Abstract
Deficits in the inwardly rectifying K+ channel, Kir4.1, cause SeSAME syndrome, autism spectrum disorders with seizures, multiple sclerosis, Huntington’s disease and Rett syndrome. Understanding how deficits in a single gene can accomplish such diverse neurological symptoms, requires that we know the functional role of Kir4.1 in the nervous system. We used whole exome sequencing (WES), channel physiology in patient-specific lymphoblastoid cells (LCLs) and established a Drosophila model to examine the functional effects of a KCNJ10 variant identified in SeSAME-like family. WES analysis of six related affected individuals, born out of consanguineous parentage, identified KCNJ10T290A variant residing within a long homozygous stretch in Chr1. The membrane potential of LCLs carrying KCNJ10T290A variant were significantly depolarized and exhibit defects in potassium permeability, thus demonstrating a loss-of-function effect. Drosophila irk2 mutant, a human homolog of Kir4.1, exhibited impairment locomotion, shortened life-span and age-dependent degeneration of dopaminergic neurons in the adult brain. Finally, neural specific expression of either Kir4.1 or Irk2 alleviate irk2 mutant phenotypes, while the Kir4.1T290A and Irk2T290A mutant proteins failed to do so. These results imply the functional conservation of Kir4.1 across species thereby elevate the potential of using Drosophila model to improve our understanding of the SeSAME syndrome. Our results emphasize the importance of Kir4.1 in regulating extracellular K+ homeostasis, which is central to patterning of nervous system during development.
Introduction
SeSAME syndrome (OMIM#612780), characterized by Seizures, Sensorineural deafness, Ataxia, Mental retardation and Electrolyte imbalance is a rare, autosomal recessive and multisystemic disorder. Otherwise known as EAST (Epilepsy, Ataxia, Sensorineural deafness, Tubulopathy) syndrome, SeSAME is predominantly caused by homozygous or compound heterozygous mutations in KCNJ10 gene (Bockenhauer et al., 2009; Scholl et al., 2009) encoding Kir4.1, an inwardly rectifying potassium channel with diverse expression patterns in multiple cell types of central and peripheral nervous system (Reichold et al., 2010; Hibino et al., 2010; Paulais et al., 2011; Thompson et al., 2011; Kelley et al., 2018; Larson et al., 2018; Song et al., 2018). Depending on tissue localization and assembly of Kir4.1 subunit, which can constitute homotetramers and/or heterotetramers with Kir5.1 (KCNJ16), these channels exhibit distinctive physiological properties (Paulais et al., 2011; Pessia et al., 2001). Kir4.1 channel play conspicuous roles in the maintenance of resting membrane potential (Kofuji et al., 2000), facilitation of glutamate uptake (Djukic et al., 2007), potassium siphoning by glial cells (Neusch et al., 2006; Song et al., 2018), cell volume and peak strength regulation of fast α-motor neurons (FαMNs; Kelley et al., 2018), axonal integrity through myelination by oligodendrocytes (Neusch et al., 2001; Schirmer et al., 2018; Larson et al., 2018) and cell migration (dehart et al., 2008). Dysfunction of Kir4.1 has been associated with a spectrum of neurodegenerative conditions like idiopathic epilepsy (Buono et al., 2004; Heuser et al., 2010; Lenzen et al., 2005), autism spectrum disorder with seizures (Sicca et al., 2011; Sicca et al., 2016; Larson et al., 2018), Huntington’s disease (Tong et al., 2014), multiple sclerosis (Srivastava et al., 2012; Brickshawana et al., 2014; Nwaobi et al., 2016; Gu et al., 2016) and Rett syndrome (Olsen et al., 2015; Kahanovitch et al., 2018).
Given the genetic redundancy and compensation effects of mammalian Kir channel proteins, invertebrate Drosophila model system offers an excellent choice to investigate the physiological functions of Kir mutant proteins. The Drosophila genome encodes only three Kir channel proteins; Irk1 (Kir2 family), Irk2 (Kir4 and Kir5 family) and Irk3 (Kir1 family) and are shown to be critical for patterning of nervous system during development (Chen and Swale, 2018). In contrast, fifteen Kir channels have been identified in mammals which are grouped into seven families based on sequence similarity and functional properties (Hibino et al., 2010). Several Kir channels (Kir2.1, Kir3, Kir4 and Kir6) that are primarily expressed in thalamus, cortex, brainstem and hippocampus are associated with epilepsy (Higashi et al., 2001; Villa and Combi, 2016). Kir channels display greater inward K+ flow at negative resting membrane potential to equilibrium potential for K+ (Ek), while at more positive membrane potentials, outward flow of K+ is inhibited by intracellular Mg2+ and polyamines (Lopatin et al., 1994). Overall Kir channel allows K+ ions to pass easily inward than outward direction.
We identified a novel homozygous mutation in KCNJ10 by WES of all six affected members of a family diagnosed with SeSAME syndrome. Next, using a series of whole-cell patch clamp experiments in patient-specific LCLs, we found that the variant compromised the membrane physiology of Kir4.1. Furthermore, to investigate its role in the Drosophila nervous system, we isolated and characterized irk2 mutant alleles. irk2 mutants exhibited shortened-life span, stereotyped locomotor patterns and age-dependent degeneration of dopaminergic neurons in the adult brain, which are in part, analogous to the neurological phenotypes observed in SeSAME syndrome. Finally, we found that the neural specific overexpression of mutant proteins; neither Irk2 nor Kir4.1 failed to alleviate irk2 whole mutant phenotypes unlike the wild type proteins demonstrating the loss-of-function of mutant Kir4.1T290A. These findings emphasize the crucial physiological role of Kir4.1 from invertebrates to mammals.
