Sequence and structural determinants of RNAPII CTD phase-separation and phosphorylation by CDK7

Abstract

The intrinsically disordered carboxy-terminal domain (CTD) of the largest subunit of RNA Polymerase II (RNAPII) consists of multiple tandem repeats of the consensus heptapeptide Y1-S2-P3-T4-S5-P6-S7. The CTD promotes liquid-liquid phase-separation (LLPS) of RNAPII in vivo. However, understanding the role of the conserved heptad residues in LLPS is hampered by the lack of direct biochemical characterization of the CTD. Here, we generated a systematic array of CTD variants to unravel the sequence-encoded molecular grammar underlying the LLPS of the human CTD. Using in vitro experiments and molecular dynamics simulations, we report that the aromaticity of tyrosine and cis-trans isomerization of prolines govern CTD phase-separation. The cis conformation of prolines and β-turns in the SPXX motif contribute to a more compact CTD ensemble, enhancing interactions among CTD residues. We further demonstrate that prolines and tyrosine in the CTD consensus sequence are required for phosphorylation by Cyclin-dependent kinase 7 (CDK7). Under phase-separation conditions, CDK7 associates with the surface of the CTD droplets, drastically accelerating phosphorylation and promoting the release of hyperphosphorylated CTD from the droplets. Our results highlight the importance of conformationally restricted local structures within spacer regions, separating uniformly spaced tyrosine stickers of the CTD heptads, which are required for CTD phase-separation.

Fig. 1. The importance of human RNAPII C-terminal domain (CTD) residues in phase-separation.

a A schematic representation of transcribing RNAPII highlighting the CTD of the largest subunit, RPB1, and the CTD heptad consensus sequence (left). Sequence logo for the human CTD heptad (right), n = number of heptad repeats. b A schematic representation of cis-trans prolyl-peptidyl isomerization of proline and hydroxyproline residues, respectively, (left) along with their free energy profiles (right). c Simplified sequences of the heptad repeat of the CTD variants used in this study. d Quantification of the in vitro liquid-liquid phase-separation (LLPS) assay, performed with the mGFP-CTD and its tyrosine variants, mGFP-CTD^Y1F and mGFP-CTD^Y1A. Each dot represents a detected droplet, the black dots indicate the median of droplet size per measurement (n = 3). The black line and error bars represent the mean ± SD of the three medians. e Same as (d) but with the mGFP-CTD and its proline variants, mGFP-CTD^P3G and mGFP-CTD^Hyp. f Representative micrographs of three independent LLPS mixing assays with mGFP-CTD (green) and mCherry-CTD (red), and with mGFP-CTD^Hyp(green) and mCherry-CTD (red), respectively. Merge = overlay of mCherry and GFP channels. Scale bar = 10 μm. (*) The intensity of all micrographs was uniformly enhanced in Fiji for better visibility. g The quantification of LLPS mixing assays shown in (f). Each dot represents a detected droplet, black dots indicate the median of droplet size (left), integrated intensity of all droplets from the mCherry (middle) or GFP (right) channels per measurement (n = 3). The black line and the error bar represent the mean ± SD of the three medians. Statistical comparison of means in datasets was determined via an unpaired, two-sided t-test (see “Methods” FM-image analysis), only p ≤ 0.05 is depicted. p-values: ((g) middle) * = 0.026, ((g) right) * = 0.019, ** = 0.0062, *** = 0.0009, respectively. h Same as (d) but with the mGFP-CTD and its serine variants, mGFP-CTD^S2A, mGFP-CTD^S5A, and mGFP-CTD^S7A. i Same as (d) but with the mGFP-CTD and its threonine variant, mGFP-CTD^T4G. All assays presented here were done in triplicates. Source data are provided as a Source Data file.

