Molecular basis for the distinct cellular functions of the Lsm1–7 and Lsm2–8 complexes

Abstract

Eukaryotes possess eight highly conserved Lsm (like Sm) proteins that assemble into circular, heteroheptameric complexes, bind RNA, and direct a diverse range of biological processes. Among the many essential functions of Lsm proteins, the cytoplasmic Lsm1–7 complex initiates mRNA decay, while the nuclear Lsm2–8 complex acts as a chaperone for U6 spliceosomal RNA. It has been unclear how these complexes perform their distinct functions while differing by only one out of seven subunits. Here, we elucidate the molecular basis for Lsm-RNA recognition and present four high-resolution structures of Lsm complexes bound to RNAs. The structures of Lsm2–8 bound to RNA identify the unique 2′,3′ cyclic phosphate end of U6 as a prime determinant of specificity. In contrast, the Lsm1–7 complex strongly discriminates against cyclic phosphates and tightly binds to oligouridylate tracts with terminal purines. Lsm5 uniquely recognizes purine bases, explaining its divergent sequence relative to other Lsm subunits. Lsm1–7 loads onto RNA from the 3′ end and removal of the Lsm1 carboxy-terminal region allows Lsm1–7 to scan along RNA, suggesting a gated mechanism for accessing internal binding sites. These data reveal the molecular basis for RNA binding by Lsm proteins, a fundamental step in the formation of molecular assemblies that are central to eukaryotic mRNA metabolism.

INTRODUCTION

The Lsm/Sm proteins are an ancient family of RNA binding proteins found in all three domains of life and have a wide range of biological functions. They are named after the autoimmune patient serum that led to their discovery (Tan and Kunkel 1966; Lerner and Steitz 1979). The Sm family includes the Sm, Lsm (“Like Sm”), and bacterial Hfq proteins (Wilusz and Wilusz 2013). These proteins share a conserved “Sm fold” consisting of an amino-terminal alpha helix followed by five antiparallel beta strands that form small beta barrels that assemble into ring-shaped hexamers or heptamers. The eukaryotic Sm proteins form heteroheptamers that interact with the major spliceosomal snRNAs U1, U2, U4, and U5, the minor spliceosomal snRNAs U4^atac, U11, U12, and telomerase RNA. The Sm-like archaeal proteins (SmAPs) are homologous to the eukaryotic Sm proteins, but their biological roles are less well understood (Mura et al. 2013). The eukaryotic Lsm proteins form at least four different six- or seven-subunit complexes (Wilusz and Wilusz 2013). In addition, a hybrid complex containing Lsm10, Lsm11, and five Sm proteins is essential for 3′ end processing of histone mRNAs (Dominski and Marzluff 2007; Sun et al. 2020).

The Lsm1–7 complex is cytoplasmic and mediates messenger-RNA (mRNA) decay, a major posttranscriptional mechanism for regulating gene expression (Bouveret et al. 2000; Tharun et al. 2000; Moore 2005; Parker 2012). Binding of the Lsm1–7 complex to mRNA is a key event in the decay pathway as it also binds the protein Pat1 (Chowdhury et al. 2007; Sharif and Conti 2013; Lobel et al. 2019). Pat1 then recruits a complex consisting of the decapping enzyme Dcp2 and its activators Dcp1, Edc1, and Edc2 (Mugridge et al. 2018). After decapping, the 5′–3′ exoribonuclease Xrn1 degrades the mRNA. Structures have been determined for the isolated S. cerevisiae Lsm1–7 complex (Sharif and Conti 2013; Zhou et al. 2014b), and the S. cerevisiae Lsm1–7 with a carboxy-terminally truncated Lsm1 bound to the carboxy-terminal domain of Pat1 (Sharif and Conti 2013; Wu et al. 2014). However, no structures are available for Lsm1–7 bound to RNA, despite the central importance of this interaction in the mRNA decay pathway. In addition to activating mRNA decapping, Lsm1–7 has many other functions including formation of phase-separated processing bodies (P-bodies) (Decker et al. 2007; Reijns et al. 2008), protecting 3′ ends of mRNA from 3′–5′ degradation by the exosome (He and Parker 2001; Tharun et al. 2005), stabilizing specific RNAs during starvation and autophagy (Gatica et al. 2019), suppressing translation of stress-activated RNAs during osmotic shock (Garre et al. 2018), and promoting translation of viral RNAs (Jungfleisch et al. 2015). Human and Schizosaccharomyces pombe Lsm1–7 complexes bind tightly to oligouridylate (hereafter, oligoU) RNAs (Lyons et al. 2014; Lobel et al. 2019). In contrast, the Lsm1–7-Pat1 complex binds tightly to oligoadenylate (hereafter, oligoA) RNAs (Chowdhury et al. 2007, 2014; Lobel et al. 2019). The molecular basis for these interactions is unknown.

The Lsm2–8 complex shares six out of seven subunits with Lsm1–7, localizes in the nucleus, and binds the 3′ ends of U6 and U6^atac snRNAs (Mayes et al. 1999; Reijns et al. 2009; Zhou et al. 2014a; Montemayor et al. 2018). U6 snRNA is transcribed by RNA polymerase III, which terminates transcription after synthesis of an oligoU tail at the end of U6 snRNA (Didychuk et al. 2018). This tail can then be elongated by the enzyme Tutase (Didychuk et al. 2018). Finally, U6 snRNA is processed by the 3′–5′ exoribonuclease Usb1 (Didychuk et al. 2018), resulting in a 2′,3′ cyclic phosphate in most organisms (Fig. 1; Lund and Dahlberg 1992). In addition, Lsm2–8 mediates nuclear mRNA decay (Kufel et al. 2004). In the case of S. pombe, Lsm2–8 is also known to play an important role in telomerase biogenesis (Tang et al. 2012).

FIGURE 1.

3′-processing of U6 snRNA alters its recognition by S. pombe Lsm2–8. (A) Processing of U6 by Usb1 leaves a 3′ cyclic phosphate group on U6, denoted as “>p.” (B) In vitro fluorescence polarization binding assays show that S. pombe Lsm2–8 preferentially binds RNA that has been processed by Usb1. All binding curves and K_d determinations in this work were performed with a restrained Hill coefficient of 1. For reference, the nonrestrained Hill coefficients, which are close to 1, are shown in the figure.

