Cognitive, Neurological, and Behavioral Features in Adults With KCNJ11 Neonatal Diabetes

KCNJ11 gene mutations are the commonest cause of permanent neonatal diabetes (PNDM), which presents in the first 6 months of life and affects 1 in 100,000 live births (1). KCNJ11 is expressed in the pancreas and brain as well as other tissues and encodes the Kir6.2 subunit of the K_ATP channel. In the pancreas, the K_ATP channel links increasing blood glucose to insulin secretion, but activating KCNJ11 mutations prevent channel closure in response to metabolically generated ATP and result in diabetes (2). Clinically, patients present in an insulin-deficient state, and prior to discovery of disease-causing variants in the KCNJ11 gene, they required insulin therapy. It was later shown that KCNJ11 PNDM could be treated with sulfonylurea tablets, which bind and close the channel, allowing insulin secretion, excellent metabolic control, and reduced glycemic variability (3). For many patients and their families, transferring from insulin to oral sulfonylureas vastly improved quality of life in relation to diabetes (4).

Central nervous system (CNS) features occur in children with KCNJ11 PNDM in addition to diabetes. These are thought to result from expression of aberrant K_ATP channels in the brain. The precise role(s) of K_ATP channels in the human CNS has not been fully elucidated, but rodent studies suggest that they play a role in glucose sensing and homeostasis as well as seizure propagation (5,6). KCNJ11 is expressed in many brain areas, but there are particularly high levels of expression in the cerebellum (7,8). The cerebellum is well-known for its role in motor learning and coordination (9), but it also has functions relating to language, executive function, and mood; furthermore, cerebellar abnormalities have been linked with autism (10,11). Documented CNS features in children with KCNJ11 mutations range from subtle neuropsychological impairments that specifically affect attention, praxis, and executive function to the severe and overt Developmental Delay, Epilepsy and Neonatal Diabetes [DEND]/intermediate DEND [iDEND] syndrome (12–15). Other associated features may include psychiatric morbidity, specifically, neurodevelopmental disorders and anxiety disorders, visuomotor impairments, and sleep disturbance (16–18). The severity of the CNS phenotype is related to the genotype. For example, the V59M mutation is frequently associated with iDEND syndrome and neurodevelopmental features, whereas the R201H mutation, previously associated with diabetes alone, has been more recently linked with subtle neuropsychological features (12). Historically, the severity of CNS features was thought to be related to the functional severity of the specific mutation in vitro, although functional interpretation also has to take into account the impact of the mutation on the open probability of the K_ATP channel, which will depend on whether it affects channel gating or ATP binding (19–22).

Sulfonylurea treatment results in partial improvement in the CNS features (23–26) and resolution of functional cerebellar and temporal lobe abnormalities on single-photon emission computed tomography (SPECT) scanning (24,27). The improvement in CNS features may be limited as a result of poor penetration of the sulfonylurea across the blood-brain barrier or active transport back out of the brain, leading to subtherapeutic concentrations in the cerebrospinal fluid (CSF) (28). This, and anecdotal clinical experience of greater CNS response with higher doses of sulfonylurea, has prompted clinical recommendations of glyburide doses of ∼1 mg/kg/day in people with severe neurological features secondary to KCNJ11 mutations (29). However, the neurobehavioral features continue to have a huge impact on families despite sulfonylurea treatment (16). This contrasts markedly with the outstanding metabolic response that changed lives by alleviating the anxiety associated with poor metabolic control (4). A key question is whether the CNS features continue to represent the major burden from KCNJ11 mutations in adult life. To date, all studies characterizing CNS features in KCNJ11 PNDM have been conducted in predominantly pediatric cohorts (12–14,16,23). However, brain development continues beyond childhood and adolescence (30,31). No study has comprehensively assessed the CNS outcomes in adults with KCNJ11 mutations.

Mutations in the INS gene are a less common cause of neonatal diabetes, accounting for ∼10% of cases (1). Heterozygous dominant negative INS mutations often affect protein synthesis, resulting in production of structurally abnormal preproinsulin and proinsulin within the β-cell, endoplasmic reticulum stress, and cell death. Individuals with these mutations also typically present with insulin deficiency but, unlike those with KCNJ11 PNDM, require lifelong treatment with replacement doses of insulin (32). The INS gene is not expressed in any significant levels in the brain; therefore, it is very unlikely that individuals with INS mutations would display a characteristic CNS phenotype as a direct result of their mutations (33). In fact, there have not been any reports of any such neurological issues—in contrast to those with KCNJ11 PNDM.

