Research ArticlePain

Cyclin-dependent–like kinase 5 is required for pain signaling in human sensory neurons and mouse models

See allHide authors and affiliations

Science Translational Medicine  08 Jul 2020:
Vol. 12, Issue 551, eaax4846
DOI: 10.1126/scitranslmed.aax4846

A painful mechanism in CDD

CDKL5 deficiency disorder (CDD) is a developmental encephalopathy caused by mutations in the cyclin-dependent–like kinase 5 (CDKL5) gene. Now, La Montanara et al. found that patients with CDD, in addition to the known symptoms, had altered pain sensitivity. Mechanistically, the authors found that CDKL5 is expressed in sensory neurons and regulates CaMKII-dependent TRPV1 signaling, thus affecting pain perception. In rodents, Cdkl5 deletion in sensory neurons resulted in reduced pain sensitivity. The results suggest that restoration of pain perception should be considered when developing therapies for treating CDD.


Cyclin-dependent–like kinase 5 (CDKL5) gene mutations lead to an X-linked disorder that is characterized by infantile epileptic encephalopathy, developmental delay, and hypotonia. However, we found that a substantial percentage of these patients also report a previously unrecognized anamnestic deficiency in pain perception. Consistent with a role in nociception, we found that CDKL5 is expressed selectively in nociceptive dorsal root ganglia (DRG) neurons in mice and in induced pluripotent stem cell (iPS)–derived human nociceptors. CDKL5-deficient mice display defective epidermal innervation, and conditional deletion of CDKL5 in DRG sensory neurons impairs nociception, phenocopying CDKL5 deficiency disorder in patients. Mechanistically, CDKL5 interacts with calcium/calmodulin-dependent protein kinase II α (CaMKIIα) to control outgrowth and transient receptor potential cation channel subfamily V member 1 (TRPV1)–dependent signaling, which are disrupted in both CDKL5 mutant murine DRG and human iPS–derived nociceptors. Together, these findings unveil a previously unrecognized role for CDKL5 in nociception, proposing an original regulatory mechanism for pain perception with implications for future therapeutics in CDKL5 deficiency disorder.


Mutations in the X-linked CDKL5 gene, encoding the cyclin-dependent–like kinase 5 (CDKL5), are associated with the CDKL5 deficiency disorder (CDD) that is characterized by early-onset intractable seizures, severe intellectual disability, and motor impairments (1, 2). CDKL5 is strongly expressed in the central nervous system (CNS), particularly in neurons of the cortex and of the hippocampus (3). Defective CDKL5 impairs proper brain development, learning and memory, and neuronal activity, including molecular processes related to neuronal depolarization and synapse formation (47). CDKL5 is activated by neuronal activity and it has also been implicated both in the phosphorylation and interaction with nuclear as well as cytoplasmic proteins potentially modulating gene expression and cytoskeleton changes ultimately affecting neuronal activity and cell survival (2, 5, 8, 9). Recently microtubule and centrosome-associated proteins have been identified as targets of CDKL5 kinase activity, including microtubule associated protein 1S (MAP1S), microtubule end-binding2 (EB2), Rho/Rac guanine nucleotide exchange factor 2 (ARHGEF2), Centrosomal Protein 131 (CEP131), and disks large homolog 5 (DLG5), whose regulation is likely to affect axonal transport, outgrowth, and synaptic plasticity (10, 11).

CDD remains a disease without a cure, whose fundamental molecular and cellular mechanisms are still in need of much investigation and whose full clinical spectrum remains only partially characterized. An important current limitation is that, unlike patients, CDD mutant mice do not develop epilepsy (4, 1214). This makes the identification of CDKL5-dependent mechanisms or potential treatments with an impact on the human phenotype very challenging. Here, we found that patients with CDD and animal models display an impairment in nociception and that CDKL5 is required for peripheral nociceptive signaling in dorsal root ganglia (DRG) neurons in mice and in induced pluripotent stem cell (iPS)–derived human nociceptors. Mechanistically, conditional deletion of Cdkl5 in DRG sensory neurons impairs nociception, phenocopying CDD in patients. Last, we showed that CDKL5 interacts with calcium/calmodulin-dependent protein kinase II α (CaMKIIα) to control outgrowth of sensory neurons and TRPV1-dependent signaling including in iPS–derived neurons from patients with CDD. Together, these findings reveal a previously unrecognized role for CDKL5 in nociception, and they suggest an original peripheral regulatory mechanism for pain processing with implications for future therapeutic interventions in CDD.


Pain perception is altered in CDD

In a recent effort to broadly characterize the clinical phenotype, variation, and natural history of CDD, the International CDKL5 Database (ICDD) ( was established in 2012. Included in the database are responses to questions given to caregivers about whether their child experiences alterations in pain perception. Analysis of this clinical data source allowed the identification of a previously unrecognized occurrence of alterations in pain perception.

Patients in the ICDD were classified by age group, gender, and mutation type (Table 1). Classification of individual CDKL5 mutations was based on predicted structural and functional phenotypic consequences similar to the groupings used in a previous study (15). In our analysis, we have grouped together mutations leading to lack of functional protein and missense/in-frame mutations within catalytic domain and Thr-Glu-Tyr (TEY) motif and compared them with mutations affecting the regulatory domains of CDKL5 such as truncations after aa172 (amino acid).

Table 1 Associations between pain perception, gender, age group, and mutation type in 202 individuals with confirmed CDKL5 mutation.

Normal, individuals with normal pain sensitivity; reduced or enhanced, individuals with reduced or enhanced pain sensitivity. n, number of individuals; RR, risk ratio; RRR, relative risk ratio; CI, confidence interval; Ref, reference category. Note: excluded individuals where both reduced and enhanced sensitivity were reported.

