Shin 2025 — Propagation of senescent phenotypes by extracellular HMGB1 is dependent on its redox state
A primary-research paper establishing that only the reduced redox form of extracellular HMGB1 (ReHMGB1) — not its terminally oxidized form — propagates senescence to bystander cells and across tissues, acting through RAGE-mediated JAK/STAT and PI3K-AKT/NF-κB signaling. Published April 2025 in Metabolism (CC BY-NC-ND open access) by the Jeon Ok Hee lab (Korea University College of Medicine) in collaboration with Jeon-Soo Shin (Yonsei), Michael J. + Irina M. Conboy (UC Berkeley, heterochronic-parabiosis lineage), and Christopher D. Wiley (Tufts Jean Mayer USDA Human Nutrition Research Center on Aging, MiDAS originator). Positions extracellular ReHMGB1 as a redox-state-dependent pro-geronic SASP factor with translational implications for senomorphic intervention; supplementary data demonstrate ReHMGB1 is elevated in human aged (70–80 yr) serum, suggesting clinical relevance.
Question
Cellular senescence spreads systemically through blood circulation, but the molecular identity of the propagating signal has remained incompletely defined. hmgb1 is a known SASP factor that exists in three biologically distinct redox states (fully-reduced ReHMGB1, disulfide DsHMGB1, and terminally oxidized OxHMGB1). The study asks: is the senescence-spreading activity of extracellular HMGB1 redox-state-dependent, and if so, can blocking it attenuate aging-related phenotypes in vivo?
Design
Cell lines + culture
- WI-38 human fetal lung fibroblasts (ATCC CCL-75) — primary model
- BJ human foreskin fibroblasts (ATCC CRL-2522)
- Primary human renal epithelial cells (ATCC PCS-400-011)
- HSKM human skeletal muscle cells (Lonza)
All cultures kept at 3 % O₂ (physiologic).
Mice
- 3-month-old (young), 15-month-old (middle-aged), and 24-month-old (old) male C57BL/6J mice (Jackson Labs Bar Harbor for 3 + 15 mo; Korea Basic Science Institute Ageing Research Facility for 24 mo).
- Approved protocols KOREA-2020-0168 and KOREA-2023-0171.
HMGB1 reagents
- ReHMGB1: N-terminus six-His-tagged recombinant human HMGB1 produced in E. coli SoluBL21 (AMSBIO AMS.C700200); purified via Ni²⁺-NTA + heparin-Sepharose; endotoxin removed by Triton X-114 phase separation; LPS < 1.0 EU/μg.
- OxHMGB1: terminally oxidized HMGB1 (catalog HM060, HMGBiotech, Milan, Italy).
- In vitro dose: 20 μg/mL HMGB1, 72 h treatment, analysis at day 4.
Paracrine senescence model (Fig 1)
WI-38 fibroblasts senesced by 20 Gy X-ray irradiation; conditioned media (CM) collected after 10 days in 40 % FBS DMEM, then diluted to 10 % FBS for treatment. Non-senescent WI-38 recipient cells treated with IR-CM either alone, with control IgG-CM, or with anti-HMGB1 antibody-treated CM (aHM-CM) for 7 days. Anti-HMGB1 mAb concentration: 2.5 μg/mL.
In vivo systemic ReHMGB1 (Fig 7)
3-month-old male C57BL/6J mice received a single IV dose of ReHMGB1 at 5 mg/kg (or PBS vehicle); analyzed at 7 days post-treatment (7 dpt). Vehicle n=6, ReHMGB1 n=8. Tissues sampled: gastrocnemius (GA), tibialis anterior (TA), heart, liver, kidney.
In vivo therapeutic HMGB1 blockade (Fig 8)
15-month-old male C57BL/6J mice received anti-HMGB1 mAb (clone 3E8, BioLegend 651402) at 0.1 mg/kg IV (or control IgG) on Day −1; 50 μL of 1.2 % BaCl₂ in saline injected unilaterally into the TA muscle on Day 0 to induce acute muscle injury; sacrifice Day 5. n=6 IgG-injured, n=6 aHMGB1-injured, n=6 uninjured.
Pharmacological pathway dissection (Fig 6)
- FPS-ZM1 (Sigma 553030), RAGE antagonist: 100 nM, 30-min pretreatment before ReHMGB1.
- Momelotinib (CYT387) (MedChemExpress HY-10961), JAK2 inhibitor: 0.3 μM.