Individuals with PNDM may have long-term CNS sequelae secondary to diabetic ketoacidosis at diagnosis, as seen in type 1 diabetes (34,35). However, cerebral edema in KCNJ11 PNDM gives rise to a pattern of neurological impairment distinct from that seen as a direct result of brain K_ATP channel dysfunction (36). More subtle neurocognitive problems also occur in the presence of diabetes per se, particularly if metabolic control is poor and diabetes is diagnosed before age 7 years (37). Further, individuals with type 2 diabetes are at increased risk of developing Alzheimer disease in later life, and this may in part be due to chronic metabolic disturbance and changes in insulin signaling (38). Indeed, there is evidence from both animal and human studies that insulin plays a key role in central processes including memory and learning (38). The nonspecific diabetes-related cognitive features could confound assessment of CNS phenotype in people with KCNJ11 mutations; however, people with INS mutations are well placed to control for them. There has been no previous detailed comparison of the CNS phenotype in people with INS and KCNJ11 mutations.

The aim of this study was to characterize the neurological and neuropsychological features in adults with KCNJ11 PNDM compared with these features in adults with INS PNDM.

Ethics approval was obtained from the National Research Ethics Service Committee South West–Exeter.

We identified 34 patients >16 years old with KCNJ11 mutations and 9 patients >16 years old with INS mutations who had received a molecular genetic diagnosis in Exeter and who had been diagnosed with permanent neonatal diabetes before 6 months of age. We approached potential participants either directly at a neonatal diabetes family event in Exeter or via the consultants in charge of their clinical care. We invited 17 individuals with KCNJ11 mutations to join the study; of these, 10 agreed to participate. However, two individuals were excluded from the analysis owing to possible confounding factors: one individual (mutation L164P) was excluded because he was taking antipsychotic medication to treat a psychotic illness at the time of the study and had had a particularly severe initial presentation with diabetic ketoacidosis and 3 days in a coma, and a second individual (mutation V59M) was excluded owing to severe neurological impairment following initial presentation with diabetic ketoacidosis (further clinical characteristics of excluded participants are available in Supplementary Table 1). We approached nine individuals with INS mutations, and four agreed to take part. All participants were from the U.K. apart from one, who was from Canada.

All participants were visited at home or assessed in the Exeter Clinical Research Facility by the same consultant neurologist and consultant clinical neuropsychologist who carried out the history taking using a standard pro forma, neurological examination, neuropsychological assessments, mood questionnaire, and neurodevelopmental screen. If possible, an informant or caregiver was also present to facilitate information gathering. The severity of intellectual impairment and behavioral disturbance in one individual (KCNJ11–8 [V59M]) meant that it was not possible for him to attempt any of the cognitive tests. Another individual (KCNJ11–6 [V252G]) did not wish to attempt the Controlled Oral Word Association Test (COWAT) and was unable to understand instructions for the Color Trails Test (CTT). In eight participants (four with KCNJ11 mutations and four with INS mutations), T2-weighted brain MRI scans were performed using a 1.5 Tesla MRI scanner. The scans were reviewed and interpreted by a radiologist and a neurologist who were blinded to the mutation status of the individuals concerned.

Participants and informants were asked for a standard medical history, the ages at which major milestones were attained, whether learning support was required, and level of education/employment.

A full neurological examination was performed. This included assessment of cranial nerves, limb tone, power, reflexes, coordination, and sensation and simple tests of motor sequencing and praxis, comprising two tests of bimanual coordination, one unilateral motor sequencing task (the Luria three hand position test), and copying unfamiliar hand positions and manual miming—both tested in each hand.

Current psychological distress was assessed using the Hospital Anxiety and Depression Scale (HADS) questionnaire. The Autism Spectrum Quotient (AQ) was administered to screen for autistic traits.

A battery of neuropsychological tests was administered to assess a variety of cognitive domains. The Wechsler Abbreviated Scale of Intelligence (WASI) was used as a brief measure of current intelligence quotient (IQ). The Verbal Paired Associates and Visual Reproduction subtests of the Wechsler Memory Scale (WMS-IV) were used to give a verbal and nonverbal (visual) measure of memory. Subtests of the Wechsler Adult Intelligence Scale, fourth edition (WAIS-IV), were administered: cancellation to assess processing speed and digit span (forward and backward) to assess working memory. Subtests of the Visual Object and Space Perception battery (VOSP) assessed visuospatial function: incomplete letters and object decision to test object perception and dot counting and cube analysis to test spatial perception. The COWAT was used to assess aspects of executive function including verbal fluency, self-monitoring, and ability to assimilate and adhere to stipulated rules. The CTT-1 and -2 were used as measures of sustained and divided attention and hand-eye motor coordination and speed. Finally, the Addenbrooke’s Cognitive Examination-Revised (ACE-R) was used as a broad screening measure of cognition, providing an assessment of the following cognitive domains: attention/orientation, memory, fluency, language, and visuospatial function.

