Epilepsy kinase CDKL5 is a DNA damage sensor which controls transcriptional activity at DNA breaks
==================================================================================================

* Taran Khanam
* Ivan Muñoz
* Florian Weiland
* Thomas Carroll
* Barbara N Borsos
* Vasiliki Pantazi
* Meghan Slean
* Miroslav Novak
* Rachel Toth
* Paul Appleton
* Tibor Pankotai
* Houjiang Zhou
* John Rouse

## Abstract

Mutation of the *CDKL5* kinase gene leads to the seizure-prone neurodevelopmental condition CDD (CDKL5 deficiency disorder) and is the most common genetic cause of childhood epilepsy. However, the phospho-targets and roles of CDKL5 are poorly understood, especially in the nucleus. We reveal CDKL5 as a sensor of DNA damage in actively transcribed regions of the nucleus, which phosphorylates transcriptional regulators such as Elongin A (ELOA) on a specific consensus motif. Recruitment of CDKL5 and ELOA to DNA damage sites, and subsequent ELOA phosphorylation, requires both active transcription and synthesis of poly–ADP ribose to which CDKL5 can bind. Critically, CDKL5 is essential for transcriptional control at DNA breaks. Therefore, CDKL5 is a DNA damage-sensing regulator of transcription, with implications for CDKL5-related human diseases.

**One sentence summary** CDKL5 kinase phosphorylates transcriptional regulators and modulates the activity of damaged transcriptional units

The protein kinases ATM, DNA-PK and ATR sense and transduce DNA damage signals, triggering a pleiotropic series of protective reactions collectively known as the DNA damage response (DDR) which prevents genome instability and disease (*1*). With the aim of expanding the repertoire of DDR kinases, we set out to find other kinases that can sense DNA damage. U–2–OS cells stably expressing mCherry–FAN1 which marks DNA damage sites, were transfected with GFP–tagged kinases individually, starting with the CMGC branch of the human kinome (Fig. S1A). BrdU–sensitized cells were laser micro-irradiated to induce a pleiotropic range of DNA lesions along a track in the nucleus. This approach revealed CDKL5 as a DNA damage sensing kinase (Fig. 1A, S1A). *CDKL5* mutations are one of the most common genetic causes of epilepsy in children (*2*), and they can lead to the severe, seizure prone neurodevelopmental disorder CDD (*3*), as well as milder syndromes (*4*). In the cytosol, CDKL5 phosphorylates microtubule regulators (*5, 6*), but the nuclear roles and targets of CDKL5 remain elusive. This prompted us to investigate how CDKL5 recognizes DNA damage, to find its nuclear targets and explore roles in DDR.

![Figure 1.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2020/12/12/2020.12.10.419747/F1.medium.gif)

[Figure 1.](http://biorxiv.org/content/early/2020/12/12/2020.12.10.419747/F1)

Figure 1. CDKL5 senses DNA damage in actively transcribed regions
**A.** BrdU–sensitized U–2–OS Flp–In T–REx stably expressing mCherry-FAN1 and GFP–NLS or GFP–CDKL5 (no NLS) were line micro-irradiated and imaged after 2min. **B.** Same as A. except that cells stably expressing GFP–NLS–CDKL5 were pre–incubated with DMSO (mock), olaparib (5 μM), talazoparib (50 nM) or PDD00017273 (0.3 μM) for 1 h prior to micro–irradiation. One of three independent experiments is shown. **C.** Same as B. except that cells were spot micro–irradiated (405 nm). **D.** Quantitation of spot intensity in C. Data represent the mean ± SEM of two independent experiments; > 50 micro–irradiated cells per point. **E.** Cells subjected to the workflow in Fig. S1C were detergent–extracted and fixed before staining with anti–GFP or fibrillarin (nucleoli). **F.** Quantification of the detergent–insoluble GFP–NLS–CDKL5 signal (minus nucleolar signal). The mean ± SD from three biological experiments is shown. Statistical significance was assessed by one-way-ANOVA-test. Asterisks ** indicate *P*–value of <0.01; ns – not significant. **G, H.** Effect of transcription inhibitors on CDKL5 recruitment after spot micro-irradiation. **I.** Stable cell lines were permeabilized and incubated with RNase A or PBS before irradiation and imaging.

