Early embryogenesis in CHDFIDD mouse model reveals facial clefts and altered cranial neurogenesis

CDK13 is one of the transcriptional kinases that regulates transcription via phosphorylation of RNA polymerase II (RNAPII) and controls alternative splicing (Bartkowiak et al., 2010; Blazek et al., 2011). Recently, a few case studies have described a variety of developmental defects in human patients carrying mutated CDK13 (Hamilton and Suri, 2019). These patients exhibit delayed development, intellectual disorders, heart and kidney defects, and craniofacial anomalies, features that – together – have been recognized as congenital heart defects, dysmorphic facial features and intellectual development disorder (CHDFIDD). In addition to the most common anomalies, patients suffer from brain anomalies, autism, seizures, limb and skeletogenesis anomalies, and other clinical presentations (Hamilton and Suri, 2019). To simulate anomalies presented in patients with CDK13 mutations, we have developed mouse models that exhibit phenotypes similar to those observed in humans (Nováková et al., 2019). We observed that Cdk13 deficiency in the mouse models causes embryonic lethality, delayed development, heart and brain abnormalities and a facial phenotype (Nováková et al., 2019).

Here, we screened the gene expression of Cdk13 across several embryonic craniofacial tissues at different developmental stages, and evaluated the alteration of developmental processes in craniofacial structures in Cdk13-hypomorph (Cdk13^tm1a/tm1a) and Cdk13-knockout (Cdk13^tm1d/tm1d) mouse embryos, with the aim to uncover the mechanisms behind the observed developmental abnormalities. To determine the modification in molecular regulations caused by Cdk13 deficiency, we also analyzed the gene expression of patterning proteins involved in craniofacial structure formation.

Moreover, previous in vitro approaches have uncovered the contribution of a Cdk13 mutation to the alteration of neuronal differentiation and neurite outgrowth through regulation of the CDK5 pathway (Chen et al., 2014). These findings are in agreement with observations of neurodevelopmental disorders in human patients carrying mutations of CDK13 (Trinh et al., 2019). Based on this evidence, we focused here on the analyses of cranial nerve growth and morphology in Cdk13-deficient animals, i.e. mice and chicken, that are associated with craniofacial structure formation, including palate morphogenesis. Further, we evaluated possible changes of neurogenesis-specific genes in Cdk13-deficient animals and tested the effect of the CDK12/13 inhibitor THZ531 on axon outgrowth from the embryonic trigeminal ganglion.

In our previous study, we observed craniofacial developmental anomalies caused by Cdk13 deficiency, which resulted either in a less severe phenotype in hypomorphic embryos [secondary palate dysmorphisms at embryonic day 15.5 (E15.5)] or in a more severe phenotype in knockout embryos (midfacial cleft at E13.5) (Nováková et al., 2019). The morphological difference between these two Cdk13 genotypes was explained by the residual expression of CDK13 in hypomorphic embryos – probably caused either by insertion of a neomycin selection cassette into the non-coding region, which has been shown to affect gene expression at both RNA and DNA levels, or by a high frequency of post-transcriptional exon shuffling within Cdk13, enabling the formation of aberrant functional Cdk13 transcripts – compared to undetectable expression of CDK13 in knockout embryos (Nováková et al., 2019). Thus, we decided to analyze all the crucial embryonic stages of Cdk13^tm1a/tm1a and Cdk13^tm1d/tm1d mice, all of which exhibited morphological anomalies of craniofacial structures including severe clefting (Fig. 1N).

Split facial prominences and, thus, a widely opened face (i.e. midfacial cleft) were detected in Cdk13^tm1d/tm1d embryos at E11.5, revealing also the development of telencephalon in the midline (Fig. 1A″). In Cdk13^tm1a/tm1a embryos at the same embryonic stage, we detected widely set facial prominences (Fig. 1A′). Measurements of distances between individual facial prominences revealed a significantly greater ratio of nasal pits distances and head width in maxillary prominence and lateral nasal prominence levels in both mutant genotypes (Fig. 2A-C). Widely set facial prominences at earlier embryonic stages then progress at E12.5 to median cleft lip in Cdk13^tm1a/tm1a embryos (Fig. 1B′,H′) but persisted in Cdk13^tm1d/tm1d embryos, leading to the development of midfacial cleft (Fig. 1B″,H″). Delayed development of the palatal shelves (Fig. 1E′,K′,K″) with cleft nasal septum (Fig. 1K′,K″) was detected in both mutant genotypes. Later, at E13.5 and E14.5, Cdk13^tm1a/tm1a embryos displayed thinner lips with increased distance between (Fig. 1F′,I′,G′J′) and delayed development of the palatal shelves (Fig. 1F′,L′,G′,M′), while Cdk13^tm1d/tm1d embryos exhibited a severe midfacial cleft (Fig. 1C″,I″,D″,J″) and persisting nasal septum cleft (Fig. 1L″,M″). Less severe anomalies were detected in the caudal palate area, in which underdeveloped palatal shelves were the most prominent feature, especially in Cdk13^tm1d/tm1d embryos (Fig. S1E-G). For Cdk13^tm1a/tm1a embryos observed at the latest embryonic stage (i.e. E16.5), improperly fused lips in the midline as well as cleft palate persisted within the rostral and caudal areas, accompanied with necrotic tissue that was visible within all structures (Fig. S1B-B″,D,D′). This confirmed embryonic lethality after E15.5, which had also been noticed previously (Nováková et al., 2019).

Based on craniofacial morphometrics, heads and individual facial prominences are generally smaller in Cdk13-deficient embryos. Thus, we quantified gene expression of CyclinD1, a gene responsible for cell cycle progression from G1 to S phase and used it as a marker of proliferation to see if there is a change in proliferation rates. Compared with control embryos, CyclinD1 expression was detected to be upregulated in almost all tissues isolated from facial prominences at E11.5 and E12.5 of Cdk13-deficient embryos (Fig. 2E). Although the palatal shelves in Cdk13-deficient animals were generally smaller and exhibited a different shape, there were no significant changes detected in CyclinD1 expression in the palatal shelves of E12.5 Cdk13^tm1a/tm1a embryos. A slight downregulation of CyclinD1 was detected in both palatal regions isolated from E14.5 Cdk13^tm1a/tm1a mice (Fig. 2E). Additional immunostaining for Ki-67 and TUNEL assays confirmed no significant changes in proliferation and apoptosis within the developing palatal shelves of Cdk13^tm1a/tm1a embryos (Fig. S2A-C).

