CDK12 Is Necessary to Promote Epidermal Differentiation Through Transcription Elongation

Abstract

Proper differentiation of the epidermis is essential to prevent water loss and to protect the body from the outside environment. Perturbations in this process can lead to a variety of skin diseases that impacts 1 in 5 people. While transcription factors that control epidermal differentiation have been well characterized, other aspects of transcription control such as elongation are poorly understood. Here we show that of the two cyclin-dependent kinases (CDK12 and CDK13), that are known to regulate transcription elongation, only CDK12 is necessary for epidermal differentiation. Depletion of CDK12 led to loss of differentiation gene expression and absence of skin barrier formation in regenerated human epidermis. CDK12 binds to genes that code for differentiation promoting transcription factors (GRHL3, KLF4, and OVOL1) and is necessary for their elongation. CDK12 is necessary for elongation by promoting Ser2 phosphorylation on the C-terminal domain of RNA polymerase II and the stabilization of binding of the elongation factor SPT6 to target genes. Our results suggest that control of transcription elongation by CDK12 plays a prominent role in adult cell fate decisions.

Graphical Abstract

In epidermal progenitor cells committed to the differentiation process, CDK12 and SPT6 promote the transcription elongation of epidermal differentiation promoting transcription factors such as GRHL3, KLF4, and OVOL1. In the absence of CDK12, there is stalling of RNA polymerase II at the transcriptional start site of GRHL3, KLF4, and OVOL1 which results in blockade of epidermal differentiation.

Significance Statement.

Skin diseases can afflict up to 20% of the population, often resulting from improper terminal differentiation. Thus understanding the regulators of epidermal differentiation is key to unlocking the mechanisms behind skin disorders. Here we show that CDK12 is responsible for promoting epidermal differentiation and in its absence differentiation does not occur. CDK12 promotes differentiation by promoting the transcription elongation of differentiation genes. These findings suggest that drugs targeting CDK12 may be useful for the treatment of skin disorders.

Introduction

Proper regulation of gene expression is not only essential for the development of an organism but also for cells to respond to stress, environmental stimulus as well as balanced growth and differentiation in the adult state. Dysregulation of gene expression has long been known to contribute to human disease most notably in tumorigenesis.^1,2 Studies on the regulation of gene expression have been primarily focused on transcription factors but they represent only one aspect of transcription. Transcription driven by RNA polymerase II (Pol II) is regulated at multiple levels including initiation, pausing, elongation, and termination with each of these stages regulated by specific cyclin-dependent kinases (CDKs) and their respective cyclins.³ The preinitiation complex (PIC) is formed when general transcription factors (GTFs) are assembled at promoters and cause the recruitment of Pol II.⁴ Once recruited, CDK7/cyclin H, which is associated with TFIIH (GTF), phosphorylates the C-terminal domain of Pol II on serine 5 (ser5) and serine 7 (ser7) to promote initiation.^5-7 Ser5 phosphorylation allows Pol II release from the PIC and transcription of short nascent RNA which allows binding of the DRB sensitivity-inducing factor (DSIF: SPT4 and SPT5) and negative elongation factor (NELF).^8-10 Binding of DSIF/NELF leads to Pol II pausing typically after transcribing 20-60 bps of RNA.^11-13 Release of paused Pol II into productive transcription occurs when positive elongation factor b (P-TEFb: CDK9 and cyclin T) phosphorylates DSIF, NELF, and serine 2 (Ser2) of Pol II’s CTD.¹⁴ Phosphorylation of SPT5 (DSIF) and NELF results in the dissociation of NELF from Pol II.¹⁵ Without the inhibitory NELF, the PAF complex and SPT6 can associate with Pol II to facilitate transcription elongation.^10,16 SPT6 is a histone chaperone that interacts with both histones H3 and H4 to disassemble and reassemble nucleosomes to allow passage of Pol II during elongation.^4,17

While the role of CDK7/cyclin H and CDK9/cyclin T in Pol II initiation and pause-release, respectively, has been well studied, the CDKs involved in promoting transcription elongation are much less characterized. Recent studies have shown that CDK12 and CDK13 perform the majority of Ser2 phosphorylation on Pol II’s CTD to regulate elongation.^18,19 It is currently unclear whether CDK12 or CDK13 have any roles in regulating cell fate decisions and tissue homeostasis. These questions have not been addressed due to the embryonic lethal phenotype in mice of knockout of these genes.^20,21 Understanding the roles of these CDKs in tissue development/homeostasis is especially pertinent as many of these are dysregulated in diseases such as cancer.^22,23

