com Major Depressive Disorder Vitiligo-hypothesis-p38alpha MAPK14 shared mediator

Key Findings

F001 — The seed molecular link rests on one hypothesis-generating in-silico study

The MAPK14 shared-mediator idea originates entirely from PMID:42418414 (2026), which integrated separate GEO datasets: MDD whole-blood (GSE98793: 128 MDD, 64 control) and vitiligo skin (GSE65127+GSE53146: 15 vitiligo, 15 control). MAPK14 was selected through a chain of correlational/heuristic steps — differential expression at a permissive threshold, STRING PPI network centrality, machine-learning feature selection, and ssGSEA immune-enrichment signatures that are inferred rather than measured cell abundances. Critically, MAPK14 is one of three co-equal machine-learning hits: the paper states, "Machine learning further prioritized three key genes: EXOSC7, KLRG1, and MAPK14." The authors themselves characterize the result cautiously: "Preliminary validation suggests MAPK14 as a potential candidate gene warranting further investigation in MDD and vitiligo."

There is no comorbid MDD–vitiligo cohort, no paired samples from the same participants, no direct p38α activity assay, and no causal perturbation. The comparison also confounds tissue (whole blood vs skin), so any "shared" signal may simply track leukocyte composition. MAPK14 is therefore not uniquely implicated even within its source paper, and none of the inference steps (coexpression, PPI centrality, ML selection, ROC performance, ssGSEA, pathway enrichment) can establish causality or mediation. EXOSC7 and KLRG1 were co-equal hits and were not independently audited — they should be evaluated on equal footing before MAPK14 is privileged.

F002 — Bidirectional epidemiological association is real, but antidepressant use lowered vitiligo risk

The population-based cohort (PMID:30528503, THIN database) establishes temporal association in both directions: MDD→vitiligo adjusted HR 1.64 (95% CI 1.43–1.87, P<.0001; n=405,397 MDD vs 5,739,048 comparators), and vitiligo→MDD HR 1.31 (before age 30) and 1.22 (age ≥30). The abstract states MDD patients "were at a 64% increased risk for vitiligo (hazard ratio 1.64, 95% confidence interval [CI] 1.43-1.87, P < .0001)."

Crucially, the same study reports that vitiligo risk "was decreased in patients using antidepressants." Antidepressants (which in vitro modulate p38, PMID:40840361) reducing vitiligo risk is more consistent with psychosocial/treatment models than with a shared pro-p38 inflammatory program that antidepressants would be expected to leave unchanged or worsen. This is association and temporal ordering of recorded diagnoses — not mediation.

F003 — Human vitiligo p38 and MDD p38α evidence are isoform- and species-discordant

The two arms do not use the same molecular entity in the same way:

• Vitiligo arm (human, pan-p38): PMID:20085492 shows activated p38 as an upstream driver of apoptosis in perilesional keratinocytes from 12 human nonsegmental-vitiligo patients — "our study demonstrates the pivotal role of p38 MAP kinase as an upstream signal of perilesional keratinocyte damage" — but this is pan-p38, not MAPK14-isoform-resolved. Supporting melanocyte models (PMID:17481858, PMID:21514118) show dopamine/oxidative-stress-induced JNK/p38 apoptosis, again pan-p38 and partly mouse (Mel-Ab) cells.
• MDD arm (mouse, p38α-specific): PMID:21835346 is the only MAPK14/p38α-isoform-specific result — "the α isoform of p38 mitogen-activated protein kinase (MAPK) was selectively inactivated" — but selective p38α deletion in mouse serotonergic neurons produces stress resilience via SERT trafficking/hyposerotonergic state. This is rodent depression-like behavior, not human MDD, and it is in neurons, not blood or skin.

Directional genetics is also null: PMID:37975615 found "none of the rigorous bidirectional MR analyses uncovered a significant causal association" between generalized vitiligo and depression. The arms therefore share a gene family label but differ in isoform specificity, species, cell type, upstream trigger, downstream substrate, and directional consequence — the signature of parallel tissue-specific p38 use, not one shared circuit.

