Expansion and isolation of human ABCB5+ MSCs
Human ABCB5+ MSCs were derived from skin samples obtained from patients undergoing abdominoplasties or other surgical interventions that provide left-over skin tissue after informed written consent was obtained. Cell production was carried out in an EU-GMP grade A cabinet in a grade B clean room under laminar air flow following a validated GMP-conforming protocol as described previously . In brief, after enzymatic digestion of the skin tissue, cells were centrifuged and expanded as unsegregated culture by serial passaging upon adherence selection in an in-house MSC-favoring medium (Ham’s F-10 supplemented with fetal calf serum, L‑glutamine, fibroblast growth factor 2 (FGF‑2), HEPES, hydrocortisone, insulin, glucose, and phorbol myristate acetate). ABCB5+ cells were isolated by antibody-coupled magnetic bead sorting using a mouse anti-human ABCB5 monoclonal antibody directed against the extracellular loop 3 of the ABCB5 molecule  (Maine Biotechnology Services, Portland, Maine; GMP purification: Bibitec, Bielefeld, Germany), cryo-preserved in CryoStor® CS10 freeze medium (BioLife Solution, Bothell, WA) containing 10% dimethyl sulfoxide and stored in the vapor phase of liquid nitrogen.
Induction of cell hypoxia
ABCB5+ MSCs (3×105) were seeded in a culture dish and placed in a hypoxia chamber, which was flushed with nitrogen-enriched gas (1% O2, 4% CO2, 95% N2) at a rate of 25–50 l/min.
ABCB5+ MSCs were seeded onto coverslips, fixed with 4% paraformaldehyde solution, permeabilized with 1% TritonTM X‑100 (Sigma-Aldrich, Taufkirchen, Germany) in phosphate-buffered saline, blocked with 0.5% bovine serum albumin in phosphate-buffered saline, and stained for HIF‑1α (for antibodies see Additional file 1: Table S2). Nuclei were counterstained with 4’,6 diamidino-2 phenylindole (DAPI) and stains microscopically (EVOSTM FloidTM cell imaging station; Life Technologies, Darmstadt, Germany) evaluated.
Quantitative real-time polymerase chain reaction (qPCR)
Total RNA was isolated using the RNeasy® Micro Kit (Qiagen, Hilden, Germany) and reverse-transcribed into cDNA using the Applied BiosystemsTM High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher, Dreieich, Germany) and the Applied BiosystemsTM SYBR Green Mastermix (Thermo Fisher) in a four-step process run in a Mastercycler® Personal thermocycler (Eppendorf, Hamburg, Germany). PCR reactions were run in triplicate in an Applied BiosystemsTM StepOne RealTimeTM PCR System (Thermo Fisher). Primer sequences are provided in Additional file 1: Table S3. Actin served as housekeeping gene. Primer quality and integrity of the amplified product was confirmed by melting curve analysis. Identity of the PCR products was confirmed by agarose gel electrophoresis. Relative quantification of transcript levels was determined using the 2−ΔΔCt algorithm.
Enzyme-linked immunosorbent assay (ELISA)
VEGF concentration in the cell culture supernatant was measured using the Invitrogen VEGF Human ELISA Kit (Thermo Fisher), according to the manufacturer’s instructions. Assays were run in triplicate.
Angiogenic trans-differentiation assay
ABCB5+ MSCs (1×106) were seeded in 24‑well culture plates and cultured for up to 96 h in culture medium supplemented with 200 ng/ml recombinant human (rh) VEGF (Sigma-Aldrich), 1000 ng/ml rhFGF‑2 (CellGenix, Freiburg, Germany) and 1000 ng/ml rh platelet-derived growth factor‑BB (PDGF‑BB; R&D Systems, Wiesbaden, Germany). Trans-differentiation and proliferation activity were assessed by CD31 and Ki67 staining, respectively (for antibodies see Additional file 1: Table S2). Nuclei were counterstained with DAPI. All experiments were performed in triplicates. Human umbilical vein endothelial cells (HUVECs; 5×105; Thermo Fisher) served as positive control.
Tube formation assay
ABCB5+ MSCs (1×105/ml or 1.5×105/ml) and HUVECs (0.5×105/ml or 1×105/ml) were seeded on GeltrexTM (Thermo Fisher)-coated culture plates and incubated at 37°C for 19–22 h (ABCB5+ MSCs) and 16–18 h (HUVECs). For examination of cell viability, cells were stained with calcein acetoxymethylester (Thermo Fisher; 1:10,000, 30 min, 37°C). Tube formation and calcein fluorescence were evaluated microscopically (EVOSTM FLoidTM cell imaging station).
