Phosphospecific Recognition of a Spt5 CTR Motif.

The transcription elongation factor Spt5 exists as a dimer with Spt4 in a complex known as DSIF in humans and contains between one and fifteen CTR repeats. The CTR motifs can be phosphorylated by the kinase Bur1/Cdk9 (24, 29, 30), which was shown to enhance recruitment of Paf1C to chromatin (19, 21, 29, 30, 38, 39) and RNA pol II processivity during transcription elongation (24, 35). Recent biochemical and in vivo evidence has demonstrated a direct interaction between the Rtf1 Plus3 domain and the CTR of Spt5 in yeast (19). We sought to characterize the human Plus3–Spt5 interaction and determine whether this interaction was phosphospecific. To clarify the interpretation of biochemical results and provide a unique binding register for our structural work, we synthesized two Spt5 peptides corresponding to residues 779–790 of human Spt5. These peptides contain one complete CTR repeat including its spacer and were constructed either with or without a phosphothreonine at position T784 (Fig. 2A). We labeled both peptides with 6-carboxyfluorescein (CF) at their N terminus and monitored binding of recombinant human Rtf1 Plus3 domain to these peptides using fluorescence anisotropy (Fig. 2B). We found that the Plus3 domain binds the phosphorylated form of the Spt5 CTR (pCTR) with a calculated equilibrium dissociation constant (K_d) of 5.45 ± 0.38 µM (Fig. 2B). Binding to unphosphorylated peptide resulted in a negligible change in anisotropy, with an estimated K_d greater than 1 mM. These data suggest that the Plus3 domain can recognize and directly bind a single phosphorylated repeat of the Spt5 CTR and strongly suggest Paf1C recruitment to RNA pol II is mediated through a direct interaction with Spt5.

Open in a separate window

Fig. 2.

Human Rtf1 Plus3 domain recognizes phosphorylated Spt5 CTR. (A) Domain architecture of human Spt5 including acidic, NusG-N-terminal (NGN) domain, KOW (Kyprides, Ouzounis, Woese) domains, and the C-terminal repeat (CTR) regions. Spt5 sequence (767–815) containing the CTR repeats is expanded, with pentapeptide CTR motifs underlined and phosphocapable threonine residues indicated with an asterisk. (B) Increasing concentrations of purified Plus3 were titrated into 20 nM of fluorescein-labeled pCTR or CTR peptide, and binding was assessed via fluorescence anisotropy. Binding measurements were performed in triplicate, and the SD was plotted. The K_d was extrapolated using nonlinear regression.

Structure of the Plus3–Spt5 pCTR Complex.

To determine the molecular basis for phosphospecific recognition of a Spt5 CTR repeat by the Rtf1 Plus3 domain, we crystallized and determined the structure of a Plus3–pCTR complex (Fig. 3A). Phases were determined using molecular replacement and the Plus3 structure as a search model. The final model was refined at 2.43-Å resolution with R_work and R_free values of 17.7 and 22.4%, respectively. Six Plus3 molecules are contained within the asymmetric unit, four of which are in complex with a CTR motif (Fig. 3A). For clarity we are describing the complex that is best resolved and whose model is most complete; however, all four complexes adopt a similar conformation (maximum rmsd of 0.8 Å over 131 Cα atoms). The two remaining Plus3 proteins are making crystal contacts that occlude the peptide binding. The conformation of the Plus3 domain is not significantly altered as a result of peptide binding, with an rmsd of 0.8 Å over 133 Cα atoms.

Open in a separate window

Fig. 3.

Rtf1 Plus3 uses an extended surface to recognize the Spt5 pCTR motif. (A) Surface view of the Plus3 domain (white) in complex with the Spt5 pCTR peptide (yellow). Omit map density (F_o-F_c, contoured at 4.0 σ, green) is shown for the peptide. The side chain of Spt5 R783 has been removed for clarity. (B) The phosphothreonine recognition pocket. The Spt5 pCTR peptide is shown in yellow with omit map density (F_o-F_c contoured at 4.0 σ) in green. Critical Plus3 residues involved in peptide recognition (white sticks) and the hydrogen bonds involved in recognition (black) are indicated. (C) The hydrophobic Spt5 pCTR-binding surface. Van der Waal interactions (magenta) between Plus3 residues (white) and the pCTR peptide (yellow) are shown. Spt5 residues pT784 and P785 are contained within the CTR motif, whereas residues M786 and Y787 are found within the spacer between CTR repeats and are making significant contacts with the Rtf1 Plus3 domain. Contacts between M786 and Plus3 residues G390 and G392 have been omitted for clarity.

The CTR peptide lies across a shallow groove formed by the β1-β2 central hairpin and the loop connecting β6 and α5 (Fig. 3A), burying 443 Ų of surface area on the Plus3 domain. The Plus3–Spt5 interface is distant from the location of methyl binding in Tudor domains and largely overlaps the conserved surface observed earlier (Fig. 1B). Interactions between the Plus3 and the Spt5 pCTR peptide can be divided into two sections, a phosphothreonine recognition interface and a hydrophobic peptide-binding interface. Recognition of the CTR phosphothreonine (pT784) is accomplished through a series of hydrogen-bonding interactions that form a cage around the phosphate. Two of the conserved positively charged residues from which the Plus3 domain derives its name are found in this interface. One of these, R366, makes a hydrogen bond with the gamma oxygen of pT784, orienting the position of the threonine residue within the binding pocket (Fig. 3B). The positioning of pT784 is also conferred by hydrophobic interactions between the gamma carbon of the pT784 residue and the aromatic ring of Y400 and the Cα positions of S443 and F441. The other canonical Plus3 residue in this interface, R388, interacts with and likely orients the side chains of Y400 and S443, both of which are making direct hydrogen bonds to the O3 position of the phosphate moiety (Fig. 3B). Combined with residue Q445, which is making a direct hydrogen bond to the O1 oxygen, this network of hydrogen-bonding interactions recognizes three of the four oxygens within the pT784, providing a mechanism for phosphospecific recognition by the Plus3 domain.

