Abstract

   The co-occurrence of insulin resistance (IR)-related metabolic
   conditions with neuropsychiatric disorders is a major public health
   challenge. Evidence of the genetic links between these phenotypes is
   emerging, but little is currently known about the genomic regions and
   biological functions that are involved. To address this, we performed
   Local Analysis of [co]Variant Association (LAVA) using large-scale
   (N = 9,725–933,970) genome-wide association studies (GWASs) results for
   three IR-related conditions (type 2 diabetes mellitus, obesity, and
   metabolic syndrome) and nine neuropsychiatric disorders. Subsequently,
   positional and expression quantitative trait locus (eQTL)-based gene
   mapping and downstream functional genomic analyses were performed on
   the significant loci. Patterns of negative and positive local genetic
   correlations (|r[g]| = 0.21–1, p[FDR] < 0.05) were identified at 109
   unique genomic regions across all phenotype pairs. Local correlations
   emerged even in the absence of global genetic correlations between
   IR-related conditions and Alzheimer’s disease, bipolar disorder, and
   Tourette’s syndrome. Genes mapped to the correlated regions showed
   enrichment in biological pathways integral to immune-inflammatory
   function, vesicle trafficking, insulin signalling, oxygen transport,
   and lipid metabolism. Colocalisation analyses further prioritised 10
   genetically correlated regions for likely harbouring shared causal
   variants, displaying high deleterious or regulatory potential. These
   variants were found within or in close proximity to genes, such as
   SLC39A8 and HLA-DRB1, that can be targeted by supplements and already
   known drugs, including omega-3/6 fatty acids, immunomodulatory,
   antihypertensive, and cholesterol-lowering drugs. Overall, our findings
   highlight the complex genetic architecture of IR-neuropsychiatric
   multimorbidity, advocating for an integrated disease model and offering
   novel insights for research and treatment strategies in this domain.

   Subject terms: Psychiatric disorders, Genomics

Introduction

   Multimorbidity, defined as the co-occurrence of multiple conditions
   within an individual, poses substantial challenges to healthcare
   systems [[79]1]. An example is the observed co-occurrence of insulin
   resistance (IR)-related metabolic conditions, such as type 2 diabetes
   mellitus (T2DM), obesity, and metabolic syndrome (MetS), with
   neuropsychiatric disorders [[80]2]. This multimorbidity contributes to
   more severe physical and mental health outcomes, leading to reduced
   treatment effectiveness and higher mortality rates [[81]3–[82]5].
   Moreover, IR is associated with detrimental effects on cognitive
   function, potentially worsening the cognitive impairment observed in
   various neuropsychiatric disorders [[83]6].

   IR manifests as reduced tissue responsiveness to insulin stimulation,
   primarily disrupting blood glucose homeostasis and inducing long-term
   micro- and macrovascular complications, as well as peripheral nervous
   system damage [[84]7]. Such a metabolic perturbation is a distinctive
   feature of T2DM, central obesity, and MetS [[85]7]. Emerging evidence
   suggests that IR shares aetiological pathways with neuropsychiatric
   disorders, including Alzheimer’s disease (AD), mood and psychotic
   disorders [[86]8–[87]10]. The connection between IR-related conditions
   and neuropsychiatric disorders is supported by compelling
   epidemiological data [[88]2, [89]11]. Indeed, bidirectional phenotypic
   associations have been found between these two nosological groups
   [[90]2]. This evidence blurs the boundaries between traditional disease
   categories, advocating for a more integrated approach to research and
   clinical management [[91]4, [92]12]. Consequently, a deeper
   comprehension of the mechanisms underlying this multimorbidity is
   essential.

   Beyond shared environmental risk factors – including poor diet,
   sedentary lifestyle, and disturbed sleep [[93]13–[94]15], which could
   also be direct manifestations of psychopathology – shared genetic
   components have been identified [[95]8]. Both IR-related conditions and
   neuropsychiatric disorders are highly heritable and polygenic [[96]16,
   [97]17], with heritability estimates, derived from twin and family
   studies, ranging from 40 to 80% [[98]18, [99]19]. Work by us and others
   disclosed global genetic correlations between neuropsychiatric
   disorders and IR-related conditions, indicative of shared genetic bases
   [[100]8, [101]10], though the effect directions were not consistent
   across all phenotype pairs. Intriguingly, two clusters of
   neuropsychiatric disorders were identified, wherein the genetics of
   IR-related conditions showed opposite directions of genetic
   correlation. The first included attention-deficit/hyperactivity
   disorder (ADHD) and major depressive disorder (MDD), which showed
   positive genetic correlations with IR-related conditions; the second
   included obsessive-compulsive disorder (OCD), anorexia nervosa (AN),
   and schizophrenia, which showed negative genetic correlations with
   IR-related conditions [[102]8]. Genetic covariance was also highlighted
   within gene sets pertinent to insulin processing, secretion, and
   signalling, suggesting that several neuropsychiatric disorders could be
   reconceptualised as “insulinopathies” of the brain [[103]8].
   Strikingly, certain neuropsychiatric disorders, such as AD and bipolar
   disorder (BD), demonstrated no global genetic correlations with
   IR-related conditions, despite previous literature suggested a shared
   pathophysiology [[104]8, [105]20]. However, global genetic correlation
   only encapsulates the average direction of genetic sharing across the
   genome, while the patterns of genetic correlations at the level of
   individual genomic regions can vary significantly [[106]21]. Local
   genetic correlation can deviate from the genome-wide average, and
   regions of strong, local genetic correlation have been reported for
   multiple traits even in the absence of genome-wide genetic correlation
   [[107]21, [108]22]. Therefore, the absence of genome-wide genetic
   correlations does not necessarily exclude shared genetics in specific
   regions, suggesting the importance to further study the possible
   genetic overlap between conditions without global genetic correlation,
   such as AD and IR-related traits [[109]8]. Importantly, dissecting the
   local patterns of genetic sharing could shed light on specific genetic
   factors involved in IR-neuropsychiatric multimorbidity and new
   potential therapeutic targets for both groups of conditions. Recent
   advances in bioinformatics have facilitated a more detailed exploration
   of the genetic overlap across distinct phenotypes. Traditional global
   genetic correlation methods, like Linkage Disequilibrium Score
   regression (LDSC), assess shared genetic architecture between
   phenotypes across the entire genome [[110]23] but may fail in
   identifying phenotype pairs that share specific genomic regions
   potentially without showing global genome-wide genetic correlation
   [[111]23]. Therefore, the utilisation of local genetic correlation
   analyses may offer more granular insights into shared genetic bases
   [[112]22].

