The early studies of disorders relied upon the methods of “classical” Mendelian laws and linkage, wherein an attempt was made to identify the segment of the genome that was shared by ill persons within an affected family. This strategy worked best when the disease was highly penetrant, was inherited in a dominant or recessive manner, and there were large multi-generation families with multiple meioses that could be analysed (Friedman et al., 2021). A crucial part of this analysis was the affectation status and accurate phenotyping. From the early analysis of large pedigrees from Iceland, to the discovery of a translocation in Chr1:11 that included the DISC1 gene as segregating with risk of psychoses, a number of studies with large family structures explored the genetic risk of psychiatric illness, with complex, and often confounding results (Sherrington et al., 1988; Millar et al., 2000; Craddock et al., 2005). We used linkage analysis, supplemented with a family-based association test (Transmission disequilibrium test), to detect evidence of shared liability in more than 50 families having multiple affected members. We observed a positive linkage and association finding at 18p11.2 for psychosis (Mukherjee et al., 2006). Gradually as more reports came in worldwide, linkage hits appeared on many chromosomal regions for psychiatric illness, though these were often not reproducible across studies, and it became apparent that there could be private mutations in particular clinical populations (Xu et al., 2009). The use of endophenotypes to be able to describe a complex disorder was thought necessary (Braff et al., 2007). However, even in pedigrees with many affected individuals, there was a lot of variation in illness presentation. The varying penetrance and heterogeneity in what was otherwise a relatively common disease led to the evolution of the common disease, common variant hypothesis (Mukherjee et al., 2002).

The DISC1 locus was first discovered as part of a chromosomal t (1; 11) (q42.1; q14.3) translocation that seemed to add to risk of both schizophrenia and bipolar disorder, in a large Scottish family (St Clair et al., 1990). This was later evaluated for association with schizophrenia by genotyping Single Nucleotide Polymorphisms (SNPs), and also sequencing by a large number of groups (Ma et al., 2018), but with conflicting results. A targeted genotyping of the SNPs in the gene resulted in only a partial gender-specific association in our sample set of 1,252 individuals (419 bipolar disorder patients, 408 schizophrenia patients and 425 controls) (Murthy et al., 2012). As large-scale genotyping methods for common variants became possible, this locus did not retain its significance across other populations, perhaps hinting again at the heterogeneity of mental illness, and its patterns of inheritance (Mathieson et al., 2012).

One interesting outcome of the Genome-wide association (GWA) method was the feasibility of comparing genotype results for multiple loci across populations and studies. This showed that the variation in population allele frequencies might affect the detection of association, across populations (Asif et al., 2021).

The experience with the APOE Ɛ4 haplotype and risk of dementia is thus exemplary. The APOE Ɛ4 haplotype, which is a combination of two missense variants, was first described by the Roses lab at the Duke University Medical Centre as a causal locus for dementia (Corder et al., 1993), and has been associated with dementia in populations across the world, though with varying effect sizes. Our studies confirmed that the frequency of the risk allele was much higher in those with dementia (20%) when compared to age-matched controls (12%) i.e., the Apoe4 carrier allele frequency was 14.8% in controls and 33% in dementia patients (Bharath et al., 2010). The global reported average proportion of APOE4 carriers is 23.9%, and varies with geography (from 19.3% to 30.0%), and ethnicity (from 19.1% to 37.5%) (Wang et al., 2021).

The intersection between variations in frequency of the APOE Ɛ4 alleles, and the prevalence of dementia thus becomes important. A recent examination of APOE Ɛ4 carrier frequency and dementia prevalence by Llibre-Guerra et al. (2023) showed a similar APOE Ɛ4 carrier frequency in Caribbean and American Hispanics of about 21%–23%, with a dementia prevalence of 9%–10% in these populations. Interestingly while the carrier frequencies for the non-Hispanic whites and African Americans are 25% and 34%, the prevalence of dementia is seen to be 3.2% and 13.3% for these two populations. The estimated dementia prevalence for adults older than 60 in India is 7.4%, with significant age and education gradients, sex and urban/rural differences, and cross-state variation (Lee et al., 2023). These differences could be due to both additional risk factors or other environmental and dietary factors.

