Personalized Scaffolds for Diabetic Foot Ulcer Healing Using Extracellular Matrix from Induced Pluripotent Stem-Reprogrammed Patient Cells

Abstract

Diabetic foot ulcers (DFU) are chronic wounds sustained by pathological fibroblasts and aberrant extracellular matrix (ECM). Porous collagen-based scaffolds (CS) have shown clinical promise for treating DFUs but may benefit from functional enhancements. Our previous work showed fibroblasts differentiated from induced pluripotent stem cells are an effective source of new ECM mimicking fetal matrix, which notably promotes scar-free healing. Likewise, functionalizing CS with this rejuvenated ECM showed potential for DFU healing. Here, we demonstrate for the first time an approach to DFU healing using biopsied cells from DFU patients, reprogramming those cells, and functionalizing CS with patient-specific ECM as a personalized acellular tissue engineered scaffold. We took a two-pronged approach: 1) direct ECM blending into scaffold fabrication; and 2) seeding scaffolds with reprogrammed fibroblasts for ECM deposition followed by decellularization. The decellularization approach reduced cell number requirements and maintained naturally deposited ECM proteins. Both approaches showed enhanced ECM deposition from DFU fibroblasts. Decellularized scaffolds additionally enhanced glycosaminoglycan deposition and subsequent vascularization. Finally, reprogrammed ECM scaffolds from patient-matched DFU fibroblasts outperformed those from healthy fibroblasts in several metrics, suggesting ECM is in fact able to redirect resident pathological fibroblasts in DFUs towards healing, and a patient-specific ECM signature may be beneficial.

Graphical Abstract

In this article, Santarella F. and colleagues produce a blended and a decellularized scaffolds starting from extracellular matrix produced by rejuvenated fibroblasts, for the first time derived by iPSC reprogramming of diabetic foot ulcers patients’ cells. Further beneficial matrix is observed upon seeding of patient match pathological fibroblasts, with a net resources reduction in decellularized scaffolds.

1. Introduction

Diabetic foot ulcers (DFU) are a major complication in the diabetic population. Of the 470 M diabetic patients, about 30 M develop a DFU, which are chronic wounds that are often refractory to therapy. This results in 15–20% of DFU cases ultimately requiring limb amputation ^[1]. Globally, the cost of DFU patients comprises a 760 billion USD market ^{[1a, 2]}. Current standard DFU therapies are focused on: control of normoxia underlying diabetes, traditional bandages, surgical debridement, and pressure-relief devices to offload the wound ^{[2d, 3]}. Unsurprisingly, there have been a wide range of tissue engineering approaches investigated, combining some or all of cell therapies, growth factors, and biomaterial support structures. Among these, examples which have shown clinical promise include: Apligraf^® (an in vitro grown ‘skin’), SkinTE^™ (an acellular foam derived from skin explant extracellular matrix-ECM) and Integra^® Omnigraft^™ (a collagenous acellular porous freeze dried scaffold) ^[4]. However, these devices had ~50% healing and thus improved tissue engineering approaches to DFUs are required ^[5].

Induced pluripotent stem (iPS) cells are a technology that enables somatic cells to become pluripotent, and then re-differentiated into any desired lineage ^[6]. Our lab was the first to demonstrate the ability to fabricate a porous tissue engineering scaffold using matrix from iPS-reprogrammed cells ^[7]; however, the source cells were derived from a commercial healthy neonatal fibroblast cell line ^[8]. These fibroblasts were cycled through iPS-reprogramming and redifferentiated into rejuvenated fibroblasts. The rejuvenated fibroblasts had enhanced ECM production, and were enriched in favorable wound healing elements (e.g., collagen III and fibronectin) ^[9], similar to those found in fetal ECM (which notably, heals without scarring ^[10]). This ECM was then used to fabricate a freeze-dried acellular scaffold using a similar technique to the clinically-approved collagen-GAG scaffolds ^[11]. This scaffold showed improved ECM deposition from DFU over collagen-GAG controls, approached significance in vascularization improvement, and had little immune reaction. A key advantage of this approach is that it allows us to take advantage of the benefits of iPS-reprogrammed cells, without the need to overcome the challenges of using these cells directly as a therapeutic ^[12].

Building on the promise of this approach, we wanted to test whether our technique could be adapted to a personalized medicine approach with pathological cells taken from a patient’s DFU. These cells could be reprogrammed and rejuvenated and used to produce ECM for development of a freeze-dried scaffold. Removing (fibroblast-rich) tissue is the standard of care in DFU treatment as these wounds are frequently debrided; thus, source cells are routinely available. Additionally, we speculated that because the previous cells were from a healthy source, the differences between the reprogrammed and source cells would be even more pronounced for DFU cells.

