Boosts ICB Response via STING Activation

CDK12/13 inactivation or attenuated expression levels are associated with elevated STING activation and improved response to ICB. We previously reported that inactivating mutations in CDK12 were associated with higher levels of intratumoral T cells in advanced prostate cancer (17). To understand the underlying mechanisms, we performed spatial transcriptomic analyses on metastatic castration-resistant prostate cancer (mCRPC) samples with either inactivated or wild-type CDK12. We found that inactivation of CDK12 was significantly associated with a tumor-specific enrichment of STING activity signature (8) (Figure 1A), along with signatures primarily composed of IFN-stimulated genes that could be induced by STING activation (8) (Figure 1A). These signatures included type I (IFN-α) and II (IFN-γ) IFN responses and antigen presentation (Figure 1A). Additionally, gene set enrichment analysis (GSEA) utilizing the MSigDB Hallmark database revealed that type I and II IFN responses were the most upregulated pathways in CDK12-inactivated tumors compared with the wild-type (Figure 1A). Mutations of CDK12 are rare in cancers, other than ovarian cancer and advanced prostate cancer. However, low expression of CDK12 was significantly associated with increased STING activity in various cancer types (Supplemental Figure 1A, top; supplemental material available online with this article; https://doi.org/10.1172/JCI193745DS1). CDK13, a paralog of CDK12, has been shown to have functional redundancy with CDK12 (13, 14, 16). Notably, low expression of CDK13 also tended to be associated with activated STING signaling (Supplemental Figure 1A, bottom). Importantly, the combined low expression of both CDK12 and CDK13 exhibited the strongest association with elevated STING activity (Figure 1B), along with significant upregulation of IFN response (Supplemental Figure 1B). Therefore, attenuated expression levels of CDK12/13 may lead to the activation of STING signaling in various cancer types.

CDK12/13 inactivation or low expression levels are associated with elevated STING activation and improved response to ICB. (A) Analyses of 10× Genomics Visium spatial transcriptomics on CDK12 mutant versus wild-type (WT) metastatic castration-resistant prostate cancer (mCRPC) samples. Top: Enrichment of the indicated pathways in tumor cells. Middle: Representative images illustrating the enrichment of STING activity signature in tumor areas. Bottom: The top 10 pathways enriched by gene set enrichment analysis (GSEA), utilizing the MSigDB Hallmark database, in tumor cells. Type I and II IFN responses are highlighted in red. Adj., adjusted. Scale bar: 200 μm. (B) Association between expression of CDK12/13 and STING activity signature in the indicated cancer types. Data were acquired from The Cancer Genome Atlas datasets. Abbreviations are defined in the legend of Supplemental Figure 1. (C) Association between pretreatment expression of CDK12/13 and overall survival in the indicated cohorts treated with ICB therapy. MI-ONCOSEQ: a pan-cancer cohort at the University of Michigan. The KM plotter data were acquired from the KM plotter database (https://kmplot.com/analysis/). (D) Association between pretreatment expression of CDK12/13 and ICB response in the melanoma and MI-ONCOSEQ ICB cohorts in C. (E) Single-cell RNA sequencing assessing expression of indicated genes or signature in tumor cells in ICB-treated cohorts. Left: Expression of CDK12/13 in tumor cells in patients with favorable versus unfavorable clinical outcomes. Right: Expression of STING activity signature in tumor cells in patients with favorable versus unfavorable clinical outcomes. Data were acquired from published research articles (see Methods). Low expression of CDK12/13 was defined as the bottom 20th percentile within each cohort. Two-tailed t tests were performed in B (with Bonferroni’s correction) and E, log-rank tests in C, and Fisher’s exact test in D. Panels B and D show box-and-whisker plots with the median (center line), 25th–75th percentiles (box), 10th–90th percentiles (whiskers), and outliers beyond the whiskers.

