Dermal Fibroblasts
Dermal fibroblasts (DFs) are mesenchymal-lineage cells embedded within the collagen/elastin-rich extracellular matrix of the dermis. They are the primary biosynthetic engine of the skin: responsible for producing and turning over collagen I, collagen III, elastin, fibrillin-1, hyaluronic acid, and proteoglycans that give skin its mechanical integrity and hydration. Beyond structural maintenance, DFs are mechanosensory (they detect and respond to ECM stiffness and tension via integrin-mediated focal adhesions and the hippo-yap-taz pathway) and participate in paracrine cross-talk with keratinocytes, melanocytes, and immune cells at the dermal-epidermal interface.
In aging, DFs are the central effector cell of skin-aging: their progressive transition into senescent, pro-inflammatory states drives the collagen fragmentation cycle, SASP-mediated local inflammation, and reduced wound healing capacity that underlie the major clinical manifestations of aged skin. See skin-aging for the full phenotype-level treatment; this page covers the DF cell-type biology and aging mechanism.
Identity
| Field | Value |
|---|---|
| Cell type | Resident mesenchymal cell of the dermis |
| Developmental origin | Neural crest (craniofacial dermis) and lateral plate / somite mesoderm (trunk dermis) |
| Proliferative capacity | Limited self-renewal; quiescent under homeostasis; reactivated by wounding or paracrine signals |
| Morphology | Elongated, spindle-shaped; prominent ER/Golgi reflecting high secretory activity; embedded in ECM via integrin-mediated focal adhesions |
Marker panels
No single marker is DF-specific; reliable identification requires combinatorial gating. Lineage-tracing caveats apply — FSP1+/S100A4+ labels other mesenchymal populations including macrophages in some contexts.
| Marker | Species | Notes |
|---|---|---|
| Vimentin | Both | Universal intermediate-filament mesenchymal marker; not DF-specific |
| PDGFRA | Both | Strongly expressed in resting DFs; best single positive marker |
| CD90 / THY1 | Both | Marks papillary-dermis-enriched fibroblast fraction; lower in reticular |
| FAP (fibroblast activation protein) | Human (activated) | Marks activated / myofibroblast state; used in imaging and therapeutic targeting |
| S100A4 / FSP1 | Both (subset) | Controversy: labels a DF subset but also expressed on macrophages; use with caution |
| DDR2 (discoidin domain receptor 2) | Mouse (primary) | Collagen receptor; broadly used as mouse CF marker; cross-reacts with other stromal cells contradictory-evidence |
| ER-TR7 | Mouse | Labels reticular DF network; does not resolve to a single human ortholog |
| COL1A1 | Human | Highly expressed; reflects core ECM-secreting function; not DF-exclusive |
Activated myofibroblast markers (acquired upon wounding, TGF-β exposure, or persistent SASP activation):
- α-SMA (ACTA2)+ — canonical myofibroblast marker
- Periostin (POSTN)+
- FAP+
Subsets — single-cell atlas (Solé-Boldo 2020)
Single-cell RNA-sequencing of human skin identified four main DF subpopulations that can be spatially mapped within the dermis 1:
| Subset | Location | Functional signature |
|---|---|---|
| Secretory-papillary | Upper (papillary) dermis | High CD90; strong ECM synthesis; active communication with undifferentiated keratinocytes at DEJ; marked by APCDD1 |
| Secretory-reticular | Deep (reticular) dermis | Lower CD90; prominent structural-scaffold and collagen-catabolic gene programs; marked by CTHRC1 |
| Pro-inflammatory | Distributed (vasculature-associated) | Enriched for cytokine/chemokine production (CCL19, APOE); expands with aging |
| Mesenchymal | Reticular dermis (near hair follicle) | Mesenchymal/progenitor signature; marked by ASPN; perifolliclular localisation |
Key aging shift: with chronological aging, fibroblasts undergo a loss of transcriptional priming — the transcriptional distinctions between subsets erode, the secretory-papillary subset loses its papillary-dermis spatial gene-expression signature, and secretory-reticular fibroblasts display a less pronounced reticular signature 1. All four subpopulations show age-related G1 cell-cycle delay and reduced collagen gene expression. Intercellular communication between DFs and undifferentiated keratinocytes is substantially reduced in aged skin — particularly interactions involving undifferentiated keratinocytes (clusters at DEJ), consistent with loss of productive DEJ maintenance 1. needs-replication — the Solé-Boldo 2020 atlas used 5 donors (2 young: 25 and 27 y/o; 3 old: 53, 69, and 70 y/o) from a single centre; subset identities and aging shifts need independent validation in a larger, diverse cohort.