CDKL5 recruitment to sites of micro–irradiation was rapid and transient (Figs. 1A–D, Movies S1, S2), reminiscent of proteins that bind poly–ADP ribose (PAR) generated by DNA damage–activated poly–ADP ribose polymerases (PARPs) (*7*). Accordingly, CDKL5 recruitment was blocked by PARP inhibitors olaparib and talazoparib (Figs. 1B–D) or by *PARP1* disruption (Fig. S1B), but prolonged by PDD00017273, an inhibitor of PARG (poly–ADP ribose glycohydrolase) which delays PAR degradation (Fig. 1B–D; Movies S1, S2) (*8*). CDKL5 was also retained on damaged chromatin after exposure to H2O2, an inducer of DNA breaks (Figs. 1E, F; S1C). Retention was prevented by olaparib, whereas nucleolar retention seen in undamaged cells was unaffected (Figs. 1E, F). Together, these data indicate that CDKL5 recruitment to DNA breaks requires synthesis of PAR, presumably by direct PAR binding although no PAR-binding motifs were detected bioinformatically. However, a series of N-terminal and C-terminal deletion constructs revealed a region between amino acids 530 and 730 necessary for CDKL5 recruitment (Figs. S2A, B), and the region 530–680 of CDKL5 was sufficient for recruitment (Figs. S2C, D). As shown in Figs. S2D–F, recombinant CDKL5 fragments corresponding to this region bound to PAR *in vitro* (Fig. S2E, F), strongly suggesting that CDKL5 senses DNA breaks by binding PAR. Accordingly, PAR was detected in CDKL5 precipitates, and vice versa, after exposure of cells to H2O2 (Figs. S2G,H).

These data show that PAR formation is required for CDKL5 recruitment, but we discovered unexpectedly that it is not sufficient. We found that the transcription inhibitors actinomycin D or α–amanitin which inhibit RNA polymerases (RNAPs) I and II, or DRB which blocks elongating RNAP II, abrogated the recruitment of CDKL5 to micro-irradiation sites. PAR synthesis and recruitment of the PAR-binding protein XRCC1 or FAN1 was unaffected (Figs 1G,H; S3A–D). Moreover, incubation of permeabilized cells with RNaseA abolished micro-irradiation tracks formed by CDKL5, but not XRCC1 or FAN1 (Fig. 1I). Therefore, CDKL5 is recruited to DNA breaks at sites of active transcription.

The data above suggested that nuclear targets of CDKL5 may be involved in transcriptional control. To identify nuclear CDKL5 targets specifically, we expressed CDKL5 wild–type (WT) or a K42R kinase–dead (KD) mutant (*5*) exclusively in the nucleus of *CKDL5*-disrupted U–2–OS cells by adding an artificial nuclear localization signal (NLS) (Fig. S4A–C). We next compared the phosphoproteome of the two cell populations, after exposure to H2O2 to induce CDKL5 retention at DNA breaks (Fig. 2A, S4D, Table S1). As well as CDKL5 itself and MAP1S and EB2 (known cytosolic substrates), screening revealed a range of proteins bearing phospho–sites that were higher in abundance in CDKL5NLS–WT cells compared with CDKL5NLS–KD cells (>1.5–fold, p<0.05) (Figs. 2B, C; Table S1). Strikingly, the phospho–acceptor [S/T] residue in almost all of the putative nuclear CDKL5 substrates lies in sequence R–P–X–[**S/T**]– [A/G/P/S] (Fig. 2C), which represents a prerequisite consensus motif for CDKL5 target phosphorylation, in agreement with previous work (*5, 6*). Gene ontology (GO) analysis showed a striking enrichment of transcription regulators (Figs. 2D, E) including EP400, a chromatin-remodeling transcriptional activator (*9*), and TTDN1, mutated in a form of tricothiodystrophy (TTD), typically caused by failure in transcription–coupled DNA repair (*10*). Elongin A (ELOA), a transcriptional elongation factor and component of an E3 ligase complex which ubiquitylates RNAPII, was a top hit (*11*).