The Cambridge Behavioural Inventory revised (CBI-R) was used to complement the information obtained from the history taking. This measure seeks the opinion of the informant, e.g., caregiver or family member, on the frequency of a range of behaviors in the domains of memory and orientation, everyday skills, self-care, abnormal behavior (e.g., tactlessness, impulsiveness), mood, unusual beliefs, altered eating habits, disturbed sleep, stereotypic and motor behaviors, and altered motivation. For each behavior, the informant assigned a score of 0–4 based on the frequency: scores of 3 (occurring daily) or 4 (occurring constantly) denote a significant behavioral deficit.

Data were analyzed using Excel 2010 and Stata 14. Qualitative data were presented descriptively. Where population normative data were available, neuropsychological test scores were converted to Z scores. For VOSP subtests, a pass was a score ≥5th population percentile. For comparison of characteristics and outcomes between the KCNJ11 and INS groups, data were analyzed using nonparametric methods (Mann-Whitney U test for numerical variables and Fisher exact test for categorical variables). Data are presented as median [range] unless otherwise stated.

Baseline clinical characteristics of the participants are outlined in Table 1; these were similar between individuals with KCNJ11 mutations and individuals with INS mutations.

Abnormalities on neurological examination were identified in seven of eight KCNJ11 participants and only one INS participant (Table 2).

Developmental histories, level of educational support, and employment history reported for each participant are described in Table 2. Developmental delay or learning difficulties were present in all KCNJ11 participants, and they continued to require high levels of support as adults. In contrast, the INS group did not report major learning difficulties, in keeping with their subsequent employment history and independence in adulthood.

Four of eight with KCNJ11 mutations, but none of the participants with INS mutations, had features of autistic spectrum disorder either via a clinical diagnosis of autism or an AQ score at or above the threshold suggestive of clinically significant autistic traits (Table 2 and Supplementary Table 3). Two individuals in the KCNJ11 group and one in the INS group required treatment either at the time of the study or in the past for depression or anxiety. HADS scores for anxiety and depression were similar in KCNJ11 versus INS participants (Supplementary Table 3). One individual in the KCNJ11 group and two individuals in the INS group scored above the HADS clinical threshold (11) for anxiety (Table 2 and Supplementary Table 3).

IQ was lower in the KCNJ11 group versus the INS group (IQ 76 [55–101], n = 7, and 111 [90–124], n = 4, P = 0.02). Three individuals in the KCNJ11 group had an IQ <70 (Supplementary Table 2) and impairments in adaptive behaviors in keeping with a clinical diagnosis of intellectual disability (39). Five of seven individuals in the KCNJ11 group scored below the clinical cut point for cognitive impairment on the ACE-R (Supplementary Fig. 1). CTT-1 scores suggested reduced attention (CTT-1 Z score −1.7 [−3.0 to −0.1], n = 6, vs. 0.4 [−1.1 to 1.2], n = 4, P = 0.03, and CTT-2 Z score −0.8 [−3.0 to 0.8], n = 6, vs. 0.7 [−1.0 to 1.2], n = 4, P = 0.13) (Fig. 2 and Supplementary Table 2).

In the KCNJ11 group but not the INS group, median scores on the WASI, WAIS, and WMS were below population average in all subtests apart from the verbal paired associates subtest of the WMS (Fig. 1). Scores were particularly low (≤2 SD below population average) in the matrix reasoning component of the WASI (Z score −3.2 [−4.8 to −0.9] vs. 0.6 [−0.7 to 0.8], P = 0.008) and the digit span component of the WAIS-IV (Z score −2.0 [−3.0 to 0.3] vs. 0 [−1.0 to 0.3], P = 0.046). Cancellation scores, although not as markedly reduced compared with population norms, were significantly lower in the KCNJ11 group (Z score −1 [−3 to 0] vs. 2.8 [0.7–3.0], P = 0.007). COWAT and VOSP scores showed a trend toward reduced executive function and visuospatial function, respectively, in the KCNJ11 group (Supplementary Table 3), although these did not reach statistical significance.

