Supplementary Figure S1. Human recombinant HMGB1 rescue the migration defect andthe reduced collagen production seen in the presence of HMGB1 knock-down cells. (a)The reduced migration rate seen in the presence of HMGB1 knock-down cells wereincreased after recombinant HMGB1 (rHMGB1) was added. Scale Bar = 200 μm. (b)Quantification of the migration rate at 24h (n=4). (c) rHMGB1 rescues the reducedcollagen production in the HMGB1 knock-down cells (n=6). Data are shown as the mean± SD. *p<0.05.

Supplementary Figure S2. Human recombinant HMGB1 activates fibroblasts in a dosedependent manner. (a) The expression of α-SMA and pro-collagen I were enhanced withthe increase of HMGB1 concentration, and 100ng/ml HMGB1 treatment showing themost effective in elevating these markers expression. Scale Bar = 200 μm. (b)Quantification of the expression of α-SMA and pro-collagen I (n=4). (c) The migrationrate of fibroblasts that treated with different concentrations of HMGB1. Scale Bar = 200μm. (d) Quantification of the migration rate at 24h (n=4). (e) The collagen synthesis offibroblasts that subjected to different concentrations of HMGB1 (n=6). Data are shown asthe mean ± SD. *p<0.05, **p<0.01.

Supplementary Figure S3. Wounds healing under circumstances of hydration showingreduced HMGB1 levels and decreased scar formation. (a) The clinical appearance of thewounds at day 0 and day 28 after wounding. Scale bar = 5 mm. (b) Scars were observedfrom the side to show the height of the scars. Scale bar = 5 mm. (c) The analysis of theoutstanding height of the scar on postoperative day 28 (n=6). (d) The analysis of SEI onday 28 (n=6). (e) Representative immunohistochemical staining for HMGB1 in scartissues. Scale Bar = 100 μm. Data are shown as the mean ± SD. *p<0.05.

Supplementary Figure S4. The synergistic funtcion of HMGB1 and S100A12 inactivating fibroblasts. (a) Compared with fibroblasts that treated with HMGB1 orS100A12 alone, the expression of α-SMA and pro-collagen I increased significantly inthe fibroblasts that treated with HMGB1 and S100A12 together. Scale Bar = 200 μm. (b)Quantification of the expression of α-SMA and pro-collagen I (n=4). (c) The migrationrate of fibroblasts that treated with HMGB1 or/and S100A12. Scale Bar = 200 μm. (d)Quantification of the migration rate at 12h (n=4). (e) The collagen synthesis of fibroblaststhat subjected to HMGB1 or/and S100A12 (n=6). Data are shown as the mean ± SD.*p<0.05 when compared with control group; #p<0.05 when compared with HMGB1 orS100A12 group.

Supplementary Figure S5. The expression of inflammatory genes. (a) HMGB1 gene wasknocked down in dermal fibroblats, and the efficiency of knockdown was confirmed bywestern blot analysis (n=4). (b) The expression of TNF-α, IL-6 and IL-8 were decreasedin kerotinocytes after knock-down of HMGB1 (n=6). (c) The expression of S100A8,S100A9 and S100A12 didn’t statistically changed after knock-down of HMGB1 (n=6).(d-f) S100A8, S100A9 and S100A12 genes were knocked down in dermal fibroblats, andthe efficiency of knockdown was confirmed by western blot analysis (n=4). (g-i) Theexpression of HMGB1 in kerotinocytes after knock-down of S100A8, S100A9 andS100A12, respectively (n=6). Data are shown as the mean ± SD. *p<0.05 whencompared with wide type group.

Supplementary Figure S6. Schematic drawing of the mechanism of epidermal HMGB1on dermal fibroblasts activation under reduced hydration.