| Dimension | Status |
|---|---|
| Subset architecture conserved across individuals? | yes — four major populations replicated across 5 donors and confirmed by RNA FISH in independent FFPE sections |
| Aging shifts replicated in mouse? | partial — Solé-Boldo 2020 notes differences from mouse fibroblast aging studies; no direct human–mouse single-cell aging atlas comparison published needs-replication |
| Replicated in independent human cohort? | not yet needs-replication |
Core functions
ECM production and remodelling
DFs are the dominant producers of the dermal ECM: collagen I and III (fibrillar scaffold, ~80% of dry dermal weight), collagen IV (basement membrane zone), elastin and fibrillin-1 (elastic network), hyaluronic acid and versican/decorin (hydrated proteoglycan gel). Under homeostasis, DFs maintain ECM turnover via matrix metalloproteinases (MMP-1, MMP-2, MMP-3, MMP-9) and their inhibitors (TIMPs) in dynamic balance.
TGF-β/Smad3 signalling is the primary anabolic driver of collagen I transcription in DFs 2. Smad3 protein levels and TGF-β responsiveness both decline in aged human DFs, contributing to the age-associated fall in collagen synthesis 2. no-fulltext-access — Purohit 2016 is not_oa in the archive; quantitative claims should be re-confirmed once PDF access is available.
Mechanosensing via integrins and c-Jun/AP-1
DFs sense ECM stiffness and tension through integrin-mediated focal adhesions. In young skin, high mechanical integrity of intact collagen fibres maintains cytoskeletal tension and fibroblast spreading. Fragmented ECM in aged skin reduces mechanical tension and fibroblast stretch, which — via α2β1 integrin signalling and elevated c-Jun/AP-1 activity — elevates MMP-1 expression and reduces collagen synthesis 3. Reduced mechanical tension also raises intracellular oxidant levels, further amplifying MMP-1 through the same AP-1 axis. This creates a mechano-oxidative-epigenetic loop driving progressive dermal atrophy 3. The hippo-yap-taz pathway has been independently implicated in fibroblast mechano-regulation (nuclear YAP/TAZ on stiff ECM → pro-anabolic programmes; cytoplasmic YAP/TAZ on fragmented ECM → reduced collagen synthesis), but this mechanism is not described in Fisher 2009 and requires a separate citation. unsourced — YAP/TAZ in aged dermal fibroblasts in vivo needs a primary-source citation.
Paracrine regulation of epidermal homeostasis
DFs secrete keratinocyte growth factor (KGF/FGF7), IGF-1, and HGF, which maintain epidermal basal-layer proliferation and stem cell activity. Age-related DF senescence and transcriptional changes reduce these trophic signals, contributing to the stem-cell-exhaustion of epidermal progenitors and the clinical feature of skin thinning.
Cellular senescence in dermal fibroblasts
Dermal fibroblasts are the most-studied senescent cell population in skin and one of the most-cited examples of how cellular senescence contributes to tissue aging 4 5.
Routes to senescence
| Route | Trigger | Notes |
|---|---|---|
| Replicative (Hayflick) | Telomere attrition from cell division | Cumulative across lifespan; dominant in intrinsic aging |
| UV-SIPS (stress-induced premature senescence) | UV-induced DNA double-strand breaks, oxidative damage | Dominant in photoaged skin; faster-onset than replicative |
| Mitochondrial oxidative stress | ROS-mediated mtDNA damage, electron transport chain dysfunction | Sod2-/- mouse model: p16^INK4a ~2-fold elevated, epidermal thinning 5 |
| Oncogene activation | RAS, RAF overactivation | Relevant in premalignant lesions; less dominant in pure aging context |
Both replicative and UV-SIPS routes converge on p21^CIP1 → p16^INK4a-mediated G1 arrest. Senescent DFs accumulate in the dermis with age; p16^INK4a immunostaining is detectable in aged human skin biopsies.
