NK Cell–Specific CDK8 Deletion Enhances Antitumor Responses

Cyclin-dependent kinase 8 (CDK8) belongs to the family of transcription regulating CDKs, and its serine/threonine kinase activity is activated by binding to cyclin C (1). CDK8 associates with cyclin C, MED12, and MED13 to form the kinase module of the mediator complex, which acts as a bridge between transcription factors and RNA polymerase II (1–5). In addition, CDK8 contributes to the activation or repression of transcription by phosphorylating multiple transcription factors. CDK8 has been implicated in the regulation of oncogenic pathways such as the p53, Wnt–β-catenin, or TGFβ pathways (4, 6–11). Overexpression or amplification of the CDK8 gene has been detected in colon cancer and leads to β-catenin hyperactivity (7). High expression of CDK8 correlates with poor prognosis of colon, breast, and ovarian cancer (9, 12, 13). These findings classified CDK8 as a bona fide oncogene, and inhibition of CDK8 is considered a promising target in antitumor therapy (14). CDK8 inhibitors are under development and several preclinical studies show encouraging effects (13, 15–18).

CDK8 regulates transcription downstream of cytokines utilizing the JAK–STAT pathway, thereby also interfering with and regulating immune responses (19). Upon binding of a ligand to a cognate receptor, JAK kinases phosphorylate tyrosine residues on STAT proteins, which then translocate to the nucleus to induce ligand–specific gene transcription (20). Once bound to DNA, STATs undergo an additional phosphorylation event on serine residues to acquire full transcriptional activity (21, 22). CDK8 is an upstream kinase of serine phosphorylation for several STATs, including STAT1–S727 (19, 23). We found a function of the CDK8-STAT1 axis in natural killer (NK) cells (23), innate lymphocytes that provide the first line of defense against transformed and infected cells. We showed that NK cells harbor a constitutive, CDK8-mediated phosphorylation of STAT1–S727, which dampens NK-cell functions. Mice lacking the S727 phosphorylation site in STAT1 (Stat1^S727A) display enhanced NK cell–mediated cytotoxicity and tumor surveillance. The finding that knockdown of CDK8 in an NK-cell line enhanced cytotoxicity and phenocopied the hyperactivity of Stat1^S727A NK cells in vitro (23) prompted us to investigate the role of CDK8 in NK cells in vivo. Here, we used two mouse models to do so; we deleted Cdk8 selectively in NKp46⁺ cells (Cdk8^fl/fl Ncr1Cre) or deleted it inducibly in all interferon-responsive cells present in multiple organs (Cdk8^fl/flMx1Cre; ref. 24). Evidence obtained in these mouse models led us to conclude that NK cells lacking CDK8 develop and mature normally but exhibit hyperactivity in vitro and enhanced tumor surveillance in vivo. The function of CDK8 appears to be NK cell–intrinsic, as the effects of CDK8 deletion were comparable between the two mouse models. Thus, reducing CDK8 expression or activity in cancer patients may both impair tumor cell viability and also increase NK cell–mediated tumor immune surveillance.

Conditional C57BL/6N-Cdk8^fl/fl (Cdk8^{tm1c(EUCOMM)Hmgu}) mice were provided by Yann Herault (IGMBC). The Cdk8^tm1c allele of the mutant was generated from mice with the Cdk8^tm1a knockout first allele (described by International Mouse Phenotyping Consortium https://www.mousephenotype.org) by excision of the lacZ-neo cassette via Flp-recombination. The conditional potential of Cdk8^fl/fl mice was activated by Cre-recombination and excision of the loxP-flanked exon 5 of Cdk8 resulting in a frameshift and the translation of a truncated CDK8 protein, which is subjected to nonsense-mediated decay. Tissue-specific recombination was induced by cross breeding of Cdk8^fl/fl with B6N-Tg(Mx1Cre);(24) and B6N-Tg(Ncr1Cre);(25) mice. All animals were on C57BL/6N background, age and gender matched (6–12 weeks) and maintained at the University of Veterinary Medicine Vienna under specific pathogen-free conditions according to Federation for Laboratory Animal Science Associations (FELASA) guidlines (2014). The animal experiments were approved by the Ethics and Animal Welfare Committee of the University of Veterinary Medicine Vienna and granted by the national authority (Austrian Federal Ministry of Science and Research) according to Section 8ff of Law for Animal Experiments under license BMWF-68.205/0218-II/3b/2012 and were conducted according to the guidelines of FELASA and ARRIVE. To inducibly delete Cdk8 in vivo, Cdk8^fl/fl and Cdk8^fl/flMx1Cre mice were injected intraperitoneally with 200 μg of poly(I:C) (InvivoGen) on days 0, 3, 6, and 9. On day 12, the deletion of Cdk8 was confirmed by PCR of DNA isolated from peripheral blood lymphocytes. Poly(I:C)-treated mice were analyzed or used for tumor challenge 7 days after the last treatment (day 16).