Supplementary Figure S7. The quantification of migration rate of fibroblasts over time. (a)The migration rate of fibroblasts co-cultured wide type HaCaT and HMGB1-knockdownHaCaT in normal or reduced hydration conditions. (b) The migration rate of fibroblaststhat treated with CM from stratified wide type HaCaT or HMGB1-knockdown HaCaTcultured in different hydration conditions. (c) The migration rate of HMGB1-treatedfibroblasts when RAGE, TLR2 or TLR4 were blocked, respectively. (d) The migrationrate of HMGB1-treated fibroblasts when RAGE, TLR2 or TLR4 were blocked,respectively or at the same time. (e) The migration rate of fibroblasts that treated withdifferent concentrations of HMGB1. (f) HMGB1 rescues the migration defect induced byknockdown of HMGB1 under different hydration conditions. (g) The migration rate offibroblasts that treated with HMGB1 or/and S100A12. (n=4). Data are shown as the mean± SD. *p<0.05 when compared at 16h and 24h. #p<0.05, ##p<0.01 when compared at 8hand 12h.

Materials and Methods

Human studies—skin biopsies and cell isolationSkin tissues were obtained from patients with hypertrophic scars who had undergone scarresection and skin grafting. The patients were informed about the purpose and proceduresof the study and voluntarily agreed to provide skin tissue samples. Written informedconsent was obtained from all donors, and all protocols were approved by the EthicalCommittee of the First Affiliated Hospital of Sun Yat-Sen University (Approval No.[2017] 059).

For dermal fibroblast isolation, we followed our previously published protocol []. Briefly,the dermis was minced and digested with collagenase type I solution (Sigma, St. Louis,MO, USA). Cells were harvested by filtering through a 100-mm cell filter (BDBiosciences, Bedford, MA, USA), and were then pelleted and cultured in Dulbecco’smodified Eagle’s medium (DMEM, Gibco, Grand Island, NY, USA) supplemented with10% fetal calf serum (Gibco). The cells were used at passages 2–4.

Cell cultureThe stratified human keratinocyte culture (SKC) model, which is also namedthree-dimensional (3D) culture of keratinocytes, was established following our previousprotocols. Briefly, 5 × 105 HaCaT cells were seeded onto 12-well transwell culture inserts(12-mm diameter, 0.4-μm pore size) (EMD Millipore, Billerica, MA, USA). The cellswere cultured in DMEM containing 10% fetal bovine serum (FBS) for 24-48 hours. Then,the medium in the upper chambers was drained, and the medium in the lower chamberswas replaced with E-medium containing 50:50 (v/v) DMEM and DMEM/F-12supplemented with 18 μM adenine, 500 ng/ml bovine pancreatic insulin, 500 ng/mlhuman Apo-transferring, 500 ng/ml triiodothyronine, 4 mM L-glutamine, 0.4 μg/mlhydrocortisone, 10 ng/ml cholera toxin, 5 ng/ml epidermal growth factor (EGF), 5% FBS,and antibiotics (penicillin and streptomycin). The E-media was changed every 2-3 days,and the stratified epithelium formed between day 10 and day 14.To further study the effect of keratinocytes on dermal fibroblast activity, thekeratinocyte-fibroblast co-culture (KFC) model and conditioned medium treatment (CMT)model were used. In the KFC model, the stratified keratinocytes that had grown on theupper chambers were co-cultured with monolayer dermal fibroblasts that had grown onthe lower chambers (Figure. 2a). In the CMT model, conditioned medium collected fromkeratinocytes was used to treat fibroblasts. Briefly, HaCaT cells were starved inserum-free medium overnight and subsequently exposed to different hydration conditionsand reagents. The medium was collected after 24 hours of incubation, diluted with freshFBS-free DMEM with a 1:20 ratio, and then applied to treat fibroblasts (Figure. 3a).

RNA interference

Cells were plated and grown to approximately 80% confluency. SMARTpool:siGENOME HMGB1 siRNA (Dharmacon, Lafayette, CO, USA) was used to knockdownthe expression of HMGB1 in HaCaT cells. SMARTpool: siGENOME AGER siRNA,siGENOME TLR2 siRNA and siGENOME TLR4 siRNA (Dharmacon) were used toknockdown the expression of RAGE, TLR2 and TLR4, respectively, in dermal fibroblasts.siGENOME Non-Targeting siRNA Pools were used as negative controls of siRNA,which were composed of four siRNAs designed to target no known genes in humans. TheDharmaFECT 1 siRNA Transfection Reagent was used to transfect HaCaT cells withHMGB1, TLR2, TLR4 or scrambled siRNA oligonucleotides. The efficiencies of geneknockdown were confirmed by western blotting analysis.