View this table:

Multinomial logistic regression was used to estimate the risk of altered pain sensitivity relative to no change in pain sensitivity (reduced pain sensitivity versus normal pain sensitivity) in the overall sample, as well as in each category of age group, gender, and mutation type (within category comparison) (Table 1). We then used the same model to evaluate the relative risk ratios (RRRs) of altered pain sensitivity in each category compared to the reference category (age group, 0 to2 years; gender, female; mutation type, truncations after aa172; between category comparison) (Table 1). Point estimates and their 95% confidence interval (CI) were reported.

We found that altered pain sensitivity in their children was reported by 53.0% (122 of 230) of caregivers. Among these, a total of 57.4% (70 of 122) specifically reported decreased pain sensitivity, 19.7% (24 of 122) specifically reported enhanced sensitivity, and 22.9% (28 of 122) reported both. Two hundred two caregivers provided specific responses including either reduced (70 of 202, 34.7%) or enhanced pain perception (24 of 202, 11.9%) (Table 1). Among those who reported either reduced or enhanced pain sensitivity, a total of 74.4% (70 of 94) reported decreased pain perception and 25.5% (24 of 94) reported increased pain perception. The likelihood of the response being specifically reduced pain perception was measured as risk ratio (RR) of reduced or enhanced pain sensitivity versus normal pain perception, whereas the RRR was used to represent the relative change in the RR in each category compared to the reference. As shown in Table 1, reduced pain perception was more likely to be reported than enhanced pain perception (RR 2.92, 95% CI 2.89, 7.00; P < 0.001). Compared with those aged 2 years or under, individuals aged 6 years and over were reported to have an increased likelihood of reduced sensitivity to pain (RRR 2.74, 95% CI 1.36, 5.53; P = 0.005) (Table 1). Compared with those with mutations in the CDKL5 regulatory domain (truncations after aa172), those with a nonfunctional protein (including missense/in-frame mutations within catalytic domain and TEY motif) did not show an increased likelihood of reduced sensitivity to pain (RRR 1.63, 95% CI 0.87, 3.06; P = 0.124) (Table 1). Similarly, we did not identify any differences by gender (RRR 1.03, 95% CI 0.44, 2.45; P = 0.941) (Table 1). Reduced was significantly more likely than enhanced pain perception at each age group [0 to 2 years old: RR 3.00, 95% CI 1.28, 7.06 (P = 0.012); 3 to 5 years old: RR 2.67, 95% CI 1.04, 6.81 (P = 0.040); 6+ years old: RR 3.00, 95% CI 1.52, 5.94 (P = 0.002)], for each gene mutation group studied [no functional protein: RR 2.94, 95% CI 1.67, 5.18 (P < 0.001); truncation after aa172: RR 2.88, 95% CI 1.29, 6.42 (P = 0.010)], and for females (RR 3.16, 95% CI 1.89, 5.29; P < 0.001) (Table 1).

CDKL5 localizes to human and murine sensory neurons and is required for nociception

Because primary nociceptors are localized in the DRG, we initially investigated whether CDKL5, so far localized to the CNS, was also expressed in the peripheral nervous system in murine and human nociceptors. Immuno and coimmunohistochemistry studies in DRG revealed that CDKL5 is expressed and mainly found in the cytoplasm of small-diameter DRG neurons (Fig. 1, A to C), where the signal becomes detectable after the age of postnatal day 45 (P45) (fig. S1, A and B). Around 90% of CDKL5-positive neurons express the nociceptive markers calcitonin gene-related peptide (CGRP) or isolectin B4 (IB4), whereas only a small percentage express parvalbumin or neurofilament 200 (NF200) (Fig. 1, D to F).

Fig. 1 CDKL5 is expressed in nociceptors.

(A) CDKL5 immunostaining in DRG from WT and Cdkl5−/y mice (20×). L-I-S, insets show high magnification of large-, intermediate-, and small-diameter DRG neurons. (B) Graphs showing ratio of signal intensity between WT and Cdkl5−/y three neuronal subpopulations. n = 3 mice each, mean with SD, Student’s t test, **P < 0.01 and *P < 0.05. (C) Graphs showing percentage of CDKL5+ neurons in the indicated sizes. n = 3 mice each. (D) Confocal immunofluorescence images of DRG biopsies from P70 WT mice showing the coexpression of CDKL5, CGRP, IB4, parvalbumin (Parv), and NF200 (20×; yellow arrowheads, coexpression; white arrow heads, expression of CDKL5 or individual cell type markers only). (E) Percentage of CDKL5+ cells coexpressing each neuronal marker (average percentage values represent average from three mice). (F) Percentage of parvalbumin+, IB4+, CGRP+, or NF200+ cells coexpressing CDKL5+ (average percentage values represent average from three mice). (G) CDKL5 and CGRP immunostaining from isogenic iPS-derived neurons after 0, 7, and 12 days of differentiation into nociceptors (63×). (H) Graphs showing the percentage of CDKL5+ cells and CDKL5+ neurons coexpressing CGRP at different stages of differentiation. n = 3, mean with SD, one-way ANOVA, Tukey’s post hoc, ****P < 0.0001 and ***P < 0.001.

CDKL5 expression in murine DRG was confirmed by reverse transcription polymerase chain reaction (RT-PCR) and immunoblotting (fig. S1, C and D). In addition, Cdkl5 mRNA neuronal expression was detected by RT-PCR from sorted green fluorescent protein–positive (GFP+) DRG neurons (fig. S1E), in line with previous single-cell studies (16, 17).

Next, to find whether CDKL5 was also expressed in human nociceptors, we differentiated iPSC obtained from human skin biopsies into nociceptors. After having confirmed the differentiation of iPSC by immunolabeling for βIII-tubulin as neuronal marker and CGRP as nociceptor marker, respectively (fig. S2, A to E), we observed that CDKL5 was indeed expressed in human nociceptors (Fig. 1, G and H).

