Epilepsy-related CDKL5 deficiency slows synaptic vesicle endocytosis in central nerve terminals

Cyclin-dependent kinase-like 5 (CDKL5) deficiency disorder (CDD) is a severe early-onset epileptic encephalopathy resulting mainly from de novo mutations in the X-linked CDKL5 gene. To determine whether loss of presynaptic CDKL5 function contributes to CDD, we examined synaptic vesicle (SV) recycling in primary hippocampal neurons generated from a Cdkl5 knockout rat model. Using a genetically-encoded reporter, we revealed that CDKL5 is selectively required for efficient SV endocytosis. We showed that CDKL5 kinase activity is both necessary and sufficient for optimal SV endocytosis, since kinase-inactive mutations failed to correct endocytosis in Cdkl5 knockout neurons, whereas the isolated CDKL5 kinase domain fully restored SV endocytosis kinetics. Finally, we demonstrated that CDKL5-mediated phosphorylation of amphiphysin 1, a putative presynaptic target, is not required for CDKL5-dependent control of SV endocytosis. Overall, our findings reveal a key presynaptic role for CDKL5 kinase activity and enhance our insight into how its dysfunction may culminate in CDD.


28
The majority of neuronal communication occurs at synapses, at which the presynapse 29 contains an abundant number of synaptic vesicles (SVs) loaded with neurotransmitters that 30 are generally released in response to neuronal activity. Following SV fusion, synchronized 31 mechanisms of SV regeneration from the presynaptic plasma membrane guarantee the 32 availability of readily releasable SVs upon repetitive firing and, hence, the fidelity of 33 neurotransmission (Cousin, 2017;Soykan et al., 2016). Neurodevelopmental disorders affect 34 more than 3 % of children worldwide and involve the disturbance of programmed brain 35 development leading to cognitive, social and motor deficits with epileptic seizures being a 36 frequently observed comorbidity (Parenti et  2009). Importantly, S293 is a major in vivo phosphorylation site on Amph1 and is 65 dephosphorylated during neuronal activity, indicating that it may be of high biological 66 importance (Craft et al., 2008). 67 In the present study, we use a novel Cdkl5 KO rat model (Simões de Oliveira et al. 2022) to 68 examine SV recycling in CDKL5-deficient hippocampal neurons. Using the genetically-encoded 69 fluorescent reporter synaptophysin-pHluorin (sypHy), we reveal that SV endocytosis is slower 70 upon loss of CDKL5, but SV exocytosis remains unaffected. Following a molecular replacement 71 strategy we demonstrate that the kinase activity of CDKL5 is both necessary and sufficient to 72 correct dysfunction in SV endocytosis. Finally, we determined that the phosphorylation status 73 of Amph1-S293 remains unaltered in CDKL5-null neurons, revealing that CDKL5 exerts its 74 effect on SV endocytosis via a distinct presynaptic substrate. Taken together, our work reveals 75 that CDKL5-mediated phosphorylation is critical for SV endocytosis efficiency, and that CDKL5 76 deficiency is responsible for presynaptic malfunction. 77

78
Endogenous CDKL5 is sorted into the presynaptic terminal 79 CDKL5 is a ubiquitous neuronal protein kinase (Rusconi et al., 2011;Schroeder et al., 2019) 80 however, its localisation at the nerve terminal has not been extensively addressed. To verify 81 that CDKL5 is present in presynaptic terminals, and therefore in the correct location to 82 influence SV recycling, a classical subcellular fractionation was performed. During this 83 protocol, an adult rat brain was subjected to homogenisation and differential centrifugation 84 to generate distinct subcellular fractions, including a crude synaptosome-(P2, mainly 85 representing the presynapse with attached postsynaptic density) and an SV-enriched (LP2) 86 fraction. Western blotting with a CDKL5-specific antibody ( Figure S1) revealed that CDKL5 was 87 present in the P2 fraction and enriched in the LP2 fraction, where the SV protein 88 synaptophysin 1 (Syp1) also accumulated ( Figure 1A). To assess whether CDKL5 