Quantitative RT-PCR (qRT-PCR)Total RNA was purified using the RNeasy Mini Kit (Qiagen) and was reverse-transcribedinto cDNA using a thermocycler (S1000, Bio-Rad) and the First Strand cDNA SynthesisKit (Fermentas) according to the manufacturer’s protocol. Real-time RT-PCR wasperformed using the SYBR qPCR mix (Toyobo) and a Real-Time PCR Detection System(Bio-Rad iQ5). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as anormalization control, and the relative gene expression was analyzed using the 2△△Ct

method. The primer sequences were as follows: IL-6 forward, 5‘-AAA TTC GGT ACATCC TCG ACGG-3’, reverse, 5‘-GGA AGG TTC AGG TTG TTT TCTG-3’; IL-8forward, 5‘-TTT TGC CAA GGA GTG CTA AAGA-3’; reverse, 5’-AAC CCT CTGCAC CCA GTT TTC-3’; TNF-α forward, 5’-ATG AGC ACT GAAAGC ATG ATCC-3’,reverse 5’-GAG GGC TGA TTA GAG AGA GGTC-3’; S100A8 forward, 5’-ATG CCGTCTACA GGGATGAC-3’, reverse 5’-ACG CCCATC TTTATCACCAG-3’；S100A9forward, 5’-GGT CAT AGA ACA CAT CAT GGA GG-3’, reverse 5’-GGC CTG GCTTAT GGT GGTG-3’; S100A12 forward, 5’-ATT GAG GGG TTAACATTAGGC TG-3’,reverse, 5’-GAT ATT CTT GAT GGT GTT TGC AAGC-3’; GAPDH forward, 5’-TGTTGC CAT CAATGACCC CTT-3’, reverse, 5’-CTC CAC GAC GTACTCAGCG-3’.

ImmunostainingFor immunohistochemical staining of tissues, samples were fixed with 4%paraformaldehyde, processed, paraffin-embedded, and sectioned at 5-μm thickness. Theactivity of endogenous peroxidase was quenched by incubation with 1% hydrogenperoxide followed by treatment with antigen retrieval solution (Dako, Carpinteria, CA,USA). Then, the sections were incubated with mouse anti-human HMGB1 (1:200dilution, Abcam, Cambridge, UK), mouse anti-human α-SMA (1:400 dilution, Abcam) at4℃ overnight, followed by incubation with secondary biotinylated secondary antibody(Vector laboratories, Burlingame, CA, USA) and avidin-biotin complex (Elite ABC kit;Vector Laboratories). Signals were visualized using 3,3’-diaminobenzidine (DAB), andnuclei were stained with hematoxylin (Sigma). Negative control experiments were

performed via omission of the primary antibody. Images were captured using a lightmicroscope (BX51WI, Olympus Co., Ltd., Tokyo, Japan).For immunofluorescence staining of cells, cells grown on the culture glass were fixedwith 4% paraformaldehyde (PFA), permeabilized with 0.3% Triton X100, blocked with10% normal goat serum, and incubated with mouse anti-human HMGB1 (1:200 dilution,Abcam), α-SMA (1:200 dilution, Abcam), pro-collagen-I (1:200 dilution, Abcam), F4/80(1:200 dilution, Abcam) and MRTF-A (1:200 dilution, Santa Cruz, Biotechnology, CA,USA). Then, the sections were incubated with Alexa Fluor 594 or Alexa Fluor 488conjugated secondary antibodies (Thermo Fisher Scientific, Rockford, IL, USA). Cellnuclei were stained with DAPI (4’6-diamidino-2 phenylindole, Sigma). Images wereobtained with a fluorescence microscope (BX51WI Olympus) and were merged usingImage-Pro Plus v. 6.0 software (Media Cybernetics, Inc, Rockville, MD, USA). Negativecontrol experiments were performed by omitting the primary antibody.

