In recent years, CDKL5 has been found to be critically involved in cell proliferation, neuronal migration, axonal outgrowth, dendritic morphogenesis, and synaptogenesis [1].

To date, at least 368 different pathogenic variants in CDKL5 have been identified and associated with an evolving CDD-related phenotypic spectrum. The main clinical features associated with this spectrum include neurodevelopmental impairment across different domains and early infantile epilepsy with frequent drug-resistant seizures. Over the last few years, genotype–phenotype correlations in CDD have been investigated due to the higher number of individuals and disticting CDKL5 mutations identified and reported in the literature. Based on the present and the previous observations, missense variants within the critical kinase domain and truncating variants disrupting the C-terminal region (after the 781st amino acid) are usually associated with a more severe phenotype, whereas missense mutations in the ATP-binding site and/or truncating variants located earlier in the protein (between amino acid 172 and amino acid 781) have been identified in previously association to milder clinical presentations [6, 12, 13]. Interestingly, the Arg178 codon has been previously identified as a mutational hotspot for missense variants, leading to a severe phenotype with refractory epilepsy and severe neurodevelopmental impairment across all domains. Notably, a relevant phenotype variability has been highlighted, even for variants affecting close regions of the protein, suggesting that the partition in A–D groups is not optimal and rather, the impact of the single variants should be considered. Specifically, the missense variant p.Trp176Arg has been associated with a milder phenotype than the nearby variants affecting the Arg178 codon [12]. Consistently, 38% of our patients with severe refractory epilepsy carried variants previously related to a more severe phenotype. However, the same variants, putatively leading to a non-functional protein, have been found in 50% of our cohort individuals, with the mildest phenotype. Therefore, we have not observed a clear genotype–phenotype correlation in our cohort, even though it represented a small sample. Further possible contributions to the phenotypes of CDKL5 mutation carriers are the effects of modifying genes and stochastic and environmental processes during development. The broad range of epileptic phenotypes associated with CDKL5 mutations identified here is reminiscent of the variable neurological phenotypes reported in association with mutations in CDKL5 reported in the literature [12–14]. However, the kinase domain appears to be crucial in protein function, and variants variably affecting the CDKL5 kinase activity are related to different phenotype severities. Additionally, catalytic domain missense variants and truncating mutations in the C-terminus led to differences in CDKL5 protein localization (constitutively in the cytoplasm and in the nucleus, respectively), with distinct implications on protein function and possibly severe effects on the resulting clinical phenotypes. Furthermore, variants leading to a substitution of physico-chemically divergent aminoacids could disrupt the protein function, also affecting its ability to bind substrates. Therefore, functional studies including quantitative polymerase chain reaction (q-PCR) and western blot (WB) from patient-derived tissues should be performed to precisely assess the impact of CDKL5 variants on protein integrity and stability as a result of different mutations. The presence of mosaicism, (and its influence on phenotypical outcomes), the impact of splicing, and the random X inactivation pattern expressed in neurons fromdistinct brain regions should be investigated as well, to better investigate disease mechanism(s) and elucidate possiblegenotype-phenotype correlations in CDD resulting from individual mutations.

Importantly, early-onset drug-resistant epilepsy is the hallmark of patients with CDKL5 mutation. In our cohort of patients, we studied the characteristics of epilepsy, with particular mention to seizure semeiology (at the onset and in the later stages of the disease) and to seizure frequency at long-term follow-up. Indeed, all patients except one were followed-up by physicians for at least 3 years from the diagnosis, allowing us to dissect the natural history of their epileptic phenotypes, without accounting for the confounding effect of the honeymoon period.

As CDD is globally considered to be drug-resistant epilepsy, we stratified patients into 4 groups according to their long-term seizure frequency: patients with good seizure control, including group 1 with no seizures (only 2 patients) and group 2 with monthly seizures (2/34); patients with intermediate prognosis including group 3 with weekly seizures (9/34) and patients with worst prognosis (group 4) that means daily seizures, from 1 to 20 seizures/day (21/34). The persistence of daily seizures (group 4) at follow-up was considered an unfavorable outcome parameter, for which we searched for possible prognostic factors at onset or clinically relevant associations.

Notably, approximately half of the patients presented seizure onset within the first 2 or 3 months of life, and this is consistent with the proportion of very early-onset CDD cases reported in different cohorts that were previously published [10, 14, 15]. However, the age at seizure onset in our cohort did not statistically relate with a higher incidence of daily seizures at follow-up appointments. Also, a late-onset of the seizures correlates with less severe epileptic presentation in terms of frequency and duration, and the patient with the latest disease onset in our cohort (Individual 7, Table 1) still experienced daily seizures poorly controlled with ASMs. Within the early-onset group (0–3 months), focal motor and non-motor seizures were the most frequent at onset, whereas in the 3–18-month-onset group, tonic seizures and spasms were more prevalent. In general, the majority of patients presented focal motor and non-motor seizures, followed by epileptic spasms, tonic seizure, and myoclonic seizures. Generalized tonic–clonic seizures were less frequent; however, they are more frequent in the later stages of the disease. The subsequent emergence of generalized epileptic seizures in patients initially devoid of such manifestations underscores a dynamic evolution within the epileptic condition. This phenomenon is shaped by a myriad of determinants with the presence of mutations in the CDKL5 gene adding an additional layer of complexity. CDKL5, involved in regulating neuronal activity and brain circuit formation, plays a significant role in epilepsy susceptibility and clinical presentation. Therefore, analyzing the evolution of seizure semiologies in patients with CDKL5 mutations requires consideration of how these specific mutations influence disease pathogenesis and progression. Disease progression remains a pivotal factor, reflecting underlying pathological changes that may lead to an expansion of epileptogenic zones or heightened neuronal hyperexcitability. Alterations in the epileptogenic substrate contribute to the shifting landscape of seizure semiologies, possibly influenced by the specific effects of CDKL5 mutations on brain development and function. Furthermore, the pharmacological efficacy of antiepileptic agents influences the clinical trajectory, as differential responses to treatment may necessitate adjustments in therapeutic regimens, potentially affecting seizure presentation. Beyond intrinsic factors, environmental and lifestyle influences continue to exert notable effects, with stress levels, sleep patterns, and exposure to potential seizure triggers serving as modulators of seizure frequency and semiology, which may interact with the underlying genetic predisposition conferred by CDKL5 mutations. Particularly, within our case series, all patients who subsequently developed generalized tonic–clonic seizures demonstrated a daily seizure frequency, except for one case characterized by weekly seizure occurrences. Approximately 40% of patients presented generalized spasms at the onset and also in subsequent periods. However, it is important to note that the presence of spasms did not correlate with a worse prognosis compared to other seizure semiologies.

