Skip to content

Genetics and neurobiology of bipolar disorder

TL;DR — Bipolar disorder is highly polygenic: modern genome-wide studies identify dozens of loci, each with small effects, and show extensive genetic sharing with schizophrenia, major depression and ADHD rather than a disorder-specific molecular lesion. The strongest convergent biology implicates neuronal and synaptic signaling, calcium-channel biology and brain-expressed regulatory genes, but polygenic scores explain too little variance for diagnosis or treatment selection. Large-scale imaging finds small, distributed group differences in cortical thickness, subcortical volume and white-matter integrity; medication, illness state and age complicate causal interpretation. CRP and some other peripheral signals differ at group level, but null or state-dependent results are common. No genetic, imaging, electrophysiological or blood test is validated as a clinical bipolar biomarker.

Genetic architecture

Family aggregation reflects many common and rare variants plus environmental and developmental influences; the evidence does not support a single “bipolar gene.” A 2021 GWAS of 41,917 cases and 371,549 controls identified 64 associated loci, with enrichment in synaptic-signaling pathways and neuronal genes expressed in prefrontal cortex and hippocampus (Mullins 2021, PMID 34002096).

The immediately preceding Psychiatric Genomics Consortium analysis included 20,352 cases and 31,358 controls, followed by 9,412 cases and 137,760 controls, and identified 30 genome-wide significant loci, 20 of them novel (Stahl 2019, PMID 31043756). The expansion from suggestive signals in earlier studies to dozens of robust loci illustrates that individual effects are small and require very large samples.

Study Ancestry/sample Main quantitative result Interpretation
Scott 2009 3,683 cases; 14,507 controls ITIH1 rs1042779 OR 1.19, P=1.8×10^-7; ANK3 replication OR 1.37, P=0.042 Historical small-sample GWAS produced mainly suggestive signals
Stahl 2019 20,352 cases; 31,358 controls plus follow-up 30 loci; 20 novel Ion channels, transporters and synaptic components emerged
Mullins 2021 41,917 cases; 371,549 controls 64 loci; 15 genes prioritized by expression integration Scale sharpened biology but did not yield a diagnostic test
Li 2021 Han Chinese discovery 1,822/4,650; replication 958/2,050 TMEM108 rs9863544 OR 0.650 (95% CI 0.559–0.756) Supports both shared and ancestry-sensitive architecture

The early European meta-analysis analyzed more than 2.3 million variants; its strongest nonsynonymous association, in ITIH1, had OR 1.19, illustrating the modest effects typical of common variants (Scott 2009, PMID 19416921). ANK3 was independently replicated in that analysis, but no single locus approaches Mendelian predictiveness.

Cross-ancestry evidence

In Han Chinese participants, a TMEM108-region variant reached genome-wide significance, while trans-ancestry correlation with European GWAS was 0.652 (SE 0.106; P=7.30×10^-10) (Li 2021, PMID 33263727). European-derived polygenic scores explained only 1.27% of liability-scale pseudo-R² in that study, demonstrating why present scores cannot confirm or exclude bipolar disorder in an individual.

The evidence base remains disproportionately European. Different allele frequencies, linkage disequilibrium and environmental structure can reduce portability even when causal biology is shared (Li 2021, PMID 33263727). Rare copy-number variation shows the same ancestry limit: in 3,915 Han Chinese bipolar cases and 7,820 matched controls, rare deletions were overrepresented and enriched in neural-signaling and dosage-sensitive genes, but of 12 European-reported CNV loci only deletions at 3q29 and 15q11.2 showed robust association in this sample (Wu 2026, PMID 42661060).

Shared liability across diagnoses

Genetic boundaries do not reproduce DSM boundaries. Across eight psychiatric disorders, analysis of 232,964 cases and 494,162 controls identified 109 loci associated with at least two disorders; 23 affected four or more, and 11 showed antagonistic effects across disorders (Cross-Disorder Group 2019, PMID 31835028).

