Introduction

Multiple protein complexes regulate transcription, in response to developmental cues, changes in environment, damage and other factors, allowing for precise control of gene expression. Among such complexes the multiprotein Mediator complex is especially important for controlling transcription, regulated by a multitude of signaling pathways through their transcription factors (TFs). The Mediator complex bridges the regulatory enhancer regions with promoters thereby recruiting RNA polymerase II (RNA Pol II) and the preinitiation complex. The Mediator possesses its own regulatory subpart – the cyclin dependent kinase module (CKM), containing either the CDK8 or the CDK19 protein kinases, in complex with their binding partner cyclin C (CcnC), as well as proteins MED12 and MED13 (Luyties and Taatjes, 2022). While CKM is a nuclear complex that regulates transcription, both CcnC (Ježek et al., 2019) and MED12 (Zhang et al., 2020) were shown to have CKM-independent cytoplasmic activities. CDK8/19 kinase activity positively regulates transcription induced by different signals (Chen et al., 2023, 2017), at least in part by mediating the RNA Pol II release from pausing (Steinparzer et al., 2019). Besides, CDK8 and CDK19 directly phosphorylate a number of transcription factors such as STATs (Martinez-Fabregas et al., 2020), SMADs (Alarcón et al., 2009) and others, thereby modulating transcription of cognate genes. CDK8/19 kinase activity also negatively regulates transcription, at least, in part through downregulating the protein levels of all the components of the Mediator and CKM (Chen et al., 2023).

Despite important CDK8/19 functions discovered in primary cell culture and cell lines (Luyties and Taatjes, 2022), the insights gained from in vivo models are limited. Mice with constitutively knocked out Cdk19 (CDK19KO) were generated as a part of the COMP project and are basically asymptomatic, fertile, and have normal lifespan. It has been shown recently that Cdk19^-/- Lin^- hematopoietic stem cells divide slower (Zhang et al., 2022), however the blood cell count was unaffected. Heterozygous mice with constitutive Cdk8 knockout were asymptomatic but, when crossed, no homozygous pups were born (Postlmayr et al., 2020; Westerling et al., 2007). Moreover, conditional Cdk8 KO is almost asymptomatic in adult mice. Minor differences were observed in colon epithelial differentiation (Dannappel et al., 2022; Prieto et al., 2023) and tumorigenesis (McCleland et al., 2015) as well as in osteoclastogenesis (Yamada et al., 2022) after tissue specific Cdk8 knockout. Although two CDK8/19 inhibitors were reported to have systemic toxicity (Clarke et al., 2016), this toxicity was subsequently found to be due to off-target effects of these inhibitors (Chen et al., 2019), and several CDK8/19 inhibitors have reached clinical trials (clinicaltrials.gov NCT03065010, NCT04021368, NCT05052255, NCT05300438). Several studies reported the existence of kinase-independent phenotypic activities for both CDK8 (Kapoor et al., 2010; Menzl et al., 2019b) and CDK19 (Audetat et al., 2017; Steinparzer et al., 2019), but the only known biochemical activity of both Cdk8 and Cdk19 that is kinase-independent is the stabilization of their binding partner CcnC (Barette et al., 2001; Chen et al., 2023). Targeted degradation or knockout of both CDK8 and CDK19 dramatically reduces the cellular levels of CcnC, whereas the knockout of either Cdk8 or Cdk19 alone has little effect on CcnC levels (Chen et al., 2023). The effects of double knockout of both Cdk8 and Cdk19 on CcnC should therefore be considered in interpreting the effect of the double knockout.

We have, for the first time, generated mice with a conditional knockout of the Cdk8 gene on the constitutive Cdk19 KO background (CDK8/19 double knockout, DKO). We have found that DKO males were completely infertile, had an undersized and dedifferentiated reproductive system. In DKO males, spermatogenesis was blocked after pachytene of meiosis I. DKO showed down-regulation of key steroid pathway genes, such as Cyp17a1, Star and Lcn2, Sc5d, Fads2 and reduced production of testosterone. Expression of key meiotic genes was also deregulated in meiosis I spermatocytes, and the progression of meiosis was halted. These results indicated a key role of the CDK8/19-CcnC complex in the maintenance of male reproductive system.

Results

Cdk8/19 DKO mice are viable and lack Cyclin C

Previously we have crossed Cdk8^fl/fl mice with Rosa-Cre/ER^T2 with tamoxifen-inducible Cre activity and demonstrated effective KO in all tissues except for ovaries and uterus (Ilchuk et al., 2022) (Fig. 1A). However, the CDK8 inactivation in the male reproductive system (testes) was efficient at the protein and DNA levels (Fig. 1B, Supplementary Fig. S1A). To investigate the effects of the double knockout of Cdk8 and Cdk19, we crossed Cdk8^fl/fl/Rosa-Cre/ER^T2 mice with Cdk19^-/- mice (Zhang et al., 2022) [MGI:5607862, Cdk19^em1(IMPC)J], which are asymptomatic and are maintained as homozygotes. We injected tamoxifen into 2 months old Cdk8^fl/fl/Cdk19^-/-/Rosa-Cre/ER^T2 male mice. The DKO mice appeared to be viable. Age matched tamoxifen-treated wild-type mice served as a control.

Figure 1.

We have also analyzed the levels of the CDK8/19 binding partner CcnC after single and double KOs in mouse embryonic fibroblasts and in mouse testes after DKO. In agreement with previous observations in human cell lines (Chen et al., 2023), CcnC was undetectable in DKO embryonic fibroblasts as well as in DKO testes, but not in the single CDK19KO (Supplementary Fig. S1B, Fig. 1C). Hence, the stabilization of CcnC by CDK8 and CDK19 occurs not only in human but also in mouse cells and tissues.