SASP composition in dermal fibroblasts
Senescent DFs secrete a pro-inflammatory SASP that drives local tissue dysfunction:
| SASP factor | Effect in dermis |
|---|---|
| IL-6 | Promotes chronic inflammation; activates STAT3 in adjacent cells |
| IL-8 / CXCL8 | Neutrophil recruitment; amplifies local inflammation |
| CXCL1 | Chemokine; feeds inflammaging loop |
| MMP-1 (collagenase-1) | Cleaves fibrillar collagen I/III → fragmentation loop 3 |
| MMP-3 (stromelysin-1) | Broad ECM degradation; can activate other MMPs |
| MMP-9 (gelatinase B) | Degrades denatured collagen / gelatin; basement membrane disruption |
This SASP profile amplifies the collagen fragmentation cycle and contributes to chronic-inflammation in the dermis.
Beneficial transient senescence in wound healing
Senescent fibroblasts and endothelial cells transiently appear at wound sites and secrete PDGF-AA (a SASP component specific to acute wound senescence), which drives myofibroblast differentiation and optimal granulation tissue formation 4. Genetic elimination of senescent cells (using p16-3MR transgenic mice + ganciclovir) significantly delays wound closure (peak delay at day 6 post 6mm punch biopsy); topical recombinant PDGF-AA fully rescues this delay 4. Persistent, age-accumulated senescence becomes pathological when immune surveillance (NK cell + macrophage clearance) declines; the same cell type that is beneficial acutely becomes a driver of chronic dysfunction when its clearance fails.
This represents a canonical example of the antagonistic pleiotropy principle at the cellular level — see cellular-senescence for broader mechanistic context.
| Dimension | Status |
|---|---|
| Pathway conserved in humans? | yes — PDGF-AA and p16 are conserved; human wound senescence documented histologically |
| Phenotype conserved in humans? | yes — impaired wound healing is a universal feature of aged skin |
| Replicated in humans (PDGF-AA mechanism)? | partial — mechanistic PDGF-AA evidence is mouse; human correlative data only needs-human-replication |
Collagen homeostasis — the self-amplifying fragmentation loop
Fisher 2009 identified a self-amplifying degradation cycle operating in aged dermis 3:
- Aged/photoaged DFs have elevated basal MMP-1 (collagenase-1; mRNA ~8-fold higher in aged vs young dermis in vivo), causing collagen I fragmentation
- Fragmented collagen reduces fibroblast stretch and mechanical tension, and raises intracellular oxidant levels (ROS approximately 3-fold higher in fragmented vs intact 3D collagen lattice cultures)
- Elevated ROS activates c-Jun/AP-1 and upregulates α2β1 integrin expression, further elevating MMP-1 transcription
- Reduced mechanical tension independently amplifies MMP-1 via the same c-Jun/AP-1 axis
- Net result: progressive collagen I loss, increasing collagen I:III ratio shift, dermal thinning
This loop is self-sustaining once established and is an example of loss-of-proteostasis in the ECM compartment. It is distinct from the SASP-driven inflammatory pathway but is synergistic with it: SASP-derived MMP-3 and MMP-9 accelerate fragmentation that feeds back into step 2. Fisher 2009 used aged human skin biopsies (young: 21–30 y/o; aged: >80 y/o) and three-dimensional collagen lattice cultures of primary adult human DFs; n=3–6 per condition.
Reduced TGF-β/Smad3-driven collagen I synthesis in aged DFs independently feeds step 4 by reducing the replenishment rate 2. needs-replication for quantitative Smad3 rescue experiments in aged human skin in vivo.
| Dimension | Status |
|---|---|
| Pathway conserved in humans? | yes — Fisher 2009 used aged human skin biopsies and cultured human DFs |
| Phenotype (collagen loss ~1%/yr) conserved? | yes — multiple independent dermal biopsy cohort studies |
| Replicated independently? | yes — multiple labs; mechanistic detail is Fisher-lab primary |
Niche signalling
TGF-β / Smad3 axis
Primary anabolic signal for collagen I and III synthesis. Originates from DFs themselves (autocrine), from activated keratinocytes, and from platelets/macrophages during wound healing. Declines in aged DFs 2. Excessive TGF-β signalling in activated myofibroblasts drives fibrotic scarring — the same pathway is anabolic at homeostatic levels and pro-fibrotic when chronically elevated.
Wnt / β-catenin
wnt-beta-catenin signalling regulates DF proliferation and differentiation. Active canonical Wnt maintains fibroblast progenitor character; dysregulation has been linked to both fibrosis (over-activation) and age-related atrophy (Wnt inhibitor secretion increasing with age). no-mechanism — the net directional change in Wnt activity in aged human DFs is not established.