The human epidermis is a fast turnover tissue that constantly relies on stem and progenitor cells residing in the basal layer to continuously supply differentiated cells to form the barrier of our skin.²⁴ Alterations in the differentiation process that leads to defective barrier formation (disrupted stratum corneum) can cause a variety of skin diseases that afflicts ~20% of the population.²⁵ Proper keratinocyte differentiation requires an exit out of the cell cycle and induction of the differentiation gene expression program. Most studies on epidermal differentiation have focused on identifying and characterizing transcription or epigenetic factors necessary for this process. Work from our lab and others have shown that key factors, such as ZNF750, KLF3, KLF4, MAF, MAFB, GRHL1, OVOL1, GRHL3, EHF, BRD4, JMJD3, and CEBP alpha/beta are essential for differentiation.^26-36 Intriguingly, our recent work demonstrated that ~30% of epidermal differentiation genes already contained paused Pol II at its transcriptional start site (TSS) in stem and progenitor cells.³⁷ Many of these poised differentiation genes with paused Pol II binding coded for the differentiation promoting transcription factors described above. Upon induction of differentiation, Pol II is released into productive elongation allowing for the expression of these transcription factors which then turns on the rest of the epidermal differentiation gene expression program. A small RNAi screen of pause-release/elongation factors identified SPT6 and PAF1 to be the key factors involved in the elongation of the differentiation-specific transcription factors as well as structural differentiation genes.³⁷ Loss of function of SPT6 or PAF1 led to inhibition of epidermal differentiation and accumulation of Pol II at the TSS of differentiation genes. These results suggest that regulation of transcription elongation plays a prominent role in cell fate decisions and suggests that the CDKs that regulate this step may potentially be important for this process.

Here, we show that of the CDKs that regulate transcription elongation, only CDK12 loss blocked differentiation of regenerated human skin. CDK12 bound to more than 8000 genes primarily localized to their TSS. CDK12 also associated with over half of the genes that SPT6 bound. These genes include key differentiation promoting transcription factors such as OVOL1, GRHL3, and KLF4 as well as differentiation-specific structural genes such as TGM1. CDK12 is necessary for Pol II Ser2 phosphorylation of its bound genes. In the absence of CDK12, Pol II Ser2 phosphorylation is lost on critical epidermal differentiation genes, which results in Pol II depletion from the gene body and buildup at the TSS. This loss of elongation causes decreased expression of crucial differentiation genes, which in turn leads to blockade of terminal epidermal differentiation. Without CDK12, elongation factors such as SPT6 are unable to bind. These results highlight the prominent role that transcription elongation factors have on somatic tissue differentiation.

CDK12 But Not CDK13 Is Necessary to Promote Epidermal Differentiation

To test whether the CDKs that regulate transcription elongation (CDK12 and CDK13), have any impact on epidermal function, each was knocked down in primary human keratinocytes using siRNAs. The knockdown cells were placed in high calcium and full confluence for 3 days to induce differentiation. Interestingly, loss of CDK13 did not impact the expression of epidermal differentiation genes (Supplementary Fig. 1A). In contrast, CDK12 knockdown (validated through 2 distinct siRNAs [CDK12-Ai and CDK12-Bi] targeting different regions of the gene) blocked the expression of critical differentiation structural genes such as KRT10, FLG, LOR, ABCA12, and TGM1 (Fig. 1A). Mutations in genes such as KRT10, FLG, ABCA12, and TGM1 lead to severe skin diseases such as epidermolytic hyperkeratosis, ichthyosis vulgaris, harlequin ichthyosis, and congenital ichthyosis, respectively.^38-41 Differentiation promoting transcription factors that we previously showed to be controlled by transcription elongation were also reduced in expression in CDK12i cells³⁷ (Fig. 1A). Control and knockdown of primary human keratinocytes were also used to regenerate human skin by seeding the cells on devitalized human dermis.^42-44 This allows the cells to establish cell-cell and cell-basement membrane contact to allow proper growth, differentiation, and stratification in 3 dimensions. This technology has been used to cure patients of junctional epidermolysis bullosa through the autologous transplantation of transgenic regenerated human skin.^45,46 Similar to cells cultured in 2D, CDK13 knockdown did not alter skin proliferation, differentiation, or morphology (Supplementary Fig. 1B-1E). No phenotypes were observed in CDK13 knockdown cells or tissue despite CDK13 being expressed on the protein level in both proliferative and differentiated cells (Supplementary Fig. 1F). CDK12 was localized in the nucleus throughout all layers of the epidermis with staining disappearing in CDK12i tissue (Supplementary Fig. 2A). CDK12 loss blocked differentiation as evidenced by a lack of stratum corneum formation (Fig. 1B). The terminally differentiated layer of the epidermis did not form due to an absence of filaggrin (FLG) and loricrin (LOR) protein expression (Fig. 1D). Early differentiation protein (keratin 10: K10) expression was also downregulated (Supplementary Fig. 2B). The proliferative capacity of the basal layer was also compromised as evidenced by the loss of Ki67-positive cells which resulted in a hypoplastic tissue (Fig. 1C, Supplementary Fig. 2B, 2C). However, the basal layer transcription factor, P63,⁴⁷ was not impacted upon CDK12 depletion (Supplementary Fig. 2D). To understand the genes that CDK12 is regulating on a genome-wide level, RNA-sequencing (RNA-Seq) was performed on CTLi and CDK12i cells cultured in differentiation conditions for 3 days. About 1130 genes (P value <.05 and ≥2-fold change) decreased in expression upon CDK12 depletion, which were enriched in genes involved in keratinocyte differentiation, skin development, and sphingolipid metabolic process (Fig. 1E, 1F, Supplementary Table 1). About 2725 genes were upregulated upon CDK12 knockdown, which were enriched in regulation of cellular macromolecule biosynthetic process and cellular response to type I interferon (Fig. 1E, 1G, Supplementary Table 1). The increased expression of interferon-related genes may be a response to the barrier loss in CDK12i cells. A comparison with the differentiation gene expression signature (differentially expressed genes upon epidermal differentiation³⁰) showed that 33.3% of the CDK12 gene signature overlapped (Supplementary Fig. 2E). Importantly, these 1285 overlapped genes were enriched for Gene Ontology (GO) terms such as epidermis development suggesting that CDK12 is vital for epidermal differentiation and may control this process through elongation (Supplementary Fig. 2F).