To directly address whether the CDKL5 expression in DRG sensory neurons was required for nociception, we took advantage of the conditional deletion of Cdkl5 in sciatic DRG neurons by injecting an adeno-associated virus serotype 5 (AAV5)–CREGFP or control GFP virus to the sciatic nerve of Cdkl5-floxed mice (Fig. 2A and fig. S3, A and B) (12). Because the cre-virus reduces CDKL5 expression selectively in DRG neurons, the strong reduction of CDKL5 signal by immunoblotting strongly supports the selective neuronal expression in DRG ganglia and the efficiency of the deletion (Fig. 2, B and C). Cdkl5 conditionally deleted mice showed an increase in paw withdrawal latency and threshold in response to noxious hot temperature (Haargraves plantar test) and noxious mechanical stimulation (Von Frey), respectively (Fig. 2, D and F). We performed the same functional tests in Cdkl5 mutant mice, an established mouse model of CDD carrying a nonsense mutation (stop codon) leading to deletion of Cdkl5-Ex6 (4). Here, we found a reduction in nociception compared to wild-type (WT) littermates that phenocopied what was observed in Cdkl5 conditionally deleted mice (Fig. 2, E and G).

Fig. 2 Conditional Cdkl5 knockdown impairs nociception.

(A) Schematic illustrating conditional Cdkl5 knockdown in DRG neurons. Depicted is the injection of AAV5 particles into the sciatic nerve of 12-week-old Cdkl5-floxed mice. (B) Immunoblotting showing CDKL5 expression from bilateral sciatic DRG (pool of four mice) 8 weeks after AAV5-GFP or AAV5-GFP-cre. βIII-Tubulin has been used as loading control (ctrl). (C) Graph showing the conditional Cdkl5 knockdown in DRG (Cdkl5-floxed mice) after GFP-AAV5 versus CRE-AAV5 injection in the sciatic nerve (n = 3, mean with SEM, Student’s t test, ***P < 0.001). a.u., arbitrary units. (D to G) Behavioral assessment of pain sensitivity with Hargreaves (thermal nociception; paw withdrawal latency) and von Frey (mechanical allodynia; paw withdrawal threshold). (D and F) n = 10 mice, mean with SD, Student’s t test, ****P < 0.0001 and ***P < 0.001. (E and G) n = 7 WT versus 6 Mut mice, mean with SD, Student’s t test, **P < 0.01 and *P < 0.05. (H) Graph showing the reaction time (time to contact and time to remove the adhesive tape from the paw) in Cdkl5 DRG conditionally deleted (cKO) versus control GFP-injected mice. n = 10 mice, mean with SD, Student’s t test, *P < 0.05. (I) Graph showing the number of posterior foot slips in each run on a grid walk in Cdkl5 DRG conditionally deleted (cKO) versus control GFP-injected mice. n = 10 mice, mean with SD, Student’s t test, *P < 0.05.

In Cdkl5 conditionally deleted mice, additional sensory modalities were unaffected as indicated by the time to contact and the time to removal during the adhesive removal test (Fig. 2H) (18) and by the thermal place preference test, evaluating the non-nociceptive thermal sensitivity (fig. S3C). Sensorimotor functions as assessed by the grid walk task (19) were also unaltered (Fig. 2I), suggesting a selective role for CDKL5 in nociception. The number of nociceptors did not differ between WT and Cdkl5 null mice (fig. S4). Together, these data suggest that CDKL5 is required for physiological nociception that relies upon the CDKL5 expression in DRG neurons.

CDKL5 is required for epidermal innervation and neurite outgrowth of human and murine sensory neurons via a CaMKII-dependent mechanism

To investigate the signaling pathways involved in CDKL5-dependent pain transmission, we decided to identify interactors of CDKL5 in vivo. To this end, we performed immunoprecipitation of CDKL5 from the murine cortex, where the kinase is highly enriched, followed by mass spectrometry of the eluate (data file S1 and fig. S5A). Considering the normalized protein ratios of each immunoprecipitation experiment (n = 6), we identified 23 proteins coimmunoprecipitating with CDKL5 (fig. S5B). Seventy-four percent of them correspond to genes associated with neuronal activity, neural development, and epilepsy (fig. S5, C to F), supporting the physiological relevance of our pool of interactors. Top-ranked CDKL5 coimmunoprecipitating proteins included CaMKIIα, putatively the strongest interactor of CDKL5, and some proteins associated with the neuronal cytoskeleton, such as Myosin Heavy Chain 10 (Myh10) (20), tubulin beta 3 (Tubb3) (21), and Dynein Cytoplasmic 1 Heavy Chain 1 (Dynch1h1) (fig. S5G) (2225). The association with CaMKIIα also suggests a role for CDKL5 in calcium-dependent signaling, neuronal activity (26), and cytoskeleton remodeling (2732).

Previous immunohistochemical studies have found that CaMKII is expressed in DRG nociceptors (33), where it plays a role in the regulation of neurite outgrowth (34). In particular, CaMKII is expressed in TRPV1-immunoreactive nociceptors, in both CGRP+ and IB4+ subpopulations, where it is required for capsaicin-mediated nociception via modulation of TRPV1; capsaicin stimulation of TRPV1 and calcium entry activate CaMKII that is, in turn, required to potentiate TRPV1 signaling (35, 36). Hence, we hypothesized that CDKL5 could be a partner of CaMKIIα in the regulation of pain both via modulation of cytoskeleton remodeling and nociception signaling pathways.

First, we examined the coexpression of CDKL5, CaMKIIα, and TRPV1 in DRG neurons by confocal microscopy, and we found that they colocalize in most DRG neurons (Fig. 3, A to C, and fig. S6). Next, we confirmed, by coimmunoprecipitation and immunoblotting from brain extracts, that CDKL5 does interact with CaMKIIα in vivo (fig. S5H). Similarly, coimmunoprecipitation experiments after CDKL5 overexpression in human embryonic kidney 293 cells, or of its inactive kinase dead form CDKL5-K42R and CaMKIIα, confirmed the interaction between the active form of CDKL5 and CaMKIIα (fig. S5I). No interaction was observed between CaMKIIα and CDKL5-K42R (fig. S5I), which showed lack of kinase activity in an in vitro kinase assay (fig. S5J), supporting the specificity of the findings. We lastly established the molecular proximity between CDKL5 and CaMKIIα in DRG neurons by using the proximity ligation assay (Fig. 3, D and E), supporting the coexpression and coimmunoprecipitation data.