Materials and Methods
Patient recruitment, genomic DNA isolation and generation of lymphoblastoid cells
The subjects were identified through the clinical services of NIMHANS. Various members of the family had been identified over many years, and the clinical information was collated. In addition, other branches of the family were identified and invited to participate in the study. The study protocol was approved by Institutional Human Ethics Committee and Institutional Stem Cell committee at National Institute of Mental Health and Neurosciences (NIMHANS), Bangalore, India. Patients were referred for biochemical evaluation and selected for further analysis by presence of seizures, ataxia, mental retardation, hearing impairment. Five ml of venous blood was collected in EDTA treated tubes (Becton-Dickinson) using standard procedures and genomic DNA was isolated using NucleoSpin® Blood L (Macherey-Nagel GmbH & Co. KG) for whole exome sequencing (WES). In addition, six ml of venous blood was collected separately in sodium heparin tubes (Becton-Dickinson) for isolation of peripheral blood mononuclear cells (PBMC) using SepMate™ tubes (STEMCELL Technologies) and histopaque-1077 (Sigma Aldrich) as a density gradient. For generation of lymphoblastoid cells (LCLs), PBMCs were transformed by Epstein Barr virus (EBV) using a standard protocol (Hui-Yuen et al., 2011). All six LCLs suspensions were cultured in RPMI 1640 medium pH 7.4 supplemented with 10% fetal bovine serum (FBS), 4 mM glutamine and streptomycin/penicillin antibiotic cocktail. Cells were maintained in a humified atmosphere of 5% CO2, 95% air at 37°C, the medium was replaced every 3rd day and cells were passaged once a week. Stabilized LCLs were further karyotyped using the G-banding approach (Bradley Howe et al., 2014), STR profiling [GenePrint® 10 System (Promega)] using blood and LCL DNA was done to confirm sample identity.
Whole exome sequencing, variant calling, quality check and annotation
The blood DNA was extracted for 10 samples [unaffected parents, (n=4), and affected offsprings, (n=6)], and the library was prepared using Nextera Rapid Capture and Expanded Exome Kits. The library was further subjected to WES, performed on Illumina Hi-Sequencer to generate pair-end reads (150bp*2). We followed whole exome sequence analysis pipeline used by Suhas et al., 2018. FastQC (v0.11.5) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc) was used for the quality of raw reads, which examine per base and per sequence quality scores, per base and per sequence GC content, per base N content and sequence length distribution. Prinseq-lite-0.20.4 tool was used to trim poor quality region (http://prinseq.sourceforge.net/) and adapterremoval-2.1.7 (Schubert et al., 2016) was used to remove adapter contamination in raw reads. Filtered reads with a quality score (Q)>20 were aligned to the human reference genome hg19 (GRCh37) using BWA (v0.5.9) (Li and Durbin, 2009). SAM to BAM conversion and sorting were done with Samtools 1.3 tool (https://sourceforge.net/projects/samtools/files/samtools/1.3/). Then the PCR duplicates were removed using PICARD tools (v1.96) (https://broadinstitute.github.io/picard/) and the INDELS were realigned using GATK (v3.6) (McKenna et al., 2010). The BAM alignment was subjected to QC using Qualimap (v2.2) (Okonechnikov et al., 2016). VarScan (v2.3.9) (Coverage=8, MAF>=0.25, p-value<0.001) was used to call for SNPs and INDELS (Koboldt et al., 2013). The quality of VCF file was checked using RTG tools 3.7.1 (https://github.com/RealTimeGenomics/rtg-tools/releases). All samples annotation was performed using ANNOVAR tool (Wang et al., 2010). Population controls (n=7) representing three religious groups (Group A, B, and C) matched for age, sex and ethnicity, were obtained from INDEX-db (Ahmed et al., 2018). All controls passed the age of risk i.e., 45 years, for neuropsychiatric illnesses, except for the outbred Parsi (religious group 3) individual (age=26), who was included as an outlier. All the controls were of southern Indian ethnic origin except for the Parsi. To validate KCNJ10 variant identified by whole exome sequencing, we performed Sanger validation using the following gene specific primers: Forward (CATTCGTTTCAGCCAGCATGC) and Reverse (TCAGACATTGCTGATGCGCA).
Assessing runs of homozygosity (ROH)
Exome-wide F-statistics was calculated using the --het option in vcftools (v0.1.5), for every sample, to investigate whether levels of heterozygosity differed between the affected siblings, unaffected parents and population controls (Danecek et al., 2011). Runs of homozygosity (ROH) was detected in all samples using --homozyg option in PLINK (v1.9) (Purcell et al., 2007). The minimum length for a tract to qualify as ROH was set to 500kb and the minimum number of variants constituting an ROH was set to 100. A maximum of 3 intervening heterozygous variants were allowed within a ROH window. ROH density was set to default i.e., an ROH must have at least one variant per 50kb, on an average. The centromeric, X, Y and mitochondrial variants were ignored during this analysis. The stretches that were shared between all the affected individuals but not observed in either of the parents or the population controls were thus notified as ROHaffected, which were identified by using a combination of intersect and subtract functions in bedtools (v2.22) (Quinlan and Hall, 2010). The variants were annotated using variant effect predictor (VEP GRCh37) (McLaren et al., 2016).
Experiments with lymphoblastoid cells Whole-cell patch clamp electrophysiology
For all electrophysiology experiments, the following LCLs were used. One healthy wild type control (not related to the family), two unaffected parental controls (III.11 and III. 12) and four affected individuals (IV.2, IV.3, IV.4, IV.5) from SeSAME-like family described in this study. Cells in suspension are resuspended in fresh RPMI culture media and dissociated to achieve at least 40% of single cells in suspension. Thereafter the cells were plated in 15mm diameter cover slips (acid-treated, ethanol washed and air dried) coated with poly-D-lysine (A-003-M EMD Millipore). Cells were kept in culture incubator (conditions as described above) for half hour before recordings. After establishment of whole-cell configuration, we immediately measured the membrane potential (Vm) of LCLs. A pulse protocol was applied in which Vm held at resting membrane potential and then stepped to test potentials between −120 to 40mV in 10mV steps for 140ms. To measure membrane currents (nA) we used a single electrode whole-cell patch clamp technique. For intracellular voltage-clamp recordings and positioning of perfusion micropipette, we used two Narishige hydraulic micomanipulators (MNW-203, Narashige Japan). Perfusion pipette with a tip diameter of 30-50 um was positioned close to the tip of patch pipette to facilitate rapid application of test solutions especially for experiments with increase in extracellular concentration of potassium. Electrodes for whole-cell patch experiments were pulled from borosilicate glass capillaries (IB150F-3, World Precision Instruments, USA) with resistance of 4-6megaohms in two step protocol using Sutter P-97 puller (Sutter Instruments Corp., Novato, CA). Micropipettes were filled with intracellular solution containing 120mM potassium D-gluconate (G4500, Sigma), 1mM MgCl2, 15mM KCl, 1mM CaCl2, 10mM EGTA, 10mM HEPES with 7.2 pH adjusted with KOH/HCl. After obtaining whole-cell mode, access resistance was 10-15megaohms. The extracellular recording solution contained 130mM NaCl2, 3mM CaCl2, 2.5mM MgCl2, 15mM HEPES with 7.4 pH adjusted with NaOH. In experiments, where LCLs were puffed with high extracellular K+, KCl was varied from 5-20 mM while NaCl was decreased to 110mM to adjust osmolarity. Recordings in LCLs, were performed using an HEKA triple patch clamp amplifiers (EPC 10 USB) at RT. To determine specificity of Kir4.1 current, we used 110μm/l BaCl2 and to block endogenous Cl- currents, we used 150μm/l niflumic acid in the bath solution. The pClamp 9 (Axon Instruments) software package was used for data acquisition and analysis. Data were analysed using Molecular Devices Clampfit 10.4 (Sunnyvale, CA) and Microsoft Excel programs. The results were presented as mean+S.E. For statistical analysis we used GraphPad Prism (San Diego, USA). To choose between parametric or non-parametric tests for normality criteria, we used Shapiro-Wilk estimator. For data sets with small N, we used non-parametric tests to avoid possible type II errors. Mean differences were statistically evaluated using ANOVA with Levene’s homogeneity of variances test and pairwise comparisons were made using Turkey adjustment. Non-parametric k independent Kruskal-Wallis test was applied with Bonferroni correction to compare the differences among means. Error bars represent +S.E.
