Yeast Spt6 Reads Multiple Phosphorylation Patterns of RNA Polymerase II C-Terminal Domain In Vitro

Abstract

Transcription elongation factor Spt6 associates with RNA polymerase II (RNAP II) via a tandem SH2 (tSH2) domain. The mechanism and significance of the RNAP II–Spt6 interaction is still unclear. Recently, it was proposed that Spt6-tSH2 is recruited via a newly described phosphorylated linker between the Rpb1 core and its C-terminal domain (CTD). Here, we report binding studies with isolated tSH2 of Spt6 (Spt6-tSH2) and Spt6 lacking the first unstructured 297 residues (Spt6ΔN) with a minimal CTD substrate of two repetitive heptads phosphorylated at different sites. The data demonstrate that Spt6 also binds the phosphorylated CTD, a site that was originally proposed as a recognition epitope. We also show that an extended CTD substrate harboring 13 repetitive heptads of the tyrosine-phosphorylated CTD binds Spt6-tSH2 and Spt6ΔN with tighter affinity than the minimal CTD substrate. The enhanced binding is achieved by avidity originating from multiple phosphorylation marks present in the CTD. Interestingly, we found that the steric effects of additional domains in the Spt6ΔN construct partially obscure the binding of the tSH2 domain to the multivalent ligand. We show that Spt6-tSH2 binds various phosphorylation patterns in the CTD and found that the studied combinations of phospho-CTD marks (1,2; 1,5; 2,4; and 2,7) all facilitate the interaction of CTD with Spt6. Our structural studies reveal a plasticity of the tSH2 binding pockets that enables the accommodation of CTDs with phosphorylation marks in different registers.

Graphical abstract

Highlights

• •High-affinity Pol II CTD-binding by Spt6 is achieved by avidity originating from multiple phosphorylation marks presented in the CTD, suggesting how phosphorylation levels fine-tune the CTD interactome.
• •Structure of RNAP II CTD bound with tandem SH2 domain of Spt6 reveals how phosphorylated CTD is recognized.
• •Isolated tSH2 of Spt6 binds the extended CTD substrate with tighter affinity than nearly full-length Spt6, suggesting that the steric effects of additional domains in Spt6 influence the binding of the tSH2 domain to the multivalent CTD ligand.

Introduction

RNA polymerase II (RNAP II) is composed of 12 subunits, where the largest Rpb1 has a flexible C-terminal domain (CTD). This unique domain of RNAPII is highly repetitive with the consensus motif Y1-S2-P3-T4-S5-P6-S7 [1]. The number and conservation of the heptads within CTD vary across species [[2], [3], [4]]. RNAP II clusters in membraneless organelles through a process mediated by the properties of the CTD that promote liquid–liquid phase separation [[5], [6], [7]]. The CTD phosphorylation stimulates the recruitment of various factors involved in the transcription, mRNA processing, and histone modifications [8]. For example, Ser2 phosphorylation rises as transcription progresses, and these phospho-CTD marks recruit termination factors at the polyadenylation site. Phosphorylation at Tyr1 suppresses the binding of termination factors, but stimulates the binding of elongation factor Spt6 [9].

Spt6 is a highly conserved essential nuclear protein [10], which is often associated with RNAP II [11,12], Spn1/Isw1 [13], and/or histones [14]. Spt6 controls normal chromatin structure [15], as well as nucleosome occupancy and positioning [16,17]. In addition, it prevents histone loss, maintains locus-specific modifications [18], and regulates H3K36 methylation [19]. During transcription, Spt6 co-localizes with RNAPII at a genome-wide scale [12,20,21], where it redeposits nucleosomes back after RNAPII passage [15,22]. Several studies also suggested a role of Spt6 in transcription initiation [17,23,24]. However, Spt6 is traditionally recognized as a transcription elongation factor [20], which stimulates the transcription rate both in vitro [25] and in vivo [26]. Additionally, Spt6 controls mRNA 3′ end processing [27] and mRNA export [28].

The C terminus of Spt6 contains a tandem SH2 (tSH2) domain. The Spt6-tSH2 domain is composed of two SH2 domains, the N-terminal (N-SH2) and C-terminal (C-SH2) domains, which intimately pack against each other. The Spt6-N-SH2 domain resembles the canonical SH2 domain including its phosphorylation recognition binding site, while the Spt6-C-SH2 is non-canonical [29,30]. Structurally, both Spt6-N-SH2 and Spt6-C-SH2 conform to the standard SH2 domain fold. The structure of free tSH2 has been solved by X-ray crystallography and NMR, and is conserved across the various organisms in which it has been determined: Antonospora locustae [29], Candida glabrata [30], Saccharomyces cerevisiae [31], and Homo sapiens [32]. Spt6 lacking the tSH2 domain results in a slow growth phenotype, transcription elongation defects, and genome-wide reduction of Spt6-RNAP II complexes over actively transcribed regions [30,33]. Spt6-tSH2 was demonstrated to be recruited to phosphorylated CTD both in vivo [20] and in vitro [28]; however, the association of tSH2 with a CTD linker rather than CTD itself has been recently reported [21]. The linker binds Spt6-tSH2 more tightly than short CTD peptides likely due to the presence of a large overall negative charge [21,32]. However, short synthetic peptides used as CTD models contain less phosphorylations than the full-length CTD. Such phosphorylations provide a highly dynamic and combinatorial regulatory network, which can regulate protein–protein interactions [34]. Moreover, the presence of multiple binding motifs may tighten the overall interaction by several orders of magnitude. For example, Tudor domains SMN and SPF30 bind dimethylarginine with milimolar affinities, but the interaction is strengthened to the micromolar range by the presence of multiple methylated residues [35].

To address the mechanism and significance of CTD–Spt6 interactions, we studied the binding of Spt6-tSH2 and nearly full-length Spt6 (Spt6ΔN) with CTD substrates. We also determined the structure of Spt6-tSH2 bound to tyrosine phosphorylated CTD diheptad (pY1-CTD)₂ by NMR. The data show that the doubly phosphorylated peptide is accommodated in an extended conformation and is recognized by Spt6 tSH2 via a two-pronged mechanism. Furthermore, our work reveals a plasticity of the tSH2 binding pockets enabling the accommodation of CTD with phosphorylations in different registers, and that the strength of the interaction increases with the phosphorylation level of CTD diheptad substrates.