![Figure 2](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2020/12/12/2020.12.10.419747/F2.medium.gif)

[Figure 2](http://biorxiv.org/content/early/2020/12/12/2020.12.10.419747/F2)

Figure 2 CDKL5 phosphorylates transcriptional regulators including ELOA
**A.** Quantitative phosphoproteomics workflow. **B.** “Sprinkler” plot of the mass spectrometry data from the experiment in A (see Table S1). **C.** List of proteins containing phosphorylation sites more abundant in cells expressing CDKL5NLS–WT versus KD (fold change > 1.5; p-site probabilities > 0.6). **D.** Protein–protein interaction network of putative CDKL5 substrates from Table S1. Confidence levels are based on the STRING database v11.0 combined score with following bins: 150–400: low confidence (blue), 400–700: medium confidence (gold), 700–900: high confidence (not encountered in this dataset), >900: very high confidence (black). P–value was calculated as 0.00068. **E.** Analysis of GO terms. Significance cut–off was set as α = 0.01 with at least 2 proteins identified in the respective group.

Extracted ion chromatogram (XIC) analysis of phospho–peptides isolated from FLAG-tagged EP400 (pSer729), ELOA (pSer311) and TTDN1 (pSer40) confirmed CDKL5-dependent phosphorylation of these proteins in cells (Fig. S5A-C). Furthermore, CDKL5 robustly phosphorylated synthetic peptides corresponding to EP400 Ser729 and ELOA Ser311 demonstrating direct phosphorylation (Fig. S5D). To further investigate ELOA phosphorylation, we generated antibodies specific for phospho-Ser311. Co-expression with WT, but not KD, CDKL5 markedly increased Ser311 phosphorylation of FLAG-ELOA, but not an ELOA Ser311Ala mutant (Fig. 3A). Strikingly, CDD–associated CDKL5 mutations, which are located predominantly in the kinase catalytic domain (*12*), severely reduced CDKL5 activity towards ELOA pSer311, whereas a series of benign variants did not (Fig. 3B).

![Fig. 3.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2020/12/12/2020.12.10.419747/F3.medium.gif)

[Fig. 3.](http://biorxiv.org/content/early/2020/12/12/2020.12.10.419747/F3)

Fig. 3. Phosphorylation of ELOA Ser311 on damaged chromatin by CDKL5
**A.** HEK293 cells were co-transfected with CDKL5 (wild type “WT” or kinase-dead “KD” K42R mutant) fused to an NLS, and FLAG-ELOA (wild type “WT” or a S311A mutant “SA”). Anti-FLAG precipitates or cell extracts were probed with the antibodies indicated. One of three independent experiments is shown. **B.** Same as A. showing a range of pathogenic (red) and benign (blue) CDKL5 variants. **C–E**. Wild type (WT), CDKL5–disrupted (CDKL5Δ/Δ) or siRNA–transfected cells were subjected to indirect immunofluorescence analysis with the indicated antibodies at laser tracks. Quantification of ELOA-pSer311 signal at the laser tracks is shown. Data represent mean ± SD. Statistical significance was assessed by one-way-ANOVA-test. Asterisks \**\*|\* indicate *P*–values of <0.0001.