Immunofluorescence
One mL of 70% confluent LCLs in suspension were collected in 1.5mL Eppendorf tube followed a brief centrifugation at 2000rpm for 3 min to collect cells down. Cells were washed briefly once with 1X Phosphate buffer Saline (PBS) and fixed in 4% paraformaldehyde (PFA: P6148, Sigma) for 10 min at RT. Remove PFA traces by washing cells twice with 1X PBS. Primary antibody incubation was done overnight at 4°C in block solution (0.2% BSA in PBS with 0.02% Triton-X100), while the secondary antibody incubation was done for 2 hrs at RT. Intermediate washing steps were done with PBT (PBS+0.02% Triton-X100) for 30 min. The following antibodies were used; hKir 4.1 (1:1000; Cat no: NBP1-20149, Novus Biologicals), Alexa Fluor 488 phalloidin (1: 1000; A12379: Invitrogen), DAPI (1:5000; D9542: Sigma) and goat anti-mouse IgG Alexa Fluor 488 (1:1000; A11001; Invitrogen). Cells were mounted using Vectashield antifade mounting medium (H-1000: Vector labs) to prevent photobleaching and left overnight at 4°C. Optical z-sectioning at 0.2 μM intervals was done using Plan-Apochromat 63x/1.40 oil objective in Zeiss Axio Observer 7 with Apotome 2 feature and Axiocam 702 monochrome camera (Carl Zeiss, Germany). Signal-to-noise ratio was improved using in-built Zeiss deconvolution module and MIP projections of 2-3 Z-stacks are presented here. Representative images reported here are from three independent experiments. For quantitative measurements, deconvoluted Z-stacks were first blinded before analysis. 3D surface rendering plugin in Imaris software is used to reduce signal-noise ratio to measure Kir4.1 punctae distribution between cytoplasm and nucleus. The respective numbers were normalized against cytoplasmic space marked by F-actin and nuclear space by DAPI signals.
Experiments with Drosophila Drosophila genetics and molecular biology
Flies were maintained on standard medium at 25°C. w1118 strain was used as wild-type control. A EP insertion in irk2 locus (G8696) obtained from Bloomington Stock Center is imprecisely excised to generate irk2 mutant alleles by standard procedures. Following fly stocks were used in this study: C155-Gal4 (Lin and Goodman, 1994), TH-Gal4 (Friggi-Grelin et al., 2003), UAS mcd8-eGFP (Jan lab). To assess the effect of irk2 mutant on the expression levels of irk2, we extracted RNA from embryos (for homozygous irk2e118) and larvae (homozygous irk2e91b and heterozygous irk2e118/e91b) and reverse-transcribed into cDNA. Following primers were used to investigate Irk2 expression levels; CGACGAGAACTCACCGCTCT and AGGGATTCCGCACACCAGCT. Full-length cDNA for Irk2 was amplified using cDNA specific primers [ATGCAGGTCCCGCTCAGCGA and GCAGCTCTCGGTGCCCTTGTAA] and Kir4.1 cDNA obtained from OriGene (SC118741). Site-directed mutagenesis was performed with Irk2 and Kir4.1 cDNA using GeneArtTM site-directed mutagenesis system (A13282: Invitrogen) to generate Irk2T290A and Kir4.1T290A cDNAs. All four cDNAs, described above, were subcloned into fly pUAST transformation vector (Brand and Perrimon, 1993) using Not1 site and used to generate transgenic flies in w1118 genetic background.
Larval locomotor assay
Five wandering third-instar larvae, staged appropriately, are placed on 1% agar plate. The total area available for larvae to crawl around is 6 cm (inner diameter of plate). Before recording their locomotor patterns, they were acclimatized for 2 min in the testing arena and recorded for 5 min using 13-megapixel Sony digital camera. The distance from agar plate to camera is 25cm and the entire chamber is enclosed to prevent external light. A LED light source was used underneath agar slabs. The video file was converted into image stacks of 3 fps for 3 min resulting in 5400 frames. Eight such videos were taken from three replicate bottles resulting in 24 videos per group and 120 individual larvae per group. These stacks of images were then analysed using wrMTrck software standardized for Drosophila larval locomotion (Brooks et al., 2016).