Results

Avidity of CTD phosphorylations stimulates Spt6-tSH2 binding

Earlier reports indicated an interaction of the tSH2 of Spt6 with the phosphorylated CTD of RNAP II from a pull-down and other in vitro binding experiments [9,[28], [29], [30], [31]]. Only two CTD heptad repeats prepared by solid-phase synthesis have previously been used as a model of phosphorylated CTD in quantitative binding experiments with Spt6. These experiments yielded a micromolar binding affinity between (pY1-CTD)₂ and Spt6-tSH2 (a K_D of ~ 3 μM) [9,30]. Here, we compared the binding affinities of (pY1-CTD)₂ and a long CTD fragment harboring 13 repetitive heptads of the CTD with approximately 13 phosphorylations at Y1 (prepared by the co-expression of c-Abl kinase and CTD [36]) to the tSH2 domain of C. glabrata Spt6 (Spt6-tSH2) and Spt6 from S. cerevisiae lacking the first unstructured 297 residues (Spt6ΔN; Figure 1(a)). The binding affinities were assayed using fluorescence anisotropy (FA) and microscale thermophoresis (MST) in conditions with a low salt buffer as reported previously [9,30]. The salt concentration only had a moderate effect in a binding study of Spt6-tSH2 with the CTD peptide (Figure S1). Remarkably, (pY1-CTD)₁₃ exhibits 2 orders of magnitude stronger binding to Spt6-tSH2 than (pY1-CTD)₂ (Figure 1(b); K_Ds 15.2 ± 0.9 nM versus 5.1 ± 0.4 μM). These enhanced nanomolar affinities are reached through avidity, originating from multiple phosphorylation marks present in the CTD. Equilibrium binding experiments with Spt6ΔN display identical binding affinity for the (pY1-CTD)₂ substrate as the isolated tSH2 of Spt6 (a K_D of 4.0 ± 0.5 μM; Figure 1(c)). However, Spt6ΔN exhibits a somewhat lower binding affinity for (pY1-CTD)₁₃ compared to the isolated tSH2 of Spt6, in the high nanomolar range (a K_D of 119 ± 28 nM; Figure 1(c)). In summary, our data suggest that short CTD peptides designed as minimal binders somewhat underestimate the avidity effects induced by the increased local concentration of multiple phosphorylations present in the repetitive motifs of the CTD (Figure 1(b)). The data also suggest that steric effects of additional domains in the Spt6ΔN construct influence the binding of the tSH2 domain to the multivalent ligand (pY1-CTD)₁₃.

Figure 1.

The role of avidity in RNAP II CTD binding. (a) Full-length Rpb1, highlighting its CTD and repetitive sequences. The indicated CTD and Spt6 constructs were used in this study. Spt6ΔN (S. cerevisiae) lacks the first unstructured 297 residues, (residues 298–1451). Spt6-tSH2 (C. glabrata; residues 1250–1444). Spt6-N-SH2 (C. glabrata; residues 1250–1354). (b) FA (left) and MST (right) titration of (pY1-CTD)₂ and (pY1-CTD)₁₃ with Spt6-tSH2 yielded K_D of 5.1 ± 0.4 μM and 15.2 ± 0.9 nM (in red), respectively. The controls with non-phoshorylated (CTD)₂ and (CTD)₁₃ are shown (in blue). (c) FA (left) and MST (right) titration of (pY1-CTD)₂ and (pY1-CTD)₁₃ with Spt6ΔN yielded K_D of 4.0 ± 0.5 μM and 119 ± 28 nM (in red), respectively. The controls with non-phoshorylated (CTD)₂ and (CTD)₁₃ are shown (in blue). Points represent the mean of three independent experiments, and error bars are the standard deviation between replicates.

Effects of phospho-CTD patterns for Spt6-tSH2 binding

Dynamic changes to the phosphorylation patterns in the CTD vary the affinity of different factors toward RNAP II during the transcription cycle, as the CTD phosphorylation patterns either stimulate or prevent binding. For example, termination factors containing the CTD-interacting domain (CID) bind only to specific phospho-CTD patterns, whereas other patterns (e.g. those with Y1 and S2 + T4 phosphorylations) prevent their binding [9,37,38]. In this way, termination machineries are not recruited in a gene body where the Y1 phosphorylation levels are high. Interestingly, some CTD phosphorylations have previously been shown to stimulate the binding of the tSH2 domain of Spt6 [9,31]. Here, we used a large library of CTD peptides with different phosphopatterns and evaluated their binding to Spt6-tSH2 using quantitative binding experiments. Interestingly, we found that Spt6-tSH2 binds various phosphorylation patterns in the CTD and found that the studied combinations of phospho-CTD marks (1,2; 1,5; 2,4; and 2,7) all facilitate CTD interaction with Spt6. (Table 1). This promiscuous binding of Spt6-tSH2 contrasts with other CTD binding modules, whose binding is usually weakened or abrogated by one or more phospho-CTD patterns [9,37,38]. In vitro binding data from FA are clustered according to the phosphorylation levels (0, 1, 2, 4 phosphorylations; Figure 2(a)). The data suggest that Spt6-tSH2 does not discriminate the type of phosphorylated residue of the CTD, nor the position or distance between multiple phosphorylated residues. For example, the (PS YpSPpTSPS YSPTSPS) peptide has two shortly spaced phosphorylations but has roughly the same affinity as peptides in which two phosphorylations are more apart from each other (e.g. (pY1-CTD)₂, (pT4-CTD)₂). This suggests a large degree of flexibility or different mechanisms of how the peptides can be accommodated at Spt6-tSH2. Similarly, the data also show that the affinity of the Spt6-tSH2–CTD interaction does not correlate to the number of CTD repeats, but rather to the number of phosphorylated residues present.

Table 1.