Activation of STING signaling has been shown to enhance the efficacy of ICB therapy (9, 21). Given the strong association between STING activity and CDK12/13 expression, we examined whether low expression of CDK12/13 predicted survival in patients treated with ICB therapy. We first analyzed RNA sequencing data from tumor samples collected prior to ICB treatment in 2 metastatic melanoma cohorts (22, 23) and found that low expression of CDK12/13 (bottom 20th percentile) was significantly associated with improved overall survival (Figure 1C). We next evaluated the predictive value of CDK12/13 expression in a pan-cancer setting. In the pan-cancer ICB cohort from the University of Michigan (MI-ONCOSEQ; n = 108; Supplemental Table 1), low pretreatment expression of CDK12/13 (bottom 20th percentile) was similarly associated with improved survival, reaching borderline statistical significance (P = 0.0519; Figure 1C), suggesting that a larger sample size may be required to achieve statistical significance. To further validate this association, we analyzed data from the Kaplan-Meier (KM) plotter database (https://kmplot.com/analysis/) (24, 25), which comprises a substantially larger ICB-treated pan-cancer population (n = 808; Supplemental Table 2). As expected, low pretreatment CDK12/13 expression (bottom 20th percentile) was significantly associated with improved survival (P = 0.0038; Figure 1C). Collectively, these findings support low CDK12/13 expression as a predictive biomarker for improved overall survival in ICB-treated patients.

Additionally, low expression of CDK12/13 significantly predicted response to ICB therapy (Figure 1D). Notably, multiple linear regression analysis revealed that low pretreatment CDK12/13 expression (Supplemental Figure 2, left) or high pretreatment STING activity signature expression (Supplemental Figure 2, right) was independently associated with improved clinical outcomes, after adjusting for age, sex, and tumor site. We further explored single-cell sequencing data from ICB-treated cohorts to specifically evaluate the expression of CDK12/13 in tumor cells (Figure 1E). As anticipated, we observed that reduced expression of CDK12/13 (Figure 1E) or elevated STING activity (Figure 1E) was significantly associated with favorable clinical outcomes in these cohorts. Collectively, these findings highlight that low expression of CDK12/13 is predictive of better clinical outcomes in cohorts treated with ICB.

Targeting CDK12/13 activates STING signaling. We next sought to determine whether targeting CDK12/13 could activate STING signaling. We previously developed a highly selective dual CDK12/13 PROTAC degrader, YJ1206, which is orally bioavailable and shows a favorable safety profile (20). In a murine prostate cancer model, Myc-CaP, we treated cells with YJ1206 and subsequently performed RNA sequencing analyses. The results revealed that the STING activity signature, as well as signatures that could be induced by STING activation (8), were significantly enriched in YJ1206-treated cells compared with controls (Figure 2A and Supplemental Figure 3A). Immunoblotting confirmed that treatment with YJ1206 markedly activated STING signaling, as evidenced by the increased levels of phospho-STING (p-STING), phospho-TBK1 (p-TBK1), and phospho-IRF3 (p-IRF3) (Figure 2B). Importantly, treatment with YJ1206 efficiently degraded CDK12/13, leading to a decrease in phosphorylation of serine 2 (Ser2) on RNAPII and an increase in γH2AX (Figure 2B and Supplemental Figure 3B), consistent with previous findings (20). To verify that the effect of YJ1206 was on target, we depleted CDK12/13 using gene-specific siRNAs. As expected, genetic depletion of CDK12/13 also resulted in STING signaling activation, decreased p-Ser2 levels, and increased γH2AX expression (Figure 2B). The activation of STING signaling was further confirmed by the elevated expression of its downstream targets, such as Ifnb1, Cxcl10, and Ccl5 (Figure 2C and Supplemental Figure 3C), as well as genes responsive to type I IFN, H2-K1 (an MHC-I gene; Figure 2C) and Cd274 (which encodes PD-L1; Supplemental Figure 3C). Importantly, similar findings were observed in additional murine cancer models, B16-F10 (melanoma) and CT26 (colon carcinoma) (Figure 2, D–F), as well as across a panel of human cancer cell lines (Figure 2, E and F, and Supplemental Figure 3D). This indicates that the phenotype was not model or strain specific. Of note, inhibition of many other oncotherapeutic targets, including the BET family of bromodomain proteins, ubiquitin-like modifier activating enzyme 1, mSWI/SNF ATPases, GSPT1, and CBP/p300, did not activate STING signaling, suggesting a distinct role of CDK12/13 degradation in activating this pathway (Supplemental Figure 3E). Moreover, inhibition of CDK12/13 with a selective CDK12/13 inhibitor, YJ5118, induced STING activity at a level comparable to that observed with the CDK12/13 degrader (Supplemental Figure 3, F and G). Collectively, these findings show that targeting CDK12/13 activates STING signaling in tumor cells.

