A bioengineered living cell construct activates metallothionein/zinc/MMP8 and inhibits TGFβ to stimulate remodeling of fibrotic venous leg ulcers

Abstract

Venous leg ulcers (VLU) represent a major clinical unmet need, impairing quality of life for millions worldwide. The bioengineered bilayered living cell construct (BLCC) is the only FDA-approved therapy demonstrating efficacy in healing chronic VLU, yet its in vivo mechanisms of action are not well understood. Previously, we reported a BLCC-mediated acute wounding response at the ulcer edge; in this study we elucidated the BLCC-specific effects on the epidermis-free ulcer bed. We conducted a randomized controlled clinical trial (ClinicalTrials.gov NCT01327937) enrolling 30 subjects with nonhealing VLUs, and performed genotyping, genomic profiling, and functional analysis on wound bed biopsies obtained at baseline and 1 week after treatment with BLCC plus compression or compression therapy (control). The VLU bed transcriptome featured processes of chronic inflammation and was strikingly enriched for fibrotic/fibrogenic pathways and gene networks. BLCC application decreased expression of profibrotic TGFß1 gene targets and increased levels of TGFß inhibitor decorin. Surprisingly, BLCC upregulated metallothioneins and fibroblast-derived MMP8 collagenase, and promoted endogenous release of MMP-activating zinc to stimulate antifibrotic remodeling, a novel mechanism of cutaneous wound healing. By activating a remodeling program in the quiescent VLU bed, BLCC application shifts nonhealing to healing phenotype. As VLU bed fibrosis correlates with poor clinical healing, findings from this study identify the chronic VLU as a fibrotic skin disease and are first to support the development and application of antifibrotic therapies as a successful treatment approach.

INTRODUCTION

Chronic venous leg ulcers (VLUs) are a widespread clinical challenge that burden the health care system, affecting over 1% of the population and incurring approximately $15 billion in associated costs annually.^1,2 Patients with VLUs suffer from diminished quality of life, chronic pain, physical disability, and increased mortality. The treatment of VLUs remains very challenging, as less than 75% of these ulcers achieve complete closure within 6 months of standard-ofcare compression therapy, and up to 25% of healed VLUs recur annually. Unfortunately, the development of novel treatments for nonhealing VLUs has been largely unsuccessful, as numerous therapies with promising results in the laboratory setting have failed to improve healing outcomes in clinical practice.³ In fact, the bioengineered bilayered living cell construct (BLCC) remains the only FDA-approved product over the past 20 years to have demonstrated efficacy in healing chronic VLUs.^4,5 The paucity of effective targeted therapies reflects an incomplete understanding of the essential biological processes that are deregulated in the VLU microenvironment and most directly impede physiologic wound healing. Moreover, there is a critical need to identify early biomarkers of therapeutic efficacy that will aid in selecting novel treatments that are most likely to succeed for downstream development in the clinical setting.

To this end, our group conducted a Phase 4 randomized clinical trial (ClinicalTrials.gov NCT01327937) designed to elucidate the mechanisms of action used by the BLCC to promote VLU closure. Using transcriptomic profiling of biopsies before and after treatment with a commercially available BLCC (Apligraf™, Organogenesis, Inc.), we previously described an inflammatory response at the ulcer edge that was distinct from the chronic inflammation present in nonhealing VLUs at baseline. Specifically, we found that a single application of the BLCC successfully recapitulated features of the acute wound repair phenotype, converting a nonhealing to a healing response within the chronic VLU microenvironment.⁶ To comprehensively characterize the BLCC’s in vivo mechanisms of action, our clinical trial design also incorporated collection of biopsies from the epidermis-free center (ulcer bed) of the chronic VLU before and 1 week after BLCC application, used to generate the VLU bed transcriptome presented in this study. A healthy extracellular matrix (ECM) dynamically participates in regulating chemical and mechanical signals important to successful wound healing.² In nonhealing VLU, the ulcer bed is characterized by disorganized ECM, marked fibrosis, and chronic inflammatory infiltrates,^7–9 all of which contribute to impaired healing. In fact, the presence of dense fibrosis and increased collagen deposition in chronic VLUs are independent predictors of a clinical nonhealing outcome.⁹ Chronic inflammation is itself a powerful perpetuator of fibrogenesis, and converging lines of evidence in other organs suggest that fibrosis is reversible once the underlying inflammatory trigger is removed.¹⁰ We therefore hypothesized and found that BLCC application, which attenuates chronic inflammation at the ulcer edge,⁶ stimulated remodeling of the ulcer bed to reverse the fibrosis that impedes VLU healing, cohesively promoting successful wound closure.