Fig. 3 Coexpression of CDKL5 with CaMKIIα and TRPV1.

(A) Confocal microscopy of immunofluorescence showing coexpression of CDKL5, CaMKIIα, and TRPV1 (20×; yellow arrowheads, coexpression; white arrowheads, expression of CDKL5 or individual cell type markers only). (B) Average diameter and SD of each CDKL5+, CaMKIIα+, or TRPV1+ population of cells (average values from three mice). (C) Percentage of CDKL5+ or CaMKIIα+ cells coexpressing each marker (average percentage values represent average from three mice). (D) Confocal images showing individual WT or Cdkl5−/y DRG neurons (63×) after proximity ligation assay (PLA) (red) and 4′,6-diamidino-2-phenylindole (DAPI) staining (blue). The presence and intensity of red dots represent the presence and degree of molecular proximity between CDKL5 and CaMKIIα. (E) Graph showing the number of PLA-positive cells in WT and Cdkl5/y DRGs. n = 3 animals, mean with SEM, Student’s t test, **P < 0.005.

We next investigated whether CDKL5-CaMKIIα signaling axis would be required for the outgrowth of DRG neuronal processes. Assessment of neurite outgrowth was carried out in both murine and human iPSC–derived nociceptors from Cdkl5 null mice and patients with CDD. We found that cultured DRG neurons from Cdkl5 mutant mice display impaired neurite outgrowth in CGRP-positive neurons (Fig. 4, A and C), and this was rescued to the WT amounts by viral-mediated Cdkl5 overexpression (Fig. 4E and fig. S7). On the contrary, outgrowth in proprioceptive parvalbumin-positive neurons remained unaffected by Cdkl5 loss of function (Fig. 4, B and D), indicating a specific role for CDKL5 in nociceptors. Impaired neurite outgrowth was also observed in iPSC-derived nociceptive DRG neurons (CGRP+) obtained from skin biopsies of patients with CDD, compared with the respective isogenic controls (Fig. 4, F and G). CGRP+ neurons represent 30 to 40% of WT and mutant DRG neurons, and more than 90% of iPSC-derived sensory neurons were βIII-tubulin+ (fig. S2D). Capsaicin-responding cells were ~60% of the KCl-responding murine DRG cultured neurons, whereas they represented more than 90% of the KCl-responding iPSC-derived neurons, both in mutant and isogenic cultures, indicating functionality of the nociceptors (fig. S2E). In addition, we found that neurite outgrowth was impaired in neurons derived from patients carrying a CDKL5 mutation compromising the kinase active site and the TEY motif but not in neurons where only the terminal part of the CDKL5 tail is lost (Fig. 4, F to H, and Table 1). Because epidermal innervation that is needed for proper nociception is a dynamic process requiring continuous cytoskeleton remodeling and outgrowth, we examined whether CDKL5 deficiency was associated with a defect in epidermal innervation.

Fig. 4 CDKL5 is required for outgrowth of human sensory neurons.

(A and B) Representative immunofluorescence of βIII-tubulin, CGRP, or parvalbumin (Parv) in WT and Cdkl5−/y DRG cultured neurons [3 days in vitro (DIV3)] showing reduced neurite outgrowth in CGRP+ but not Parv+ Cdkl5−/y neurons. (C) Graph showing reduced neurite outgrowth in CGRP+ neurons from Cdkl5−/y mice (n = 5 biological replicates, mean with SD, Student’s t test, ***P < 0.001). (D) Graph showing neurite outgrowth in Parv+ neurons from WT and Cdkl5−/y mice. n = 5 biological replicates, mean with SD, Student’s t test, P > 0.05. ns, not significant. (E) Effect of Cdkl5 overexpression on neurite outgrowth in CGRP+ neurons from Cdkl5−/y mice (n = 5 biological replicates, mean with SD, one-way ANOVA, Tukey’s post hoc, **P < 0.005). (F) Representative images of CGRP immunostaining of iPS-derived sensory neurons from three patients with CDD carrying unique mutations (E55fsX74; G155fsX197; S855X) compared to their isogenic controls (20×). (G) CGRP immunostaining was used to measure total average neurite outgrowth (NeuronJ) of nociceptive neurons. n = 5 independent experiments, total neurite outgrowth per cell. Mean with SD, Student’s t test, *P < 0.05 and **P < 0.01. (H) Schematic representation of human CDKL5 protein adapted, (2) with the catalytic domain in blue [adenosine 5′-triphosphate (ATP)–binding site, amino acids (aa) 19 to 43; kinase active site, amino acids 131 to 143; TEY motif, amino acids 169 to 171] and the C-terminal tail in white (nuclear localization signal 1, amino acids 312 to 315; nuclear localization signal 2, amino acids 784 to 789; nuclear export signal, amino acids 836 to 845).

Immunohistochemical analysis of skin innervation revealed impairment in epidermal but not dermal innervation in CDKL5 mutant versus WT adult mice (Fig. 5, A to E). We also measured epidermal innervation from P4 to adulthood (P70) and found that Cdkl5 mutant mice display deficient epidermal innervation only in adulthood (P70) and not at earlier stages (P4 or P16; Fig. 5B). In support of these data, we found, by immunoblotting of skin lysates, that the specific neuronal and axonal proteins PGP9.5 (Protein Gene Product 9.5 ), CGRP, and βIII-tubulin are reduced in Cdkl5 mutants as compared to WT adult mice (Fig. 5, F and G). However, consistent with the selective reduction of nociceptive fibers in the epidermis of Cdkl5 mutant mice, the expression of these axonal proteins is not reduced in sciatic nerve lysates (Fig. 5, H and I) or after immunostaining for CGRP (Fig. 5, J and K) in Cdkl5 mutant mice.