Figure 1.

CDK12 is necessary for epidermal differentiation. (A) Primary human keratinocytes were knocked down with control (CTLi) or CDK12 (CDK12-Ai and CDK12-Bi are distinct siRNAs that target 2 different regions of the gene) siRNAs and placed in differentiation conditions for 3 days (full confluence and 1.2 mM calcium). The mRNA levels of each gene (x-axis) were measured using RT-qPCR. qPCR results were normalized to L32 levels. (B) Regenerated human skin using three-dimensional organotypic cultures made from CTLi or CDK12i cells were harvested after 4 days of culture. Hematoxylin and eosin staining of regenerated CTLi and CDK12i human skin. The dashed black lines denote the basement membrane zone and scale bar = 20 µm. (C) Quantification of tissue thickness from (B). (D) Immunostaining of late differentiation proteins filaggrin (FLG: red) and loricrin (LOR: green) are shown (day 4 regenerated human skin). Nuclei are shown in blue (Hoechst staining). The dashed white lines denote the basement membrane zone and scale bar = 20 µm. (E) RNA-Seq analysis of CTL and CDK12 knockdown primary human keratinocytes differentiated for 3 days (full confluence and 1.2 mM calcium). RNA-Seq results were performed in duplicates. About 2725 genes were upregulated (red) and 1130 genes were downregulated (blue) upon CDK12 knockdown. Heatmap is shown in Log2 scale. (F) Gene Ontology (GO) terms for the 1130 genes downregulated in CDK12i cells. (G) GO terms for the 2725 genes with increased expression upon CDK12 depletion. N = 3 independent experiments for Figure 1 unless otherwise indicated. Mean values are shown with error bars = SD. **P < .01, ***P < .001, ****P < .0001 (t test). Abbreviations: CDK12, cyclin-dependent kinase 12; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; siRNAs, small interfering RNA.

CDK12 Binds and Regulates Epidermal Differentiation Genes

To determine whether CDK12 directly regulated the expression of epidermal differentiation genes, we performed ChIP-Seq on CDK12 in differentiated human keratinocytes. CDK12 bound to 12 230 peaks with the vast majority (88%) of the binding sites mapping back to genic regions (promoter, intron, exon, 5ʹ-UTR, TTS, and 3ʹ-UTR) (Fig. 2A). The 12 230 CDK12-bound sites mapped back to 8630 genes which were enriched for GO terms such as “negative regulation of transcription” and “ubiquitin dependent protein catabolic process” (Fig. 2B, Supplementary Table 2). Since CDK12 has previously⁴⁸ been implicated in regulating transcription elongation through Pol II, we compared our previously³⁷ generated Pol II ChIP-Seq data to CDK12. The CDK12- and Pol II-binding profiles were similar at CDK12-bound genes (Fig. 2C, 2D). There were also notable differences in binding as CDK12 binding was primarily concentrated at the TSS of genes whereas Pol II binding also had another peak near the transcriptional end site (TES) (Fig. 2D). 40.6% of the genes increased upon CDK12 loss were bound by CDK12 and were enriched in GO terms such as positive regulation of cell morphogenesis and negative regulation of epithelial cell proliferation (Fig. 2E, 2F). 490 (43.4%) of the genes decreased upon CDK12 knockdown were also bound by CDK12 and were enriched for the establishment of skin barrier and positive regulation of epidermis development GO terms (Fig. 2E, 2G). Since CDK12 bound to genes both upregulated and downregulated upon CDK12 loss, we analyzed the distribution of CDK12 binding on those genes to determine if its binding can explain the changes in gene expression. Interestingly, the 490 genes that were decreased upon CDK12 knockdown had much higher levels of CDK12 binding to the gene body than the 1105 genes that were upregulated (Fig. 2H). In contrast, the upregulated genes had increased levels of CDK12 at the TSS as compared to the 490 downregulated genes (Fig. 2H). This suggests that gene body accumulation of CDK12 is important for the expression of its target genes. The genes targeted by CDK12 in the gene body included those that code for key transcription factors (GRHL3, OVOL1, KLF4) that promotes epidermal differentiation as well as structural differentiation genes (TGM1) suggesting that CDK12 may regulate their elongation (Supplementary Fig. 3A-3D).