Live imaging of adult locomotion using flyBowl assay and vertical climbing assay
The schematic of testing arena and the methods to assay adult locomotion are reported in Simon and Dickinson, 2010 (flyBowl assay) and Aggarwal et al., 2018 (vertical climbing assay). We used 3-5 days old flies, aged at 23°C at 12-12 LD cycle. The videos were captured at 250FPS at 0.055mm/pixel resolution using Pointgrey camera (13Y3M) with a Canon 18-55ES lens at 55um/pixel resolution. Total distance traveled: Distance traveled by a fly is equal to the sum of all track lengths in body length units (BLUs) in a given time duration. For a genotype, total distance traveled in a given time is calculated as the mean of total distance traveled by each fly of that genotype. Average speed: For each track, the speed of the fly is calculated as the average instantaneous speed for that track. Further, mean speed of a fly is given by the mean of speeds for all tracks. Finally, for a genotype, average speed is calculated as the mean of mean speed of each fly. For fly bowl Assay: The average speed of the fly was measured by dividing total distance traveled by a fly in a given amount of time. The videos were captured at 15FPS at 0.064mm/pixel resolution. Live imaging done using FlyCon Tra (Fly Contour based Tracker) software (Aggarwal et al., 2018). Statistical analysis of all raw data gathered over time for both fly bowl and vertical climbing assay was done using GraphPad Prism (San Diego, USA). Graphs were plotted using matplotlib library in Python. Unpaired t-test (for normal distribution) or Mann-Whitney test (for non-normal distribution) was used for comparison between two genotypes. For comparison between more than two genotypes, we used two-way repeated measures ANOVA, post-hoc Sidak’s multiple comparisons. Data is reported as mean+SEM for all experiments.
Results
Clinical and biochemical features of family with SeSAME syndrome
Six affected individuals were identified from the relatives of an index subject (IV.2) who had sought help with the symptoms of seizures, ataxia and developmental delay. The clinical features were broadly similar to SeSAME syndrome but without electrolyte imbalance (Fig. 1A; Table 1). Born through two consanguineous unions with no perinatal or postnatal problems, all affected individuals developed tonic-clonic seizures at 4-5 months of age and had delayed speech and motor development. Cerebellar symptoms (gait ataxia, intentional tremors and dysdiadochokinesia) were manifested from early childhood. The gait ataxia was progressive in nature, resulting in severe disability and later being confined to wheel chairs [IV. 2-5]. Dysmorphic facies, dysarthria, brisk deep tendon reflexes (DTRs), bilateral ankle clonus and an extensor Babinski response were evident in all of them. All the patients showed certain characteristic dysmorphic facial features like prominent supraorbital ridges, thick eyebrows, deep set eyes, epicanthal fold, low set ears, prominent antihelix, prominent nasal tip and thick lips (Fig. 1B). Behavioural abnormalities like stereotypies, hyperactivity, anger outbursts and psychotic symptoms were also observed (Table 1). They also had hearing impairment, and audiometry measures revealed bilateral mild to severe sensory neural hearing loss. Motor nerve conduction velocities from patients (V.1-2) were normal. The EEG from patients (V.1-2) showed generalised seizure discharges before treatment (Figure 1C), which became normal after treatment with anti-epileptic drugs. The other four members (IV.2-5) remained seizure free for several years on medication. MRI from IV.2 showed enlarged basal ganglia and cerebellar atrophy (Fig. 1D). The remaining members of the family were clinically unaffected.
Various biochemical parameters were evaluated for the six affected individuals (Supplementary Table 1). The raw values of each parameter were unit variance normalized (−1.85 to +1.95) across six samples and hierarchically clustered (Fig. 1F). Serum creatinine levels were highest in III. 11 and lowest in III. 12. Urinary sodium, serum potassium, serum calcium, serum homocysteine and LDL/HDL ratio were higher in IV.2 compared to remaining individuals. The serum chloride, urinary chloride, plasma ammonia, plasma lactate, vitamin B12, total cholesterol, urinary potassium, blood urea and urinary calcium were relatively higher in III. 12 compared to remaining individuals. Aryl sulfatase and urinary creatinine levels were higher in IV. 2 and IV.3. Calcium/creatinine and alkaline phosphatase levels were higher in IV.4 and IV.5 compared to remaining individuals.
Novel KCNJ10T290A variant residing in a ROH of Chr1 segregate in all patients
To identify the crucial disease-causing loci, we performed WES and analyzed exome-wide levels of homozygosity in all family members. A novel variant KCNJ10T290A seems to be cosegregating with a long homozygous stretch in Chr1 in all affected individuals of this kindred (Fig. 2A). Deleterious genetic effects of inbreeding are evident in children’s born out of consanguineous unions with a relatively higher burden of homozygous alleles (Bittles and Black, 2010; Shawky et al., 2013; Corry, 2014; Fareed and Afzali, 2014). These effects have been implicated to influence the evolution of mental illness and neurodevelopmental disorders (Bittles and Black, 2010). Since SeSAME syndrome follows autosomal recessive (AR) inheritance and the role of homozygous alleles in AR illness has been well established (Sund et al., 2013; Kelmemi et al., 2015; Wang et al., 2016; Martin et al., 2018), we used two independent approaches to identify the causative variants segregating with the illness in the SeSAME family; assessing the exome-wide runs of homozygosity (ROH), referred as ROH method and assessing the variants based on their allele frequencies, referred as non-ROH method.
Variant prioritization using ROH method
We analyzed the exome-wide levels of homozygosity for all samples within the pedigree including seven population controls (see materials and methods). Principal Component Analysis (PCA) of the exome-wide F-statistics separates the family members (n=6) from the population controls (n=7), explaining for an overall variance of 49.6%. All samples (both familial and population) within the two clusters, fell within their 95% confidence ellipses, except for 2 controls representing the relatively admixed communities (Fig. 2B). The SeSAME family alone was subjected to PCA in which the cases (n=4) formed a cluster and the unaffected parents (n=2) fell outside the 95% confidence ellipse (Supplementary Fig. 2), explaining the intra-familial variance in homozygosity. The ROH within the exomes of the individuals in the pedigree and the population controls were identified. A total of 56 homozygous stretches (either overlapping or unique) were identified in all cases and controls, of which 44 stretches belonged to the four affected siblings and the remaining were distributed between unaffected parents and population controls (Supplementary Table 2). Nevertheless, no ROH was detected in a subset of population controls. The burden of ROHs witnessed in the cases as compared to controls could be attributed to their consanguineous parentage. Of the ROHs identified in total, five stretches were explicitly shared between all the affected siblings but not observed in the unaffected parents and population controls, which will henceforth be notified as ROHaffected (Table 2). The ROHaffected consists of a union set of 5329 variants across all the cases and controls, of which any given variant was observed in at least one sample. Since the disorder follows an autosomal recessive (AR) inheritance pattern, of the 5329 variants, we identified those that were heterozygous (HET) in both unaffected parents, but homozygous (HOM) in all of the affected siblings. Seventy-eight such variants, belonging to 47 genes, were identified and all of them mapped to Chr 1 (Supplementary Table 3). This skewed observation could not be attributed to the length of Chr 1 for three reasons: i) the method used to compute ROH uses a sliding window approach which essentially removes the bias induced by the length of the chromosome; ii) the same Chr 1 ROH was not observed in either of the controls; iii) no ROH was observed in Chr 2 despite its genomic length being comparable to that of Chr 1. Of the 78 variants only three missense variants i.e., i) Chr1:158368964-C-T (OR10T2) ii) Chr1:160011455-T-C (KCNJ10) and iii) Chr1:161495040-C-T (HSPA6), were predicted to be deleterious by both SIFT (Vaser et al., 2016) and PolyPhen (Adzhubei et al., 2015).