Spermatogenesis is blocked in DKO males

To investigate DKO effects we performed necropsy with subsequent H&E staining 8 weeks post tamoxifen treatment (Fig. 1D). The most striking results were observed in the urogenital system. The prostate, testes and the epididymis were significantly smaller compared to single KO and tamoxifen-treated wild-type mice (Fig. 1E), the diameter of testicular tubules was diminished (Fig. 1F). The germinal epithelium of the tubules was presented by Sertoli cells with typical large light nuclei and some spermatogonia or spermatocytes. Cells at the postmeiotic stages of spermatogenesis were completely absent. Among the spermatocytes there were only early prophase cells. In contrast, we detected cells with colchicine-like apoptotic bodies (Fig. 1F,G, yellow arrows). Thus, in these mice the spermatogenesis progenitor cells were present but their differentiation and/or meiosis was blocked. In Sertoli cells the vacuoles are the sign of the loss of contact between these cells and spermatogenic elements (Fig. 1F,G). Leydig cells were significantly smaller and were almost depleted of secretory vacuoles suggesting the lack of hormonal activity. The epididymis contained empty tubules; the epithelium appeared undifferentiated, that is, neither typical borders between epithelial cells nor clear Golgi zone were visible. The prostate and the epididymis were also atrophic (Fig. 1F). These changes were detectable from 2 weeks after tamoxifen treatment and persistent for at least 6 months (Fig. 1D). However, they were not observed in single KOs (Supplementary Fig. S1C).

DKO and CDK8KO mice are infertile

To prove the DKO infertility we assessed their sexual behavior and number of pups they fathered. As tamoxifen has an impact on the male reproductive system (Willems et al., 2011), the experiments were performed at least 6 weeks post injection and we used tamoxifen-treated C57BL/6 mice as a control. Four groups were enrolled in the experiment: tamoxifen treated C57BL/6 (WT+Tam), tamoxifen treated Cdk8^fl/flRosa-Cre/ER^T2 (CDK8KO), Cdk8^fl/flCdk19^-/- Rosa-Cre/ER^T2 (CDK19KO) and tamoxifen treated Cdk8^fl/flCdk19^-/-Rosa-Cre/ER^T2 (DKO). Three males in each group were kept separately with two outbred CD1 females each for 3 months. Copulative plugs and the number of pups were checked 5 days a week (Fig. 1H). Tamoxifen-treated wild-type mice demonstrated normal fertility: 10 plugs were detected and 86 pups were born. The CDK19KO group showed slightly reduced fertility: 32 pups along with an increased number of the plugs (16) probably caused by the lower rate of pregnancy onset. Surprisingly, only two plugs and no pups were observed in the CDK8KO group despite normal appearance and behavior of males, and no microscopic abnormalities in the urogenital system (Supplementary Fig S1). Neither plugs nor pups were detectable in the DKO cohort (Fig. 1H). These experiments showed that both CDK8KO and DKO are infertile, and the cause of CDK8KO infertility is likely to be the lack of sexual behavior.

Spermatogenic cells of DKO mice are unable to advance through meiosis I prophase

To obtain a quantitative spermatogenesis pattern we performed cell cycle analysis by flow cytometry with propidium iodide staining. We examined wild-type, single KOs and DKO animals sacrificed two months after tamoxifen treatment. Cell cycle analysis revealed striking differences in DKO mice, with disappearance of elongated spermatids, almost all round spermatids and massive cell death (Fig. 2A,B). All other groups had the same normal cell distribution (Fig. 2B).

Figure 2.

Noticeably, the 4n population and cells in the S-phase were not impaired by DKO indicating that DKO spermatogonia successfully entered meiosis but could not produce haploid spermatids. This ongoing “meiotic catastrophe” led to almost full depopulation of the testes in DKO mice (Fig. 2С). Due to fast transition from meiosis I to meiosis II it is difficult to detect secondary spermatocytes by flow cytometry. To specify the stage of CDK8/19-dependent spermatogenesis failure we performed phospho-γH2A.X and SYCP3 immunofluorescent staining of seminiferous tubules. Phospho-γH2A.X is a DNA damage marker which marks double-strand breaks in leptotene, zygotene and in sex chromosomes during the pachytene. SYCP3 is a component of the synaptonemal complex, expressed during meiosis I prophase, which forms distinct patterns in the late zygotene/pachytene. In the wild-type tubules all the meiosis stages were visible (Fig. 2D). At the same time almost all meiotic cells in DKO were blocked in the pachytene (Fig. 2D), indicating that DKO cells enter meiosis but cannot traverse through meiosis I prophase and subsequently undergo cell death.

Single cell RNA sequencing

To investigate molecular mechanisms of CDK8/19 mediated alterations in meiosis we performed single cell RNA sequencing (scRNAseq) of the testes. We analyzed cell suspensions from testes of two tamoxifen-treated wild-type С57/BL/6 and two DKO animals.

The analysis of cell type composition confirmed the reduced number of post pachytene spermatocytes and almost complete absence of spermatids while meiotic entry, leptotene and zygotene cell numbers remained unaffected (Fig. 3A,B). At the same time the percentage of undifferentiated spermatogonial stem cells in DKO animals was unchanged, nor was the proportion of Leydig cells altered. Sertoli cells became the most abundant cell type in DKO mice with reduced testes cellularity, however their absolute number didn’t change significantly.

Figure 3.

Differential gene expression

Differentiation of sperm lineage is a tightly controlled process, where changes in functioning in a certain type of cells can affect the viability and differentiation of other cells. To understand the underlying mechanism of meiotic cell death in DKOs we compared gene expression in each type of cell clusters present both in wild-type and knockout animals.

Leydig cells

Leydig cells are the primary source for testosterone in the testes, a hormone required for spermatogenesis. scRNAseq analysis showed significant transcriptomic changes in Leydig cells. Among 126 genes that were differentially expressed (|log2FC|>0.4; p<0.01) in WT vs DKO Leydig cells, 30 were down-regulated, however among 14 strongly affected genes with |log₂FC|>1 11 were down-regulated. Strikingly, 9 of 14 strongly downregulated genes were associated with lipid (specifically steroid) metabolism (Supplementary table S1) and 3 of these 9 genes were linked to male infertility (Aherrahrou et al., 2020; Song, 2007; Stoffel et al., 2008).