In this feasibility study, a specific impaired patient-sourced fibroblasts from diabetic foot ulcers (DFU – extensively described as DFU 8 clone previously ^[13]), were used as source of ECM material. This DFU patient was selected based on the highly impairment in regeneration and ECM production. DFU fibroblasts are then reprogrammed using iPS technology and redifferentiated into rejuvenated fibroblasts named ‘iDFU’. These fibroblasts have previously been shown to produce a rejuvenated extracellular matrix and exploited as “fabric entities” to produce ECM material for our scaffolds ^{[13c, 13d, 14]}. Briefly, we have chosen for this feasibility study, one clone (named i2DFU8 previously). ECM elements enriched in these cells includes those also found in fetal skin ECM (i.e.: collagen III and GAG, giving benefit to the wound healing ^[15]), and these cells had enhanced healing in a human skin equivalent model. We hypothesized, its potential is linked to its matrix and therefore was chosen to be exploited. Following the previous proteomics analysis of the matrix of these rejuvenated fibroblasts we have chosen 7 ECM markers relevant for regeneration to analyze and found increased ^{[13a, 13d, 16]}: GAGs, collagen I (main component of ECM ^[17]), collagen III (collagen type found in increased levels during scarless wound healing ^[18]), collagen IV (basal lamina ^[19]), fibronectin 1 (primordial ECM deposited in wound healing ^{[13c, 15c]}, LAMA1 (present in mature skin ^[20]) and LAMA5 (staminal niche ^[21]). We have analyzed these components and fabricated a porous freeze-dried scaffold using their ECM and analyzed the potential of this scaffold for diabetic wound repair. To assess our scaffolds in vitro, we chose to monitor key processes impaired in the DFU healing process (e.g., cell proliferation, ECM production, vascularization potential) ^{[13b, 15a]}. To our knowledge, this is the first-time patient-specific iPS cell reprogramming to generate a source ECM for scaffold fabrication has been used for any application for tissue engineering. Since sourcing sufficient cell quantities for a personalized medicine approach can be a challenge and the freeze-drying process has inefficiencies and can cause material loss, we wanted to develop a second fabrication approach that uses fewer cells. To do this, we seeded iDFU or control cells directly onto collagen-GAG scaffolds and cultured them until they had deposited ECM within the scaffold. This approach has the advantage that the ECM is organized in its natural 3D structure by fibroblasts. Following decellularization, we tested whether the iDFU-produced proteins remained within the scaffold and then tested the scaffolds for their wound repair potential using the same screening as the freeze-dried scaffolds.

Thus, the specific focus of this manuscript is finding an engineering solution to show the feasibility of using DFU patient fibroblast as precursors for iDFU rejuvenated fibroblasts, and then exploiting iDFU as a source of rejuvenated material for porous scaffold fabrication in 2 different ways. These scaffolds were analyzed in vitro for ECM deposition and in vitro vascularization as indicators of their regenerative potential.

2. Results & Discussion

2.1. Diabetic foot ulcer fibroblasts are rejuvenated following induced pluripotent stem cell reprogramming

DFU fibroblasts have an impaired regeneration and their pathological ECM has reduced content of collagens such as collagen type III ^{[13b, 13c]}. Recently, iPS reprogramming has been used to rejuvenate cells and have them produce a more juvenile matrix ^[22]. For example, iDFU ECM has an increased content of GAGs and Collagen III (present in young matrices which heal scar free ^[10f]) and the incorporation of cells into a human skin equivalent with DFU impairment induced a faster regeneration ^{[9a, 13d, 16a]}. We hypothesized that the ECM was playing an active role in DFU regeneration and that this ECM could be used as a source material to produce an acellular scaffold that drives regeneration. Our previous results showed that beneficial extracellular matrix (ECM) elements produced by fibroblasts cycled through iPS reprogramming form promising scaffolds for diabetic wound repair ^[8a]. These beneficial ECM elements are also produced from iPS-reprogrammed fibroblasts from patient DFU ^[13d]. This motivated the exploration of ECM from patient-derived cells as a source material for our scaffolds, with a potential for this technique to be used in a personalized medicine approach. To demonstrate the potential of this approach, in this study, one patient-sourced fibroblast from diabetic foot ulcers (DFU – previously described extensively as “DFU8” in ^{[13a, 14]}), were used. We chose these patient-derived DFU cells as they are known to have impaired regeneration capabilities and ECM production ^{[13c, 13d]}. From these DFU fibroblasts, their iPS reprogrammed counterpart, iDFU, were produced and analyzed for their suitability for this approach. Following previous publications we chose the iDFU clone (rejuvenated cells previously known as ‘i2DFU8’) with high collagen III presence in its ECM and that had demonstrated successful re-epithelialization in a wounded human skin equivalent model ^{[13b, 13d]}. We hypothesized, its pro-repair potential is linked to its extracellular matrix and wanted to test whether this could be used for enhanced personalized scaffold fabrication. We found this specific cell pair suitable to quantify possible regeneration signs in vitro, given the contrast between the pathological DFU and reprogrammed iDFU ^[13d].