The mouse lymphoma cell lines YAC-1 (26) and RMA-S (kindly provided by Prof. A. Cerwenka; ref. 27) and the v-abl⁺ leukemic cell line (B6-4; generated in the laboratory of Prof. Veronika Sexl) were cultured in RPMI1640 (Sigma) complete medium and mouse melanoma cell line B16F10 (kindly provided by Thomas Felzmann; ref. 28) was cultured in DMEM (Sigma) complete medium, both containing 10% FCS (Bio & Sell), 100 U/mL penicillin, 100 μg/mL streptomycin (Sigma) and 50 μmol/L 2-mercaptoethanol (Sigma). All cell lines were passaged up to 10 times, authenticated by flow cytometry (last authentication August 2017), and tested for mycoplasma contamination by the Mycoplasma Detection Kit-Quick Test (Biomake, last test August 2017).

Cdk8^fl/fl, Cdk8^fl/flMx1Cre and Cdk8^fl/flNcr1Cre mice were challenged with 5 × 10⁴ B16F10 cells by intravenous (i.v.) injection into the tail vein. After 23 days, the mice were sacrificed and the pulmonary tumor nodules were counted by three independent researchers in a blinded manner. In the Abelson Murine Leukemia virus (A-MuLV) model newborn Cdk8^fl/fl and Cdk8^fl/flNcr1Cre mice were injected subcutaneously (s.c.) into the neck fold with 100 μL of replication-incompetent ecotropic retrovirus encoding for v-abl as described previously (29). Mice were checked daily for disease onset. At the first sign of the disease, the mice were sacrificed, body and spleen weight was determined, and the white blood cell count was analyzed using the Scil Vet ABC (Scil Animal Care). Additionally, spleen, bone marrow, and blood were analyzed for the infiltration of B cells (CD19⁺B220⁺), T cells (CD3⁺), and NK cells (CD3⁻NKp46⁺) by flow cytometry. In the v-abl model, Cdk8^fl/fl and Cdk8^fl/flNcr1Cre mice were injected s.c. with 10⁶v-abl⁺ cells into both flanks and the tumor onset was controlled every other day. Ten days after injection, the mice were sacrificed and the tumor weight was determined. For flow cytometric analysis of tumor infiltrating NK cells, tumors were cut into 2 to 5 mm² pieces and digested for 40 minutes at 37°C in Collagenase D (1 mg/mL; Roche) and DNAse I (20 μg/mL; Roche) before filtration to obtain singe cell suspensions.

NK cells were isolated from spleen single-cell suspensions using DX5-labeled MACS beads according to the manufacturer's instructions (Miltenyi Biotec). NK cells were expanded in RPMI1640 complete medium supplemented with 5,000 U/mL rhIL2 (Proleukin, Novartis) for 7 days. The purity was analyzed by flow cytometry and was 85% to 95% CD3⁻NK1.1⁺ of living cells. For Western blot analysis of lytic molecules, expanded NK cells were stimulated with 5 ng/mL rmIL12 (R&D Systems) or 50 ng/mL rmIL15 (PeproTech) for 2 hours. For the growth curve analysis, cells were counted daily using a cell counting chamber (Neubauer). For antibody- or cytokine-induced IFNγ, granzyme B (GZMB), and perforin production 2.5 × 10⁶ to 5 × 10⁶ freshly isolated splenocytes were seeded in RPMI1640 complete medium on anti-NK1.1 (PK136; 10 μg/mL) precoated tubes or in RPMI1640 complete medium supplemented with 5,000 U/mL IL2 and 5 ng/mL IL12 or 5,000 U/mL IL2, 5 ng/mL IL12 and 50 ng/mL IL15 for 4 hours. BD GolgiStop (BD Bioscience) was added 1 hour after the start of stimulation at 37°C. After an additional 3 hours of incubation, the cells were stained for CD3, DX5, NKp46, IFNγ, GZMB, and perforin and analyzed by flow cytometry. For the functional analysis of tumor-infiltrating NK cells, tumor single-cell suspensions were incubated in RPMI complete medium or RPMI complete medium supplemented with 5,000 U/mL IL2, 5 ng/mL IL12, and 50 ng/mL IL15 for 4 hours in the presence of Brefeldin A (eBioscience). The IFNγ production of CD3⁻NKp46⁺ cells was analyzed by flow cytometry.