Fig. 5 CDKL5 is required for epidermal innervation.

(A) Immunohistochemical localization of nerve fibers in the derm and in the epiderm of 50-μm murine skin sections (P70), using PGP9.5 as specific axonal marker (20×). Sections were counterstained for nuclei in Mayer’s hematoxylin. (B) The density of the intraepidermal nerve fibers (IENF per millimeter), crossing the dermal-epidermal junction, was analyzed in WT and Cdkl5−/y mice at different developmental stages (P4, P16, and P70). n = 3, mean with SD, two-way ANOVA, Sidak’s post hoc, **P < 0.005. (C) Immunofluorescence of skin biopsies from P70 WT and Cdkl5−/y mice. Representative micrographs where collagen IV identifies the epidermal basal lamina and βIII-tubulin identifies IENFs (20×). (D) Individual IENF crossing the dermal-epidermal junction were analyzed. n = 4, mean with SD, Student’s t test, ****P < 0.0001. (E) The intradermal neuronal network (DNet; μm/mm) was measured by using NeuronJ software (magnification, ×20). N = 4, mean with SD, Student’s t test, P > 0.05. (F) Western blotting showing the expression of neuronal markers (PGP9.5, βIII-tubulin, and CGRP) from skin lysates of P16 and P70 WT and Cdkl5−/y mice. (G) The immunoblotting bands have been quantified by densitometry after normalization with GAPDH (glyceraldehyde-3-phosphate dehydrogenase) used as internal loading control. n = 3 to 4 independent experiments, mean with SD, two-way ANOVA, Sidak’s post hoc, *P < 0.05, **P < 0.01, and ***P < 0.005. (H) Western blotting showing the expression of PGP9.5, βIII-tubulin, and CGRP from sciatic nerve lysates in WT and Cdkl5−/y mice. GAPDH was used as loading control. (I) Densitometry of immunoblotting bands from P70 WT and Cdkl5−/y mice was performed after normalization to GAPDH. n = 3 independent experiments, mean with SD, Student’s t test, P > 0.05. (J) Representative CGRP immunofluorescence of sciatic nerve sections from P70 WT and Cdkl5−/y mice (10×). (K) Graphs showing quantification of the immunofluorescence signal in WT and Cdkl5−/y mice. n = 4 mice per group, mean with SD, Student’s t test, P > 0.05.

Next, we asked whether CaMKII activity was required for CDKL5-dependent neurite outgrowth. We found that DRG outgrowth was impaired after administration of the specific CaMKII enzymatic inhibitor KN93, whereas its inert structural analog KN92 did not have an effect (Fig. 6, A and B). KN93 did not further reduce outgrowth in CDKL5 mutant cells (Fig. 6, A and B), indicating that CDKL5 and CaMKII are likely to belong to the same signaling pathway. Cdkl5 overexpression rescued outgrowth defects in Cdkl5 mutant knockout DRG neurons, but this effect was blocked by KN93 (Fig. 6, A and B). These data together suggest that CDKL5-dependent DRG outgrowth relies on CaMKII.

Fig. 6 CDKL5/CaMKII signaling is required for capsaicin signaling and outgrowth of sensory neurons.

(A) Representative immunofluorescence of DRG cultured neurons (DIV3, from P70 WT and Cdkl5−/y mice), immunostained for GFP or HA (hemagglutinin) after neurons had been infected with AAV9 particles expressing GFP or HA-Cdkl5107 at DIV1 (20×). DRG neurons were also treated with 0.5 μM CaMKII inhibitor KN93 or its inert structural analog KN92. (B) Graphs showing neurite outgrowth of GFP- or HA-positive cells that was normalized to the control WT-GFP. n = 9, biological replicates, mean with SD, one-way ANOVA, Tukey’s post hoc, ****P < 0.0001, ***P < 0.001, and **P < 0.005. (C) Calcium imaging (Fura-2) from iPS-derived nociceptors carrying three different CDKL5 mutations (E55fsX74; G155fsX197; S855X) compared with the respective isogenic controls. Graphs showing quantification of intracellular calcium in response to capsaicin. n = 5, average single-cell Fura-2 normalized signal per well (~30 cells per well). Mean with SD, Student’s t test, **P < 0.01 and *P < 0.05. (D) Representative transient traces of the calcium amount in DIV3 WT and Cdkl5−/y neurons (Fura-2 tracks). (E to G) Calcium imaging in DIV3 cultured Cdkl5−/y DRG neurons (Fura-2) after capsaicin and after Cdkl5 overexpression (sspTR-CBh-HA-Cdkl5107 versus GFP AAV9 particles, infection at DIV1) (E). The same experimental conditions as (D) after pretreatment with KN93 or KN92 in WT and Cdkl5−/y neurons (F), overexpressing Cdkl5 or GFP (G). n = 5, average single-cell Fura-2 normalized signal per well (~30 cells per well), mean with SD, one-way ANOVA, Tukey’s post hoc, ****P < 0.0001, ***P < 0.001, and **P < 0.005. (H) Calcium imaging in DRG explants from WT and Cdkl5−/y mice. Graphs showing quantification of intracellular calcium in response to capsaicin (Fluo-4). n = 5, average single-cell Fura-4 normalized signal per well (~30 cells per well), mean with SD, Student’s t test, **P < 0.01.