Western blotThe samples were homogenized, and radioimmunoprecipitation assay (RIPA) buffer(Invitrogen, Carlsbad, CA, USA) was used for the total protein extractions. A Nuclearand Cytoplasmic Protein Extraction Kit (Beyotime Biotechnology, Shanghai, China) wasused for nuclear and cytosolic protein extraction. Equal amounts of protein wereseparated by SDS-PAGE (polyacrylamide gel electrophoresis), transferred ontonitrocellulose membranes (Amersham, Chalfont, UK) for immunoblotting, and incubatedwith 5% nonfat dry milk for blocking. Mouse anti-human HMGB1 (1:3000 dilution,Abcam), mouse anti-human collagen I (1:5000 dilution, Abcam), mouse anti-humanMRTF-A (1:5000 dilution, Santa Cruz), mouse anti-human Lamin B1 (1:5000 dilution,Proteintech, Rosemont, IL, USA), mouse anti-human α-tubulin (1:5000 dilution, Sigma)and mouse anti-human β-actin (1:5000 dilution, Sigma) were used as primary antibodies.A horseradish peroxide (HRP)-conjugated antibody (1:5,000 dilution, Vector) was used asa secondary antibody. The signals were detected and visualized using an EnhancedChemiluminescence (ECL) detection kit (GE Healthcare Bio-Sciences, Piscataway, NJ,USA). Band intensity was quantified by densitometry using Image J software (NationalInstitutes of Health, Bethesda, MD, USA).

Co-immunoprecipitation assayThe binding of MRTF-A with SRF was detected using a co-immunoprecipitation assay.Briefly, the pre-cooled RIPA buffer containing protease inhibitors (Roche, Indianapolis,IN, USA) was used to lyse cells. The lysates of cells were precleared by incubation withprotein A cross-linked to agarose (Sigma), followed by incubation with an anti-MRTF-Aor SRF antibody. The immunocomplexes were collected by centrifugation and thenincubated with a specific antibody against SRF or MRTF-A (1:1000, Abcam). The blotswere subsequently incubated with an HRP-conjugated secondary antibody (1:2000,Sigma) and measured with the electro-chemiluminescence system (Amersham).

Chromatin Immunoprecipitation (ChIP) AssaysA ChIP experiment was performed using the EZ-ChIP kit (Millipore) according to themanufacturer ’s instructions. Briefly, fibroblasts were treated with freshly prepared 1%formaldehyde to crosslink proteins to DNA. The crosslinked chromatin was thensonicated and immunoprecipitated with 5 μg of an anti-SRF polyclonal antibody (G-20X, Santa Cruz) and a normal mouse IgG antibody, which was used as a negative control.After elution of protein/DNA complexes, the crosslink was reversed. Recovered andinput DNA was quantified by real-time PCR. The primers were designed using PrimerExpress Software (Applied Biosystems) as follows: α-SMA promoter sequence, 5′-AGTTTT GTG CTG AGG TCC CTA TAT G-3′and 5′-TTC CCAAAC AAG GAG CAAAGA-3′. KPNA2, whose promoter has no potential SRF binding sites, was used as a negativecontrol. KPNA2 promoter sequence: Forward: 5' -TCC CTC CCA TAG TAG CCA GA-3'. Reverse: 5' -GGC GAC AGC CTT AAA CAA AT -3'. Enrichment was calculated asthe percentage of immunoprecipitated DNA relative to the input.

Luciferase reporter assaysCells were plated on 6-well plates and co-transfected with 1.0 μg of α-SMA pro-Luc orempty (pGL3-Basic) plasmids together with 0.05 mg of Renilla luciferase (pRL-TK,Promega, Madison, WI, USA) using Lipofectamine 2000 (Life Technology, Carlsbad, CA,USA). After 48 hours, the cells were lysed in Passive Lysis Buffer (Promega). Luciferaseactivity was measured using the Dual-Luciferase Reporter Assay System (Promega) witha GloMax Luminometer (Promega) according to the manufacturer's instructions. Theresults were calculated as the ratio of firefly luciferase/Renilla luciferase activity. Theexperiments were performed in triplicate.

