Genome-wide functional CRISPR-Cas9 screen reveals CDK7 as a vulnerability in HNSCC cells
In an attempt to uncover essential genes with therapeutic potential, we conducted a CRISPR-Cas9 knockout screen across five different HNSCC cell lines: FaDu, UT-SCC38, HCA-LSC1, UT-SCC42B and Detroit 562, which recapitulate the genetic and clinical diversity typically found in HNSCC (Supplementary Fig. 1a, b). We used a comprehensive whole-genome library, targeting 18,010 genes with 90,709 different gRNAs. The HNSCC cell lines were transduced using a low multiplicity of infection (MOI) of 0.3 and subsequently selected with puromycin to ensure efficient integration of the gRNAs. Cells were cultured for 25 days, after which genomic DNA was extracted, and gRNAs were amplified and sequenced (Fig. 1a and Supplementary Fig. 1c-e). Guide-RNA depletion and enrichment was analyzed using MAGeCK software to identify negatively and positively regulated genes, indicative of a role in cell viability.
Cell-line-specific and commonly essential genes were determined based on their false discovery rate (FDR) score. Only genes with FDR < 0.10 were considered for further analysis, thus identifying between 500 and 1500 essential genes per cell line (Fig. 1b). A focus on essential genes across multiple cell lines revealed 228 genes consistently essential in all five analyzed HNSCC cell lines (Fig. 1c and Supplementary Data 1), including well-known oncogenes and drivers of HNSCC cell proliferation such as MYC and CCND1. Pathway analysis of the common essential genes showed a significant enrichment in MYC and E2F targets and cell cycle checkpoint-related gene sets (Fig. 1d).
Given the translational potential of CDK inhibitors, we focused our attention on the essentiality of CDK proteins as targetable and druggable candidates. CDK1 and CDK7 emerged as the most crucial genes across all five HNSCC cell lines, while other CDKs such as CDK2 or CDK5 were found to be non-essential, or showed selectivity to certain cell lines, such as CDK4, CDK6 or CDK9 (Fig. 1e). CDK1 is the only CDK in mammals that is essential for cell cycle progression. However, to date, CDK1 inhibitors have been limited by high toxicity and insufficient clinical efficacy. Interestingly, CDK7-selective inhibitors have been recently developed and are under current active testing in clinical trials. Moreover, CPTAC-HNSCC data analysis showed that CDK7 protein levels significantly increased in tumors compared to normal tissue counterparts (Supplementary Fig. 1f) and DepMap data analysis demonstrated that HNSCC cell lines are among the most sensitive to CDK7 inhibition (Supplementary Fig. 1g). On this basis, we selected CDK7 for further investigation due to its critical role in cell cycle progression and transcription, as well as its function as an upstream regulator of other CDKs, including CDK1 itself. Single gRNA analysis revealed that every gRNA targeting CDK7 was depleted at the final time point in all HNSCC cell lines (Fig. 1f).
To validate the essentiality of CDK7 as a candidate gene in our screen, we first analyzed CDK7 protein levels in our HNSCC cell line panel (Fig. 2a, b). We next performed targeted CRISPR knockouts of CDK7 (CDK7 KO) using two distinct single gRNAs in three HNSCC cell lines: FaDu, UT-SCC38 and HCA-LSC1, which harbored different endogenous levels of CDK7. We confirmed that these two gRNAs robustly reduced CDK7 protein levels in all three HNSCC cell lines (Fig. 2c). Subsequently, we carried out competitive co-culture assays between control and KO cells to validate the results of our screen (Fig. 2d). In these assays, CDK7 KO cells were significantly outcompeted by control cells in FaDu (Fig. 2e), UT-SCC38 (Fig. 2f) and HCA-LSC1 cells (Fig. 2g), thus confirming a significant reduction in cell viability upon CDK7 depletion. These results highlight the critical role of CDK7 in maintaining cell proliferation/survival and validate our initial findings from the CRISPR screen.