Single-cell suspensions of splenocytes or bone marrow were prepared. For blood analysis, the erythrocytes were lysed using BD FACS Lysing Solution according to manufacturer's protocol (BD Bioscience). For the detection of intracellular proteins cells were fixed and permeabilized with BD Cytofix/Cytoperm Fixation/Permeabilization Solution Kit (BD Bioscience) according to the manufacturer's instructions. The antibodies (clones) targeting following proteins were purchased from eBioscience: CD3 (17A2), CD3ϵ (145-2C11), CD11b (M1/70), CD16/CD32 (93), CD27 (LG.7F9), CD49b (DX5), CD122 (5H4), CD226 (10E5), Gr-1 (RB6-8C5), GZMB (NGZB), IFNγ (XMG1.2), KLRG1 (2F1), NKG2D (CX5), NKp46 (29A1.4), NK1.1 (PK136), perforin (eBioOMAK), and Ter119 (TER-119). Antibodies (clones) targeting following proteins were purchased from BD Bioscience: CD19 (1D3) and B220 (RA3-6B2). Flow cytometry experiments were performed on a BD FACSCanto II (BD Bioscience) and analyzed using BD FACSDiva V8.0 or FlowJo V10 software. CD3⁻NK1.1⁺ cells were sorted from splenocyte single-cell suspensions using BD FACSAria III.

Cell lysis, SDS-PAGE, and Western blots were performed as described previously (23). The detection of chemiluminescence was performed using Clarity Western ECL substrate (BioRad) and the ChemiDoc^T XRS+ Molecular Imager (BioRad) and analyzed and quantified by Image Lab software (BioRad). The following Cell Signaling Technology antibodies were used: anti-CDK8 (#4106), anti–pSTAT1-S727 (#9177), anti-STAT1 (#9172), anti-perforin (#3693), and anti-GZMB (#4275). Anti-CDK19 (#HPA007053) was purchased from Sigma-Aldrich. Anti–β-actin (Santa Cruz #47778) was used as loading control.

For in vitro cytotoxicity assays, DX5-MACS–sorted NK cells were expanded for 7 days in IL2 as described above and mixed at indicated effector:target ratios with carboxyfluorescein diacetate succinimidyl ester (CFSE, Molecular Probes, CellTrace CFSE Cell Proliferation Kit) labeled target cells. For the ex vivo cytotoxicity, assay freshly isolated splenocytes were mixed with CFSE-labeled target cells at ratios of 100:1, 50:1, and 10:1. After 3 to 4 hours of incubation at 37°C, the specific target cell lysis was assessed by flow cytometry as previously described (30).

The organs were fixed with Roti-Histofix (Roth) and embedded in paraffin. Sections (3 μm) were stained with hematoxylin and eosin (H&E; Microm HMS 740 Robot Stainer; Thermo Scientific) and scanned with a Zeiss AxioImager Z1.

Unpaired t tests and Kaplan–Meier plot analysis by the log-rank test were performed using GraphPad Prism version 5.00 (GraphPad Software). Statistical significance is indicated for each experiment (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