As we observed distinct changes in facial and palatal morphogenesis in Cdk13-deficient embryos, we asked if there is localized expression of Cdk13 mRNA within certain areas during craniofacial development. We evaluated gene expression of Cdk13 on frontal sections to reveal possible differences in its distribution along the labio-lingual axis of the palatal shelves. In embryos between E12 and E14, the Cdk13 signal was spread evenly within the palatal mesenchyme and epithelium of the palatal shelves (Fig. 3A-C). Later, at E15, we detected increased expression of Cdk13 in the palatal mesenchyme, close to the region of fusion in the craniofacial midline (Fig. 3D). In the developing snout, intensity of the Cdk13 signal was similar in the surface epithelium and mesenchyme. A visibly stronger signal was detected in the forebrain (Fig. 3F) and nasal epithelium at E11 (Fig. 3F″,F″′), and in the maxillary mesenchyme at E12 (Fig. 3G). At later stages (E13, E14), Cdk13 mRNA was evenly spread in the developing snout in both mesenchyme and epithelium (Fig. S3). Quantification of relative Cdk13 expression by qPCR revealed decreasing levels in the palatal area throughout the development (Fig. 3H) and increasing levels in the maxillary, mandibular and nasal prominences between E11.5 and E12.5 (Fig. 3I).

To uncover possible differences of Cdk13 distribution along the rostral-caudal axis of the developing palate, we analyzed its expression in the sagittal sections. During earlier developmental stages (E11-E14), Cdk13 was expressed in both rostral and caudal areas of the developing palatal shelves, with higher Cdk13 signal intensity detected in the rostral region compared to that in caudal areas (Figs S4A-D and S5). Changes emerged in older embryos (E15-E16), where we observed strong expression of Cdk13 in the palatal epithelium and adjacent mesenchyme, especially in the developing palatal rugae, both in the rostral and caudal regions (Fig. S4E,F).

Next, we asked how CDK13 protein is distributed in cells in vitro. We analyzed its expression in cultured mouse embryonic fibroblasts (MEFs) and cells derived from DRGs (dorsal root ganglia) of adult mice as these cell types typically form long cytoplasmic processes.

In MEF cells, CDK13 was localized to the nuclear area and enriched in the cytoplasm around the nucleus (Fig. 4A). However, CDK13 protein was also detected within cellular outgrowths (Fig. 4A′-A″′), including their most apical tips, where it was located together with actin filaments (Fig. 4A′). In cells isolated from DRGs, CDK13 was detected in the nuclear area, with highest levels detected in the cytoplasm adjacent to nuclei (Fig. 4B). As observed in cultured MEFs, we also detected CDK13 within long neural processes along the neurofilaments that were forming the outgrowths (Fig. 4B′,B″). CDK13 in cytoplasm as analyzed by immunofluorescence staining was further evaluated by cellular fractionation followed by western blotting for CDK13. As expected, CDK13 was detected in the nuclear fraction of primary MEF cells; however, significant levels of CDK13 were also detected in the cytoplasmic fraction (Fig. 4C). Moreover, the mouse fibroblast cell line NIH3T3 was used to assess the subcellular localization of CDK13, and CDK13 was found in both cytoplasmic and nuclear cellular fractions, with higher levels in the nuclear fraction (Fig. 4D).

Hypomorphic mutation causes production of a truncated form of CDK13, with preserved N-terminal domain (Nováková et al., 2019). This form of CDK13 aggregated in the cellular processes together with deposits of F-actin in E12.5 Cdk13^tm1a/tm1a MEFs (Fig. S6A). Similarly, aggregates of truncated CDK13 were located in protrusions of cells isolated from E12.5 Cdk13^tm1a/tm1a DRGs (Fig. S6B). Additionally, live cell imaging of MEF cells revealed different numbers of cellular protrusions per cell in Cdk13^tm1d/tm1d cells (Fig. S6C, top) compared to control cells. This was associated with the more-stretched morphology of cultured Cdk13^tm1d/tm1d cells (Fig. S6D), their weaker adherence to culture plates and easier release from plates during trypsin digestion (data not shown). However, cellular velocity (Fig. S6C, bottom) and the percentage of the cellular surface being protruded (Fig. S6C, middle) showed no statistically significant difference between either genotype in MEF cultures.

This specific cellular localization of the CDK13 to cellular protrusions indicates a potential cytoplasmic role of the CDK13 protein. However, when we performed detection of ICC signal by polyclonal anti-CDK13 antibodies, we revealed a reduction of the CDK13-specific signal in nuclei (see arrowheads in Fig. S6E) but not in the cytoplasm surrounding the nuclei or the cellular protrusions of Cdk13^tm1d/tm1d cells (Fig. S6E), indicating primary effect of Cdk13 deficiency in nuclei.

Development of the craniofacial region is closely associated with the development of cranial nerves. A large area is innervated by the trigeminal nerve comprising three main branches, i.e. the maxillary, mandibular and ophthalmic branch (Higashiyama and Kuratani, 2014). We detected complex spatial Cdk13 gene expression in the developing maxillary nerve in cells that ensheath bundles of nerve fibers (Fig. S7A-E′). As the CDK13 has been demonstrated to regulate neurite outgrowth (Chen et al., 2014), we hypothesized that alteration of cranial nerve growth occurs during early craniofacial development in Cdk13-deficient animals.

First, we visualized outgrowth of cranial nerves by whole-mount immunohistochemistry analysis of neurofilaments to uncover possible alterations of general morphology of trigeminal nerve and associated nerves. Alterations regarding the outgrowth of several cranial nerves were detected in Cdk13^tm1a/tm1a and Cdk13^tm1d/tm1d embryos (Fig. 5B,C′), with obvious hypoplasia of maxillary, mandibular and ophthalmic nerves, which were reduced in length. These nerves originate from the trigeminal ganglion (TG), which we also found to be abridged in the mutant embryos (Fig. 5B′,C′). Only the frontal nerve (frN) was not morphologically altered (Fig. 5B′,C′). This confirms that Cdk13 deficiency negatively affects neurogenesis during embryonic development.

As observed by using whole-mount immunohistochemistry, changes of nerve protrusions were even more obvious on transversal sections of E11.5 Cdk13^tm1d/tm1d embryos. In the rostral region, we observed missing maxillary nerves in the developing maxillary prominence (Fig. 5D′,E′;,Fig. S8A′,B′), hypoplastic MnNs in the mandibular prominence (Fig. 5E′) and reduced size of the TG (Fig. S7F,G) confirming anomalies regarding nerve growth in Cdk13-deficient animals.