The molecular basis for the Lsm2–8 interaction with U6 snRNA terminating in a cyclic phosphate has yet to be elucidated, despite existing cryo-EM structures of the human U4/U6.U5 tri-snRNP (Agafonov et al. 2016) and pre-B complex (Charenton et al. 2019), both of which contain the Lsm2–8 complex bound to U6 snRNA. In these structures, the U6 snRNA-Lsm2–8 complex can only be resolved to resolutions of ∼10 Å, due to its location on the surface of these complexes and/or dynamic motions. The S. cerevisiae Lsm2–8 structure has been determined (Zhou et al. 2014a; Montemayor et al. 2018) but is unlikely to be representative of most eukaryotes due to significant differences in Lsm8 protein sequence and U6 snRNA posttranscriptional modification (Didychuk et al. 2017). Budding yeasts have Usb1 enzymes that open the 2′,3′ cyclic phosphate to produce a 3′ phosphate, and their Lsm8 proteins have evolved a unique carboxy-terminal extension in order to interact with this 3′ phosphate through electrostatic interactions (Didychuk et al. 2017; Montemayor et al. 2018). In contrast, the Lsm8 proteins of most eukaryotes, from S. pombe to humans, do not have this Lsm8 extension and have U6 snRNAs with terminal 2′,3′ cyclic phosphates. The relative specificity of the Lsm2–8 interaction with U6 versus other RNAs is also not well understood.

Here, we investigate the binding properties of Lsm1–7 and Lsm2–8 complexes from S. pombe. We show that Lsm2–8 specifically recognizes the terminal 2′,3′ cyclic phosphate while still retaining the ability to bind unmodified RNA. We reveal the molecular basis for Lsm2–8 binding by determining high-resolution structures of Lsm2–8 complexes bound to two RNAs, one that is unmodified and one that has a 2′,3′ cyclic phosphate. Next, we describe the RNA binding properties of the Lsm1–7 complex utilizing a panel of RNAs of varying lengths and sequences. We report two high-resolution structures of Lsm1–7 bound to high affinity RNA targets that explain the molecular basis of Lsm1–7-RNA recognition. We demonstrate that Lsm1–7 loads unidirectionally onto RNA from the 3′ end, using a binding mechanism in which specificity is bestowed by the Lsm1 carboxy-terminal region. Removal of this region enables 3′–5′ scanning and leads to a model for allosteric regulation of mRNA decay. In summary, the RNA binding activities of Lsm1–7 and Lsm2–8 are remarkably distinct despite similar quaternary structures that differ by only one subunit.

RESULTS

The Lsm2–8 complex specifically recognizes oligoU with a 2′,3′ cyclic phosphate

Schizosaccharomyces pombe Lsm proteins were coexpressed in E. coli and orthogonal affinity tags were used to purify the Lsm1–7 and Lsm2–8 complexes (Supplemental Fig. 1). For Lsm2–8, we demonstrated that the complex can bind to U6 snRNA and further associate with protein Prp24 to form the complete U6 snRNP, a known interaction previously observed for the S. cerevisiae U6 snRNP (Montemayor et al. 2018). We wished to determine if the Lsm2–8 ring specifically recognizes the 2′,3′ cyclic phosphate group at the end of U6 snRNA, and if so, how. We therefore used fluorescence polarization to measure and compare the binding affinities of Lsm2–8 for oligoribonucleotides corresponding to the 3′ oligoU tail of U6 terminating with a 2′,3′ cyclic phosphate group, a 3′ phosphate, or an unmodified 3′ hydroxyl (Fig. 1B; Supplemental Table 1).

We find that Lsm2–8 binds tightly to an RNA oligonucleotide with a 2′,3′ cyclic phosphate group, and fourfold less tightly to the unmodified RNA with a 3′ hydroxyl (K_d = 26 and 100 nM, respectively) (Fig. 1B). This fourfold difference suggests specific recognition of the 2′,3′ cyclic phosphate group. Interestingly, Lsm2–8 binds to an oligonucleotide terminating in a 3′ phosphate least tightly of all with a K_d of 152 nM. Thus, the Lsm2–8 complex preferentially binds to oligonucleotides terminating in a 2′,3′ cyclic phosphate group while still retaining ability to bind to unmodified RNA, and can somehow strongly discriminate between 2′,3′ cyclic phosphate and noncyclic 3′ phosphate groups.

Structures of Lsm2–8-RNA complexes

We crystallized the Lsm2–8 complex bound to UUUUU-3′-OH and UUUUU > p and determined their structures by X-ray diffraction to resolutions of 1.9 and 2.3 Å, respectively (Table 1). In the Lsm2–8 complex bound to the UUUUU-3′-OH RNA, the 5′ uridine is bound by Lsm4, and the next three uridines are bound sequentially by Lsm8, Lsm2, and Lsm3 (Fig. 2A,C). Uridine binding involves stacking and an extensive hydrogen bonding network as previously described (Fig. 2E; Zhou et al. 2014a; Montemayor et al. 2018). The last uridine with the 3′-OH group is disordered except for its 5′ phosphate as evidenced by weak electron density for both the ribose and uracil groups (Supplemental Fig. 2a). We confirmed that the RNA remains completely intact after prolonged incubation with Lsm2–8, further supporting the idea that the last uridine is covalently attached but disordered in the structure (Supplemental Fig. 2b).

TABLE 1.

Diffraction data collection and structure refinement statistics

FIGURE 2.

Structures of S. pombe Lsm2–8 bound to pentauridylate with a 2′,3′ OH or 2′,3′ cyclic phosphate. (A) Overview of Lsm2–8 bound to “unprocessed” U6 snRNA 3′ terminus. (B) Overview of Lsm2–8 bound to “mature” U6 snRNA 3′ terminus. (C) Detail of Lsm2–8 interface with “unprocessed” RNA. The 3′ terminal uridine is disordered and thus not visible in the final electron density maps. (D) In contrast, the mature U6 3′ end, with a 2′,3′ cyclic phosphate, shows electron density for the terminal nucleotide. (E) The Sm-like pocket in Lsm3 binds RNA as observed previously in other Lsm2–8 complexes from S. cerevisiae. (F) In contrast, the 3′ uridine cyclic phosphate has a unique binding mechanism relative to the other four uridines, including a stacking interaction with the carboxy-terminal histidine of Lsm8.

In contrast to the Lsm2–8 complex with unmodified RNA, the structure of the Lsm2–8 ring bound to UUUUU > p reveals that the last nucleotide is highly ordered (Supplemental Fig. 2c,d). The cyclic phosphate causes the RNA chain to make a sharp turn of nearly 180°, which positions the terminal nucleobase to stack over the last histidine at the carboxyl terminus of Lsm8. The terminal uracil base adopts an unusual syn conformation and forms a hydrogen bond to the Lsm3–Asp83 side chain (Fig. 2F). Lsm3–Arg27 forms a hydrogen bond to a nonbridging oxygen on the terminal cyclic phosphate group and makes a salt bridge with the Lsm8 carboxy-terminal histidine carboxyl group. This binding mechanism is markedly different from that in the S. cerevisiae Lsm2–8 complex with a noncyclic phosphate, where the terminal uridine adopts an anti conformation and lacks direct contacts with the corresponding arginine and aspartate residues (Supplemental Fig. 3; Montemayor et al. 2018).