According to the GO analysis (Fig. 3C, Supplementary table S2), the most prominent changes were downregulation of lipid metabolism and the steroid hormone biosynthesis. The most significantly down-regulated gene was Cyp17a1, showing 10-fold decrease (Fig. 3D). The Cyp17a1 gene in male mice is mainly expressed in Leydig cells (Missaghian et al., 2009). The product of this gene catalyzes the key reaction of the steroidogenic pathway. Mice with constitutive Cyp17a1 KO are phenotypically female (Aherrahrou et al., 2020). Downregulated Lcn2 (Fig. 3D) specifically expressed in Leydig cells codes for a protein of the lipocalin family that transports small hydrophobic molecules including steroids and is affected by fertility manipulations (Kang et al., 2017; Yanai et al., 2021). Another downregulated gene, Fads2 (Fig. 3D), is a key enzyme for biosynthesis of highly unsaturated fatty acids. Its KO causes infertility in males by arresting the spermatogonial cycle at the stage of the round spermatids (Stroud et al., 2009). The Sult1e1 (Fig. 3D) gene encodes sulfotransferase which inactivates estrogens and its deficiency leads to hyperplasia of Leydig cells and atrophy of seminiferous tubules (Song, 2007). The Star, Sc5d, Pld3 and Apoc1 (Fig. 3D) genes, also involved in lipid metabolism, were down-regulated more than 2-fold. Interestingly, the Hsd3b6 (Fig. 3D) is a surprisingly upregulated gene associated with steroid metabolism in Leydig cells.

Not only testosterone production was decreased in double knockouts. Expression of Insl3 gene, coding for an important peptide hormone secreted by Leydig cells (Esteban-Lopez and Agoulnik, 2020), was also reduced (Supplementary table S3).

Among up-regulated genes, were several transcription factors governing steroidogenesis in Leydig cells (de Mattos et al., 2022; Ye et al., 2017). One of them, Nr2f2 (Fig. 3D) encodes COUP-TFII which regulates genes immediately involved in steroidogenesis (Cyp17a1, Hsd3b1, Cyp11a1 and Akr1c14) alongside several other Leydig-specific genes (Insl3, Amhr2) (Fig. 3D). Another gene, Cebpb, encodes TF C/EBPβ which activates Star transcription and stimulates expression of Nr4a1 (also known as Nur77) which is itself a key regulator of steroidogenesis in Leydig cells.

The Kit gene (down-regulated), encoding a tyrosine kinase receptor, and Kitl gene (up-regulated), encoding its ligand, were also present among differentially expressed genes. Both of these genes play an important role in spermatogenesis (knockouts and mutants are infertile), Leydig cell maturation and steroidogenesis (Liu et al., 2017; Ye et al., 2017).

We hypothesize that the observed upregulation of several steroidogenesis regulators may be a part of a compensation feedback loop in response to the decreased testosterone level.

Sertoli cells

ScRNAseq (Fig. 3A, B) shows that Sertoli cell fraction is greatly increased in DKO specimens, however this is mostly due to the reduction of germ cell number. At the same time, analysis of cell cycle-specific genes revealed that in DKO Sertoli cells can re-enter the cell cycle (Fig. 4A). In particular, genes involved in G0-S transition and G2-M transition were upregulated.

Figure 4.

According to downregulation of the steroid hormone biosynthesis in Leydig cells we expected to see in Sertoli cells patterns similar to those in mice with Sertoli cell specific KOs of androgen receptor (AR) – SCARKO (Sertoli Cells Androgen Receptor KnockOut) (De Gendt et al., 2014; Larose et al., 2020) and luteinizing hormone (LH) receptor – LURKO (Griffin et al., 2010). However, it must be noted that these mice have constitutive receptor KOs and do not fully develop their reproductive system, whereas our KO model mice develop normally and become fertile before the knockout induction.

Among DEGs we identified a number of upregulated and downregulated genes consistent with their changes in SCARKO (Defb45, Myl6, Tmsb4x, Espn, Ldhb among top DEGs). At the same time, we compared our DEG set with those published earlier (De Gendt et al., 2014; Larose et al., 2020) and observed little overlap between our data and published results as well as between the two published papers themselves (Supplementary Fig. S3).

However, a much better agreement was found between our data and the study by deGendt et.al. (De Gendt et al., 2014) at the level of GO Molecular Function and Cellular Compartment categories.

According to the GO Cellular Compartment (Supplementary table S2), both up- and down-regulated genes are associated with cytoskeleton (Vim, Actb, Actg1) (Show et al., 2003), apical-basal axis (Cdc42) (Heinrich et al., 2021), cellular membrane (Atp1a1, Gja1) (Rajamanickam et al., 2017; Sridharan et al., 2007) and intercellular contacts (Espn, Anxa2, S100a10) (Chojnacka et al., 2017; Willems et al., 2010). These closely related structures play an important role in Sertoli cell functioning: formation of the blood-testicular barrier (BTB) and intercommunication with developing germ cells (Fig. 4B) (Wong and Cheng, 2009). Therefore, our findings on differential gene expression in Sertoli cells are in agreement with data obtained in mice with impaired AR signaling. Among the top enriched biological processes several up-regulated stress response pathways were also found, indicating deterioration of the Sertoli cells in the absence of testosterone (Fig. 4D).

This molecular evidence of Sertoli cell spatial organization disturbance and disruption of cell contacts is in good agreement with the patterns observed during histological examination.