We first confirmed DFU and iDFU cells identity as fibroblasts via flow cytometry, which is consistent with our previously published work (Figure S1–S4) ^{[9a, 9b, 13c, 13d, 14, 16a, 23]}. We then analyzed if their ECM is a suitable material for scaffold fabrication.

On a per cell basis, iDFU cells are more efficient at producing matrix than control cells and their pathological counterpart, DFU, and the RNA of iDFU cells is elevated for matrix elements that are consistent with fetal ECM (see Figure 1)^[10c]. We recovered 50mg of dry weight ECM for primary patient fibroblasts (both normal foot fibroblasts and pathological DFU fibroblasts) and about 10mg for reprogrammed iDFU fibroblasts (Figure 1A). When compared to DNA content, iDFU have about 3 times less DNA compared to primary fibroblasts (Normal fibroblasts and DFU). Yet, iDFU samples had the most collagen, and when normalized to DNA content, there was at least 3 times higher content of ECM collagen and GAG versus control fibroblasts or DFU (Figure 1A). In our previous work with a neonatal commercial fibroblast line (BJ Cells), we also saw higher ECM production when these cells were iPS reprogrammed versus the source BJ cells.

Figure 1.

Patient derived fibroblasts Normal (black), DFU (red) and iDFU (green) gene and protein expression after 3 weeks in 2D in flasks differentiation media + 100 μg/mL AA, 64k cells/cm². (A) dry lyophilized monolayer weight per flasks; total collagens (I to IV); total GAG; total DNA; and collagens and GAGs normalized to DNA. (B) Expression of structural ECM protein genes RNA, normalized to relative GAPDH of control fibroblasts at time zero. Graphs show averages ± standard error mean, one-way ANOVA + Tukey’s, *=p<0.05 (n=3).

The messenger RNA of the major ECM proteins show that patient’s iDFU fibroblasts outperform DFU and healthy fibroblasts in terms of expression of collagen I, III, IV, fibronectin 1, LAMA1 and LAMA5 –key components of skin ECM (Figure 1B). The relative blot quantification detects on average higher collagen III and fibronectin presence in iDFU ECM (non-significant); equivalent amounts of collagen I to normal cells; and higher collagen I than DFU. Collagen III is elevated in scarless fetal regeneration making it desirable for inclusion in skin regeneration scaffolds (Figure S1) ^{[9a, 10c–e, 13a, 13c, 15b]}. These results are in line with the expectation of having more GAGs and collagen III for reprogrammed iDFU fibroblasts, while DFU have an impaired ECM production and less collagen III content; as per previous data generated on cell lines, fibroblasts cycled through iPS reprogramming produces a rejuvenated matrix signature ^{[13c, 13d]}.

2.2. iDFU-derived ECM can be used to fabricate a scaffold with matrix elements that are beneficial for skin regeneration.