   In this study, we aimed to dissect the genetic overlap between three
   IR-related metabolic conditions – namely, obesity, T2DM, and MetS - and
   nine psychiatric disorders by examining their pairwise patterns of
   local genetic correlation throughout semi-independent regions across
   the genome. Any shared genomic region was further explored using
   positional and expression quantitative trait locus (eQTL)-based gene
   mapping techniques. This was followed by a functional annotation of the
   mapped genes, enabling a deeper exploration of biological mechanisms
   underlying IR-neuropsychiatric multimorbidity. Lastly, we investigated
   the shared (likely) causal variants possibly driving the
   pathophysiology of this multimorbidity.

Materials and methods

Input datasets

   We leveraged publicly available summary statistics from the largest
   genome-wide association studies (GWASs) on the three most prevalent
   IR-related conditions, namely obesity, MetS, and T2DM
   (n = 244,890–933,970), and nine neuropsychiatric disorders, including
   AD, ADHD, AN, autism spectrum disorder (ASD), BD, MDD, OCD,
   schizophrenia, and Tourette’s syndrome (TS) (n = 9,725–933,970). These
   neuropsychiatric disorders were chosen because they are the best
   genetically characterised by the Psychiatric Genomics Consortium
   [[113]24]. Further details, including sample size of each GWAS, are
   reported in Table [114]1. To maintain consistency in genetic data,
   analyses were confined to individuals of European ancestry, employing
   the human genome build GRCh37/hg19 as a reference. All statistical
   analyses were performed using R v4.2.1 (2022-06-23).

Table 1.

   Characteristics of genome-wide association study (GWAS) summary
   statistics used as input for Local Analysis of [Co]variant Association
   (LAVA) and follow-up genomic analyses included in this study.
   Phenotype Authors Year PMID Ancestry N Cases Controls N[eff]
   MetS Lind 2019 31589552 European 291,107 59,677 231,430 189,772.81
   Obesity Watanabe et al. 2019 31427789 European 244,890 9805 235,085
   37,649.69
   T2DM Mahajan et al. 2022 35551307 European 933,970 80,154 853,816
   293,100.50
   AD Wightman et al. 2021 34493870 European 762,917 86,531 676,386
   306,866.18
   ADHD Demontis et al. 2023 36702997 European 225,534 38,691 186,843
   128,213.80
   AN Watson et al. 2019 31308545 European 72,517 16,992 55,525 52,041.91
   ASD Grove et al. 2019 30804558 European 46,350 18,381 27,969 44,366.62
   BD Mullins et al. 2021 34002096 European 413,466 41,917 371,549
   150,669.89
   OCD IOCDF-GC/OCGAS 2018 28761083 European 9725 2688 7037 7,780.14
   MDD Howard et al. 2019 30718901 European 500,199 170,756 329,443
   449,855.91
   Schizophrenia Trubetskoy et al. 2022 35396580 European 130,644 53,386
   77,258 126,281.98
   TS Yu et al. 2019 30818990 European 14,307 4819 9488 12,783.30
   [115]Open in a new tab

   MetS metabolic syndrome, T2DM type 2 diabetes mellitus, AD Alzheimer’s
   disease, ADHD attention-deficit/ hyperactivity disorder, AN anorexia
   nervosa, ASD autism spectrum disorder, BD bipolar disorder, MDD major
   depressive disorder, OCD obsessive-compulsive disorder, IOCDF-GC/OCGAS
   International OCD Foundation Genetics Collaborative/OCD Collaborative
   Genetics Association Studies, TS Tourette’s syndrome, PMID PubMed ID, N
   total sample size, Neff effective sample size
   [Neff = 4/(1/Cases + 1/Controls)].

Local genetic correlation analyses

   We utilised the R package LAVA (Local Analysis of [co]Variant
   Association) ([116]https://github.com/josefin-werme/LAVA) to perform
   pairwise local genetic correlation analyses between the three
   IR-related conditions and the nine neuropsychiatric disorders
   [[117]22]. Compared to traditional global correlation analysis methods
   [[118]23], LAVA estimates the genetic correlation at smaller genomic
   loci, which provides a more fine-grained overview of the genetic
   overlap between traits. In addition to providing insight into the
   potentially heterogeneous nature of the shared association patterns
   across the genome, LAVA allows identification of the regions from which
   the pleiotropy is originating [[119]22]. Further details regarding the
   LAVA analytical steps are provided in the [120]Supplementary
   information. Given the total number of bivariate tests performed across
   all phenotype pairs, local genetic correlations were deemed as
   statistically significant at a maximum acceptable false discovery rate
   (FDR) of q = 0.05, following the approach of Hindley, Frei [[121]25].