YAP/TAZ — mechanotransduction
hippo-yap-taz pathway integrates ECM mechanical signals. Nuclear YAP/TAZ in response to stiff intact ECM drives collagen synthesis. Cytoplasmic YAP/TAZ on fragmented/loose ECM reduces anabolic output. This provides a plausible molecular link between ECM fragmentation (upstream) and collagen synthesis decline (downstream), complementary to the Fisher 2009 c-Jun/AP-1/integrin model. However, YAP/TAZ mechanotransduction in aged dermal fibroblasts in vivo is not established by Fisher 2009 (which does not address the Hippo pathway). A primary source for YAP/TAZ in aged DFs is needed. unsourced
Paracrine inputs from keratinocytes
Keratinocytes supply IL-1α, IL-1β, and KGF to DFs at the DEJ. IL-1α is a key SASP amplifier — paracrine from keratinocytes or autocrine from DFs themselves — that sustains NF-κB activity and MMP expression. With aging, DEJ cross-talk diminishes (rete ridge flattening reduces contact area; Solé-Boldo 2020 ligand-receptor predicted interaction analysis confirms reduced communication) 1.
Single-cell aging signature
The Solé-Boldo 2020 human skin single-cell RNA-seq atlas provides the most comprehensive transcriptomic characterization of DF aging in human tissue 1:
- Study design: 5 donors (2 young: 25 and 27 y/o; 3 old: 53, 69, and 70 y/o); >15,000 skin cells total; >5,000 fibroblasts analysed; sun-protected inguinoiliac region biopsies (4mm punch); scRNA-seq via 10X Genomics
- Four DF subsets with distinct spatial localisation and functional programmes: secretory-reticular (cluster #1, CTHRC1+), pro-inflammatory (cluster #2, CCL19+/APOE+), secretory-papillary (cluster #3, APCDD1+/COL18A1+), and mesenchymal (cluster #9, ASPN+); spatially validated by RNA FISH in FFPE sections
- Age-related loss of fibroblast priming: all four subpopulations show reduced expression of their defining functional-annotation gene sets in aged donors, consistent with de-differentiation or loss of niche-instructed identity; aged fibroblasts in secretory-papillary cluster present less papillary and more reticular expression signatures
- All subpopulations show G1 cell-cycle delay in aged donors
- Predicted ligand-receptor interactions between DFs and undifferentiated keratinocytes (at DEJ) are substantially reduced in aged donors — particularly pronounced in the two oldest donors (≥69 y/o); providing a cellular basis for the rete-ridge flattening and reduced trophic support seen histologically
- Old fibroblasts express genes encoding skin aging-associated secreted proteins (SAASP per Wulupa 2014) — differential expression in aged clusters
Frontmatter single-cell-aging-signature summary: “Loss of transcriptional priming across four secretory-papillary/secretory-reticular/pro-inflammatory/mesenchymal subsets; reduced intercellular communication with undifferentiated keratinocytes; age-associated gain of pro-inflammatory transcriptional state (Solé-Boldo 2020)”
needs-replication — single-center atlas; independent validation in a more diverse cohort and cross-comparison with mouse Tabula Muris Senis data would strengthen these conclusions.
Therapeutic angles
Senolytic clearance
Senolytic agents that clear p16^INK4a+ cells are under investigation for skin aging. In mouse models, removing senescent cells improves wound healing 4 and reduces SASP-driven tissue damage. fisetin (100 mg/kg oral; Yousefzadeh 2018) and dasatinib + quercetin (D+Q) are the most-studied senolytics in skin contexts, though published skin-specific endpoints in human RCTs are not yet available as of 2026-05-19. See senolytics and skin-aging for therapeutic-landscape context.
needs-human-replication — no completed, published human RCT has used skin-specific aging endpoints as primary outcomes in a senolytic trial.
Retinoid stimulation of collagen synthesis
Topical tretinoin (all-trans retinoic acid) is the best-characterised pharmacological intervention for skin aging: it suppresses the UV-induced AP-1/NF-κB-MMP axis (reducing MMP-1, MMP-3, MMP-9 induction by 50–80%) and stimulates collagen I and III synthesis via RAR-β in DFs 6. Mechanism in DFs: AP-1 transrepression counteracts the MMP-driven fragmentation loop. See skin-aging for the full clinical-evidence profile.