is targeted exclusively to nerve terminals or displays a more diffuse 95 axonal distribution, we performed coefficient of variance (CV) analysis. Hippocampal neurons 96 were transfected with either CDKL5 fused to the fluorescent protein mCerulean (mCer-97 CDKL5), Syp1-mCer or the empty mCer vector and were then immunolabelled for the 98 presence of the fluorescent tag ( Figure 1B). SV proteins, such as Syp1, are anticipated to have 99 a punctate distribution along the axon and therefore a higher CV value. In contrast, lower CV 100 values indicate a homogeneous distribution of a protein along the axon. In agreement, mCer-101 Syp1 displayed a localised distribution along the axon and a high CV value, in agreement with 102 previous results (Gordon and Cousin, 2013). Quantification of the distribution profile of CDKL5 103 in axonal segments indicated a CV value similar to the empty mCer vector ( Figure 1C). 104 Therefore, CDKL5 is diffusely distributed along the axon, including presynaptic terminals. 105 blotting. Initially, we confirmed the absence of CDKL5 in lysates of KO neurons (Figure 2A). 114 We then analysed a range of presynaptic molecules including proteins important for SV 115 recycling, such as clathrin heavy chain (CHC), dynamin 1 (Dyn1), endophilin A1, alter the total protein level of any candidate, or the phosphorylation status (and thus activity) 124 of either GSK3 or Akt when compared to wild-type (WT) controls (Figure 2A). Therefore CDKL5 125 KO neurons do not display overt alterations in presynaptic proteins or signalling cascades. 126

Loss of CDKL5 does not influence the levels of presynaptic proteins or the number of
Next, we investigated whether the lack of CDKL5 led to a reduced number of presynaptic 127 terminals. To achieve this, WT and CDKL5 KO neurons were double-stained for two distinct 128 presynaptic markers, synaptic vesicle protein 2A (SV2A) and VGLUT1, to assess the number of 129 presynaptic boutons and excitatory presynaptic subtypes, respectively. There were no 130 genotype-specific differences in SV2A-and VGLUT1-positive puncta along neuronal processes 131 ( Figure 2B). Therefore, there is no effect of the absence of CDKL5 on either the number of 132 total or excitatory presynaptic terminals ( Figure 2C, D). Overall, this data reveals that the 133 formation and maintenance of nerve terminals in rat primary neuronal cultures is not affected 134 upon CDKL5 deficiency. 135

Loss of CDKL5 impairs SV regeneration but does not influence SV exocytosis 136
The presynaptic localisation of CDKL5 suggests that CDKL5 is implicated in SV recycling. 137 Indeed, phenotypes reported in mice lacking CDKL5, such as altered frequency of 138 spontaneous and miniature postsynaptic currents (mPSCs) ( unquenched upon stimulus-dependent SV exocytosis and exposure to the cell surface, and 146 re-quenched following endocytosis and SV acidification ( Figure 3A). To determine the 147 potential contribution of CDKL5 to SV recycling across a range of stimulus intensities, primary 148 hippocampal neurons derived from CDKL5 KO rats or WT littermate controls were transfected 149 with sypHy and stimulated with action potential (AP) trains of either 5 Hz or 10 Hz (both 300 150 APs) or 40 Hz (400 APs) ( Figure 3B, E, H). To quantify for the extent of activity-dependent SV 151 exocytosis, the amount of sypHy fluorescence during stimulation was measured as a 152 proportion of the total fluorescence within the presynapse revealed by perfusion with NH 4 Cl 153 that allows for an estimation of the total recycling SV pool. We found that the extent of SV 154 exocytosis remained unaltered between genotypes across all stimulation frequencies 155 investigated ( Figure 3C, F, I). To confirm this phenotype, we next measured the rate of sypHy 156 fluorescence increase during prolonged stimulation (10 Hz for 90 s) in the presence of 157 bafilomycin A1. Bafilomycin A1 is a V-type ATPase inhibitor, and therefore removes any 158 potential contribution from SV endocytosis to the sypHy response during the