A 50:50 mix of collagen I (bovine tendon origin) and ECM from patient-derived cells was blended and freeze dried according to a previously optimized protocol to produce tissue engineering scaffolds ^{[23a, 24]}. Scaffolds activated with DFU ECM were called Coll-DFU, while scaffolds with iDFU ECM were labelled Coll-iDFU, and compared to the clinically relevant control scaffold containing collagen-glycosaminoglycan (CG) ^{[4a, 25]}. To our knowledge this is the first time a porous freeze-dried scaffold has been generated using ECM from patient cells that have been iPS-reprogrammed. Inclusion of fibroblast ECM in the slurry did not macroscopically affect the scaffolds after freeze drying (Figure S5A) but after PBS re-hydration Coll-DFU scaffolds appeared more transparent (Figure S5B) and after 3 weeks of culture with fibroblasts Coll-DFU scaffolds had significantly contracted (Figure S5C), while Coll-iDFU were not as evidently affected. The patient ECM functionalized scaffold Coll-DFU had significantly smaller pores (~ 40 μm) than Coll-iDFU (~ 70 μm), and both had a finer microstructure compared to the CG control (150 μm) (Figure 2A–B). This was consistent with the cell line ECM functionalized scaffolds that we previously published ^[23a]. Patient ECM-activated scaffolds – Coll-DFU and Coll-iDFU – had a lower Young’s modulus (both 0.5 KPa versus CG control ca. 1 KPa) (Figure 2C) consistent with our previous work, and this was attributed to a lower collagen I content ^[23a]. GAG levels were equivalent between Coll-DFU and Coll-iDFU but were 8 times lower than the levels introduced during fabrication for controls (Figure 2D). Immunofluorescence analysis shows inclusion of a diversity of ECM proteins (chosen among those selected via proteomics and transcriptomics analysis in previous studies ^{[13a, 13d, 16b]}) uniformly distributed in the ECM activated scaffolds, with a significant increase in the presence of collagen III, IV and fibronectin 1 for Coll-iDFU compared with CG controls (Figure 2E–F). These proteins are also elevated in younger skin and in fetal skin (which heals scar free) and are desirable in DFU healing to increase ECM integrity and diversity ^{[10c–f, 13c, 15b, 15c, 18]}. Of note, fibronectin 1 is highly present in our scaffold, and fibronectin is abundant during the initial stages of wound healing ^{[13c, 15c]}. With scaffolds successfully fabricated from iDFU ECM, we next tested their suitability for wound repair by seeding them with non-diabetic patient foot fibroblasts (note, these were not from the same patient) and pathological DFU fibroblasts from the same patient as those used to generate ECM for scaffold fabrication.

Figure 2.

Characteristics of cross-linked patient ECM- activated scaffolds CG (empty control- black), Coll-DFU (collagen blended with DFU ECM - red) and Coll-iDFU (collagen blended with iDFU ECM - green). (A) Scanning electron microscopy of scaffolds, scale bar=100μm; (B) Pore size quantification of 150 pores in per scaffold type (n=3). (C) Compression test Young’s modulus of fabricated scaffolds (n=6). (D) GAGs quantification after scaffold fabrication (n=3). (E) Fluorescence images of sectioned scaffolds for quantification of collagen I (green), collagen III (red), collagen IV (green), fibronectin (green), LAMA1 (green), LAMA5 (green) and dapi (blue), scale bar 100 μm. (F) Integrated fluorescence density expressed as arbitrary units [A.U.] analysis of ECM components present in fabricated scaffolds. Graphs represent averages ± standard error mean, one-way ANOVA+Tukey’s, *=p<0.05.

2.3. iDFU ECM-activated scaffolds enhance matrix deposition most effectively when seeded with patient-matched pathological DFU cells

In order to prove iDFU derived ECM was capable of inducing a successful matrix deposition by seeded DFU, we have monitored key elements like Collagen III, fibronectin 1 and GAGs, as they are key to heathy scarless regeneration and they are impaired in DFU matrix deposition ^{[10c, 15, 26]}. Following cell seeding onto the scaffolds, all scaffolds demonstrated the ability to support cell growth at same level as controls and comparable to previous studies keeping the same readout and methods previously developed ^{[5b, 23a, 27]} (Figure 3A). While the GAG levels decreased versus their starting values for control fibroblasts and DFU seeded on CG (as previously seen ^{[23a, 27]}), GAG levels for Coll-DFU and Coll-iDFU remained consistent with unseeded scaffolds (Figure 3B). In future work it will be interesting to determine whether the GAG in our ECM-activated scaffolds being of human origin – versus shark origin for the CG – plays a role in this, or whether the lower level of GAGs is desirable for regeneration ^{[25, 28]}. In order to analyse protein deposition induced by the scaffolds after 3 weeks seeding (a readout that can also be monitored and quantified subsequently in vivo), we opted to quantify the ECM produced via immunofluorescence quantification ^{[23a, 27]}. When we analyzed protein distribution across the scaffold, we found that the diverse ECM components were evenly distributed. Coll-iDFU seeded with DFU cells is the most effective group for inducing a high presence of collagen I (main component of ECM ^[17]), collagen III (collagen type found in increased levels during scarless wound healing ^[18]) and LAMA1 (present in mature skin ^[20]) (Figure 3C–D). In the same group a reduction of fibronectin 1 is seen ^{[13c, 15c]}. Collagen IV (basal lamina ^[19]) and LAMA5 (staminal niche ^[21]), are not significantly different to the other groups, although LAMA5 appears elevated for the DFU seeded on Coll-iDFU. Although RNA data confirms a generalized increased effect on ECM protein transcription from the Coll-iDFU seeded with both normal and DFU fibroblasts (Figure S6) – correlating with the increased presence of protein in immunofluorescence data – we were more focused on the protein produced. This, is more functional than the PCR readouts at the transcription level that represent a specific snapshot in time ^[29]. These data indicate that the DFU cells can be induced to outperform control cells – despite them being ‘healthier’ i.e., non-diabetic – when they are placed in the appropriate extracellular matrix. This is a key observation motivating how essential the appropriate extracellular matrix is for DFU repair, and contradicting the dogma that these cells are irreversibly damaged. We can speculate that DFU fibroblasts recognize their own rejuvenated-ECM in Coll-iDFU, responding with enhanced deposition, and this reinforces the opportunity of utilizing a personalized medicine approach in the clinic. There has recently been evidence that previously insulted skin cells show enhanced healing potential, and perhaps this relationship is expanded to aspects of an extracellular matrix ‘memory’ ^{[30] [31]}.