RNA binding properties of the Lsm1–7 complex

Lsm1–7 shares six out of seven subunits with Lsm2–8 and has all but one of the uridine binding pockets present in Lsm2–8. The S. cerevisiae Lsm1–7 structure in the absence of RNA has been determined and displays an overall structure that is very similar to Lsm2–8 (Fig. 3). We therefore reasoned that Lsm1–7 and Lsm2–8 might bind to similar RNA sequences. However, when superimposing the structures of S. cerevisiae Lsm2–8 bound to RNA and Lsm1–7, there is steric clash between the 3′-end of the RNA and the carboxy-terminal domain of Lsm1 (Fig. 3B). This may explain why the S. cerevisiae Lsm1–7 ring alone has been reported to not bind to RNA (Chowdhury et al. 2014).

FIGURE 3.

The elongated carboxy-terminal region of Lsm1 in the S. pombe Lsm1–7 ring attenuates RNA binding. (A) Structure of the S. cerevisiae Lsm1–7 ring in the absence of bound RNA. The carboxy-terminal region of Lsm1 crosses the distal face of the ring, occluding the central pore. (B) Superposition of U6 snRNA from the homologous S. cerevisiae Lsm2–8 ring with the S. cerevisiae Lsm1–7 ring. The alignment in this figure was achieved by superposition of Lsm2, Lsm3, and Lsm6 components of the rings, which share a pairwise r.m.s.d of 0.5 Å. A steric clash is visible between the 3′ uridine phosphate and the carboxy-terminal region of Lsm1. (C) In vitro fluorescence polarization binding assays show that S. pombe Lsm1–7 tightly binds to polyuridylate tracts without accessory proteins, like Pat1. In contrast, S. pombe Lsm1–7 lacks detectable affinity for polyadenylate. (D) Designed truncations of the Lsm1 carboxyl terminus. Lsm1 is colored blue to red from the amino- and carboxyl-termini, respectively, and regions selected for truncation are annotated in the figure, resulting in truncation of the last 12 residues that fold back into the pore, or truncation of the entire helical region that spans the distal face of the ring. (E) Binding assays showing that deletion of the Lsm1 carboxy-terminal region generally enhances binding affinity for RNAs that harbor polyuridine tracts, and the relative enhancement is greatest for the weakest binding RNAs. (F) A strong preference for an adenosine 3′ terminus over a uridine cyclic phosphate 3′ terminus is diminished upon deletion of the carboxy-terminal 12 residues of Lsm1.

Since S. pombe Lsm1–7 has been observed to bind to oligoU RNAs as well as oligoA RNAs in the presence of Pat1 (Lobel et al. 2019), we first analyzed binding to four 15mers: U15, U14A, U13AA, and A15 under stringent binding conditions including 0.1 mg/mL competitor tRNA, 0.1 mg/mL sodium heparin, and 0.1 mg/mL BSA. We find that U15, U14A, and U13AA all bind tightly to Lsm1–7; in contrast, A15 does not bind in our assay (Fig. 3C). Owing to the observed potential for steric clash between RNA and the carboxyl terminus of Lsm1, we also compared RNA binding to a variant of Lsm1–7 in which the last 56 amino acids of Lsm1 were removed (Lsm1_Δ56C–7) (Fig. 3D). Lsm1_Δ56C–7 binds even more tightly to the oligoU RNAs and, like Lsm1–7, also does not bind to A15 (Fig. 3E). Upon observing high affinity binding of Lsm1–7 to oligoU RNAs, we assayed RNAs with shorter U-tracts for binding. Lsm1–7 and Lsm1_Δ56C–7 binds tightly to the oligo CCCCCUUUUUA (K_d = 29.7 and 13 nM, respectively), which is comparable to the binding observed for longer oligoU tracts (Supplemental Table 1). The oligo ACCCAUUUUU binds less tightly than U15, indicating that the length of the oligoU tract plays a role in binding affinity.

In general, the Lsm1_Δ56C–7 construct has higher affinity for RNAs than Lsm1–7 (Fig. 3E). When the oligoU tract is at or near the 3′ end, the difference in binding affinities between Lsm1–7 and Lsm1_Δ56C-7 is approximately twofold. However, if the oligoU tract is followed by two or more 3′ nucleotides, the binding enhancement afforded by the Lsm1 carboxy-terminal deletion becomes much more significant (five- to ninefold). For example, the oligonucleotide UUUUUACCCCC binds poorly to Lsm1–7 (K_d = 189 nM) and tightly to Lsm1_Δ56C–7 (K_d = 34 nM) (Fig. 3E). These data indicate that high affinity binding sites for the Lsm1–7 complex must be at the 3′ termini of RNA. On the other hand, the carboxy-terminally truncated Lsm1–7(Lsm1_Δ56C–7) complex can bind tightly to the UUUUA sequence irrespective of its position in the RNA. Thus, the Lsm1 carboxy-terminal extension functions to prevent high affinity binding to RNA, except for 3′ terminal sites. We further tested the effect of 3′-end nucleotide identity on binding and found that tetrauridylate followed by a single adenosine binds more tightly than an RNA with a pentauridylate tract (Fig. 3E).

Interestingly, we observe that the carboxy-terminal 12 residues of Lsm1 strongly discriminates against cyclic phosphate RNAs. Comparison of a monoadenylated RNA and a cyclic phosphate RNA shows an approximate 14-fold reduction in binding affinity for the cyclic phosphate RNA, which is substantially attenuated by deleting only a small region from the carboxyl terminus of Lsm1 (Fig. 3D,F). We conclude that the carboxy-terminal 12 amino acids of Lsm1 are important for the binding specificity of Lsm1–7.

Structures of Lsm1–7 bound to RNA

We were unable to obtain crystals of wild-type Lsm1–7 in complex with RNA. Since deletion of the carboxyl terminus of Lsm1 generally enhances binding affinity (Fig. 3E), we therefore crystallized the Lsm1_Δ56C–7 variant of the Lsm1–7 complex bound to the U-tract RNAs UUUUUA and AUUUUG and determined their structures to resolutions of 1.8 and 2.1 Å, respectively (Table 1; Fig. 4). In the structure of UUUUUA bound to Lsm1_Δ56C–7, the first uridine is disordered and not visible in the electron density (Fig. 4A,C; Supplemental Fig. 4). The next four uridines occupy pockets in Lsm4, Lsm1, Lsm2, and Lsm3, respectively. The terminal adenosine reaches across Lsm6 to form a hydrogen bond to Asn66 of Lsm5 (Fig. 4E). This is the first observation of Lsm5 interacting with RNA; Lsm5 is not utilized in the Lsm2–8 structures. Furthermore, Lsm5 is the only Lsm protein out of the eight studied here that has a noncanonical nucleobase binding pocket because it is missing the arginine that typically forms a cation-pi stack with uracil (Fig. 2E; Zhou et al. 2014a,b; Montemayor et al. 2018). In place of arginine, Lsm5 has an asparagine (Asn66) that forms a hydrogen bond to the terminal adenine. Interestingly, this noncanonical asparagine in the Lsm5 binding pocket is highly conserved (Supplemental Data File 4).