We discovered that the VLU bed is characterized by pathologic inflammation (that includes both innate and adaptive immunity) and fibrosis evident at the genomic level resulting from upregulation of many profibrotic factors, fibrogenic myofibroblasts and canonical TGFß signaling. We found that a single BLCC treatment triggers a shift from chronic inflammation toward an acute (pro-healing) wounding response, which is coordinately associated with reversal of fibrosis in the ulcer bed. Furthermore, we discovered a novel BLCC-regulated antifibrotic remodeling of the VLU bed that involves induction of a series of metallothioneins, in vivo increase in zinc levels and increase of zinc-dependent MMP8 collagenase in fibroblasts, facilitating remodeling of the ulcer bed to promote healing in VLU.

MATERIALS AND METHODS

Experimental design

Patients presenting to the wound clinic at the University of Miami with VLUs were recruited for the study (ClinicalTrials.gov NCT01327937). Written informed consent was obtained from all participants prior to inclusion in the study. The study was approved by the Institutional Review Board of the University of Miami. To identify nonhealing VLUs for enrollment, subjects first underwent screening for 28 days during which they received standard-of-care compression therapy consisting of a foam dressing and a four-layered compression bandage system. Exclusion criteria were: subjects who require vacuumassisted closure therapy on or after Week 0 study visit; subjects with arterial disease as determined by an Ankle Brachial Index measurement of <0.65; subjects with any systemic or congenital condition like uncontrolled diabetes mellitus, positive HIV test, or any disorder(s) that may compromise wound healing; subjects with carcinomas located at the target ulcer with biopsy confirmed active malignancy (subjects with other carcinoma locations were not excluded from entry into the study); subjects receiving, or have received at any time within 30 days prior to screening visit, noninhaled corticosteroids except topical steroids not at the target ulcer (inhaled steroid therapy is acceptable on study), immunosuppressive agents, radiation therapy, hemodialysis, peritoneal dialysis, or chemotherapy, anticipated use of the above agents or therapies excludes subject from entry into the study; clinical vasculitis, severe rheumatoid arthritis, and other collagen vascular diseases; signs and symptoms of cellulitis or osteomyelitis at the target ulcer; avascular target ulcer beds and ulcers of mixed etiology such as arterial disease with VLU were excluded; target ulcer with exposed bone, tendon, or fascia; known hypersensitivity to bovine collagen or to the components of the BLCC agarose shipping medium; subject enrolled in any wound or investigational study (drug, biologic, or device) for any disease within 30 days of the screening visit; subject previously treated with Apligraf, Dermagraft or any other cell therapy at the target ulcer site within 30 days of the screening visit; subject with a history of alcohol or substance abuse within the previous year, which could interfere with study compliance such as inability to attend scheduled study visits; subjects who are a current smoker or has not ceased smoking 6 months prior to the screening visit, or in the opinion of the Investigator, has a smoking history that may compromise wound healing; subject who, in the opinion of the Investigator, for any reason other than those listed above, will not be able to complete the study per protocol; target ulcer has decreased in size by ≥40% from screening to Week 0; confirmed gene expression overlap between the subject’s cells (buccal swab) and the cells contained in BLCC. At the conclusion of the screening period, the following inclusion criteria were used: participants aged 18 or older with ulcers of >5 cm² that had not decreased in area by >40%, with no clinical signs of infection, were randomized or assigned to either the control group (n = 8) receiving standardof-care compression therapy changed weekly by the Investigator, or to the treatment group (n = 11) receiving weekly BLCC applications (Organogenesis, Inc., Canton, MA) along with compression therapy. The BLCC was fenestrated prior to application using a #11 blade with six fenestrations per 44 cm². 3 mm biopsy specimens were obtained from the center of the nonhealing VLU bed at the time of randomization as well as on Day 7 of treatment (Week 0 and Week 1, respectively). Prior to biopsy, ulcers were debrided and local lidocaine injection was used for anesthesia. Patients were monitored weekly in the University of Miami wound clinic, where VLU size and appearance were recorded for 12 weeks or until wound closure was achieved. Full details of the clinical trial design and execution, including outcome measures and inclusion/exclusion criteria, are as previously described.⁶ See also Supporting Information Tables S1 and S2 for subject demographics and inclusion in analysis of primary outcome regarding effects on the VLU bed.

Sample processing

Samples were embedded in OCT compound (Fisher Scientific, Waltham, MA) and/or stored in formalin for paraffin embedding, stored in RNAlater (Ambion/Applied Biosystems, Foster City, CA) for RNA isolation and/or snap-frozen for protein isolation. To confirm that the tissue specimens obtained from the BLCC treatment group exclusively contained VLU wound tissue and not residual cells of the BLCC construct, genotyping of a portion of each Week 1 biopsy in the BLCC treatment group was performed by an outside laboratory (Esoterix Clinical Trials Services, Burlington, NC).

