The culture was then induced at OD 600 = 0.4 by adding galactose to 1% final concentration. Both WT and mutant proteins were purified using the same method: cells were harvested, washed, and resuspended in two volumes of extract buffer (50mM Tris, pH 8, 450mM NaCl, 100 mM KCl, 10% glycerol, 1mM PMSF, 1μg/ml leupeptin, 1μg/ml pepstatin A). Cells were disrupted by vortexing with glass beads and crude extract was clarified by centrifugation at 4˚C, first for 10 min at 10,000 g and then for 20 minutes at 25,000 g. Imidazole was added to 10mM and clarified lysate was incubated with 0.5ml Ni-NTA Agarose beads (Invitrogen, R901-15) equilibrated in extract buffer for 1 hour at 4˚C. The beads were washed twice with 10ml of extract buffer containing 10mM and 20mM imidazole respectively and the protein was eluted in 0.25ml fractions of extract buffer supplemented with 250mM imidazole. Fractions containing protein were pooled and diluted in 9 volumes of Q buffer (40mM Tris-HCl, pH 7.0, 10% glycerol and 1mM DTT). Bound proteins were eluted with a 20ml linear NaCl gradient (50mM to 1M) in Q buffer. Fractions containing Nmd3 were pooled and dialyzed in 20mM Tris, pH7.5, 150mM KOAc, 1mM DTT and 10% glycerol. Dialyzed protein was stored at -80˚C. Efl1-internal His8: Yeast Efl1 protein with an internal 8xHis tag in domain II of the protein was purified from Codon plus (RIL) E. coli cells (Stratagene) with pAJ3114. 1 liter of bacterial culture was grown to OD 600 of 0.5 and induced with 1mM IPTG for 4 hours at 30˚C. Cells were harvested and washed with Lysis buffer (40mM Tris, pH 8.0, 500mM NaCl and 10% glycerol). The cell pellet was resuspended in 40ml Lysis buffer (supplemented with 5mM βME, 1mM PMSF and 1μM each leupeptin and pepstatin) and disrupted by sonication. Lysates were cleared by centrifugation at 20,000 g for 20 minutes. Imidazole was added to 10mM. Clarified lysate was bound to a 1ml Ni-NTA column (HisTrap HP, GE Healthcare) by pumping at 1ml/min. The column was first washed with 10ml lysis buffer supplemented with 10mM imidazole and 5mM βME and then with low salt buffer (40mM Tris, pH 8.0, 50mM NaCl, 10% glycerol, 20mM imidazole and 5mM βME). Bound protein was eluted with low salt buffer with 125mM imidazole. Fractions containing protein were pooled and bound to 2ml Source Q (GE Healthcare) column pre-equilibrated with buffer A (100mM Tris pH 8.0, 50mM NaCl, 2mM DTT and 0.2mM EDTA). Column was washed with 10ml buffer A. Bound proteins were eluted with a 20ml linear NaCl gradient (50mM to 1M) in buffer A. Fractions containing Efl1 were pooled and dialyzed in 20mM Tris, pH 7.5, 150mM KOAc, 1mM DTT and 10% glycerol. Dialyzed protein was stored at -80˚C.
Sdo1-kemptide-6His: Yeast Sdo1 protein with C-terminal kemptide and 6xHis tag was purified from Codon plus (RIL) E. coli cells (Stratagene) with pAJ2982. 1 liter of bacterial culture was grown to OD 600 of 0.5 and induced with 1mM IPTG for 4 hours at 30˚C. Cells were harvested and washed with Lysis buffer (40mM Tris, pH 8.0, 500mM NaCl and 10% glycerol). The cell pellet was resuspended in 40ml Lysis buffer (supplemented with 5mM βME, 1mM PMSF and 1μM each leupeptin and pepstatin) and disrupted by sonication. Lysates were cleared by centrifugation at 20,000 g for 20 minutes. Imidazole was added to 10mM. Clarified lysate was bound to a 2 ml Ni-NTA column (HisTrap HP, GE Healthcare) by pumping at 1ml/min. The column was first washed with 20ml lysis buffer supplemented with 10mM imidazole and 5mM βME and then with 20ml lysis buffer supplemented with 30mM imidazole and 5mM βME. Bound protein was eluted using lysis buffer with 250mM imidazole in 0.67ml fractions. Fractions containing Sdo1 protein were pooled and loaded on 150ml Sephacryl S-200 (Pharmacia) gel filtration column pre-equilibrated in S-200 buffer (40mM Tris-HCl pH 8.0, 150mM NaCl, 10% glycerol, 5mM βME). Bound protein was eluted in S-200 buffer. Fractions containing Sdo1 were pooled and dialyzed in 20mM Tris, pH 7.5, 150mM KOAc, 1mM DTT and 10% glycerol. Dialyzed protein was stored at -80˚C.

Purification of 60S subunits from yeast
Yeast strains AJY2781 and AJY2846 were grown in 2L of YPD to OD600 of 1.0. Cells were transferred to ice for 1h before harvesting by centrifugation at 4˚C. Cell pellets were washed and resuspended in 6ml of binding buffer (20mM HEPES-KOH, pH 7.6, 60mM NH 4 Cl, 5mM Mg(oAc) 2 , 2mM DTT, 1mM PMSF). Cells were disrupted by glass bead lysis and cleared by centrifugation at 30,000 g for 25 min at 4˚C. Active 80S ribosomes were purified from cell extracts using cysteine-charged sulfolink columns as described [41]. To dissociate subunits, ribosome pellets were resuspended in 1ml elution buffer (20mM HEPES-KOH, pH 7.6, 60mM NH 4 Cl, 500mM KCl, 10mM Mg(oAc) 2 , 2mM DTT, 1mM PMSF and 1uM each leupeptin and pepstatin) and incubated at 37˚C for 30 min after the addition of puromycin and GTP, each to 1mM final concentration. The sample was then centrifuged through 10-30% sucrose gradients in elution buffer for 4h at 32,000 rpm in an SW40 rotor (Beckman Coulter). Fractions containing the 60S and 40S peaks were pooled separately and concentrated using an Amicon Ultra-4 50 K (Millipore), and buffer was changed to Ribosome storage buffer (20mM HEPES-KOH, pH 7.6, 100mM KCl, 5mM Mg(oAc) 2 , 2mM DTT, 250mM sucrose).
Supporting information S1 Fig. Complementation of NMD3 mutants. 10-fold serial dilutions of AJY3249 (P GAL -NMD3) cells with empty vector, or vector containing either WT (pAJ409) or mutant NMD3 (pAJ2805 or pAJ3609). Cells were plated on galactose-containing media (left) to allow expression of WT NMD3 from the genome and compared to cells plated on glucose-containing media (right) to shut down expression of genomic NMD3. (TIFF) S2 Fig. Complementation of TIF6 mutants. 10-fold serial dilutions of AJY1700 (tif6Δ with P GAL10 -TIF6-myc URA3 plasmid) cells with empty vector or vector containing either wild-type (pAJ2846) or mutant TIF6 (pAJ2833 or pAJ3401). Cells were grown on galactose-containing media (left) to allow expression of the P GAL10 -TIF6-myc URA3 plasmid. The URA3 plasmid was shuffled out by plating cells on glucose-containing media with 5FOA (right).   Johnson.