Positional and eQTL gene mapping

   The biomaRt R package (version 2.54.1) (10.18129/B9.bioc.biomaRt)
   [[122]26] was used to annotate single-nucleotide polymorphisms (SNPs)
   within each genetically correlated region and positionally map them to
   genes. We used the Ensembl database (release 109, GRCh37/hg19, homo
   sapiens) as a reference for gene annotations. We defined filters to
   specify the genomic regions of interest based on their location
   (chromosome number, start and end positions).

   For the eQTL-based gene mapping, the loci2path R package (version
   1.3.1) (10.18129/B9.bioc.loci2path) [[123]27] was used to identify
   eQTLs within the genetically correlated regions that may influence gene
   expression in 13 cortical, subcortical, and cerebellar brain regions
   (i.e., total brain cortex, frontal cortex BA9, hippocampus,
   hypothalamus, amygdala, anterior cingulate cortex BA24, caudate,
   nucleus accumbens, putamen, cervical spinal cord, substantia nigra,
   cerebellar hemisphere, cerebellum). We obtained the eQTL data from the
   Genotype-Tissue Expression (GTEx) project (GTEx V8, GRCh38/hg38)
   ([124]https://gtexportal.org/home/dataset) and restricted our analysis
   to brain tissues due to their relevance to neuropsychiatric disorders.
   Prior to the analysis, we lifted the eQTL coordinates to the
   GRCh37/hg19 genomic build using the UCSC LiftOver tool
   ([125]https://genome-store.ucsc.edu) to align with the used GWAS
   summary statistics.

Functional annotation of genetically correlated regions

   Functional annotation analyses were conducted separately for each
   phenotype pair where genetically correlated regions were found. We
   employed the GENE2FUNC module within the Functional Mapping and
   Annotation of Genome-Wide Association Studies (FUMA) platform
   [[126]28], using default parameters and multiple testing correction
   [[127]28]. This approach served to examine important properties of the
   mapped genes, such as their tissue-specific and temporal expression
   profiles, enrichment in predefined gene sets, potential as drug
   targets, and previous trait/disease associations. Detailed information
   on the methods applied for these analyses are presented in the
   [128]Supplementary information.

   To contextualise our findings within the broader landscape of known
   disease associations, we also investigated the overrepresentation of
   the identified genes within those previously associated with traits or
   diseases by querying the NHGRI-EBI GWAS Catalog [[129]29].

Colocalisation analyses

   To identify the specific shared causal variants within each region
   showing local genetic correlation, we conducted robust Bayesian
   colocalisation analyses through the coloc R package [[130]30] and the
   Sum of Single Effects (SuSiE) regression framework [[131]31]
   ([132]https://chr1swallace.github.io/coloc/articles/a06_SuSiE.html).
   Notably, these approaches allow for simultaneous evaluation of multiple
   causal genetic variants within a genomic region and are therefore not
   limited by the single causal variant assumption that traditional
   colocalisation methods use. The input genomic regions were those
   showing evidence of local genetic correlation between each pair of
   IR-related conditions and neuropsychiatric disorders. The detailed
   methodology is reported in the [133]Supplementary information.

Functional annotation of 95% credible sets of shared causal variants

   We employed the SNPnexus web server ([134]https://www.snp-nexus.org/)
   to further characterise the functional significance of the likely
   causal variants identified by colocalisation [[135]32]. This tool
   integrates a wealth of genomic and functional annotation resources to
   elucidate the potential biological consequences of variants on gene
   structure, regulation, and function. The analysis encompassed several
   annotation categories, including gene annotations, regulatory elements
   (e.g., miRBASE, CpG islands), and non-coding scoring (i.e.,
   deleteriousness Combined Annotation Dependent Depletion [CADD] scores),
   along with pathway enrichment analysis of credible set variants
   [[136]32]. A detailed description of these steps is provided in the
   [137]Supplementary information.

   Finally, the drugs/compounds that target genes mapped to likely causal
   variants were sourced from GeneCards, independent from their approved
   or investigational status. GeneCards is an online platform that gathers
   information from multiple databases including DrugBank, PharmaGKB,
   ClinicalTrials, DGIdb, the Human Metabolome Database, and Novoseek
   [[138]33].

Results

Local patterns of genetic correlation between IR-related conditions and
neuropsychiatric disorders

   For each pair consisting of an IR-related condition and a
   neuropsychiatric disorder, bivariate local genetic correlation was
   evaluated in all genomic regions for which both phenotypes exhibited a
   univariate signal at p < 1 × 10^−4, resulting in a total of 2251 tests.
   Of note, only 19.6% of the regions with significant local SNP-based
   heritability (h^2[SNP]) for both phenotypes showed a bivariate
   p < 0.05, indicating that significant local h^2[SNP] is often present
   without any local correlation signal between neuropsychiatric and
   IR-related conditions. After FDR correction, moderate to high degrees
   of local genetic correlations (|r[g]| = 0.21–1, p[FDR] < 0.05) were
   identified for 20 of the 27 phenotype pairs examined, across 109 unique
   semi-independent genomic regions (see Fig. [139]1 and Table [140]2).
   Noteworthy, local genetic correlations also emerged between IR-related
   conditions and neuropsychiatric disorders that had not shown
   significant global genetic correlations, namely AD, BD, and TS
   [[141]8]. In total, 128 FDR-significant local genetic correlations were
   identified, of which 75 with a positive direction of the effect and 53
   with a negative direction (Table [142]2; detailed results are provided
   in Table [143]S1; see also Fig. [144]1b, c). For 59 (46.1%) of the 128
   local correlations, the 95% confidence intervals (CIs) for the
   explained variance included the value 1, consistent with a scenario
   where the local genetic signal for those phenotype pairs is entirely
   shared (Table [145]2). Interestingly, exclusively positive local
   genetic correlations were found between IR-related conditions and
   ADHD/MDD, while those detected between IR-related conditions and AN
   were all negative. No local genetic correlation was found between ASD
   and IR-related conditions. Conversely, a combination of positive and
   negative local genetic correlations was detected between all the other
   IR-related and neuropsychiatric conditions (Fig. [146]1, Table [147]2),
   of which all but MetS-schizophrenia had no previous evidence of global
   genetic overlap (see Table [148]2).

