Nature Article

\documentclass[11pt]{article}
\usepackage[margin=1in]{geometry}
\usepackage{times}
\usepackage{graphicx}
\usepackage[super,sort&compress,comma]{natbib}
\usepackage{amsmath,amssymb}
\usepackage{setspace}
\usepackage{lineno}
\usepackage{booktabs}
\usepackage{hyperref}
\doublespacing
\linenumbers

\title{\textbf{\Large Structural basis of selective kinase inhibition reveals an allosteric regulatory mechanism}}
\author{
  First Author$^{1}$,
  Second Author$^{2}$,
  Third Author$^{1,2,*}$
}
\date{}

\begin{document}
\maketitle

\noindent
$^{1}$Department of Structural Biology, Stanford University School of Medicine, Stanford, CA 94305, USA.\\
$^{2}$Department of Chemistry, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.\\
$^{*}$Correspondence to: [email protected]

\vspace{1em}
\noindent\textbf{Summary paragraph.}
Protein kinases regulate nearly every aspect of cellular signalling, and their dysregulation is implicated in cancer, autoimmune disorders, and neurodegeneration. Despite the clinical success of kinase inhibitors, achieving selectivity among the more than 500 human kinases remains a central challenge due to the high conservation of the ATP-binding site. Here we report the crystal structures of cyclin-dependent kinase 7 (CDK7) bound to a series of covalent inhibitors at resolutions between 1.8 and 2.3~\AA. The structures reveal an unexpected allosteric pocket formed by displacement of the activation-loop DFG motif, which is exploited by the most selective compound in the series. Biochemical assays confirm that this compound inhibits CDK7 with an IC$_{50}$ of 3.2~nM while showing more than 1{,}000-fold selectivity over the closely related CDK2 and CDK9. Hydrogen-deuterium exchange mass spectrometry demonstrates that inhibitor binding stabilizes the inactive C-helix-out conformation, providing a mechanistic basis for selectivity. These findings establish a structural framework for the rational design of selective CDK7 inhibitors with therapeutic potential.

\section*{Main}
Cyclin-dependent kinases (CDKs) are serine/threonine kinases that control cell-cycle progression and transcriptional regulation~\cite{ref1}. CDK7, as part of the CDK-activating kinase (CAK) complex, phosphorylates the activation loop of cell-cycle CDKs and the C-terminal domain of RNA polymerase II, placing it at the nexus of proliferation and gene expression~\cite{ref2}. Elevated CDK7 activity has been observed in triple-negative breast cancer, acute myeloid leukaemia, and small-cell lung cancer, motivating the development of CDK7-targeted therapies~\cite{ref3}.

First-generation CDK7 inhibitors achieved nanomolar potency but suffered from off-target activity against CDK2 and CDK9, leading to dose-limiting toxicities in clinical trials. The structural basis for this cross-reactivity has remained unclear, in part because prior crystallographic studies captured only the active DFG-in conformation of the kinase domain. We hypothesized that covalent inhibitors targeting a non-conserved cysteine residue (Cys312) near the active site could stabilize an alternative conformation amenable to selective targeting.

To test this hypothesis, we synthesized a focused library of 28 acrylamide-based compounds and determined co-crystal structures of the five most potent analogues bound to CDK7. The highest-resolution structure (compound~\textbf{5}, 1.8~\AA) revealed that the inhibitor displaces the DFG motif by 4.7~\AA{} relative to the apo structure, opening an allosteric pocket lined by non-conserved residues Leu83, Val94, and Phe101 (Fig.~1). This pocket is absent in CDK2 and CDK9, where the corresponding residues are bulkier (Ile, Leu, and Tyr, respectively), providing a structural explanation for the observed selectivity.

\begin{figure}[t]
\centering
\fbox{\parbox[c][7cm][c]{0.85\linewidth}{\centering Crystal structure of CDK7 bound to compound~\textbf{5} (1.8~\AA{} resolution). The allosteric pocket formed by DFG-out displacement is highlighted in blue. Key hydrogen bonds to Cys312 and Glu85 are shown as dashed lines.}}
\caption{{\bf Structural basis of CDK7-selective inhibition.} The covalent bond to Cys312 anchors compound~\textbf{5} in the ATP-binding site, while the pyridine substituent extends into the allosteric pocket. Residues lining the pocket are shown as sticks; electron density ($2F_o - F_c$, contoured at 1.0$\sigma$) is shown as blue mesh around the inhibitor.}
\label{fig:structure}
\end{figure}