Figure 2.

CDK12 binds to genes coding for epidermal differentiation proteins. (A) Genomic localization of the 12 230 CDK12-bound peaks. The percentage of CDK12 binding to each genomic region is shown. CDK12 ChIP-Seq was performed in primary human keratinocytes cultured in differentiation conditions (full confluence and 1.2 mM calcium: day 3). N = 2. (B) Gene Ontology terms of the 8630 genes that CDK12 binds. (C) Heatmap of CDK12 and RNA Pol II ChIP-Seq (day 3 differentiated cells) on sites bound by CDK12. x-Axis shows −2 kb to +2 kb from the TSS. Heatmap is shown ranked by decreasing CDK12 binding. (D) Metagene plot of CDK12 (green) and RNA Pol II (orange) ChIP-Seq reads at CDK12-bound regions. y-Axis is shown as read count per million reads and x-axis is distance along with CDK12-bound genes. TSS is the transcriptional start site and TES is the transcriptional end site. (E) Venn diagram of CDK12-bound genes (CDK12 ChIP-Seq) with genes upregulated or downregulated upon CDK12 depletion. (F) GO terms of the CDK12-bound genes that overlap with genes increased upon CDK12 loss. (G) GO terms of the 490 decreased genes upon CDK12 knockdown that overlap with CDK12-bound genes. (H) Metagene plot of CDK12 ChIP-Seq reads on the 490 (orange) genes downregulated and 1105 (green) genes upregulated upon CDK12 knockdown. y-Axis is shown as read count per million reads and x-axis is distance along with CDK12-bound genes. TSS is the transcriptional start site and TES is the transcriptional end site. Abbreviations: CDK12, cyclin-dependent kinase 12; ChIP-Seq, chromatin immunoprecipitation sequencing; GO, Gene Ontology.

CDK12 Promotes Elongation of Differentiation Genes through Pol II Ser2 Phosphorylation and Binding of the Elongation Factor SPT6

To gain insight into the mechanism of how CDK12 controls epidermal differentiation, we performed Pol II Ser2p (phosphorylation) ChIP-Seq in control and CDK12 knockdown-differentiated cells (Supplementary Table 3A, 3B). Loss of CDK12 led to a large depletion of Pol II Ser2p at genes where CDK12 bound (Fig. 3A). At genes where CDK12 did not bind, there was much less change in Pol II Ser2p binding between control and CDK12 knockdown groups (Fig. 3B). There was no change in total levels of Pol II Ser2p in CDK12i cells suggesting that CDK12 specifically phosphorylates Pol II Ser2 on genes that it binds (Fig. 3C). These genes include OVOL1, KLF4, TGM1, and GRHL3 where CDK12 loss led to a reduction in Pol II Ser2p binding throughout the gene including the TSS (Fig. 3D-3G). Since Pol II Ser2p has been correlated with transcription elongation, loss of it may potentially cause total Pol II depletion from the gene body and accumulation at the TSS. To test this, total Pol II ChIP was performed in CTLi- and CDK12i-differentiated cells. The absence of CDK12 resulted in loss of total Pol II binding from the gene bodies of OVOL1, KLF4, TGM1, and GRHL3 and buildup at the TSS (Fig. 4A, 4B). Importantly no impacts were seen with Pol II binding on housekeeping gene PGK1, which is a gene that CDK12 does not bind (Fig. 4A, 4B, Supplementary Table 2). This suggests that CDK12 depletion leads to failure to elongate differentiation genes and thus subsequent buildup of Pol II at the TSS of those genes. In addition to promoting elongation through Pol II Ser2 phosphorylation, CDK12 may also be required for the stabilization of elongation factors binding to differentiation genes. Since we previously demonstrated³⁷ that the elongation factor SPT6 was necessary for epidermal differentiation, we compared the SPT6-bound genes to CDK12. Nearly 60% (3623/6307) of the genes bound by SPT6³⁷ were also bound by CDK12 (Fig. 4C). In addition, 30.5% (345/1130) of the genes downregulated in CDK12i cells also overlapped with the genes decreased in SPT6 knockdown cells³⁷ (Fig. 4D). These 345 overlapped downregulated genes were enriched for skin development and keratinocyte differentiation genes suggesting that CDK12 and SPT6 are in the same pathway and that CDK12 may be necessary for stabilizing SPT6 binding to target genes (Fig. 4E). To examine whether CDK12 is necessary for the stabilization of SPT6 binding to differentiation genes, SPT6 ChIP was performed in differentiated CTLi and CDK12i cells. In control cells, SPT6 bound robustly to OVOL1, GRHL3, KLF4, and TGM1 (Fig. 4F). In contrast, CDK12 knockdown abolished SPT6 binding to those genes (Fig. 4F). The diminished binding is not due to a loss of SPT6 levels as CDK12 knockdown did not alter protein levels of SPT6 (Fig. 4G). These results suggest that CDK12 promotes elongation of epidermal differentiation genes through the phosphorylation of Pol II Ser2 and stabilization of elongation factor binding.