F004 — A p38 inhibitor (losmapimod) failed in a controlled human MDD trial

The strongest causal test in the MDD arm is negative. PMID:24699061 evaluated losmapimod (GW856553, an oral p38α/β inhibitor) 7.5 mg BID for 6 weeks in two RCTs in clinically diagnosed MDD enriched for anergia/psychomotor retardation. An underpowered, early-terminated Study 574 (n=24) nominally favored drug (Bech difference −4.10; 95% CI −7.36, −0.83; p=0.017), but the pre-registered Bayesian confirmatory Study 009 (n=128) "showed no advantage for losmapimod" (Bech difference +1.11; 95% credible interval −0.22, 2.50), with no significant cytokine biomarker changes. The authors conclude: "In conclusion 7.5 mg BID losmapimod was not effective in MDD." Because losmapimod inhibits p38α (MAPK14) — albeit not isoform-selectively — this is direct human evidence against a causal, druggable p38 role in MDD.

F005 — Vitiligo's mental-health comorbidity is largely nonspecific, matching atopic dermatitis

If a vitiligo-specific p38 mediator drove MDD, vitiligo's psychiatric burden should exceed that of other inflammatory dermatoses. It does not. PMID:41781039 (German DAK claims, N=2,885,984; 4,631 vitiligo cases, ICD-10 L80, propensity-matched 1:3) reports depressive-episode prevalence of 8.9–19.2% and, in matched comparison with atopic dermatitis, "only a few and inconsistent differences were observed" (more pronounced differences appeared only vs psoriasis). The companion claims analysis PMID:42051022 reports modest, shared risk: depression RR 1.23 (95% CI 1.15–1.32), anxiety RR 1.32 (1.19–1.47), explicitly noting these were "linked to skin diseases like vitiligo, atopic dermatitis, or psoriasis." Modest effect sizes (RR ~1.2–1.3) shared across dermatoses support nonspecific inflammatory/psychosocial burden rather than a vitiligo-specific molecular mechanism. Note these two claims analyses share the DAK data source and likely overlapping cohorts — treat as one data source, not two independent replications.

F006 — No genetic evidence for MAPK14 as a shared pleiotropic locus

Vitiligo susceptibility is immune/oxidative/melanogenic, not MAPK14-centered. The genetics review PMID:39890561 states GWAS "have identified over 50 susceptibility loci, including key genes within the MHC region and those involved in immunity, oxidative stress, and melanogenesis" — MAPK14 is not among the highlighted genes. The newest multi-omics vitiligo mediator is TAPBP via STAT2/interferon–antigen-presentation (PMID:41884389): "Cross-omic analysis identified TAPBP as a top candidate, which is significantly upregulated in lesional melanocytes and core to the antigen-processing network" — again, not MAPK14. Targeted PubMed searches (2026-07-26) for an MDD–vitiligo cross-trait genetic-correlation/colocalization/LDSC study and for a MAPK14 depression GWAS locus returned no papers, and the only genetic disease-to-disease test (bidirectional MR, PMID:37975615) was null. There is currently no locus-level colocalization, shared eQTL/pQTL, or genetic-correlation evidence placing MAPK14 as a shared pleiotropic driver. (Null MR does not, by itself, exclude shared pleiotropy — but no cross-trait analysis supports it either.)

F007 — Psychosocial mediation is well-supported for the vitiligo→MDD arm

The systematic review PMID:34554406 (168 studies, 1979–2021) reports depression in 0.1–62.3% and anxiety in 1.9–67.9% of vitiligo patients, with significantly higher burden associated with "female sex, visible or genital lesions, age < 30 years (particularly adolescents), and greater body surface area involvement, among others," and lesion concealment as the commonest coping strategy. The narrative review PMID:42082425 (2026) concurs that "Disease visibility, particularly involvement of the face or genital regions; greater body surface area involvement; active disease progression; and longer disease duration emerged as key clinical factors associated with worse outcomes." This visibility/age gradient mirrors the age-stratified vitiligo→MDD risk (HR 1.31 if <30y vs 1.22 if ≥30y) in PMID:30528503, consistent with psychosocial rather than molecular mediation.

Mechanistic Model / Interpretation

The evidence supports two independent (parallel) p38 pathways plus a psychosocial/nonspecific-inflammation overlay, not a single shared mediator.