Biopsies obtained from VLU wound bed were formalinfixed and paraffin-embedded. Five-micrometer-thick sections were stained using hematoxylin & eosin following standard protocol and assessed for the presence of dermis to confirm characteristic VLU dermal morphology.

RNA isolation

Following tissue homogenization, RNA isolation and purification was performed using miRNeasy MiniKit (Qiagen Inc., Valencia, CA) with incorporation of on-column DNase treatment as per manufacturer’s protocol. RNA was quantified with Nanodrop and integrity was assessed with the 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Patients for which the RNA Integrity Number of Week 0 and Week 1 VLU wound bed biopsies exceeded the preset threshold of 4 were submitted for microarray analysis. In total, paired microarray-quality biopsies were obtained for five controls and seven BLCC-treated patients.

Gene expression microarrays

Microarray experiments were performed at the University of Miami’s Hussman Institute for Human Genomics Core Facility. RNA was amplified, fragmented and hybridized to arrays using GeneChip 3’ IVT Express kit following manufacturer’s protocol (Affymetrix, Santa Clara, CA). 100 ng of total RNA were used as input for the Ambion WT Expression Kit (Ambion, Austin, TX) to produce labeled single-stranded cDNA. Labeled products were hybridized to Affymetrix Human Genome U133 Plus 2.0 Arrays. (Affymetrix). The Fluidics 450 station, GeneChip® Operating Software and GeneChip® Scanner 3,000 7G (Affymetrix) were used to perform staining, washing and scanning of the arrays. For quality control, cRNA, cDNA, and fragmentation products were analyzed using the Nano 6,000 Bioanalyzer (Agilent).

Microarray data analysis

Microarrays of healthy dermis immediately after epidermal removal during split-thickness grafting (n = 6)¹¹ were downloaded from NCBI’s Gene Expression Omnibus, accession number GSE28914. Microarray files from healthy dermis and from BLCC and control-treated VLUs were imported into Genespring 13.0 (Agilent Technologies), normalized using the GC Robust Multiarray Average (GCRMA) algorithm, and filtered for quality control, using Genespring’s Biological Analysis Workflow. The processed data were then compared by paired t test (VLU bed before vs. after treatment) or moderated t test with Benjamini-Hochberg correction (VLU bed vs. healthy dermis). Fold change thresholds of >2 for individual genes and > 1.5 for pathway analysis were applied.

Acute dermal profiles

Microarray profiles of acutely injured dermis from healthy donors immediately postepidermal stripping were downloaded from data deposited in the NCBI Gene Expression Omnibus database (www.ncbi.nlm.nih.gov/geo) under GEO Accession Series GSE28914. Donor demographics and details of sample collection and processing were previously described.¹¹ Raw (.CEL) microarray files were normalized and summarized using the GCRMA algorithm along with the microarray data obtained from chronic VLUs in this clinical trial. Microarray files utilized to establish the VLU bed transcriptome are listed in Supporting Information Table S3.

QPCR validation

RNA reverse transcription to cDNA was performed using QScript cDNA synthesis kit (Quanta Biosciences, Gaithersburg, MD). QPCR reactions were performed using Perfecta SYBR Green Supermix (Quanta Biosciences) on BioRad CFX instrument, using default settings. Gene expression was normalized to ARPC2 internal control housekeeping gene. Fold change was calculated using the delta–delta Ct method. For all QPCR validation studies, technical triplicates were included. Primer sequences for the panel of metallothionein transcripts were obtained from a previous publication.¹² ARPC2 primer sequences were (5′to3′): Forward- TCCGGGACTACCTGCACTAC, Reverse- GGTTCAGCACCTTGAGGAAG. TGFB2 primer sequences were (5^′ to 3′): Forward- AAAGCCAGAGTGCCTGAACA, Reverse- AGCGCTGGGTTGGAGATG. TIMP3 primer sequences were (5′ to 3′): Forward- GTGCAACTTCGTGGAGAGGT, Reverse- CAGGTAGTAGCAGGACT TGATCTTG.

Western blotting

Samples were homogenized and 25 μg of protein per sample was resolved on 4–20% Criterion TGX precast gels (BioRad, Hercules, CA) and transferred onto PVDF membranes (BioRad). Membranes were blocked with I-Block (Applied Biosystems) in PBS/ 0.1% Tween-20, and probed with antimetallothionein (Abcam, San Francisco, CA; Cat. No. ab12228), anti-MMP8 (Abcam Cat. No. ab81286), or antidecorin (Santa Cruz Biotechnology; Cat. No. sc-22,753). Membranes were incubated with horseradish peroxidaseconjugated secondary antibodies (Cell Signaling, Danvers, MA) and developed using an ECL Prime chemiluminescence detection system according to manufacturer’s protocol (GE Healthcare, Aurora, OH). For loading controls, anti- β-Actin and anti-ARPC2 antibodies were used (Sigma Aldrich, St. Louis, MO). Gels were scanned and bands were quantified using Gel Doc (Biorad) and the ImageJ software.

