Immunohistochemical analysis of Gottron’s biopsies of DM patients demonstrated a similar distribution of papillary dermal C4S as in all other DM skin lesions, but with a striking density exceeding that of samples from other locations. Compared to healthy control biopsies (Figure 1a), samples taken from non-Gottron’s DM lesions (Figure 1b) exhibited a mean 2.9-fold increased density of C4S. Compared to healthy control IP biopsies (Figure 1c), samples from Gottron’s (Figure 1d) exhibited a 4.6-fold increased density of C4S (both p<0.001). Gottron’s had a mean 1.6-fold C4S density compared to biopsies from other non-Gottron’s DM lesions (p<0.01). C4S was not present in the epidermis of any samples.

We next investigated the Gottron’s biopsies for dermal expression of proteins that interact with CS. Many proteins covalently or non-covalently attach to CS, but of these proteins, only the variant 7 (v7) isoform of CD44 has also been causally implicated in autoimmunity (^{Farkas et al., 2005}; ^{Hoffmann et al., 2007}; ^{Marhaba et al., 2003}; ^{Wittig et al., 2000}; ^{Wittig et al., 2002}). The v7 domain, as well as one region within the ‘standard’ CD44 (CD44s) that lacks any of the variant domains, act as covalent attachment sites for CS, while CD44v6-7 has also been shown to non-covalently bind CS (^{Keller et al., 2007}; ^{Sleeman et al., 1997}). Thus, we examined our samples for expression of CD44s and variant isoforms CD44v3, v6 and v7 (^{Kim and Werth, 2011}). Consistent with previous reports, we observed strong epidermal expression of CD44s, CD44v3, and CD44v6 in healthy control skin, particularly within the basal layer and stratum spinosum, but scant dermal matrix expression beyond staining of infiltrating leukocytes (^{Leigh et al., 1996}; ^{Seiter et al., 1998}; ^{Seiter et al., 1996}). Non-Gottron’s and Gottron’s DM lesions exhibited qualitatively stronger CD44s and CD44v3 staining within vessels compared to healthy controls; however dermal staining patterns of CD44s, CD44v3, and CD44v6 in healthy controls, non-Gottron’s and Gottron’s did not show significant differences (data not shown). We found no detectable CD44v6 in any dermal sample, despite abundant pericellular keratinocyte staining in the epidermis in the same slides, which served as an internal positive control.

In contrast, Gottron’s biopsies demonstrated a mean 4.5-fold denser accumulation of CD44v7 within the papillary dermal matrix compared to healthy control non-IP biopsies (p<0.001), and a mean 2.5-fold CD44v7 density compared to the dermis of non-Gottron’s lesions (p<0.01) (Figure 2). Non-Gottron’s lesions did not differ from controls and this differential expression was specific to CD44v7.

To study the effect of location versus disease in our patient samples, we acquired biopsies from the skin overlying the fourth proximal IP (PIP) joint of five healthy volunteers with no history of autoimmune disease. Immunohistochemistry of healthy IP skin was compared with Gottron’s samples to reveal a 4.5-fold greater density of C4S (p<0.001) and a 2.8-fold greater density of CD44v7 (p<0.01) in the IP dermis of patients with disease. For within-person comparisons, we also obtained a second biopsy from either the shoulder or the arm of two healthy volunteers. C4S density was not significantly different in the skin overlying the IP joint compared to non-IP skin of either volunteer. CD44v7 density in IP skin, however, demonstrated a modest but significant 1.6-fold (t-test p=0.0149) and 1.4-fold (t-test p=0.0307) increase compared to that in non-IP skin in the two volunteers, respectively.