Fig. 1. Local genetic correlations between neuropsychiatric and insulin
resistance-related conditions.

   Fig. 1
   [149]Open in a new tab

   a Chord diagram representing the network of local genetic correlations
   between insulin resistance-related conditions and neuropsychiatric
   disorders. A higher width of a ribbon reflects a higher number of
   shared genetically correlated loci between two phenotypes, highlighting
   a substantial polygenic overlap and suggesting potential shared
   pathophysiological mechanisms between them. The colours of the ribbons
   are used purely for visual distinction and do not imply any additional
   significance or categorisation. Labels for insulin resistance–related
   conditions are displayed in dark cyan, while those for neuropsychiatric
   disorders are shown in black. b Bar plot presenting the number of local
   genetic correlations identified between neuropsychiatric disorders and
   insulin resistance-related conditions. Each bar corresponds to a
   different neuropsychiatric disorder, segmented by the direction of
   effect of local genetic correlations, with blue indicating negative and
   red indicating positive local genetic correlations between
   neuropsychiatric disorders and insulin resistance-related conditions.
   The height of each bar reflects the quantity of local genetic
   correlations detected for each disorder. c Network visualisation of
   local genetic correlations between a spectrum of neuropsychiatric
   disorders and insulin resistance-related conditions. Nodes represent
   distinct phenotypes for which local bivariate genetic correlations were
   evaluated; nodes for insulin resistance–related conditions are labeled
   in dark cyan, while those for neuropsychiatric disorders are labeled in
   black. Edges connecting the nodes vary in width proportionally to the
   number of local genetic correlations identified between phenotype
   pairs. Edge colour denotes the direction of the genetic correlation
   estimate, with red indicating a positive correlation and blue
   indicating a negative correlation. Abbreviations: AD, Alzheimer’s
   disease; ADHD, attention-deficit/hyperactivity disorder; AN, anorexia
   nervosa; ASD, autism spectrum disorder; BD, bipolar disorder; MDD,
   major depressive disorder; MetS, metabolic syndrome, OCD,
   obsessive-compulsive disorder; T2DM, type 2 diabetes mellitus; SCZ,
   schizophrenia; TS, Tourette’s syndrome.

Table 2.

   Summary of local genetic correlations between neuropsychiatric
   disorders and insulin resistance-related conditions.
   Neuropsychiatric disorders Insulin resistance-related conditions
   Previous evidence of bivariate global genetic correlation N loci with
   significant local h^2[SNP] simultaneously for both phenotypes (% of
   loci tested) N genetically correlated loci^b (% of loci with
   significant^a local h^2[SNP] for both phenotypes) N positively
   correlated loci (% of correlated loci for the pair)^b N negatively
   correlated loci (% of correlated loci for the pair)^b N genetic
   correlations estimates whose confidence intervals included 1 (% of
   correlated loci for the pair)^b
   AD MetS NO 306 (12.3%) 13 (4.2%) 6 (46.2%) 7 (53.8%) 2 (15.4%)
   AD Obesity NO 34 (1.4%) 1 (2.9%) 1 (100.0%) 0 (0.0%) 1 (100.0%)
   AD T2DM NO 214 (8.6%) 16 (7.5%) 12 (75.0%) 4 (25.0%) 3 (18.8%)
   ADHD MetS YES 94 (3.8%) 10 (10.6%) 10 (100.0%) 0 (0.0%) 9 (90.0%)
   ADHD Obesity YES 30 (1.2%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
   ADHD T2DM YES 88 (3.5%) 12 (13.6%) 12 (100.0%) 0 (0.0%) 7 (58.3%)
   AN MetS YES 65 (2.6%) 1 (1.5%) 0 (0.0%) 1 (100.0%) 1 (100.0%)
   AN Obesity YES 40 (1.6%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
   AN T2DM YES 73 (2.9%) 3 (4.1%) 0 (0.0%) 3 (100.0%) 2 (66.7%)
   ASD MetS YES 21 (0.8%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
   ASD Obesity YES 13 (0.5%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
   ASD T2DM YES 12 (0.5%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
   BD MetS NO 121 (4.8%) 10 (8.3%) 6 (60.0%) 4 (40.0%) 0 (0.0%)
   BD Obesity NO 53 (2.1%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
   BD T2DM NO 107 (4.3%) 6 (5.6%) 1 (16.7%) 5 (83.3%) 1 (16.7%)
   MDD MetS YES 95 (3.8%) 5 (5.3%) 5 (100.0%) 0 (0.0%) 3 (60.0%)
   MDD Obesity YES 45 (1.8%) 4 (8.9%) 4 (100.0%) 0 (0.0%) 3 (75.0%)
   MDD T2DM YES 83 (3.3%) 8 (9.6%) 8 (100.0%) 0 (0.0%) 4 (50.0%)
   OCD MetS YES 71 (2.8%) 3 (4.2%) 0 (0.0%) 3 (100.0%) 2 (66.7%)
   OCD Obesity YES 40 (1.6%) 1 (2.5%) 0 (0.0%) 1 (100.0%) 0 (0.0%)
   OCD T2DM YES 54 (2.2%) 1 (1.9%) 1 (100.0%) 0 (0.0%) 0 (0.0%)
   Schizophrenia MetS YES 200 (8.0%) 10 (5.0%) 2 (20.0%) 8 (80.0%) 6
   (60.0%)
   Schizophrenia Obesity NO 89 (3.6%) 1 (1.1%) 0 (0.0%) 1 (100.0%) 1
   (100.0%)
   Schizophrenia T2DM NO 172 (6.9%) 18 (10.5%) 4 (22.2%) 14 (77.8%) 10
   (55.6%)
   TS MetS NO 57 (2.3%) 1 (1.8%) 1 (100.0%) 0 (0.0%) 1 (100.0%)
   TS Obesity NO 27 (1.1%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)
   TS T2DM NO 47 (1.9%) 4 (8.5%) 2 (50.0%) 2 (50.0%) 3 (75.0%)
   TOTAL 2251 128 (5.7%) 75 (58.6%) 53 (41.4%) 59 (46.1%)
   [150]Open in a new tab