Targeting CDK12/13 activates STING signaling. (A) Enrichment of the indicated pathways in Myc-CaP cells treated with YJ1206 at 1 μM for 24 hours. Adj., adjusted. (B) Immunoblot of the indicated proteins in Myc-CaP cells treated with YJ1206 at increasing concentrations for 4 hours, or siRNAs targeting Cdk12 and/or Cdk13. Nontargeting siRNA was used as control. GAPDH was used as a loading control. (C) Top: Analysis of the indicated gene expression by RT-qPCR in Myc-CaP cells treated with YJ1206 at 1 μM for 15 hours, or siRNAs targeting Cdk12 and/or Cdk13. Nontargeting siRNA was used as control. Bottom left: IFN-β ELISA results in Myc-CaP cells treated as described above. Bottom right: Flow cytometry assessing surface MHC-I expression in Myc-CaP cells treated as described above. (D) Immunoblot of the noted proteins in B16-F10 cells treated with YJ1206 at increasing concentrations for 4 hours. (E and F) Flow cytometry median fluorescence intensity (MFI) quantifications of surface MHC-I (E) or PD-L1 (F) in the indicated cells treated with YJ1206 at 1 μM or 3 μM for 15 hours. (G) Quantification of immunofluorescence DNA/RNA hybrid (red) staining in Myc-CaP cells treated with 1 μM YJ1206 for 4 hours or siRNA targeting Cdk12 and/or Cdk13, with/without RNase H. Representative images are in Supplemental Figure 4D. Nontargeting siRNA was used as control. Forty (siRNA treatment) or 20 (YJ1206 or DMSO treatment) cells were used per data point. Data in C, E, and F are displayed as mean ± SD of triplicate experiments. Data in G are presented as box-and-whisker plots, with the median (center line), 25th–75th percentiles (box), and minimum to maximum values (whiskers). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by 2-tailed t test. NS, not significant. Bonferroni’s correction was applied for multiple comparisons in C and E–G.

Activation of the STING pathway may lead to activation of NF-κB signaling (5), which may upregulate MHC-I expression (26, 27). To identify key components involved in MHC-I upregulation via CDK12/13 antagonism, we knocked out Cgas, Sting1 (encoding STING), Ifnar1 (encoding IFNAR1), or Rela (encoding p65; Supplemental Figure 4A) in cancer cell lines. As anticipated, knockout (KO) of Cgas, Sting1, or Ifnar1 abolished MHC-I induction by CDK12/13 antagonism in various cancer models (Supplemental Figure 4B). Interestingly, while Rela KO significantly reduced MHC-I expression — consistent with its known role as an MHC-I regulator — CDK12/13 antagonism still significantly induced MHC-I expression in Rela-KO cells (Supplemental Figure 4C). These findings suggest that the cGAS/STING pathway and type I IFN response are crucial for MHC-I upregulation mediated by CDK12/13 antagonism, whereas NF-κB signaling is not essential.

Transcription replication conflicts (TRCs) have been linked to loss of the transcription elongation kinase CDK12 (13). R-loops (RNA-DNA hybrids) formed as a result of TRCs have been attributed to triggering the release of cytosolic DNAs, which activate STING signaling via cGAS (28, 29). Therefore, we hypothesized that genetic depletion or pharmacological degradation of CDK12/13 may lead to R-loop formation. We, thus, depleted Cdk12, Cdk13, or both in tumor cells with siRNAs and assessed R-loops with the S9.6 antibody (30, 31), the specificity of which was confirmed using RNase H treatment (Figure 2G and Supplemental Figure 4D). Cdk12 depletion resulted in increased R-loop formation (Figure 2G and Supplemental Figure 4D). Similarly, depletion of Cdk13, the paralog of Cdk12, also led to elevated R-loop levels (Figure 2G and Supplemental Figure 4D). Notably, simultaneous depletion of both Cdk12 and Cdk13 resulted in the highest increase in R-loop formation (Figure 2G and Supplemental Figure 4D). This observation was confirmed through treatment with the CDK12/13 degrader, YJ1206 (Figure 2G and Supplemental Figure 4D). Thus, genetic depletion or pharmacological degradation of CDK12/13 gives rise to a drastic increase in R-loop formation.