Comparison Genetic finding Clinical implication
Bipolar I–schizophrenia Bipolar I correlated strongly with schizophrenia, driven partly by psychosis Psychotic features do not imply a wholly separate genetic disease
Bipolar II–major depression Bipolar II correlated more strongly with MDD than bipolar I did Depressive polarity has a partly shared inherited basis
Bipolar disorder–ADHD SNP genetic correlation 0.64 overall Comorbidity is not only diagnostic confusion
Early-onset bipolar–ADHD SNP genetic correlation 0.71 Early onset may concentrate shared neurodevelopmental liability

The bipolar–ADHD meta-analysis included 4,609 ADHD cases, 9,650 bipolar cases and 21,363 controls; genetic correlations were 0.64 overall and 0.71 for early-onset bipolar disorder (van Hulzen 2017, PMID 27890468). These results complement, but do not replace, longitudinal clinical differentiation.

The 2019 bipolar GWAS found bipolar I more genetically correlated with schizophrenia and bipolar II more correlated with major depression (Stahl 2019, PMID 31043756). Subtype labels therefore capture some genetic structure, but correlation is neither identity nor a patient-level classifier.

From loci to pathways

Risk alleles are enriched in neuronal, synaptic and brain-expressed genes. The 2021 GWAS also found enrichment among targets of antipsychotics, calcium-channel blockers, antiepileptics and anesthetics, and prioritized druggable genes including HTR6, MCHR1, DCLK3 and FURIN (Mullins 2021, PMID 34002096).

Pathway enrichment in the 2019 GWAS included insulin-secretion regulation and endocannabinoid signaling (Stahl 2019, PMID 31043756). Such enrichment is hypothesis-generating: it does not show that an existing drug acting on a pathway will prevent or treat bipolar episodes.

The cross-disorder analysis found pleiotropic loci in genes expressed in brain from the prenatal second trimester onward and enriched for neurodevelopmental functions (Cross-Disorder Group 2019, PMID 31835028). This supports a developmental component without implying that bipolar disorder is fixed or clinically manifest from childhood.

Structural neuroimaging

The ENIGMA Bipolar Disorder Working Group pools standardized data from more than 150 researchers, 20 countries and 55 institutions. Its large-scale results show distributed reductions in cortical thickness, subcortical volume and white-matter integrity rather than one pathognomonic lesion (Ching 2022, PMID 32725849).

Modality/finding Evidence base Result Limitation
Resting-state functional imaging 51 studies; 1,842 bipolar; 2,190 controls Differences in frontal, insular, striatal, temporal, cerebellar and precuneus activity State, medication and analytic heterogeneity
Voxel-based morphometry 83 studies; 2,790 bipolar; 3,690 controls Lower volume in insular-temporal, frontal, cingulate, thalamic and fusiform regions Group-level overlap is extensive
Youth amygdala volume 11 studies SMD −0.74 (95% CI −1.36 to −0.15) Small studies and age dependence
Adult amygdala volume 11-study synthesis SMD 0.20 (95% CI −0.31 to 0.73) No significant case-control difference

The multimodal meta-analysis found partial functional/structural convergence in insula-temporal cortex, frontostriatal-thalamic circuitry and default-mode regions (Chen 2022, PMID 36093787). The spatial distribution is biologically plausible for emotion, salience and cognitive control, but the abstract did not report patient-level sensitivity or specificity.

Youth, but not adult, amygdala volume differed significantly from controls in an earlier meta-analysis (Pfeifer 2008, PMID 18827720). This age interaction cautions against describing a static “bipolar brain.”

Lithium exposure, illness duration, episode burden and cardiometabolic illness can each correlate with brain measures. ENIGMA explicitly treats medication and clinical-risk profiles as modifiers rather than nuisances that can always be removed statistically (Ching 2022, PMID 32725849).

Functional and electrophysiological signals

Auditory P50 sensory gating was altered in 16 bipolar studies comprising 975 patients, and also in first-degree relatives; episode state modified the bipolar signal and medication tended to improve it (Atagun 2020, PMID 32361172). Because similar abnormalities occur in schizophrenia and relatives, P50 is a transdiagnostic research phenotype rather than a diagnostic assay.