BRAIN / NEURAL COMPARTMENT (MDD arm)
  chronic stress / neuroinflammation
      │
      ▼
  serotonergic neurons ──[p38α / MAPK14, isoform-specific]──► SERT trafficking → hyposerotonergic state
      │  (MOUSE ONLY, deletion → RESILIENCE; PMID:21835346)
      ▼
  depression-like behavior  ── human MDD?  ◄─ p38 inhibitor losmapimod NULL in RCT (PMID:24699061)

SKIN COMPARTMENT (vitiligo arm)
  oxidative stress (H2O2 / dopamine)
      │
      ▼
  melanocytes + perilesional keratinocytes ──[pan-p38, NOT isoform-resolved]──► p53/NF-κB apoptosis
      │  (HUMAN NSV keratinocytes + mouse/human melanocytes; PMID:20085492, 17481858, 21514118)
      ▼
  melanocyte loss → depigmentation

BRIDGE EDGE (proposed shared mediator)
  ??? shared upstream state / circulating mediator / homologous cell program ???
      │
      ▼
  ✗ MISSING: no comorbid cohort, no paired tissue, no colocalization,
    null directional MR, isoform/species discordance

PSYCHOSOCIAL / NONSPECIFIC OVERLAY (best-supported real link)
  visible/extensive vitiligo, younger age, stigma → distress → MDD
  (PMID:34554406, 42082425, 30528503) — shared across dermatoses (PMID:41781039, 42051022)

Causal-chain grading:

Cell-state map — is any state truly shared? The neural substrate (SERT trafficking in serotonergic neurons) and the skin substrate (p53/NF-κB apoptosis in melanocytes/keratinocytes) are analogous (both engage p38 family kinases under stress) but not shared: different upstream triggers (psychosocial stress vs oxidative/dopamine), different cells, different substrates, and different isoform resolution. The circulating-immune "bridge" candidate (whole-blood MAPK14 transcript, PMID:42418414) is speculative and confounded by leukocyte composition. No truly shared MAPK14-dependent state is demonstrated.

Evidence Base

Rodent neuroinflammation p38 papers (PMID:41707798, PMID:41786069, PMID:42134655) reinforce a neural p38 route in mouse depression-like behavior but are pan-p38/pan-MAPK and cannot be upgraded to human MDD or a shared mediator.

PMID:42418414 Replication / Method Audit

No independent GEO re-analysis was performed (datasets not supplied). This is a methodological audit of the seed design as reported.

Reproduces independently: MDD–vitiligo comorbidity exists (PMID:30528503); p38 is stress-activated in immune cells and skin (non-specific). Does NOT reproduce: MAPK14 as a specific shared driver over EXOSC7/KLRG1 or generic inflammatory markers; any shared cell-state; MAPK14 direction/magnitude/adjusted significance; any causal or mediation claim. GSE52790 and GSE80009 were not examined (not provided) — flagged as open replication tasks.

Model-Comparison Table

Limitations and Knowledge Gaps

1. No comorbid human cohort exists with clinically adjudicated MDD and subtype/activity-defined vitiligo carrying MAPK14 RNA, p38α protein, phospho-p38 activity, or downstream substrate measurements. This is the central missing bridge study; its absence is documented from targeted PubMed searches, not a comprehensive multi-database systematic review.
2. Isoform specificity is unresolved in vitiligo: all skin evidence is pan-p38, so MAPK14 (p38α) vs MAPK11/12/13 contributions are unknown. No isoform-selective genetic perturbation or kinase-dead rescue exists for either arm's human context.
3. Transcript ≠ protein ≠ activity: the seed relies on bulk transcript abundance; p38α protein, phosphorylation, and substrate activity were never measured across compartments.
4. The MDD isoform-specific evidence is mouse-only and points to resilience-on-deletion, complicating a straightforward "p38α drives MDD" narrative.
5. Confounders — cell mixture, medications, disease activity, batch, healthcare contact, smoking, BMI, socioeconomic factors — are largely uncontrolled in the seed and unaddressed for the molecular hypothesis.
6. Null MR does not exclude shared pleiotropy — no cross-trait LDSC/colocalization has been performed, so the pleiotropy model is untested rather than refuted (though MAPK14-specific pleiotropy is unsupported by known vitiligo genetics).
7. Search scope: single database (PubMed); Embase, GWAS Catalog, Open Targets, ClinicalTrials.gov, and preprint servers were not queried. "No evidence found" applies only to explicitly searched scopes.