Variants prioritization using non-ROH method
To identify other putative deleterious variants segregating within the family, which could have otherwise been ignored by the ROH based method, we identified all the exonic and splice variants (including non-synonymous, stop gain and stop loss) segregating within the family. The common variants i.e., those with a minor allele frequency (MAF)>0.01 in 1KG_all (1000 Genomes Project) and ExAC_all (Exome Aggregation Consortium) databases, were excluded from the analysis (Clarke et al., 2012; Lek et al., 2016) We identified seven variants belonging to seven genes (Supplementary Table 4). Interestingly, all the seven variants were located within Chr1:151288779-161088292, which was a subset of ROHaffected. Amongst the seven variants, Chr1:160011455-T-C [KCNJ10] was an obvious overlap. The remaining six variants fell on PI4KB, RORC, FLG2, FCRL1, PVRL4 and NIT1 genes. Apart from the KCNJ10 variant, none were predicted to be deleterious by both SIFT and PolyPhen. However, three of the remaining six variants (Chr1:151288779-T-C [PI4KB], Chr1:161049499-G-A [PVRL4] and Chr1:161088292-A-G [NIT1]) were predicted to be deleterious by either of the two algorithms, and were seemingly rare with MAFs of 0, 3E-3 and 5E-3 respectively, in South Asian populations according to the gnomAD database (Lek et al., 2016). Finally, the zygosity of the KCNJ10T290A variant was confirmed for all individuals in the family (III.11-12, IV.2-5, IV.9-10 and V.1-2) by Sanger sequencing (Fig. 2C).
Thus, of the union set of nine putative deleterious variants (three based on ROH method and seven based on allele frequencies) segregating within the family, the KCNJ10 gene was shortlisted for functional analysis to unravel the molecular impact of the variant for following reasons: i) KCNJ10, the candidate gene known to cause SeSAME syndrome (Celmina et al., 2018); ii) the variant reported in the patients is novel; iii) this was the only deleterious variant identified by both methods and iv) the variant resides on the cytoplasmic C-terminal tail (Fig. 2D) which is strongly conserved through evolution (Fig. 2E).
Kir4.1T290A disrupts channel properties in patient-derived LCLs
LCLs have been routinely used as a surrogate in vitro cell model to investigate cellular mechanisms of neurodevelopmental psychiatric disorders (Giuliano et al., 2014; Kumar et al., 2016). To investigate the functional role of Kir4.1T290A, we generated patient-specific LCLs, validated by karyotype for six members of SeSAME family. All six LCLs are free from both numeric and structural chromosomal abnormalities (Supplementary Fig. 1).
The barium-sensitive inwardly-rectifying K+ current in LCLs measured by whole-cell patch clamp was substantially compromised in all affected individuals. Kir4.1T290A significantly depolarized membranes of LCL and showed deficits in clearance of extracellular K+. To determine whether LCLs express functionally active endogenous Kir4.1 protein, we used immunofluorescence (IF), western blot and electrophysiology (Fig. 3). In parental controls, Kir4.1 is in close proximity with the actin-rich plasma membrane, diffusely discernible in the cytoplasm and enriched in the nuclear membrane and nucleus (Fig. 3A). However, in all affected individuals, we observed an increased punctate distribution of Kir4.1 in the cytoplasm but with no apparent disparity in the nucleus and nuclear membrane (Fig. 3B). To confirm the IF findings, western blot analysis showed a substantial increase in the expression of Kir4.1 in all affected individuals compared with unaffected parents (Fig. 3C and 3D). These findings suggest an unstable nature of Kir4.1T290A in all affected individuals.
To confirm whether the endogenous Kir4.1 expressed in LCLs is functionally active and elicit detectable inward-rectifying potassium currents in vitro, we performed whole-cell patch clamp recordings in response to voltage-steps from −120 to 40mV in 10mV, from a holding potential of −30mV both in the presence and absence of 110μM barium, a selective Kir channel blocker. Baseline current discharges from two heterozygous parental controls (III.11: −0.89+0.086, n=18, p=1.114 and III. 12: −0.86+0.049, n=16, p=1.347) were not significantly different from wild type controls (−0.85+0.046, n=17) (Fig. 3E). In contrast, the average barium-sensitive current densities were substantially decreased in all three control LCLs tested, in heterozygous parents (III. 11: −0.64+0.041, n=15, p=2.1E-4 and III.12: −0.60+0.086, n=14, p=1.8E-4) and wild type (−0.63+0.104, n=14, p=2.5E-4) compared with their respective baseline discharges, implying the specificity of K+ currents recorded from endogenous Kir channels (Fig. 3E).
We recorded the resting membrane potential of LCLs from patients (Fig. 3F). Average membranes voltages from all patients (IV.2: −30mV+3.640, n=18, p=1.3E-5; IV.3: −32mV+2.156, n=20, p=2.4E-5; IV.4: −31mV±3.083, n=17, p=1.7E-4; IV.5: −24mV±2.817, n=20, p=2.8E-5) were significantly hyperpolarized as compared to wild type (WT:-55mV+4.102, n= 24) and parental controls (III.11: −51mV+3.842, n=21 and III.12: −50mV+4.21, n=19). In wholecell voltage clamp, membrane current amplitudes were measured in all family members at both positive and negative potentials than the K+ equilibrium potential (Ek) (Fig. 3G and 3H). The mean current densities as a function of voltage (pA/pF) measured in all those expressing the mutant channel were markedly smaller than wild type and parental controls (Fig. 3H). One major facet of the Kir4.1 channel is to clear extracellular K+ thereby showing stronger rectification. To test the K+ clearance ability of LCLs, we clamped the cells at their resting membrane potential, with and without 110μM barium, and measured the elicited membrane current discharges upon induced K+ steps (from 5-20 mM). Overall, barium-sensitive currents from all patients were significantly reduced when compared to both parental and wild type controls (Fig. 3I).