Osteopontin is a known binding ligand for CD44v7, and this specific interaction has been implicated in chronic inflammation (^{Denhardt et al., 2001}). Osteopontin is present in the spinous and granular layers of normal epidermis, and around sweat glands, hair follicles, and sebaceous glands of the dermis (^{Chang et al., 2008}). Our examination of healthy control skin revealed similar patterns (Figure 3a). However, compared to location-matched healthy controls, biopsies taken from non-Gottron’s (Figure 3b) and Gottron’s (Figure 3c) exhibited increased accumulation of osteopontin diffusely distributed throughout the dermis. Quantitative analysis of dermal osteopontin revealed a mean 3.3-fold greater density in non-Gottron’s samples compared to healthy control samples (p<0.01) and a 2.5-fold osteopontin density in Gottron’s compared to healthy controls (p<0.05) (Figure 3d). Osteopontin in Gottron’s was not found to be significantly different from that in non-Gottron’s DM lesional skin, nor did osteopontin in healthy control IP skin differ from healthy control non-IP skin.

Because mechanical stretching has been reported to induce expression of CD44v7 and v8 in porcine ocular trabecular meshwork cells (^{Keller et al., 2007}), we hypothesized that the increased CD44v7 seen in the IP skin of DM patients and healthy volunteers is induced by chronic stretching of the skin over joints. Thus, we stretched primary dermal fibroblasts on cell culture membrane inserts and then determined CD44v7 mRNA expression by quantitative real-time PCR (RT-PCR). We found that mechanical stretching of dermal fibroblasts induced a significant 1.6-fold increase in CD44v7 mRNA expression at six hours (p<0.01), which returned to a baseline expression by twelve hours (Figure 4a). CD44v7 protein in the supernatant of stretched fibroblasts at 6 hours increased 3.3 ± 1.3 times relative to control fibroblasts (p<0.05) (Figure 4b,c). Osteopontin alone did not induce CD44v7 mRNA (data not shown). These results were reproducible in three repeat experiments. In contrast, mechanical stretching did not have a significant effect on fibroblastic osteopontin mRNA expression.

Interferon-γ has been reported to stimulate osteopontin expression in monocytes (^{Li et al., 2003}), but this induction has not been studied in dermal fibroblasts. Because interferon-γ is present in the skin of DM patients (^{Caproni et al., 2005}; ^{Wenzel et al., 2005}) and mutations in the interferon-γ pathway have been associated with DM (^{Chinoy et al., 2007}), we hypothesized a role for interferon-γ in the dysregulation of osteopontin expression in the skin. Treatment of human primary dermal fibroblasts with interferon-γ (10ng/mL) induced a 4.4-fold mean OPN mRNA expression compared to untreated control cells in vitro (p<0.001) (Figure 5a). Interferon-γ (10ng/mL) induced an 8.1-fold increase in OPN protein in fibroblast supernatant (p<0.001) (Figure 5b). Osteopontin mRNA and protein in the fibroblast supernatant were not induced by the pro-inflammatory cytokines IL-1α or interferon-α, indicating specificity (Figure 5a, 5b). Neither interferon-γ nor osteopontin treatment had a significant effect on fibroblast CD44v7 mRNA expression (data not shown).

Based on these findings, we hypothesized that the increased CD44v7 in association with its ligand osteopontin could play a crucial role in the localized inflammation of Gottron’s papules. To study the functional effect of increased CD44v7 ligation with osteopontin on immune cell recruitment, we incubated dermal fibroblast monolayers overnight, without or with stretching, without or with recombinant human osteopontin (rOPN), followed by rinsing. Fluorescently labeled THP-1 monocytes were then co-cultured with the prepared fibroblast monolayers, followed by rinsing to remove non-adherent monocytes, and the adherent monocytes were visualized with an inverted fluorescent microscope. We found that overnight incubation of fibroblast monolayers in medium supplemented with rOPN had no effect on THP-1 adhesion in the absence of stretch (Figure 6). Stretching the fibroblasts in the absence of rOPN supplementation resulted in an increase in the number of adherent THP-1 cells 1.8 times unstretched, no-rOPN controls (Figure 6, p<0.01). Stretching fibroblasts overnight in medium supplemented with rOPN resulted in an increase in THP-1 adhesion 2.2 times that of stretched fibroblasts without rOPN (p<0.001) and 2.3 times that of unstretched, no-OPN controls (p<0.001).