CDKL5-dependent phosphorylation of endogenous ELOA Ser311 was evident at sites of DNA damage. Signal intensity was reduced by depletion of ELOA (Fig. 3C) or by incubation of cells with lambda-phosphatase or the ELOA Ser311 phosphopeptide antigen (Fig. S6A), thereby confirming antibody specificity. ELOA phosphorylation was reduced by disruption or depletion of *CDKL5* (Figs. 3D, E), or by olaparib or DRB which block CDKL5 recruitment (Fig. 3D). We wondered if ELOA is recruited to DNA damage sites by a similar mechanism to the ELOA kinase. In agreement with this idea ELOA recruitment to micro-irradiation tracks was rapid, transient, and inhibited by olaparib, α-amanitin and DRB (Fig. S6B–D). Similar results were obtained for other nuclear CDKL5 substrates such as ZNF592 and ZAP3 (Fig. S6B-D) but not EP400 (data not shown). These data reveal CDKL5-dependent phosphorylation of substrates such as ELOA at DNA damage sites, involving a common mechanism of recruitment of both kinase and substrate.

The ontological enrichment for transcription regulators among the nuclear CDKL5 substrates suggested a role in transcriptional control at DNA damage sites. Breaks in genomic DNA silence adjacent genes (*13–15*), and we tested a role for CDKL5, first using a reporter system in which a cluster of FokI nuclease– mediated DSB, induced upstream of a doxycycline inducible reporter gene, silences the reporter cassette (Fig. S7A) (*13*). Depletion of CDKL5 weakens silencing of the reporter cassette, similar to depletion of ATM or ZMYND8 (Fig. 4A; Figs. S7A–C) (*13, 14*). Second, we took advantage of a system where inducible overexpression of the site-specific meganuclease I-PpoI that cuts 14– 30 times in the human genome, results in silencing of genes that sustain DSBs within or nearby. As shown in Fig. 4B, CDKL5 depletion largely abolished the I– PpoI–induced silencing of *SLCO5a1* and *RYR2* genes reported previously (*15*).

![Fig. 4.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2020/12/12/2020.12.10.419747/F4.medium.gif)

[Fig. 4.](http://biorxiv.org/content/early/2020/12/12/2020.12.10.419747/F4)

Fig. 4. CDKL5 facilitates transcriptional silencing at DNA breaks
**A.** U–2–OS cells (263 IFII; Fig. S7A) were transfected with the siRNAs indicated. After addition of doxycycline, transcription was monitored in cells ± induction of FokI by quantification of YFP(–MS2) foci (*left*). 150 cells were analysed per condition per experiment. The mean ± SD from four independent experiments is shown. (*Right*) Quantitative RT–PCR analysis of YFP-MS2 mRNA. Data represent mean ± SD. **B**. Quantitative PCR with reverse transcription (qRT-PCR) analysis of *SLCO5a1* and *RYR2* expression levels (left) and cutting efficiencies (right) in U–2–OS HA-ER-I-PpoI cells depleted of CDKL5, at the times indicated after inducing I–PpoI. The mean ± SD from qPCR replicates of three independent experiments is shown Statistical significance for all the data was assessed by two-way-ANOVA-test **p<0.01, \***|p<0.001 and \**\*|\*p<0.0001; ns – not significant. **C.** Schematic diagram: CDKL5 functions in nucleus and cytosol.

Our study reveals that CDKL5 senses DNA breaks at sites of ongoing transcription, through binding to PAR, and switches off transcription nearby (Fig. 4C). How transcriptional activity at DSB is sensed by CDKL5 remains to be investigated. Once bound to damaged DNA, CDKL5 phosphorylates transcriptional regulators including ELOA, and facilitates the silencing of genes harboring DNA breaks; it is likely that phosphorylation of multiple CDKL5 substrates participates in silencing. For example, CDKL5 phosphorylation of ELOA could influence transcriptional elongation rates, but this remains to be ascertained. CDKL5 was recently linked to transcriptional control in a different context when it was reported to promote renal injury in mice exposed to toxic insults through up-regulating SOX9-dependent genes (*16*). We are interested in the possibility that CDKL5 controls transcription even in the absence of toxic insult, perhaps by sensing the transient DSBs induced by topoisomerases which are known to regulate transcription and impact on brain function (*17*). This will be interesting to investigate, especially as it could be particularly relevant to the pathogenesis of epilepsy and CDD.