Readouts
SA-β-gal (Cell Signaling 9860); EdU (Invitrogen C10337); p16^INK4a + p21 immunofluorescence; RT-qPCR (β-actin housekeeping); Western blot; WST-1 proliferation (Takara MK400); Incucyte SX1 live-cell phase confluence; cytokine antibody array (Full Moon Biosystems); multiplex ELISA (R&D Systems HS600C + HSTA00E, Arigo, Chondrex); bulk RNA-seq (Illumina NovaSeq 6000, paired-end 100 bp; STAR + HTSeq + DESeq2; |log₂FC| ≥ 1, p-adj < 0.05); GSEA + GO/KEGG (DAVID + clusterProfiler); STRING + Cytoscape MCODE + TRRUST TF prediction.
Behavioral endpoints (Fig 8J): grip strength (Bioseb force gauge), rotarod latency to fall (0–40 rpm over 300 s), treadmill work (1 % incline, 5–25 m/min, mild electric shock).
Statistics: two-tailed unpaired t-tests with Welch’s correction; one-way ANOVA + Dunnett’s or Tukey’s; Mann–Whitney for non-normal; Benjamini–Hochberg for multiple comparisons. Prism v10.1.1 + R v4.3.3.
Findings
1. HMGB1 release from senescent cells mediates paracrine senescence (Fig 1)
Conditioned media from IR-senescent WI-38 (IR-CM) induced senescence in naïve WI-38 (↑SA-β-gal, ↓EdU). Adding anti-HMGB1 mAb at 2.5 μg/mL to the CM (aHM-CM) markedly attenuated this paracrine senescence:
| Readout | IR-CM | IgG-CM | aHM-CM | p (aHM vs IgG) |
|---|---|---|---|---|
| SA-β-gal positive ratio | ~9 % | ~9 % | ~4.5 % | p < 0.01 |
| EdU positive cells | ~60 % | ~60 % | ~70 % | p < 0.05 |
| Nuclear HMGB1 intensity (recipient cells) | low | low | restored | p < 0.05 |
| IL-6 in CM (ELISA) | ~55 pg/mL/cell | ~55 | ~37 | p < 0.01 |
| TNF-α in CM (ELISA) | ~25 pg/mL/cell | ~25 | ~18 | p < 0.05 |
mRNA (RT-qPCR): aHM-CM significantly reduced p16, p21, p53, IL6, VCAM1, CCL2 in recipient WI-38; CXCL1 was not significantly different. Western blot: p21 protein p < 0.01 and p53 protein p < 0.01 both reduced. n = 3–6 independent experiments per assay.
A supplementary arm (Supp Fig 1) confirmed: mouse dermal fibroblasts (MDFs) from young and old mice exposed to old-mouse serum with anti-HMGB1 antibody showed reduced 3MR luminescence (senescent-cell reporter) and reduced p16, Il6, Mmp3.
2. ReHMGB1 — but not OxHMGB1 — induces senescence (Fig 2)
WI-38 treated with 20 μg/mL ReHMGB1, OxHMGB1, or 20 Gy IR (positive control) for 72 h, analyzed day 4:
| Readout | PBS | ReHMGB1 | OxHMGB1 | IR | ReHMGB1 vs PBS |
|---|---|---|---|---|---|
| WST-1 proliferation | 1.0 | ~0.7 | ~0.95 (ns) | ~0.2 | p < 0.001 |
| SA-β-gal positive ratio | ~1 % | ~5 % | ~2 % (ns) | ~80 % | p < 0.001 |
| EdU positive cells | ~65 % | ~15 % | ~60 % (ns) | ~15 % | p < 0.0001 |
| p21+ cells (% of DAPI) | ~20 % | ~60 % | ~22 % (ns) | ~90 % | p < 0.001 |
mRNA upregulation by ReHMGB1 (not by OxHMGB1): p16, p21, IL6, IL8, CXCL1, SERPINE1, TNF-α, CCL2, VCAM1, TP53, MMP13 (all p < 0.05 to p < 0.001). The IL8 fold-change reached ~20× control.
Generalization: ReHMGB1 induced SA-β-gal + reduced EdU in WI-38, BJ foreskin, renal epithelial, and HSKM cells across “effective concentration range” (Supp Fig 3). OxHMGB1 inert across all four cell types.
3. ReHMGB1 transcriptomically resembles IR-senescence; OxHMGB1 diverges (Fig 3)
Bulk RNA-seq (n ≥ 3 replicates per arm):
- ReHMGB1: 1,087 DEGs (601 up, 486 down) at |log₂FC| ≥ 1, p-adj < 0.05.