We have attempted to identify other modifying variants related to the risk of dementia. The CLU (Clusterin/ApoJ)—rs11136000) and PICALM (phosphatidylinositol binding clathrin assembly protein)—rs3851179) loci were reported in a two-stage genome-wide association study of AD involving over 16,000 individuals from Europe and the United States (Harold et al., 2009). These results were not replicated in our modest sample of AD cases and age-matched controls; we observed an MAF of 0.29 for CLU (rs11136000) and 0.43 for PICALM (rs3851179) in controls (Shankarappa et al., 2017). We were soon joined by the other groups that studied the East Asian, Turkish, Polish and African American populations who also did not find an association at these loci (Logue et al., 2011; Klimkowicz-Mrowiec et al., 2013; Sen et al., 2015; Han et al., 2018). There appeared to be a particular allele frequency that was showing association, especially in the case of the CLU locus. It was concluded that perhaps the same variant might not show an association but a larger haplotype analysis might be useful to determine whether the gene locus was important in Alzheimer’s disease.

Analysis of the region around the APOE Ɛ4 locus identified a polymorphism in the TOMM40 gene which codes for a protein in the outer mitochondrial membrane (Roses et al., 2010). The variation in the length of the polyT tract in the sixth intron of this gene (resulting in alleles S, L, VL) showed a significant association with dementia (Gottschalk et al., 2014). While the TOMM40 gene biology suggests that it could contribute independently to the risk of dementia, its proximity to the APOE gene could lead to a spurious association in genetic analyses. The results have varied across populations depending upon the linkage disequilibrium between the TOMM40 locus (rs10524523) and the APOE locus. In the European population, 0.8% of the APOE Ɛ4 (−) and 94.2% of APOE Ɛ4 (+) had the ‘L’ allele. In the African-American population, 1.1% of the APOE Ɛ4 (−) had L, but only 47.8% of the APOE Ɛ4 (+) had the L allele (Yu et al., 2007). We thus genotyped the TOMM40 locus in our dementia population along with the APOE Ɛ4 status.

Populations may differ in their genetic predisposition to damage caused by environmental factors. Alcohol related disease may thus be viewed as a particular natural experiment, wherein the risk of disease is contingent upon exposure. The two major enzymes involved in alcohol metabolism show significant ethnic variation. Alcohol dehydrogenase (ADH), converts alcohol to acetaldehyde, which is then converted to acetate by aldehyde dehydrogenase (ALDH). Seven known ADH genes encode enzymes that catalyze the conversion of alcohol to aldehyde (Osier et al., 2002). However, the relative kinetics of the enzyme variants in an individual might dictate the ability to use large quantities of alcohol and not feel uncomfortable. This could add to the risk of disease, as a feeling of discomfort would discourage excessive drinking, and perhaps be protective. Differences in the frequency of variants in the ADH and ALDH genes contribute to the “flushing” response in East Asians, and they are thought to be more sensitive to the effects of alcohol, as compared to Europeans. Recent reports from East Asia have linked the ALDH2 allele to flushing and discomfort on alcohol consumption (Lee et al., 2014). Interestingly, Indians were thought to be more sensitive to alcohol, as compared to Europeans, in early Greek writings from 1,500 years ago (Lucian of Samosata, 2018). A study in our population, however, could not detect the ALDH2 isoform associated with increased aldehyde accumulation as observed in East Asians. The protective allele may thus be rare in the Indian population. These differences in the oxidative pathways could also impact other metabolic processes, and lead to adverse consequences on the liver (Li et al., 2015). Most patients who use alcohol develop liver steatosis over time, although further progression to cirrhosis is not universal. A reliable prediction of susceptibility to liver complications from alcohol abuse would thus be useful.