## Supporting information

Supplementary Figure legends, Experimental procedures, supplementary references [[supplements/419747_file02.docx]](pending:yes)

Figure S1 [[supplements/419747_file03.jpg]](pending:yes)

Figure S2 [[supplements/419747_file04.jpg]](pending:yes)

Figure S3 [[supplements/419747_file05.jpg]](pending:yes)

Figure S4 [[supplements/419747_file06.jpg]](pending:yes)

Figure S5 [[supplements/419747_file07.jpg]](pending:yes)

Figure S6 [[supplements/419747_file08.jpg]](pending:yes)

Figure S7 [[supplements/419747_file09.jpg]](pending:yes)

Movie S1 [[supplements/419747_file10.avi]](pending:yes)

Movie S2 [[supplements/419747_file11.avi]](pending:yes)

Table S1 [[supplements/419747_file12.xlsx]](pending:yes)

Table S2 [[supplements/419747_file13.xlsx]](pending:yes)

## Acknowledgements

We thank the technical support of the MRC-PPU including the DNA Sequencing Service, Tissue Culture team, Reagents and Services team, and the PPU Mass-Spectrometry team. We also thank Fiona Brown and James Hastie for ELOA antibody production and purification. We’re grateful to Jessica Downs and Roger Greenberg for the U–2–OS cells harbouring the FokI silencing reporter, to Evi Soutoglou for U–2–OS HA-ER-I-PpoI cells, and to Nick Lakin for *PARP*-disrupted U–2–OS cells. We’re grateful to Graeme Ball for help with microscopy data analysis. We thank Luis Sanchez-Pulido and Chris Ponting for help with bioinformatic analyses. We are grateful to members of the Rouse lab for useful discussions. This work was supported by the Medical Research Council (grant number MC_UU_12016/1; TK, IM, JR) and the pharmaceutical companies supporting the Division of Signal Transduction Therapy Unit (Boehringer-Ingelheim, GlaxoSmithKline, and Merck KGaA).

## Footnotes

*   Author affiliations updated and ORCID included for all the authors. Middle name 'N' added for the author Barbara Borsos.It is now Barbara N Borsos Supplemental files updated with added references. There are now 30 references in the revised version A url added for Validation TMT https://doi.org/10.5281/zenodo.4311494 

*   [https://doi.org/10.5281/zenodo.4311475](https://doi.org/10.5281/zenodo.4311475) 

*   [https://doi.org/10.5281/zenodo.4311494](https://doi.org/10.5281/zenodo.4311494) 

*   [http://www.ebi.ac.uk/pride/archive/projects/PXD022975](http://www.ebi.ac.uk/pride/archive/projects/PXD022975) 

*   [http://www.ebi.ac.uk/pride/archive/projects/PXD022916](http://www.ebi.ac.uk/pride/archive/projects/PXD022916) 

*   Received December 10, 2020.
*   Revision received December 12, 2020.
*   Accepted December 12, 2020.


*   © 2020, Posted by Cold Spring Harbor Laboratory

The copyright holder for this pre-print is the author. All rights reserved. The material may not be redistributed, re-used or adapted without the author's permission.

## References

1.  1. A. N. Blackford,  S. P. Jackson, ATM, ATR, and DNA-PK: The Trinity at the Heart of the DNA Damage Response. Mol Cell 66, 801–817 (2017).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1016/j.molcel.2017.05.015&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=28622525&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom) 

2.  2. J. D. Symonds et al., Incidence and phenotypes of childhood-onset genetic epilepsies: a prospective population-based national cohort. Brain 142, 2303–2318 (2019).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1093/brain/awz195&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom) 

3.  3. S. Fehr et al., The CDKL5 disorder is an independent clinical entity associated with early-onset encephalopathy. Eur J Hum Genet 21, 266–273 (2013).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1038/ejhg.2012.156&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=22872100&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom) 

4.  4. C. I. MacKay et al., Expanding the phenotype of the CDKL5 deficiency disorder: Are seizures mandatory? Am J Med Genet A, (2020).
    