Immunofluorescence

5–7 μm thick sections of paraffin-embedded specimens were de-paraffinized in xylene, rehydrated in alcohol, and washed with phosphate-buffered saline (PBS) (Fisher Scientific). Antigen retrieval was performed at 95°C with sodium citrate (Sigma). Tissue sections were blocked for 1 hour with 5% bovine serum albumin (Sigma) and incubated with primary antibodies overnight. Primary antibodies used were as follows: anti-IBSP (rabbit polyclonal; 1:50 dilution; Abcam Cat. No. ab52128), antimetallothionein (mouse monoclonal; 1:00 dilution; Abcam Cat. No. ab12228), anti-CD45 (mouse monoclonal; 1:100 dilution; DAKO Cat. No. M0701), anti- MMP8 (rabbit monoclonal; 1:00 dilution; Abcam Cat. No. ab81286), and anti-αSMA (mouse monoclonal; 1:250 dilution; Sigma Aldrich Cat. No. A5228). Slides were incubated with FITC or TRITC-labeled secondary antibodies (1:500 dilution; Invitrogen, Carlsbad, CA) for 1 hour at room temperature and mounted with mounting media containing propidium iodide to visualize cell nuclei. Sections were viewed using a Nikon Eclipse E400 microscope and digital images were collected with QImaging camera and the NIS Elements BR 3.1 software.

Zinc quantification

Zinc quantification was performed on protein extracted from snap-frozen tissue using a colorimetric assay kit (Abcam, ab102507) per manufacturer’s instructions; absorption was measured at 560 nm.

Pathway analysis

Gene set enrichment analysis was performed using Genespring’s Gene Ontology tool. Additional pathway analysis and network generation were performed using Ingenuity Pathway Analysis (IPA; Qiagen; www.ingenuity.com).

Statistical analysis

For comparison of VLU subjects’ response to treatment, a ratio paired t test was used. For comparisons of control and BLCC-treated groups, an upaired t test (if data were normally distributed) or Mann–Whitney U test (if data were not normally distributed) was used. For all analyses, p-values of less than 0.05 were considered statistically significant. Statistics for microarray data were performed as described under “Microarray Data Analysis.” Statistical tools within the IPA software package employed Fisher’s Exact Test to detect the reported significantly enriched pathways, biologic processes, and upstream regulators; in all cases, enrichment p-values were Benjamini-Hochbergcorrected for multiple testing. Gene ontology enrichment p-values were calculated within the Genespring 13.0 software package which utilizes Broad Institute’s Gene Set Enrichment Analysis algorithms. Pathway and upstream regulator statistics were calculated within the Ingenuity software package, as follows: Enrichment p-values were calculated using Fisher’s exact test with Benjamini-Hochberg correction for multiple testing. Pathway activation was assessed by a Z-score calculation reflecting consistency of directionality of gene expression changes observed after treatment, as compared with directionality of relationship reported in the scientific literature. Statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software) unless otherwise stated.

RESULTS

The chronic VLU bed transcriptome contains hallmarks of fibrosis

Patients with nonhealing VLUs were randomized to receive compression therapy alone or to compression plus BLCC (Figure 1, Supporting Information Table S1) as previously described.⁶ Punch biopsies were obtained from the epidermis-free ulcer center (ulcer bed) at baseline (Week 0) and after one week of treatment (Week 1). To exclude the possibility that BLCC (as opposed to wound tissue) was evaluated, DNA genotypes from Week 1 biopsies were examined and compared to cellular genotypes of the BLCC used, as previously described.⁶ DNA genotyping excluded the presence of BLCC donor cells in Week 1 patient biopsies. Specimens suitable for genomic analysis at both time points were subjected to transcriptomic profiling by microarray followed by comprehensive pathway analysis (Supporting Information Table S2).

Figure 1.

CONSORT diagram. Patients included in VLU bed study as part of ClinicalTrials.gov NCT01327937.