To demonstrate the specific effect of fibroblast CD44v7 on THP-1 adhesion, we incubated prepared fibroblast monolayers with anti-CD44v7 blocking antibody or isotype control IgGκ for one hour immediately before THP-1 co-culture. Stretched, rOPN-treated monolayers blocked with anti-CD44v7 antibody demonstrated a 41% reduction in THP-1 adhesion compared to identically prepared fibroblast monolayers incubated with control isotype IgGκ (Figure 6, p<0.001).

Lesional skin biopsies from patients with dermatomyositis (DM, n=10) were obtained in the autoimmune skin disease clinic at Penn. Five biopsies were taken from Gottron’s papules over the PIP or MCP joints and five unmatched biopsies were taken from active lesions on the arm, shoulder, or chest. Healthy control skin was provided by volunteers (n=10) without a history of autoimmune skin disease; five biopsies were from skin overlying the PIP of the third digit, and five biopsies were from sun-protected arm. Two volunteers provided both a PIP and arm or shoulder sample. All biopsies were obtained with written, informed patient consent and adherence to the Helsinki Guidelines according to our approved protocol under the UPenn Institutional Review Board. Biopsies were fixed in 10% neutral buffered formalin overnight, embedded in paraffin, and cut into 4-μm sections.

Immunostaining for chondroitin-4-sulfate was performed on human skin as described previously (^{Chang et al., 2011a}) by predigestion with chondroitinase AC II (Sigma, St. Louis MO), followed by incubation with an anti-deltaDi-4S monoclonal antibody that recognizes a terminal glucuronic acid residue adjacent to N-acetylgalactosamine-4-sulphate at the non-reducing end of chondroitin sulfate chains in proteoglycans produced after chondroitinase digestion of chondroitin sulphate chains (clone 2-B-6, Seikagaku Biobusiness Corporation, Japan). We also performed immunostaining using monoclonal antibodies against a fusion protein of glutathione S-transferase with individual variant domains (v3, v6, and v7) of human CD44 (AbD Serotec, Raleigh NC). Staining for CD44 v6 required heat-induced antigen retrieval, whereby slides were incubated for ten minutes at 70°C in Target Retrieval Solution (Dako, Carpinteria CA). Osteopontin was detected using a rabbit polyclonal antibody recognizing full-length as well as thrombin and MMP-cleaved osteopontin (AbCam, Cambridge MA). Sections were then incubated with biotinylated secondary antibody and streptavidin-horseradish peroxidase (Dako, Carpinteria CA), followed by Vector NovaRED substrate (Vector, Burlingame CA). Negative controls included mouse IgGκ isotype antibody (Sigma, St. Louis MO) and PBS in place of chondroitinase AC II enzyme digestion.

All high-power fields in three sections of each 3-5mm skin biopsy were examined at 10X magnification. Among all fields examined, one high-powered field appearing most representative of all sections was selected from each biopsy to measure matrix intensity, and this field was photographed at 10X magnification. An intensity channel was applied to photograph files using commercial software to isolate staining intensity (Image-Pro Plus v5.0). One-hundred 5 m circular manual tags were selected by systematic uniform random sampling of the area of interest (e.g. within the upper dermis for CD44v7) (Mayhew 2008). The mean intensity of staining within the total area selected by the manual tags is reported as the representative staining intensity value for each specimen.

Normal adult, human fibroblasts were obtained from the American Type Culture Collection (Rockville MD). Mechanical stretching experiments were performed by seeding primary human dermal fibroblasts onto membrane inserts (6-well format, 23.1-mm diameter, 3.0-μm pore size, BD Falcon #353091, BD Biosciences, Franklin Lakes, NJ) and culturing the cells to confluence (^{Keller et al., 2007}). Next, a smooth 5-mm glass bead was placed in the dish beneath the center of the insert membrane, and constant weight was applied to the lid of the insert to force the edge of the insert down onto the bottom of the dish. This produced distortion of the membrane, increasing its surface area and thereby mechanically stretching the cells and their extracellular matrix. Stretching was continued for 6 or 12 hours, and then each stretched membrane insert was cut out, vortexed in RNAeasy lysate buffer, and RNA purified, followed by quantitative PCR (q-PCR).