We have previously shown that CDK8 phosphorylates STAT1–S727 in NK cells. NK cells lacking this phosphorylation site display increased cytotoxicity and tumor surveillance (23). To thoroughly analyze the function of CDK8 in NK cells we generated mice lacking Cdk8 specifically in the NK-cell compartment (Cdk8^fl/flNcr1Cre) and mice with a poly(I:C)-inducible deletion of Cdk8 in interferon-responsive cells (Cdk8^fl/flMx1Cre). The expression of CDK8 in freshly isolated Cdk8^fl/flNcr1Cre NK cells and in vivo poly(I:C) pretreated Cdk8^fl/flMx1Cre splenocytes was reduced (Fig. 1A). In both mouse models, deletion of CDK8 had no impact on the frequency and total numbers of NK cells in the bone marrow (Fig. 1B; Supplementary Fig. S1A) and spleen (Fig. 1C; Supplementary Fig. S1B). NK cells develop in the bone marrow and further mature in the periphery. CDK8 did not affect the distribution of individual developmental stages of NK cells in the bone marrow (Fig. 1D). Similarly, maturation of NK cells in the spleen was unchanged, as the percentage of NK cells in various maturation stages as defined by expression of CD27 and CD11b was comparable between Cdk8^fl/flNcr1Cre, Cdk8^fl/flMx1Cre and their Cdk8^fl/fl littermate controls (Fig. 1E; Supplementary Fig. S1C).

Activity of NK cells is controlled by a balance between activating and inhibitory receptors. We thus investigated the expression of the maturation-associated inhibitory receptor KLRG1 (31) as well as DNAM-1 and NKG2D, two activating receptors involved in antitumor responses (32), on the surface of Cdk8^fl/flNcr1Cre and control NK cells. We found no differences in the frequency or expression of these receptors (Fig. 1F; Supplementary Table S1). Poly(I:C) treatment increased the frequency of KLRG1⁺ cells and the expression of DNAM-1 but this effect did not rely on CDK8 expression (Fig. 1F; Supplementary Table S1). We concluded that neither NK cell–intrinsic nor –extrinsic CDK8 influences NK-cell development, maturation, and expression of NK-cell receptors.

To investigate the impact of CDK8 on NK-cell function in vitro, we expanded NK cells in the presence of IL2 for 7 days. The deletion of CDK8 was confirmed (Fig. 2A). CDK8 deletion in expanded NK cells was more efficient than in freshly isolated and unstimulated NK cells, as NK cells in vitro upregulate CDK8 (Fig. 2A; Supplementary Fig. S2A). Although CDK8 has been described as an upstream kinase of STAT1–S727 (19, 23), we did not observe an impact of CDK8 deficiency on STAT1–S727 phosphorylation in IL2-expanded NK cells (Fig. 2A). We reasoned that the CDK8 paralog CDK19 may compensate for CDK8 and thus for STAT1 phosphorylation. Indeed, we found enhanced expression of CDK19 in IL2-cultured Cdk8^fl/flNcr1Cre NK cells (Fig. 2B). Loss of CDK8 did not alter the growth of NK cells when stimulated with IL2 (Fig. 2C). Similarly, IFNγ production after short-term stimulation of ex vivo NK cells with IL12 or IL12 + IL15 was unaffected by the decrease in CDK8 expression. Stimulation with anti-NK1.1 resulted in a 10% decrease of IFNγ production by Cdk8^fl/flNcr1Cre NK cells (Fig. 2D). This effect was not caused by differential expression of NK1.1 (Supplementary Fig. S2B).

The most pronounced changes upon CDK8 deletion were observed when we studied components of the lytic machinery. Expression of perforin was 2-fold upregulated in IL2 expanded CDK8-deficient NK cells in response to IL12 or IL15 (Fig. 2E; Supplementary Fig. S2C). This effect extended to ex vivo IL2 + IL12 + IL15 activated NK cells, confirming that our observations are not restricted to IL2 stimulation in vitro (Fig. 2F). The expression of GZMB was also 10% enhanced upon CDK8 deletion, but this effect was restricted to IL2-expanded NK cells (Fig. 2E and F; Supplementary Fig. S2C). Upon IL2 expansion, the increase in expression of effector molecules in CDK8-deficient NK cells translated into enhanced cytotoxicity against YAC-1 target cells in vitro (Fig. 2G). The enhanced killing capacity was not detected against B16F10 melanoma and v-abl⁺ leukemic target cells (Fig. 2G). This may suggest that the increased abundance of effector molecules observed in CDK8-deficient NK cells is not sufficient to increase their cytotoxic capacity against all target cell lines. To test whether Cdk8^fl/flNcr1Cre NK cells show increased activity without stimulation, we performed an ex vivo cytotoxicity assay using whole splenocytes from naïve mice. Irrespective of CDK8 expression, we detected no killing of RMA-S target cells after 4 hours (Supplementary Fig. S2D). In conclusion, CDK8 restrains the expression of the lytic machinery in cytokine-activated NK cells, which leads to an enhanced cytotoxic capacity toward YAC-1 but not B16F10 or v-abl⁺ target cells.