Fig. 2. MD simulations of RNAPII CTD constructs.

a Average intramolecular distance (r_ij) as a function of residue separation (|i-j|) for CTD^cons (gray circles), variants CTD^Y1F (orange triangles), and CTD^Y1A (green diamonds) in CG single chain simulations. The values of the Flory scaling exponent (ν; see “Methods” for details) are also shown (maximum SD is 0.003). The red dashed line represents the scaling behavior of an ideal chain. b Representative snapshots from CG condensed phase simulations of CTD^cons, variants CTD^Y1F and CTD^Y1A. Each molecule is colored differently using van der Waals spheres. Solvent omitted for clarity. c Density maps of the carbon atoms in the N- and C-termini capping groups for CTD^cons (left) and the double cis proline variant CTD^cisP3,6 (right). Two representative conformations of di-heptad constructs are shown (light blue tubes) with trans (green) and cis (orange) prolines. d Distributions of the radius of gyration for the constructs in (c). Mean ± SD is shown. e Per-residue intramolecular interaction energy maps for the constructs in (c). Values matching or exceeding –7.0 kJ/mol are colored in dark red. SD (0.003-2.586). f Representative structures of a compact (top left) and extended (top right) SPXX-motif, as defined by the distance (dashed line) between the Cα atoms of residues at positions 1 and 4 (green spheres). Backbone, tube; atoms, sticks. The distributions of the Cα-Cα distances for the CTD^cons, CTD^S2A, CTD^T4G, and CTD^S7A variants are also shown. The three SPXX motifs within a single di-heptad are displayed separately from top to bottom. In each plot, we provide the ratio between the area under the curve (AUC) up to 0.7 nm cutoff value (red dashed line) and the AUC beyond the cutoff. g Distributions of the Cα-Cα distances same as in (f) for CTD^cons and all cis proline variants. Note that shown results from all-atom simulations (c–g) were averaged over three replicas and all di-heptad constructs. In (d, f, and g) the distributions for CTD^cons are reprised as dashed black lines for ease of comparison. Source data are provided as a Source Data file.

Fig. 3. The impact of CTD heptad residues on phosphorylation by CDK7 and CTD phosphorylation analysis under phase-separation conditions.

a The initial reaction velocities of the human CDK7 complex with the mGFP-CTD substrate and its variants relative to the wild-type CTD. Black dots indicate the individual measurements (n = 3). Bars, mean; error bars, ± SD. b Representative micrographs of three independent in vitro LLPS kinase assays with 5 μM mGFP-CTD (green) and 0.5 mM ATP after 5 and 10 min with indicated concentrations of the CDK7 complex. Scale bar = 10 μm. c Quantifications of experiments shown in (b). Each dot represents a detected droplet, black dots indicate median of droplet size per measurement (n = 3). The black line and error bar represent the mean ± SD of the three medians. d Representative snapshots from CG condensed phase simulations of CTD molecules phosphorylated at S5. Rendering same as (b) in Fig. 2. e The CDK7 complex sedimentation assay in the presence of mGFP-CTD. A scan of representative SDS-PAGE (top)—(M) molecular weight marker; (I) input; (S) supernatant; (P) pellet. Quantifications of the sedimentation assays (bottom) (n = 3), mean ± SD. f Representative micrographs of at least two independent in vitro LLPS assays with mGFP-CTD (green) and increasing concentration of the CDK7 complex labeled with Alexa594 (red). Merge = overlay of GFP and Alexa 594 channels. Scale bar = 1 μm. (*) The intensity of all micrographs was uniformly enhanced in Fiji for better visibility. g Super-resolution micrographs show the CDK7 complex (red) localization at the surface of the mGFP-CTD (green) droplets. h Kinase assays with mGFP-CTD at different concentrations and 0.5 mM ATP with the CDK7 complex (at 5 nM), in the presence and absence of 10% dextran (n = 3), respectively, mean ± SD. i (left) Schematic representation of the sedimentation kinase assay workflow. i (right) Quantification of the sedimentation kinase assay (n = 3), mean ± SD. Statistical significance for (e, h, i) was determined by unpaired, two-sided t-test. p-values: (e) * = 0.030, (h) ** = 0.0093, * = 0.0292, **** ≤ 0.0001, (i) ** = 0.0012, **** = 0.0002. Source data are provided as a Source Data file.

Fig. 4. General overview of RNAPII CTD sequence and structural features, which have influence on RNAPII CTD phase-separation.

A model of a droplet containing the CTD of RNAPII highlights the interacting tyrosine sticker residues separated by spacers (middle). Spacers can adopt distinct β-turns depending on the cis-trans proline isomerization (left). RNAPII CTD droplet dissolution upon phosphorylation by the CDK7 complex (right). Letters in colored circles designate amino acid residues of the CTD heptad repeat. Peptides were drawn in ChemSketch (Freeware) 2024.1.1 (Advanced Chemistry Development, Inc).