The most characteristic feature found in CDD was the presence of a variety of epileptic seizure patterns, which may occur sequentially throughout during lifetime (e.g., onset with spasms, later stages characterized by tonic or focal seizures) or coexist in a given phase of the disease. In fact, about half patients already showed more than one seizure type between focal, tonic, and spasm seizures at the onset of the disease. In some cases, different seizure types could even occur simultaneously in the same electroclinical event, as in the case of HTSS, a sequence characterized by hypermotor seizures, tonic seizures and epileptic spasms, or by analogous sequences described as HTSS-like, in which we can find these 3 phases in different order or just the tonic seizure and spasms clustering together. According to our experience, patients presenting with HTSS or HTSS-like at onset or later in life were characterized by a higher incidence of multiple daily seizures at long-term follow-up. However, due to the overall severity of the CDD and the small sample size, this comparison did not achieve statistical significance. Nevertheless, our experience still leads us to hypothesize that such sequences could serve as potential indicators of a more severe phenotype. Indeed, the greater phenotypic variability in seizures among patients with CDD may reflect a more pronounced involvement of the Central Nervous System (CNS) and, consequently, a higher prevalence of drug-resistant epilepsy among them. Unfortunately, due to the small sample size of males with CDD in our cohort, the phenotypic variables analyzed in this study did not show any correlation with gender.

Regarding EEG features, all patients showed focal, bifocal, or multifocal paroxysmal discharges at the onset or later during lifetime. About 26% of patients showed hypsarrhythmia (classical or modified) at the onset or in subsequent periods and all of them presented spasms alone or in association with other kind of seizures. However, among patients with spasms (16/34), hypsarrhythmia was reported in approximately 60%, a slightly higher value compared to what reported by Demarest and colleagues (47%), but still lower compared to children with infantile spasms of other etiologies (82%) [16]. A recently published study on epileptic spasms in children with CDD found that a poorer response to treatment among CDD patients with spasms was associated with delayed treatment initiation [17]. However, in our case series, no EEG abnormalities, including hypsarrhythmia, were found to correlate with seizure frequency at follow-up.

As we expected, all the cases included in our study showed moderate to severe psychomotor development delay, except one case of developmental regression after the initial achievement of milestones, underlining the difference with Rett syndrome [18, 19].

Similarly, all CDD patients had a variable degree of CVI. Due to the observational nature of the study, it was not possible to standardize the collection of neurodevelopment information and the evaluation of CVI degree for a statistical evaluation. However, from the descriptive data collected, we noted that the only three children with a milder developmental delay (moderate) were the only ones to present a lower degree of visual involvement, with the presence of eye fixation (even if intermittent), eye follow, and absence of nystagmus or dysconjugate gaze. The prognostic value of CVI had already been postulated in the 2019 by Demarest and colleagues, who demonstrated a correlation between the degree of CVI and the severity of developmental impairment, but they did not study the relationship with the epilepsy course [10]. A previous work found out that the ability to walk and use of language were associated with lower rates of current seizures, regardless of the gene variant [3]. Conversely, in our case series, the 3 patients with the least visual involvement and milder psychomotor delay, showed daily current seizures, suggesting an independent trend of these two variables. Obviously, further correlation studies are needed to clear this relationship. To better compare results, it would be necessary using psychomotor development assessment scales specific for CDD that take into account the slow, albeit constant, achievement of milestones during lifetime, as well as to better assess CVI degree, standardized clinical data (i.e., eye fixation and follow-up, nystagmus) should be accompanied with instrumental information (visual evoked potentials).

Our experience confirmed the drug-resistant nature of CDD [20]. No ASM could be considered effective in achieving seizure freedom, neither as monotherapy nor in add on, as expected from literature [21]. In our cohort, up to 30 different ASMs have been attempted to control seizures. At the time of the last evaluation, the most used ASMs were VPA, CLB, CBD, GVG, LTG, LEV, and LCM. The higher percentage of withdrawal interested CZP and other benzodiazepines except CLB. This trend was noted even for medications that had limited and temporary efficacy, such as VGB, ACTH, PHT, PB, TPM, and methylprednisolone in pulse. Interestingly, among our patients, CBD seemed to have a partial effect although none of patients achieved seizure freedom after CBD administration. However, confirming its effectiveness, CBD had the lower withdrawal percentage at ascertaiment compared to the other ASMs.