Retinal optical coherence tomography has also been explored as an accessible CNS proxy. A meta-analysis of 820 patient eyes and 904 control eyes found peripapillary retinal nerve-fiber-layer thinning across schizophrenia and bipolar samples (overall SMD −0.74), but diagnostic groups were combined for key analyses (Lizano 2020, PMID 31112601). This does not establish bipolar-specific retinal pathology.

Inflammation and metabolic signaling

Peripheral inflammation is among the most replicated but least specific findings. Across 11 studies and 1,618 participants, CRP was higher in bipolar disorder than controls (SMD 0.39, 95% CI 0.24–0.55) (Dargél 2015, PMID 25742201).

Mood state CRP SMD versus controls 95% CI Interpretation
Mania 0.73 0.44 to 1.02 Largest observed signal
Euthymia 0.26 0.01 to 0.51 Small residual association
Bipolar depression 0.28 −0.17 to 0.73 Not statistically significant

The same CRP analysis found no association with lithium or antipsychotic use, but residual confounding by BMI, smoking, infection and physical illness remains possible (Dargél 2015, PMID 25742201).

Not all candidate inflammatory-metabolic markers replicate. In 11 studies with 1,118 participants, leptin did not differ significantly in mania (g −0.99, 95% CI −2.43 to 0.43), depression (g 0.17, −0.45 to 0.79), or euthymia (g 0.03, −0.39 to 0.46) (Fernandes 2016, PMID 27065008).

Erythrocyte DHA was lower in six small case-control studies totaling 118 bipolar I patients and 147 controls, while EPA was only a trend and omega-6 fatty acids did not differ (McNamara 2016, PMID 27087497). The small, observational evidence cannot distinguish dietary exposure, metabolic consequence or causal mechanism.

Sleep and circadian organization

Sleep is both a phenotype and a potential episode trigger, but much evidence is clinically derived rather than molecular. A meta-analysis of 10 studies and 1,824 patients estimated hypersomnia in 29.9% (95% CI 25.8–34.1%; I²=59.2%) (Grigolon 2019, PMID 30611064).

Lifestyle trials provide indirect mechanistic evidence: sleep-focused interventions improved depressive symptoms with SMD −0.80 (95% CI −1.21 to −0.39), although only 18 lifestyle studies entered the systematic review and interventions were heterogeneous (Simjanoski 2023, PMID 37263531).

Chronotype is not bipolar-specific. A five-study meta-analysis found schizophrenia participants more evening-oriented than controls but not different from bipolar I participants, supporting a transdiagnostic rather than diagnostic signal (Linke 2021, PMID 34332427).

Neuroprogression: useful model, unsettled mechanism

The concept of “neuroprogression” proposes that recurrent episodes, inflammation, oxidative stress and medical comorbidity progressively alter function and biology. Observational associations between episode burden, cognition, functioning and biological markers are consistent with this model, but reverse causation and treatment exposure are difficult to exclude.

Insulin resistance may mark a clinically adverse subgroup: impaired glucose metabolism was associated with chronic course (OR 2.96, 95% CI 1.69–5.17), rapid cycling (OR 2.88, 1.59–5.21), and poor mood-stabilizer response (OR 6.74, 1.04–43.54), but the review included only 10 reports and 1,183 people (Miola 2023, PMID 37086806).

Childhood maltreatment also predicts earlier onset, rapid cycling and greater episode burden, demonstrating that worse longitudinal biology cannot be attributed to inherited risk alone (Agnew-Blais 2016, PMID 26873185).