Germ cells

As primary spermatocytes of double knockout mice stop their differentiation in pachytene, we compared gene expression in two groups of germ cells: undifferentiated spermatogonia and meiosis-entry (leptotene-zygotene-pachytene) spermatocytes. 1107 DEGs were identified for undifferentiated cells, 1650 for spermatocytes, and 874 of these genes were common. Almost all the DEGs were up-regulated (99.2%). This overlap may be explained by the fact that undifferentiated and differentiated spermatogonia represent a continuous spectrum with gradually changing transcription patterns. Unique genes for the Early Spermatogonia cluster were attributed by GO as genes related to RNA processing and biosynthesis and ribosome biogenesis, whereas no stem cell specific pathways were found. We conclude that Cdk8/19 knockout and disruption of steroid biosynthesis and testosterone production have no significant impact specifically on early spermatogonia.

Most cells in knockout testes do not progress to stages after pachytene. Accordingly, apoptosis-related pathways, stress response pathways and the autophagy pathway were upregulated (Supplementary table S2, Supplementary Fig. S4). More specifically, a set of meiosis-specific genes was also up-regulated. These genes were related to chromosome pairing, DNA recombination and histone modification, which are known to be involved in meiosis. We hypothesize that this effect might be caused by compensatory mechanisms triggered by inability to proceed through meiosis.

Validation of single cell RNA data

To further prove the steroid hypothesis, we performed CYP17A1 immunostaining of testicular cross sections and western blotting, as well as direct testosterone measurement in the blood via LC-MS and cell culture experiments with ex vivo Leydig cells.

In control testes, CYP17A1 protein was localized outside of seminiferous tubules, in Leydig cells. In DKO animals, CYP17A1 staining was completely absent (Fig. 5A), despite Leydig cells being present in histological sections (Fig. 1G). WB (Fig. 5B) further confirming this finding.

Figure 5.

CYP17A1 catalyzes several key reactions leading to transformation of pregnenolone into testosterone. We measured the concentration of the key androgen hormone - testosterone in the blood of wild-type mice, single KOs and DKO. Only DKO showed significantly lower levels of testosterone, consistent with only this genotype presenting abnormal spermatogenesis (Fig. 5C).

Changes in steroid synthesis in Leydig cells can be caused by several mechanisms – specifically CDK8/19/Cyclin C can regulate CYP17A1 expression in Leydig cells themselves, or CDK8/19/Cyclin C may be involved in regulation of brain hormonal signals, which control testosterone production. To test the latter hypothesis we measured the concentration of luteinizing hormone in blood serum. The results showed no difference in LH level through all the genotypes (Fig. 5D). This means that CDK8/19/Cyclin C acts directly in Leydig cells to regulate steroidogenesis.

Spermatogenesis in DKO mice slightly recovers with time

To address the question of the persistence of spermatogenesis failure we performed histological and flow cytometry analysis of mice 5 months after tamoxifen-treatment. A certain percentage of 1n cells (primarily round spermatids) have reappeared 3-5 months after tamoxifen treatment (Fig. 6A), nevertheless the total number of cells and 1n cells remained low compared to control (Fig. 6B). H&E staining of prostate and epididymis also revealed slight alleviation of the phenotype, but no mature sperm (Fig. 6С). Immunofluorescent staining of seminiferous tubules confirmed this finding demonstrating solitary tubules with postmeiotic cells (Fig. 6D). However, CYP17A1 was not restored in Leydig cells (Fig. 6E).

Figure 6.

Pharmacological inhibition of CDK8/19 does not affect spermatogenesis

Recent findings have established that CDK8/19 may have kinase-independent functions (Steinparzer et al., 2019) including stabilization of Cyclin C by CDK8/19 (Chen et al., 2023). Depletion of CDK8/19 in testes leads to CCNC degradation (Fig. 1C). To examine if the observed phenotype is related to the kinase activity of CDK8/19 or is kinase-independent we treated WT mice with a pharmacological CDK8/19 inhibitor – SNX631-6 (Li et al., 2023). 11 weeks old male C57BL/6 mice were treated with SNX631-6 medicated chow (500 ppm, 40-60 mg/kg/day dosage on average) or control diet for 3 weeks. No abnormal clinical observations were identified in treated mice through the whole treatment period. Testes were weighed and fixed in 10% formalin for H&E staining. There were no significant differences in testes weight between control and treated group and histology analysis did not reveal any noticeable difference (Fig. 7A,B), in contrast to the DKO testes. However, the drug concentration in the testes measured by LCMS/MS was lower than in blood, suggesting the effects of the blood-testis barrier, but still ∼10 times higher than the drug’s IC₅₀ in the cell-based assay - 50 ng/ml. (Fig. 7C).

Figure 7.

The effects of DKO and CDK8/19 kinase inhibition were compared in ex vivo primary Leydig cell culture from WT mice and Cdk8^fl/flCdk19^-/-Cre-Rosa mice. We treated the first with CDK8/19 inhibitor Senexin B and the latter with hydroxytamoxifen to induce Cdk8^-/- in vitro. Then we compared expression of Cyp17a1, Star and Fads genes downregulated according to scRNA sequencing data in treated and untreated cells (Fig. 7D). Star and Fads were downregulated in OH-Tam treated cells, whereas basal Cyp17a1 expression level was too low to make any conclusion. At the same time, CDK8/19 inhibition didn’t affect expression levels of Star and Fads, suggesting that the effects of the DKO were kinase-independent.

Discussion

The Mediator complex is a key part of transcriptional regulation through signaling pathways. The enzymatic component of the Mediator-associated CKM module – CDK8 and its paralog CDK19, have emerged as co-regulators in several signal-activated transcriptional pathways, such as Wnt, TGF-beta, androgen and estrogen signaling, STATs, response to serum growth factors and others (reviewed in (Menzl et al., 2019a)). Despite that, the knockout of Cdk8 is only critical in embryogenesis (Westerling et al., 2007), with no phenotype in adult organisms (McCleland et al., 2015), while Cdk19 is wholly dispensable. The same tendency is observed for the knockout of MED subunits and other CKM proteins: constitutive KOs are often embryonically lethal (except for paralog of MED12 - MED12L), while tissue-specific knockouts in adult animals rarely have a severe phenotype (Ilchuk et al., 2023).