Neural expression of Kir4.1T290A did not rescue Drosophila irk2 mutant phenotypes
Of the three Kir channel proteins expressed in Drosophila, BLASTp search identified Irk2 (CG4370), as the single closest homolog of human Kir4.1. Amino acid sequence alignment of Irk2 with Kir4.1 showed identical domain organization with 20% identity and 36% similarity (Fig. 4A). Enhancer-promoter (EP) element G8696 residing in the intronic region of Irk2 genomic locus was imprecisely excised to generate a 2348 bp and 2092 bp deletion from the EP insertion site to the pore-forming intramembrane domain, yielding two irk2 mutant alleles; irk2e118 and irk2e91b (Fig. 4B). Homozygous irk2e118 mutants are late embryonic lethal and transcriptional null allele, whereas homozygous irk2e91b mutants hatched viable and fertile adults with 51+12% reduction in Irk2 mRNA levels relative to wildtype and hemizygous genetic controls (Fig. 4C). However, the transheterozygotes (irk2e118/ irk2e91b); henceforth referred as irk2 mutant, exhibited reduction in Irk2 mRNA levels (72+15%) and shortened life span (Fig. 4D). Furthermore, to confirm the specificity of irk2 whole mutant phenotypes, we performed rescue experiments using the Gal4-UAS system with Irk2WT, Kir4.1WT, irk2T290A and Kir4.1T290A channel proteins. Motor neuron-specific overexpression of either Irk2WT or Kir4.1WT, substantially rescued embryonic lethality of irk2e118 homozygous mutants and shortened life span of heteroallelic irk2 mutant. The neural specific overexpression of mutant proteins irk2T290A and Kir4.1T290A, in contrast, failed to ameliorate irk2 mutant phenotypes (Fig. 4E).
In addition, we assayed larval locomotion to study the inherent functional capacity of Irk2 channels in motor neurons. Live imaging of larval locomotion over time followed by automated tracing revealed substantial restricted movement of irk2 compared to WT controls (Fig. 5A). irk2 (n=100) mutants displayed altered peristaltic behaviour (26+6.8) and speed (5.9+0.2) compared to WT (n=100; 75+5.9, p<0.00024 and 2.4+0.32, p<0.00017, respectively). Both peristaltic and speed defects are substantially improved by restoring Irk2WT channel expression (57+7, n=100, p<0.00035 and 5.4+0.16, n=100, p<0.00054) and Kir4.1WT (59+5.3, n=100, p<0.00017 and 5.8+0.23, n=100, p<0.00068) in neurons (Fig. 5B). Nonetheless, neural expression of mutant proteins viz., irk2T290A and Kir4.1T290A failed to suppress the locomotor defects of irk2 implying the functional significance of T290A mutation segregating in SeSAME family (Fig. 5).
irk2 mutant flies displayed impaired locomotion
The motor neurons in adult flies are necessary for complex movements (Bidaye et al., 2014; Enriquez et al., 2015; Gowda et al., 2018) and we tested the persistence of early larval phenotypes in young adults. We used high-resolution live-tracking adult locomotor assays in two independent platforms; a vertical fly-climbing assay to measure fly’s innate response to gravity and its role in patterns of locomotion, thereby indirectly assaying the drive to move (Fig. 6A; Aggarwal et al., 2018) whereas horizontal fly-bowl assay which measures locomotion in a cue free environment (Fig. 6D; Simon and Dickinson, 2010). In the vertical fly-climbing assay, 5-day old flies were provided with regular stimulus for locomotion by rotating the arena for fly climbing every 15 seconds (Aggarwal et al., 2018). The innate geotactic response of the flies makes them always climb towards the top of the arena, and when, after 15 seconds, the arena is rotated 180 degrees, the flies which are on the top are now pushed downwards providing them with the geotactic stimulus. This regular geotactic stimulus, delivered every 15 seconds is processed very reliably in the controls, with an evident climbing event for each rotation. irk2 mutants, are positively geotactic in the first minute (p=0.02; data not shown). The total number of tracks as well as the distance remain unchanged between wild type and irk2 mutants (p=0.3788; p=0.4656). However, the average time spent by flies per track was higher in irk2 mutants (7.98+0.44 seconds) compared with wild type (5.83+0.65 seconds, p=0.0183, unpaired student’s t-test, n=10; Fig. 6B). Furthermore, irk2 mutants were slow climbers (1.56 + 0.15 BLU/s) compared to controls (2.36 +0.22 BLU/s, p=0.0095, unpaired student’s t-test, n=10) (Fig. 6C; Supplementary Video 1). In the flyBowl assay, irk2 mutants walked significantly slower (2.49+0.28 mm/s) as compared to wild type (4.24+0.35mm/s, p=0.002, unpaired Kolmogorov-Smirnov test, Fig. 6E). In this cue free environment, the irk2 mutants showed a sharp decline in locomotor activity post 6 minutes into the assay compared to wild type flies (p<0.01 for all the time points, two-way repeated measures ANOVA, post-hoc Sidak’s multiple comparisons, n>35, Supplementary Video 2).
Progressive loss of dopaminergic neurons in the irk2 mutant brain
We know that specific loss of dopaminergic neurons in the aging brain lead to motor deficits in many neurodegenerative conditions especially in Parkinsonism and Ataxia syndrome. In addition, these patients manifested tremors, ataxia, progressive spasticity and basal ganglia and cerebella dysfunction. We therefore explored whether a similar pathology occurs in irk2 mutants. To visualize DA neurons in 30 days old adults, we used TH (tyrosine hydroxylase)-specific Gal4 line (Friggi-Grelin et al., 2003) fused to membrane-tethered eGFP. Cells positive for anti-GFP were consider for analysis. We focused on six major clusters in central brain, which are annotated hereafter as paired posterior lateral 1 and 2 (PPL1 and PPL2); paired posterior media 1 and 2 (PPM1/2), paired posterior medial 3 (PPM3) and paired anterior lateral (PAL) (Fig. 7A). Paired anterior media (PAM) cluster was excluded from our analysis, since TH-Gal4 does not reliably label these cells. The neurons in all four DA clusters per hemisphere were significantly fewer in irk2 mutant flies, as compared to age-matched controls (Fig. 7B). This suggests that inhibiting Irk2 channel function promotes degeneration of DA clusters. To test whether the observed decrease in DA neurons is due to age-associated neurodegeneration rather an improper neural development, we analysed DA clusters in these mutants at an early age. On day 3, irk2 mutants showed no loss DA neurons compared to age-matched control flies (Fig. 7C). These results, strongly suggest that loss of irk2 function is sufficient to accelerate age-dependent progressive loss of dopaminergic neuronal clusters which could plausibly be attributed to locomotor defects and shortened-life span.