What is not clinically ready

Candidate Replicated group association? Individual clinical test? Main barrier
Common-variant polygenic score Yes No Low variance explained; ancestry portability
Structural MRI pattern Yes No Small distributed effects; medication/state confounding
Functional MRI pattern Partial No Analytic and state heterogeneity
CRP Yes, state-dependent No Nonspecific to bipolar disorder
Leptin No consistent difference No BMI/age confounding and null pooled results
P50 sensory gating Yes, transdiagnostic No Shared with schizophrenia and relatives
Sleep/chronotype Yes, transdiagnostic No Consequence, trigger and trait are difficult to separate

The strongest conclusion is therefore architectural: bipolar disorder arises from distributed, overlapping biological liabilities. Current evidence can inform mechanisms and stratified research, but diagnosis remains longitudinal and clinical (Mullins 2021, PMID 34002096; Ching 2022, PMID 32725849).

Mechanistic convergence and its limits

Bioenergetic evidence is compartment-specific. A review identified 12 lactate studies: five of six brain-spectroscopy studies and both cerebrospinal-fluid studies reported higher lactate, whereas two peripheral studies did not. The five-study brain meta-analysis sat at the significance boundary (Z=1.97, P=.05), with I²=86% and possible publication bias (Kuang 2018, PMID 29726068). This supports a mitochondrial hypothesis in a subset; it does not make peripheral lactate a diagnostic marker.

Intervention data are compatible with that pathway but do not validate it. Thirteen randomized trials of diverse “mitochondrial modulators” yielded a pooled depressive-symptom SMD of −0.48 (95% CI −0.83 to −0.14; I²=75%); the N-acetylcysteine subgroup was −0.88 (−1.48 to −0.27; I²=81%), with small samples and pharmacologically heterogeneous agents (Liang 2022, PMID 35013098). A treatment-class effect assembled after the fact cannot identify the operative mitochondrial target.

Circadian machinery offers another bridge: core clock transcriptional loops regulate sleep/activity timing and also intersect glucose and thyroid physiology, making rhythm disruption a plausible shared mechanism for bipolar disorder and endocrine comorbidity (Yan 2022, PMID 36683994). Plausibility remains ahead of causal mediation evidence.

Neuroprogression is a subgroup hypothesis

A 114-study review found clinical and imaging signals consistent with progression in a subset, particularly with more manic episodes, early trauma and comorbidity, but emphasized variable trajectories and predominantly cross-sectional evidence (Passos 2016, PMID 27097559). Inflammatory models propose episode-related cytokine elevation, reduced neurotrophic support, glial dysfunction and neuroendocrine effects as interacting processes (Muneer 2016, PMID 26766943). The unresolved issue is direction: repeated episodes may alter these systems, but inflammatory and metabolic burden can also precede or amplify episodes.

Open questions

  • Can multi-ancestry polygenic scores and rare CNV burden gain enough calibration to predict onset, polarity or treatment response rather than merely case-control status (Li 2021, PMID 33263727; Wu 2026, PMID 42661060)?
  • Which imaging differences precede illness onset, and which reflect episodes, medication or physical comorbidity (Ching 2022, PMID 32725849)?
  • Are CRP-defined subgroups stable enough for prospective anti-inflammatory-treatment enrichment (Dargél 2015, PMID 25742201)?
  • Does preventing insulin resistance improve mood course, or is impaired glucose metabolism mainly a marker of severe illness (Miola 2023, PMID 37086806)?
  • Can repeated within-person sleep, activity and immune measurements separate episode triggers from consequences (Simjanoski 2023, PMID 37263531)?