Even though comprehensive genomic analyses of the two selected CDK7-targeting gRNAs revealed no predicted high-risk off-target effects in coding regions (Supplementary Fig. 2a), we performed rescue experiments to further rule out potential non-specific effects. Specifically, we introduced a codon-optimized, CRISPR-resistant exogenous CDK7 construct designed to evade recognition by the targeting gRNAs, into three HNSCC cell lines from our panel. Western blot analysis confirmed efficient and selective depletion of endogenous CDK7 protein by both gRNAs across all three HNSCC cell lines, without affecting the exogenously expressed CDK7 (Supplementary Fig. 2b). Subsequently, competitive growth assays further demonstrated that exogenous expression of the codon-optimized CDK7 was effective and fully rescued the loss of cellular fitness caused by CDK7 knockout (Supplementary Fig. 2c-e). Together these results validate the specificity of our gRNAs and further support the essential role of CDK7 in HNSCC cell survival.
Furthermore, we assessed the impact of CDK7 KO on well-known downstream targets and cell cycle regulators. We found that CDK7 depletion led to reduced phosphorylation levels of the Carboxy-terminal domain (CTD) of RNA Polymerase II (Ser5), retinoblastoma protein (Rb) (Ser780), and CDK1 T-loop phosphorylation (Thr161) in the three HNSCC cell lines tested. Additionally, we observed a reduction in the total protein levels of Rb and CDK1 with the extent of these effects varying in a cell line-specific manner (Fig. 2h, i). These data demonstrate that CDK7 depletion effectively abrogates HNSCC cell proliferation by targeting several key downstream effectors and regulators of cell cycle control and transcription.
The robust effects of genetic CDK7 KO prompted us to evaluate the therapeutic potential of available CDK7-selective inhibitors in preclinical HNSCC models. We selected two distinct compounds: YKL-5-124, a selective covalent CDK7 inhibitor, and samuraciclib (CT7001, ICEC0942), an orally bioavailable, ATP-competitive inhibitor currently undergoing Phase I/II clinical trials in cancer patients. We first analyzed the effects of selective CDK7 inhibitors on the viability of a panel of five HNSCC cell lines. Both YKL-5-124 and samuraciclib (Fig. 3a) significantly reduced HNSCC cell viability, although lower IC50 values were observed for YKL-5-124 (ranging from 35 to 100 nM) than samuraciclib (range 30 to 200 nM) after 5 days of treatment (Fig. 3b).
Remarkably, both CDK7-selective inhibitors were more effective in decreasing HNSCC cell viability than the CDK4/6 inhibitor palbociclib (>10-fold higher IC50 values) (Fig. 3a, b). Analogous results were observed in colony formation assays performed over 14 days of treatment. Lower doses of YKL-5-124 and samuraciclib (ranging from 5 to 25 nM) completely abolished HNSCC cell proliferation, whereas considerably higher concentrations of palbociclib were required to achieve a comparable effect (Fig. 3c). Comparable results were obtained by testing two additional HNSCC cell lines derived from primary oral tumors (UT-SCC2 and Cal-33), showing similar IC50 values and anti-proliferative effects to the rest of our HNSCC panel from larynx and pharynx origin (Supplementary Fig. 3a, b).
We next investigated the immediate molecular effects of CDK7 inhibition on the phosphorylation of downstream protein substrates. HNSCC cells were treated with 1 µM of either YKL-5-124 or samuraciclib for 24 h to minimize potential compensatory mechanisms upon prolonged treatment. This resulted in decreased phosphorylation levels of CTD-RNA Polymerase II (Ser5), Rb (Ser780), and CDK1 (Thr161) in most HNSCC cell lines. Similarly, CDK7 inhibition resulted in decreased total CDK1 protein levels in a cell-line-dependent manner, mirroring the changes previously observed by CDK7 KO. CDK7 protein levels remained unchanged or slightly increased upon treatment with CDK7 inhibitors. Notably, a shift in CDK7 electrophoretic mobility was detected upon treatment with YKL-5-124 in all HNSCC, likely due to the covalent binding of this compound (Fig. 3d and Supplementary Fig. 3c-f).