Discussion
We are able to identify a novel variant in the KCNJ10 gene as a cause for the SeSAME syndrome. The biological processes involved were explored using genomics, channel physiology from patient-derived LCLs and behavioural abnormalities in Drosophila that are influenced by age-dependent degeneration of dopaminergic neurons. These processes are perhaps conserved across species, and could provide a useful platform for translational neuroscience (Ryczko and Dubuc, 2017; Splinter et al, 2018).
Why renal tubulopathy is spared in a subset of SeSAME cases?
All six affected individuals in SeSAME family reported here displayed relatively uniform neurological and psychiatric manifestations in the form of ataxia, focal epilepsy, delayed developmental milestones, loss of intellectual and sensory-neural hearing but with no apparent electrolyte imbalance. These clinical features specific to SeSAME-like pedigrees (Dhaibani et al., 2018) is also evident in children with autism-spectrum disorders with epilepsy (Sicca et al., 2011; Sicca et al., 2016). In fact, several modern-day mammals like Jack Russell Terriers (Gilliam et al., 2014), Belgian Shepherd dogs (Mauri et al., 2017) and Malinois dogs (Van Poucke et al., 2017) experienced this SeSAME-like phenotype with KCNJ10 mutations. Therefore, it remains obscure how and why certain KCNJ10 variants fail to manifest electrolyte imbalances in SeSAME syndrome. There could be two possibilities for this discrepancy. First, it’s possible that certain KCNJ10 mutations can affect CNS functions independently of other organ systems. It is conceivable that astrocytes and microglial cells of nervous system are highly sensitive to dysregulation of potassium homeostasis, while basolateral membrane in the distal nephron may be impervious to this effect (Djukic et al., 2007). Another possibility is that same KCNJ10 variants could behave differently between CNS and kidney, since the channel activity depends largely on the formation of heterotetramers with other Kir entities (Kir5.1), cell type specificity, gating mechanisms and influence of cell surface signaling receptors (Reichold et al., 2010; deHart et al., 2008). Finally, it is unclear whether renal electrolyte deficit is a progressive impairment that develops over time, or a direct effect of the mutation, which necessitates further investigations and follow-up clinical evaluations.
Is ROH a cause or an effect of recessive inheritance in SeSAME syndrome?
The KCNJ10 is the single most identified candidate gene for the SeSAME syndrome, with 21 mutations from 27 patients reported till date, of which 11 were from consanguineous unions (Celmina et al., 2018). Autosomal recessive neuropsychiatric illnesses have also been witnessed in children born out of consanguineous marriages and some of them carry long stretches of homozygous segments within their genomes (Melhem et al., 2016; Gandin et al., 2015; Bittles and Black, 2010). Since the affected individuals detected to have SeSAME syndrome were born out of consanguineous parentage, we primarily used the ROH based approach to identify those variants present within the homozygous stretches shared between all those affected. This approach identified three deleterious variants belonging to KCNJ10, OR10Γ2 and HSPA6 genes. The allele frequencies of OR10Γ2 (18%) and the HSPA6 (9%) variants in the ExAC South Asian (SAS) populations are fairly high and hence unlikely to contribute to illness. However, it is interesting to note the previous association of the reported HSPA6 variant in patients with sensory disturbances (Kobayashi et al., 2013). This suggests that although gene of major effect are probably the primary drivers of the illness, the diversity in the clinical presentation is possibly an outcome of the complex genetic interactions between common and rare variants of varying effect sizes. As an alternative, we also used a non-ROH approach which is purely based on the allele frequencies to identify homozygous rare coding variants that could have been missed out by the previous approach. The fact that both methods autonomously converged to a deleterious novel KCNJ1CF29 variant, co-segregating with the illness is remarkable. Given the clinical phenotypic diversity and for additional reasons as discussed above, we suggest two possibilities broadly classified into intrinsic and extrinsic factors. Intrinsic factors include recombination hot-spots, defects in DNA repair, chromatin remodelling and yet unidentified intra-cellular signaling events, that favour to the occurrence of ROH, co-segregating with the illness. Another factor could be the clan structure of the family, which indicates a high degree of endogamy. Therefore, an interplay between these two factors could influence the clinical diversity of SeSAME syndrome.
Another possibility is that individual ROHs might play key role in spatial-temporal regulation of gene expression within cell types that are sensitive to K+ homeostasis (Christofidou et al., 2015). The difference in the expression of Kir4.1 among unaffected siblings in our SeSAME pedigree also highlights the role of ROH in gene regulation. Therefore, it would be useful to investigate the functional consequences of homozygosity in expression of genes within the ROH and/or in close proximity especially in cell types that are relevant to the pathophysiology of SeSAME syndrome.
Usually for every pregnancy in autosomal recessive disorders, there is a probability of 0.25 that the offspring(s) will inherit two copies of the disease gene and will therefore exhibit the phenotype (Ellard et al., 2015). However, in a clinical setting this distribution is skewed more towards almost all affected individuals in the same generation, than one would rather expect by chance, especially in children born to consanguineous unions. Thus, this skewed observation needs to be addressed at holistic paradigms by developing bio-physical and mathematical models to understand the physics and governing dynamics (Nash J.F, 1950) of the intra-cellular events, influencing the silent recombination choices of homologous chromosomes.