FIGURE 4.

The structure of S. pombe Lsm1–7 bound to RNA. (A) Overview of Lsm1–7 bound to UUUUA. (B) Overview of Lsm1–7 bound to AUUUUG. (C) RNA-binding interface of Lsm1–7 bound to UUUUA, showing that four uracil bases bind in the same manner as in the Lsm2–8 complex, while the adenosine binds into non-Sm pocket in Lsm5. (D) Lsm1–7 bound to AUUUUG, showing a similar binding mechanism as in C. (E,F) Detailed view of the non-Sm Lsm5 binding pocket occupied by adenine or guanine. In both cases, the 3′ purine is coordinated through hydrogen bonding with Lsm5–Asn66 and stacking with Lsm5–Asn68.

In the structure of Lsm1_Δ56C–7 bound to AUUUUG, all nucleotides are visible within the electron density map. The path of the RNA around the interior of the Lsm1–7 torus is a right-handed helix, with the phosphodiester backbone in the center and the nucleobases splayed out into binding pockets. The six nucleotides of the RNA make an almost complete turn with the last nucleotides 6 Å below the first (Fig. 4D,F). The 5′ adenosine interacts with the noncanonical binding pocket on Lsm5 by stacking on Tyr39 (Fig. 4F). The 3′ G also interacts with Lsm5 by forming a hydrogen bond to Asn66 in a manner that is similar to the interaction observed for the terminal adenine of UUUUUA, except the asparagine side chain donates a hydrogen bond to the guanine oxygen instead of an adenine nitrogen (Fig. 4E,F). The structures explain why oligonucleotides harboring these sequences bind with nearly identical affinities to Lsm1_Δ56C–7 (Fig. 3E).

Since the terminal purines are both within hydrogen bonding distance of Lsm5 Asn66 and are in van der Waals contact with Lsm5 Asn68 (Fig. 4E,F), we tested the contribution of these interactions to binding by mutating Lsm5 Asn66 and Asn68 to alanine. The double mutant has a small but measurable effect upon binding RNA (K_d = 23 vs. 41 nM) (Supplemental Table 1). We also tested the binding specificity of the double mutant by comparing its ability to bind an RNA terminating with a cyclic phosphate versus the terminal adenosine. While Lsm1–7 can discriminate between these RNAs with a 14-fold difference in affinity, the double mutant shows only a sevenfold difference, corresponding to a twofold loss in binding specificity (Supplemental Table 1).

Lsm1–7 loads onto RNA from 3′ ends and is blocked by secondary structure

In the crystal structures, the 5′ end of the RNA hexamer is nearest the “proximal face” of Lsm1–7 while the 3′ end is nearest the “distal face” (Fig. 5A), the same orientation as in the Sm ring where the bound RNA passes completely through the center of the torus. This similarity suggests that single stranded RNA may be able to pass through the Lsm1–7 torus. Moreover, we reasoned that binding may be affected by adjacent secondary structure, particularly if the complex were to load onto RNA in a directional manner. We therefore created RNA constructs with UUUUA binding sites and stem–loop structures at either the 5′ or 3′ termini or both (Fig. 5B). We find that both Lsm1–7 and Lsm1_Δ56C–7 can only bind tightly to the RNA with a 5′ stem–loop and single stranded 3′ end (K_d = 70 and 32 nM, respectively) (Fig. 5C). When the 5′ end is single stranded with a hairpin at the 3′ end, binding is severely weakened by 25-fold or more (K_d > 1 µM). This loss in binding affinity cannot be attributed to the 5′ cytidines, which do not affect binding to single stranded RNA (Fig. 3E). With hairpins at both ends, the binding is also severely weakened. We therefore conclude that both Lsm1–7 and Lsm1_Δ56C–7 load directionally onto the 3′ ends of single-stranded RNAs, and that binding is effectively blocked by downstream secondary structure. Comparison of surface electrostatics in Lsm1_Δ56C–7 shows the proximal face has more electropositive surface area than the distal face of the ring (Supplemental Fig. 5a), which may facilitate initial contact with RNA. Furthermore, comparison to the human Sm ring bound to U4 snRNA (Leung et al. 2011; Li et al. 2016) shows how directional threading of the RNA through the ring may be facilitated by conserved distal face contacts, as several residues in SmD1 and Sm2 that interact with RNA are also present in Lsm rings (Supplemental Fig. 5b).

FIGURE 5.

Lsm1–7 loads single stranded RNA from the 3′ end. (A) Cut-away view of the pore of Lsm1–7 bound to AUUUUG. Lsm3 and Lsm6 are omitted for clarity. (B) Synthetic RNAs used for binding tests, in which the 3′ or 5′ end is blocked from binding (or threading through) the pore of the Lsm1–7 ring by the presence of the hairpin structure. (C) Wild-type Lsm1–7 tightly binds the 5′ hairpin RNA and has weak binding for RNAs harboring a 3′ hairpin. Deletion of the carboxy-terminal region of Lsm1 does not significantly alter the binding specificity with respect to 3′ or 5′ hairpins.

DISCUSSION

Structural basis for RNA recognition by Lsm2–8

Phylogenetic analyses suggest that the Lsm/Sm family of proteins arose approximately 2.5 billion years ago through two waves of gene duplication, the first wave resulting in the Lsm genes from which the Sm proteins subsequently diversified (Veretnik et al. 2009). As Lsm and Sm proteins interactions are common to all spliceosomal RNAs, it is likely that these interactions arose early during the evolution of the spliceosome. The S. pombe Lsm2–8 complex is highly similar to human and other organisms (Supplemental Data File 4), so we expect the structural and functional data presented here to be broadly general in eukaryotic biology.

The molecular basis for Lsm2–8 binding is rationalized by the unique structural features observed in the Lsm2–8 complex bound to UUUUU > p. The 2′,3′ cyclic phosphate group bends the RNA chain around and stabilizes an unusual syn conformation of the terminal uracil base. Quantum chemical calculations of 2′,3′ cyclic UMP indicate that the cyclic phosphate stabilizes the syn conformation of uridine to be significantly more favored over the anti conformation (Grabarkiewicz and Hoffmann 2006). This syn conformation of the terminal uridine facilitates stacking on the Lsm8 carboxy-terminal histidine and allows for a hydrogen bond to Lsm3 Asp83, which could not happen in an anti conformation. Consistent with its central role in directly binding to the cyclic phosphate, Lsm3 Arg27 is 100% conserved (Montemayor et al. 2018). The carboxy-terminal histidine residue of Lsm8 that stacks with the terminal syn uracil nucleobase is also strongly conserved in almost all eukaryotes (Supplemental Data File 4). It is therefore clear that the carboxyl terminus of Lsm8 and nearby residues are important for specifically recognizing posttranscriptionally modified U6 snRNA (Didychuk et al. 2017; Montemayor et al. 2018).