   For each neuropsychiatric disorder, the table presents the presence or
   absence of global genetic correlation with the specific insulin
   resistance-related condition, significant loci with h^2[SNP] for both
   phenotypes tested, and the directionality of local genetic correlations
   with T2DM, Metabolic Syndrome (MetS), and obesity. The table further
   provides the number of local genetic correlations with confidence
   intervals that include 1, indicating potential for completely
   overlapping genetic influences at specific loci.

   AD Alzheimer’s disease, ADHD attention-deficit/ hyperactivity disorder,
   AN anorexia nervosa, ASD autism spectrum disorder, BD bipolar disorder,
   h^2[SNP], single-nucleotide polymorphism (SNP)-based heritability, MetS
   metabolic syndrome, MDD major depressive disorder, N number of, OCD
   obsessive-compulsive disorder, T2DM type 2 diabetes mellitus, TS
   Tourette’s syndrome.

   ^ap < 1 × 10^−4; ^bFalse discovery rate (FDR) at maximum q = 0.05.

   Furthermore, fifteen out of the 109 unique regions were associated with
   more than one phenotypic pair (Table [151]S1; we refer to these here as
   hotspots). The major hotspots showing significant bivariate local r[g]s
   between multiple phenotypic pairs were the chr2:59251997-60775066
   (between T2DM-ADHD, MetS-AN, MetS-MDD), chr6:31320269-31427209
   (MetS-AD, T2DM-AD, T2DM-schizophrenia), and chr16:29043178-31384210
   genomic regions (MetS-schizophrenia, obesity-schizophrenia,
   T2DM-schizophrenia) (see Table [152]S1 and Fig. [153]S1). Notably,
   11.71% of the genetically correlated regions detected here (15/128) are
   located in the Major Histocompatibility Complex (MHC) region
   (chr6:28477797-33448354). All r[g]s detected in the MHC were between
   T2DM/MetS and either schizophrenia, AD, or BP, with prevalence of a
   negative direction of the effect (Table [154]S1).

Genes underlying insulin resistance-neuropsychiatric multimorbidity

   In the regions where we detected significant local genetic
   correlations, we identified a total of 1455 distinct genes through
   eQTL-based mapping, and 1495 unique protein-coding genes through
   positional mapping across all phenotype pairs (Table [155]S2–[156]3).
   Notably, the pseudogene CYP21A1P was recurrently eQTL-mapped across
   multiple phenotype pairs (AD-T2DM, AD-MetS, BD-T2DM,
   schizophrenia-T2DM). In total, 140 genes were mapped for at least three
   phenotype pairs, indicating a potentially broader relevance in the
   genetic architecture of IR-neuropsychiatric multimorbidity (Table
   [157]S3). Within this subset, 20 genes, all located within the MHC
   region, were involved in immune-inflammation and vesicle
   metabolism/trafficking (e.g., HLA-B, MICA, C4A, C4B, AGER, BTNL2,
   HLA-DRA, HLA-DRB1, HLA-DQA1, PSMB8, HLA-DRB5, and FLOT1), and four
   genes were involved in insulin signalling and secretion (i.e., STX1A,
   FLOT1, MAPK3, and PHKG2) (see Table [158]S3).

Functional annotation of the identified regions

   Considering the genes mapped to the regions showing local correlations,
   411 gene sets were significantly enriched (Table [159]S4).
   Immune-related pathways were prominently represented for multiple
   phenotype pairs (i.e., AD-MetS/T2DM, BD-T2DM, TS-T2DM,
   schizophrenia-T2DM). Other biological pathways related to oxygen
   transport, lipid metabolism (including omega-3 and omega-6
   polyunsaturated fatty acid levels (PUFAs)), embryonic/placental
   development, insulin receptor/phosphoinositide 3-kinase (PI3K), and
   vesicular function/secretion were enriched across different phenotype
   pairs (Table [160]S4). Pharmacogenomic markers, notably genes
   genome-wide significantly associated with response to metformin (i.e.,
   STX1B, STX4, ZNF668), were enriched in regions shared between
   schizophrenia and MetS, obesity, and T2DM (Table [161]S5).

   In a more granular examination, we also evaluated enrichment of
   life-stage-specific expression profiles for genes mapped to the
   genetically correlated regions (Tables [162]S7–[163]8). Specifically,
   regions correlated between schizophrenia and obesity featured genes
   upregulated at 19 weeks post-conception. Conversely, regions associated
   with the schizophrenia-MetS pair exhibited a distinct pattern, with
   genes showing downregulation in brain samples from individuals at age
   11. Furthermore, regions of overlap between OCD and MetS held genes
   upregulated in early adulthood brain tissues, while the genes in the
   overlapping regions marking the OCD-obesity pair exhibited gene
   downregulation in late childhood.