Importance of KCNJ10 variant characterization in SeSAME syndrome
Functional characterization of exogenously expressed mutant Kir4.1 has been attained in a variety of isolated cell types like astrocytoma cells (Sicca et al., 2016), C6 glioma cells (Méndez-González et al., 2016), CHO and HEK 293 cells (Reichold et al., 2010; Williams et al., 2010), Drosophila S2 cells (Döring et al., 2002), kidney-derived ts201 cells (Méndez-González et al., 2016) and Cosm6 (Kucheryavykh et al., 2007). However, the sensitivity and specificity of physiological read outs from Kir ion channels depends inherently on factors like cellular milieu, auxiliary subunits and formation of subunits for heterooligomeric assembly in the cell type of choice (Döring et al., 2002; Neusch et al., 2006; Fakler et al., 1996a and 1996b). Another major challenge is to extrapolate the findings from cloned Kir channels in isolated cell types to their native effects in disease-relevant tissues, which most likely yield weak currents and rectification (Olsen et al., 2006). Therefore, in vitro cellular models that preserves the native environment to measure authentic Kir4.1 channel activity will serve as an excellent cell of choice to investigate functional effects. We therefore addressed these issues, using patient-derived LCLs which harbour the risk variant.
Since the KCNJ10T290A variant observed in this SeSAME family, is a potential threonine phosphorylation residue, we performed in silico analysis to explore this possibility. Using Group-based Prediction system (GPS; Xue et al., 2008), we speculate that T290 could potentially be phosphorylated (GPS score 9.4 > cutoff 6.78 by Interleukin 1 Receptor Associated Kinase 4, IRAK4) at native conditions. Nonetheless, T290A phosphorylation has not been reported in PhosphoSitePlus. In light of the in-silico predictions and in vitro functional implications of KCNJ10T290A variant using patient-derived LCLs followed by whole-cell patch clamp (see results section), we predict the likelihood of T290 phosphorylation either at a very low stoichiometry or under specific signaling stimuli. However future experiments will assist in discriminating between these possibilities.
Genetic manipulation techniques and phenotypic similarities among different categories of K+ channel mutants in Drosophila (Littleton and Ganetsky, 2000; Bellen et al., 2010), offer an excellent opportunity to investigate the functional role of Kir ion channels at organismal level. Apart from in vitro approaches, animal models showed greater potential in deciphering the cell-type specific role of Kir4.1 in brain function. Justifying the importance of Kir4.1 in K+ homeostasis, Kir4.1 knockout mouse, Xenopus oocytes and zebrafish mimics a subset of SeSAME symptoms observed in humans (Djukic et al., 2007; Neusch et al., 2001). In fact, of the three Kir channels identified in Drosophila, reduced expression of irk2 in the CNS lead to defects in spike discharges and adult lethality suggesting an alternate model to investigate cellular functions of Kir4.1 (Chen and Swale, 2018). In addition, irk2 channels are constitutively active in Drosophila S2 cells, derived from the macrophage-like lineage, and functionally associate with sulphonyl urea receptors (SUR). In this study, we explored the role of Irk2 in CNS at cellular level. Irk2 is essential for survival of adult dopaminergic neurons and for coordinated rhythmic locomotion in larval and adult motor neurons. This relies on complex interplay between central and peripheral nervous systems where the central pattern generator (CPG) neurons at neuromeres regulate peristaltic movement of peripheral body wall muscles (Hasegawa et al., 2016; Kohsaka et al., 2017). Therefore, it appears that Irk2 ion conductance pathway in Drosophila is responsible for regulating K+ gradient and glutamate homeostasis at tripartite synapses of neuromuscular junctions (NMJ) ultimately controlling synaptic activity and neurotransmitter release. In addition, it would be valuable to dissect the role of Irk2 in synaptic vesicle turnover at NMJ given the role of K+ inward rectification in regulating endocytosis (Feliciangeli et al., 2010). Finally, our findings provide a proof-of-concept approach to explore the cell type specific function of Kir4.1 in Drosophila nervous system to better understand mechanisms of K+ spatial buffering and synaptic architecture which is an essential feature for proper synaptic transmission and plasticity.
Finally, recapitulating the biology of complex neuropsychiatric disorders in cellular and animal models, to identify downstream mechanisms is of heuristic value. Using patient-specific cells we were able to show the dysfunction of the primary disease causative gene in SeSAME syndrome, which when defective can cause quantifiable degenerative symptoms in Drosophila model system. As we constantly expand the identification of candidate genes for neuropsychiatric syndromes, these trans-disciplinary approaches will benefit our understanding of genotype-phenotype correlations at mechanistic level.
Funding
These experiments were supported by Ramalingaswami re-entry fellowship (RLF/DBT/2015) and ADBS (BT/PR17316/MED/31/326/2015) from Department of Biotechnology; Early career grant (ECR/2015/000468) from Department of Science and Technology and National Centre for Biological Sciences, Tata Institute of Fundamental Research.
Authors contributions
Conceptualization: RK
Special methodology: RKN, AC, RPM, VVP, SJ and JS
Investigation: RKN, AC, AA, AGS, SS, SP, MD, NR, MP, PSB, BV, MN, BM and RK
Manuscript writing: RKN, AC and RK
Data visualization and analysis: RKN, AC, AA and RK
Supervision: RK
Competing interests
The authors declare no competing or financial interests.
Acknowledgement
The authors are immensely grateful to all members of SeSAME family for their participation and constant involvement in this study. This work was primarily carried out by NIMHANS in collaboration with Prof. K. VijayRaghavan’s lab at National Centre for Biological Sciences, Tata Institute of Fundamental Research (NCBS – TIFR). We thank Jan’s lab (UCLA), Prof. LS Shashidhara’s lab (IISER, Pune), Edward Giniger’s lab (NINDS), Bloomington Stock centre for generous sharing of fly stocks and reagents and Gautham (NIMHANS) for facial description of SeSAME kindreds. We are grateful to Prof. K.VijayaRaghavan for helpful discussion during the course of investigation. We thank ADBS genomic team for sharing the WES analysis pipeline. This work was generously supported with funds from Ramalingaswami re-entry fellowship (RLF/DBT/2015), ADBS (BT/PR17316/MED/31/326/2015) from Department of Biotechnology and Department of Science and Technology (ECR/2015/000468) and National Centre for Biological Sciences, Tata Institute of Fundamental Research. Finally, the authors are thankful to all members of the lab and the ADBS consortium for suggestions and discussions during the course of investigation.
Footnotes
* Address for correspondence Ramakrishnan Kannan, PhD, National Institute of Mental Health and Neurosciences (NIMHANS), Bengaluru, Karnataka, India PIN:560029 Phone: 091-80-26995791 Fax:+91-80-26564830 Email: kannan{at}nimhans.edu.in