Although S. pombe Lsm2–8 binds most tightly to oligoU with a 2′,3′ cyclic phosphate, it also must recognize heterogenous RNA sequences and different 3′-ends in vivo. For example, Lsm2–8 binds to telomerase RNA (TER1), which has a stretch of 3–6 uridines ending with a 3′-OH. Consistent with this, Lsm2–8 still binds to oligoU RNA ending in 3′-OH, with a K_d of 100 nM. We find that Lsm2–8 also binds tightly to 3′ monoadenylated RNA (Supplemental Table 1). Interestingly, we note that 3′ monoadenylated U6 is present in human cells and is the preferred substrate for Usb1 (Hilcenko et al. 2013; Shukla and Parker 2017; Nomura et al. 2018).

Unique RNA binding properties of the Lsm1–7 complex

In contrast to Lsm2–8, the Lsm1–7 complex strongly discriminates against RNAs with a 2′,3′ cyclic phosphate by 14-fold (Fig. 3F; Supplemental Table 1). This may be to avoid certain RNAs in the cytoplasm; for example, tRNA splicing proceeds through a cyclic phosphate intermediate (Shigematsu et al. 2018). Our binding measurements indicate that Lsm1–7 binds tightly to RNAs containing poly(U) or UUUUR and loads onto RNA from the 3′ end. This is consistent with previous reports of Lsm1–7 binding to UA rich regions of viral genomes to regulate translation (Galao et al. 2010; Jungfleisch et al. 2015). Lsm1–7 binds tightly to the sequence AUUUUR (Fig. 3E), and this sequence is highly reminiscent of sequences found in 3′-UTR regions that undergo adenine/uridine-rich element (ARE) mRNA decay (Barreau et al. 2005; von Roretz et al. 2011). Consistent with this, knockdown of Lsm1 inhibits ARE-mediated decay (Lin et al. 2007) and depletion of the Lsm1–7 binding protein Pat1b up-regulates mRNAs with AREs (Vindry et al. 2017).

Lsm1–7 preferentially loads onto single-stranded RNA from the 3′ end. The trajectory of the RNA in Lsm1_Δ56C–7 suggests that the RNA can thread completely through the ring, provided the Lsm1 carboxy-terminal helix is removed or displaced. Indeed, our binding data show that RNAs with high affinity binding sites followed by additional 3′ nucleotides (e.g., AAA or CCCCC) bind well to Lsm1_Δ56C–7 but not Lsm1–7. This is likely due to steric clash involving the Lsm1 carboxy-terminal region (Fig. 3B). The Lsm1 carboxy-terminal region may act as a gate to allow threading of the RNA through the ring, followed by scanning via a one-dimensional search process, or sliding (Berg et al. 1981; Halford and Marko 2004). Accordingly, there are conserved residues in the Lsm and Sm rings, the latter of which have previously been observed to interact with “threaded” single stranded RNA (Supplemental Fig. 5).

A model for Pat1 stimulation of RNA binding by Lsm1–7

Taken together, our data suggest an allosteric model for Pat1-mediated stimulation of Lsm1–7 binding (Chowdhury et al. 2014; Lobel et al. 2019). In this model, Pat1 binding displaces the Lsm1 carboxy-terminal domain to stimulate RNA binding affinity and relax specificity (Fig. 6). This model is consistent with existing structural data indicating that the carboxy-terminal domain of Pat1 binds to Lsm2 and Lsm3 (Sharif and Conti 2013; Wu et al. 2014), close to where the Lsm1 carboxy-terminal helix reaches across to dock on the other side of the ring (Fig. 6A). This positioning would place the middle domain of Pat1, which is not present in the structure but is required for high affinity binding to oligoA RNA (Lobel et al. 2019), very close to the carboxy-terminal helix of Lsm1. Our model is also consistent with the observation that mutation of the Pat1 interaction surface with Lsm2–Lsm3 destroys the ability of Pat1 to stimulate the RNA-binding activity of S. pombe Lsm1–7 (Lobel et al. 2019) and leads to defects in mRNA degradation in S. cerevisiae (Sharif and Conti 2013; Wu et al. 2014). In S. cerevisiae, the carboxy-terminal alpha-helical extension of Lsm1 is required for high affinity binding to oligoA RNA in the presence of Pat1 (Chowdhury et al. 2012).

FIGURE 6.

Proposed model for Pat1-regulated gating of the Lsm1–7 RNA-binding pore. (A) Relative orientations of RNA, Lsm1–7 and Pat1 proteins, derived from structures of S. pombe Lsm1–7 with RNA and S. cerevisiae Lsm1–7 with the carboxy-terminal domain of Pat1 (Sharif and Conti 2013). (B) Model for RNA binding specificity in Lsm rings. The Lsm2–8 ring contains a uridine-cyclic phosphate binding site that includes the carboxyl terminus of Lsm8 and endows specificity for Usb1 processed U6 snRNA. (C) In contrast, the carboxy-terminal region of Lsm1 in the Lsm1–7 ring antagonizes association of the ring with uridine-cyclic phosphate terminated RNA, while allowing association of the ring with 3′ monopurinated polyuridylate tracts. (D) Deletion of the carboxy-terminal region of Lsm1 allows the ring to associate with a broader range of RNAs. In vivo, the role of Lsm1 truncation may be mimicked by displacement of the carboxyl terminus by association of the ring with the Pat1 cofactor. It remains to be seen if RNA is capable of threading all the way through the pore of the RNA, as depicted here, or rather enters and exits the pore on the proximal face alone.

In summary, the Lsm proteins are essential RNA binding proteins that initiate the formation of molecular assemblies involved in major pathways of gene expression, including pre-mRNA splicing and mRNA decay. Understanding how these proteins recognize their RNA targets is therefore an important aspect of eukaryotic biology. We quantitatively demonstrate that the Lsm1–7 and Lsm2–8 complexes achieve strikingly different functional properties, despite similar quaternary structures and sharing of subunits. By elucidating the structures of these complexes bound to RNA, we establish the molecular basis for RNA–protein interactions that are fundamental to eukaryotic gene expression.