STING activation induced by CDK12/13 inactivation is TRC dependent. We next examined whether CDK12/13 degradation also resulted in formation of TRCs, using the proximity ligation assay (PLA) to detect the physical proximity of proliferating cell nuclear antigen (PCNA) to RNAPII (32). We observed that depletion of either Cdk12 or Cdk13 yielded a relatively moderate increase in TRC formation (Figure 3, A and B, and Supplemental Figure 5A). In contrast, simultaneous depletion of both Cdk12 and Cdk13 led to a drastic increase in TRC formation (Figure 3, A and B). This was further validated with the dual CDK12/13 PROTAC degrader (Figure 3, A and B) or depleting Cdk12/13 using CRISPR-Cas9 (Supplemental Figure 5B). The specificity of the PLA was confirmed with 5,6-dichloro-1-β-D-ribofuranosyl-benzimidazole (DRB) or triptolide treatment, which blocked transcription (32) (Figure 3, A and B). Collectively, our findings indicate that depletion or degradation of CDK12/13 leads to TRC formation.

STING activation induced by CDK12/13 inactivation is TRC dependent. (A and B) Representative images (A) or quantification (B) of immunofluorescence assessing PCNA-RNAPII PLA foci in Myc-CaP cells treated with siRNA targeting Cdk12 and/or Cdk13, or YJ1206 at 3 μM for 4 hours, with or without DRB or triptolide treatment. (C–E) Representative images (C) or quantification (D and E) of dsDNA (C, left, and D) and ssDNA (C, right, and E) in Myc-CaP cells treated with 3 μM YJ1206 for 4 hours, with or without DRB treatment. Scale bars: 5 μm. (F) ELISA measuring cGAMP levels in Myc-CaP cells treated with YJ1206 at the indicated concentrations. (G) Immunoblot of the noted proteins in Myc-CaP cells treated with YJ1206 at 1 μM, with or without DRB for 4 hours. Data in D and E are presented as box-and-whisker plots, with the median (center line), 25th–75th percentiles (box), and minimum to maximum values (whiskers). Data are displayed as mean ± SEM in F of triplicate experiments. One hundred cells were used per data point in B, D, and E. ****P < 0.0001 by 2-tailed t test. NS, not significant. Bonferroni’s correction was applied for multiple comparisons.

Head-on TRCs, whereby the transcription and replication machinery move toward each other, are known to promote R-loop formation (33, 34), which can contribute to genomic instability (34–36). We treated cells in which CDK12/13 were degraded with DRB, which abolished TRC formation (Figure 3, A and B), and measured the R-loop levels. As anticipated, DRB treatment also eliminated the increased R-loop levels resulting from CDK12/13 degradation (Supplemental Figure 5, C and D). To establish the causal link between TRCs and R-loop formation, we further inhibited DNA replication using aphidicolin (37) or hydroxyurea (38). As anticipated, treatment with aphidicolin or hydroxyurea abolished TRCs induced by CDK12/13 antagonism (Supplemental Figure 6A). Notably, R-loop accumulation triggered by CDK12/13 antagonism was also strongly diminished under these conditions (Supplemental Figure 6A). These findings support a model in which CDK12/13 antagonism promotes R-loop formation secondary to TRC induction.

