🧬 Aging Wiki

gpnmb

Properties1
aliasesGPNMB, transmembrane glycoprotein NMB, osteoactivin, DC-HIL, HGFIN, glycoprotein nmb

15 min read

Partially verified 2026-05-29 (claude). Verified: UniProt Q14956 identity fields (REST); Rose 2010 (ADAM10/angiogenic shedding, abstract); Pahl 2010 (ESRD fold-changes 8.6×/17.5×, n=21+22, abstract); Saade 2021 (review, local PDF); Chen 2024 Gut (local PDF); METRIC trial n/PFS (Crossref abstract); Tyshkovskiy 2026 aging/mortality claims (pre-verified study page). Not PDF-verified (closed-access or archive-pending): Suda 2021, Wang 2025, Colombo 2019, Chung 2014, Chung 2011, Liang 2026, Kim 2026, Liu 2024, Kilavuz 2025, Yardley 2015, Zhang 2013, Yu 2020, van der Vliet 2026. Quantitative claims from these sources should be treated as provisional. mr-causal-evidence corrected to partial (two conflicting PD-specific MR studies now exist — see Gaps section).

GPNMB (transmembrane glycoprotein NMB)

GPNMB (also osteoactivin, DC-HIL, HGFIN) is a type I transmembrane glycoprotein whose ectodomain is proteolytically shed as a soluble fragment by ADAM10. It is highly expressed by macrophages, microglia, and other myeloid cells, melanocytes, osteoblasts, and many tumours, and is strongly induced under lysosomal/phagocytic stress and tissue damage, where it has broadly immunosuppressive and tissue-repair-associated functions. In aging biology it has emerged as one of the most consistent conserved transcriptomic and proteomic markers of mammalian ageing, mortality, and chronic disease — and uniquely, as a validated antigen displayed on the surface of senescent endothelial cells.

Identity

FieldValue
UniProtQ14956 (GPNMB_HUMAN) — Swiss-Prot (manually curated)
NCBI Gene10457
HGNC symbolGPNMB
EnsemblENSG00000136235 (versioned:.15 per GTEx v8/v10)
Mouse orthologGpnmb (one-to-one)
GenAgeNot in GenAge as of 2026-05-29 needs-genage-check

Molecular biology

GPNMB is a type I single-pass transmembrane glycoprotein (UniProt Q14956) belonging to the PMEL/NMB family. The canonical isoform is 572 amino acids; a shorter 560 amino acid isoform is generated by alternative splicing.1 Its domain architecture comprises:

  • Signal peptide (residues 1–22) — cleaved co-translationally to yield the mature protein
  • Extracellular domain / ectodomain (residues 23–498) — contains a PKD (polycystic kidney disease) domain (residues 240–327) implicated in protein–protein interactions and ligand binding (heparin and syndecan binding confirmed)
  • Single transmembrane helix (residues 499–519)
  • Cytoplasmic tail (residues 520–572) — contains phosphoserine at position 542

The protein carries 12 N-linked glycosylation sites and is extensively glycosylated, contributing to the apparent molecular weight exceeding the predicted ~65 kDa.2

Ectodomain shedding. The membrane-anchored form is cleaved at the ectodomain by the metalloprotease ADAM10 (a disintegrin and metalloprotease domain-containing protein 10), releasing a soluble extracellular domain fragment. The shed fragment retains angiogenic activity in vitro (promotes endothelial cell migration) and acts as a paracrine factor in the tumour microenvironment.3 Shedding is enhanced by cellular stress, inflammation, and receptor activation signals — providing a plausible mechanism by which GPNMB elevations appear in plasma under ageing and disease.

Subcellular localisation. GPNMB localises to the cell membrane, melanosome membrane (stages I–IV), and early endosome membrane — consistent with its induction under lysosomal lipid-handling and phagocytic stress.2

Family context. GPNMB is paralogous to PMEL (premelanosome protein) and shares the PKD domain architecture. Both proteins localise to melanosomes; PMEL is the primary structural scaffold, while GPNMB appears more broadly expressed across myeloid and stromal cells.