The wider availability of genomic screens now allows us to interrogate families with multiple affected members through other methods. Whole exome sequences (WES) and whole genome sequences can provide additional insight, especially on a case-by-case, or even a familial basis (Kato et al., 2023). The exome constitutes less than 3% of the genome, but since it directly impacts protein function, variants may have a much larger impact on biological processes, and thus susceptibility to disease. Such a penetrant coding variant may be sufficient in itself, or interact with a background of common variants (Kato et al., 2023).

About 10% of dementia patients have another first-degree relative with dementia (Reitz et al., 2023). We did a WES study, on a set of such individuals, and were able to identify several rare deleterious genetic variations, in the coding region of genes involved in amyloid signalling (PSEN1, PLAT and SORL1) and other dementia-associated pathways (Syama et al., 2018). We detected a different base change for a previously reported mutation in the PSEN1 gene (TGG to TGT/C) resulting in an identical W165C missense mutation, highlighting the possibility of different DNA variants across families or populations causing changes in the same risk regions of candidate genes.

We also used WES in a set of families with multiple affected members with schizophrenia, bipolar disorder, dementia and OCD and identified many plausible risk variants in familial severe mental illness (SMI) (Figure 1) (Ganesh et al., 2022). We detect potentially deleterious variants in genes and pathways that have been implicated in Mendelian disorders, and also previous GWA studies (Ganesh et al., 2022). Interestingly some of these individuals also shared some of the clinical characteristics of the same Mendelian disorders (unpublished observations). Thus, many risk alleles for genetic diseases seem to produce a spectrum of syndromes, depending on whether they occur as heterozygous, homozygous or compound heterozygotes.

This possible intersection with the distribution of variants, across populations and within a gene and protein at particular points of vulnerability, is exemplified by our observations on the PLA2G6 gene. The Phospholipase A2 Group 6 (PLA2G6) gene located on chromosome 22q13.11 consists of 17 exons and encodes 806-amino acid VIA calcium-independent phospholipase A2 protein (Chu et al., 2020). Homozygous missense variants (e.g., p.D331Y, p.R741Q), are often linked to atypical INAD (infantile neuroaxonal dystrophy) and early onset levodopa-responsive parkinsonism, with variable brain iron accumulation (Roeben et al., 2023). Many reports also found neuropsychiatric symptoms or behavioral changes in the initial presentations of patients with PLA2G6 mutations (Chu et al., 2020). Abnormalities in phospholipid biosynthesis, membrane remodelling and related processes have been implicated in many neuropsychiatric syndromes (Minghui et al., 2022). Increased phospholipase A2 (PLA2) activity has been reported in schizophrenia, and treatment with antipsychotic drugs reduced the enzyme activity to levels similar to those in control (Gattaz et al., 1987). Thus, it was crucial to study variants of PLA2G6 in SMI.

The studies on the molecular genetics of neuropsychiatric illness now have to encompass a wide spectrum, from issues of penetrance to the range of minor allele frequencies. This may require a reconsideration of many of the prevailing models of classifying and understanding the impact of genetic variations, as recently discussed by Yao et al. (2023). We describe a few encounters with these phenomena in our work (summarised in Abstract Figure), which may have made this quest seem worthwhile.

At NIMHANS, our sample set of 464 unaffected adults older than 60, had an APOE Ɛ4 carrier allele frequency of 18.3%. On genotyping the TOMM40 poly A locus in our unaffected elderly population (N = 165) we found, 3.6% of APOE Ɛ4 (−) and 37% of APOE Ɛ4 (+) had the “L” allele. As shown in Table 1, a comparison of individuals with dementia (N = 151) and age-matched controls (N = 165) showed a significantly higher occurrence of the “L” allele with APOE Ɛ4 carriers and dementia (p < 0.0001). Further, even among the APOE Ɛ4 non-carriers, the “L” allele frequency was significantly higher in cases than in controls (p < 0.0008) showing that the L allele on its own may also be a risk factor for dementia in our sample set. We also checked for LD across these two genes with the 3 markers (rs10524523, rs429358, rs7412) in our population. The linkage between the TOMM40 locus and APOE though present, appeared to be much weaker compared to that reported in European populations (Yu et al., 2007).