CDKL5 is required for TRPV1/CaMKII-dependent signaling and nociception in sensory neurons and in vivo

Because TRPV1 activity relies on CaMKII-dependent interaction and phosphorylation (35, 36), we hypothesized that TRPV1 signaling might require CDKL5. Therefore, we aimed to address whether CDKL5 and CaMKII regulate TRPV1-dependent calcium signaling after stimulation with the well-established ligand capsaicin, which activates nociception by engaging with its receptor TRPV1 (3739). We used capsaicin as specific activator of TRPV1, and we measured calcium influx in iPS-derived nociceptors to find that capsaicin-induced increase in calcium is impaired in neurons carrying CDKL5 mutations involving the kinase active domain (E55fsX74) and the TEY motif (G155fsX197) (Fig. 6C). Similarly, capsaicin-induced calcium responses were impaired in Cdkl5 mutant murine DRG explants and cultured neurons, where we established that AAV9-Cdkl5 overexpression rescued the defect in calcium response (Fig. 6, D, E, and H). When we blocked CaMKII signaling in cultured WT DRG neurons by using KN93, in contrast to its inert structural analog KN92, we observed, as expected, that KN93 inhibited the activation of TRPV1 after administration of capsaicin. Inhibition was similar to what was observed in Cdkl5 mutant cells, where the CaMKII inhibitor did not further decrease calcium amount (Fig. 6E). Delivery of KN93 after Cdkl5 overexpression in Cdkl5 mutant DRG neurons blocked the rescue in calcium mobilization after capsaicin delivery (Fig. 6F). Together, these data confirm our initial hypothesis that CaMKII is required for CDKL5-dependent calcium responses after capsaicin.

To further test the role of CaMKII signaling in CDKL5 function, we infected cultured WT and Cdkl5 mutant DRG neurons with AAV particles expressing the CaMKII–autocamtide-2–related inhibitor peptide (CaMKII-AIP) plasmid (40) or an active Cdkl5 plasmid to rescue CDKL5 function in null neurons. We then measured neurite outgrowth and calcium after capsaicin stimulation (fig. S8).

CaMKII-AIP–infected neurons showed impaired neurite outgrowth (fig. S8, A and B) and reduced capsaicin-dependent calcium mobilization (fig. S8, C and D) to a similar degree as Cdkl5 mutant knockout DRG neurons. Both outgrowth and calcium were not further impaired in Cdkl5 mutant knockout DRG neurons after CaMKII-AIP infection, suggesting functional interaction of the two proteins. Last, neurite outgrowth and capsaicin-dependent calcium induction were rescued by Cdkl5 overexpression (fig. S8, A to D); however, this failed when CaMKII was inhibited (fig. S8, A to D), indicating that CaMKII is required for CDKL5 activity.

Last, we hypothesized that capsaicin signaling would also be compromised in vivo, leading to biochemical and behavioral impairments in Cdkl5-deficient mice. In line with this hypothesis, we found that both the licking behavior and cell signaling response to intradermal injection of capsaicin were impaired in vivo in Cdkl5−/y mutant mice and after conditional deletion of Cdkl5 specifically in sciatic DRG sensory neurons. The time spent licking the right paw after injection of capsaicin (1 μg) was reduced both in Cdkl5−/y mice and in AAV5-CRE–treated versus AAV5-GFP–treated Cdkl5-floxed mice (vehicle was injected in the left paw) (Fig. 7, A to C). Consistently, 5 min after capsaicin injection in the right paw, pCaMKIIα was up-regulated in DRG on the injected side in WT but not in Cdkl5−/y mice, supporting a defective signaling response (Fig. 7, D to F). We confirmed defective transmission of the stimulus in Cdkl5−/y mice also to the right dorsal horn of the spinal cord by measuring phosphorylated extracellular signal–regulated kinase 1/2 (pERK1/2) expression, typically activated by capsaicin-TRPV1 signaling in lamina I and II of the dorsal horn (41). pERK1/2 was induced only in WT but not in Cdkl5−/y mice after intradermal injection of capsaicin (Fig. 7, G to I) (42, 43).

Fig. 7 Cdkl5 deletion impairs capsaicin-mediated nociceptive signaling and licking behavior.

(A and B) Bar graphs showing the time spent licking the right paw after the injection of 1 μg of capsaicin in the posterior right paw versus vehicle injected in the left paw [(A) WT versus Cdkl5−/y; (B) GFP versus CRE AAV5 intrasciatically injected Cdkl5-floxed mice). (A) n = 5 mice each group, mean with SEM, two-way ANOVA, Sidak’s post hoc, ****P < 0.0001. (B) n = 8 mice each group, mean with SEM, two-way ANOVA, Sidak’s post hoc, *P < 0.05. (C) Fisher’s combined probability test (false discovery rate, meta-analysis adjusted for multiple testing) of the P values from the time courses in (A and B) (one statistical test for each time point). (D) Representative confocal immunofluorescence images for phospho–T286-CaMKIIα in WT and Cdkl5−/y sciatic left and right DRG (63×). Tissue was fixed immediately after the capsaicin injection test in the right paw (vehicle in the left). (E and F) Bar graphs showing the percentage number of positive pCaMKIIα cells (E) and the pCaMKIIα average signal intensity in WT versus Cdkl5−/y left and right DRG small cells after normalization to WT (F). n = 5 mice per group, mean with SD, one-way ANOVA, Tukey’s post hoc, ****P < 0.0001 and ***P < 0.001. (G) Schematic illustrating a cross section of the lumbar spinal cord emphasizing the left (WT and Mut) and right (WT and Mut) dorsal horns (after injection of vehicle in the left paw and capsaicin in the right paw) where nociceptive neurons make a synapse in the lamina I and II (L, left; R, right). (H) Representative confocal immunofluorescence images for pERK1/2 in WT and Cdkl5−/y spinal dorsal horns (10×). Tissue was fixed 5 min after capsaicin injection in the posterior right paw versus vehicle injected in the left paw. (I) Graph showing pERK1/2 signal intensity in spinal lamina I and II (right side versus left side) in Cdkl5−/y mice versus WT. n = 5 mice per group, mean with SD, Student’s t test, **P < 0.01.