Figure 3.

CDK12 is necessary for Pol II Ser2 phosphorylation at CDK12-bound genes. (A) RNA polymerase II Ser2 phosphorylation (Pol II Ser2p) ChIP-Seq at regions bound by CDK12 in CTLi (green)- and CDK12i (orange)-differentiated cells (day 3: full confluence and 1.2 mM calcium). y-Axis is shown as read count per million reads and x-axis is distance along with genes. TSS is the transcriptional start site and TES is the transcriptional end site. Pol II Ser2p ChIP-Seq in CTLi and CDK12i cells were performed in duplicates. (B) Metagene plot of Pol II Ser2p ChIP-Seq at regions not bound by CDK12 in CTLi (green)- and CDK12i (orange)-differentiated cells. (C) Western blot of Pol II Ser2p in CTLi- and CDK12i-differentiated cells. Actin was used as a loading control. N = 3. Representative blots are shown. (D-G) Gene tracks of OVOL1 (D), KLF4 (E), TGM1 (F), and GRHL3 (G). Pol II Ser2p ChIP-Seq is shown in CTLi (blue) and CDK12i (red) cells. y-Axis shows reads per million (RPM) and black bar over gene tracks represent significant peaks. x-Axis shows position along the gene. Abbreviations: CDK12, cyclin-dependent kinase 12; ChIP-Seq, chromatin immunoprecipitation sequencing.

Figure 4.

CDK12 is necessary for SPT6 binding to epidermal differentiation genes. (A) RNA polymerase II (Pol II) ChIP on CTLi- and CDK12i-differentiated cells (day 3: full confluence and 1.2 mM calcium). qPCR was used to determine the amount of binding to genes listed on the x-axis. Primers were targeted toward the gene body of each gene. Results are plotted as a percentage of input. (B) Pol II binding at the TSS of each gene listed on the x-axis in CTLi- and CDK12i-differentiated cells. (C) Venn diagram of genes bound by SPT6 (SPT6 ChIP-Seq) and CDK12 (CDK12 ChIP-Seq). (D) Overlap of the CDK12i decreased genes with the SPT6i downregulated genes. (E) Gene Ontology of the 345 overlapped genes from (D). (F) SPT6 ChIP on CTLi- and CDK12i-differentiated cells (day 3: full confluence and 1.2 mM calcium). qPCR was used to determine the amount of binding to genes listed on the x-axis. Primers were targeted toward the gene body of each gene. Results are plotted as a percentage of input. (G) Western blot of SPT6 in CTLi- and CDK12i-differentiated cells. Actin was used as a loading control. Representative blots are shown. N = 3 independent experiments for Fig. 4A, 4B, 4F, 4G. Mean values are shown with error bars = SD. *P < .05, **P < .01 (t test). Abbreviations: CDK12, cyclin-dependent kinase 12; NS, not significant; qPCR, quantitative polymerase chain reaction; TSS, transcriptional start site.

Discussion

The role of CDK12 or CDK13 in adult tissue homeostasis has not been characterized due to the embryonic lethality of murine knockout models.^20,21 This is an important issue that needs to be addressed since CDK12/CDK13 inhibitors have been proposed to be used in combination with PARP inhibitors (to induce synthetic lethality) for the patients with breast cancer.^49,50 CDK12 knockout leads to embryonic lethality by E6.5.²⁰ In murine embryonic stem cells CDK12 or CDK13 depletion results in spontaneous differentiation due to loss of pluripotency gene expression.⁵¹ Our results show an opposite phenotype in somatic tissue in that CDK12 but not CDK13 is essential for epidermal differentiation. In the absence of CDK12, regenerated human skin fails to differentiate including an absence of stratum corneum formation. CDK12 depletion also led to a thinner epidermis and loss of Ki67-positive proliferative cells in the basal layer. This phenotype is likely due to CDK12 also being expressed in the basal layer of the epidermis (Supplementary Fig. 2A) and its known role in promoting mammalian proliferation by phosphorylating cyclin E1.⁵² Global gene expression profiling of CDK12 knockdown cells showed that 2725 genes were upregulated and 1130 genes were decreased in expression. The genes that were increased in expression upon CDK12 depletion were enriched in inflammation-related GO terms. This is a phenomenon we have observed in other genes (ZNF750, CBP, or BRD4) where the loss of function phenotype is the inhibition of epidermal differentiation.^26,35,36 This shared upregulation of inflammatory gene signatures suggests that there is a potential feedback mechanism that senses a breach in the barrier. Thus, upregulation of the inflammatory response may serve to combat any pathogens that would get into the skin due to the loss of barrier formation. Deciphering this mechanism may be an area of future investigation. The 1130 downregulated genes were enriched in keratinocyte differentiation and skin development GO terms. Furthermore, a third of the CDK12 regulated overlapped with the differentiation gene signature. These results indicate that CDK12 controls the differentiation gene expression program.