- OxHMGB1: 613 DEGs (295 up overlap with Re; 154 down overlap).
- 55 DEGs in CellAge senescence gene set; 35 in cell-cycle regulator set.
- ReHMGB1-specific (vs Ox) upregulated: VCAM1, ICAM1, CCL2, CXCL10, ATF3, IGFBP family, RELB; downregulated oxidative-stress genes RRM2, MYBL2, FOXM1.
- Bystander senescence marker p15^INK4b^ (CDKN2B) upregulated specifically by ReHMGB1.
- Against published IR-senescence dataset GSE13027: ReHMGB1 DEG signature closely resembled IR-induced senescence; OxHMGB1 diverged significantly.
- Versus IR, ReHMGB1 shared 89 upregulated DEGs (incl. VCAM1, IGFBP5, IL15, TGFB1, MMP2, MMP12, PDGFRB) — but exhibited ReHMGB1-specific signatures (STAT5A, COL14A1, IL12A, IRF1).
- GO-BP enrichment: cell migration, ECM organization, signal transduction (concordant with fibrosis + ossification processes).
- KEGG enrichment: cytokine-cytokine receptor interaction, PI3K-AKT, JAK-STAT, TNF signaling.
- OxHMGB1 activated interferon-response GO terms — qualitatively distinct from senescence-driving ReHMGB1.
4. Cytokine-array secretome confirms paracrine-senescence-competent SASP (Fig 4)
CM from ReHMGB1-treated WI-38 enriched in canonical SASP factors: IGFBP5, IL15, TGFB1, MMP2, GSK3A, IGF1R, PDGFRB (plus FGF2, TNFRSF1B, VCAM1, IL10, CX3CL1, IGFBP1/2/3, RETN, TIMP1, CTNNA1, TGFBR2, IGFBP3, MPO). Profile closely resembled IR-senescent CM; OxHMGB1 CM lacked SASP enrichment. KEGG: AGE-RAGE diabetic-complications pathway, MAPK, cytokine-cytokine interaction, FOXO; GO MF: TGF-β I/II/III receptor binding, IGF I/II binding.
5. Mechanism: ReHMGB1 → RAGE → JAK/STAT + PI3K-AKT/NF-κB (Figs 5 + 6)
GSEA (normalized enrichment scores, p = 0.000 each):
| Pathway | NES | Direction |
|---|---|---|
| Cytokine-cytokine receptor interaction | +1.50 | up in Re |
| JAK-STAT signaling | +1.48 | up in Re |
| Cell cycle | −1.81 | down in Re |
| DNA replication | −1.93 | down in Re |
TRRUST TF predictions for ReHMGB1-up DEGs: RELA, NFKB1, STAT1 (plus STAT3, STAT5A, XBP1, REL, ATF4, KLF4, CEBPA, EP300, SP1, EGR1, MYC, JUN, ZFP36); down: E2F1/E2F3/E2F4, PTTG1, TP53, IRF1.
Western blot (n = 3) in ReHMGB1 vs OxHMGB1 vs PBS:
- ↑ RAGE protein (p < 0.01 Re vs PBS)
- ↑ p-AKT (p < 0.05)
- ↑ p-NF-κB p65 (p < 0.05)
- ↑ p-JAK2 (p < 0.05)
- ↑ p-STAT1 (p < 0.01)
Pharmacological rescue (Fig 6): pre-treating ReHMGB1-stimulated WI-38 with either FPS-ZM1 (100 nM, RAGE antagonist) or Momelotinib (0.3 μM, JAK2 inhibitor) restored proliferation (Incucyte phase confluence) and reduced SA-β-gal (p < 0.05) while increasing EdU+ (p < 0.01). Both abrogated ReHMGB1-induced senescence, confirming the RAGE → JAK2 → STAT1 axis is functionally necessary.
6. Systemic ReHMGB1 in young mice induces multi-tissue senescence (Fig 7)
Single IV bolus 5 mg/kg ReHMGB1 in 3-mo C57BL/6J, sampled 7 dpt:
Serum cytokines (multiplex ELISA): ↑ IL-6 (p < 0.05), ↑ IL-1β (p < 0.05), TNF-α ns.