Our data reveal a previously unknown function of CDKL5 in the regulation of primary nociception in the peripheral nervous system as shown by impaired nociception after conditional deletion of Cdkl5 in DRG sensory neurons. Specifically, we found that CDKL5 is required for CaMKII-dependent TRPV1 signaling and outgrowth of sensory neurons in both CDD murine and human neurons. In vivo, in a CDD animal model, this translates into impaired capsaicin-dependent nociceptive signaling and behavioral responses as well as in reduced epidermal innervation.

The anatomical, mechanistic, and functional data all together support the initial anamnestic evidence that more than 34% of CDKL5 patients have defective pain perception. The clinical questionnaire that we used should not be considered as an exhaustive clinical piece of data per se but rather a clinical indication that generated a hypothesis guiding further experiments in animal models and human iPSC. In a much smaller percentage of patients with enhanced pain perception, cortical hyperexcitability, which is a typical feature of these patients, might lead to dysfunctional central processing of nociception. This might be in line with disorders affecting cognition and neuronal activity in the CNS such as autism spectrum disorders (ASDs) and Rett syndrome (RTT), where it has been proposed that altered pain sensitivity is related to a dysfunctional mode of cortical pain processing, rather than to defective primary peripheral nociception (4447). Therefore, it is possible that modest impairment in peripheral pain perception and marked cortical hyperexcitability might coexist in a subset of patients leading to the net effect of enhanced pain perception. Alternatively and similar to the protein SHANK3 that regulates TRPV1 and heat pain as a loss of function (48) and touch sensation as a gain of function in ASD (49), we cannot rule out that CDKL5 might have a dual loss and gain-of-function role in specific sensory neuron subtypes.

CDD and RTT are distinct diseases caused by mutations in separate independent genes, CDKL5 and methyl CpG binding protein 2 (MeCP2), respectively. Although in vitro data suggested that the two might interact (1, 50, 51), functional interaction in vivo remains unclear.

Recently, genes such as MeCP2 and CaMKIIα, which have been implicated in the pathogenesis of ASD (5254) and that are highly expressed in the CNS, have also been localized to the peripheral nervous system where they regulate sensory modalities (40, 55). However, a role in nociception had not been described so far.

Limitations of our work include the need for future comprehensive clinical and neurophysiological investigation to fully characterize pain perception in patients with CDD and for the assessment of potential additional CDKL5-dependent mechanisms that might control pain perception and nociception. In addition, it is presently not known whether CDKL5 plays a role in inflammatory or in traumatic painful conditions or whether reduced epidermal innervation is directly responsible for impaired nociception. Last, although impairment in epidermal innervation only occurred in adult mice, the human clinical data includes infants and toddlers. Published evidence suggests that skin innervation is already complete at about 1.5 years of age (56); however, it would be useful to generate data allowing for a direct comparison between the maturation of skin innervation in humans and mice.

In summary, the importance of these findings lies on (i) the identification of a previously unrecognized regulatory mechanism for nociception that relies on CDKL5, (ii) the discovery of impaired nociception as a previously uncharacterized symptom in CDD, and (iii) the usefulness of nociception as therapeutic outcome measure in animal models of CDD. Because animal models of CDKL5 deficiency do not develop seizures, monitoring pain responses during therapeutic interventions might be a unique opportunity to test disease-modifying treatment in preclinical settings before clinical trials. Last, our data suggest that gene therapy or other interventions in patients with CDKL5 deficiency should be directed not only to the brain but also to DRG to restore nociceptive input in patients with impaired pain perception.


Study design

We investigated the role of CDKL5 in nociception in both human and murine sensory neurons as well as in Cdkl5 mutant mice by performing both mechanistic and behavioral studies. All surgical and experimental procedures on rodents were carried out in accordance with the U.K. Animals (Scientific Procedures) Act 1986 and approved by the veterinarian and ethical committee of Imperial College London. Animals were assigned randomly to experimental groups, and surgeries were carried out in a random block design. All analyses were performed by the same experimenter who was blinded to the experimental groups. All behavioral testing and analyses were performed by an observer blinded to the experimental groups. n values represent the number of animals in the experiment, and each experiment contained a minimum of three technical replicates.


We used clinical data that have been collected in the ICDD ( that was established in 2012. Ethics approval was obtained from the Human Research Ethics Committee, University of Western Australia. This questionnaire contains a broad range of clinical information on patients with CDD provided by caregivers. It collects information from families/caregivers of a child with CDD mainly through completion of an online questionnaire. The data collected are comprehensive and include questions on developmental milestones, functional abilities, epilepsy, sleep, and other aspects of medical history including details of genetic mutation, medication, hospital admissions, and behavior. Here, we used clinical data related to pain perception that are reported in Table 1.


All animal procedures were approved by the Imperial College London Ethics Committee and were performed in accordance with the U.K. Animals Scientific Procedures Act (1986). C57BL/6 background mice lacking Cdkl5 exon 6 (4) and Cdkl5 floxed C57BL/6 background mice containing loxP sites flanking Cdkl5 exon 4 (12) and WT littermates or C57BL/6 (Harlan) mice ranging from days 4 to 70 of age were used for all experiments. Mice were anesthetized with isoflurane (3% induction and 2% maintenance), and buprenorphine (0.1 mg kg−1) and carprofen (5 mg kg−1) were administered perioperatively as analgesic. The experimenter was blind to the genotype of animals and to treatment given to each animal.

Injection of viral vectors into the sciatic nerve

A total of 2.5 μl of each viral vector [AAV5-GFP SignaGen SL100819 or AAV5-Cre SignaGen SL100821, titer 3.06 × 1013 genome copies (GC)/ml] was injected into the sciatic nerve of adult Cdkl5-floxed mice (12) with a Hamilton syringe and Hamilton needle (NDL small RN ga34/15 mm/pst45o). After 2 months, the animals were exposed to specific behavioral tests and then sacrificed to evaluate the silencing of Cdkl5 gene in sciatic DRG by Western blot.