We also detected alterations in the expression patterns of genes and proteins strongly associated with craniofacial development, and specifically enriched in mesenchymal (SOX9, Msx1, Foxd1) and epithelial structures (SOX2, Meis2, Shh) in this region. In E11.5 Cdk13^tm1d/tm1d embryos, we observed very low levels of the neural crest marker SOX9 in the mesenchyme of lateral nasal prominences and maxillary prominences, including an area with undeveloped maxillary nerves (Fig. 5D′; Fig. S8A′). However, the pattern of SOX2, one of the regulators of neural cells differentiation and peripheral nervous system development (Adameyko et al., 2012), was similar to that in wild-type (WT) embryos, only missing in the area of the prospective maxillary nerve (Fig. 5E′; Fig. S8B′). Expression of Meis2, a gene involved in cranial neural crest development (Machon et al., 2015), was absent from the neural tube in the rostral area (Fig. 6A, see arrow at the top, indicating nt) and less expressed in the same tissue more caudally. Moreover, Meis2 expression was detected caudally, close to the facial midline in knockout (Fig. 6A, asterisks) but not in WT mice. Conversely, Meis2 was enriched in the mesenchyme of the rostral medial nasal prominences (Fig. 6A). Msx1, a key player in craniofacial development (Levi et al., 2006), was absent from the mesenchyme of the medial nasal prominences (asterisks) and also from the brain close to its midline (Fig. 6B). However, analysis of genes necessary for determination and development of facial primordia (Jeong et al., 2004), revealed that expression of Shh was enhanced in the ventral neural tube within the rostral region but also, to a similar extend, in the developing brain more caudally (Fig. 6C, arrows). Its altered expression pattern in the epithelium covering the forming stomodeal cavity was detected rostrally (Fig. 6C). Moreover, enhanced expression of Foxd1, which is expressed downstream of Shh, in the developing mandibular prominences points to altered development of the lower jaw (Fig. 6D). Finally, expression of Fgf8 was absent in the epithelium of the developing primitive oral cavity close to the head midline (Fig. 6E, arrows and asterisks) and its expression was reduced in the developing brain (Fig. 6E, asterisks). This confirmed that Cdk13 deficiency leads to deregulation of these key developmental genes already in early craniofacial development, subsequently resulting in severe facial clefting.

To further evaluate alterations in the expression of genes relevant to neurogenesis during palatogenesis in Cdk13^tm1a/tm1a embryos, we used PCR array analysis to simultaneously investigate the expression profile of 84 neurogenesis-specific genes. In separately dissected tissues from rostral and caudal palatal shelves (Fig. 7A), we first compared expression of neurogenesis-specific genes in rostral and caudal palatal regions (Table S1-S4).

Of all genes analyzed this way (Fig. 7A), 15 genes (Ache, Apoe, Bmp2, Dcx, Fgf2, Heyl, Mef2c, Neurod1, Neurog1, Nog, Nrg1, Ntn1, Pou4f1, S100a6 and Sox3) were found to be significantly enriched in caudal compared to rostral palatal shelves in control (WT) embryos at E12.5 (Fig. 7B, top left panel). The same comparison between rostral and caudal palatal shelves in Cdk13^tm1a/tm1a embryos at E12.5 determined a preserved gene expression ratio only for Bmp2 – probably caused by the downregulation of expression of caudal palatal shelve-dominant genes (Ache, Dcx, Ntn1, Pou4f1, Bdnf, Bmp2, Ep300, Fgf2, Heyl, Mef2c, Kmt2a, Neurog1, Nf1, Nog, Notch2, Nrg1 and S100a6) in the caudal palatal shelves (Fig. 7C, right panel). By contrast, a change in ratio was detected for Hey2 and Ntn1, whose expression was – conversely – enriched in the rostral palatal shelves (Fig. 7B, top right panel), caused by extreme upregulation of both genes in mutant embryos (Fig. 7C, left panel).

In control embryos at a later stage, i.e. at E14.5, ten genes were enriched in rostral palatal shelves (Artn, Bcl2, Bmp2, Gdnf, Hey1, Neurog2, Pax3, Pou3f3, Slit2 and Tgfb1) and 15 genes in caudal palatal shelves (Ache, Apoe, App, Cdk5r1, Dvl3, Fgf2, Map2, Ndn, Nf1, Ntf3, Ntn1, Pafah1b1, Pou4f1, Robo1 and Rtn4), (Fig. 7B, bottom left panel). Again, the same comparison between rostral and caudal palatal shelves, this time in E14.5 Cdk13^tm1a/tm1a embryos, revealed a preserved gene expression ratio for 11 genes (Ache, App, Bcl2, Fgf2, Hey1, Nf1, Ntn1, Pax3, Pou3f3, Slit2 and Tgfb1) in E14.5 Cdk13^tm1a/tm1a embryos. Generally, deregulation of the neurogenesis-specific genes in the E14.5 palatal shelves was not as substantial (Fig. 7D) as was in case for E12.5 (Fig. 7C). But, in both observed embryonic stages, gene expression was generally downregulated in the caudal compared to the rostral region, whereas in the rostral region both up- and downregulation of gene expression was detected (Fig. 7C,D).

Moreover, altered expression patterns of palatal genes and proteins were detected later at E12.5 and E14.5. Cdk13-deficiency leads to noticeable reduction of Foxd1 expression in the mesenchyme of snout midline (Fig. 7F,G), Msx1 reduction in the rostral palatal shelves (Fig. 7J,K) and also reduction of Meis2 in the caudal palatal shelves (Fig. 7L,M), especially at E12.5. Levels and distribution of both SOX9 and SOX2 were similar to those observed for WT animals (Fig. 7H,I).

To address the effect of CDK13 depletion on transcription, we performed chromatin immunoprecipitation (ChIP) with an antibody specifically recognizing RNA polymerase II (RNAPII) and IgG as a negative control (Fig. S9). For two genes selected from the PCR array analysis (Fig. 7), i.e. Pou4f1 and Ntn1, we decided to evaluate the presence of RNAPII within their promoter regions in MEF cells obtained from Cdk13^+/+ and Cdk13^tm1d/tm1d mice. Increased levels of RNAPII within either promoter of Pou4f1 and Ntn1 were observed in comparison to IgG control (Fig. S9A,B). In the case of the Pou4f1 promoter, we detected less RNAPII in MEF cells obtained from Cdk13^tm1d/tm1d compared to those obtained from Cdk13^+/+ mice (Fig. S9A). In contrast, RNAPII was associated more with the Ntn1 promoter in Cdk13^tm1d/tm1d MEFs than in Cdk13^+/+ MEF cells (Fig. S9B). We then decided to compare the presence of RNAPII on these promoters as a percentage of RNAPII compared to IgG control. A slight decrease of RNAPII was observed for the Pou4f1 promoter in Cdk13^tm1d/tm1d MEF cells compared with that in Cdk13^+/+ MEFs (Fig. S9C). In contrast, higher occupancy of RNAPII was demonstrated for the Ntn1 promoter in Cdk13^tm1d/tm1d MEFs when compared with Cdk13^+/+ MEFs (Fig. S9D). Although the statistical difference for RNAPII occupancy at these promoters was not significant, the obtained observation correlates with decreased and increased mRNA levels of Pou4f1 and Ntn1 genes as detected by using PCR expression screen (Fig. 7). However, future research regarding these findings will be necessary.