   Detailed results for gene set analysis, spatio-temporal expression
   specificity of the mapped genes, and druggable gene annotations are
   reported in Table [164]S4–[165]S10.

Shared causal variants between insulin resistance-related conditions and
neuropsychiatric disorders

   Of the 128 regions identified with local r[g], colocalisation analyses
   successfully pinpointed the likely causal variants driving this
   association in 10 regions (see Table [166]S11–[167]12). For
   comprehensive functional annotations of the 95% credible set variants
   within these 10 regions see Tables [168]S13–[169]S23.

   Notably, one region on chromosome 4 and two on chromosome 6 showed the
   highest posterior probability for colocalisation, linking schizophrenia
   with MetS and AD with T2DM, respectively (Table [170]S11–[171]12, Fig.
   [172]S2–[173]4). The schizophrenia-MetS relationship implicated the
   rs13107325 variant in the SLC39A8 gene, which modulates the activity of
   the miRNA hsa-miR-374b-5p (Tables [174]S12–[175]S14). For the AD-T2DM
   pair, the likely causal variants were rs9271608 and rs9275599, mapped
   to the HLA-DRB1 and MTCO3P1 genes, respectively. According to
   GeneCards, HLA-DRB1 is targeted by immunosuppressive and
   anti-inflammatory drugs (e.g., azathioprine, lapatinib, interferons-β,
   and acetylsalicylic acid), as well as by statins and psychotropic drugs
   (e.g., carbamazepine, clozapine, and lamotrigine) (Table [176]S23).

   Further seven regions had good support for colocalisation
   ([177]Supplementary information); these regions showed local genetic
   correlations for the AD-T2DM, MDD-T2DM, BD-MetS, and schizophrenia-MetS
   pairs (Table [178]S11–[179]12). Most of the identified variants were
   observed within or near genes pivotal to immune function, vesicle/small
   molecules trafficking, lipid metabolism, organ development, retinoic
   acid signalling, and DNA repair/apoptosis (Tables [180]S17–[181]18).
   The variants often had high CADD PHRED scores, suggesting highly
   deleterious effects (Tables [182]S13). Genes mapping to these variants,
   like the HLA-DQB1 and FADS1/2 genes, are targeted by existing drugs and
   supplements, such as antihypertensive drugs, omega-3/6 PUFAs, and
   vitamin A (Table [183]S23).

Discussion

   In this study, we examined the genetic relationship between IR-related
   conditions - specifically, obesity, T2DM, and MetS - and nine
   neuropsychiatric disorders by investigating the pairwise patterns of
   local genetic correlation across the genome. At the same time, we
   explored the specific genetic factors and biological mechanisms
   underlying their multimorbidity. The results presented here offer novel
   insights into the shared genetic aetiology between these phenotypes,
   unveiling a complex pattern of both positive and negative local genetic
   correlations. For the first time, we demonstrated that even in the
   absence of global genetic correlations, significant local correlations
   exist (i.e., between AD, BD, TS and IR-related conditions). These
   findings expand the results of previous studies [[184]8, [185]10], with
   important implications for understanding the pathophysiology of these
   disorders and for developing targeted therapeutic interventions
   addressing IR-psychiatric multimorbidity. We identified 128 local
   genetic correlations across 109 unique genomic regions. Notably, the
   MHC region emerged as a particularly significant contributor in terms
   of shared genetic signal, as confirmed by enrichment in biological
   pathways related to immune function. In addition, genes mapped to the
   genetically correlated regions showed enrichment in pathways involved
   in lipid metabolism, insulin signalling, and vesicular function, among
   others.

   Regarding the directions of the detected genetic correlations, we
   observed exclusively positive local genetic correlations for ADHD and
   MDD with IR-related conditions, indicating synergistic genetic effects
   that predispose to both neuropsychiatric symptoms and IR-related
   conditions. Our enrichment analyses of the genes mapped to these
   regions suggest that the genetic overlap might be mediated by genes
   involved in extracellular matrix organisation, vesicle trafficking, and
   oxygen transport/oxidative processes. These pathways are involved in
   both brain function and metabolic regulation [[186]34–[187]36]. In
   particular, extracellular matrix molecules are implicated in synaptic
   plasticity and homeostasis [[188]34] and may also influence tissue
   insulin sensitivity [[189]37]. Similarly, vesicle trafficking, integral
   to synaptic function and neurotransmission, could be a central theme
   where neuronal communication and insulin signalling intersect,
   contributing to the multimorbidity of the conditions [[190]36].
   Conversely, we identified exclusively negative correlations between AN
   and IR-related conditions. These results align with the distinct
   phenotypic characteristics of AN, including increased insulin
   sensitivity and metabolic alterations related to undernutrition, which
   differ markedly from other neuropsychiatric disorders [[191]38,
   [192]39].

   While phenotypic overlap of AD and BD with IR-related conditions has
   been frequently reported (e.g., Wimberley, Horsdal [[193]2], Santiago
   and Potashkin [[194]40])), previous genetic analyses did not find
   global genetic correlations between these phenotypes [[195]8]. This may
   have occurred due to the averaging effect of global analyses. Our
   study, which is the first to report significant local genetic
   correlations between AD, BD, TS and IR-related conditions, suggests
   that positive and negative local correlations could neutralise each
   other in global correlation analyses, a phenomenon observed in other
   recent studies [[196]41, [197]42]. These heterogeneous patterns of
   genetic overlap could also point towards aetiologically distinct
   subgroups that warrant further exploration with deep phenotyping and
   functional validation. Such analyses could bring us closer towards
   precision medicine, offering the potential for personalised healthcare
   and improved treatment success [[198]43].