Binding affinities of Spt6-tSH2 for various peptides using FA

	KD (μM)
(CTD)2	375.0 ± 69
(pY1-CTD)1	38.7 ± 2.4
PS(YSPTSPS)(pYSPTSPS)	36.8 ± 2.6
PS(pYSPTSPS)(YSPTSPS)	45.4 ± 1.7
PS(YpSPpTSPS)(YSPTSPS)	5.7 ± 0.4
(pY1-CTD)2	5.1 ± 0.4
(pT4-CTD)2	7.2 ± 0.3
(pY1pS5-CTD)2	0.3 ± 0.1
(pY1pS2-CTD)2	0.3 ± 0.1
(pS2pS7-CTD)2	1.0 ± 0.1
NIM	3.3 ± 0.3

Figure 2.

Promiscuity of Spt6-tSH2 interaction with various CTD peptides. (a) Equilibrium binding for Spt6-tSH2 with various peptides monitored by FA. The peptides are divided into groups with regard to the number of phosphorylated residues present. NIM (Nrd1 Interacting Motif) is a RNAP II phospho-CTD motif mimic. The symbols and color coding of data points for the individual peptides are shown. The binding curves are fitted by non-linear regression to a single binding site model [39]. Standard errors are generated from a set of three independent measurements. (b) Correlation between experimental and in silico (Faunus) calculations of K_D with respect to the peptide net charge. The trendline (brown) is a guideline for the relation between values of K_D determined by experimental and theoretical approaches (blue triangles). Peptide net charge with respect to the experimental K_D also shows a linear relation (red asterisk). The peptides used in the calculation (from top right to left down): PSYSPTSPS, PS(pY)SPTSPS with charge − 2, a set of doubly phosphorylated diheptads (PSpYSPTSPSpYSPTSPS; PSYpSPTSPSYpSPTSPS, PSYSPpTSPSYSPpTSPS; PSYSPTpSPSYSPTpSPS) with charge − 4, NIM (DDEDGYNPYTL) with charge − 5, a set of quadruply phosphorylated diheptads (PSpYpSPTSPSpYpSPTSPS; PSpYPSTpSPSpYSPTpSPS) with charge − 8, Rpb1 linkers from Sdano et al. [21] (pS1493; pT1471 + pS1493; pY1473 + pS1493) with charges − 8 and − 9, respectively. For more details, see Materials and Methods.

Hill and co-workers recently reported a strong interaction of Spt6-tSH2 with Rpb1 linker, with a K_D of ~ 4 nM [21], the strongest binding of Spt6 so far. Based on our results above, the strength of the binding might be explained by its large negative charge (− 9 e). To support our hypothesis, we assayed a non-native substrate Nrd1 Interaction Motif (NIM). NIM is a phospho-CTD mimic from the Trf4 protein with the sequence DDDEDGYNPYTL, implicated in the transcription termination of non-coding genes in S. cerevisiae [40]. The NIM peptide has a negative charge of (− 5 e) and yielded a similar binding affinity to the native CTD ligands (Table 1; Figure 2(a)). Moreover, we found a linear correlation between both experimental and in silico dissociation constants and the net charge of the ligand (Figure 2(b)). In silico dissociation constants were calculated using the Monte Carlo method with a coarse-grained model, which was shown to correctly capture the electrostatic protein interactions (for details, see the Materials and Methods section) [41,42]. Thus, it is likely that both the CTD and the linker are involved in the recruitment of Spt6.

Structure of Spt6-tSH2 in complex with phospho-CTD

Even though several groups have worked on obtaining the complex of Spt6-tSH2–CTD using X-ray crystallography [[29], [30], [31], [32]], none of their attempts were successful. However, they identified several key residues located on the N- and CTD of the Spt6-tSH2 that are highly conserved and, therefore, were suggested to play a role in the interaction with the CTD (Figure 3(e)).

Figure 3.

Structure of Spt6-tSH2 in complex with (pY1-CTD)₂ peptide. The structure of Spt6-tSH2 in complex with (pY1-CTD)₂ is shown in ribbon representation. The Spt6-N-SH2 and Spt6-C-SH2 subdomains are colored in dark cyan and dark magenta, respectively. (a) and (b) are related by a 180° rotation around the vertical axis. The residues responsible for interaction with the CTD peptide are shown as sticks, and the phosphoresidue-binding pockets are indicated with cyan circles. The CTD is shown as sticks (yellow). (c) Spt6-tSH2 structure with electrostatic surface generated by Chimera v1.12. The CTD peptide position is shown as ribbons for the major conformation (in yellow) and the minor conformation (in red). The binding pockets #1–3 are indicated with black circles. (d) Overlay of Spt6-tSH2 domains bound to the CTD (solved here) and Rpb1 linker (1468–1500; PDB ID 5vko). The linker peptide, the CTD peptide, and the structure of Spt6-tSH2 are shown in blue, yellow, and gray, respectively. The N and C termini of the protein and peptide are indicated. (e) Alignment of Spt6-tSH2 amino acid sequence with corresponding regions in other eukaryotic Spt6 sequences (S. glabrata, S. cerevisiae, D. renio, H. sapiens, X. tropicalis). The residues involved in the interaction with the CTD within binding pockets #1 and #2 (black stars) compared to those described by Sun et al. [30] (black triangles) and Sdano et al. [21] (black dots). The additional binding pocket #3 (black stars) identified in this study is shown. The sequence alignment was produced with Clustal Omega v1.2.4 [43] and colored with ESPrint Server v3.0 [44].