In contrast to the nonhealing VLU edge, the transcriptome of the epidermis-free ulcer bed is not well- characterized. To establish a baseline for our posttreatment profiles, we compared VLU bed profiles with the study participants at Week 0 with dermal profiles following acute dermal wounding¹¹ (Supporting Information Table S3). In line with the prominent histopathologic features of chronic inflammatory infiltrates and fibrosis evident in VLU bed specimens,^7–9 Ingenuity™ pathway analysis (IPA) of >4,000 differentially expressed genes identified highly enriched and activated processes of inflammatory response (p = 3.95 × 10⁻²¹, activation score + 3.79) and fibrogenesis (p = 1.15 × 10⁻¹¹, activation score + 2.96) in the chronic ulcer bed (Supporting Information Figure S1). The activated inflammatory response consisted of genes involved in signaling pathways of both innate and adaptive immunity (Figure 2A). The presence of fibrosis was underscored by an enriched network of collagens and secreted matricellular proteins in conjunction with upregulation of many profibrotic factors (Figure 2B), including fibronectin (FN1), tenascin (TNC), osteopontin (SPP1), connective tissue and hepatocyte growth factors (CTGF, HGF), and PAI-1 (SERPINE1).^13–17 In addition, α-SMA (alpha-smooth muscle actin; ACTA2), a marker of fibrogenic myofibroblasts,¹⁸ was upregulated ~18-fold, which correlated with highly positive staining in VLU bed sections in vivo (Figure 2C). Global enrichment (P = 6.51 × 10⁻⁴⁷) and activation (Z = +5.98) of the upstream regulator TGFB1 was supported by >250 differentially expressed genes in the nonhealing VLU bed (Supporting Information Table S4), several of which participate in profibrotic canonical TGFß signaling¹⁹ (Figure 2D). Interestingly, IPA Toxicity analysis identified pathologic processes of hepatic fibrosis and liver damage among the regulated genes in the chronic ulcer bed (Figure 2E).

Figure 2.

Inflammation and fibrosis in the chronic VLU bed transcriptome. (A) Gene networks of innate and adaptive immunity and(B) altered collagens and upregulated profibrotic factors present in nonhealing wound bed specimens from VLU patients (n = 6) in comparison with dermis from healthy donors immediately postwounding (n = 6). (C) Immunofluorescence staining of alpha-smooth muscle actin (red) with overlying DAPI staining of nuclei (blue) in nonhealing VLU bed sections. Image is representative of n = 4 study subjects. (D) Upregulated TGFB signaling pathway and a subset of target genes involved in fibrotic processes in the chronic VLU bed. Fold expression change over healthy donor dermis is indicated under each gene. (E) IPA Toxicity analysis. Percent overlap of VLU bed genes with IPA toxicity related lists. Overlap p-values were calculated using Fisher’s exact test.

Genomic profiles of BLCC-treated VLU shift toward acutely injured dermis

We next asked whether the pathologic inflammation and fibrosis evident at the gene expression level in the nonhealing VLU bed were affected by BLCC application. Analysis of Week 0 and Week 1 microarray profiles identified 72 genes uniquely regulated in BLCC-treated VLU in comparison with VLU treated with compression therapy alone; an additional three genes were regulated in opposite directions in the BLCC and Control groups (Supporting Information Figure S2 and Table S5). Among these we identified a subset of genes which, in line with the trend observed at the BLCC-treated VLU edge,⁶ shifted expression toward that of the acutely injured dermis (Supporting Information Figure S3A–B). Similar shifts in gene expression recapitulated the acute dermal phenotype at the level of wound-healing processes, with predicted pro-healing effects on epithelial proliferation, cytoskeletal formation, apoptosis, and others (Supporting Information Figure S3C).

BLCC application triggers inhibition of TGFβ signaling associated with reversal of fibrosis

We postulated that the BLCC, by triggering a shift from chronic inflammation toward an acute (pro-healing) wounding response, is coordinately associated with reversal of fibrogenic pathways in the ulcer bed. To test this, we focused on TGFβ, the prototypical profibrotic signaling pathway with evidence of activation in the nonhealing VLU bed (Figure 2D). Following BLCC treatment, TGFB1 shifted to a predicted state of inhibition (Z =−2.6), supported by expression changes in 18 downstream genes (Figure 3A), including downregulation of TGFB2 (Figure 3B). Another target, tenascin C, an ECM signaling molecule driving organ fibrosis,^13,20 which had displayed 24-fold-elevated levels in the nonhealing ulcer bed pretreatment, significantly decreased in BLCC-treated VLU (Figure 3C). We also confirmed BLCC-triggered downregulation of TIMP3, another TGFB1 target and inhibitor of ECM remodeling²¹ (Figure 3D). We further found that protein expression of the leucine-rich proteoglycan decorin, a natural inhibitor of TGFB1 that directly interacts to abrogate its profibrotic signaling,²² was induced post-BLCC treatment (Figure 3E). Finally, we performed unbiased principal component analysis on a set of TGFB superfamily genes, the majority of which did not individually meet statistical thresholds for differential regulation in our microarrays, and identified a cumulative shift in expression after BLCC application (Supporting Information Figure S4), further supporting modulation of TGFB signaling by BLCC.

Figure 3.