To gain an understanding of the genes that CDK12 directly regulates to enable differentiation to occur, we mapped the genome-wide binding sites of CDK12. Our data in primary human skin cells demonstrated that CDK12 bound to 8630 genes. Interestingly, CDK12 bound to a similar percentage of genes that were increased or decreased in expression upon CDK12 knockdown. The binding of CDK12 either to the gene body or TSS seems to correlate with whether the gene will be downregulated or upregulated upon CDK12 loss. Genes that were downregulated upon CDK12 depletion had more CDK12 binding to the gene body. Conversely upregulated genes upon CDK12 knockdown had higher levels of CDK12 binding to the TSS than downregulated genes. This agrees with the model of transcription elongation as CDK12 phosphorylation activity is required in the gene body to promote elongation.^53,54

Of the 8630 genes that CDK12 bound only 490 of those genes were downregulated upon CDK12 loss. Importantly, these 490 CDK12-bound genes that decreased in expression upon CDK12 depletion were enriched for skin barrier and epidermis development genes suggesting that differentiation genes were more sensitive to perturbations in CDK12 levels. This may be due to the cell fate change where differentiation genes are highly induced and thus may be more reliant on transcription elongation. In addition, we had previously shown that many of the differentiation genes including GRHL3, TGM1, and OVOL1 already had paused Pol II at its TSS in epidermal stem and progenitor cells which progressed to active elongation upon induction of differentiation.³⁷ This suggests that paused genes may be more reliant on CDK12 to promote elongation. Thus during cell fate transitions paused and highly induced genes are more susceptible to perturbations in CDK12 levels. In addition, redundancy in kinases may also explain why only a small portion of CDK12-bound genes are downregulated upon CDK12 loss. It has been shown that inhibition of CDK12 in certain contexts resulted in a global loss of Pol II Ser2 phosphorylation while in others no impacts were observed. In HEK293 cells, CDK12 inhibition caused a loss of global Pol II Ser2 phosphorylation.⁵⁴ In contrast, our work here and studies shown in HCT116 cells observed no impacts on global Pol II Ser2 phosphorylation upon CDK12 inhibition.^55,56 The global sensitivity of Pol II Ser2 phosphorylation may be due to the cell types used as cell types such as HCT116 cells and primary human keratinocytes may have redundant kinases such as CDK13 and BRD4 that can phosphorylate global levels of Ser2.^18,57,58 This redundancy may help explain why only a small portion of the CDK12-bound genes are differentially expressed upon CDK12 loss.

Next, we wanted to determine how CDK12 impacted Pol II Ser2p levels. On CDK12-bound genes, there was a large reduction in Pol II Ser2p levels in CDK12 knockdown cells whereas there was minimal impact on genes not bound by CDK12. Loss of Pol II Ser2p was found on critical differentiation promoting transcription factors such as KLF4, GRHL3, and OVOL1. Consistent with other studies,⁵⁴ CDK12 depletion led to decreased levels of Pol II Ser2p throughout the gene including the TSS. This in turn led to a reduction of total Pol II found along with the gene bodies of KLF4, GRHL3, TGM1, and OVOL1. The reduction in the gene bodies caused a buildup of total Pol II at the TSS of those genes resulting in a blockade of transcription elongation.

In addition to promoting Pol II Ser2 phosphorylation, CDK12 may also be important for stabilizing elongation factor binding to target genes. We previously showed that SPT6 is necessary to promote the transcriptional elongation of epidermal differentiation genes and thereby could be dependent on CDK12 for binding to target genes.³⁷ Supporting this, a majority (57.4%) of the SPT6-bound genes overlapped with CDK12, and 30.5% (345/1130) of the CDK12i-downregulated genes overlapped with the genes decreased upon SPT6 knockdown. Most importantly, knockdown of CDK12 resulted in decreased SPT6 binding to its target genes including KLF4, GRHL3, TGM1, and OVOL1. It should be noted that we did not determine on a genome-wide level all the genes that require the presence of CDK12 in order to allow SPT6 to bind. It is likely there are genes that require SPT6 for elongation but not CDK12 since there are potentially redundant kinases as discussed above. CDK12 may potentially phosphorylate SPT6 to allow contact with other members of the elongation complex and stabilize it. It was previously shown that THZ531, a small molecule inhibitor of CDK12- and CDK13-impacted SPT6 phosphorylation.⁵⁹ Our results here demonstrate that CDK12 and SPT6 act in the same pathway to promote epidermal differentiation whereas other elongation factors are necessary for other aspects of skin physiology. For example, we previously showed that the FACT complex (SSRP1 and SPT16) is not essential for epidermal differentiation but is necessary for transcription elongation of MAP2K3 to mediate the skin inflammatory response.^37,60 In addition, ELL which is part of the super elongation complex has no impact on differentiation of the skin but is essential for the proliferation of epidermal stem and progenitor cells.^8,61 Thus, the elongation factors that have been previously identified through biochemical assays to promote transcription elongation and thought to act in the same pathway actually have different gene targets and functions in tissue. In the future, it will be important to characterize each elongation factor’s function in adult tissue since many are altered in diseases such as cancer.³