TGF-β pathway modulation
Pro-anabolic TGF-β supplementation or Smad3 re-activation in aged DFs is a candidate therapeutic strategy to rescue collagen synthesis. No clinical-grade modality has been validated for skin aging specifically; pirfenidone (anti-fibrotic TGF-β inhibitor) has the opposite goal and would worsen atrophy. no-mechanism — the optimal intervention point in the TGF-β/Smad3 axis for aged DF rescue has not been established.
Limitations and gaps
| Gap | Tag | Notes |
|---|---|---|
| Subset abundance quantification in vivo | needs-replication | Solé-Boldo 2020 is single-center; absolute subset frequency + aging shift not yet confirmed in independent cohort |
| Human PDGF-AA wound mechanism | needs-human-replication | Demaria 2014 is mouse; human wound biopsy PDGF-AA data limited to correlational |
| Smad3 rescue in vivo (aged human skin) | needs-replication | Purohit 2016 in-vitro; no in-vivo aged human skin intervention data |
| Purohit 2016 quantitative claims | no-fulltext-access | Closed-access (not_oa); archive confirms not_oa; do not rely on specific quantitative claims until PDF access |
| Wnt aging directionality | no-mechanism | Net change in canonical Wnt signalling in aged human DFs not established |
| FSP1/S100A4 marker specificity | contradictory-evidence | Labels DFs and macrophages; not reliable as sole positive marker |
| Senolytic RCTs with skin endpoints | needs-human-replication | No published RCT uses skin aging as primary senolytic endpoint |
| DF senescence burden quantification | needs-replication | p16^INK4a+ DF frequencies in large human age-stratified dermis cohorts not robustly established |
Cross-references
- skin-aging — phenotype hub; full clinical manifestations and therapeutic landscape
- dermis — tissue-niche context (implicit stub — dermis page not yet seeded; R38 substrate)
- cellular-senescence — upstream hallmark; p16/p21 pathway; antagonistic-pleiotropy framing
- loss-of-proteostasis — collagen fragmentation loop; ECM proteostasis failure
- chronic-inflammation — DF SASP drives dermal inflammaging; IL-6, IL-8, MMP axis
- altered-intercellular-communication — reduced DF-keratinocyte cross-talk with aging
- tgf-beta — primary anabolic axis for DF collagen synthesis; declines in aging
- wnt-beta-catenin — DF proliferation and differentiation regulator
- hippo-yap-taz — mechanosensing pathway; links ECM fragmentation to collagen synthesis decline
- sasp — SASP composition in senescent DFs; IL-6, MMPs
- replicative-senescence — replicative senescence route in DFs
- fisetin — senolytic studied in skin aging contexts
- quercetin — senolytic (D+Q combination); needs-human-replication for skin endpoints
- dasatinib — senolytic (D+Q combination)
- senolytics — class page for senolytic interventions
- cardiac-fibroblasts — structural analog page (closest in-wiki prototype)
- hair-greying — sister aging phenotype in skin; planned page (implicit stub)
Footnotes
Footnotes
-
doi:10.1038/s42003-020-0922-4 · Solé-Boldo L, Raddatz G, Schütz S et al. · 2020 · Commun Biol 3:188 · in-vivo (human tissue scRNA-seq) · n=5 donors (2 young 25–27 y/o; 3 old 53–70 y/o); >15,000 cells total; >5,000 fibroblasts · model: 4mm punch biopsies from sun-protected inguinoiliac skin of healthy male Caucasian donors · 493 citations · 100th percentile · Identifies four DF subsets: secretory-reticular (CTHRC1+), pro-inflammatory (CCL19+/APOE+), secretory-papillary (APCDD1+/COL18A1+), mesenchymal (ASPN+); spatially validated by RNA FISH; demonstrates age-associated loss of transcriptional priming (all subsets lose functional-annotation gene expression), G1 cell-cycle delay in all aged subpopulations, and reduced predicted ligand-receptor interactions between DFs and undifferentiated keratinocytes at the DEJ needs-replication (single-center, 5 donors, male-only) · local PDF downloaded: (local PDF) ↩ ↩2 ↩3 ↩4 ↩5