MATERIALS AND METHODS

Production of recombinant Lsm2–8

Codon optimized open reading frames (Genscript) for S. pombe Lsm proteins were cloned into modified variants of the pQLink expression system (Scheich et al. 2007). In the initial single open reading frame plasmids, Lsm2, Lsm5, and Lsm8 lacked purification tags. Lsm3 and Lsm6 harbored amino-terminal glutathione S-transferase and maltose-binding tags, respectively, with an intervening tobacco etch virus (TEV) protease site. Lsm4 and Lsm7 harbored noncleavable strep-II or hexahistidine tags, respectively, at their carboxy-termini. After cloning individual ORFs into individual plasmids, the multi-ORF expression system was assembled into a single plasmid through ligation independent cloning (Scheich et al. 2007) to yield a final plasmid with ORFs assembled in order Lsm6, Lsm3, Lsm2, Lsm8, Lsm4, Lsm7, and Lsm5. The sequence of the final plasmid was confirmed by Sanger sequencing.

The resulting S. pombe Lsm2–8 expression plasmid was transformed into E. coli BL21(DE3) STAR pLysS cells (Invitrogen # C602003). Cells were grown in terrific broth (RPI # T15100-5000.0) containing ampicillin with shaking at 37°C until the OD₆₀₀ was approximately 2, at which point isopropyl β-d-1-thiogalactopyranoside was added to a final concentration of 1 mM and the cells were allowed to continue growing with shaking at 37°C for another 24 h, at which point the OD₆₀₀ was approximately 16. Cells were harvested by centrifugation and the resulting cell pellet was resuspended in 0.025 volumes of buffer A (500 mM NaCl, 50 mM HEPES acid, 50 mM imidazole base, 10% glycerol, 1 mM TCEP, final pH ∼ 7) per original volume of cell culture (i.e., 100 mL of resuspension buffer per 4 L of original cell culture). Protease inhibitors (EMD Millipore # 539137-10VL) were added, as were DNase I (Sigma # DN25-100MG) and lysozyme (Sigma # 62971-10G-F) at a final concentration of 0.02 and 0.5 mg/mL, respectively. The resuspended cells were subjected to a single freeze-thaw cycle prior to sonication and clearance of cell debris by centrifugation.

The soluble fraction was first purified by immobilized metal affinity chromatography (Qiagen # 30230) using gravity flow at room temperature. Buffer A was used for washing and buffer A supplemented with 500 mM imidazole pH ∼ 7 was used for step elution. The pooled eluate was then dialyzed overnight at 4°C with 20 kDa MWCO membranes (Pierce # 66012) into buffer B (100 mM NaCl, 10 mM HEPES acid, 10 mM sodium HEPES, 10% glycerol, 1 mM TCEP-HCl, pH ∼ 7.0). Centrifugation was used to remove any precipitation that formed during dialysis. Soluble protein was further purified via glutathione agarose chromatography (GenScript # L00206) with gravity flow at room temperature in fresh buffer B. Step elution was performed in buffer B supplemented with 50 mM HEPES acid, 50 mM sodium HEPES, and 10 mM reduced glutathione. One milligram of TEV protease was added to the eluate, which was then dialyzed overnight as above, but at room temperature against 1 L of buffer C (150 mM NaCl, 50 mM tris base, 50 mM tris-HCl, 1 mM trisodium EDTA, 1 mM TCEP-HCl, 1 mM sodium azide, pH ∼ 8.0). Dialyzed protein was then purified using gravity flow chromatography at room temperature on Strep-Tactin agarose (IBA GmBH # 2-1208-025) in fresh buffer C, with step elution in buffer C supplemented with 2.5 mM biotin. The resulting eluate was manually applied at room temperature to a 1 mL HiTrapQ anion exchange column (GE Healthcare # 29051325) that had been preequilibrated in buffer B. The column was then attached to an AKTA chromatography system at 4°C and elution accomplished by applying a linear gradient of NaCl up to 2 M in buffer B. The Lsm2–8 complex desorbed from the column at ∼250 mM NaCl. Peak fractions were collected, pooled, and the concentration of the purified complex was determined by UV absorbance and the estimated molar extinction coefficient of 46,300 M⁻¹ cm⁻¹ at 280 nm (Wilkins et al. 1999). Final purity of the Lsm2–8 complex was determined by 20% (29:1) tris-tricine SDS-PAGE. The presence of all protein components was further confirmed by electrospray ionization mass spectrometry at the UW-Madison Biotechnology Center Mass Spectrometry Facility, where the amino-terminal methionine of Lsm5 and Lsm8 was found to be missing, and Lsm7 lacked either 1 or 2 amino-terminal residues. Protein samples were either stored at 4°C or frozen as 100 µL aliquots in liquid nitrogen before long-term storage at −80°C.

Production of recombinant Lsm1–7

The Lsm1–7 complex was produced as described for Lsm2–8, with the following exceptions. The multi-ORF expression system was assembled into a single plasmid through ligation independent cloning as described for Lsm2–8, with the ORFs assembled in order Lsm6, Lsm3, Lsm2, Lsm1, Lsm4, Lsm7, and Lsm5. Lsm7 lacked a purification tag.

Protein expression, purification and storage were performed as with the above Lsm2–8 complex, with the exception that Strep-Tactin chromatography was not used and instead the eluate from glutathione agarose chromatography was loaded directly onto a HiTrapQ column after overnight incubation with TEV protease at room temperature. Peak fractions were collected, pooled, and the concentration of the purified complex was determined by UV absorbance and the estimated molar extinction coefficient at 280 nm (Wilkins et al. 1999). Electrospray ionization mass spectrometry showed that all protein chains were full-length with the exception of Lsm5 and Lsm7, which lacked 1 and 1–2 residues from their amino-termini, respectively.

In order to prepare the Lsm5–N66A/N68A or Lsm1 carboxy-terminal truncation mutants of Lsm1–7, a modified inverse PCR mutagenesis protocol was used on the multi-ORF Lsm1–7 expression plasmid. PCR amplification was performed in “GC” optimized Pfusion buffer (NEB # M0532S) using a gradient of annealing temperatures and oligonucleotides that anneal only within the unique open reading frame regions of each Lsm locus. Amplicons of correct length were gel purified prior to phosphorylation and ligation. All mutant clones were confirmed to be of correct length and sequence by analytical restriction enzyme digestion and Sanger sequencing.

Production of recombinant Prp24

A codon optimized open reading frame (Genscript) for S. pombe Prp24 was cloned into a modified variant of plasmid pET3a, encoding an octahistidine tag, biotin acceptor peptide sequence (Beckett et al. 1999), and tobacco etch virus (TEV) protease site located upstream of the Prp24 ORF. The sequence of the final plasmid was confirmed by Sanger sequencing.

The S. pombe Prp24 was produced as described above for the Lsm proteins with the following exceptions. Cells were harvested by centrifugation and the resulting cell pellet was resuspended in 0.1 volumes of buffer A (500 mM NaCl, 50 mM HEPES acid, 50 mM imidazole base, 10% glycerol, 1 mM TCEP, final pH ∼ 7) per original volume of cell culture (i.e., 100 mL of resuspension buffer per 1 L of original cell culture).