Tissue p21 (Cdkn1a) mRNA:
| Tissue | Δp21 vs Veh | p |
|---|---|---|
| Gastrocnemius (GA) | ↑ | < 0.05 |
| Tibialis anterior (TA) | ↑ | < 0.05 |
| Liver | ↑ | < 0.05 |
| Heart | ↑ (trend) | ns |
| Kidney | ↑ (trend) | ns |
Additional TA mRNA: ↑ p15 (Cdkn2b, p < 0.05), ↑ Cxcl10, ↑ Timp1 (p < 0.05). GA: ↑ Timp1 (p < 0.05). Liver: ↑ Tnf-α (p < 0.05), ↑ Mmp13 (p < 0.05), ↑ Timp1 (p < 0.05).
TA myofiber histology: mean cross-sectional area (CSA) not significantly different at 7 dpt (no atrophy yet), but HMGB1+ myonuclei decreased (p < 0.01) — consistent with nuclear-to-extracellular HMGB1 export — and p16+/p21+ dual-positive cells increased (p < 0.001).
Pharmacokinetics (pull-down assay of ReHMGB1 in serum): undetectable at 0 h baseline; peaked at 6 h post-injection (p < 0.001 vs 0 h); still detectable at 24 h (p < 0.05). The discussion notes the published serum half-life of reduced HMGB1 is ~17 min (Zandarashvili 2013 NMR study), indicating extracellular ReHMGB1 must engage receptors rapidly during a narrow active window.
7. Aged-serum ReHMGB1 elevation — human + mouse (Supp Figs 5 + 6)
- In 24-mo aged C57BL/6 mice: total HMGB1 and reduced HMGB1 both significantly elevated in serum vs 3-mo controls.
- In human serum: ReHMGB1 significantly elevated in older (70–80 yr) vs younger (~40 yr) individuals; total HMGB1 trended up but was non-significant.
This is the most clinically translatable finding of the paper: aged human serum carries elevated levels of the geronic redox form, making it a candidate plasma biomarker of inflammaging tempo as well as a plausible mechanistic node in the aging-blood-impairs-tissue heterochronic phenotype.
8. HMGB1 blockade in middle-aged muscle injury rescues regeneration (Fig 8)
15-mo male C57BL/6J + BaCl₂ TA injury, 0.1 mg/kg anti-HMGB1 3E8 mAb IV at Day −1, sacrifice Day 5. n=6 per arm.
| Readout | Uninjured | IgG-injured | aHMGB1-injured | p (aHMGB1 vs IgG) |
|---|---|---|---|---|
| Muscle mass (mg/g) | ~1.5 | ~1.0 | ~1.0 | ns |
| Serum HMGB1 (ELISA) | ~250 pg/mL | ~3,500 | ~1,000 | < 0.05 |
| SA-β-gal+ cells / unit area | ~0 | ~20 | ~10 | < 0.05 |
| Mean fiber CSA (μm²) | ~3,000 | ~500 | ~1,000 | < 0.01 |
| % centralized nuclei | ~2 % | ~80 % | ~50 % | < 0.05 |
| MyoD+ cells (%) | ~0 % | ~22 % | ~32 % | < 0.05 |
| p21+ cells (%) | ~1 % | ~20 % | ~10 % | < 0.05 |
Behavioral readouts at 4 dpt (Fig 8J):
| Test | Uninjured | IgG-injured | aHMGB1-injured | p (aHMGB1 vs IgG) |
|---|---|---|---|---|
| Grip strength (g/g) | ~4.0 | ~2.8 | ~3.9 | < 0.01 |
| Rotarod latency (s) | ~110 | ~50 | ~85 | < 0.05 |
| Treadmill work (KJ) | ~4 | ~2.5 | ~6.5 | < 0.05 |
Interpretation: anti-HMGB1 mAb blockade in middle-aged mice reduced systemic + tissue senescence, lowered systemic HMGB1, increased myogenic progenitor MyoD+ recruitment, partially restored myofiber CSA, and fully restored grip strength to uninjured levels — i.e., a single low-dose mAb pre-treatment delivers a senomorphic effect with functional readout in a clinically relevant injury context.
Why it matters
- First molecularly explicit mechanism for systemic senescence propagation via circulating SASP. Prior work (Acosta 2008, Kuilman 2008 — see sasp) mapped CXCR2/IL-8 and IL-6 as local paracrine mediators; how the signal travels across the bloodstream to remote tissues had been a major open question (#gap/no-mechanism on altered-intercellular-communication until this paper). Shin 2025 nominates circulating ReHMGB1 → RAGE → JAK/STAT + PI3K-AKT/NF-κB.