-
purohit-2016-smad3-fibroblasts · PMID 27132061 · doi:10.1016/j.jdermsci.2016.04.004 · Purohit T, He T, Qin Z, Li T, Fisher GJ, Yan Y, Voorhees JJ, Quan T · 2016 · J Dermatol Sci 83(1):80-83 · Letter · in-vitro · model: human dermal fibroblasts, young vs aged donors · volume/issue/pages confirmed via PubMed efetch + Crossref (2026-05-19); archive: not_oa (closed-access) — content unverifiable until PDF access obtained no-fulltext-access. Do not rely on specific quantitative claims. ↩ ↩2 ↩3 ↩4
-
fisher-2009-collagen-fragmentation-mmp · doi:10.2353/ajpath.2009.080599 · Fisher GJ, Quan T, Purohit T, Shao Y, Cho MK, He T, Varani J, Kang S, Voorhees JJ · 2009 · Am J Pathol 174(1):101-114 · in-vivo + in-vitro · n=4–7 per group (in vivo biopsies: young 21–30 y/o; aged >80 y/o); n=3–5 per condition (lattice cultures) · model: sun-protected buttock biopsies from young vs aged volunteers + 3D collagen lattice cultures of primary adult human DFs · 444 citations · 100th percentile · MMP-1 mRNA ~8-fold elevated in aged human dermis in vivo; collagen fragmentation (via MMP-1 treatment of 3D lattices) raises ROS ~3-fold and elevates MMP-1 via c-Jun/AP-1 and α2β1 integrin axis; MitoQ10 (1 nmol/L) reduces oxidant levels and MMP-1 mRNA by ~30–40%; establishes self-amplifying collagen-fragmentation–oxidative stress–MMP-1 cycle in aged skin · Note: the YAP/TAZ pathway is NOT described in this paper; mechanosensing mechanism is c-Jun/AP-1 + α2β1 integrin · local PDF downloaded: (local PDF) ↩ ↩2 ↩3 ↩4
-
demaria-2014-senescent-cells-wound-healing · doi:10.1016/j.devcel.2014.11.012 · Demaria M, Ohtani N, Youssef SA et al. · 2014 · Dev Cell 31(6):722-733 · in-vivo + in-vitro · n≥4/group · model: p16-3MR transgenic mouse (6mm punch biopsy to dorsal flank); p16/p21 DKO mice; primary mouse skin fibroblasts/endothelial cells · 1,889 citations · 100th percentile · Senescent fibroblasts and endothelial cells transiently appear at wound sites and secrete PDGF-AA; GCV-mediated senescent cell elimination delays wound closure (peak at day 6); topical recombinant PDGF-AA rescues this delay; senescent cells do not drive wound closure via contraction; establishes dual beneficial (transient) / pathological (persistent) role of dermal senescence · local PDF downloaded: (local PDF) ↩ ↩2 ↩3 ↩4
-
velarde-2012-mitochondria-skin-senescence · doi:10.18632/aging.100423 · Velarde MC, Flynn JM, Day NU, Melov S, Campisi J · 2012 · Aging (Albany NY) 4(1):3-12 · in-vivo + in-vitro · n=6–13/age group (WT); n=8–9 (Sod2-/-) · model: Sod2-/- mouse skin (EUK-189 maintained) + cultured human keratinocytes/fibroblasts (rotenone) · 276 citations · 100th percentile · Mitochondrial oxidative damage from SOD2 deficiency drives p16^INK4a elevation (~2-fold), epidermal thinning, and stratum corneum thickening; human keratinocytes more sensitive than DFs to rotenone-induced senescence needs-human-replication · local PDF downloaded: (local PDF) ↩ ↩2
-
fisher-1996-photoaging-ap1-mmp · doi:10.1038/379335a0 · Fisher GJ et al. · 1996 · Nature 379:335-339 · in-vivo · n=6–17 per timepoint (mRNA); n=9–10 (AP-1/NF-κB, protein, activity) · model: adult Caucasian human buttock skin, in-vivo UVB 2 MED · 1,395 citations · 100th percentile · Sub-erythemogenic UVB (≥0.01 MED) activates AP-1 and NF-κB → MMP-1, MMP-3, MMP-9 upregulation; MMP-2 not induced; all-trans retinoic acid pretreatment reduces AP-1 binding ~70% and MMP-1/MMP-9 mRNA/protein/activity by 50–80% · local PDF downloaded: (local PDF) ↩