Cell-type expression and function

GPNMB displays a strikingly cell-type-specific induction pattern despite its broad tissue expression in bulk RNA-seq. At the cell-type level, high expression is a hallmark of:

Macrophages and myeloid cells. GPNMB / HGFIN is strongly upregulated when monocytes differentiate into tissue macrophages, with expression rising further under pathological stimuli.4 In end-stage renal disease, circulating monocytes show 8.6-fold GPNMB upregulation and macrophage-differentiated cells show 17.5-fold upregulation vs healthy controls, indicating GPNMB as a marker of myeloid activation and phagocytic load.4 Saade et al. 2021 reviewed consistent evidence that GPNMB is highly expressed in both macrophages and microglia and is induced by pro-inflammatory stimuli and pathological states including neurodegeneration.1

Lipid-associated macrophages (LAMs). GPNMB marks the lipid-associated macrophage (LAM) / lipid-laden foam-cell phenotype in multiple disease contexts. Wang et al. 2025 showed that myeloid-specific Gpnmb-knockout in mice prevents formation of hepatic LAMs and significantly reduces steatosis and fibrosis in metabolic dysfunction-associated steatotic liver disease (MASLD/NASH).5 This establishes GPNMB not merely as a LAM marker but as a functional driver of monocyte-to-LAM differentiation in lipid-disease settings. GPNMB+ foamy microglia in MS lesions have been independently identified alongside disrupted oxylipin metabolism.6

Disease-associated microglia (DAM). GPNMB is a defining marker of the disease-associated microglia and lipid-associated microglia states identified in Alzheimer’s disease, Parkinson’s disease, and other neurodegeneration contexts. In PS19 (tauopathy) mice, microglial GPNMB mediates transfer of functional mitochondria to astrocytes via extracellular vesicles; Gpnmb ablation worsens cognitive impairment.7 In Parkinson’s substantia nigra, GPNMB is elevated across Braak stages and proposed as a biomarker of disease severity.8

Osteoblasts / osteoclasts (“osteoactivin”). The alias “osteoactivin” reflects GPNMB’s role in bone biology: it is expressed in osteoblasts and promotes osteoblast differentiation and bone mineralisation. Human monocyte-derived cells induced toward osteogenic phenotypes (“monoosteophils”) express high GPNMB/osteoactivin alongside osteogenic markers.9 In aged rodents, reduced osteoactivin expression in bone tissue correlates with impaired osteoblastogenesis and bone loss.10

Dendritic cells (DC-HIL). The alias “DC-HIL” (dendritic cell–associated heparan sulphate proteoglycan-dependent integrin ligand) reflects GPNMB’s expression on plasmacytoid and conventional dendritic cells, where it functions as an immune checkpoint ligand.

Senescent endothelial cells. GPNMB is a seno-antigen — a surface antigen enriched on senescent vascular endothelial cells. Suda et al. 2021 identified GPNMB by transcriptome analysis of senescent endothelial cells and validated it as an immunotherapy target: a senolytic vaccine directed against GPNMB+ cells cleared senescent cells, improved metabolic phenotypes in HFD-fed mice, and extended lifespan in progeroid mice.11 This is the most direct evidence linking GPNMB surface expression to the senescent cell burden.

DimensionStatus
Pathway conserved in humans?yes — macrophage/myeloid induction conserved; LAM phenotype confirmed in human MASLD
Phenotype conserved in humans?yes — UK Biobank plasma GPNMB associates with mortality and multimorbidity (see below)
Replicated in humans?yes — multiple biomarker studies; plasma association robust at n=50,117

Immunosuppressive function (DC-HIL / syndecan-4 axis)

GPNMB functions as an immune checkpoint ligand via its interaction with syndecan-4 (SDC4), a heparan sulphate proteoglycan expressed on activated T cells.