In the KCNJ11 group, 6 individuals had severe behavioral features, which clustered in the domains of everyday skills (5 of 6), stereotypic behavior (5 of 6), memory and orientation (4 of 6), abnormal behavior (3 of 6), mood (3 of 6), and motivation (3 of 6). Specific everyday skills highlighted included writing (3 of 5) and dealing with money/bills (2 of 5). The most frequent stereotypic behaviors were being rigid/fixed (3 of 5) and having fixed routines (4 of 5). Poor concentration was highlighted as a specific feature in all four individuals who had memory and orientation problems. Two individuals had significant difficulties with self-care, and two reported disturbed sleep.

Structural brain MRI was normal in participants with KCNJ11 mutations. Of INS control subjects, two had normal results and two had minor abnormalities that were not clinically significant (Table 2).

Performance on the cognitive tests was better in the two individuals with the R201H mutation. These individuals consistently had scores equal to or greater than the KCNJ11 group medians (Figs. 1 and 2 and Supplementary Table 2), and no significant behavioral features were reported.

We have characterized for the first time the profile of neurological, neuropsychological, and behavioral features present in adults with PNDM due to KCNJ11 mutations. The key features were learning difficulties, features of autism spectrum disorder, subtle motor deficits affecting coordination and motor sequencing, and reduced IQ. Specific cognitive domains most affected were perceptual/nonverbal reasoning, working memory, and attention, with a trend toward executive dysfunction and impaired visuospatial abilities. Verbal paired associate memory was relatively preserved. The impact on everyday functioning was significant; two participants were severely impaired, requiring support with activities of daily living. A comparison group of patients with neonatal diabetes of similar duration due to INS mutations did not show any of these specific features, indicating that they are unlikely to be a nonspecific effect of metabolic disturbance from birth. Furthermore, as both groups were insulin treated from diagnosis as infants, the differences observed are unlikely to have been influenced by variation in the timing or duration of action of insulin on insulin and insulin-like growth factor receptors in the brain.

Our findings are consistent with studies in pediatric cohorts with KCNJ11 neonatal diabetes. Specifically, the motor features noted on neurological examination, together with impairments of attention, working memory, visuospatial ability, and executive function, are consistent with the previously reported high prevalence of developmental coordination disorder, inattention, executive dysfunction, and poor visuomotor performance in children with KCNJ11 mutations (12–14,17,18,23,40). Autism spectrum disorder features in four individuals with KCNJ11 mutations are consistent with previous research reporting high rates of neurodevelopmental disorders in affected children (16,18,41). Dyspraxia, visuomotor impairment, autism, and impaired executive function may be related to the high levels of expression of dysfunctional K_ATP channels in the cerebellum in KCNJ11 PNDM (7,10).

Importantly, the abnormal findings that we report in the KCNJ11 group were mutation specific; the two individuals with the R201H mutation had no overt features and only some subtle abnormalities on neuropsychological testing (Supplementary Table 2). As a result, both were able to live independently and support themselves financially. Those with more severe features required high levels of support from family members and professionals in health care, education, and social care. This is consistent with previous studies showing that the severity of the CNS phenotype is related to the specific mutation; for example, the V59M mutation results in more severe features, greater impairment in daily living skills (13), and greater impact on families (16). Interestingly, however, there was a relatively good level of social integration from all patients with KCNJ11 mutations even when the neurobehavioral features were severe.

Some of our findings contrast with previous research. To our knowledge, choreiform movements have not previously been associated with KCNJ11 PNDM but were observed in one individual with the V59M mutation in our study. We did not identify abnormal tone in our cohort, which contrasts with the hypotonia previously reported, particularly in the context of DEND/iDEND syndrome (42). This may be explained by seven of eight individuals in our study being sulfonylurea treated; improvement in tone to near normal following transfer from insulin to sulfonylureas has been observed in a recent study of children with KCNJ11 PNDM (23). Similarly, improvement of visuospatial abilities and attention following transfer to sulfonylureas was noted in the pediatric study (23), which could account for the attention deficits and visuospatial impairment being less marked in our cohort of sulfonylurea-treated adults than might have been expected given previous descriptions (17,23). Our neuroimaging findings contrast with this pediatric study in which there were nonspecific findings in 12 of 17 who underwent brain MRI, largely comprising white matter abnormalities (23). However, these scans were performed at baseline prior to transfer to sulfonylureas (23). It is not known whether the abnormalities would have improved after a period of sulfonylurea treatment, as has been shown in SPECT studies (24,27).