Hypothetically CDK8 and CDK19 can compensate for each other in KO animals and mitigate the phenotype in single KOs. To confirm this we produced the first ubiquitous knockout of Cdk8/19 in 2 months old animals. Previously described animals with conditional knockout of Cdk8 in the intestine on Cdk19-/- background presented with a mild phenotype, moderately affecting the number of specialized secreting cells (Dannappel et al., 2022). Our DKO mice had a phenotype evident at the histological level: they lacked haploid spermatids and mature spermatozoa (Fig. 1F), with primary spermatocytes blocked in pachytene of meiosis I (Fig. 2). Surprisingly, Cdk8-/- males previously described as asymptomatic revealed their infertility due to the lack of sexual behavior, despite the absence of morphological changes in the urogenital system (Fig. 1E).

Mice with Med1^fl/fl/Vasa-Cre tissue specific KO had an opposite phenotype to the one described here: prophase I passage was accelerated in these mice and spermatocytes entered zygotene/pachytene stage prematurely (Huszar et al., 2015). The difference between these effects is quite interesting in the context of the concept that association with CKM attenuates Mediator-dependent transcription either by sterically preventing TF binding (Freitas et al., 2022), or by post-transcriptional regulation of Mediator complex subunits (Chen et al., 2023; Poss et al., 2016). As far as we know, besides the current work, Huszar et al., 2015 is the only paper establishing a connection between any components of the Mediator complex and spermatogenesis.

We demonstrated a disruption of steroid biosynthesis in Leydig cells which could play the key role for the observed phenotype. A number of works (reviewed in (Ilchuk et al., 2023)) affirm a significant role of Mediator complex in lipid metabolism and biosynthesis as one of the main cofactors for main adipogenic TFs such as C/EBPα, PPARγ and SREBP.

We detected a decrease in steroidogenic gene expression by scRNAseq (primarily Cyp17a1), and then confirmed this finding by WB, immunofluorescence, and direct testosterone measurement. This corresponds well with perturbed Leydig cells morphology (reduced size and secretory vacuoles depletion) and explains observed atrophy of prostate and epididymis (Fig. 1F). It is worth mentioning that constitutive Cyp17a1 knockout in mice leads to female phenotype in XY mice (Aherrahrou et al., 2020). Besides Cyp17a1, several other steroid metabolism related genes were downregulated (Lcn, Fads, Star, Sc5d, Sult1e1). Interestingly, the main feature of Fads (Stoffel et al., 2008) and Sult1e1 (Song, 2007) KOs is male infertility, caused by spermatogenesis arrest.

There is much evidence that meiosis in spermatocytes depends on AR signaling (Wang et al., 2022). Several papers established a connection between AR signaling, and the Mediator complex (Russo et al., 2019). MED1 and MED19 directly interact with the AR and regulate its transcriptional activity (Jin et al., 2012; Weber et al., 2021). Additional evidence of the role of the Mediator complex and CDK8/19 in the urogenital system comes from studies of their role in prostate cancer, where CDK19 is elevated together with the AR signaling (Becker et al., 2020).

The DKO has also dramatically impacted Sertoli cells. The SCARKO mice were engineered two decades ago and show morphological abnormalities (i.e, reduced testes, epididymis and prostate, meiotic arrest during meiosis I prophase) similar to DKO (Chang et al., 2004; De Gendt et al., 2004). The scRNAseq technique used to elucidate the molecular mechanism of the meiotic arrest revealed a number of similarities with DKO data for Sertoli cells, especially in terms of pathways (Cao et al., 2021). Thus, genes associated with cytoskeleton, cellular membrane and cell-cell contacts were enriched among the DEGs (Fig. 4B). Nevertheless, the overlap of genes affected in the SCARKO and DKO mice was low (Supplementary Fig. S3). However, an important difference between our model and SCARKO is that Cdk8 knockout is induced after puberty. Another important consideration is that Sertoli cells are not the only cell type impacting spermatogenesis through AR signaling. AR signaling in peritubular myoid cells is also essential for successful spermatogenesis (Welsh et al., 2009). In this regard, probably a more relevant model for comparison with our results is described in (Stanton et al., 2012) where androgen signaling was suppressed in grown-up rats by low dose testosterone, or its combination with an AR antagonist. In this paper the group with the most complete inhibition of AR signaling displayed increased apoptosis in spermatocytes and full blockage of second meiotic division.

Just as in (Stanton et al., 2012), genes differentially expressed in spermatocytes blocked in prophase I can be divided in the following groups: (1) genes with known roles in meiosis, especially DNA synapse, homologous recombination and DNA repair, (2) genes associated with cellular stress and apoptosis, and (3) genes with roles in RNA processing and splicing. Changes in the first group can be explained by the compensatory mechanisms and changes in the second group correlate well with the observed extensive cell death of spermatocytes, changes in the third group remain enigmatic.

Noteworthy, 99.2% of spermatocyte DEGs with |logFc|>0.4 were upregulated genes, which is very similar to the findings of Pelish et al. (Pelish et al., 2015) and agrees with the negative regulation of the Mediator complex by CDK8/19 (Chen et al., 2023).

Based on these results we can conclude that DKO causes perturbance in different testicular cell types: Leydig cells, Sertoli cells and spermatocytes. We hypothesize that all these changes are caused by disruption of testosterone synthesis in Leydig cells. Data for other cell types are consistent with this hypothesis, however there is still a possibility that CDK8/19 have distinct roles in other testicular cell types which are not visible against the background of the testosterone reduction. However, our data are not sufficient to make the conclusion about the role of CDK8/19/Cyclin C directly in meiosis, besides its hormonal regulation. To answer this question, definitive mouse strains with cell-type specific KO in Sertoli cells and meiotic spermatocytes would be needed.