Mechanistically, GPNMB (as DC-HIL) on dendritic cells and myeloid cells engages syndecan-4 on activated T cells, transducing inhibitory signals that suppress T-cell proliferation and effector function. Ligation of DC-HIL induces phosphorylation of its intracellular immunoreceptor tyrosine-based activation motif (ITAM), triggering IFN-γ and iNOS production that suppresses T-cell activation.12 GPNMB shed into the tumour microenvironment as extracellular vesicles also engages syndecan-4 on CD8+ T cells, suppressing their activation.13 In Sézary syndrome, malignant T cells overexpress syndecan-4 bearing specialised heparan sulphate modifications that enhance DC-HIL binding and suppress cytotoxic responses.14

In the myeloid-derived suppressor cell (Mo-MDSC) context, GPNMB expression marks functionally suppressive MDSCs; blocking DC-HIL/syndecan-4 is proposed as a combination checkpoint target alongside PD-1/PD-L1.15

This immunosuppressive axis is relevant to aging biology: GPNMB induction in tissue macrophages and DAM under chronic disease/age-related stress could contribute to local T-cell suppression, potentially impairing immunosurveillance of senescent and cancerous cells — a candidate effector mechanism for the GPNMB → chronic-inflammation → cellular-senescence link.

Why GPNMB matters for aging

In the multi-species transcriptomic-clock analysis of Tyshkovskiy et al. 2026, Gpnmb was among the top genes positively associated with chronological age, expected mortality, and (negatively) with maximum lifespan, conserved across rodents and primates 16. It is one of the three headline universal mortality genes alongside CDKN1A and LGALS3:

  • Cross-species conservation — up-regulated with ageing in both rodents and primates; a conserved cross-species ageing/mortality biomarker.16
  • Chronic disease — Gpnmb was significantly up-regulated in at least 5 of 9 rodent chronic-disease models (Alzheimer’s/5xFAD, CKD, ischaemic stroke, NASH, diabetic nephropathy) and contributed strongly to disease-associated mortality-tAge acceleration.17
  • UK Biobank plasma validation — circulating GPNMB protein concentration (n=50,117) was positively associated with all-cause mortality and a broad multimorbidity spectrum (cardiac arrest, heart failure, liver cirrhosis, diabetes, kidney failure, depression, atherosclerosis, sleep disorders) after adjustment for age and sex.18

This positions GPNMB not merely as a correlate of cellular-senescence and chronic-inflammation but as a circulating, human-validated mortality and multimorbidity biomarker — see transcriptomic-clock-tage.

Senolytic target

Suda et al. 2021 provides a mechanistic entry point: if GPNMB surface expression marks senescent endothelial cells, then the rising plasma GPNMB signal with age may reflect, at least in part, the growing senescent cell burden.11 The senolytic vaccine approach (immunising against GPNMB+ cells) reduced senescence markers in adipose tissue and extended lifespan in progeroid mice — establishing GPNMB as a tractable senolytic antigen, though not yet validated in normally-ageing mice or humans. needs-human-replication

Independent biomarker support

Liu et al. 2024 confirmed positive age-correlation of serum and urine GPNMB in a cross-sectional human cohort (n=473, age 25–91), with serum GPNMB negatively associated with skeletal muscle mass index and IGF-1 — consistent with a SASP/inflammatory tone hypothesis.19

Lysosomal stress and neurodegeneration

GPNMB’s induction under lysosomal lipid-handling stress makes it a sensitive readout of GBA/glucocerebrosidase pathway dysfunction. In Gaucher disease (types 1 and 3), plasma GPNMB is substantially elevated and remains elevated even after prolonged enzyme replacement therapy, proposing GPNMB as a marker of residual disease activity.20 However, GPNMB did not differentiate between neuronopathic Gaucher subtypes or idiopathic Parkinson’s disease.20

GBA heterozygous loss-of-function mutations are the most common genetic risk factor for Parkinson’s disease. The GPNMB elevation in GBA-related diseases, combined with its marking of DAM states in Parkinson’s substantia nigra, suggests GPNMB may report on a shared lysosomal-stress/neuroinflammatory axis across GBA → lysosomal dysfunction → microglial activation → GPNMB induction.