Sulfonylurea treatment may influence the CNS phenotype in KCNJ11 PNDM. Two studies have suggested that an earlier age of initiation of sulfonylureas can lead to better CNS outcomes (17,23). We were unable to assess this in our study because median age at transfer to sulfonylureas was 18 years (range 11–34). However, the persistence of CNS features in some patients even after early initiation of treatment (16,23) suggests that other factors are involved. Specifically, active transport of glyburide out of the brain across the blood-brain barrier, as has been demonstrated in a rodent model (28), may result in suboptimal concentrations in the CSF, thereby limiting therapeutic efficacy in the human CNS. Anecdotal clinical experience suggests that this can be partially addressed by increasing the dose of glyburide to ∼1 mg/kg/day; however, there have been no cases of complete resolution of CNS features in a patient with iDEND. Another possible reason for the partial response is that pathways that can fully restore K_ATP channel function in other tissues are not available in the CNS to interact with brain K_ATP channels. For example, restoration of pancreatic K_ATP channel function resulting in excellent glycemic control with sulfonylurea treatment is dependent on the activity of incretin hormones (3). Furthermore, there is a theoretical impact of insulin deficiency in utero or C-peptide deficiency prior to sulfonylurea transfer on the brain as an indirect consequence of KCNJ11 mutations, but more studies are needed to explore this in humans. Indeed, given the complexities of human neurodevelopmental processes, it is likely that several factors contribute in some way to the response of CNS features to sulfonylureas in KCNJ11 PNDM.

This study has important strengths. It is the first to assess in detail the CNS manifestations of KCNJ11 mutations in adults and to control for the nonspecific effects of PNDM by comparing the features in individuals with INS mutations. Limitations of the study, which relate to the rarity of the disease, are the small number of individuals in each group, the broad range of mutations studied, and the variable timing of initiation and duration of treatment with sulfonylureas in the KCNJ11 group. Studies in larger cohorts with single specific mutations would be valuable. Furthermore, exploration of the impact of treatment-specific factors, such as age of initiation, dose, and CNS handling of sulfonylureas in humans, on CNS features in KCNJ11 PNDM is warranted.

The CNS phenotype in adults with KCNJ11 mutations comprises learning difficulty, autistic features, subtle motor dysfunction, moderately reduced IQ, and impaired attention, perceptual reasoning, and working memory. The severity of these features varies with the causative mutation. They persist despite long-term sulfonylurea therapy, at least when this is started after the first decade of life, and represent the major burden from KCNJ11 mutations once glycemia is well controlled on sulfonylureas. These CNS features are not present in individuals with INS mutations, which indicates that they occur not as a result of the lifelong metabolic disturbance imposed by PNDM but, rather, as a consequence of impaired K_ATP channel function in the brain. Clinicians in adult and pediatric medicine should be aware of the potential impact of CNS features in patients with KCNJ11 mutations and should consider multidisciplinary management to ensure appropriate support is provided.

Acknowledgments. The authors thank the individuals with neonatal diabetes and their families who participated in the study; staff in the Exeter Clinical Research Facility who supported the study; Dr. James Taylor, consultant radiologist, who interpreted the MRI scans; and Dr. Jon Fulford, who provided technical support with MRI scanning.

Funding. A.T.H. is supported by a Wellcome Trust Senior Investigator award (grant 098395/Z/12/Z). P.B. has a Sir George Alberti Clinical Research Training Fellowship funded by Diabetes UK (grant 16/0005407). S.E.F. has a Sir Henry Dale Fellowship jointly funded by the Wellcome Trust and the Royal Society (grant 105636/Z/14/Z). M.H.S. and B.A.K. are supported by the Exeter National Institute for Health Research Clinical Research Facility.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. P.B. drafted the article. P.B., J.D., L.T., M.H.S., B.A.K., T.J.F., S.E.F., A.C., A.T.H., and A.Z. approved the final version prior to submission. P.B., J.D., L.T., M.H.S., T.J.F., S.E.F., A.T.H., and A.Z. contributed to analysis and interpretation of data. P.B., J.D., L.T., M.H.S., S.E.F., A.T.H., and A.Z. contributed to data acquisition. J.D., L.T., M.H.S., B.A.K., T.J.F., S.E.F., A.C., A.T.H., and A.Z. critically revised the manuscript. L.T., B.A.K., A.C., A.T.H., and A.Z. contributed to the conception and design of the study. P.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.