R-loops have been found to induce the release of cytosolic ssDNAs (29) and dsDNAs (28) to activate STING via 2′3′-cGAMP (cGAMP) production (28). As expected, CDK12/13 degradation resulted in elevated levels of both cytosolic dsDNAs and ssDNAs (Figure 3, C–E, and Supplemental Figure 6B), as well as increased cGAMP production (Figure 3F). In contrast, elevation in cytosolic dsRNAs was not detected upon CDK12/13 degradation (Supplemental Figure 6B, bottom). Importantly, DRB treatment, which abolished TRC formation (Figure 3, A and B), fully eliminated the increase in cytosolic DNAs and, thus, reversed the YJ1206-mediated activation of STING signaling (Figure 3G). Taken together, these data show that targeting of CDK12/13 induces TRCs and R-loop formation, which in turn leads to the release of cytosolic DNAs and production of cGAMP, activating STING signaling.

CDK12/13 inactivation activates antitumor immunity and enhances response to ICB. Given that activation of STING enhances the efficacy of ICB therapy (9, 21), we next examined whether inactivating CDK12/13 could improve response to ICB. We depleted Cdk12, Cdk13, or both in the Myc-CaP tumor model (Supplemental Figure 7A, top). Notably, simultaneous depletion of both Cdk12 and Cdk13 led to the most profound improvement in response to anti–PD-1 treatment (Figure 4A and Supplemental Figure 7A, bottom), while depletion of either Cdk12 or Cdk13 alone led to relatively moderate enhancement of anti–PD-1 efficacy (Figure 4A). Furthermore, Cdk12/13 depletion significantly delayed tumor growth (Supplemental Figure 7A, bottom), extending the time to endpoint from 13 days in controls to 29 days after treatment (Figure 4A). To confirm that the tumor inhibitory effect was immune dependent, we evaluated tumor growth in both immunodeficient and immunocompetent mice. Although Cdk12/13 depletion exerted a direct tumor inhibitory effect in immunodeficient mice (Supplemental Figure 7B), consistent with previous studies (13), the tumor growth delay was significantly more pronounced in immunocompetent mice (Supplemental Figure 7B). These findings indicate that an intact immune system is essential for the full antitumor response mediated by Cdk12/13 depletion.

CDK12/13 inactivation activates antitumor immunity and enhances response to ICB. (A) Growth curves of subcutaneous (s.c.) tumors derived from Myc-CaP cells with or without Cdk12, Cdk13, or Cdk12/13 depletion in FVB mice (n = 4–5 mice per group) treated with IgG or anti–PD-1 (α-PD-1). (B) Growth curves of s.c. tumors derived from the Cdk12KO-sgp53 tumor cells in C57BL/6 mice (n = 5–6 mice per group) treated with vehicle, α-PD-1, YJ1206, or the combination of α-PD-1 and YJ1206 (combo). (C) Growth curves of s.c. tumors derived from Myc-CaP cells in FVB mice (n = 5–6 mice per group) treated with vehicle, α-PD-1, cGAMP, or the combination of α-PD-1 and cGAMP (combo). Intratumoral injections were performed for cGAMP administration. (D and E) Growth curves of s.c. tumors derived from the specified tumor cells in the indicated mice (n = 5–10 mice per group) treated with vehicle, α-PD-1, YJ1206, or the combination of α-PD-1 and YJ1206 (combo). (F) Assessment by CombPDX for synergism of α-PD-1 and YJ1206 in the indicated models treated in B, D, and E. (G) Body weight of the indicated mice after treatment by the indicated agents. YJ1206 was administered orally at a dose of 100 mg/kg, 3 times per week, and α-PD-1 was administered intraperitoneally at a dose of 200 μg/mouse every 3 days. Data are displayed as mean ± SEM in A–E and as mean ± SD in G. Significance in B–E was determined by 2-way ANOVA. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Bonferroni’s correction was applied for multiple comparisons.