Soluble protein was manually applied at room temperature to a 5 mL HiTrapQ anion exchange column (GE Healthcare # 29051325) that had been preequilibrated in buffer B. The column was then attached to an AKTA chromatography system at 4°C and elution accomplished by applying a linear gradient of NaCl up to 2 M in buffer B. The peak fractions were collected, pooled, diluted twofold against fresh buffer B and then manually applied at room temperature to a 5 mL HiTrap Heparin cation exchange column (GE Healthcare # 17040701) that had been preequilibrated in buffer B. The column was then attached to an AKTA chromatography system at 4°C and elution accomplished by applying a linear gradient of NaCl up to 2 M in buffer B. The peak fractions were pooled and the final concentration of protein was determined by UV absorbance and the estimated molar extinction coefficient of 114,600 M⁻¹ cm⁻¹ at 280 nm (Wilkins et al. 1999). Protein samples at ∼6 mg/mL were either stored at 4°C, or frozen as 100 µL aliquots in liquid nitrogen before long-term storage at −80°C.

Synthesis and purification of RNA

In vitro transcription with T7 RNA polymerase was used to synthesize mature S. pombe U6 nucleotides 1–100 (Tani and Ohshima 1989) from a linearized pUC57 plasmid template harboring a T7 promoter (TTCTAATACGACTCACTATA) and a minimal 56 nucleotide HDV ribozyme “drz-Mtgn-3” (Riccitelli et al. 2014) to ensure a homogenous 3′ end with a cyclic phosphate. A 20 mL transcription reaction contained ∼0.25 mg/mL linearized plasmid template, ∼0.5 mg/mL T7 RNA polymerase, 5 mM each of ATP, GTP, CTP and UTP; 100 mM tris, 100 mM Tris-HCl, 40 mM MgCl₂, 5 mM DTT, 1 mM spermidine trihydrochloride, and 0.01% (v/v) Triton X-100. Synthesis of RNA was performed at 37°C for 2 h. Trisodium EDTA pH 8.0 was then added to a final concentration of 100 mM to halt transcription and solubilize accumulated magnesium pyrophosphate. The transcription was then concentrated to ∼400 µL with 10 kDa MWCO spin filters (Amicon # UFC901008) prior to addition of 6 mL of 100% formamide and subsequent electrophoresis on a denaturing 10% (19:1) polyacrylamide gel containing containing 8 M urea, 100 mM tris base, 100 mM boric acid, and 1 mM EDTA acid. Full-length U6 RNA was identified by UV shadowing, extracted by scalpel, and removed from the gel matrix by passive diffusion overnight at room temperature with gently shaking into a solution containing 300 mM sodium acetate, 50 mM HCl, 1 mM EDTA, 1 mM sodium azide, pH ∼ 5.6. Soluble RNA was separated from solid acrylamide by filtration through 0.22 micron filters and then manually applied to a 5 mL HiTrap Q column (GE Healthcare # 29051325) that had been equilibrated in buffer D (300 mM NaCl, 10 mM KH₂PO₄, 10 mM K₂HPO₄, 1 mM EDTA acid, 1 mM sodium azide, pH ∼ 7). Bound RNA was washed with 20 mL of buffer D and then eluted in a single step with buffer D adjusted to 2 M NaCl. Eluted RNA was pooled and concentration and buffer exchange were accomplished by three iterations of 10-fold concentration (Amicon # UFC501008) and redilution against buffer E (100 mM KCl, 20 mM bis-tris, 10 mM HCl, 1 mM EDTA, 1 mM sodium azide, pH ∼ 6.5). The final RNA concentration was estimated using UV absorption and an anticipated extinction coefficient of 1,223,990 M⁻¹ cm⁻¹ at 260 nm (Kibbe 2007).

In order to generate a pentauridylate RNA with a 3′ cyclic phosphate for cocrystallization with S. pombe Lsm2–8, an RNA oligo with sequence 5′-UUUUUA-3′ was purchased from Integrated DNA Technologies (IDT) and treated with human Usb1, which rapidly removes 3′ adenosine residues from oligoU tracts and leaves a 3′ cyclic phosphate that is not subjected to ring opening as with S. cerevisiae Usb1 (Didychuk et al. 2017; Nomura et al. 2018). Human Usb1 was prepared exactly as described elsewhere (Nomura et al. 2018). The conversion protocol involved resuspension of an RNA pellet from IDT in buffer E to a final concentration of 2.7 mM. A limiting amount of enzyme was used to ensure the product RNA was predominantly that of a single cleavage reaction. This was accomplished by adding 180 µL of RNA at 2.7 mM to 180 µL of human Usb1 at 32 µM in buffer F (100 mM NaCl, 10 mM HEPES acid, 10 mM sodium HEPES base, 40% glycerol, 1 mM TCEP-HCl, 1 mM trisodium EDTA, pH ∼ 7) and incubation at 37°C for 1 h. The RNA was purified by 20% polyacrylamide denaturing gel electrophoresis and anion exchange chromatography as above, with the exception that HiTrapQ wash buffer contained 100 mM NaCl, 10 mM HEPES acid, and 10 mM sodium HEPES base, and the step elution buffer contained 1 M instead of 2 M NaCl. The eluted RNA was pooled and not subsequently adjusted prior to addition to Lsm2–8 (see below). The concentration of RNA was determined using an approximate molar extinction coefficient of 50,000 M⁻¹ cm⁻¹ at 260 nm (Kibbe 2007). All other RNAs used for cocrystallization experiments were purchased from IDT and resuspended in buffer E without further modification prior to addition to protein (see below).

The 5′-FAM labeled RNAs used for fluorescence polarization (S. pombe U6 nucleotides 91–100 and similar) were purchased from Integrated DNA Technologies and purified by urea PAGE and ion exchange as above for full-length U6, with the following exceptions: UV shadowing was not used (or required) to visualize the RNA after denaturing electrophoresis, concentration of RNA after anion-exchange chromatography was performed with 3 kDa MWCO spin filters (Amicon # UFC500308), and the final RNA concentration was estimated using UV absorption and an anticipated extinction coefficient of 75,000 M⁻¹ cm⁻¹ at 495 nm (Kibbe 2007). In the case of the cyclic phosphate probe, a limiting amount of human Usb1 was added to an adenosine terminated RNA as above.