Mechanical allodynia test

Mechanical allodynia was quantified by measuring the hind paw withdrawal response to von Frey filament stimulation. Animals were placed in methacrylate cylinders (20 cm high, 9 cm diameter) with a wire grid bottom through which the von Frey filaments (North Coast Medical Inc.) with a bending force in the range of 0.16g to 8g were applied by using a modified version of the up-down paradigm, as previously reported (57). The filament of 0.16g was used first and the 8g filament was used as a cutoff. Then, the strength of the next filament was decreased or increased according to the response. The threshold of response was calculated using a regression curve generated on the basis of the sequence of filament number (numbered from 1 to 9) versus log10 of their strength in gr; an interpolation of the number of filament plus or minus a correction factor of 0.5 according to the response was then used to calculate the threshold of response in gr. Clear paw withdrawal, shaking, or licking of the paw was considered as a nociceptive-like response. Both hind paws were tested and averaged. Animals were allowed to habituate for 1 hour before testing to allow an appropriate behavioral immobility.

Hargreaves test (plantar test)

Paw withdrawal latency in response to radiant heat was measured using the plantar test apparatus (Ugo Basile). Briefly, the mice were placed in methacrylate cylinders (20 cm high × 9 cm diameter) positioned on a glass surface. The heat source was positioned under the plantar surface of the hind paw and activated with a radiant light beam, the intensity of which was chosen in preliminary studies to give baseline latencies from 8 to 9 s in control mice. A cutoff time of 15 s was used to prevent tissue damage in the absence of response. The mean paw withdrawal latencies from both hind paws were determined from the average of five separate trials, taken at 2-min intervals to prevent thermal sensitization and behavioral disturbances. Animals were habituated to the environment for 1 hour before the experiment to become quiet and to allow testing (58).

Grid walk

Mice were allowed to run the grid walk [50 cm by 5 cm plastic grid (1 cm by 1 cm) placed between two vertical 40-cm-high wood blocks] three times per session. The total number of steps and missteps per run for each hind paw was analyzed by a blinded investigator.

Adhesive removal

Small adhesive stimuli (6-mm round adhesive labels) were placed on both hind paws of the mouse; the time to make first contact with both forepaws and the time to completely remove the adhesive were recorded for each paw. Each mouse underwent three trials. All testing was performed by a blinded investigator after a period of acclimatization.

Thermal place preference test

Unrestrained mice were allowed to move freely between two connected plate compartments (Ugo Basile), each set at a given temperature, thus choosing their preferred temperature environment. Mice were placed on one plate set to 30°C, which is known to be a thermoneutral temperature for mice, whereas the other plate was set to either neutral 30°C or nonnoxious temperatures of either 23° or 37°C. Each animal was monitored for 10 min, recording the time spent in each plate compartment. Each temperature range was tested at the same time of day on each testing day, randomly interchanging plate compartments to avoid environmental bias. All testing was performed by a blinded investigator after a period of acclimatization.

Statistical analysis

GraphPad Prism 7 was used for statistical analysis. Data were plotted as means ± SD or SEM. Statistical significance was set at P < 0.05. All experiments were performed in triplicate unless otherwise specified. Asterisks indicate a significant difference analyzed by analysis of variance (ANOVA) with Tukey or Sidak post hoc test, Student’s t test as indicated for normal distributions, parametric two-tailed test, and Fisher’s combined probability test. All data analyses were performed blinded to the experimental group.


Materials and Methods

Fig. S1. CDKL5 expression.

Fig. S2. Differentiation of nociceptors in CDKL5 mutant neurons.

Fig. S3. Cdkl5 conditional deletion in DRG neurons and thermal preference test.

Fig. S4. CGRP-, IB4-, and CaMKIIα-positive DRG neurons in WT and Cdkl5 mutant mice.

Fig. S5. CDKL5 immunoprecipitation with mass spectrometry and CaMKIIα coimmunoprecipitation.

Fig. S6. Coexpression of CDKL5, CaMKIIα, and TRPV1 in DRG neurons.

Fig. S7. CDKL5 overexpression in DRG.

Fig. S8. CDKL5-dependent neurite outgrowth and calcium mobilization in DRG neurons.

Data file S1. CDKL5 coimmunoprecipitated proteins by mass spectrometry.

Data file S2. Raw data.

References (5965)


Acknowledgments: We would like to thank A. Freiwald for assistance with the proteomics data analysis (proteomics facility in Mainz), S. B. McMahon (Kings College London) for insightful discussions, E. Bradbury (Kings College London) for the support with the thermal preference test, and M. Silvestre (Crick Institute London) for assistance with mouse colonies. Funding: This work was supported by start-up funds from the Division of Brain Sciences, BRC funds, Imperial College London (S.D.G.); Wings for Life (S.D.G.); The Rosetrees Trust (S.D.G.); the NIHR Imperial Biomedical Research Centre (S.D.G.); International Foundation for CDKL5 Research (H.L., J.D., and K.W.); and NHMRC (H.L.). Author contributions: P.L.M. designed and performed experiments and data analysis and wrote the paper. A.H. designed and performed experiments and data analysis. L.L.B. and T.P. designed experiments and provided mice for experiments. T.H.H., F.D.V., G.K., J.C., Y.G., Q.A.M., and N.G. performed experiments. I.P., K.W., and J.D. performed data analysis. K.B. performed experiments and data analysis; S.K.U. provided experimental advice, edited the paper, and provided mice for experiments. H.L. designed experiments, performed data analysis, and edited the paper. H.Y. and D.S.M. provided research material for experiments. N.I. designed experiments, provided experimental advice, and edited the paper. N.D.M. provided experimental advice, edited the paper, and provided mice and research material for experiments. S.D.G. designed experiments, provided funding, and wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD013316. All data associated with this study are in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article