BLCC inhibits TGFB signaling in the chronic VLU bed. (A) IPA upstream regulator network identifying targets of TGFB1 that are up- or downregulated in BLCC-treated VLUs supporting a predicted inhibited state of TGFB1, as supported by IPAcalculated Z-score. (B) QPCR of TGFB2 expression in n = 7 pairs of chronic VLU bed specimens at baseline (Week 0, “WO”) and one week post-BLCC treatment (Week 1, “W1”). ** = P < 0.005 by ratio paired t test. (C) Box- and- whisker plots of tenascin C microarray probe expression intensity in the healthy donor dermis immediately postwounding (“Acute dermis,” n = 6) and in seven pairs of VLU before and 1 week after BLCC treatment (“W0” and “W1”). (D) QPCR of TIMP3 expression in n = 7 VLU before and after BLCC treatment. Mean ± SEM are shown; * = p < 0.05 by ratio paired t test. (E) Western blot and quantification of decorin expression in n = 7 VLC before and after BLCC treatment. Normalization was performed to ARPC2 expression. Men ± SEM are show; * = p < 0.05 by ratio paired t test.

BLCC activates ECM remodeling

We further examined our profiles for evidence of novel BLCC-regulated antifibrotic remodeling of the VLU bed. The most highly regulated BLCC gene was integrin-binding sialoprotein 2 (IBSP), a structural bone ECM protein with regenerative activity²³ (with no previously described role in fibrosis), which was induced in response to BLCC but was not regulated in controls treated with compression therapy alone (Figure 4A and B). At the pathway level, gene ontology analysis further supported BLCC-mediated effects on ECM remodeling processes, with enrichment of “extracellular matrix” (12 genes; p = 6.6 × 10⁻⁵), “regulation of growth” (10 genes; p = 1.2 × 10⁻²), “regulation of cell adhesion” (7 genes; p = 1.6 × 10⁻²), “response to wounding” (10 genes; p = 2.5 × 10⁻²), and “regulation of proteolysis” (10 genes; p = 2.6 × 10⁻²) processes (Figure 4C). Unexpectedly, we observed highly significant enrichment of the “response to zinc ion” process (p = 1.4 × 10⁻⁹) which involved induction of a series of metallothionein transcripts following BLCC application.

Figure 4.

Extracellular matrix remodeling response to BLCC application. (A) Integrin-binding sialoprotein (IBSP) expression by microarray in chronic VLU patients (paired Week 1 vs. Week 0 biopsies) treated with standard-of-care compression (Control, n = 5) and BLCC (n = 7). ** = p < 0.005 by unpaired two-tailed t test with Welch’s correction. Dotted line denotes fold change threshold of 2. (B) Immunofluorescence of IBSP protein (red) in ulcer bed sections before and after BLCC treatment (Week 0 and Week 1, respectively). Images are representative of findings in n = 5 paired BLCC-treated VLUs. (C) Gene ontology analysis of enriched biological processes among genes displaying fold change >1.5 one week post- treatment with BLCC. Select overrepresented functions are listed in columns, and corresponding BLCC-regulated genes are shaded in gray.

Induction of metallothioneins, zinc release, and MMP8 collagenase following BLCC treatment

Metallothioneins act as modulators of local cellular zinc supply to regulate the activities of zinc-dependent proteins, including matrix metalloproteinases (MMP).²⁴ Using QPCR for a panel of alternatively spliced metallothionein transcripts,¹² we validated our microarray findings in an expanded pool of pre- and posttreatment VLU biopsies from eight control and eight BLCC-treated subjects. We confirmed induction of MT1H, MT1X, and MT2A transcripts in the BLCC-treated group (Figure 5A), which correlated with increased metallothionein protein expression in the VLU bed (Figure 5B and C). Moreover, we identified in vivo increase in zinc levels following BLCC application (Figure 5D).

Figure 5.

BLCC induces metallothionins, zinc release, and MMP8 expression in the VLU bed. (A) QPCR of MT1H, MT1X, and MT2A metallothioneins in n = 8 BLCC-treated VLU as compared with n = 8 control-treated VLU. * = p < 0.05 by Mann–Whitney U-test. (B) Western blot of metallothionein expression in chronic VLU before and after BLCC treatment. Band intensities of type 1 metallothionein proteins (6–20 kDa) were quantified relative to expression at Week 0. (C) Metallothionein expression by immunofluorescence (green) in VLU bed sections and DAPI staining of nuclei (blue). Images are representative of n = 5 BLCC-treated VLU patients. (D) Zinc quantification (absorbance at 560 nm) in n = 4 VLU before and after BLCC treatment. * = p < 0.05 by paired t test. (E) Western blotting of MMP8 expression (MW 70 kDa) post-BLCC treatment; images representative of n = 5 BLCC-treated and n = 3 control-treated VLU. * = p < 0.05 by unpaired t test with Welch’s correction. (F) Immunofluorescence staining of MMP8 (red) and CD45 (green) expression. Images representative of n = 5 BLCC-treated VLU.