The functional consequences of CDK7 pharmacological inhibition were also investigated on cell cycle dynamics and apoptosis. YKL-5-124 treatment led to cell accumulation in S and/or G2/M phases depending on the cell line analyzed, and samuraciclib treatment led to cell accumulation in G2/M phase in all five HNSCC cell lines (Fig. 3e, f and Supplementary Fig. 4a, b). This was also accompanied by a significant induction of apoptosis, especially after samuraciclib treatment, in all HNSCC cell lines tested (Fig. 3g, h and Supplementary Fig. 4c). Concordant to these findings, increased levels of cleaved PARP were also detected by western blot upon YKL-5-124 and samuraciclib treatment (Fig. 3d and Supplementary Fig. 3c-f). Concomitantly, selective CDK7 inhibition led to a significant increase in DNA damage, as indicated by elevated γH2AX levels in HNSCC cells treated with either YKL-5-124 or samuraciclib compared to control cells (Fig. 3i, j).
In order to delineate the global transcriptional changes caused by CDK7 inhibition, RNA-seq experiments were performed in FaDu cells treated with either YKL-5-124 or samuraciclib for 48 h. As represented by volcano plots, a total of 804 and 1,147 genes were found significantly up- and downregulated by YKL-5-124, respectively (Fig. 4a), and 630 and 989 genes were significantly up- and downregulated upon samuraciclib treatment, respectively (Fig. 4b). In addition, Venn diagrams depict the unique and overlapping changes by YKL-5-124 and samuraciclib, with a total of 645 common downregulated genes (Fig. 4c) and 376 common upregulated genes (Fig. 4d). GSEA using Reactome pathways database of common downregulated genes showed a predominance of gene sets related to cell cycle and mitotic pathways, as well as upregulation of specific gene sets by both compounds (Fig. 4e). GSEA of Hallmark gene sets further revealed a common downregulation of genes critical for cell cycle progression by both CDK7 inhibitors, including MYC and E2F targets, and G2/M checkpoint genes (Fig. 4f and Supplementary Fig. 5a). In addition, Supplementary Fig. 6 depicts gene expression changes across the cell cycle pathway upon treatment with CDK7 inhibitors. Moreover, genes involved in DNA repair were also found to be significantly and commonly downregulated by both compounds, suggesting a broader impact on cellular regulatory mechanisms essential for cell division and genome stability (Fig. 4f). Furthermore, YKL-5-124 treatment led to the downregulation of oncogenic pathways such as mTORC1, TNFα, and KRAS signaling, suggesting differential effects possibly due to varying mechanisms of action between these two CDK7 inhibitors (Supplementary Fig. 5b).
Interestingly, a number of essential genes identified in our genome-wide CRISPR screen (most of which undruggable targets) were significantly downregulated by CDK7 inhibition (Fig. 4g and Supplementary Data 2), suggesting a common downregulated profile of genetic vulnerabilities. According to these findings, CDK7 inhibition emerges as a promising therapeutic strategy to effectively and broadly target genetic vulnerabilities in HNSCC.
We next examined the effects of both genetic and pharmacological inhibition of CDK7 on tumor growth in vivo. Deletion of CDK7 dramatically impaired the tumor growth of FaDu and HCA-LSC1 cell lines subcutaneously injected in immunodeficient mice (Fig. 5a), reaching final tumor volume reductions of 60 and 75% respectively (Fig. 5b). Noteworthy, treatment with YKL-5-124 and samuraciclib in a therapeutic setting (Fig. 5c, d) was also demonstrated to be effective at reducing the tumor volumes of both FaDu and HCA-LSC1 xenografts (Fig. 5e, f). Notably, YKL-5-124 exhibited a more potent antitumor activity, showing tumor reductions up to 74% after 12 days of treatment (Fig. 5g). By contrast, samuraciclib led to lower and delayed antitumor effects with tumor reductions of 40% after 12 days of treatment (Fig. 5g). Remarkably, none of these compounds caused any sign of adverse effects on the mice or body weight reduction (Supplementary Fig. 7).