The effects of DKO on spermatogenesis were maximal at 1-2 months after tamoxifen administration. At later time points we detected partial restoration of spermatogenesis, and signs of re-differentiation of the epididymis and activity of secreting cells in the prostate. This restoration did not lead to the production of mature sperm, and the total number of haploid cells increased insignificantly. There are several possible explanations for such an effect. It is unlikely that restoration happened in cells, which did not activate Cre-Lox, as the absence of CDK8 has been confirmed by us through PCR (Supplementary Fig. S1A). Additionally, no restoration of CYP17A1 was found in 5-month-old animals. There is a possibility that transcription of genes previously regulated by CDK8/19/Cyclin C was rewired through other transcriptional regulators. An intriguing possibility is activation of a compensatory AR-independent pathway that was recently shown in a zebrafish model (Zhai et al., 2022). An argument against such hypothesis would be the absence of such compensation in Cyp17a1 KO animals or SCARKO mice, but in the former animals the male reproductive system does not develop at all, and in the latter, non-canonical AR signaling through Src kinase may lead to progression to round spermatid stage (Cooke and Walker, 2021).

The luteinizing hormone produced in the pituitary gland is a major regulator of steroidogenesis in Leydig cells. It was shown that LH regulates expression level of Cyp17a1 and several other steroidogenic genes (Li et al., 2021; Ma et al., 2004). As CDK8KO in our model is ubiquitous, infertility could be caused by decrease of LH level due to improper regulation in the brain. However, we measured LH level in the serum and found it unaffected by single or double KOs (Fig 5D). That means that CDK8/19/Cyclin C are required for steroidogenic genes’ transcription in Leydig cells. And indeed, in our ex vivo experiment with Leydig primary cell culture, even without addition of LH, expression of Star and Fads was downregulated upon in vitro induction of CDK8KO by hydroxytamoxifen on the CDK19KO background.

The results of the present study reveal for the first time the role of CDK8/19/Cyclin C in the function of the male reproductive system. The key question regarding the results described here is whether the phenotypic effects are due to the inhibition of CDK8/19 kinase activity or to the loss of Cyclin C, which is protected by CDK8 and CDK19 from proteasome-mediated degradation (Chen et al., 2023), or other kinase-independent functions of CDK8/19. Here we show that Cyclin C essentially disappears both from the embryo fibroblasts and from the testes of the DKO mice. Given the known CDK8/19-independent functions of Cyclin C (Ježek et al., 2019), and the effect of Cyclin C knockout on lipid accumulation (Song et al., 2022), the likelihood of Cyclin C rather than CDK8/19 mediating the observed effects should be investigated. As we have shown, 30-day treatment of 11 weeks old mice with a potent CDK8/19 kinase inhibitor produced no anatomical changes in the testes. Furthermore, ex vivo treatment of Leydig cells with a CDK8/19 inhibitor did not reproduce the DKO effect on gene expression, suggesting that the effects of DKO were kinase-independent and possibly mediated by Cyclin C. A definitive answer to the role of Cyclin C vs CDK8/19 kinase activity will require a phenotypic comparison of DKO mice with mice expressing kinase-inactive mutants of CDK8/19 or inducible Cyclin C knockout.

Materials and Methods

Animals and conditional DKO

Cdk8^fl/fl and R26-Cre-ER^T2 (Jax:008463; B6.129-Gt(ROSA)26Sor^{tm1(cre/ERT2)Tyj/J}) provenance and genotyping procedures are described (Ilchuk et al., 2022). Cdk19^-/- (Cdk19^em1(IMPC)J) were purchased from Jackson Laboratory. Mice were genotyped by real-time PCR using oligonucleotides listed in Supplementary table S4.

Animals were maintained under controlled room conditions (22–24°C and a 14 h light: 10 h dark photoperiod). Mice were given access to food and water ad libitum. All manipulations with animals were performed according to the Local Bioethical Committee recommendations and according to the Declaration of Helsinki (1996). Tamoxifen treatment was performed as described previously: 8-10 week old males were injected with 3 mg of tamoxifen daily for 7 consecutive days (Ilchuk et al., 2022). Tamoxifen used for KO induction is an estrogen receptor modulator and can affect the mouse organism, especially fertility and hormonal levels, per se (Willems et al., 2011). To ensure that observed effects are caused by the KOs and not by tamoxifen treatment we used tamoxifen-treated controls in every experiment and conducted all experiments but one (Fig. 1D) at least one month after tamoxifen treatment. Mice were sacrificed by cervical dislocation. To evaluate fertility, males were kept with two CD-1 outbred female mice each. Copulative plugs and newborn pups were monitored daily in the morning.

Histology

For primary tissue examination organs were fixed in 10% formaldehyde, paraffin embedded, sectioned on a microtome and stained with H&E.

For immunofluorescent staining frozen sections were used. Organs were collected in tissue freezing medium (Leica Biosystems, Germany). Sections were cut 3 μm thick in the cryostat Leica CM1850 at −20°C. Slides were incubated for 2 min in 4% PFA and washed with PBS three times for 10 min. Sections were permeabilized in 0.03% Triton-X100 (VWR Life Science, Radnor, PA), 0.1% rabbit serum diluted in PBS at 4oC overnight and washed again with PBS. Sections were stained with primary antibodies (Supplementary Table S3) at a dilution of 1:100 for an hour, washed in PBS, and incubated with secondary antibodies (1:100). Slides were stained with 5 ng/ml DAPI (Thermo Fisher Sci., Waltham, MA), mounted with Mowiol® 4-88 (Sigma-Aldrich) and examined on Leica STELLARIS confocal microscope (Leica Microsystems, Germany).