Our results coupled with work from other labs suggest that upon pause-release, NELF leaves the Pol II complex which allows the recruitment of the PAF complex and SPT6.^15,16 The PAF complex then recruits CDK12.⁵³ Upon recruitment, CDK12 phosphorylates Pol II Ser2 to promote elongation.⁵³ CDK12 may potentially also phosphorylate SPT6 to trigger the stability of the complex.⁵⁹ These events lead to productive transcription elongation of epidermal differentiation genes. Tellier et al⁵⁴ showed that inhibition of CDK12 activity led to a loss of SPT6 and PAF complex from chromatin and the elongation complex suggesting that its kinase activity is required for stabilizing the elongation complex. Similarly here we showed that CDK12 is required for stabilizing the binding of SPT6 to a subset of CDK12-bound target genes. We also previously showed that PAF1 depletion impaired epidermal differentiation similar to the SPT6 and CDK12 knockdown phenotype.³⁷ These results suggest that the transcription elongation complex composed of CDK12, SPT6, and PAF1 is critical to promote epidermal differentiation.

Methods

Primers and siRNA Sequences

All primers and siRNA sequences can be found in the Primers and siRNA Sequences (Supplementary material).

Cell Culture

Primary human epidermal keratinocytes (derived from neonatal foreskin) were cultured in EpiLife medium (Thermo Fisher Scientific: MEPI500CA) supplemented with human keratinocyte growth supplement (Thermo Fisher Scientific: S1001K) and pen/strep. Proliferating, non-differentiated keratinocytes were cultured in subconfluent conditions. To induce epidermal differentiation, primary human keratinocytes were plated at full confluence in the presence of 1.2 mM calcium for 3 days.

Regenerated Human Skin

For the organotypic skin cultures, 1 million control, CDK12, or CDK13 knockdown cells were seeded onto devitalized human dermis to regenerate human epidermis.^42,44,62,63 Human dermis was purchased from the New York Firefighters Skin Bank. The seeded cells were raised to the air-liquid interface to promote stratification and differentiation. Tissue was harvested 4 days after initial seeding.

Gene Knockdown

siRNAs targeting human CDK12 (Dharmacon D-004031-01 and D-004031-02) CDK13 (Life Technologies s16398) or control siRNAs (final concentration 10 nM) were transfected into keratinocytes using Lipofectamine RNAiMAX (Thermo Fisher Scientific: 13778-500) reagent according to the manufacturer’s protocol and incubated for 18 hours.

RNA Isolation and RT-qPCR

Total RNA from cells was extracted using the GeneJET RNA purification kit (Thermo Fisher Scientific: K0732) and quantified using a Nanodrop. About 1 µg of total RNA was reversed transcribed using the Maxima cDNA synthesis kit (Thermo Fisher Scientific: K1642). Quantitative PCR was performed using the Bio-Rad LFX96 real-time system. L32 was used as the internal control for normalization.

Western Blotting

About 20 µg of the cell lysates were used for immunoblotting and resolved on 10% SDS-PAGE and transferred to PVDF membranes. Primary antibodies used include beta-actin (Santa Cruz: Sc-47778) at 1:5000, RNA pol II Ser2p (Active Motif: 91115) at 1:1000, CDK13 (Bethyl: A301-458A) at 1:500, and SPT6 (Bethyl: A300-801A) at 1:1000. Secondary antibodies including IRDye 800CW (LI-COR: 926-32212) donkey anti-mouse and IRDye 680RD (LI-COR: 926-68073) donkey anti-rabbit were used at 1:10 000.

Histology and Immunofluorescence

Regenerated human skin sections were fixed in 4% paraformaldehyde for 11 minutes followed by blocking in phosphate-buffered saline (PBS) with 2.5% normal goat serum, 0.3% Triton X-100, and 2% bovine serum albumin for 30 minutes. Primary antibodies used were CDK12 (Sigma: HPA008038) at 1:300, P63 (Abcam: Ab124762) at 1:300, Loricrin (Abcam: Ab198994) at 1:1000, Filaggrin (Abcam: Ab3137) at 1:200, MKi67 (Abcam: Ab16667) at 1:300, and Keratin 10 (Abcam: Ab9025) at 1:500. The secondary antibodies used were Alexa Fluor 555-conjugated goat anti-mouse IgG (Thermo Fisher Scientific: A11029) or Alexa Fluor 488-conjugated donkey anti-rabbit IgG (Thermo Fisher Scientific: A21206) both at 1:500. Nuclear dye, Hoechst 33342 (Thermo Fisher Scientific: H3570) was used at 1:1000.