Druggability (aging-context tier 2)

GPNMB reaches druggability tier 2 (high-quality clinical-stage probe in oncology, not yet aging-validated) on the aging-context convention:

Glembatumumab vedotin (CDX-011) is an antibody–drug conjugate (ADC) consisting of an anti-GPNMB monoclonal antibody conjugated to monomethyl auristatin E (MMAE), a cytotoxic payload. It was developed as a GPNMB-targeted cancer therapy on the premise that GPNMB overexpression in tumours would direct payload delivery.

  • EMERGE phase 2 (JCO 2015) — 124 patients; primary endpoint not met in the overall breast cancer population, but suggested activity in the GPNMB-overexpressing TNBC subset.21
  • METRIC phase 2b (NPJ Breast Cancer 2021) — 327 patients (213 GV : 92 capecitabine, 2:1 randomisation) in gpNMB-overexpressing, previously-treated TNBC. Primary endpoint not met: median PFS 2.9 months (GV) vs 2.8 months (capecitabine). Development of glembatumumab vedotin as a TNBC therapy was discontinued after METRIC.22

Why tier 2, not tier 1. No FDA-approved drug targets GPNMB. No clinically-validated probe exists for any aging indication. Glembatumumab vedotin demonstrated clinical exposure and pharmacodynamic target engagement, qualifying the target for tier 2, but the failure of the METRIC trial and absence of aging-indication development means this remains oncology-only clinical history. An aging-validated probe (e.g., a senolytic antibody or ADC targeting GPNMB+ senescent cells) does not yet exist. no-aging-validated-probe

Senolytic immunotherapy angle. The Suda 2021 senolytic vaccine approach represents a conceptually different druggability pathway — mobilising the immune system against GPNMB+ senescent cells rather than directly killing GPNMB-expressing cells with a cytotoxic payload. This has not advanced to human clinical trials as of 2026-05-29. clinical-trials-active: 0 (no trials explicitly targeting GPNMB for aging/senolytic indication found on ClinicalTrials.gov v2 API search 2026-05-29). needs-clinical-trials-recheck

R25 recency search results (protein, 18-month cadence)

Search performed 2026-05-29 via PubMed eutils. Query terms: GPNMB + (aging/ageing/senescence/mortality/biomarker), date window 2024-01-01 to 2026-05-29 (122 results); GPNMB + (meta-analysis/RCT/systematic review), 2021-2026 (2 RCT-context results). High-priority hits triaged:

  • Must-include: Suda 2021 (Nature Aging, n=mice, senolytic vaccination — GPNMB seno-antigen), Liu 2024 (Heliyon, n=473 cross-sectional, age-biomarker), Wang 2025 (Int Immunopharmacol, MASLD/LAM mechanism), Kilavuz 2025 (Orphanet JRD, GBA/Gaucher context), Vahdat 2021 METRIC RCT (n=327, drug development failed)
  • Two RCT-filter hits (PMID 41850233 = tofersen/ALS/CSF-proteomics; PMID 40435316 = semorinemab/AD) — both Phase 2/3 RCTs where GPNMB emerged as a CSF pharmacodynamic marker in neurodegeneration. These are consistent with the neuroinflammation picture but do not test GPNMB as an intervention target; integrated as supporting context.
  • MR studies found (new): Li et al. 2025 (PMID 40214956, Mol Neurobiol) — null MR result for GPNMB and PD risk/progression. Guo et al. 2026 (PMID 41851136, NPJ Parkinson’s Dis) — MR/SMR/colocalisation supports causal role in PD bone-brain axis. Results are PD-specific and conflicting; no MR study for aging/mortality outcome. mr-causal-evidence updated from not-tested to partial.
  • No contradictions to training-era citations found. METRIC 2021 confirms the negative trial outcome that was already noted in training knowledge.