To evaluate the impact and relevance of our in vitro findings, we used an established in vivo model for NK cell–mediated tumor surveillance (25) and injected the poly(I:C) pretreated Cdk8^fl/fl and Cdk8^fl/flMx1Cre mice i.v. with B16F10 melanoma cells. Tumor burden in the lungs was analyzed 23 days thereafter. The number of pulmonary tumor nodules in mice lacking CDK8 in interferon-responsive cells was diminished (Fig. 3A). To confirm that the observed effect on tumor surveillance is mediated by NK cells, we injected the B16F10 cells i.v. into Cdk8^fl/flNcr1Cre mice. Mice with diminished CDK8 in NKp46⁺ cells were better able to suppress formation of B16F10 lung tumor nodules than were littermate controls (Fig. 3B). This led us to conclude that the radical decrease in CDK8 expression in the NK-cell compartment results in enhanced in vivo tumor surveillance despite the absence of observable changes in vitro. In vitro changes were probably masked by experimental conditions including prolonged cytokine stimulation. Similar effects were seen when we injected v-abl⁺ transformed cells s.c. into Cdk8^fl/fl and Cdk8^fl/flNcr1Cre mice. Again, depletion of CDK8 in NK cells restrained tumor growth, as Cdk8^fl/flNcr1Cre mice showed smaller tumors than their littermate controls (Fig. 4A and B). To investigate the mechanism underlying the improved tumor surveillance, we analyzed NK-cell infiltration into the solid tumors by flow cytometry. Indeed, Cdk8^fl/flNcr1Cre NK cells infiltrated the tumors more efficiently (Fig. 4C and D). However, we could not detect differences in IFNγ production by tumor infiltrating NK cells between the Cdk8^fl/fl and Cdk8^fl/flNcr1Cre mice (Fig. 4E and F). We thus concluded that enhanced NK-cell functionality as well as increased frequency of NK cells in the tumor environment both contribute to improved control of tumor growth in vivo.

Although tumor transplantation models are well defined, fast, and versatile, they fail to mimic the slow growth of a tumor in vivo. To study the influence of CDK8 in such a setting, we injected newborn Cdk8^fl/fl and Cdk8^fl/flNcr1Cre mice with the Abelson Murine Leukemia virus (A-MuLV). The virus induces a slowly progressing pro-B cell leukemia that is controlled in an NK cell–dependent manner (33). As depicted in Fig. 5A, mice expressing little CDK8 in NK cells showed prolonged survival. Nonetheless, at the time of sacrifice, the symptoms of leukemia were comparable in both groups, as assessed by spleen weight, white blood cell count (Fig. 5B), and infiltration of malignant B cells in the bone marrow (Fig. 5C and D). In this model, the enhanced tumor surveillance was not paralleled by an increased number of immune cells; the frequency of T and NK cells in bone marrow, spleen, and blood of the leukemic mice was comparable (Fig. 5E). This led us to conclude that in the slowly evolving leukemia model, the increased capability of Cdk8^fl/flNcr1Cre NK cells to kill emerging tumor cells resulted in prolonged survival of the affected mice (Fig. 5A). Our results suggest that targeting CDK8 could improve NK-cell function in the context of tumor immunotherapy.

Immunotherapy can improve life expectancy and survival of patients suffering from cancer. A greater variety of anticancer treatment strategies, including those that target the immune system, is needed (34). We here show that deletion of CDK8 in NK cells enhanced antitumor responses against B16F10 melanoma cells and v-abl–transformed lymphoma cells. In addition, CDK8 deletion prolonged disease latency in a slowly progressing model system for leukemia. These results define CDK8 as worth targeting in our battle against cancer.