Proposed Follow-up Experiments / Discriminating Studies

Ranked by efficiency in separating the competing models:

1. Cross-trait LDSC + MAPK14-locus colocalization (highest value, lowest cost). Vitiligo GWAS × MDD GWAS global/local genetic correlation, locus colocalization, and shared eQTL/pQTL at MAPK14, distinguishing colocalization from linkage and sample overlap. Summary statistics already exist. Expected under shared pleiotropy: colocalization at MAPK14; under parallel/nonspecific: none.
2. Comorbid paired-tissue profiling. Treatment-naive MDD-only, active-NSV-only, comorbid, matched disease-free, and a visible-inflammatory-skin comparator (atopic dermatitis). Single-cell + surface-protein/phospho-p38α (CyTOF) in blood and paired lesional/nonlesional skin. Under shared mediator: one phospho-p38α+ state elevated in both arms, maximal in comorbid subjects; under parallel/nonspecific: compartment-specific signatures indistinguishable from atopic dermatitis.
3. Isoform-resolved perturbation with rescue. Cell-restricted MAPK14 CRISPR knockout/interference in human iPSC-derived serotonergic neurons/glia and in melanocytes/keratinocytes/CD8 T cells, with selective p38α pharmacology plus CRISPR-resistant wild-type vs kinase-dead MAPK14 rescue and dose-response. Establishes isoform specificity, causal necessity, and on-target vs off-target effects per compartment.
4. Longitudinal mediation with equalized surveillance. Prospective repeated phenotyping measuring MAPK14/phospho-p38 and cytokines before onset of the second condition, with mediation adjusting for visibility/severity, stigma, stress, socioeconomic factors, smoking, BMI, autoimmune comorbidity, medications, and healthcare utilization; include negative-control outcomes and active comparators to address surveillance/treatment bias.
5. Vitiligo activity RCT of a selective p38α inhibitor with a phospho-p38 target-engagement biomarker, mirroring the informative null MDD RCT to test the vitiligo-arm causal edge.

Prespecified decision criteria: Shared mediation requires a demonstrated shared MAPK14+ cell-state (study 2) and MAPK14-locus colocalization (study 1) and isoform-specific necessity in both arms (study 3). Absence of a shared state but arm-specific necessity → parallel p38. Vitiligo≈AD signature → nonspecific. Effect abolished by visibility/stigma adjustment → psychosocial. Attenuation by equalized surveillance → surveillance bias. Null coloc + null MR → reject shared-pleiotropy/directional-genetic.

Curation Leads (require independent DisMech verification)

• Disposition: Keep mdd_vitiligo_acquired_mapk14_shared_mediator as a KNOWLEDGE_GAP discussion, optionally narrowed to parallel tissue-specific p38 models. Do not promote to EMERGING mechanistic hypothesis; consider deprecating the "single shared mediator" framing.
• Node/biomarker/edge/module: Evidence does NOT justify a MAPK14 pathophysiology node, biomarker, causal edge, or conserved cross-disease module for this pair. At most, MAPK14 is a hypothesis-level candidate marker in one in-silico study.
• Seed lead: PMID:42418414 — "Machine learning further prioritized three key genes: EXOSC7, KLRG1, and MAPK14."
• MDD refutation lead: PMID:24699061 — "In conclusion 7.5 mg BID losmapimod was not effective in MDD."
• Vitiligo refinement: p38 evidence is pan-p38, ex-vivo/in-vitro (PMID:20085492); genetics point to HLA/immune/oxidative/TAPBP (PMID:39890561, PMID:41884389), not MAPK14.
• Alternative candidates: EXOSC7 and KLRG1 were co-equal ML hits and were not independently audited — evaluate on equal footing before privileging MAPK14.
• Overlap flag: PMID:41781039 and PMID:42051022 share the DAK claims source — treat as one data source.

Supported vs Refuted Hypotheses (summary)

• Supported: MDD–vitiligo comorbidity is real, bidirectional, and temporally ordered (PMID:30528503). p38 is genuinely activated in vitiligo skin (pan-p38, PMID:20085492) and p38α is mechanistically active in mouse serotonergic stress circuitry (PMID:21835346). Psychosocial mediation of vitiligo→MDD is well-supported (PMID:34554406, 42082425).
• Refuted / contradicted (as stated): (i) p38 inhibition treats human MDD — refuted (PMID:24699061). (ii) Vitiligo has a specific MDD-linking signature beyond other skin diseases — refuted (PMID:41781039). (iii) Genetic directional causation vitiligo↔depression — null (PMID:37975615). (iv) MAPK14 as an established shared pleiotropic locus — unsupported (PMID:39890561, 41884389).
• Unresolved: whether a narrow, cell-state-restricted MAPK14 program contributes to each disease independently (parallel model) remains open and testable.

Citation Manifest

All identifiers, quotes, and statistics are drawn from the abstracts/snippets recorded during the investigation and remain leads for independent DisMech curator validation.

Artifacts

• OpenScientist final report
• OpenScientist final report