To further validate the therapeutic potential of CDK7 inhibition in a more advanced preclinical HNSCC model, we assessed the efficacy of samuraciclib and YKL-5-124 in a patient-derived xenograft (PDX) model (Fig. 6a). CDK7 protein expression in the PDX model was confirmed by IHC (Supplementary Fig. 8a). Both inhibitors led to a significant reduction in tumor growth compared to vehicle-treated controls, thereby confirming their potent antitumor activity in this clinically relevant setting (Fig. 6b). Importantly, both treatments were well tolerated in mice. No signs of systemic toxicity were observed, indicated by normal hematological parameters (Supplementary Fig. 8b) as well as the absence of histopathological abnormalities in liver tissue (Supplementary Fig. 8c). These data collectively support the in vivo safety profile of CDK7 inhibition.
Global transcriptomic analysis of PDX-derived tumors after 5 days of treatment further demonstrated distinct gene expression profiles between treated and control groups, as shown by principal component analysis (Fig. 6c). Differential expression analysis identified a substantial number of upregulated and downregulated genes following both samuraciclib and YKL-5-124 treatment (Fig. 6d, e, middle). Samuraciclib treatment significantly downregulated gene sets involved in cell cycle regulation, DNA replication, and MYC targets (Fig. 6d, left), while upregulated genes were enriched in stress response, p53 signaling, inflammatory pathways, and epithelial-to-mesenchymal transition (EMT) (Fig. 6d, right). YKL-5-124 induced a comparable transcriptional response, including strong downregulation of E2F target genes and clear upregulation of EMT-associated programs (Fig. 6e). However, YKL-5-124 appears to cause more widespread expression changes probably reflecting tumor sample variability or compound-specific effects.
Immunohistochemical analysis of PDX-derived tumors further contributed to demonstrate a consistent significant reduction in the proliferation markers Ki67 and phospho-histone H3 (pH3) in tumors treated with either compound (Fig. 6f, g).
Finally, the therapeutic potential of CDK7 inhibitors was also tested in two HNSCC patient-derived organoid (PDO) models harboring endogenous CDK7 protein expression (Fig. 7a). Treatment of already formed PDOs with either YKL-5-124 or samuraciclib was effective in abrogating the growth of both HNSCC PDOs tested (Fig. 7b). The effects of CDK7-selective inhibitors were compared to those of the CDK4/6 inhibitor palbociclib (Fig. 7c). IC50s ranged from 329 to 397 nM for YKL-5-124 and 399 to 666 nM for samuraciclib, but were higher for palbociclib (Fig. 7d). Interestingly, PDO_84 showed resistance to palbociclib, while PDO_55 exhibited a much lower sensitivity to palbociclib compared to both CDK7 inhibitors (IC50 of 2 µM) (Fig. 7c, d). These data are in line with our results when testing these three compounds on the proliferation and viability of HNSCC cell lines (Fig. 3a-c). Representative images of PDO treatments are shown in Fig. 7e.
Also relevant from a therapeutic point of view, we explored the ability of CDK7-selective inhibitors to effectively target and eradicate cancer stem cells (CSC)-enriched HNSCC tumorsphere cultures. Clonal sphere-forming ability in non-adherent serum-free culture conditions is a hallmark of self-renewal and CSC-related phenotype. This is also the basis for the formation and growth of tumor organoids, which are 3D structures cultivated from patient-derived stem cells. Thus, we found that YKL-5-124 and samuraciclib were both effective in reducing the viability of CSC-enriched tumorsphere cultures in FaDu and UT-SCC38 cells, in a dose-dependent manner (Supplementary Fig. 9).