Hindlimb ischemia (HLI) induction and post-surgical care
Male OF1 mice (Charles River Laboratories, Saint-Germain-Nuelles, France) were anesthetized with 2% isoflurane in 100% oxygen, and the inner faces of both hindlimbs were carefully shaved. After local disinfection, an about 1‑cm skin incision was made on the left hindlimb from the inguinal region to the bifurcation region of the femoral artery into the saphenous and popliteal artery. The femoral artery and vein were dissected from the nerve. The femoral artery/vein block was ligated proximally by two 8‑0 ties placed just distally from the superficial epigastric artery, and distally by two 8‑0 ties placed just proximally from the bifurcation of the femoral artery into the saphenous and the popliteal artery. After cutting the femoral artery/vein block between the two proximal and between the two distal ties, the femoral artery/vein block was removed. When necessary, major branches such as the lateral circumflex femoral artery were ligated to avoid bleeding.
Thereafter, subcutaneous tissue and skin were closed with non-resorbable sutures or clamped with titanium micro clips (WDT, Garbsen, Germany). Postoperative care included pain management by injection of buprenorphine (Buprenovet, Bayer; 0.1 mg/kg) once directly after surgery or flunixin meglumine (2.5 mg/kg twice daily) during 3 days and daily local wound care with an antiseptic healing cream (Dermaflon, Pfizer).
Injection of ABCB5+ MSCs
On the day after surgery, mice were anesthetized with isoflurane to receive intramuscular injections at the ischemic limb of human ABCB5+ MSCs suspended in Ringer’s lactate solution containing 2.5% human serum albumin and 0.4% glucose at concentrations between 1×106 and 1×108 cells/ml, as required. Cell doses, injection volumes and sites are given in the Results section.
Blood perfusion measurement
Animals were anesthetized with isoflurane and placed on a warming platform in a supine position for imaging at the internal face of the thighs. Hindlimb blod flow was measured before and immediately after surgery (day 1) and on days 3, 5, 7, 14, 21 and 28 by real-time laser doppler blood perfusion imaging (LDPI; PeriCam PSI, Perimed Instruments). The scanned area covered an ellipse framing internal face of the thigh. Blood perfusion was expressed as the ratio between LDPI values in the left (ischemic) and right (non-ischemic) limb.
Histopathology and immunohistochemistry
Animals were sacrificed by CO2 inhalation and the thigh and gastrocnemius muscles preserved in 10% neutral buffered formalin solution, which was replaced after 24–48 h by 70% ethanol. Fixed tissues wee embedded in paraffin wax, cut to 2–4 µm thickness, stained with hematoxylin and eosin, and inspected by conventional light microscopy. Neovascularization was semi-quantitatively quantified as 0 = none, 1 = minimal capillary proliferation, focal, 1–3 buds, 2 = groups of 4–7 capillaries with supporting fibroblastic structures, 3 = broad band, and 4 = extensive band of capillaries with supporting fibroblastic structures. Immunohistochemical staining for CD31 was performed using rabbit anti-human/mouse CD31 (ab28364, Abcam; dilution 1:50) and dextran polymer-horseradish peroxidase-labelled anti-rabbit IgG (DAKO EnVision®+, K4010, Agilent) for detection. CD31 expression was semi-quantitatively quantified as 0 = none, 1 = minimal, 2 = slight, and 3 = moderate by two independent investigators who were blinded to the treatment.
One-way ANOVA followed by Dunnett’s test was used to compare LDPI ratios within versus baseline and neovascularization and CD 31 expression in the cell-treated groups versus control.
Adults (18–85 years) with diabetes mellitus type 2 (hemoglobin A1c <11%) were eligible if they had a neuropathic diabetic plantar foot ulcer (Wagner grade 1 or 2, 1–50 cm2), confirmed by vibration sense testing (128‑Hz Rydel-Seiffer tuning fork) without presence of significant arterial disease (ankle-brachial index ≥0.7 or transcutaneous oxygen pressure >40 mmHg or as per Doppler ultrasonography).
Main exclusion criteria were acute Charcot foot, active osteomyelitis, treatment-requiring ulcer infection, adjacent or chronic skin disorders, skin malignancies, acute or untreated deep vein thrombosis, need for hemodialysis, surgical procedures within 2 months and use of active wound care agents within 2 weeks prior to treatment, and current use of systemic immunosuppressants, cytotoxics or glucocorticoids.
The study was a national, multicenter (eight sites in Germany), open-label, single-arm, phase I/IIa trial comprising three periods: standard-of-care screening (≥ 6 weeks), treatment and efficacy follow-up (weeks 1–12), and safety follow-up period (until end of month 12). The trial was performed in accordance with the Declaration of Helsinki and local regulations and approved by the ethical committees of all participating study sites. Patients gave written informed consent prior to trial participation.