Of the zinc-dependent MMPs, MMP8 collagenase was most highly upregulated in the VLU bed following BLCC application (Supporting Information Figure S2B). Consistent with transcriptional profiles, MMP8 protein expression markedly increased in BLCC but not control (standard-of-care compression)-treated VLUs (Figure 5E). Since MMP8 is traditionally produced, stored, and secreted by neutrophils, we performed dual immunofluorescence for the pan-leukocytic CD45 surface marker (Figure 5F). Surprisingly, we found that MMP8 did not localize with CD45-positive cells, and instead was expressed by fibroblasts situated among fibrotic bundles of collagen and surrounded by the CD45+ inflammatory infiltrate (Supporting Information Figure S5).

DISCUSSION

Our study reports the first Phase 4 clinical trial designed to investigate the mechanisms of action of a clinically efficacious cell-based therapy (BLCC) in promoting the healing of chronic VLUs. In this report, we identify transcriptional changes in the VLU bed that constitute a BLCC application response, which is collectively characterized by antifibrotic, pro-remodeling effects (Figure 6). These findings are in line with clinical observations that the presence of dense fibrosis correlates with poor VLU healing outcomes,⁹ though targeting fibrosis represents a new paradigm in the treatment of chronic VLUs. Our study thus supports the development of products promoting global remodeling of the VLU bed and further provides specific alterations in gene expression and shifts in profibrotic pathway activities that may be utilized as early surrogate markers of therapeutic efficacy in vivo. This is of utmost importance to the chronic wound-healing field, in which the development of novel therapies has been impeded by preclinical models that do not adequately recapitulate the chronic ulcer microenvironment.²⁵ Indeed, no new treatments have met FDA endpoints for clinical efficacy since the BLCC was introduced 20 years ago.

Figure 6.

Model of antifibrotic BLCC effects on the nonhealing VLU bed to promote wound closure. The chronic VLU bed is characterized by unresolved inflammation which perpetuates fibrogenesis, as evidenced at the transcriptional level by activated TGFβ signaling and elevated markers of fibrosis including tenascin C. BLCC application reverses chronic inflammation and decreases fibrosis of the VLU bed by dampening TGFβ signaling and by stimulating matrix remodeling through activation of zinc-dependent collagenases and inhibition of TIMPs.

In this work, we characterized the transcriptome of the nonhealing VLU bed and identified the presence of activated TGFβ signaling there. TGFβ is the prototypic driver of fibrosis in many organs and directly and indirectly promotes fibrosis by activating and sustaining myofibroblasts, stimulating excessive ECM synthesis, and inhibiting ECM degradation and remodeling.¹⁹ We further found that treatment with the BLCC (but not with compression therapy alone) produced a shift of TGFB1 signaling to an inhibited state as evidenced by changes in its downstream targets. Tenascin C, a TGFB target that increases ECM production by stimulating myofibroblasts and maintaining the pro-inflammatory (and thus profibrotic) milieu,²⁰ was highly upregulated in the chronic VLU bed at baseline and was significantly reversed (downregulated) in response to BLCC application. Interestingly, it is a detectable biomarker in serum, tissue, and blood in other fibrotic diseases, including those of the skin,^13,20 and may thus function as an indicator of BLCC activity. BLCC application also modulated TGFB1 target genes involved in antifibrotic remodeling, including protease inhibitor TIMP3. We confirmed TIMP3 downregulation following BLCC application, permitting MMP-mediated matrix proteolysis.²¹ Finally, we found that BLCC application induced transcript and protein levels of decorin, a powerful natural inhibitor of TGFβ signaling that binds to TGFβ and physically prevents its interaction with fibrogenic receptors.²² Endogenous decorin plays a role in acute wound healing²⁶ and also retards fibrosis in other tissues²²; it is a therapy under consideration for postburn hypertrophic scarring.²⁷

Beyond its effects on TGFβ signaling, BLCC application unexpectedly triggered a metallothionein-driven cascade predicted to promote ECM remodeling. By chaperoning the zinc ion, metallothioneins have a demonstrated role in tissue repair, as they regulate the local availability of zinc for critical wound-healing processes.^28,29 Downregulation of MT1H, MT1X and MT2A in dermal fibroblasts contributes to fibrotic keloid scar formation,³⁰ while increased MT2A levels accelerates postburn wound healing.^29,31 In BLCC-treated VLU specimens, metallothionein induction was accompanied by endogenous release of zinc which was detectable in vivo. The concept that an allogeneic cell therapy can mediate local zinc levels may have relevant biological implications.