Hematoxylin and Eosin Staining

Sectioned tissue derived from regenerated human skin was fixed with 10% formalin solution (Sigma HT5012) for 12 minutes. Sections were then dipped in 0.25% Triton X-100 in PBS for 5 minutes. Hemotoxylin (Vector H-3401) staining was performed for 8 minutes, rinsed in water, and then dipped in acid alcohol (1% HCL in 70% ethanol). After subsequent rinsing, the sections were dipped in 0.2% ammonia water for 1 minute, rinsed again, and then dipped in 95% ethanol. Eosin (Richard-Allan Scientific 71304) staining was performed for 30 seconds followed by 95% ethanol rinsing for 1 minute. Sections were then put into 100% ethanol for 4 minutes followed by 2 minutes in Xylene.

RNA-sequencing (RNA-seq) and Library Preparation

Control and CDK12i cells were placed in differentiation conditions (full confluence and 1.2 mM calcium) for 3 days. Total RNA was isolated using the GeneJET RNA (Thermo Fisher Scientific: K0732) purification kit and quantified by Nanodrop for control and CDK12i cells. RNA-seq was performed using the Illumina NovaSeq S4 machine at the Institute of Genomic Medicine core facility at UCSD. ~30 million reads per sample were obtained using pair-ended 100 base long reads.

Chromatin Immunoprecipitation Sequencing (ChIP-seq), Library Preparation, and ChIP-qPCR

Ten million cells and 5 µg of antibody were used for each antibody pulldown experiment for ChIP.^36,43,64 ChIP was performed using the following antibodies: SPT6 (Bethyl: A300-801A), RNA Pol II (Active Motif: 91151), RNA Pol II Ser2 (Active Motif: 91115), CDK12 (Bethyl: A301-679A and LSBio: LS-A7350), Rabbit IgG (Millipore: 12-370) and mouse IgG (Abcam: Ab18413). Cells for the RNA Pol II/RNA Pol II Ser2p ChIP-qPCR or ChIP-Seq were fixed at a final concentration of 1% formaldehyde. Cells for the SPT6 or CDK12 ChIP-qPCR or ChIP-Seq were fixed in both formaldehyde (1% final concentration, Thermo Fisher Scientific: 28908) and disuccinimidyl glutarate (2 mM final concentration, Thermo Fisher Scientific: 20593). qPCR results are represented as a percentage of input DNA.

For ChIP-Seq, the ChIP DNA library was prepared using the TruSeq DNA sample prep kit (Illumina). Sequencing was done on the HiSeq 4000 System (Illumina) using single 1 × 75 reads at the Institute for Genomic Medicine Core, UCSD.

RNA-seq Analysis

Reads were aligned to the GENCODE v19 transcriptome hg19 using TopHat2 with default settings.⁶⁵ Differential expression among samples was calculated using analysis of variance (ANOVA) from the Partek Genomic Suite (Partek Incorporated). Analysis of the read count distribution indicated that a threshold of 10 reads per gene generally separated expressed from unexpressed genes, so all genes with fewer than 10 reads were excluded from ANOVA analysis. Gene lists for significantly upregulated or downregulated genes were created using FDR <0.05 and ≥2-fold change. Enriched GO terms for RNA-seq differentially expressed gene sets were identified using Enrichr.^66,67 Heatmaps for the RNA-seq data were generated using Partek’s Genomic Suite (http://www.partek.com/partek-genomics-suite/).

ChIP-seq Analysis

The ChIP-seq reads were processed by the ENCODE Transcription Factor and Histone ChIP-Seq processing pipeline (https://github.com/ENCODE-DCC/chip-seq-pipeline2) on our local workstation. The reads were first trimmed based on the quality score before alignment to reference hg19; Upon alignment and deduplication, the peak-calling was then carried out by MACS2.2.4 with a cutoff q value of 0.05.^68,69 The heatmaps for the ChIP-Seq data were generated using the ngs.plot.⁷⁰ Gene tracks were visualized using the UCSC genome browser along with annotation tracks.

Funding

This work was supported by grants from the National Institutes of Health (NIH R01AR072590, R01AR066530, and R01CA225463) to G.L.S.

Conflict of Interest

The authors declared no potential conflict of interest.

Author Contributions

J.L.: conceived of the project, designed the experiments, wrote the paper, performed the experiments, performed the bioinformatics analysis. M.T., Y.C.: performed the experiments. S.L.: made the graphical abstract. G.L.S.: conceived of the project, designed the experiments, wrote the paper. All authors contributed to the reading and editing of the manuscript.

Data Availability

The datasets generated from this study including RNA-Seq and ChIP-Seq data have been deposited in GEO (GSE166407).