ELISASecreted HMGB1 in the culture medium was measured using a human HMGB1 ELISAKit (Shino-Test Corporation, Tokyo, Japan) according to the manufacturer’srecommendations (Millipore) with minor modifications. Briefly, 100 μl of samples wasincubated with HMGB1 in pre-coated 96-well plates. Purified recombinant HMGB1protein was used to generate the standard curve. Then, the plates were incubated with adetection antibody and an HRP-conjugated antibody successively. The signal wasvisualized by TMB (3,3′,5,5′-Tetramethylbenzidine) substrate, and the reaction wasterminated by 0.35 mol/L sulfuric acid. The optical density was read at 450 nm using amicroplate reader (Thermo Fisher Scientific).In addition, the LDH activity in the medium was determined by a specific commercial kit(Sigma) to normalize the expression of HMGB1. Briefly, 50 µl of each supernatant weretransferred to a 96-well plate, supplemented with 50 µl of reconstituted substrate mix,and then incubated in the dark at room temperature for 30 min. Then, 50 µl of an acidsolution was added to each well to stop the reaction. The absorbance of each solution was

read at 490 nm with a microplate spectrophotometer (Bio-Rad). The total LDH activityrelease was obtained by culturing cells in the presence of 1% Triton X-100, and thepercentage of LDH release was calculated. Then, secreted HMGB1 was normalized to theLDH activity in the culture medium.

Collagen synthesisTo quantify the collagen synthesis of fibroblasts in different conditions, [3H] prolineincorporation into acetic acid-soluble proteins was detected as described previously.Briefly, fibroblasts were seeded onto 12-well plates and grown to subconfluency. Afterstarving in serum-free medium overnight, the cells were subsequently exposed todifferent hydration conditions and reagents. Then, the cells were incubated with freshserum-free medium containing β-aminoproprionitrile, ascorbate and [3H]-proline, and thereaction was stopped 24 hours later with the addition of acetic acid. After incubation at4°C overnight, culture supernatants were collected and the protein precipitates werecollected by centrifugation, washed with NaCl and dissolved in acetic acid. [3H] prolineincorporation was determined by liquid scintillation counting in a scintillation counter(CanberraPackard, Meriden, CT, USA).

In vitro cell migration assayAn in vitro wound healing assay was performed to evaluate the migration rate offibroblasts in different conditions. 95% confluent cells that were grown on 12-well plateswere treated with 10 µg/ml mitomycin C at 37°C for 3 hours. A scratch was created inthe cell layer using a pipette tip. At least three non-overlapping fields in each scratchwere monitored, and the average width of each scratch was measured at 0, 8, 16 and 24hours.

In vivo study of the wounds that healed in different hydration conditionsFour female New Zealand white rabbits (weighing 3–4 kg) were used. Briefly, the hairof the ventral side of each rabbit ear was removed using an electric shaver followed bydepilatory cream application (Nair). Six 8-mm full-thickness punch biopsies werecollected down to the bare cartilage on the ventral side of the ear. In order to establishdifferent hydration conditions, the wounds that were treated with vaseline, and coveredwith breathable polyurethane dressing (Tegaderms) were used as hydration group. Thewounds that were directly exposed to air without any treatment were used as control(reduced hydration). Wounds were treated everyday, and the scar tissues were harvestedat day 28.

HistologyTissues harvested from rabbit ears were fixed with 4% paraformaldehyde overnight,followed by dehydration and embedding with paraffin. Serial 5-mm sections were cut andstained with hematoxylin and eosin (H&E) for observation of histological changes. The

images of histological slides of rabbit ear scars were photographed and analyzed using amicroscope (BX51 WI Olympus). For scar formation quantification, the scar elevationindex (SEI) and the ratio of newly formed collagen to old collagen were calculated.

Statistical analysisThe data in this study are expressed as the mean ± standard deviation (SD). The numberof independent replicates of every part of an experiment is represented using the letter ‘n’in the figure legend. The differences between experimental groups were compared usingthe paired Student’s t-test. All statistical analyses were performed using SPSS 17.0software (SPSS, Chicago, IL). P values lower than 0.05 were considered significantlydifferent.