We hypothesised that gene expression of TOMM40 could be influenced by the genotype at this intron 6 locus and carried out gene expression studies. However, in a study of 45 LCLs, two-thirds of which were derived from AD patients, there was no influence of the TOMM40 allele on gene expression. The APOE Ɛ4 carrier status also did not have any effect on TOMM40 gene expression in LCLs (Supplementary Figure S1). It is, however, possible that there may be tissue-specific effects of this polymorphism which are not seen in LCLs.

We compared individuals who had developed cirrhosis (AUDC+), to others who had not (AUDC-ve); though both had been using alcohol. The demographic and clinical characteristics of participants (N = 236) showed that the two groups did not differ in age, but those in the AUDC + ve group had used slightly lesser units of alcohol (13 ± 6) compared to the AUDC-ve group (16 ± 7) (p = 0.01) (Supplementary Table S2). The AUDC + ve group had also been drinking for a comparatively shorter duration (16 ± 7 years vs. 18 ± 8 years) and had a later age of onset of drinking (29 ± 8 years vs. 23 ± 7 years) (p = <0.001). AUDC + ve group showed more evidence of liver disease, in the form of higher concentrations of serum total and direct bilirubin, Alkaline Phosphatase (ALP), Gamma- Glutamyl Transferase (GGT) and lower concentrations of total protein, albumin, and haemoglobin levels, when compared to AUDC-ve group.

In order to check for a genetic predisposition if any, for adverse consequences of alcohol exposure we computed the genetic risk score (GRS) for each individual at the candidate loci examined (Shankarappa et al., 2022). GRS calculation showed a significantly higher genetic risk score in the overall AUD group (Mean ± SD = 1.4 ± 0.8) compared to the control population group (SAS) (Mean ± SD = 0.82 ± 0.76) (p < 0.01). Further, subgroup analysis showed that the AUDC + ve group had a higher GRS (Mean ± SD = 1.3 ± 1.0) than the AUDC-ve group (Mean ± SD = 1.1 ± 1.0) though this was not statistically significant. Thus it appears that effect sizes taken from other studies could successfully predict adverse outcomes of risk allele combinations in our population. However, a more accurate calculation might have to include computed variables to account for cross-population genetic structure. Such a computed GRS could predict the risk of developing cirrhosis on alcohol exposure and have applications in personalised medicine.

We screened our whole exome sequencing data of familial severe mental illness families for PLA2G6 variants (Figure 2A). WES in 310 individuals from SMI families and population controls, revealed 851 instances (720 intronic, 57 exonic, 56 5′UTR, 18 3′UTR) of variants in the PLA2G6 gene. Ten exonic variants (4 synonymous and 6 non-synonymous) were seen in 57 individuals. Six non-synonymous variants were found in heterozygous state (Table 2). SIFT and MutationTaster predicted p.H117R, p.I256V and p.D377Y as deleterious. The variant p.Asp377Tyr was found in 3 siblings with schizophrenia, all of whom had pronounced parkinsonian symptoms on antipsychotic treatment. To assess the possible impact of these variants on drug binding, molecular docking analysis with the AlphaFold structure of the protein and antipsychotic medications, chlorpromazine, and risperidone was done. The preliminary molecular docking analysis suggests that while both chlorpromazine and risperidone bind at three predicted drug binding sites, risperidone has a greater binding affinity (Figure 2B). In addition, the variants p.Arg741Gln, p.Arg741Trp (near site 2) and p.Asp377Tyr (near site 3, which is the interface region between the catalytic domain and the ankyrin repeat) are in close proximity to these predicted binding sites. Which could have an impact in the antipsychotic binding affinity of the particular variant protein in addition to alteration in its biological function. It is perhaps not inconceivable that in schizophrenia patients with such heterozygous PLA2G6 variants, antipsychotic treatment can trigger parkinsonian symptoms.