To confirm the effect of CDK13 deactivation on craniofacial structure formation, we used the synthetic CDK13 inhibitor THZ531 to treat early maxillary prominences of Gallus gallus embryos. It is necessary to note that THZ531 is also targeting CDK12, which shares several functions with CDK13 and, in developmental processes and under certain conditions, they can replace each other. Therefore, the usage of this inhibitor should prevent also compensatory mechanisms induced by CDK13 deficiency.

We injected 1 mM THZ531 into the craniofacial mesenchyme at Hamburger–Hamilton (HH) developmental stage 10 (HH10), the postoptic region at HH15 and into maxillary prominences at HH20. Then, chicken embryos were allowed to develop for 4-6 days (Fig. 8A). This led to the absence or deficiency of maxillary prominences, resulting in the unilateral cleft lip (Fig. 8B,C). Injection of THZ531 at earlier stages resulted in a stronger phenotype with high deficiency in craniofacial tissues (Fig. S10).

Furthermore, the effect of Cdk13 deficiency on cranial nerve development as shown earlier (Fig. 5) was additionally tested by chemical inhibition of CDK13 (together with CDK12) in functional ex vivo cultivation experiment using TG explants (Fig. 8). TGs were dissected from E12 WT mouse embryos and cultured, while culture medium was supplemented with THZ531 (100 nM or 300 nM). A significant reduction in the formation of neurite outgrowths was observed in TGs when cultured with different concentrations of the CDK13 inhibitor THZ531 compared to control group (Fig. 8E-G), revealing the importance of both CDK13 and CDK12 in development of cranial nerves.

Here, we determined a key contribution of CDK13 to craniofacial morphogenesis and neurogenesis, where the phenotype of Cdk13-deficient animals included a smaller head and disturbed facial morphology, including facial clefts. In our mutants, we found altered expression patterns of Msx1, Meis2, Shh, Foxd1 and Fgf8 in craniofacial structures. Downregulation or upregulation of these genes or their downstream targets causes similar phenotypic craniofacial phenotypes, including facial clefts or reduced growth of facial nerves (Levi et al., 2006; Machon et al., 2015; Jeong et al., 2004). Pronounced similarities in the morphology of craniofacial structures with Cdk13-knockout embryos has previously been observed in mice embryos with reduced functional expression levels of Fgf8. These animals also exhibited abnormal midline separation with nasal prominences set widely apart (Griffin et al., 2013). Such morphological changes indicated insufficiency in the development of neural crest cells (NCCs), from which all these affected craniofacial structures originate – which we confirmed by identifying reduced protein expression of the NCC marker SOX9 in facial prominences of Cdk13-deficient animals. Moreover, we detected significant downregulation of Mef2c (in all the palatal shelve regions, except rostral E12.5 palatal shelves). Its conditional downregulation, specifically in NCCs, has previously been found to result in craniofacial anomalies and neonatal lethality (Verzi et al., 2007).

Similar to patients with mutations in CDK13 or genes encoding other associated proteins, such as cyclin M, cyclin K, CDK10 (Colas, 2020) and CDK5RAP2 (Yigit et al., 2015), Cdk13-deficient mice produce neurogenic anomalies. These patients exhibit not just craniofacial anomalies but also affected neural tissues. Mutation in cyclin M results in optic nerve hypoplasia, mutation in cyclin K led to developmental delay and intellectual disabilities, and mutations in CDK10 cause intellectual disabilities connected with language and learning disorders (Colas, 2020). Patients with loss-of-function mutation of CDK5 display craniofacial anomalies other than facial clefts (short forehead, full cheeks, micrognathia), and also suffer of lissencephaly, cerebellar hypoplasia, microcephaly, intellectual disabilities, speech delay or autistic features (Colas, 2020), similar to patients with CDK13 mutations (Hamilton and Suri, 2019).

CDK5 has a specific role in neural tissues physiology and regulates several neural processes, such as axonal transport, migration and synaptic vesicle endocytosis, and is also important for regulation of axon and neurite outgrowth (Shah and Rossie, 2018). Its knockout leads to perinatal lethality in mice, connected with deficient neuronal migration and impaired axonal transport of neurofilaments (Ohshima et al., 1996). Importantly, in vitro experiments using cortical neurons has demonstrated the role of CDK13 and also CDK12 in the regulation of neurite outgrowth through a common signaling pathway that involves modulation of Cdk5 at RNA level (Chen et al., 2014). In our Cdk13-deficient models, we uncovered significant downregulation of genes encoding CDK5 regulatory subunits, such as of Cdk5rap2 and Cdk5r1, in the caudal palatal shelves at E14.5, confirming the CDK13-CDK5 functional association observed in in vitro experiments (Chen et al., 2014). Moreover, CDK5RAP2 mutations in human patients cause Seckel syndrome, manifested by microcephaly and cognitive problems (Yigit et al., 2015), and CDK5R1 was proposed to be a candidate gene in patients affected by NF1 microdeletion syndrome, which mutations lead to non-syndromic intellectual disability and undefined facial deformities (Venturin et al., 2006).

In this study, we found the CDK12/CDK13 inhibitor THZ531 to inhibit neurite outgrowths from the TG (Fig. 8E,G). Even though one could argue that the observed effect might be a result of CDK12 and CDK13 together, it is broadly accepted, that inhibition of CDK12 primarily targets DNA-damage response (DDR) pathways and pre-replication complex assembly (Pilarova et al., 2020). The process of neurite outgrowth takes place in nondividing differentiated cells of the TG; therefore, it is unlikely that altered outgrowth is the result of CDK12 inhibition due to blocking of the cell cycle in S-phase. Also, there is no significant impact of CDK12 inhibition of the DDR pathway since cells with lesser outgrowth of neurites do not exhibit any signs of DDR stress. Thus, the documented reduction in neurite outgrowth is likely a result of CDK13 inhibition through a different mechanism or, alternatively, a combinatory effect of CDK12 and CDK13 inhibition through regulation of CDK5 function in neural cytoskeleton organization; however, this needs further experimental evidence.

CDK13 is ubiquitously expressed in the murine snout, developing palatal shelves and cranial nerves and its mutation leads to prominent facial clefts. Craniofacial structures, including the palate, develop thanks to large contribution of NCCs. Growing peripheral nerves and glial cell types originate mostly from migrating NCCs and Schwann cell precursors (SCPs) (Furlan and Adameyko, 2018) but there are few exceptions, among them sensory neurons, which innervate maxillary region including the palate. These neurons originate in mesencephalic trigeminal nucleus in the CNS, and their development is regulated by FGF8 and Pou4f1 (Hunter et al., 2001), in agreement to that study, expression of both genes was detected altered in Cdk13-deficient embryos by us.