The second portion of the Plus3–pCTR interface is hydrophobic, and unexpectedly, we found it primarily contained Spt5 residues from the spacer region between repeats (Figs. 2A and ​and3C).3C). The size of this spacer in humans is often conserved and typically contains residues with hydrophobic character (Fig. S2). In the complex we crystallized, Plus3 residues Y400, P398, N393, G390, and G392 combine to form a small pocket that accommodates Spt5 spacer residue M786. The first spacer position is occupied by a large hydrophobic residue such as Met, His, or Leu, any of which could be accommodated within this pocket. The second spacer position, Y787 in our peptide, is interacting through its aromatic ring with a hydrophobic surface on Rtf1 formed by residues F441, F437, and I391. Although the tyrosine in our peptide is the consensus residue at this position, it seems quite reasonable that the Plus3 surface could accommodate glutamine and histidine residues which are also found in this position. Outside of the phosphothreonine recognition surface, the central organizing residue on the Plus3 surface is F441. This residue makes van der Waals contacts with Spt5 residues P785, M786, and Y787 (Fig. 3C). Interactions with P785 are with the Cβ and Cγ positions explaining its preference for this cyclic residue, whereas interactions with M786 and Y787 are largely through the peptide backbone allowing for some flexibility in CTR sequence recognition. Together the structure of the Rtf1 Plus3 domain in complex with a pCTR peptide reveals the basis for phosphothreonine-specific recognition of Spt5, whereas interactions outside of pT784 can accommodate the observed variation in CTR sequence.

Identification of Critical Residues Within the Plus3 Domain That Mediate pCTR Binding.

The binding interface observed in our Plus3–pCTR complex is separate from both the proposed DNA-binding surface identified by NMR chemical perturbation studies (37) and the canonical methyllysine-binding site in Tudor domains (43), suggesting the Plus3 domain has evolved a separate binding surface specifically to interact with Spt5 (Fig. 1B). We next tested the importance of residues at the Plus3–pCTR interface for peptide binding in solution. Guided by our structural data, we used site-directed mutagenesis to alter amino acids that lie on the Plus3 surface, avoiding residues that may play a significant role in conferring protein stability. We tested the ability of purified WT and mutant Plus3 domains to interact with pCTR peptide using differential scanning fluorimetry (45, 46). In this assay, a hydrophobic fluorescent dye is used to monitor the change in melting temperature (ΔT_m) conferred by the addition of ligand. We observe a ΔT_m of 1.5 °C when using WT Plus3 and pCTR peptide compared with a ΔT_m of 0.5 °C when using WT Plus3 and an unphosphorylated Spt5 peptide (Fig. 4A). In the absence of peptide, the thermal denaturation profiles of Plus3 variants were similar to WT, indicating that overall stability was unaffected for any of the proteins tested (Fig. S3). A R366A substitution, however, abolished pCTR binding, suggesting an important role in peptide binding for this residue. This role may be to orient the side chain of Spt5 T784, thereby positioning the attached phosphate group into the hydrogen-bonding cage and suggesting why the effect of a R366A mutant was stronger than that of Q445A which is also hydrogen-bonding to pT784 (Fig. 4A). Alanine substitutions within the hydrophobic interface at positions F437 and F441 significantly compromised peptide binding, whereas a R435A variant had a subtle effect, all consistent with their role in peptide recognition observed in the structure. Substitutions in conserved residues K413 and F440 located outside the binding interface had no effect (Fig. 4A and Fig. S3), further validating predictions based on the structure.

Open in a separate window

Fig. 4.

Substitutions of residues at the Plus3–pCTR interface affect peptide binding. (A) Increase in the unfolding temperature for Plus3 upon the addition of CTR peptide as measured by differential scanning fluorimetry. White bar indicate measurements performed using WT Plus3 protein with the indicated peptide, gray bars used the indicated Plus3 variant with pCTR peptide. Asterisks denote P values of <0.05. Error bars represent SD from 12 replicates. (B) Binding isotherms for human Plus3 or the indicated Plus3 variant to 6-CF-labeled pCTR peptide. Increasing concentrations of purified Plus3 variants were titrated into 20 nM of pCTR peptide, and binding was assessed via fluorescence anisotropy. Experiments were performed in triplicate, and the SD was plotted.

We used fluorescence anisotropy to test the importance of residues at the Plus3–pCTR interface using an alternative approach. R366A and F441A substitutions both severely reduced pCTR binding (Fig. 4B), in agreement with the results from differential scanning fluorimetry. These results suggest that both hydrogen bonding and hydrophobic interactions are critical for proper pCTR binding. Titrations using a F437A variant of Plus3 resulted in reduced affinity for pCTR, consistent with an important but not critical role in peptide recognition. Together, these data demonstrate that the pCTR-binding surface observed in the crystal is also being used for peptide binding in solution. Further it demonstrates that both the phosphothreonine recognition cage and the hydrophobic binding interface centered on F441 are critical determinants for Spt5 binding.

The authors thank B. Van Houten, B. Tomson, C. MacGillivray, and A. Berman for technical assistance and advice. Funding support was provided by National Institutes of Health Grant R01 GM52593 (to K.M.A.).