Due to the unique binding characteristics of Spt6-tSH2 to phospho-CTD peptides, we set out to investigate this phenomenon at atomic level using NMR spectroscopy. We determined the solution structure of the tSH2 domain (1249–1444) of Spt6 (Figure 3(a) and (b)) in complex with a double repeat of the CTD phosphorylated on Y1 ((pY1-CTD)₂; PSpYSPTSPSpYSPTSPS). A combination of NOE distance restraints and PRE (paramagnetic relaxation enhancement) data was used to generate the structural model (see Materials and Methods). A total of 2080 intramolecular, structurally meaningful restraints derived from the 3D ¹³C- and ¹⁵N-edited nuclear Overhauser enhancement spectroscopy (NOESY) experiments and 31 intermolecular contacts identified in the 3D F1-¹³C/¹⁵N-filtered NOESY-[¹³C,¹H]-heteronuclear single quantum coherence (HSQC), were used in the calculation of the structure of the complex. An ensemble of 20 structures with the lowest energy were selected as the representative ensemble (see Table S1). The overall shape and aggregation state of the NMR complex was further validated with SAXS data. The SAXS experimental data fit well to the theoretical back-calculated scattering data from the NMR structure with 1:1 stoichiometry (Figure S2H). The N- and C-terminal SH2 domains are arranged in a tail-to-head manner and are connected via a rigid linker that forms a long α-helix (continuous connection αB^N-αA^C). The overall fold of both domains resembles the canonical SH2 fold, consisting of three antiparallel β-sheets and two α-helices. Two additional short antiparallel β-strands (βΕ^N-βF^N and βΕ^C-βF^C) are inserted between the β-sheets and the terminal α-helices. These additional structural elements provide interdomain contacts that stabilize the mutual orientation of the N- and C-terminal SH2 domains. The tandem architecture is further stabilized by the long α-helix, which provides a number of hydrophobic residues to create a hydrophobic core of the tSH2. Next, the two domains are stabilized by contacts between the βC^N-βD^N and βB^C-βC^C loops. Although both the N- and C-terminal SH2 domains resemble the canonical fold of the SH2 domain, they represent an integral fold that cannot be separated. We were only able to express and purify the N-terminal SH2, whereas the C-terminal SH2 failed to adopt the correct fold in the absence of its N-terminal partner. The architecture of tSH2 with tight contacts between the N- and C-terminal SH2 domains is unique, as other consecutive SH2 domains are either loosely associated or connected by flexible linkers (Figure S2C and D).

The (pY1-CTD)₂ peptide is accommodated in an extended conformation and is recognized by Spt6-tSH2 via a two-pronged mechanism. The phosphate group of the upstream pY1 is inserted deeply into the canonical pocket of Spt6-N-SH2 (Pocket #1) and contacts the side-chains of residues R1281, S1284, R1285, and H1303. The downstream region of the (pY1-CTD)₂ peptide wraps around the C-terminal SH2. Therefore, the phosphate group of the downstream pY1 is recognized in the non-canonical pocket of Spt6-C-SH2 (Pocket #2) formed by K1354 and K1434. (Figures 3(a) and (b) and S2A). The non-canonical pocket is located on the opposite side of the β-sheet to the canonical one (Figures 3(a) and (b) and S2A). Interestingly, the non-canonical C-terminal SH2 domain contains an additional positively charged binding pocket (Pocket #3; K1360, K1361), which allows the accommodation of phosphorylated residues in alternative conformations (Figures 3(c) and S2A). Our NMR data for the tSH2 domain of Spt6 in complex with (pY1-CTD)₂ contain two sets of distinct intermolecular NOEs. One set corresponds to the major conformation, which is described above. The second set corresponds to a minor conformation in which the downstream region of (pY1-CTD)₂ is bound in a different conformation (Figures 3(c) and S2A). The existence of multiple binding sites in the non-canonical C-terminal SH2 domain of Spt6 explains our FA binding data in which differently phosphorylated CTD peptides are bound with similar binding affinities. Furthermore, the perturbation of one binding pocket does not abrogate the overall binding affinity, due to the existence of alternative binding modes in different binding pockets (Table 2, Figure 4). However, the interaction with (pY1-CTD)₂ is diminished upon deletion of the C-terminal SH2 (Figure 4). The symmetrical experiment with the C-terminal SH2 could not be done, as it requires the N-terminal SH2 for the correct folding.

Table 2.

Binding affinities for wt and its variants based on FA experiments

	KD (μM)
Wild-type	5.1 ± 0.4
R1281A	17.6 ± 1.0
S1283A	5.4 ± 0.2
S1284A	8.2 ± 0.1
R1285A	12.0 ± 0.3
E1324A	1.4 ± 0.1
K1354E	16.3 ± 1.7
K1360A	12.4 ± 0.5
K1434A	24.0 ± 2.7
R1281A/S1283A	35.5 ± 1.9
R1281A/S1284A	11.4 ± 1.0
R1281A/R1285A	11.6 ± 0.9
R1281A/S1283A/R1285A	16.5 ± 0.5
R1281E/S1284A/R1285E	48.3 ± 4.7
Spt6-N-SH2	313.2 ± 12.5

Figure 4.

Critical residues for Spt6-tSH2 and CTD interaction. (a) Normalized binding isotherms for Spt6-tSH2 and its variants with the (pY1-CTD)₂ peptide. The equilibrium binding of Spt6-tSH2 and its variants with the (pY1-CTD)₂ peptide was monitored using FA. The symbols and color coding of data points for the individual Spt6-tSH2 variants are shown. The binding curves are fitted by non-linear regression to a single binding site model [39]. Standard errors are generated from a set of three independent measurements. (b) The same data as in (a) shown as logarithmic bar graph of K_D of the wild-type and its variants (single/double/triple amino acid substitutions and Spt6-N-SH2).

Comparison of Spt6-tSH2 bound to CTD with other SH2 complexes

The binding of the phosphorylated tyrosine in Pocket #1 resembles that of the canonical single SH2 domains in terms of the topology and the residues involved (Figure S2B and E). In the canonical SH2 domain, such as the SH2 domain of Src (Figure S2B), the phosphotyrosine residue is deeply embedded into the binding pocket formed by the invariant arginine (R175), histidine (H201), serine (S177), glutamate (E178), and threonine (T179). The aromatic ring is further coordinated by lysine (K204) and arginine (R155). In contrast to the canonical phosphotyrosine-binding pocket of SH2 [45,46], the Pocket #1 of Spt6 lacks the residues that coordinate the aromatic ring of phosphotyrosine. This makes the Pocket #1 of Spt6-tSH2 larger and less stringent with regard to the selectivity of substrates, and can tolerate other phosphorylated residues (e.g. pS and pT). Furthermore, Spt6-N-SH2 also does not bind the flanking residues of the phosphotyrosine, a typical feature observed in the canonical SH2 [45,46]. As a result, the Pocket #1 of Spt6-tSH2 is less stringent with regard to the selectivity of substrates.