\begin{table}[t]
\centering
\caption{{\bf Inhibitory activity and selectivity of lead compounds against CDK family members.} IC$_{50}$ values (nM) determined by radioactive kinase assay; selectivity ratio is IC$_{50}$(off-target)/IC$_{50}$(CDK7).}
\label{tab:ic50}
\begin{tabular}{lccccc}
\toprule
Compound & CDK7 & CDK2 & CDK9 & Selectivity (CDK2) & Selectivity (CDK9) \\
\midrule
\textbf{1} & 45 & 120 & 89 & 2.7$\times$ & 2.0$\times$ \\
\textbf{3} & 12 & 980 & 340 & 82$\times$ & 28$\times$ \\
\textbf{5} & 3.2 & $>$5{,}000 & 4{,}100 & $>$1{,}500$\times$ & 1{,}280$\times$ \\
\textbf{7} & 8.4 & 3{,}200 & 1{,}700 & 380$\times$ & 200$\times$ \\
\textbf{9} & 22 & 1{,}400 & 560 & 64$\times$ & 25$\times$ \\
\bottomrule
\end{tabular}
\end{table}

Kinase selectivity assays across a panel of 468 kinases confirmed that compound~\textbf{5} is among the most selective CDK inhibitors reported to date, with an S-score(10) of 0.005 (Table~\ref{tab:ic50}). Hydrogen-deuterium exchange mass spectrometry (HDX-MS) revealed that compound~\textbf{5} binding reduces deuterium uptake in the $\alpha$C-helix region by 40\% relative to the apo enzyme, consistent with stabilization of the inactive C-helix-out conformation observed crystallographically.

In cellular assays, compound~\textbf{5} selectively suppressed phosphorylation of the RNA Pol~II CTD (Ser5) with an EC$_{50}$ of 18~nM in MDA-MB-231 triple-negative breast cancer cells, without affecting CDK2-dependent Rb phosphorylation at concentrations up to 10~$\mu$M. Colony formation was reduced by 85\% at 50~nM, and synergy was observed with the PARP inhibitor olaparib (combination index = 0.35).

These results demonstrate that covalent targeting of Cys312 combined with exploitation of a conformationally gated allosteric pocket enables exquisite selectivity among closely related CDKs. The structural framework presented here should guide the design of next-generation CDK7 inhibitors with improved pharmacokinetic properties for clinical development.

Several aspects of this work merit discussion. The allosteric pocket identified here is formed only upon DFG-out displacement, explaining why it was not detected in previous crystallographic studies that captured the active conformation. This conformational selectivity mechanism is conceptually distinct from Type~II inhibitors that also exploit the DFG-out state but bind to a pocket adjacent to the gatekeeper residue rather than the activation loop. The covalent bond to Cys312 provides an additional layer of selectivity, as this residue is present in only 11 of the 518 human kinases.

A limitation of the current study is that pharmacokinetic properties of compound~\textbf{5} were not optimized; preliminary microsomal stability assays indicate a half-life of approximately 25~minutes in human liver microsomes, suggesting that prodrug strategies or metabolic stabilization will be required for oral dosing. Additionally, while the ternary complex with cyclin~H and MAT1 provides the most physiologically relevant structural context, the resolution of these structures is limited by crystal packing contacts that partially occlude the allosteric pocket.

\section*{Methods}

\subsection*{Protein expression and purification}
Human CDK7 (residues 1--346) in complex with cyclin H (residues 1--323) and MAT1 (residues 229--309) was co-expressed in Sf9 insect cells using the Bac-to-Bac system. The ternary complex was purified by Ni-NTA affinity chromatography followed by size-exclusion chromatography on a Superdex 200 column equilibrated in 25~mM Tris-HCl pH~7.5, 150~mM NaCl, and 1~mM TCEP.

\subsection*{Crystallography}
Crystals of CDK7--cyclin H--MAT1 were grown by hanging-drop vapor diffusion at 18$^\circ$C. Compounds were soaked into pre-formed crystals at 1~mM for 4~hours. Diffraction data were collected at beamline 12-2, Stanford Synchrotron Radiation Lightsource, and processed with XDS. Structures were solved by molecular replacement using Phaser with PDB entry 1UA2 as the search model. Refinement was performed with PHENIX, and models were validated with MolProbity.

\subsection*{Kinase activity assays}
IC$_{50}$ values were determined using a radioactive $^{33}$P-ATP filter-binding assay with a CTD heptapeptide substrate. Reactions contained 10~nM CDK7--cyclin H--MAT1, 10~$\mu$M ATP, and varying concentrations of inhibitor. Selectivity profiling across 468 kinases was performed by DiscoverX using the scanMAX platform.

\subsection*{HDX-MS}
Hydrogen-deuterium exchange was initiated by diluting protein (10~$\mu$M) 10-fold into D$_2$O buffer. Exchange was quenched at 10~s, 1~min, 10~min, and 60~min by addition of ice-cold quench buffer (pH~2.5). Peptic peptides were analyzed on an Orbitrap Exploris 480 mass spectrometer coupled to an M-class UPLC system. Deuterium uptake was quantified using HDExaminer v3.1 and mapped onto the crystal structure.