Along with these findings, we detected deregulated expression of genes that are typically associated with the TG in the developing palatal shelves. Expression of Nrg1 (Meyer et al., 1997), a gene encoding an EGF receptor ligand, which in turn induces axon growth (Onesto et al., 2021) was downregulated in the palatal shelves at E12.5. Interestingly, Nrg1 deficiency in mice also leads to maxillary dysmorphology (Waddington et al., 2017), an alteration from which palatal shelves are developing. Another gene usually detected in the TG is the neural transcription factor Neurog1 that previously has also been associated with the palatal shelves (Visel et al., 2007). Its downregulation in TGs was observed as a result of Pou4f1 downregulation (Lanier et al., 2009). Similarly, we found Pou4f1 and Neurog1 deregulated in palatal shelves of mice at E12.5. However, deregulation of Neurog1 expression corresponded with expression changes of Ntn1, a gene encoding a secreted factor that is responsible for the guiding of developing peripheral motor axons (Serafini et al., 1996), and the proposed risk factor for development of the non-syndromic cleft lip and cleft palate (Leslie et al., 2017). Expression of both genes was highly upregulated in rostral palatal shelves and strongly downregulated in caudal palatal shelves, which indicates a tendency to rescue nerve growth to rostral regions of the developing face. Fewer nerves and their reduced development in the facial region of Cdk13-deficient embryos is further accompanied with significantly downregulated expression of the Ache gene in the caudal palatal shelves at E12.5. Ache encodes an acetylcholinesterase necessary for the degradation of acetylcholine, leading to termination of signal transduction at neuromuscular junctions, which is also important for neurite outgrowth and its elongation (Olivera et al., 2003).

In Cdk13-deficient animals, we also determined changes in the expression of genes involved in neuronal migration, possibly leading to alteration of nerve growth. One of such genes, whose expression was detected to be downregulated in all analyzed palatal shelve regions, is Pafah1b1. This gene encodes a non-catalytic subunit of an enzyme responsible for activation of GTPases and actin polymerization at the leading edge of locomoting neurons making it important for neural migration (Kholmanskikh et al., 2003). Its mutation in humans frequently results in lissencephaly and Miller–Dieker syndrome (Liu et al., 2021), a disease with very phenotypic manifestations similar to those of Cdk13 deficiency, including growth delay and cerebral, cardiovascular, facial and limb anomalies (Yingling et al., 2003).

CDK13 was originally discovered as a nuclear factor participating in the regulation of transcription. We were able to detect CDK13 in the nucleus by immunofluorescence and cellular fractionation; however, we also determined significant levels of CDK13 in the cytoplasm. This might seem a rather surprising observation but presence of CDK13 in the cytoplasm of various cancer cells has been already documented by another group (Wu et al., 2023). Importantly, CDK13 can phosphorylate the intracellular domain of the transmembrane protein SERINC5 within the cytoplasm (Chai et al., 2021). These observations indicate that CDK13 can be involved not just in transcription or RNA processing but also in other cellular processes in cytoplasm – although this needs further analyses.

Our findings uncovered that Cdk13-deficiency triggers changes in the expression of neurogenesis-specific genes in the palatal shelves and the development of hypoplastic cranial nerves in vivo, confirming a previous in-vitro study using cortical neurons (Chen et al., 2014). However, a functional connection between anomalies in cranial nerve development, which, in turn, lead to facial malformations, has been detected only in few cases. Such non-canonical function of nerves as moderators of facial morphogenesis has been proven in Möbius syndrome, where either defective cranial nerves (Rizos et al., 1998; Tomas-Roca et al., 2015) or the entire defective rhombencephalon (Verzijl et al., 2003) can cause facial malformations. Other disorders of the peripheral nervous system that lead to facial anomalies are hereditary sensory and autonomic neuropathy type IV (HSAN IV) (Gao et al., 2013; Louryan et al., 1995; Nakajima et al., 2020) and Parry–Romberg syndrome with hemifacial atrophy accompanied by various neurological pathologies (Vix et al., 2015). Also, downregulated expression of the gene Nrg1, which was observed also by us, leads to altered maxillary development (Waddington et al., 2017), as this gene encodes neurogenin ligands with a positive axon growth potential. However, any functional association between the development of hypoplastic peripheral cranial nerves and the development of craniofacial clefts is still missing.

CDK13 is crucial for the development of several tissues and organs, including craniofacial structures. Here, we determined its role during development of facial structures in a mouse model of CHDFIDD, where loss-of-function in Cdk13 results in cleft lip/palate and the formation of midfacial clefts, accompanied by the deregulated expression pattern of certain genes and of proteins indispensable for proper functioning of major signaling pathways. Cdk13-deficient animals exhibited altered neurogenesis accompanied with distorted expression of neurogenesis-specific genes leading to the development of hypoplastic cranial nerves. As our analyses uncovered the cytoplasmic localization of CDK13 in vitro, further investigation is necessary to explore its potential roles beyond transcriptional regulation.

Heterozygous hypomorphic (Cdk13^{tm1a(EUCOM)Hmgu}) and knockout (Cdk13^tm1d) mice were obtained from the Infrafrontier Research Infrastructure – Mouse disease Models (https://www.infrafrontier.eu/services/rodent-model-generation/#MouseModelsGeneration). Mice were generated at the Transgenic and Archiving Module, CCP (IMG, Prague, Czech Republic).

Breeding and genotyping protocols were performed as previously described (see Nováková et al., 2019). In summary, the Cdk13^tm1a allele consists of a splicing acceptor that is surrounded by two FRT sites within intron 2. Moreover, it has two loxP sites within intron 2 and one loxP site within intron 4. The Cdk13^tm1d allele lacks exons 3 and 4, resulting in a non-functional allele of the Cdk13 gene (Nováková et al., 2019).

All animal procedures were performed in strict accordance with the Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee (Masaryk University, Brno, Czechia, No. MSMT-34505/2020-7).

Mouse embryonic fibroblasts (MEFs) were grown in conventional Dulbecco's modified Eagle medium (DMEM) supplemented with 4.5 g/l glucose and 20% fetal calf serum (FCS), the NIH3T3 cells were grown in DMEM supplemented with 4.5 g/l glucose and 10% FCS. Cell cultures were maintained at 37°C under 5% CO₂.