   Multiple genomic regions (15 out of 109) showed significant
   correlations for more than one phenotype pairs, implying a potentially
   more prominent and ubiquitous role in the IR-neuropsychiatric
   multimorbidity. Among the recurring regions, chr2:59251997-60775066,
   mapping to the BCL11A gene, was implicated in the correlation of T2DM
   with ADHD, MetS with AN, and MetS with MDD. BCL11A codes for a
   transcription factor essential for B cell function and haematopoiesis,
   as well as for neuronal development, regulating processes such as
   neurogenesis/axonogenesis, and neuronal migration [[199]44, [200]45].
   BCL11A variants have also been associated with neurodevelopmental
   disorders and impaired cognition, as well as with IR in in vivo and in
   vitro studies [[201]45–[202]47]. Among other genes that were mapped
   across at least three phenotypic pairs, some (i.e., STX1A, FLOT1,
   MAPK3, and PHKG2) are pivotal in insulin signalling and secretion
   [[203]48–[204]50]. These findings strengthen a molecular basis for
   linking neuropsychiatric disorders to altered insulin function [[205]8,
   [206]50, [207]51]), which has also been tied to cognitive deficits,
   anhedonia, and reward processing alterations [[208]3, [209]4, [210]6].

   Over 11% of the correlated genomic regions were located within the MHC
   region (chr6:28477797-33448354), where extensive pleiotropy has been
   demonstrated previously [[211]22, [212]52]. This region is renowned for
   its high gene density, polymorphism, and involvement in
   immune-inflammatory responses [[213]53]. The influence of the MHC
   region extends beyond autoimmune and infectious diseases
   susceptibility, being also associated with neuropsychiatric disorders,
   such as ASD, schizophrenia, and BD [[214]54]. Our findings point to a
   plausible genetic link between IR-related metabolic dysfunction,
   immune-inflammatory dysfunction, and neuropsychiatric disorders. This
   is consistent with previous findings indicating that both central and
   peripheral inflammation may mediate the link between IR and
   neuropsychiatric conditions [[215]55, [216]56]. Inflammation may also
   impair brain insulin signalling, potentially resulting in
   neurobehavioural consequences [[217]57]. Notably, most of the local
   genetic correlations identified within the MHC region showed a negative
   direction of effect. We cannot provide a clear explanation of this
   finding, but it may lie in the balance of pro-inflammatory and
   anti-inflammatory factors in immune response, in which MHC genes play a
   role [[218]54]. Additionally, MHC class I (MHC-I) molecules,
   traditionally associated with immune functions, have also been
   implicated in synapse pruning, a process important for refining neural
   circuits during development [[219]58]. MHC-I molecules are expressed in
   neurons and modulate microglia-mediated synapse elimination by marking
   less active synapses for phagocytosis [[220]59, [221]60]. This
   activity-dependent mechanism shapes functional neuronal networks and
   has been implicated in pathological synapse loss in neurodegenerative
   conditions [[222]59, [223]61]. Dysregulated MHC-I signalling can lead
   to aberrant synaptic pruning, implicated in disorders such as
   schizophrenia and ASD [[224]58]. Hence, the dual role of MHC-encoded
   molecules in both immune modulation and synaptic plasticity, as well as
   the potential differential expression of genes in the MHC region across
   different tissues and the lifespan may help explain the observed
   negative genetic correlations [[225]62]. Experimental validation of our
   findings will be necessary to determine the exact functional
   implication of the observed genetic associations.

   Relatedly, our study identified multiple genes related to the human
   leukocyte antigen (HLA) system, innate immunity, and immunomodulation
   (i.e., HLA-B, HLA-DRA, HLA-DRB1, HLA-DQA1, HLA-DRB5, MICA, C4A, C4B,
   AGER, PSMB8, and BTNL2), supporting their possible influence on
   IR-neuropsychiatric multimorbidity. Of note, immunomodulatory drugs
   (e.g., non-steroidal anti-inflammatory drugs and monoclonal antibodies)
   have shown some efficacy as add-on treatments in psychoses and MDD, and
   might have higher efficacy in people with IR-neuropsychiatric
   multimorbidity [[226]63, [227]64]. Another gene recurrently mapped
   across various phenotype pairs was the CYP21A1P pseudogene, located
   within the MHC region. Intergenic recombination of CYP21A1P leads to
   altered glucocorticoid and androgen production [[228]65];
   glucocorticoids have anti-inflammatory/immunosuppressive effects, and
   regulate glucose metabolism and the body’s stress response [[229]66].
   Specifically, glucocorticoids counteract insulin by decreasing
   peripheral glucose uptake and stimulating hepatic gluconeogenesis,
   leading to IR under conditions of excessive release, such as in chronic
   stress [[230]67]. Prolonged exposure to glucocorticoids can induce
   neurotoxic effects, possibly involved in the development of psychiatric
   disorders [[231]68, [232]69]. These hormones also modulate the
   serotonergic system, which is strongly implicated in psychiatric
   disorders and insulin signalling [[233]70, [234]71]. Interestingly,
   gene set enrichments within correlated regions between schizophrenia
   and IR-related conditions were related to the response to metformin, a
   frontline oral medication for T2DM. This implies a potential overlap in
   therapeutic targets between schizophrenia and T2DM, which could lead to
   a reassessment of treatment strategies for these patients. Previous
   randomised-controlled trials (RCTs) confirmed the efficacy of metformin
   in counteracting antipsychotic-induced metabolic side effects in
   individuals with psychoses [[235]72, [236]73], while improving
   psychiatric and cognitive symptoms in the same population [[237]74].