Flow cytometry spermatogenesis assay

The protocol modified from (Jeyaraj et al., 2003) was used. In testicular cells, this method allows to distinguish between 4n primary spermatocytes that entered meiosis, 2n cells (Leidyg, Sertoli and other somatic cells, spermatogonia stem cells and secondary spermatocytes, that completed first meiotic division) and two 1n groups: round spermatids and elongated spermatids after DNA compaction as well as apoptotic (subG1) cells (Fig. 2A, left) (Jeyaraj et al., 2003; Pereira et al., 2016). The testes of mature mice were placed in phosphate buffered saline (PBS), the tunica albuginea was opened, and the contents were transferred to a solution of 0.5 µg/ml collagenase IV (PanEco, Russia) in PBS and incubated for 15 min at 32°C with shaking at 220 rpm. The seminiferous tubules were washed twice with 1 U DNAse (NE Biolabs, Ipswich, MA) in PBS, transferred to 0.01% trypsin (PanEco), and shaked for 15 min at 32°C, 220 rpm. Trypsin was inactivated with 0.01% bovine serum albumin in PBS; cells were thoroughly resuspended, passed through 70 μm nylon cages (Wuxi NEST Biotechnology, China) and reconstituted in 1 ml PBS. For flow cytometry cells were fixed with 0.75% PFA for 15 minutes at 37°C, then washed in PBS. Cells were lysed in the buffer containing 50 μg/ml propidium iodide (PI), 100 μg/ml RNAse A, 0.1% sodium citrate, 0.3% NP-40 (VWR Life Science) for 30 min at 4°C in the dark followed by flow cytometry analysis on a CytoFlex 26 (Beckman Coulter, Indianapolis, IN) in PE-A and PerCP-A channels. At least 10,000 fluorescent ‘events’ were collected per each sample. Data were analyzed using CytExpert software (Beckman Coulter).

10x Chromium library preparation and sequencing

2 R26-Cre-ER^T2 and 2 Cdk8^fl/flCdk19^-/-R26-Cre-ER^T2 tamoxifen-treated male mice at 7th week after activation were sacrificed and tissue cell suspension was prepared as described above. The phenotype was confirmed by cell cycle analysis of a portion of cell suspension used for the 10X library preparation. Flow cytometry analysis showed complete absence of haploid cells in one DKO animal while the other had a subpopulation of round but not elongated spermatids (Supplementary Fig. S2).

Single cell 3’ v3 kit and Chromium controller (10X Genomics) (Zheng et al., 2017) were used to generate GEMs for further processing. RT and other stages of single cell library preparation were performed as per manufacturer instructions, with the exception of AMPure XP Reagent used instead of SPRIselect. These sample libraries were sequenced with Illumina Novaseq6000.

Single cell data processing

Raw fastq files generation, alignment and read filtering were done by 10x Genomics Cell Ranger 6.1.1 with default settings. Read quality was additionally assessed with FastQC. Reads were aligned to the GRCm38 genome, with an average 95% alignment rate. Raw counts, generated by Cell Ranger 6.1.1, were piped to Seurat (Hao et al., 2021), where additional filtering was applied. Cells with aberrant UMI vs. genes relations were filtered with pagoda2 genes vs. molecule filter (https://github.com/kharchenkolab/pagoda2), where minimal counts were set as 1000 and minimal genes as 500. Next, cells with high percentages of mitochondrial genes expressions per cell were excluded, threshold for filtering set as 25%. After that Scrublet doublet (Wolock et al., 2019) score was calculated, cells with scores higher than 0.20 were marked as doublets and excluded. Each sample had more than 5000 high quality cells remaining after filtering. These cells were normalized with the default Seurat method and piped to further filtering and analysis.

To account for ambient gene expression, DecontX (Yang et al., 2020) was used, for which cluster-assigned data is required. For stable cluster assignment, a reference dataset was formed of WT samples, which were integrated via Harmony (Korsunsky et al., 2018) and then clustered with Seurat. Clustering was checked against testis cell type markers (Cao et al., 2021; Hermann et al., 2018) to ensure selection of true clusters that would reiterate across samples. Cluster of cells belonging to the IGB_2 WT sample only, which did not correspond to any cell type and was missing essential housekeeping genes, was discarded as a cluster of low quality cells with no biological relevance. This reference dataset was used for automatic cluster mapping with Seurat for all samples, and received labels were used with DecontX to lessen share of ambient gene expression in data.

Single cell data analysis

Corrected counts for all cells were piped back to Seurat. Then normalized WT and double knockout samples were integrated with Harmony to account for samples’ batch effects. 3480 variable genes were used for PCA and following integration, none of them mitochondrial or ribosomal. Harmony-integrated data was UMAP dimension-reduced and clustered with Seurat. Visualizations were also made with the Seurat package. Cell type was assigned according to their marker genes (Cao et al., 2021; Hermann et al., 2018). Both cells and microenvironment in the testis were detected in wild-type as well as in KO samples.

Spermatogonia cell cluster was reintegrated and reclustered separately for discovery of spermatogonia subtypes. Integration and clusterization was done as before, with 798 features at the start of the process. Spermatogonia subclusters were assigned types according to literature-curated marker genes (Cao et al., 2021; Hermann et al., 2018). Differential expression tests (negative binomial Seurat implementation) were conducted for each chosen cell cluster with WT compared against double knockout, batch effect between samples accounted for with the linear model. All mitochondria, ribosomal and high ambient genes (by DecontX) were excluded from testing. Genes with absolute logFC < 0.4 and genes expressed in less than 20% cells in class were excluded either. Additionally, after testing, all genes with no or almost no expression in double knockout, high expressions in WT, and high expression in WT spermatids were excluded as their differential expression might be caused by different ambient composition in WT, heavily influenced by larger spermatid cell proportion. For Sertoli cells, a small subcluster, potentially caused by technical effects, was excluded before testing.

Assignment of groups and cell count

The clustering analysis was performed according to markers described in previous studies (reviewed in (Suzuki, 2023)). This analysis allowed us to identify nine populations with respective genes: Sertoli cells (Wt1, Sox9, NR5a1, Clu), Leydig cells (Hsd3b1, Cyp17a1), macrophages (Cd74, C1qa, Cd45, Ccl5), T-cells, fibroblasts (Pdfrgfra, Cd34, Dcn, Gcn), and the differentiating sperm lineage divided into 5 clusters: undifferentiated/early spermatogonia (Crabp1, Tcea3, Kit), meiotic entry (Stra8, Rhox13) leptotene/zygotene (Sycp3, Dmc1), two clusters for pachytene (Piwil1, Id4) and pachytene/diplotene (Acr, Pgk2) (Supplementary Fig. S5).