For cytokine stimulation, fibroblasts were plated in 100-mm diameter dishes and grown to confluence. Fibroblasts were treated for six hours with 10ng interferon-γ/mL (R&D Systems, Minneapolis, MN), 100U interferon-α/2 mL (PBL Interferon Source, Piscataway NJ), or 20ng recombinant human interleukin (IL)-1α/mL, before lysis with RNAeasy mini kit lysate buffer and processing RNA (Qiagen, Valencia CA). Experiments were performed in triplicate and repeated three times with reproducible results.

Quantification of specific mRNAs was performed as previously described (^{Chang et al., 2011a}). Total RNA was reverse transcribed using the SuperScript® First-Strand Synthesis System for RT-PCR (Invitrogen, Carlsbad CA) with random hexamer primers. Real-time PCR was performed using Taqman® Gene Expression Assays for human osteopontin (OPN) (Taqman Assay ID Hs00960942_m1) and PPIA (cyclophilin A), the housekeeping control (Taqman Assay ID Hs99999904_m1) (Applied Biosystems, Foster City CA). Custom-made sequences for detection of human CD44 exon 11 transcripts, which encode exclusively the v7 motif, were 5′-tgc aag gaa gga caa cac caa-3′ (sense primer) and 5′-ggg tgt gag att ggg ttg aag aaa t-3′ (antisense primer). All reactions were performed in triplicate. The Ct method was employed to analyze the differential gene expression in cytokine-treated fibroblast cultures compared to controls, as described previously (^{Chang et al., 2011b}).

Serum-free supernatant from control and stretched fibroblasts were collected after 6 hours. The supernatants were centrifuged at 10,000g for 1 minute. Equal amounts of protein (10 μg) were separated on a 4-12% bis-Tris gel (Invitrogen), transferred to a membrane (Millipore MA, USA), and stained with 1:2000 dilution of mouse anti-human CD44v7 (AbD Serotec UK). For loading control, the washed membrane was stained with 0.1% Coomassie R-350 (GE Healthcare). Western blot results were recorded by X-ray films (Kodak Scientific Imaging USA) and quantified by ImageJ (^{Collins, 2007}).

Conditioned media from interferon-γ, IL-1α, or interferon-α-treated cultured fibroblasts were collected and concentrated two-fold. The protein was quantitated by CB-Protein-Assay-Kit (G-Biosciences, St Louis, MO) and volume was adjusted to 100 ul, so that equal protein was used for each sample. Osteopontin protein was quantitated by commercial ELISA (R&D Systems Inc).

Primary dermal fibroblasts were cultured to confluence on 23.1-mm polyethylene terephthalate membrane inserts and then stretched, or not, with 5-mm glass beads. Here, stretching lasted overnight and was combined with treatment with 0 or 5.0μg of recombinant OPN/ml. Unbound OPN was rinsed with warmed PBS prior to the assay of monocyte adherence. Fibroblasts were maintained in serum-free, phenol red-free DMEM for the remainder of the assay. For inhibition studies, 25 g of mouse anti-human CD44 variant 7 (AbD Serotec, Raleigh NC) per mL or mouse anti-human IgGk isotype antibody control (Sigma, St. Louis MO) per mL was added to stretched monolayers after rinsing of unbound OPN for 1 hour, then washed. Living THP-1 monocytes labeled with 5-chloromethylfluorescein diacetate (CellTracker™ Green CMFDA, Invitrogen, Eugene, OR) were added (5×10⁵/insert), followed by a 30-min incubation at 37°C. Unbound monocytes were then removed by four rinses, and the number of monocytes firmly bound to the fibroblast monolayers was counted under fluorescent microscopy in 6 random high-power fields near the centers of the stretched membranes, where the stretching was the greatest. Experiments were repeated four times with reproducible results.

Comparison of several groups simultaneously was performed by initially using analysis of variance (ANOVA; GraphPad Prism v5.01). When ANOVA indicated differences amongst the groups, pairwise comparisons between experimental groups were performed using the Student-Newman-Keuls q statistic and are indicated in Figures 1^-6. Unless otherwise indicated, summary statistics are reported as mean ± SEM.