\subsection*{Cell-based assays}
MDA-MB-231 cells were maintained in DMEM supplemented with 10\% FBS at 37$^\circ$C and 5\% CO$_2$. For CTD phosphorylation assays, cells were treated with compound for 4~hours, lysed in RIPA buffer, and analyzed by western blot with anti-phospho-Ser5 RPB1 antibody (Cell Signaling \#13523). Colony formation assays were performed by seeding 500 cells per well in 6-well plates, treating with compound or DMSO for 14~days, and staining with crystal violet. Combination indices for olaparib synergy were calculated using the Chou--Talalay method.

\subsection*{Statistical analysis}
IC$_{50}$ values were determined by nonlinear regression of dose--response curves using GraphPad Prism~10. All biochemical measurements were performed in triplicate and are reported as mean $\pm$ standard deviation. Selectivity scores (S-scores) were calculated as the fraction of kinases with percent inhibition $>$90\% at 1~$\mu$M.

\section*{Extended Data}

\subsection*{Extended Data Fig.~1}
Electron density maps ($2F_o - F_c$ at 1.0$\sigma$) for all five inhibitor-bound structures, showing unambiguous placement of each compound in the active site. Omit maps ($F_o - F_c$ at 3.0$\sigma$) calculated prior to ligand placement confirm the binding poses.

\subsection*{Extended Data Fig.~2}
HDX-MS difference plots (compound~\textbf{5} minus apo) mapped onto the CDK7 structure. Regions showing $>$10\% protection from exchange are colored blue; regions showing deprotection are colored red. The $\alpha$C-helix and activation loop show the largest protection, consistent with conformational stabilization upon inhibitor binding.

\subsection*{Extended Data Table~1}
X-ray data collection and refinement statistics for all five co-crystal structures, including space group, unit cell dimensions, resolution, $R_{\text{work}}$/$R_{\text{free}}$, and Ramachandran statistics.

\subsection*{Extended Data Table~2}
Kinome-wide selectivity panel results for compound~\textbf{5} at 1~$\mu$M. Of 468 kinases tested, only CDK7 showed $>$90\% inhibition (99.2\%). Four additional kinases showed 50--90\% inhibition: GSK3$\beta$ (62\%), DYRK1A (58\%), CLK2 (54\%), and HIPK2 (51\%).

\subsection*{Data availability}
Atomic coordinates and structure factors have been deposited in the Protein Data Bank under accession codes XXXX--XXXX. Source data for all figures and tables are provided with this paper. Additional datasets are available from the corresponding author upon reasonable request.

\subsection*{Code availability}
Custom scripts for HDX-MS data analysis are available at \url{https://github.com/example/cdk7-hdx}.

\section*{References}
\bibliographystyle{naturemag}
\begin{thebibliography}{9}
\bibitem{ref1}
  Malumbres, M. Cyclin-dependent kinases. \emph{Genome Biol.} \textbf{15}, 122 (2014).
\bibitem{ref2}
  Fisher, R.P. Cdk7: a kinase at the core of transcription and the cell cycle. \emph{J. Cell Sci.} \textbf{118}, 5171--5180 (2005).
\bibitem{ref3}
  Kwiatkowski, N. et al. Targeting transcription regulation in cancer with a covalent CDK7 inhibitor. \emph{Nature} \textbf{511}, 616--620 (2014).
\bibitem{ref4}
  Endicott, J.A., Noble, M.E.M. \& Johnson, L.N. The structural basis for control of eukaryotic protein kinases. \emph{Annu. Rev. Biochem.} \textbf{81}, 587--613 (2012).
\bibitem{ref5}
  Hu, S. et al. Discovery of a highly selective CDK7 inhibitor with clinical activity in models of therapy-resistant cancer. \emph{Nat. Chem. Biol.} \textbf{15}, 792--800 (2019).
\end{thebibliography}

\vspace{1em}
\vspace{1em}
\noindent\textbf{Acknowledgements.} We thank the staff at SSRL beamline 12-2 for assistance with data collection, J.~Smith for assistance with compound synthesis, and members of the structural biology group for helpful discussions. This work was supported by NIH grants R01-CA234567 and P30-CA124435, a Damon Runyon Cancer Research Foundation fellowship (to B.B.), and a Stanford Bio-X seed grant.

\noindent\textbf{Author contributions.} A.A.\ conceived and designed the study, determined crystal structures, and wrote the manuscript. B.B.\ designed and synthesized all inhibitor compounds and performed biochemical kinase assays. C.C.\ conducted HDX-MS experiments, performed cellular assays, and analyzed data. All authors contributed to data interpretation and manuscript revision.

\noindent\textbf{Competing interests.} A patent application covering compounds described in this study has been filed by Stanford University (application no. PCT/US2024/XXXXXX). The authors declare no other competing interests.

\noindent\textbf{Correspondence.} Correspondence and requests for materials should be addressed to [email protected].

\end{document}