stimulation by 159 blocking SV acidification (Sankaranarayanan and Ryan, 2001). When this experiment was 160 performed, no difference was observed in either the rate of the sypHy fluorescence increase 161 (SV exocytosis rate) or the extent of the sypHy response (SV recycling pool size) between WT 162 and CDKL5 KO neurons ( Figure S2A, B, C  (Sankaranarayanan and Ryan, 2000). To quantify the kinetics of SV 169 retrieval, the sypHy stimulation peak was normalised, and the amount of sypHy remaining to 170 be retrieved 2 minutes after termination of stimulation was measured. This parameter was 171 used for consistency across protocols, since in specific cases the decay kinetics were not 172 mono-exponential (rendering time constant measurements redundant). CDKL5 KO neurons 173 consistently displayed slower SV endocytosis across all frequencies examined when 174 compared to WT, suggesting that CDKL5 is important for optimal SV endocytosis ( Figure 3D, 175 G, J). Interestingly, the requirement for CDKL5 appeared to be more prominent at lower 176 stimulation frequencies. 177 To confirm that this phenotype was due to slowed SV endocytosis and not dysfunctional SV 178 acidification, we determined the kinetics of SV acidification using an acid-pulse protocol 179 (Granseth et al., 2006). In this protocol, an impermeant acid buffer (pH 5.5) is perfused 180 immediately after stimulation to quench all surface sypHy, which exclusively reveals the 181 sypHy signal inside recently retrieved SVs (where the quenching rate can be calculated). In 182 this protocol WT and CDKL5 KO neurons expressing sypHy are perfused with acid buffer both 183 prior to stimulation (to reveal an initial baseline) and immediately after stimulation (10 Hz, 30 184 s, to reveal the quenching rate inside SVs) ( Figure S2D). No significant difference in the SV 185 acidification rate in neurons lacking CDKL5 compared to WT neurons was apparent ( Figure  186 S2E), confirming that the slowing in the post-stimulus sypHy fluorescence decay in CDKL5 KO 187 neurons was due to impaired SV endocytosis. 188 CDD is a disorder of early life, and a therefore key question to address is whether defects can 189 be rescued by the re-introduction of the gene, or whether the altered circuit activity in its 190 absence renders gene correction redundant. To address this in our system, we determined 191 whether expression of WT CDKL5 in KO neurons could correct SV endocytosis deficits. Both 192 CDKL5 KO and WT littermate controls were co-transfected with sypHy and either mCer- CDKL5 193 or an empty mCer vector and stimulated with either 300 APs at 10 Hz or 400 APs at 40 Hz. 194 Analysis of the post-stimulus sypHy response showed that expression of mCer-CDKL5 fully 195 restored the kinetics of SV endocytosis after 10 Hz stimulation and partially after 40 Hz. 196 Importantly, mCer-CDKL5 overexpression had no impact on SV endocytosis kinetics in WT 197 neurons, indicating that increased levels of the protein kinase had no dominant negative 198 effect ( Figure 4A-D). Thus, expression of CDKL5 can restore presynaptic defects observed in 199 KO neurons. previously. In contrast, neither of the CDKL5 mutants were able to correct the SV endocytosis 213 defect ( Figure 5C, E). The absence of rescue was not due to their low expression, since this 214 was equivalent to the exogenously-expressed WT enzyme ( Figure S3). These data reveal that 215 the protein kinase activity of CDKL5 is essential for optimal SV endocytosis kinetics and also 216 associates CDKL5 pathology with defective SV recycling. 