The necessity of targeting both CDK12 and CDK13 to achieve optimal enhancement of ICB response was further validated with an additional prostate cancer syngeneic model harboring Cdk12 loss and Trp53 depletion (Cdk12KO-sgp53), which was previously established (13). We found that beyond the genetic loss of Cdk12, degradation of CDK13 by YJ1206 treatment (Supplemental Figure 7C) demonstrated antitumor effects and strongly improved ICB efficacy (Figure 4B). We next evaluated the combination of YJ1206 and anti–PD-1 across various syngeneic models, including Myc-CaP (prostate cancer), B16-F10 (melanoma), CT26 (colon cancer), and LLC (lung cancer), after confirming that activation of STING by intratumoral administration of cGAMP led to strong suppression of tumor growth and enhanced ICB efficacy (Figure 4C). We observed that degradation of CDK12/13 by YJ1206 (Supplemental Figure 7D) displayed antitumor effects and significantly improved anti–PD-1 efficacy in these models, which are otherwise insensitive to anti–PD-1 monotherapy (Figure 4, D and E, and Supplemental Figure 7E). Importantly, combination treatment with YJ1206 and anti–PD-1 demonstrated a strong synergistic effect in all models tested (Figure 4F). Consistent with the reported favorable safety profile (20), we observed no body weight loss in mice treated with YJ1206 or the combination of YJ1206 and anti–PD-1 (Figure 4G). Administration of high-dose YJ1206 also led to significant tumor suppression in immunodeficient mice (Supplemental Figure 8A), accompanied by a marked increase in tumor cell apoptosis (Supplemental Figure 8B), consistent with the literature (20). Nevertheless, in line with the Cdk12/13 depletion data (Supplemental Figure 7B), YJ1206 treatment exhibited substantially stronger control of tumor growth in immunocompetent mice compared with immunodeficient mice (Supplemental Figure 8A). Together, our data show that targeting CDK12/13 delays tumor growth and enhances response to anti–PD-1 across various preclinical models.

Antitumor activity of CDK12/13 degradation is CD8+ T cell and STING dependent. In tumors from mice treated with YJ1206, protein levels of phosphorylated RPA2 (p-RPA2-S33) and γH2AX, markers associated with TRC (39, 40), were markedly elevated (Supplemental Figure 8, C and D), supporting the conclusion that YJ1206 also induces TRC formation in vivo. We next examined whether tumors from YJ1206-treated animals showed increased levels of STING signaling. RNA sequencing revealed that STING activity signatures, as well as signatures that could be induced by STING activation (8), were significantly enriched in tumors from mice treated with YJ1206 compared with vehicle (Figure 5, A and B). Histological staining further revealed that administration of YJ1206 significantly elevated the levels of p-STING and p-IRF3 in tumor samples (Figure 5, C and D), demonstrating that YJ1206 treatment induced STING activation in vivo. STING activation is known to promote MHC-I expression (9), and as expected, tumor-specific MHC-I expression was significantly enhanced following YJ1206 treatment (Figure 5, C and D).

Antitumor activity of CDK12/13 degradation is STING dependent. (A) Enrichment of the indicated pathways in Myc-CaP subcutaneous (s.c.) tumors from FVB mice treated with YJ1206 compared to vehicle control. Adj., adjusted. (B) All pathways significantly enriched by GSEA, utilizing the MSigDB Hallmark database, in Myc-CaP s.c. tumors from FVB mice treated with YJ1206 compared to vehicle control. Type I and II IFN responses are highlighted in red. Adj., adjusted. (C and D) Representative images (C) or quantification (D) of immunofluorescence (for p-STING and MHC-I) or immunohistochemistry (for p-IRF3) performed on Myc-CaP tumor tissues from mice treated with vehicle, anti–PD-1 (α-PD-1), YJ1206, or the combination (combo) of α-PD-1 and YJ1206 (n = 4 mice per group). Scale bar: 40 μm. (E) Left: Growth curves of s.c. tumors derived from CT26 cells with or without Sting1 KO, in BALB/c mice treated with the combination (combo) of α-PD-1 and YJ1206. Right: Growth curves of s.c. tumors derived from CT26 cells with Sting1 KO, in Sting1-KO BALB/c mice treated with the combination (combo) of α-PD-1 and YJ1206. YJ1206 was administered orally at a dose of 25 mg/kg, 3 times per week, and α-PD-1 was administered intraperitoneally at a dose of 100 μg/mouse every 3 days (n = 5 mice per group). Data are displayed as mean ± SEM. Significance in D was determined by 2-tailed t test and by 2-way ANOVA in E. Bonferroni’s correction was applied for multiple comparisons in D and E.