Gaps

  • Whether elevated GPNMB is causal for mortality or a damage/clearance readout marker is unresolved; UK Biobank association is observational, not Mendelian-randomized. mr-causal-evidence: partial — two PD-specific MR studies exist with conflicting results (Li 2025 null; Guo 2026 positive for PD bone-brain axis); no MR for aging/mortality outcome. contradictory-evidence no-mechanism
  • gtex-aging-correlation (age-stratified Spearman ρ) not yet populated from GTEx v10 age-stratified API. needs-gtex-age-stratified
  • GPNMB is not in GenAge as of 2026-05-29. needs-genage-check
  • Senolytic vaccine approach (Suda 2021) demonstrated in progeroid mice only; not replicated in normally-ageing mice or humans. needs-human-replication
  • GBA/GPNMB axis: whether GPNMB elevation in GBA-variant Parkinson’s risk is mechanistic or incidental is unresolved. no-mechanism

Footnotes

Footnotes

  1. doi:10.3389/fimmu.2021.674739 · Saade M et al. · Front Immunol 2021 · review · macrophage/microglia GPNMB expression and anti-/pro-inflammatory roles across disease states. PDF verified 2026-05-29 (local archive). Confirms 572 aa canonical / 560 aa alternative-splice isoform; ADAM10 as sheddase; GPNMB expression in macrophages, microglia, and DAM; broadly anti-inflammatory role with pro-inflammatory exceptions. ↩ ↩2

  2. UniProt Q14956 (GPNMB_HUMAN), Swiss-Prot manually curated entry, accessed 2026-05-29 · protein function, domain annotation, subcellular localisation, PTM sites, glycosylation. Background claim — pending verifier PDF cross-check. ↩ ↩2

  3. doi:10.1371/journal.pone.0012093 · Rose AA et al. · PLoS One 2010 · in-vitro + in-vivo (breast cancer model) · n=cell lines + tumour samples · ADAM10 identified as primary GPNMB sheddase; shed ECD promotes endothelial migration in vitro. Background claim — archive pending download. ↩

  4. doi:10.2215/CJN.03390509 · Pahl MV et al. · Clin J Am Soc Nephrol 2010 · observational · n=21 hemodialysis + 22 controls · GPNMB 8.6× up in monocytes, 17.5× in macrophages in end-stage renal disease vs healthy controls. Background claim — archive pending download. ↩ ↩2

  5. doi:10.1016/j.intimp.2025.114554 · Wang J et al. · Int Immunopharmacol 2025 · in-vivo (mouse, myeloid-specific Gpnmb-KO) · Gpnmb-KO prevents hepatic LAM formation, reduces steatosis and fibrosis in MASLD. Background claim — archive pending. ↩

  6. doi:10.1038/s41593-026-02302-3 · van der Vliet et al. · Nat Neurosci 2026 · PMID 42168651 · in-vivo (MS lesion human tissue + mouse) · GPNMB+ foamy microglia in MS lesions with disrupted lipid metabolism and oxylipin accumulation. Background claim — DOI confirmed via PubMed ESummary; archive not checked. ↩

  7. doi:10.1038/s41593-026-02317-w · Liang C et al. · Nat Neurosci 2026 · in-vivo (PS19 tauopathy mouse) · microglial GPNMB mediates mitochondria transfer to astrocytes via extracellular vesicles; Gpnmb ablation exacerbates cognitive impairment. Background claim — archive not checked. ↩

  8. doi:10.5607/en25051 · Kim et al. · Exp Neurobiol 2026 · PMID 42130054 · multi-omics, human post-mortem substantia nigra · GPNMB elevated across Braak PD stages; proposed biomarker of disease severity. Background claim — DOI confirmed via PubMed ESummary; archive not checked. ↩

  9. doi:10.1371/journal.pone.0067649 · Zhang Z & Shively JE · PLoS One 2013 · in-vivo (NOD/SCID mouse femoral defect) · human monoosteophils expressing high GPNMB/osteoactivin achieve bone repair in 14 days; GPNMB correlates with osteogenic differentiation phenotype. Background claim — archive pending. ↩

  10. doi:10.1016/j.biopha.2020.110253 · Yu et al. · Biomed Pharmacother 2020 · in-vivo (aged rat + d-galactose model) · PMID 32497865 · cycloastragenol increases osteoactivin/GPNMB in bone, alleviates age-related bone loss via osteoblast enhancement. Background claim — archive pending. ↩