The structure of Spt6-tSH2 bound to CTD is similar to the recently reported structure of Spt6-tSH2 in complex with the Rpb1 linker. The RMSD deviation of C. glabrata Spt6-tSH2 and S. cerevisiae Spt6-tSH2 from the complexes determined by NMR and X-ray, respectively, is 1.14 Å (Figure S2G). The CTD and Rpb1 linker are accommodated at Spt6-tSH2 in a similar manner (Figure 3(d)). The CTD has two phosphotyrosines separated by seven residues, whereas the Rpb1 linker contains two phosphorylated residues (T1471 and S1493) separated by 21 amino acids. Despite the different positions and residue types that are phosphorylated in them, the two peptides are accommodated in the same orientation and the phospho-residues are inserted in the Pocket #1 and Pocket #2 of Spt6-tSH2. Whereas the phospho-CTD is exclusively recognized by the two-pronged mechanism, the binding of the Rpb1 linker by Spt6-tSH2 involves additional residues, mainly the hydrophobic L1478 and F1492, and polar N1475, E1476, N1481, and D1488 residues. In Pocket #1, the structures show a slightly different recognition of the upstream pY (of the CTD) and the upstream pT (of the linker). With the Rpb1 linker, the phosphothreonine (pT1471) and also the neighboring tyrosine (+ 2 residue) are inserted in Pocket #1, which is considered to be a molecular mimic of phosphotyrosine [21]. In contrast, the pY residue of the CTD substrate is recognized by the Spt6-SH2 similarly as in the canonical SH2 domains (Figure S2B) [45]. It involves the invariant R1281, S1283, and S1284 residues, and R/K-conserved residue R1285. Mutations of these residues weakened the interaction between the tyrosine-phosphorylated CTD double-repeat and Spt6-tSH2 (Figure 4). With the CTD substrate, the peptide is wrapped around Spt6-C-tSH2 in a more extended conformation compared to the Spt6-tSH2-Rpb1 linker complex, and the aromatic ring of the downstream phosphotyrosine is deeply embedded into the Pocket #2 (Figure S2A and B) formed by several aromatic and hydrophobic residues (Y1380, F1382, L1394, F1396, W1407, F1433, L1437). The recognition of the downstream phosphorylated residue in Pocket #2 is also achieved in a similar manner in both complex structures (Figure S2A and B). The invariant K1354 and K1424 coordinate the phosphate groups of pY and pS1493 of the CTD and linker substrates, respectively. The phosphate moiety is additionally stabilized by Y1380. Perturbation of this pocket resulted in lower affinities toward the phosphorylated substrates (Figure 4). Interestingly, both Spt6 complexes have their substrate peptides accommodated in the opposite orientation to a substrate bound to the tSH2 domain of ZAP-70 (Figures 3(d) and S2F). With the ZAP-70 complex, both SH2 domains canonically recognize the peptide stretches with phosphotyrosines that are connected by a short α-helix (Figure S2F).

Discussion

Previous studies reported the structures of the free tSH2 domains from yeast [30,31]. Based on the earlier data [28], it was generally assumed that the tSH2 of Spt6 associates with the phosphorylated CTD. However, no structural information on the Spt6-tSH2-CTD complex has been available so far, therefore we set out to investigate the structure of tSH2 in complex with the doubly phosphorylated CTD. In the meantime, a new structural model has been proposed in which not the CTD, but the linker between the Rpb1 core and its CTD is suggested to be a novel binding partner for Spt6-tSH2 [21]. The structure of Spt6-tSH2 bound to the Rpb1 linker is derived from S. cerevisiae, in which the linker is approximately ~ 75 residues long and largely unstructured. However, the linker in other yeasts such as S. pombe is longer (~ 100 residues) than in S. cerevisiae, is partially structured, and forms additional contacts with Rpb4/7 [47]. Therefore, the interaction of the linker with Rpb4/7 could spatially restrict the binding of Spt6. Recent Cryo-EM data of a human elongating complex showed a low-resolution electron density for Spt6-tSH2, suggesting local flexibility of this domain and possible recruitment via both the linker and the CTD [32]. Furthermore, Buratowski and colleagues showed that the linker must be connected to Rpb1 to function, whereas CTD can be fused to RNAP II subunits as long as it is tethered near its original position [48].

The binding strength is generally evaluated with the use of short peptides as the substrate models [9,21,30,31,38,40], but in the context of the CTD and its multiple repeats, the overall dissociation constant might be different due to the avidity effects. Our MST experiments with the phosphorylated 13-repeats of CTD show a strengthening of the affinity from the μM to the nM level (Figure 1). Tripsianes and colleagues [34,35] reported a similar 3 orders of magnitude increase in affinity for Tudor domains if peptides with multiple methylated arginines are introduced in the assay. The same effect is observed when multiple proline-rich motifs are present in the OCRE domain [49]. Therefore, the enhanced nanomolar affinities obtained by avidity originating from multiple phosphorylation marks present in the CTD suggest that the CTD may participate in the recruitment of Spt6. It is also possible that the 2 orders of magnitude stronger binding of Spt6-tSH2 to (pY1-CTD)₁₃ reflects the fact that more than one Spt6-tSH2 is bound to the phosphorylated 13-repeats of CTD.