Treatment consisted of up to two topical applications of 2×106 allogeneic ABCB5+ MSCs (suspended in Ringer’s lactate solution containing 2.5% human serum albumin and 0.4% glucose ) per cm2 wound area on day 0 and at week 6. The cells were manufactured as a GMP-conforming standardized ATMP (for main product release data see Additional file S1: Table S4). Originally, only one cell application was planned. The second application was amended to the protocol only after data from a first-in-human trial on chronic venous ulcers  suggested that a second cell dose at 6 weeks after the first cell dose might provide additional benefit for chronic wound healing. Cell application could be preceded by an optional wound debridement at the investigator’s discretion followed by waiting until the bleeding had entirely stopped. For cell application, a suspension containing 1×107 ABCB5+ MSCs/ml was applied onto the wound surface, delivering 2×106 ABCB5+ MSCs/cm2 wound surface area. Thereafter, the cells were allowed to settle for 15–30 min, optionally fixed in place with fibrin gel (Tisseel®; Baxter, Unterschleißheim, Germany), and then the wound was covered with a waterproof film dressing (Tegaderm™; 3M, Neuss, Germany). On the following day (≥ 12 hours after cell application), the film dressing was replaced by a microbe-binding dressing (Cutimed® Sorbact® tamponade or compress; BSN, Hamburg, Germany), which was changed again 1–2 days later. Additionally, patients received standard care until week 12 including glycemic control, ulcer debridement, appropriate wound dressings (i.e. microbe-binding tamponade of cavities and exudate-absorbent foam dressing for coverage), and antibiotics if required. All patients had to use offloading devices including cast devices or individually fitted therapeutic footwear [19, 54, 55].
Primary efficacy endpoint was percent wound surface area reduction at week 12 or last available post-baseline measurement. Secondary efficacy endpoints were percent and absolute wound surface area reduction at predefined visits, proportion of patients achieving complete and 30% wound closure, time to complete and to 30% wound closure, granulation, epithelialization, wound exudation, time to amputation at the target leg, pain and life quality. Safety outcome measures included adverse events (during the whole study period) and vital signs, changes in physical examination findings and time to amputation of the target leg (during efficacy follow-up).
Wound surface area determination followed a multi-step approach combining computerized evaluation (PictZar® planimetry software; BioVisual, Elmwood Park, NJ, USA; 98% accuracy, 94% inter-rater reliability, 98% intra-rater reliability according to a validation and reliability study ) of standardized photographs and depth measurements using a wound measuring probe, to account for the typical three-dimensional shape of DFU wounds; i.e. consisting of wound floor, side wall and, occasionally, not visible tunneling or undermining areas. For details of the measuring and calculation algorithm see Additional file 2: Methods S1. Formation of granulation and epithelial tissue was estimated by the investigator in % of wound area from standardized wound photographs. Wound exudation was rated by the investigator as low (dry), moderate (moist), and high (wet) according to the criteria defined by the World Union of Wound Healing Societies . Pain was rated by the patient using a 0–10‑point numerical rating scale with 0 = no and 10 = worst imaginable pain. Quality of life was assessed using the participant-reported Short Form (36) Health Survey (SF‑36) and Dermatology Life Quality Index (DLQI) questionnaires.
Enrolment followed a Simon optimal two-stage design with responders defined as patients presenting with at least 30% wound surface area reduction at week 12. The sample size required to achieve 80% power at 5% significance level was calculated using PASS 13 software (NCSS, East Kaysville, UT, USA) to be 37 patients. This enabled the option to terminate the trial if ≤6 or ≥14 of the first 18 treated patients were responders. As in an interim analysis 12 of 18 patients emerged as responders, recruitment was continued. However, by force of the emerging COVID‑19 pandemic, the trial was prematurely completed. At that time, 23 patients had been treated.
Safety assessments were performed on the safety analysis set, which included all patients who received at least one cell dose. Efficacy assessments were performed on the full analysis set (FAS), which included all patients of the safety analysis set who underwent wound surface area assessments at baseline and at least one post-baseline visit, and on the per-protocol set (PP), which included all patients of the FAS who had no major protocol deviations.
If not otherwise stated, normally (D'Agostino–Pearson normality test) distributed parameters are presented as mean ± standard deviation, and non-normally distributed parameters as median and interquartile range (IQR). Statistical significance of percent wound surface area changes from baseline was tested against the null hypothesis (median change = 0) using a two‑sided Wilcoxon signed rank test. Time to complete wound closure, time to 30% wound surface area reduction and time to amputation at the target leg were analyzed using the Kaplan–Meier method.