217 The kinase activity of CDKL5 is necessary and sufficient for optimal SV endocytosis 218 We have revealed an essential requirement for the enzymatic activity of CDKL5 in SV 219 endocytosis. However a key question to address is whether this activity is both necessary and 220 sufficient to correct SV endocytosis dysfunction in CDKL5 KO neurons. To address this, we 221 examined whether expression of the isolated protein kinase domain was sufficient to correct 222 presynaptic function in CDKL5 KO neurons. To determine this, we generated mCer-tagged 223 deletion mutants of CDKL5 comprising either the kinase domain (ΔC; aa 1-297) or the C-224 terminal tail (Δkinase; aa 298-960) ( Figure 6A). Primary cultures of hippocampal CDKL5 KO 225 neurons were co-transfected with sypHy and either full-length CDKL5 or one of the deletion 226 mutants. Double immunostaining of primary cultured hippocampal neurons for GFP and 227 endogenous CDKL5 suggested that Δkinase was expressed to higher levels than WT, whereas 228 ΔC could not be quantified due to the absence of an antibody epitope ( Figure S3). SV 229 endocytosis kinetics were assessed by monitoring sypHy fluorescence after stimulation with 230 300 APs at 10 Hz or 400 APs at 40 Hz ( Figure 6B, D). We observed that the isolated kinase 231 domain was sufficient to rescue SV endocytosis kinetics similarly to full-length CDKL5 at both 232 stimulus intensities (Figure 6C, E). In contrast, the isolated C-terminus could not, suggesting 233 that this region cannot support SV endocytosis in the absence of the protein kinase domain. 234 Therefore, the ability of the isolated CDKL5 protein kinase domain to correct presynaptic 235 function reveals that it is both necessary and sufficient to rescue SV endocytosis, and that the 236 C-terminal tail is dispensable for this role. 237

CDKL5-mediated phosphorylation at Amph1-S293 is not required for SV regeneration 238
Since the kinase activity of CDKL5 is necessary for optimal SV endocytosis, this suggests that 239 there is at least one CDKL5 substrate at the presynapse that mediates this role. with each other, as it would be anticipated for an enzyme to interact with its substrates, even 244 transiently. We demonstrated reciprocal co-immunoprecipitation of Amph1 and CDKL5 from 245 rat brain lysates ( Figure 7A). This indicates that CDKL5 binds to Amph1 in vivo, and hence 246 supports that Amph1 may be a CDKL5 substrate. 247 Previous studies determined Amph1-S293 as the residue phosphorylated by mediated phosphorylation of Amph1, we generated a rabbit polyclonal phospho-specific 254 antibody against Amph1-S293 ( Figure S4A). To validate this antibody, we generated 255 recombinant GST-conjugated constructs of the central region of WT Amph1 that 256 encompassed this site (residues 248-620, GST-Amph1) and two phospho-mutants, a null (GST-257 S293A) and a mimetic (GST-S293E) and assessed its specificity by Western blotting. This 258 approach revealed that the pAmph1-S293 antibody reacted exclusively with the phospho-259 mimetic GST-S293E (Figure S4B), suggesting that the phospho-antibody is highly specific for 260 phosphorylated Amph1-S293. 261 Amph1 undergoes dephosphorylation coupled to neuronal activity (Bauerfeind et  depolarising conditions, failed to prevent Amph1-S293 dephosphorylation ( Figure S4C). 277 Additionally, we examined the impact of electrical field stimulation, during which neurons 278 were stimulated with 300 APs at 10 Hz or 400 APs at 40 Hz in the presence or absence of the 279 antagonists AP5 and CNQX (which prevent postsynaptic activity or recurrent spontaneous 280 activity). We observed that Amph1-S293 was dephosphorylated after stimulation at both 281 frequencies ( Figure S4D) To assess whether Amph1-S293 is a CDKL5 substrate, WT and CDKL5 KO neuronal cultures 287 were stimulated with 50 mM KCl and allowed to repolarise for different periods of increased 288 duration to determine whether the absence of CDKL5 impacted on rephosphorylation of this 289 residue ( Figure 7B). there was no significant change in the phosphorylation levels of Amph1-S293 either before, 296 during or after the KCl stimulus when compared to WT controls ( Figure 7C). In contrast, 297 phosphorylation of MAP1S-S900 was eliminated in CDKL5 KO neurons in all conditions. This 298 supports the conclusion that Amph1-S293 is not directly phosphorylated by CDKL5 in vivo and, 299 therefore, this phospho-site does not play a significant role in the slowing of SV endocytosis 300 due to CDKL5 deficiency. 301