- Redox switch is the actionable handle. HMGB1’s three cysteines switch the protein between cytokine-active reduced form, chemokine-active disulfide form (TLR4-engaging per Venereau 2012), and inactive sulfonyl form. The senescence-propagation activity is selective for the reduced form, meaning systemic redox state (plasma thiol/disulfide balance, glutathione status, ROS load) gates whether circulating HMGB1 is geronic. Aged human serum has elevated ReHMGB1.
- Senomorphic, not senolytic. HMGB1 blockade in the muscle-regeneration model suggests a non-cytotoxic alternative to senolytic clearance — disrupt the propagation signal rather than kill the senescent source. Existing clinical HMGB1-blocking strategies (glycyrrhizin, anti-HMGB1 mAbs, recombinant thrombomodulin) become candidates for aging repurposing. Pharmacological rescue with FPS-ZM1 (RAGE antagonist) or Momelotinib (JAK2 inhibitor at the FDA-approved myelofibrosis drug) establishes existing approved-drug-class probes that could be repurposed.
- Plasma biomarker candidate. Elevated ReHMGB1 in 70–80 yr human serum (Supp Fig 6) suggests a redox-form-specific HMGB1 assay could index inflammaging tempo orthogonally to bulk total-HMGB1.
Extrapolation table
| Dimension | Status |
|---|---|
| Pathway conserved in humans? | yes (HMGB1, RAGE, JAK/STAT, NF-κB all canonical; supplementary data demonstrate aged human serum ReHMGB1 elevation) |
| Phenotype conserved in humans? | partial (cell-culture work uses human WI-38/BJ/HSKM; in-vivo phenotype mouse only) |
| Replicated in humans? | no (first-in-class mechanism; no interventional human trial of HMGB1 blockade for aging indication) |
Connections to the wiki
- hmgb1 — protein page; this study is the canonical citation for redox-form selectivity of senescence propagation
- rage — receptor; canonical extracellular HMGB1 receptor; this study confirms RAGE → JAK2 → STAT1 axis with FPS-ZM1 pharmacological probe
- sasp — adds HMGB1 to DAMP-class SASP factors with explicit redox-state-dependent activity; complements existing CXCR2/IL-8 (Acosta 2008) and IL-6 (Kuilman 2008) axes
- chronic-inflammation — extends HMGB1 role beyond NLRP3/pyroptosis cytokine-priming to direct senescence propagation
- altered-intercellular-communication — explicit mechanism for systemic senescence spread (the hallmark’s central but mechanistically underspecified phenomenon)
- jak-stat-pathway · nf-kb — downstream effectors
- satellite-cells — muscle regeneration arm of Fig 8 (MyoD+ recruitment improved)
- sarcopenia · frailty — middle-aged muscle injury rescue is the most clinically translatable phenotype tested
Limitations
- Mouse-only in vivo. No human translation arm (though human aged-serum ReHMGB1 elevation is shown observationally in Supp Fig 6).
- Redox-form pharmacology is hard. Discussion explicitly notes “redox modulation of HMGB1 in vivo remains technically challenging due to its rapid oxidation, with a serum half-life of ~17 min” — therapeutic strategies cannot deliver “stable ReHMGB1”; intervention must target downstream RAGE/JAK or sequester all HMGB1 regardless of redox form.
- Antibody is not isoform-specific. The 3E8 mAb targets all HMGB1 isoforms. Authors note: “isoform-specific tools will be crucial for dissecting functional differences and guiding therapeutic strategies.”
- Single-dose, short-window. In vivo systemic ReHMGB1 was a single IV bolus with 7 dpt readout; no chronic-administration arm. The therapeutic mAb arm was a single Day −1 dose. Long-term safety + efficacy unknown.
- Bulk RNA-seq, not single-cell. Does not resolve recipient-cell-type heterogeneity of susceptibility to ReHMGB1 in tissue.
- 15-mo vs 24-mo gap. Discussion: “comparison with younger mice in future studies would be important to better understand the age-dependent nature of HMGB1-driven senescence under redox regulation.” The therapeutic arm used 15-mo mice; the aging-blood arm used 24-mo. No direct head-to-head.
- muscle-mass-difference is non-significant. Anti-HMGB1 rescued fiber CSA and behavioral readouts, but gross muscle mass did not differ — the rescue is structural-quality + functional, not bulk regrowth.