Both Spt6-tSH2 and Spt6ΔN bind the minimal CTD substrate (pY1-CTD)₂ with the same affinity (K_D ≈ 4–5 μM). This suggests that the tSH2 domain is the essential region of Spt6 for contacting the CTD. Interestingly, Spt6-tSH2 binds the extended CTD substrate (pY1-CTD)₁₃ with tighter affinity (K_D ≈ 15 nM) than the nearly full-length protein (Spt6ΔN; K_D ≈ 119 nM) (Figure 1). The reduced binding of the nearly full-length protein (Spt6ΔN) relative to the isolated domain (Spt6-tSH2) for the extended multivalent ligand (pY1-CTD)₁₃ suggests that additional domains in the Spt6ΔN construct sterically obscure binding to the repetitive phosphorylated sites [50]. In other words, it is possible that the Spt6 core sterically blocks additional tSH2 domain interactions within the multivalent ligand (pY1-CTD)13 (via the steric exclusion of ligand sites). This would explain why there is no observed difference in affinity between Spt6-tSH2 and Spt6ΔN for the minimal CTD substrate (pY1-CTD)2, which can only accommodate one tSH2 domain at a time.

The Spt6-tSH2 domain does not show a clear preference for a particular phosphorylation mark, nor does it induce a specific peptide conformation upon binding. The recognition of the substrate resembles the binding of a capping enzyme, in which only the phosphorylated residues are recognized [51]. In contrast, the CID domains found in Nrd1 [52], Rtt103 [38,53] or Pcf11 [53] induce a β-turn conformation of the CTD with extensive contacts with downstream residues. Additionally, CIDs have a strong preference for a particular phosphorylation mark, namely pS2, pT4, and pS5. Several laboratories performed biochemical assays to provide evidence of interaction with peptides derived from pS2- [30], pS5-, and pY1-CTD alone and in combination with pS2 and pS5 [9]. Our data are in agreement with those of previous studies, which have binding strengths in the range of μM affinities. Furthermore, our data show a non-discriminating behavior for various phosphate marks. The binding tightens with the increasing number of phosphorylation marks present at the two-repeat CTD substrate. This drop in recognition stringency for a particular residue can be explained by the replacement of positively charged residues (Pocket #1) responsible for the coordination of the aromatic ring of phosphotyrosine in the canonical SH2s (Figure S2B).

Single amino-acid substitutions of tSH2 do not show a high drop in affinity, although there are exceptions, e.g. the invariant arginine R1281A (Pocket #1) and lysines K1354 and K1434A (Pocket #2) (Figure 4). Additional amino acid substitutions are necessary to observe a greater effect on the CTD–Spt6 interaction. The need to introduce additional amino acid substitutions to further diminish the interaction is well reported for SH2 domains, due to the interchangeability of the hydrogen-bonding network as the interaction is more complex than the recognition of phosphomarks [29]. Our triple amino acid substitution R1281E/S1283A/R1285E impairs the strength of the CTD–Spt6 interaction 10 × compared to the wild-type (Figure 4). In contrast, deleting the Spt6-C-SH2 domain abolishes the binding completely (Figure 4), which is consistent with previous observations [30]. Single amino acid substitutions in Pocket #1 have been previously tested in vivo and only exhibited a mild effect on cell fitness [21,31]. Multiple amino acid substitutions in Pocket #2 caused a moderate phenotype [21]. This effect was stronger when mutations of Pocket #1 and #2 were combined [21]. We aimed to perform in vitro and in vivo studies with all three pockets inactivated to recapitulate severe phenotypes observed by the truncation of Spt6-tSH2 [29,30,33]. However, our experiments with multiple amino acid substitutions in all three pockets resulted in insoluble protein variants.

The phosphorylation patterns change rapidly on five out of seven residues of the CTD repeat during the transcription cycle, as reported by ChIP assays. As a result, a specific set of factors is recruited by these phosphorylation patterns during transcription. For example, the recruitment of termination factors is stimulated by the pS2 marks, but inhibited by the pY1 marks [9,20]. Interestingly, Spt6 does not display this narrow selectivity, but is instead able to recognize multiple combinations of CTD patterns. This behavior is further supported by genome-wide analysis and mass spec pull-downs [54,55]. The additional phosphogroups are likely to be accommodated into the existing binding pockets (Pocket #1–3). The distances among binding pockets are not equal, and their combination may enable the accommodation of various phospho-CTD patterns (Figure S4). The robustness of the recognition of phosphorylated substrates is further illustrated by the effect of point mutations. Previous works showed that single and multiple-point mutations in binding pockets had only a minor effect on the binding affinity [29,30]. Here, we created a triple amino acid substitution in Pocket #1, and the binding to the CTD only decreased 10 × compared to the wild-type, as the other two binding pockets of Spt6-tSH2 could substitute for its binding capacity. To remove the interchangeability of binding pockets and ability to accommodate different phosphopatterns, we deleted the entire Spt6-C-SH2 domain with its Pocket #2 and #3. The resulting construct with a single binding pocket (Spt6-N-SH2 domain) exhibited a low binding affinity toward phosphorylated CTD (Figure 4 and Table 2).

The mapping of phosphorylation sites on the tandem repeats of the RNAP II CTD showed that the phosphomarks are evenly dispersed throughout the length of the CTD [56,57]. Our data suggest that the multiple phosphorylation marks present in the CTD strongly enhance binding affinities through avidity effects. This may play a key role in tethering the RNAP II CTD to factors that process the nascent transcript and modify the chromatin template. The tSH2 domain of Spt6 has recently been found to accommodate a cognate pY ligand provided in two parts: a tyrosine side-chain and a phosphate of neighboring phosphothreonine [21]. This remarkable plasticity of the tSH2 domain is further documented by our structural and binding data, which suggest why Spt6 acts as a general transcription factor associated with different stages of transcription by RNAP II, having different phospho-CTD patterns. The tSH2 domain of Spt6 is an unusual domain with multiple binding sites, which can accommodate different CTD phosphopatterns. This observation is consistent with the genome-wide localization of Spt6, which broadly associates with RNAP II during transcription elongation [21], supporting possible interactions with a wide range of phosphorylation marks, including Y1, S2, T4, S5, and S7 [54,55].