MEF cells isolated from WT and KO mouse embryos were cultured in DMEM as explained above. Cells were seeded on ibidiTreat μ-Slide 8 Well (80826, Ibidi) at concentration 50000/ml. After adhesion of cells, DMEM was changed to Opti-MEM (11058021, Gibco, Thermo Fisher Scientific) without Phenol Red supplemented with 1% fetal bovine serum (FBS). Cells were stained with F-actin-specific dye and membrane-specific dye 30 min before scanning. Live Imaging was performed using a Leica SP8 Confocal microscope (Leica, Germany) equipped with a CO₂-controlled and tempered transparent chamber. Automated detection and tracking of filopodia, and automated quantification of cell migration was carried out as described by Barry et al. (2015). Final analysis was performed using GraphPad (GraphPad Software, Boston, MA, USA).

Mouse embryos (control and Cdk13^tm1a/tm1a) were fixed in 4% paraformaldehyde, washed in distilled water and dehydrated through a graded series (30-100%) of ethanol solutions. Later, samples were dried out using the CPD 030 Critical Point Dryer (BAL-TEC) and shadowed by using gold in a metal shadowing apparatus Balzers SCD040 (Balzers, Liechtenstein). Images were taken with the TESCAN Vega TS 5136 XM scanning electron microscope (SEM) (Tescan, Czech Republic), using one embryo per stage (E12.5, E13.5, E14.5, E16.5) with the representative phenotype.

Frontal images of E11.5 embryonic heads were taken using a Leica S6D stereoscope with the DFC295 camera (both Leica, Germany). Individual measurements (mx – distance between edges of the maxillary prominences; lnp – distance between edges of the lateral nasal prominences; pits – distance between individual nasal pits) were performed in AxioVision 4.8 software (Zeiss, Germany) using length measurement tool. Distance ratios were calculated as distance between nasal pits÷distance mx, and distance between nasal pits÷distance lnp. Graphs and statistical significance were performed in GraphPad (GraphPad Software, Boston, MA, USA). Measurements were performed in at least four different embryos of all three genotypes (Cdk13^+/+, Cdk13^tm1a/tm1a and Cdk13^tm1d/tm1d).

Mouse embryonic tissues were fixed in 4% PFA overnight. Specimens were then embedded in paraffin and cut in transverse and sagittal planes at 5 µm sections. For immunohistochemistry staining, sections were deparaffinized in xylen and rehydrated in an ethanol series (100%, 96%, 70%). Antigen retrieval was performed either in 1% citrate buffer pH6 or in DAKO antigen retrieval solution pH9 (S1699, DAKO Agilent, USA) at 97.5°C.

For protein localization, we incubated sections with primary antibody (Table S5) for 1 h at room temperature or overnight at 4°C. Antibodies against the following proteins were used: 2H3 (1:50, Nefm, AB_2314897, Developmental Studies Hybridoma Bank), SOX2 (1:100, 2748s, Cell Signaling), SOX9 (1:100, HPA001758, Sigma), Ki67 (1:200, RBK027, Zytomed systems). Then sections were incubated with the following secondary antibodies (1:200) for 30 min at room temperature: anti-mouse Alexa-Fluor 488 (A11001), and anti-rabbit Alexa-Fluor 594 (A11037, both Thermo Fisher Scientific, USA). Ki-67-positive cells were detected by anti-rabbit secondary antibody, ABC binding (PK-6101, Vector laboratories) and followed by application of the DAB (K3468, Dako) chromogenic system.

For DNA staining (nuclei), sections were mounted using Fluoroshield with DAPI (F6057, Sigma, Merck, Germany). If DRAQ5 (62251, Thermo Fisher Scientific, USA) was used for nuclei staining, sections were mounted with Fluoroshield (F6182, Sigma, Merck, Germany). Pictures were taken on confocal microscopes Leica SP8 (Leica, Germany) and Zeiss LSM800 (Zeiss, Germany). Nuclei in DAB-stained sections were counterstained with hematoxilyn. Sections were photographed under bright-field illumination with the Leica DMLB2 compound microscope (Leica, Germany).

Mitotic index on Ki-67-stained sections in the palatal shelves, was counted as a ratio between Ki-67-positive cells (brown) and total number of cells (i.e. Ki-67-negative plus Ki-67-positive cells). Cells were counted independently in mesenchyme and epithelium of the palatal shelves. Cells were counted on four sections (both left and right palatal shelves) in three embryos for each genotype (Cdk13^+/+ and Cdk13^tm1a/tm1a).

MEFs and cells from embryonic DRGs (dorsal root ganglia) were isolated from E12.5 embryonic mice. Tissues were enzymatically processed using Dispase II (D4693, Sigma, Merck, Germany) for 1 h at 37°C while shaking. Cells were then centrifuged, filtered through 40 µm Cell strainer (431750, Corning), seeded on glass inserts and left to grow until 70–80% confluency in DMEM (D6546, Sigma, Merck, Germany). Adult DRG cells were isolated from DRGs from adult mice. DRGs were enzymatically processed using Collagenase IV (LS0004188, PAN Biotech) for 6 h at 37°C. Tissues were resuspended every hour by pipetting. Cells were then filtered through Cell strainer and seeded on glass inserts and left to grow and form long outgrowths in Neurobasal cultivation medium (21103-49, Gibco). Cells were then fixed in 4% PFA for 15 min.

For protein localization, we incubated cells on glass inserts with primary antibody for 1 h at room temperature. The following antibodies were used: anti-CDK13 (1:100, HPA059241, Sigma, Merck, Germany), anti-CDK13 N-TERM (1:100, SAB1302350, Sigma), anti-CDK13 (1:150, PA5-63692, Invitrogen, Thermo Fisher Scientific, USA), anti-F-actin (1:100, A12379, Alexa-Fluor 488™ phalloidin, Thermo Fisher Scientific), anti-2H3 (1:50, Nefm, AB_2314897, Developmental Studies Hybridoma Bank). Then sections were incubated with the following secondary antibodies (1:200) for 30 min at room temperature: anti-mouse Alexa-Fluor 488 (A11001), and anti-rabbit Alexa-Fluor 594 (A11037, both Thermo Fisher Scientific, USA). Glass inserts with cells were mounted on glass slides with Fluoroshield with DAPI (F6057, Sigma, Merck, Germany). Pictures were taken on confocal microscope Leica SP8 (Leica, Germany). Cells were cultivated from at least three different embryos of all three genotypes (Cdk13^+/+, Cdk13^tm1a/tm1a and Cdk13^tm1d/tm1d).