   Another significant finding of this study was the identification of
   colocalisation signals. Among the 128 regions demonstrating local
   genetic correlation, 10 regions were prioritised for their high
   posterior probabilities of harbouring the same causal variants shared
   between IR-related conditions and neuropsychiatric disorders. This was
   instrumental for further elucidating shared pathophysiological
   mechanisms and novel potential drug targets for IR-neuropsychiatric
   multimorbidity [[238]75, [239]76]. The two most likely shared causal
   variants were located in the chr4:102544804-104384534 and
   chr6:32586785-32629239/chr6:32682214-32897998 regions, suggesting novel
   cross-links between schizophrenia and MetS, and AD and T2DM,
   respectively. The identified shared causal variant (rs13107325) between
   schizophrenia and MetS maps to the SLC39A8 gene, encoding the ZIP8
   metal cation transporter. Previous studies demonstrated its association
   with altered brain manganese levels and protein complexity in
   schizophrenia, brain morphology and dendritic spine density, as well as
   a broader impact on various conditions, including developmental,
   neuropsychiatric and cardio-metabolic diseases/traits
   [[240]77–[241]80]. Our findings also highlight SLC39A8’s potential as a
   therapeutic target via zinc chloride/sulphate [[242]81]. Interestingly,
   RCTs have shown beneficial effects of zinc sulphate in reducing
   symptoms of ADHD, MDD, and SCZ [[243]82–[244]84], as well as improving
   glucose handling in prediabetes [[245]85]. In the AD-T2DM context, the
   rs9271608 variant mapping to the HLA-DRB1 gene presented compelling
   causal candidacy, pointing to the potential for immunosuppressive drugs
   such as azathioprine, lapatinib, and interferons-β to influence AD-T2DM
   manifestations. The administration of intranasal treatment with
   interferon-β was shown to improve anxious/depressive-like behaviours by
   modulating microglia polarisation in AD rat models [[246]86]. Of note,
   the rs9271608 also shows broad biological relevance as it is active as
   a promoter across numerous cell types and tissues, including various
   immune and neuronal progenitors [[247]87]. The remaining regions of
   notable colocalisation underpin associations between AD and T2DM, MDD
   and T2DM, BD and MetS, and schizophrenia and MetS, hinting at potential
   targetable mechanisms for current drugs and supplements, including
   antihypertensive drugs, omega-3/6 PUFAs, vitamin A [[248]81]. Several
   antihypertensive drugs have been associated with a reduced risk of
   depression [[249]88], and omega-3 PUFAs showed beneficial effects on
   depression symptoms in a meta-analysis of RCTs [[250]89]. Genes
   associated with omega-3/omega-6 PUFAs were enriched when considering
   the regions showing correlation between BD and MetS, in line with their
   relevance in the multimorbidity. Finally, vitamin A inhibits amyloid β
   protein deposition, tau phosphorylation, neuronal degeneration and
   improves spatial learning and memory in AD mouse models [[251]90]. It
   is worth noting that a significant local genetic correlation without
   detectable colocalisation does not necessarily mean that there are no
   shared causal variants; this may reflect limitations in the power of
   the colocalisation analysis, particularly in scenarios with complex
   patterns of associations, which are often observed in highly polygenic
   traits [[252]22].

   Our study should be viewed considering some limitations. Although our
   study may serve as a starting point by highlighting potential shared
   causal variants and suggesting biological mechanisms through which
   shared genetic regions might impact both mental and metabolic health,
   the functional interpretation of our findings remains largely
   speculative; future in vitro and animal model studies will be necessary
   to validate our findings and provide clearer mechanistic insights into
   the underlying biology. The high linkage disequilibrium in the MHC
   region may have led to spurious pleiotropy, not necessarily implying
   the presence of the same shared causal SNPs [[253]91]. Rare genetic
   variants were not considered, and population-specific effects may not
   be adequately captured by our analyses, which were limited to European
   ancestry. While the available GWAS summary statistics were generally
   obtained in samples of adequate size for this kind of study, the GWAS
   summary statistics for OCD were based on a relatively small sample
   size, potentially influencing the number of significant local genetic
   correlations detected by LAVA.

   In conclusion, our study provides novel insights into the shared
   genetic underpinnings of neuropsychiatric and IR-related conditions,
   challenging traditional notions of their separate pathophysiology. Our
   results support a more integrated disease model, and the need to move
   beyond the conventional view of distinct aetiologies. The implications
   of our findings extend to clinical practice, emphasising the need for a
   holistic approach in the screening and management of
   IR-neuropsychiatric multimorbidity. For example, the importance of
   lifestyle interventions for both metabolic and psychiatric health, and
   for developing pharmacological treatments that target both conditions.
   The discovery of shared causal variants, particularly in genes like
   SLC39A8 and HLA-DRB1, opens new avenues for targeted therapeutic
   interventions. The convergence of genetic findings on mechanisms
   related to immune-inflammation, insulin signalling, lipid metabolism,
   vesicle trafficking, among others, provides a compelling direction for
   future research. Overall, our study not only unveils the shared genetic
   landscape of neuropsychiatric and IR-related conditions, but also
   establishes a foundation for integrated research and treatment
   approaches, contributing to a paradigm shift towards comprehensive care
   strategies that address the issue of multimorbidity.

Supplementary information

   [254]Supplementary Information^ (60.3KB, docx)
   [255]Supplementary Tables^ (6.3MB, xlsx)
   [256]Supplementary Figures^ (471.7KB, docx)

Acknowledgements