  11. doi:10.1038/s43587-021-00151-2 · Suda M et al. · Nat Aging 2021 · in-vivo (mouse, progeroid model + HFD) · GPNMB identified as seno-antigen on senescent endothelial cells; senolytic vaccine cleared GPNMB+ cells, improved metabolic phenotypes, extended lifespan in progeroid mice. archive: not_oa — pending verifier download. ↩ ↩2

  12. doi:10.1038/jid.2014.254 · Chung JS et al. · J Invest Dermatol 2014 · in-vivo (mouse melanoma) · DC-HIL on CD11b+Gr1+ myelomonocytic cells is primary T-cell suppressor; ITAM phosphorylation mediates IFN-γ/iNOS induction; DC-HIL knockout or antibody blockade markedly reduced melanoma growth. Background claim — archive pending download. ↩

  13. doi:10.1136/gutjnl-2024-331903 · Chen A et al. · Gut 2024 · in-vitro/in-vivo (HCC) · n=95 HCC patient samples (TCGA + in-house cohorts); n=6/group in vivo · GPNMB in small EVs binds syndecan-4 (SDC4) on CD8+ T cells, suppressing activation; FTO stabilises GPNMB via m6A demethylation; blocking FTO/GPNMB/SDC4 axis with FTO inhibitor CS2 enhances anti-PD-1 and sorafenib efficacy. PDF verified 2026-05-29 (local archive). ↩

  14. doi:10.1182/blood-2010-08-302034 · Chung JS et al. · Blood 2011 · in-vitro · PMID 21252093 · Sézary syndrome T cells overexpress syndecan-4 with enhanced heparan sulfate modifications enabling DC-HIL binding and surface TGF-β trapping. Background claim — DOI confirmed via PubMed ESummary; archive not checked. ↩

  15. doi:10.1158/1078-0432.CCR-18-2334 · Colombo MP et al. · Clin Cancer Res 2019 · commentary · GPNMB/DC-HIL on Mo-MDSCs as targetable immune checkpoint; DC-HIL/syndecan-4 interaction; proposed combination with PD-1/PD-L1 blockade. Background claim — archive not_oa. ↩

  16. tyshkovskiy-2026-universal-transcriptomic-hallmarks · n=11,165 transcriptomes, 4 species · meta-analysis · elastic-net clock coefficients + mixed-effects gene-trait associations · model: mouse/rat/macaque/human, multi-tissue ↩ ↩2

  17. tyshkovskiy-2026-universal-transcriptomic-hallmarks · 9 rodent disease datasets · ANOVA, P_adj<0.05 · model: mouse/rat disease models (5xFAD, CKD, stroke, NASH, diabetic nephropathy) ↩

  18. tyshkovskiy-2026-universal-transcriptomic-hallmarks · UK Biobank Olink plasma proteomics, n=50,117 · Cox proportional-hazards, adjusted age+sex, P_adj<0.05 · model: human ↩

  19. doi:10.1016/j.heliyon.2024.e36574 · Liu YY et al. · Heliyon 2024 · cross-sectional · n=473 (age 25–91) · serum and urine GPNMB positively correlated with age; serum GPNMB negatively associated with skeletal muscle mass index and IGF-1. Background claim — archive pending. ↩

  20. doi:10.1186/s13023-025-04054-y · Kilavuz S et al. · Orphanet J Rare Dis 2025 · cross-sectional · n=Gaucher disease types 1&3 + PD + controls · plasma GPNMB elevated in GD, remains elevated after ERT; does not differentiate neuronopathic GD subtypes or idiopathic PD. Background claim — archive pending. ↩ ↩2

  21. doi:10.1200/JCO.2014.56.2959 · Yardley DA et al. · J Clin Oncol 2015 · randomized phase 2 (EMERGE) · n=124 · primary endpoint not met in overall population; activity signal in GPNMB-overexpressing TNBC. Background claim — archive pending. ↩

  22. doi:10.1038/s41523-021-00244-6 · Vahdat LT et al. · NPJ Breast Cancer 2021 · randomized phase 2b (METRIC) · n=327 (213 GV : 92 capecitabine) · primary endpoint not met: PFS 2.9 vs 2.8 months; glembatumumab vedotin development in TNBC discontinued. Background claim — archive pending. ↩

Graph View

Backlinks