Mouse embryos were dissected, fixed in 4% PFA while rotating at 4°C for 4 h. Embryos were then postfixed in graded methanol dilutions (25%, 50%, 75%, 100%). Embryos were then bleached in a mixture of hydrogen peroxide, DMSO and methanol for 24 h, and postfixed in combination of DMSO and methanol. Embryos were incubated with primary antibody to stain neurofilaments (2H3, AB_2314897, Developmental Studies Hybridoma Bank) for 7 days while rotating, followed by secondary antibody incubation (anti-mouse Alexa-Fluor 488, A11001) for 2 days while rotating. Embryos were finally cleared in a mixture of benzyl benzoate and benzyl alcohol until they got transparent. For microscopy, embryos were placed on Nunc^TM glass bottom dishes (150680, Thermo Fisher Scientific) and imaged by using a Zeiss AxioZoom.V16-Apotome2 (Zeiss, Germany) at CELLIM (Core Facility Cellular Imaging, CEITEC, Masaryk University, Brno, Czech Republic). Whole-mount immunodetection of the neurofilaments was performed in three different embryos of all three genotypes (Cdk13^+/+, Cdk13^tm1a/tm1a and Cdk13^tm1d/tm1d) at E11.5.

MEF and MIH3T3 cells were cultured on a 100 mm culture plate until 75% confluency in a 37°C incubator under 5% CO₂. Cells were washed with ice-cold PBS and collected into 1 ml of ice-cold PBS with a scraper, transferred to an 1.5 ml Eppendorf tube and centrifuged at 300 g and 4°C for 4 min. Cell pellets were resuspended in 200 μl of TMK buffer (25 mM Tris-HCl pH 7.4; 1 mM MgCl₂; 5 mM KCl) and 150 μl TMK+1% NP-40 was added to the tube. Incubation on ice was for 5 min, followed by centrifugation at 250 g and 4°C for 5 min. Supernatant (cytoplasmic fraction) was transferred to the new 1.5 ml Eppendorf tube. Pellet (nuclei) was resuspended in 500 μl of buffer S1 (0.25 M saccharose; 10 mM MgCl₂) and transferred to the new Eppendorf tube on top of 500 μl of buffer S2 (0.35 M saccharose; 0.5 mM MgCl₂), followed by centrifugation at 1400 g and 4°C for 5 min. Supernatant was removed and the pellet was resuspended in 50 μl of ice-cold PBS. Protein concentration was assessed for cytoplasmic and nuclear samples, and an appropriate volume of 3×Laemmli buffer was added to the given samples. The samples were boiled for 5 min at 100°C, followed by western blot procedure as published by Blazek et al. (2011). Antibodies used for western blotting were: CDK13 (Merck, cat. no.: HPA059241), PARP (Cell Signaling Technology, cat. no.: 9542S), α-Tubulin (Cell Signaling Technology, cat. no.: 7291S), Lamin B (Santa Cruz Technology, cat. no.: sc-6217) and GAPDH (Santa Cruz Technology, cat. no.: sc-23233).

We employed the same chromatin immunoprecipitation protocol as published by Blazek et al. (2011), with few modifications: MEF cells were washed with PBS and crosslinked with 1% formaldehyde/PBS solution for 10 min at room temperature. Crosslinking was quenched for 5 min with 125 mM glycine (final concentration). Cells were washed twice with ice-cold PBS and lysed in sonication buffer (0.5% SDS, 20 mM Tris-HCl pH 8.0, 2 mM EDTA, 0.5 mM EGTA, 0.5 mM PMSF, protease inhibitor). For immunoprecipitation, protein extracts were pre-cleared with ChIP Grade G agarose beads (Cell Signaling Technology, cat. no.: 9007S) and then incubated overnight with anti-RNAPII antibody [Cell Signaling Technology, Rpb1 NTD (D8L4Y), cat. no.: 14958S; 1:300], followed by 2 h incubation with ChIP Grade G agarose beads. The beads were washed once with low-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl pH 8.0, 150 mM NaCl) followed by triple washing with the same buffer containing 500 mM NaCl, one wash with lithium buffer (2 mM EDTA, 20 mM Tris pH 8, 250 mM LiCl, 1% NP-40 and 1% natrium deoxycholate) and two washes with TE buffer. We eluted the immunoprecipitates with 1% SDS and 100 mM natrium bicarbonate at room temperature for 15 min and crosslinking was reversed by incubation with 200 mM NaCl for 5 h at 65°C. Proteins were digested with proteinase K (Sigma), and DNA was extracted with phenol:chloroform:isoamyl alcohol (25:24:1) and precipitated with isopropanol overnight. A fraction of precipitated DNA was used in qPCR reactions with the KAPA SYBR FAST qPCR Master Mix (2×) optimized for LightCycler^® 480 (Merck, KK4611). The amplifications were run using the Roche LC480II LightCycler under the following conditions: initial activation step at 94°C for 5 min, followed by 45 cycles at 95°C for 10 s, 60°C for 20 s and 72°C for 15 s. For the Pou4f1 promoter, primers forward (Fw): 5′-AGAAATGCGCTGTGGATGAT-3′ and reverse (Re): 5′-TCCCGAGTAGAAAGCACACA-3′ were used as published by Tang et al. (2020). For the Ntn1 promoter, primers Fw: 5′-GTCGGCAAATTTTCTCCAAA-3′ and Re: 5′-GTTGCATCCTTTCACCCACT-3′ were used as published by Kaneko et al. (2014).

PCR arrays analyses were performed on tissues isolated from embryonic mice, i.e. rostral and caudal parts of palatal shelves at E12.5 and E14.5. One sample was obtained by pooling tissue from two or three embryos, three biological replicates were used for each stage and genotypes were analyzed. Total RNA was extracted using the RNeasy Plus Mini Kit (74136, Qiagen, Germany) according to the manufacturer′s instructions. Total RNA concentration and purity were measured using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, USA). First-strand cDNA was synthesized using the gb Reverse Transcription Kit (3012, Generi Biotech, Czech Republic) according to the manufacturer′s instructions. RT² Profiler™ PCR Array Mouse Neurogenesis (330231, Qiagen, Germany) was performed according to the manufacturer′s instructions on a LightCycler 96 (Roche, Germany). C_T values were exported to an Excel file to create a table of C_T values. This table was then uploaded on to the data analysis web portal at http://www.qiagen.com/geneglobe. Samples were assigned to controls and test groups. C_T values were normalized to Gapdh as a reference gene. The above data analysis web portal calculated fold change/regulation using the delta delta C_T method, in which delta C_T is calculated between gene of interest (GOI) and an average of reference genes (HKG), followed by delta-delta C_T calculations [delta C_T (Test Group)-delta C_T (Control Group)]. Fold Change was then calculated using 2^ (-delta delta C_T) formula. P values were calculated based on a Student's t-test of the replicate 2^(- Delta C_T) values for each gene in control and test groups. The P-value calculation used is based on parametric, unpaired, two-sample equal variance, two-tailed distribution.