Hormone measurement

To measure the concentration of hormones in the serum, blood was taken from the left ventricle of the heart, transferred to sterile test tubes and left at 4°C overnight. Measurement of luteinizing hormone was performed using ELISA (CEA441Mu, CloudClone, China). The testosterone concentration was measured by the LC-MS method as described previously (Povaliaeva et al. 2020). 17α-Hydroxyprogesterone-d8 were used as standard, proteins were precipitated by adding ZnSO4 and MeOH. The measurement was performed using an on-line extraction method with Agilent Bond Elut C18 cartridges as trap column and Waters Aquity UPLC BEH C18 column, Agilent 1290 Infinity II LC and and AB Sciex Triple Quad 5500 mass-spectrometer.

Immunoblotting

Protein lysis and western blotting were performed as described in (Ilchuk et al., 2022). Briefly, cells were lysed in RIPA buffer supplemented with protein inhibitor cocktail (Sigma-Aldrich, St. Louis, MO). Total protein concentration was quantified by the Bradford method. Absorbance at 560 nm was measured with a CLARIOstar Plate Reader (BMG Labtech, Germany). Proteins were separated by SDS-PAGE and transferred onto 0.2 μm nitrocellulose membrane (Bio-Rad, Hercules, CA). After blocking with 5% skimmed milk, membranes were treated with primary antibodies (Supplementary table S4) and incubated at 4oC overnight. Membranes were washed with Tris-borate saline with Tween 20 (TBS-T) and incubated for 1 h at room temperature with secondary antibodies (Supplementary table S4). Membranes were visualized with the Clarity Western ECL Substrate (Bio-Rad) using a iBright FL1500 Imaging System (Invitrogen, Waltham, MA).

Leydig cell culture

Seminiferous tubules were isolated as described above for flow cytometry but after collagenase IV (1 ug/ml) digestion the supernatant was passed through 70 μm nylon cages and centrifuged at 1000 rpm. The cells were resuspended in the medium for Leydig cells (low glucose DMEM (PanEco) with 10% fetal bovine serum (FBS, HyClone, GE Healthcare Life Sciences, Chicago, IL), 1% penicillin/streptomycin (PanEco), 0.3 mg/ml L-glutamine (PanEco)) and seeded on Costar® 6-well plates (Corning, New York, NY) pre-coated by 0.01% poly-L-lysyne (Sigma-Aldrich, P0899), incubated at 37°C in 5% CO2 for 1 h and carefully washed with the medium. After 24 h cells were treated with 0.56% potassium chloride hypotonic solution for 5 min to remove unattached cells. On the 7th day 1 µM Senexin B (SenB, Senex Biotechnology, Columbia, SC) or 1 µM 4-hydroxytamoxifen (4-OHT, Sigma-Aldrich) was added to the cells. After 7 days of incubation, the cells were lysed with ExtractRNA (Evrogen, Russia).

qRT PCR

Total RNA from cell suspension was extracted with ExtractRNA (Evrogen) and quantified using NanoDrop. Equal amounts of RNA were reverse-transcribed to generate cDNA using Superscript II reverse transcription Supermix for qRT-PCR (Invitrogen). qRT-PCR was then performed with SYBR PCR Mix (Evrogen) using the Thermal Cycler system (Bio-Rad). The data thus obtained were analyzed following the comparative (ΔΔCt) method. Actb was used as a housekeeping gene. Sequences of primers used are listed in Supplementary table S4.

Statistics

The normality of data was tested using the Shapiro–Wilk test, all datasets met the condition for normality (p > 0.05). One-way or two-way analysis of variance (ANOVA) followed by Holm-Sidak’s post hoc test for multiple comparisons was used (GraphPad Prism 8; GraphPad Software, San Diego, CA). P value <0.05 was taken as evidence of statistical significance.

Declaration of interests

I.B.R is Founder and President and M.C. is consultant of Senex Biotechnology, Inc. Other authors declare no conflict of interest.

Acknowledgements

Animal experiments, single cell RNA sequencing and histology were funded by Russian Science Foundation project #22-15-00227. Flow cytometry and confocal microscopy experiments were performed at the equipment of Center for Precision Genome Editing and Genetic Technologies for Biomedicine and funded by the Ministry of Science and Higher Education of the Russian Federation (075-15-2019-1661). We thank Dr. Andrey Yu. Kulibin for providing anti-SYCP3 antibodies and Tamara A. Kiriukhina for assistance in isolating of primary MEF culture.

Author contributions

A.B., E.V., N.S.., V.N.M., A.T., D.K., A.K., M.U., A.N., J.L., M.C., V.A.M. and V.T. carried out the experiments. A.B. wrote the manuscript with support from V.T. and E.V. E.V. designed the figures. E.V., A.K., I.R., M.C. and A.S. edited the manuscript. A.B., V.T., E.V., V.N.M. and Y.S. analysed the data. Z.A., V.B., D.O. and E.A. designed the computational framework and analysed the scRNA data. I.R. and A.S. conceived the original idea. A.B., I.R., A.S. and Y.S acquired the funding. A.B. and V.T. designed experiments, interpreted data and were the project administrators.

List of abbreviations

• 4-OHT: 4-hydroxytamoxifen

• AR: androgen receptor

• BTB: blood-testicular barrier

• CcnC: Cyclin C

• CKM: cyclin dependent kinase module

• DEG: differentially expressed genes

• DKO: double knockout

• LH: luteinizing hormone

• PI: propidium iodide

• RNA Pol II: RNA polymerase II

• scRNAseq: single cell RNA sequencing

• SenB: Senexin B

• TF: transcription factor