Taken together, our data suggest that not only the interaction of Spt6 with the Rpb1-CTD linker [21], but also the interactions with different CTD phospho-patterns may be important for the recruitment and/or retention of Spt6 on genes. In an alternative model, the tSH2 domain may initially recruit Spt6 to the elongation complex through its interactions with the Rpb1 linker or the phospho-CTD patterns, but then the interactions between the Spt6 core and Rpb4/7 stalk on the Rpb1 side of RNAPII [32] play a role in retaining Spt6 in the elongation complex. It is also possible that Spt6 undergoes dynamic conformational changes as it associates with histones co-transcriptionally, and the promiscuity of the Spt6-tSH2 domain could aid in the maintenance of the interaction with RNAPII.

Materials and Methods

Expression and purification

Plasmid pET28b, with thrombin cleavable N-terminal His-tag and sequence of tSH2 domain (Spt6-tSH2) from C. glabrata, was donated by P. Cramer. The plasmid was transformed into the BL21-CodosPlus (DE3)-RIL bacterial strain. Site-specific mutants were generated using a QuikChange site-directed mutagenesis kit (Agilent) and verified by DNA sequencing. The Spt6-tSH2 was expressed overnight at 16 °C. Cells were harvested, resuspended in 50 mM Na₂HPO₄, 500 mM NaCl, 20 mM BME, 20 mM Imidazole, pH 7.3, and sonicated. The cell lysate was cleared by centrifugation. The protein was loaded on a HisTrap FF crude column (GE Healthcare) and eluted with buffer containing 500 mM imidazole. The buffer was replaced with a low-salt 50 mM Hepes, 10 mM NaCl, 20 mM BME (pH 7.3) buffer using a HiPrep26/10 desalting column (GE Healthcare). The sample was loaded into a HiTrap SP HP column (GE Healthcare) and eluted with a 10 mM to 1 M NaCl gradient. The eluted protein was further purified in a HiLoad Superdex 75 in 50 mM Hepes, 150 mM NaCl, 20 mM BME (pH 7). The protein was concentrated using Vivaspin 20 (Sartorius) with a 10-kDa cutoff. The N-terminal SH2 domain (1250–1354), Spt6-N-SH2, was sub-cloned into the pET-22b vector with a C-terminal His-tag and purified using affinity chromatography with a HisTrap FF crude (GE Healthcare) column followed by gel filtration in a HiLoad Superdex 75 16/600 (GE Healthcare). For the purpose of NMR and SAXS measurements, the construct for Spt6-tSH2 was re-cloned into pET28b containing the N-terminal His-tag with SUMO. The SUMO tag was removed with on-column cleavage in a HisTrap FF crude (GE Healthcare) column followed by gel filtration in a HiLoad Superdex 75 16/600 (GE Healthcare). The Spt6ΔN construct from S. cerevisiae (residues 298–1451) was cloned into the pET151-D/TOPO vector. Expression and purification followed the protocol described elsewhere [31].

Preparation of peptides

The (pY1-CTD)₁₃ peptide was prepared by the co-expression of phosphotyrosine kinase c-Abl and SUMO-tagged CTD with subsequent purification via affinity and ion-exchange chromatography as described elsewhere [36]. To increase the homogeneity of the phosphorylated CTD sample, we only collected the middle section of the separated fraction eluted from the anion exchange chromatography column (HiTrap Q HP). The control (CTD)₁₃ peptide was N-terminally fused with a SUMO tag in the pET28b vector. Expression and purification followed the same procedure as for the (pY1-CTD)₁₃ peptide without the ion-exchange chromatography.

Fluorescence anisotropy

FA was measured in a Horiba Jobin-Yvon FluoroMax 3 spectrometer. The CTD-peptides were synthetised by the Clonestar peptide service (Brno, CZ) with the following sequences: (CTD)₂ = 5,6-FAM-PSYSPTSPSYSPTSPS, (pY1-CTD)₁ = 5,6-FAM-PSpYSPTSPS, 5,6-FAM-PSpYSPTSPSYSPTSPS, 5,6-FAM-PSYSPTSPSpYSPTSPS, (pY1-CTD)₂ = 5,6-FAM-PSpYSPTSPSpYSPTSPS, (pY1pS2-CTD)₂ = 5,6-FAM-PSpYpSPTSPSpYpSPTSPS, (pY1pS5-CTD)₂ = 5,6-FAM-PSpYSPTpSPSpYSPTpSPS, 5,6-FAM-PSYpSPpTSPSYSPTSPS, (pT4-CTD)₂. = 5,6-FAM-PSYSPpTSPSYSPpTSPS, (pS2pS7-CTD)₂. = 5,6-FAM-PSYpSPTSPpSYpSPpTSPpS and the NIM peptide [40]. In FA measurements, the purified tSH2 and its variants were titrated against the CTD-peptides as previously reported [9]. Each data point is an average of measurements in triplicate. The data were analyzed by non-linear least-squares regression [39].

Microscale thermophoresis

MST were performed at RT in a Monolith NT.115 (NanoTemper Technologies), with excitation power 50% and MST power 40% under the same conditions as in FA measurements. The (pY1-CTD)₁₃ and (CTD)₁₃ peptides were labeled with a RED-NHS protein labeling kit (NanoTemper Technologies) according to the manufacturer's instructions with the exception of the incubation period (overnight). The MST measurements were performed in 50 mM Hepes, 10 mM NaCl (pH 7), 0.05% Tween 20, and in 50 mM Hepes, 150 mM NaCl (pH 7), 0.05% Tween 20. The concentration of the CTD in the assay was 50 nM. The data were fitted with four-parameter logistic regression. We were able to fit our data with models for 1, 2, and 3 binding sites, and they all provided a similar K_D.