Effective wound closure requires the controlled activity of collagenolytic MMPs,³² and MMP8 is the predominant collagenase in acutely healing wounds.³³ It is classically secreted by neutrophils in response to inflammatory stimuli³⁴ and has been reported as increased in chronic ulcer tissue and wound fluid.^33,35 MMP8 was also found increased in in blood monocytes, lung macrophages, and airway epithelial cells from patients with idiopathic pulmonary fibrosis,³⁶ but not in the fibroblasts. We did detect some basal MMP8 protein expression in chronic VLUs at baseline (pretreatment), which was significantly induced by BLCC application but not by compression therapy alone. Unexpectedly, though, the BLCC-stimulated MMP8 was not produced by neutrophils, but rather by ulcer bed fibroblasts. As fibroblasts have been shown to produce MMPs in other contexts,³⁷ our data support the notion that MMP8 may have distinct biological functions that may depend of the cell source, whereby fibroblast-derived MMP8 may have beneficial antifibrotic effects. We also recognize the limitation of our study—the MMP8 activity in the tissue was not assessed due to limited amount of biomaterial. Taken together, though, our findings support a novel BLCC-triggered metallothionein/zinc/MMP8 cascade in the chronic VLU bed with possible antifibrotic impact.

Pathologic fibrogenesis is remarkably similar at the cellular level across different organs. IPA analysis initially identified a transcriptome-level parallel between hepatic fibrosis and fibrosis of the chronic VLU bed at baseline (pretreatment). The liver responds to injury with a classical wound-healing response of inflammation followed by remodeling, which progresses to fibrosis (cirrhosis) when the inflammatory injury signals are sustained. But as a regenerative organ, the liver also has a remarkable capacity for reversal of fibrosis.¹⁰ It was therefore striking that we observed further parallels between the antifibrotic BLCC mechanisms identified in this study and those described in models of regressing liver fibrosis. These included anti-TGFβ approaches, stimulation of endogenous decorin,²² fibrosis-attenuating deficiency of tenascin C,³⁸ and MMP8 reversal of cirrhosis.³⁹ Finally, expression of metallothioneins has been shown to reverse the fibrotic liver phenotype, specifically through upregulation of MMP collagenases.⁴⁰

Inflammation and fibrosis are linked almost universally not only in the liver but in other models of organ fibrosis, in which a sustained immune response to injury perpetuates fibrogenesis, which may be reversible upon removal of the chronic inflammatory signals.¹⁰ This clinical trial identified a BLCC-triggered shift of chronic inflammation to acute pro-healing inflammatory wound phenotype that predominated at the VLU edge⁶ and was accompanied by initiation of an antifibrotic gene program in the ulcer bed. The combined modulation of inflammation and fibrosis by the BLCC may account for the unique success in contrast to other VLU treatments which have failed to demonstrate efficacy. To this end, it is likely that recapitulating individual elements of the BLCC’s effects, for instance TGFß inhibition alone, will not be sufficient to promote VLU healing. Furthermore, the concept that two distinct locations (edge and bed) within a nonhealing wound have unique properties and specific response to treatment has practical implications. It highlights the importance of examining both the ulcer edge and bed in future clinical trials, as the responses in each location are quite genomically distinct. Moreover, it underscores the complexity of chronic wound environment and cell-type specific properties and cross-talk required to achieve successful wound closure.

There are some limitations to this clinical trial. First, we cannot determine whether the effects of BLCC treatment are direct signals from the allogeneic cells of the construct or an indirect consequence of its placement on endogenous factors present in the chronic VLU. The BLCC construct disappears within 1–2 weeks, yet its clinical efficacy is evident months after application,⁴ suggesting that irrespective of whether its initial influences were direct, it must affect a sustained response that is perpetuated by resident VLU cells to lead to wound closure. Second, the lack of animal models that accurately recapitulate the complex features of chronic VLUs found in patients limits our ability to validate our findings at the mechanistic level.^25,41 Third, limitations in our clinical trial design did not allow for prospective validation of findings. Instead, future studies are needed to track and correlate the genomic indicators of the pro-healing, antifibrotic response identified in the VLU bed in this trial with clinical progression and outcomes in VLU, for their utilization as biomarkers during the process of VLU closure.

In summary, data from our clinical trial provide new insights regarding molecular pathophysiology of the ulcer bed and how and why the BLCC is efficacious in healing chronic VLUs. Specifically, we have identified universal markers of enhanced matrix remodeling and reversal of fibrosis that indicate a successful BLCC treatment response in the VLU bed. Applying these findings to the selection, testing and development of new treatments will hopefully pave the way for identifying therapies that benefit patients suffering from nonhealing VLUs.