    

5.  5. I. M. Munoz et al., Phosphoproteomic screening identifies physiological substrates of the CDKL5 kinase. EMBO J 37, (2018).
    
    

6.  6. L. L. Baltussen et al., Chemical genetic identification of CDKL5 substrates reveals its role in neuronal microtubule dynamics. EMBO J 37, (2018).
    
    

7.  7. T. Eisemann,  J. M. Pascal, Poly(ADP-ribose) polymerase enzymes and the maintenance of genome integrity. Cell Mol Life Sci 77, 19–33 (2020).
    
    

8.  8. D. I. James et al., First-in-Class Chemical Probes against Poly(ADP-ribose) Glycohydrolase (PARG) Inhibit DNA Repair with Differential Pharmacology to Olaparib. ACS Chem Biol 11, 3179–3190 (2016).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1021/acschembio.6b00609&link_type=DOI) 

9.  9. S. K. Pradhan et al., EP400 Deposits H3.3 into Promoters and Enhancers during Gene Activation. Mol Cell 61, 27–38 (2016).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1016/j.molcel.2015.10.039&link_type=DOI) 

10. 10. E. R. Heller et al., Mutations in the TTDN1 gene are associated with a distinct trichothiodystrophy phenotype. J Invest Dermatol 135, 734–741 (2015).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1038/jid.2014.440&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=25290684&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom) 

11. 11. R. C. Conaway,  J. W. Conaway, The hunt for RNA polymerase II elongation factors: a historical perspective. Nat Struct Mol Biol 26, 771–776 (2019).
    
    

12. 12. R. Krishnaraj,  G. Ho,  J. Christodoulou, RettBASE: Rett syndrome database update. Hum Mutat 38, 922–931 (2017).
    
    

13. 13. N. M. Shanbhag,  I. U. Rafalska-Metcalf,  C. Balane-Bolivar,  S. M. Janicki,  R. A. Greenberg, ATM-dependent chromatin changes silence transcription in cis to DNA double-strand breaks. Cell 141, 970–981 (2010).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1016/j.cell.2010.04.038&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=20550933&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=000278618800012&link_type=ISI) 

14. 14. F. Gong et al., Screen identifies bromodomain protein ZMYND8 in chromatin recognition of transcription-associated DNA damage that promotes homologous recombination. Genes Dev 29, 197–211 (2015).
    
    [Abstract/FREE Full Text](http://biorxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiZ2VuZXNkZXYiO3M6NToicmVzaWQiO3M6ODoiMjkvMi8xOTciO3M6NDoiYXRvbSI7czo0ODoiL2Jpb3J4aXYvZWFybHkvMjAyMC8xMi8xMi8yMDIwLjEyLjEwLjQxOTc0Ny5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

15. 15. T. Pankotai,  C. Bonhomme,  D. Chen,  E. Soutoglou, DNAPKcs-dependent arrest of RNA polymerase II transcription in the presence of DNA breaks. Nat Struct Mol Biol 19, 276–282 (2012).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1038/nsmb.2224&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=22343725&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom) 

16. 16. J. Y. Kim et al., A kinome-wide screen identifies a CDKL5-SOX9 regulatory axis in epithelial cell death and kidney injury. Nat Commun 11, 1924 (2020).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1038/s41467-020-15638-6&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=32317630&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom) 

17. 17. P. J. McKinnon, Topoisomerases and the regulation of neural function. Nat Rev Neurosci 17, 673–679 (2016).
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1038/nrn.2016.101&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=27630045&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2020%2F12%2F12%2F2020.12.10.419747.atom)