We also knocked out Sting1 (the gene encoding STING) in tumor cells (Supplemental Figure 4A and Supplemental Figure 8E) and found that loss of Sting1 partially rescued tumor growth under the combinatorial treatment of YJ1206 and anti–PD-1 (Supplemental Figure 8F and Figure 5E. To evaluate the role of host-derived STING, we performed the same combination treatment in Sting1-KO mice. In this context, tumors with Sting1 deletion were resistant to low-dose YJ1206 combined with anti–PD-1 (Figure 5E). These findings collectively highlight that both tumor-intrinsic and host STING are essential for the full therapeutic efficacy of combined YJ1206 and anti–PD-1 treatment. This supports the concept that tumor-derived DNA can be transferred into the cytosol of antigen-presenting cells, thereby activating STING signaling in antigen-presenting cells and promoting CD8+ T cell–mediated antitumor immunity (41–44).

In agreement with this, immune profiling (Supplemental Figure 9, A–E) revealed that CD8+ T cells and dendritic cells were among the most significantly increased populations of cells in tumors with Cdk12/13 depletion compared with control (Figure 6A). Moreover, activated CD8+ T cells were also significantly increased (Figure 6B). Consistently, a significant increase in activated CD8+ T cells was also observed in tumors from YJ1206-treated mice (Figure 6C). Furthermore, while anti–PD-1 monotherapy only mildly elevated the proportion of active CD8+ T cells, the combination of anti–PD-1 and YJ1206 led to the most pronounced increase in intratumoral activated CD8+ T cells (Figure 6C). Importantly, the absolute number and fraction of tumor antigen–specific CD8+ T cells were significantly increased in tumors from mice treated with YJ1206 alone or in combination with anti–PD-1 (Figure 6D and Supplemental Figure 9F). To further assess the functional importance of CD8+ T cells, we treated mice with an anti-CD8 antibody, which significantly reduced the therapeutic efficacy of the YJ1206 and anti–PD-1 combination (Figure 6E). Collectively, these findings underscore the essential role of CD8+ T cells in mediating the antitumor effects of YJ1206 and anti–PD-1 combination therapy.

Antitumor activity of CDK12/13 degradation is CD8+ T cell dependent. (A) Quantification of flow cytometry showing the absolute number of the indicated immune cell populations in subcutaneous (s.c.) tumors derived from Myc-CaP cells with or without Cdk12/13 depletion (sgCdk12/13). DCs, dendritic cells; PMN-MDSCs, polymorphonuclear myeloid-derived suppressor cells; M-MDSCs, monocytic myeloid-derived suppressor cells. (B) Quantification of flow cytometry showing the absolute number of IFN-γ+ or TNF-α+ CD8+ T cells in tumors from A. (C) Representative images (left) or quantification (right) of flow cytometry measuring the proportion of IFN-γ+ and TNF-α+ CD8+ T cells in the indicated tumor models treated with vehicle, α-PD-1, YJ1206, or the combination (combo) of α-PD-1 and YJ1206 (n = 5–9 mice per group). (D) Left: Gating strategy for CD8+ T cells specific to TRP2 in flow cytometry. Right: Quantification of flow cytometry measuring the absolute number of CD8+ T cells specific to TRP2 in the indicated tumor models treated as in C (n = 6–7 mice per group). (E) Growth curves of s.c. tumors derived from Myc-CaP cells in FVB mice treated with the combination (combo) of α-PD-1 and YJ1206, following CD8+ T cell depletion by anti-CD8 antibody treatment (α-CD8a). Mice without CD8+ T cell depletion (IgG treated) were used as control (n = 5–10 mice per group). YJ1206 (100 mg/kg) was given orally 3 times per week, and α-PD-1 (200 μg/mouse) was administered intraperitoneally every 3 days. Data in A and E are displayed as mean ± SEM. Data in B–D are presented as box-and-whisker plots, with the median (center line), 25th–75th percentiles (box), and minimum to maximum values (whiskers). Significance in A–D was determined by 2-tailed t test and by 2-way ANOVA in E. NS, not significant. Bonferroni’s correction was applied for multiple comparisons.