Mouse embryos were fixed in 4% PFA and fixation time differed based on the stage. The tissues were then dehydrated in an ethanol series, embedded in paraffin, and 5 µm transverse sections were obtained. The sections were deparaffinized in xylene and dehydrated in 96% ethanol. To detect the expression of certain genes, we used RNA in situ hybridization assay (RNAScope Multiplex Fluorescent v2 Assay kit, 323 110, ACD Bio, USA) in formalin-fixed paraffin-embedded tissues according to the manufacturer's instructions. All reactions, which required incubation at 40°C, were carried out using the HybEZTM II oven (ACD Bio, USA). Probes for Cdk13 (895581), Fgf8 (313411), FoxD1 (495501), Meis2 (436371), Msx1 (421841) and Shh (314361) (all ACD Bio, USA) were used. For negative control staining, a probe diluent (300041; ACD Bio) was used instead of a probe. The hybridized probes were visualized using the TSA-Plus Cyanine 3 system (NEL744001KT, Perkin-Elmer, USA), according to the manufacturer's protocol. DAPI (323 108, ACD Bio, USA) was used to stain nuclei. Pictures were obtained with the Leica SP8 confocal microscope (Leica, Germany).

RNAscope^® Probe Mm-Cdk13 (#895581) binds to nucleotides 3312– 4317 (last couple of exons and the 3′UTR) of mouse Cdk13 (NM_001081058.2). If deletion of exons 3 and 4 (Cdk13^tm1d/tm1d) is unstable, it is possible to have a signal (some weak signal in the form of dots is present in both hypomorphic and knockout embryonic palatal shelves).

Rostral and caudal parts of the palatal shelves at E12.5, E14.5 and E16.5, as well as maxillary (mx), mandibular (md) and frontonasal prominences (fnp) at E11.5 and E12.5 were dissected from WT mouse embryos to quantify the differences in Cdk13 expression during development of the facial structures. Individual parts of the palatal shelves and facial prominences were dissected from at least three different embryos for each stage. Proliferation rate changes in the Cdk13-deficient embryos were assessed using CyclinD1 gene expression in tissues isolated from rostral and caudal palatal shelves (Cdk13^tm1a E12.5, E14.5) and lips (Cdk13^tm1d E12.5). Total RNA was extracted using the RNeasy Plus Mini Kit (74136, Qiagen, Germany) according to the manufacturer's instructions. Total RNA concentration and purity was measured using a NanoDrop One (Thermo Fisher Scientific, USA). First-strand cDNA was synthesized using gb Reverse Transcription Kit (3012, Generi Biotech, Czech Republic) according to the manufacturer's instructions. TaqMan probe was used to quantify Cdk13 (Mm01164725_m1, Thermo Fisher Scientific) and CyclinD1 (Mm00432359_m1, Thermo Fisher Scientific) gene expression. The RT-PCR reaction was performed on LightCycler 96 (Roche, Germany). The comparative C_T method was used for analysis.

Fertilized Gallus gallus (chicken) eggs were bought at a farm (Integra, Zabcice, Czech Republic) and incubated in a humidified incubator at 37°C. At Hamburger–Hamilton (HH) developmental stage 20 (HH20), 1 mM CDK12/13 inhibitor THZ531 (SML2619, Sigma Aldrich, Merck, Germany) was injected into the maxillary prominences (Fig. 8) using a glass capillary attached to an Eppendorf FemtoJet 4i Microinjector (Eppendorf, Germany) supplied with a manipulator (Leica, Germany). Moreover, we injected the inhibitor into the craniofacial mesenchyme and the postoptic region of chicken embryos at HH10 and HH15, respectively (Fig. 8). 10% Trypan Blue was used as a contrasting dye. After injections, embryos were incubated for another 96 h, and then killed and fixed in 4% paraformaldehyde. Frontal and lateral images were taken using a Leica S6D stereoscope with a DFC295 camera (both Leica, Germany).

The MTT test was used to establish the toxicity of the CDK12/13 inhibitor THZ531. Cells isolated from the trigeminal ganglion (TG) were cultivated for 4 days in 100 μl of DMEM supplemented with 10% FBS, 1% penicilin/streptomycin, 1% L-glutamine in a 96-well plate (100,000 cells/ml). Cells were treated every day with different concentrations of THZ531 (i.e. 50, 100, 300, 500, 750, 1000 and 1500 nM). Untreated cells were used as controls. After 4 days, medium was discarded and replaced with 50 μl of DMEM without FBS followed by addition of 50 μl of MTT Reagent (MTT Cell Proliferation Kit Ab211091, Abcam), 50 μl of common culture medium and 50 μl of MTT Reagent were added to empty wells and used as background controls. The plate was then wrapped in foil and incubated for 3.5 h at 37°C under 5% CO₂. Afterwards, 150 μl of MTT solvent (MTT Cell Proliferation Kit Ab211091, Abcam) was added to the samples. Absorbance was measured at 590 nm and cell survival rates were established as the absorbance index, with higher absorbance meaning better cell survival, lower absorbance meaning worse cell survival (Fig. S7H).

TGs were dissected from mouse embryos at E12 and washed in ice-cold Neurobasal medium (21103049, Gibco, Thermo Fisher Scientific, USA). Each TG was placed in a separate well on a culture well plate into 10 μl drop of Matrigel (356231, Corning, USA) and left 20 min to polymerize in the tissue culture incubator. Neurobasal medium supplemented with 100 nM or 300 nM THZ531 (SML2619, Sigma Aldrich, Merck, Germany) was added. Images were taken after adding cultivation medium (0 h) and again after 24 h by using an Olympus IX71 inverted microscope (Olympus, Japan). Neurite outgrowth was measured using ImageJ software (NIH, USA) while comparing areas covered by neurites on pictures taken under different culture conditions.

Changes of the ability of TGs to produce neurite outgrowths (in %) were calculated as following: neurite outgrowth area (total area−TG area)/total area×100.

Apoptotic cells were detected by using the TUNEL assay (ApopTag Peroxidase In Situ Apoptosis Detection Kit, cat. no. S7101, Chemicon, Temecula, USA). Nuclei were counterstained with hematoxylin. Sections were photographed under bright-field illumination with a Leica DMLB2 compound microscope (Leica, Germany). TUNEL assays were performed using three different embryos of the Cdk13^+/+ and Cdk13^tm1a/tm1a genotypes.

Data were evaluated for statistical significance in GraphPad (GraphPad Software, Boston, MA, USA) using unpaired two-tailed Student's t-tests. Differences were considered to be significant at *P<0.05, **P<0.01 and **P<0.001. Errors are indicated as the ±standard error of the mean (±s.e.m.).

We acknowledge the core facility CELLIM supported by the Czech-BioImaging large RI project (LM2018129 funded by MEYS CR).