NMR experiments and structure calculation

All the NMR spectra of the tSH2 domain of Spt6 were recorded in Bruker AVANCE 700, 850 and 950 MHz spectrometers equipped with cryoprobes at a sample temperature of 37 °C using 0.5 mM of uniformly ¹⁵N, ¹³C-labeled protein sample in 50 mM Hepes, 10 mM NaCl, 20 mM BME (pH 7) (90% H₂O/10% D₂O). The unlabeled peptide consisting of two repeats of the CTD heptad (PSpYSPTSPSpYSPTSPS, pY = phosphorylated tyrosine) was titrated to the free tSH2 to a final ratio of 1:1.5 (protein:peptide). The assignment of the backbone resonances of the Spt6-tSH2 complexed with CTD was performed as described previously using the 3D experiments HNCA, HN(CO)CA, CBCA(CO)NH, and HNCACB [58]. Protein backbone and side-chain atom assignment was completed up to a level of 78% in total. The missing 22% of resonances mostly originate from unstructured loops, terminal regions, and residues that suffered from unfavorable chemical exchange properties. The spectra were processed in Topspin 2.1 (Bruker BioSpin) and analyzed in Sparky 3.0 (Goddard and Kellner, University of California, San Francisco). Proton and carbon side-chain resonances were assigned using a 4D version of HCCH TOCSY [59] that was measured with a non-uniform sampling [60]. Acquired data were processed and analyzed analogously as described previously [61,62].

All the intramolecular restraints were derived from 3D ¹⁵N- and ¹³C-edited NOESY experiments recorded in a 950-MHz spectrometer. The intermolecular constraints between the protein and the peptide were determined from the 3D F1-¹³C/¹⁵N-filtered NOESY-[¹³C,¹H]-HSQC experiments [63,64] with a mixing time of 150 and 400 ms in an 850-MHz spectrometer. The NOEs were semi-quantitatively classified based on their intensities in the 3D NOESY spectra. The initial structure calculations were performed in CYANA 3.97 using the automated NOE assignment module [65]. The structure defined by the automatically assigned intramolecular NOE distances and manually assigned intermolecular NOE distances was further refined in AMBER16 [66] using an ff14SB [67] force field with the extension for the phosphorylated tyrosine [68] and a simulated annealing protocol described previously [69,70]. Out of the 50 calculated structures, the 20 with the lowest energy were selected as a representative ensemble.

To determine the orientation of the CTD on the tSH2 domain, a peptide with a paramagnetic tag on its N-terminal part (TOAC-GPSpYSPTSPSpYSPTSPS-PEG₂-RR-amid, TOAC = 2,2,6,6-tetramethylpiperidine-1-oxyl-4-amino-4-carboxylic acid, PEG = polyethylene glycol) was titrated to the isotopically labeled tSH2 to a final ratio of 1:1 (paramagnetic sample). The intensity of the peaks in ¹H–¹⁵N HSQC was measured and compared to the intensity of the peaks in the spectrum after the quenching of the paramagnetic tag by the addition of ascorbic acid (diamagnetic sample). The intensity of the peaks in the diamagnetic spectrum was referenced using the spectra of tSH2 titrated with unlabeled peptide in order to avoid biasing the intensities by the free label in the solution or the label itself. The ratio of the intensity of the peak in the paramagnetic and diamagnetic spectra was then converted to the distance between the nitrogen of the amide group of the amino acid and the oxygen atom of the TOAC label, as published elsewhere [71,72]. These distances were later used as restraints in CYANA 3.97 [65] calculations in order to determine the position of the TOAC label.

SAXS analysis

Samples for SAXS were prepared by mixing 0.1 mM Spt6-tSH2 in a 1:1.5 ratio with (pY1-CTD)₂. Samples were measured in the BioSAXS-1000, Rigaku at CEITEC. Data were collected at X-ray beam wavelength λ = 1.54 Å. The sample to detector (PILATUS 100K; Dectris Ltd.) distance was 0.485 m, covering a scattering vector range from 0.008 to 0.65 Å⁻¹. For the buffer and sample, one 2D image was collected with an exposure time of 60 min. Data are plotted using Gnuplot 4.6. Evaluation of the solution scattering of the atomic models and the fitting to experimental data were performed with CRYSOL [73].

Monte Carlo simulations

To assess the correlation between the binding affinity and the overall charge of the CTD peptide, we performed Metropolis Monte Carlo (MC) simulations using the Faunus framework [74]. A spherical cell with a radius of 200 nm contained one copy of the tSH2 domain and one copy of CTD. Each protein is described with an implicit-solvent coarse-grained model, where every residue is treated as a spherical bead (located at the center-of-mass of the residue) with a radius derived from the amino acid molecular weight and the common density of 0.9 g/ml. The N and C termini of both proteins were represented as separate beads. Harmonic bonds with a spring constant of 0.5 kT/Ų and an equilibrium distance of 4.9 Å connected neighboring beads. No additional angular or dihedral restraints were applied. Each residue interacted with the Lennard–Jones potential with an averaged attraction depth of 0.35 kT and with a charge based on its protonation state (allowed to change during the simulation). The solvent was treated as a dielectric continuum using the Debye–Hückel approximation [[75], [76], [77]] for the interaction of charged residues.

The coarse-grained model of the Spt6-tSH2 domain is based on the NMR solution structure solved within this project, while the structure of CTD was a random coil. The tSH2 domain was placed in the middle of the simulation sphere with all degrees of motion frozen. The CTD peptide was allowed to move with standard rotation-translation moves (single-residue translation; whole protein translation and rotation); Crankshaft; Pivot rotations [78]. Each amino acid was allowed to change its protonation state by titration move, in which protons are exchanged between the residues and the solution. The energy associated with the exchange is determined by the change in local electrostatic energy using the pKa of the isolated amino acid and pH of the system (set to neutral pH 7) [79]. Titratable residues with their pKa values are as follows: C terminus (2.6), Asp (4.0), Glu (4.4), His (6.3), N terminus (7.5), Tyr (9.6), Lys (10.4), Cys (10.8), Arg (12.0), pSer (2.19, 5.78), pThr (2.19, 6.1), and pTyr (2.19, 5.9) [80]. The total number of moves was 2.5 × 10⁹ for each simulation. The moves were chosen randomly. The simulations were performed in an NVT ensemble at 298 K. The dissociation constants were calculated from the averaged probability distribution of the ligand around the tSH2 domain.

Accession numbers

The atomic coordinates and restraints for the NMR ensemble of the Spt6-tSH2–CTD complex have been deposited in Protein Data Bank, PDB ID code: 6QTC.

Edited by Karolin Luger

Appendix A. Supplementary data

Supplementary material