We have previously shown that mice harboring a point mutation in a STAT1 serine phosphorylation site (Stat1^S727A) show increased protection against tumor spread by enhanced NK cell–mediated cytotoxicity (23). We and others have identified CDK8 as an upstream kinase for STAT1–S727 phosphorylation (19, 23). CDK8 knockdown in an immortalized NK cell line also increases cytotoxicity in vitro (23). Our findings described in this study largely recapitulated and extended our previous observations with one exception: we were intrigued to see that the STAT1–S727 phosphorylation is preserved upon CDK8 deletion in IL2-expanded NK cells. In the long-term absence of CDK8 in NK cells, as in a knockout mouse model, CDK19, which is a paralog kinase of CDK8, becomes upregulated and might phosphorylate STAT1–S727. It is unknown if and how CDK19 influences NK-cell biology. The compensatory upregulation seems to be cell type specific as no such effects have been observed in human colon cancer cell lines upon CDK8 knockout (35). Even though CDK19 may compensate for the lack of CDK8, we do not know how CDK8 deletion in the presence of an intact STAT1–S727 phosphorylation enhances cytotoxicity. An intact STAT1 transactivation domain is critical for efficient recruitment of CDK8 (19). Therefore, we speculate that the enhanced cytotoxicity of Stat1^S727A NK cells recapitulates the reduced recruitment of CDK8 to specific STAT1-dependent genes. Alternatively, the effects of CDK8 deletion may be unrelated to STAT1-dependent transcription and the enhanced NK-cell surveillance may be caused by other mechanisms. CDK8, which interacts with a variety of transcription factors (1), also phosphorylates SMAD proteins downstream of the cytokine TGFβ (8), a suppressor of NK-cell function (36). CDK8 may regulate NK-cell activity in a SMAD-dependent manner and the deletion of CDK8 could impair TGFβ signaling. Moreover, CDK8 deficiency drives epigenetic changes in chromatin: loss of CDK8 in intestinal cells causes a reduction in histone trimethylation on H3K27 due to the repression of EZH2 activity (37). As EZH2-deficient NK cells show enhanced cytotoxicity against tumor cells (38), we speculate that CDK8-deficient NK cells also harbor changes in H3K27 trimethylation that could drive increased NK-cell activity. Besides acting on specific transcription factors, CDK8 regulates transcription in a global manner by taking part in the mediator complex together with MED12, MED13, and cyclin C (1). It remains unclear whether the effects of CDK8 deletion in NK cells involve the mediator complex. No information is available on other mediator subunits for NK-cell biology. MED12 mutations have been described in post-transplant T lymphoproliferative disorders (39) and cyclin C has been identified as tumor suppressor in T-cell acute lymphoblastic leukemia (40), suggesting a complex role of the mediator complex in the regulation of T lymphocytes.

CDK8 is deregulated in several types of cancer, including breast and colon cancer, and is thus a promising therapeutic target (14). In summary, we have shown CDK8 does not influence NK cell development and proliferation, but instead suppresses NK-cell antitumor activity in vivo. Although the molecular mechanisms behind these effects remain unclear, our results implicate that targeting CDK8 in cancer patients holds high therapeutic potential. Blocking or degrading CDK8 will not only act on the tumor cells but also enhance NK-cell lytic responses.

No potential conflicts of interest were disclosed.

Conception and design: A. Witalisz-Siepracka, D. Gotthardt, V. Sexl

Development of methodology: D. Gotthardt, V. Sexl

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Witalisz-Siepracka, D. Gotthardt, M. Prchal-Murphy, Z. Didara, I. Menzl, L. Edlinger, E.M. Putz

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Witalisz-Siepracka, D. Gotthardt, Z. Didara, L. Edlinger, V. Sexl

Writing, review, and/or revision of the manuscript: A. Witalisz-Siepracka, D. Gotthardt, I. Menzl, E.M. Putz, V. Sexl

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Didara, D. Prinz

Study supervision: V. Sexl

The work was supported by the Austrian Science Fund FWF (grant P28571 to V. Sexl, the Schroedinger Fellowship J-3635 to E.M. Putz, and the PhD program “Inflammation and Immunity” FWF W1212 to V. Sexl).

We would like to thank the European Conditional Mouse Mutagenesis Consortium for providing the Cdk8^{tm1a(EUCOMM)Hmgu} mouse strain and Mohammed Selloum (Yann Herault Group) for providing the Cdk8^{tm1c(EUCOMM)Hmgu} mouse strain. We are grateful to Thomas Rülicke and Thomas Kolbe for their help in acquiring the mice and to all members of the mouse facility. We thank all members of the SFB F61, in particular Thomas Decker and Mathias Müller, for discussions and scientific input. We are indebted to Sabine Fajmann and Petra Kudweis for excellent technical assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.