Fluorescence polarization binding assays

All fluorescence polarization binding assays were performed in buffer H (100 mM NaCl, 20 mM tris pH 8.2, 10 mM MgCl₂, 1 mM TCEP HCl, 1 mM sodium azide, 0.1 mg/mL tRNA [Roche # 12172120], 0.1 mg/mL BSA [Ambion # AM2616], and 0.1 mg/mL sodium heparin [Sigma # N4784-250MG], pH ∼ 8) in black 96 well microplates (Greiner Bio-One # 655209) and imaged on a Tecan Infinite M1000Pro using an excitation wavelength of 470 nm and emission wavelength of 519 nm. For each sample, 100 µL of RNA at 2 nM was added to 100 µL of protein at a defined concentration between 0.4 nM and 5 µM. Fluorescence polarization was measured in duplicate from two independent titrations using different protein concentrations. Binding curves were fit using nonlinear regression in GraphPad Prism 4 to the following four parameter equation: FP = FP_min + (FP_max − FP_min)/(1 + 10^((logK_d − log[protein])*H)), where FP_min and FP_max are the minimum and maximum polarizations, K_d is the binding dissociation constant, and H is the Hill coefficient. H was constrained to be 1 during nonlinear regression. Depicted binding curves are normalized to FP_min and FP_max. All raw binding data are provided in Supplemental Data File 1.

Crystallization and structure determination of Lsm2–8 complexes

The Lsm2–8/RNA complexes were reconstituted by adding crude RNA from IDT to protein in an approximately twofold stoichiometric excess. The complexes were then dialyzed overnight at 4°C against 1 L of buffer I (50 mM NaCl, 10 mM MgCl₂, 10 mM tris-HCl, 10 mM bis-tris base, 1 mM TCEP-HCl, pH ∼ 7) using 20 kDa MWCO membranes (Pierce # 66012). The dialyzed complexes were concentrated to ∼10 mg/mL with 50 kDa MWCO spin filters (Amicon # UFC505008) prior to high-throughput crystallization screening on a Mosquito crystallization robot (TTP Labtech). Initial crystallization hits were obtained exclusively using the precipitant pentaerythritol propoxylate (5/4 PO/OH) in a MIDAS screen (Molecular Dimensions # MD1-59). Crystals were optimized by hanging drop vapor diffusion at 16°C, using 2 µL of Lsm2–8/RNA complexes mixed with 2 µL of crystallization reagent containing 100–200 mM KCl, 50–100 mM HEPES pH 7.4, 35% pentaerythritol propoxylate (5/4 PO/OH). Crystals were vitrified by direct immersion into liquid nitrogen.

Diffraction data were integrated using XDS (Kabsch 2010). Space group determination was performed in POINTLESS (Evans 2011). STARANISO (Tickle et al. 2017) was used for merging and ellipsoidal truncation of the anisotropic diffraction data. Phenix.xtriage was used to assay potential twinning in the diffraction data (Adams et al. 2010). Initial phases were determined by molecular replacement using Phaser (McCoy et al. 2007).

For the 3′ diol terminated structure (PDB 6PPN), three diffraction data sets were collected from two isomorphous crystals at 100 K on beamline 24-ID-E at the Advanced Photon Source. Molecular replacement was used to obtain initial phases with initial search templates PDB 4EMG (S. pombe Lsm3) (Wu et al. 2012), PDB 4EMH (S. pombe Lsm4) (Wu et al. 2012), PDB 4EMK (S. pombe Lsm5/6/7) (Wu et al. 2012), and homology models (Marti-Renom et al. 2000; Zimmermann et al. 2018) of S. pombe Lsm2 and Lsm8 that were constructed from the corresponding S. cerevisiae orthologs (PDB 4C92 and 4M7D, respectively) (Sharif and Conti 2013; Zhou et al. 2014a). Molecular replacement used a single search model in which the above seven proteins were placed in a fixed orientation relative to one another to resemble the known architecture of the S. cerevisiae Lsm rings (Sharif and Conti 2013; Zhou et al. 2014a; Montemayor et al. 2018). Structure refinement was performed in Phenix.refine (Adams et al. 2010; Afonine et al. 2012) using secondary structure restraints and TLS parameterization, with iterative rounds of manual model building in Coot (Emsley and Cowtan 2004; Emsley et al. 2010) and additional automated refinement in Phenix.refine.

The 2′,3′-cyclic phosphate terminated structure (PDB 6PPP) was determined as above, but using addition of a different RNA prior to dialysis and using the above S. pombe Lsm2–8 structure for molecular replacement. Twelve diffraction data sets were collected from two isomorphous crystals at 100 K on beamline 21-ID-D at the Advanced Photon Source. Structure refinement was performed in Phenix.refine (Adams et al. 2010; Afonine et al. 2012) using secondary structure restraints and TLS parameterization, in combination with reference model restraints to the best resolved Lsm2–8 ring in PDB 6PPN.

Crystallization and structure determination of Lsm1–7 complexes

Schizosaccharomyces pombe Lsm1–7 complexes lacking the carboxy-terminal 56 residues of Lsm1 were reconstituted with RNA as above, adding 5′-UUUUUA-3′ for PDB 6PPQ or 5′-AUUUUG-3′ for PDB 6PPV. Crystals were obtained by mixing 0.2 µL of complex with 0.2 µL of the following mixture: 20 mM sodium formate, 20 mM ammonium acetate, 20 mM trisodium citrate, 20 mM sodium potassium tartrate, 20 mM sodium oxamate, 100 mM sodium HEPES base, 100 mM MOPS acid, 10% PEG 8000, and 20% ethylene glycol. Crystals were vitrified by direct immersion into liquid nitrogen.

The 3′ adenosine terminated structure (PDB 6PPQ) was determined by merging three diffraction data sets collected from a single crystal at 100 K on beamline 21-ID-D at the Advanced Photon Source. PDB 6PPN was used for molecular replacement. Structure refinement was performed in Phenix.refine (Adams et al. 2010; Afonine et al. 2012) using secondary structure restraints. The 3′ adenosine binding pocket was first identified by residual Fo–Fc density after modeling and refining tetrauridylate into the four typical Sm-like pockets in Lsm1–7. Subsequent to incorporating the 3′ adenosine, the final electron density maps exhibited residual Fo–Fc density that could not be remediated by deletion of the adenosine or changing the identity of the adenosine to uridine. We therefore conclude the remaining Fo–Fc density is due to unmodeled dynamics in the 3′ adenosine binding pocket.

The 3′ guanosine terminated structure (PDB 6PPV) was determined from a single diffraction data set collected from a single crystal at 100 K on beamline 21-ID-D at the Advanced Photon Source. PDB 6PPQ was used for molecular replacement and refinement was conducted as above. The 3′ guanosine binding pocket did not exhibit residual Fo–Fc density as above for adenosine.

For all structures presented here, simulated annealing omit maps were prepared to confirm the presence of bound RNA in the final models deposited into the Protein Data Bank. All figures were generated with PyMOL (http://www.pymol.org). Structural biology applications used in this project were compiled and configured by SBGrid (Morin et al. 2013). Electrostatic surface potentials were calculated using APBS (Baker et al. 2001) as implemented in PyMOL. All final coordinate sets and structure factors with calculated phases are provided in Supplemental Data File 2. A PyMol session with annotation matching that used throughout the manuscript is provided as Supplemental Data File 3.

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.