References

  1. Mullins N, et al. Genome-wide association study of more than 40,000 bipolar disorder cases provides new insights into the underlying biology. Nature Genetics. 2021. PMID 34002096.
  2. Stahl EA, et al. Genome-wide association study identifies 30 loci associated with bipolar disorder. Nature Genetics. 2019. PMID 31043756.
  3. Li HJ, et al. Novel Risk Loci Associated With Genetic Risk for Bipolar Disorder Among Han Chinese Individuals: A Genome-Wide Association Study and Meta-analysis. JAMA Psychiatry. 2021. PMID 33263727.
  4. Cross-Disorder Group of the Psychiatric Genomics Consortium. Genomic Relationships, Novel Loci, and Pleiotropic Mechanisms across Eight Psychiatric Disorders. Cell. 2019. PMID 31835028.
  5. van Hulzen KJE, et al. Genetic Overlap Between Attention-Deficit/Hyperactivity Disorder and Bipolar Disorder: Evidence From Genome-wide Association Study Meta-analysis. Biological Psychiatry. 2017. PMID 27890468.
  6. Scott LJ, et al. Genome-wide association and meta-analysis of bipolar disorder in individuals of European ancestry. Proceedings of the National Academy of Sciences USA. 2009. PMID 19416921.
  7. Ching CRK, et al. What we learn about bipolar disorder from large-scale neuroimaging: Findings and future directions from the ENIGMA Bipolar Disorder Working Group. Human Brain Mapping. 2022. PMID 32725849.
  8. Chen G, et al. Functional and structural brain differences in bipolar disorder: a multimodal meta-analysis of neuroimaging studies. Psychological Medicine. 2022. PMID 36093787.
  9. Pfeifer JC, et al. Meta-analysis of amygdala volumes in children and adolescents with bipolar disorder. Journal of the American Academy of Child & Adolescent Psychiatry. 2008. PMID 18827720.
  10. Atagun MI, et al. Meta-analysis of auditory P50 sensory gating in schizophrenia and bipolar disorder. Psychiatry Research: Neuroimaging. 2020. PMID 32361172.
  11. Lizano P, et al. A Meta-analysis of Retinal Cytoarchitectural Abnormalities in Schizophrenia and Bipolar Disorder. Schizophrenia Bulletin. 2020. PMID 31112601.
  12. Dargél AA, et al. C-reactive protein alterations in bipolar disorder: a meta-analysis. Journal of Clinical Psychiatry. 2015. PMID 25742201.
  13. Fernandes BS, et al. Leptin in bipolar disorder: A systematic review and meta-analysis. European Psychiatry. 2016. PMID 27065008.
  14. McNamara RK, et al. Meta-analysis of erythrocyte polyunsaturated fatty acid biostatus in bipolar disorder. Bipolar Disorders. 2016. PMID 27087497.
  15. Grigolon RB, et al. Hypersomnia and Bipolar Disorder: A systematic review and meta-analysis of proportion. Journal of Affective Disorders. 2019. PMID 30611064.
  16. Simjanoski M, et al. Lifestyle interventions for bipolar disorders: A systematic review and meta-analysis. Neuroscience & Biobehavioral Reviews. 2023. PMID 37263531.
  17. Linke M, et al. Chronotype in individuals with schizophrenia: A meta-analysis. Schizophrenia Research. 2021. PMID 34332427.
  18. Miola A, et al. Insulin resistance in bipolar disorder: A systematic review of illness course and clinical correlates. Journal of Affective Disorders. 2023. PMID 37086806.
  19. Agnew-Blais J, et al. Childhood maltreatment and unfavourable clinical outcomes in bipolar disorder: a systematic review and meta-analysis. Lancet Psychiatry. 2016. PMID 26873185.
  20. Wu Y, et al. Higher frequencies of CNV deletions in bipolar disorder among Chinese Population. Mol Psychiatry. 2026. PMID 42661060.
  21. Yan X, et al. Circadian rhythm disruptions: a possible link of bipolar disorder and endocrine comorbidities. Front Psychiatry. 2022;13:1065754. PMID 36683994
  22. Kuang H, et al. Lactate in bipolar disorder: a systematic review and meta-analysis. Psychiatry Clin Neurosci. 2018;72:546–555. PMID 29726068
  23. Liang L, et al. Mitochondrial modulators in the treatment of bipolar depression: a systematic review and meta-analysis. Transl Psychiatry. 2022;12:4. PMID 35013098
  24. Passos IC, et al. Areas of controversy in neuroprogression in bipolar disorder. Acta Psychiatr Scand. 2016;134:91–103. PMID 27097559
  25. Muneer A. Bipolar disorder: role of inflammation and the development of disease biomarkers. Psychiatry Investig. 2016;13:18–33. PMID 26766943