Figure 3.

Analysis of 3 weeks seeded CG (black), Coll-DFU (red) and Coll-iDFU (green) scaffolds with healthy control fibroblasts (not patient matched) and pathological DFU (patient matched). (A) Total DNA and (B) GAGs quantifications of seeded scaffolds. (C) Fluorescence images of sectioned scaffolds for quantification of collagen I (green), collagen III (red), collagen IV (green), fibronectin (green), LAMA1 (green), LAMA5 (green) and dapi (blue), scale bar 100 μm. (D) Integrated fluorescence analysis of ECM components present in fabricated scaffolds. Graphs represent averages ± standard error mean, one-way ANOVA +Tukey’s, *=p<0.05.

2.4. Coll-DFU and Coll-iDFU scaffolds allow endothelial cells to form vessel-like structures comparable to CG controls

Vascularization experiments are designed to check whether the novel scaffolds supports vascularization at a similar or higher level to the already successful control CG scaffold, and can inform us on whether the deposited ECM elicits a favorable response ^[32]. To further characterize the regenerative potential of Coll-iDFU we tested their ability to induce vascularization in-vitro following a standardized protocol seeding with human umbilical vein endothelial cells (HUVEC) and mesenchymal stromal cells (which act as pericytes) ^{[23a, 33]}. Results revealed the ability of our ECM-activated scaffolds to match CG controls (Figure 4A–B). The conditioned media from fibroblast seeded scaffolds yielded vessel formation at same level as controls (vessel numbers and length), consistent with VEGF release data (Figure 4C–G). In summary, this data suggests that the vascularization potential for Coll-DFU or Coll-iDFU is equivalent to the clinically approved CG control scaffold.

Figure 4.

In-vitro vascularization of reprogrammed ECM scaffolds. (A) Number of vessel-like structures formed by HUVEC in co-culture with MSC on CG (black), Coll-DFU (red) and Coll-iDFU (green). One-way ANOVA+Tukeys post-test, n.s. = p>0.05 (n=6), mean±SEM. (B) Immunofluorescence of vascularized scaffolds. Blue = DAPI and red = phalloidin. Red are considered confirmed vessel-like structures. Scale bar 300μm. (C) VEGF ELISA quantification (ANOVA + Tukey’s post-hoc) of 3 week seeded scaffolds with Control or DFU fibroblasts (* = p<0.05; n=3; Mean ± SEM). (D) Matrigel assay on HUVEC seeded for 24h with conditioned media from CG, U-Coll and iU-Coll seeded with control or DFU fibroblasts (3 weeks). (E) Quantification of number of vessels formed in matrigel, (F) average length and (G) total length of the vessels average ±SEM 1-way ANOVA+Tukeys post-test *=p<0.05 (n=3).

2.5. Development of a decellularized scaffold

While the data and analysis of Coll-iDFU was promising in terms of its functional potential for a personalized medicine approach, the fact that the iPS reprogrammed iDFU fibroblasts were less proliferative motivated us to develop a technique that would require fewer cells. In addition, subjecting the 2D-produced proteins to blending and freeze-drying may disrupt protein folding and activity. To test if unblended ECM is more effective than a blended scaffold technique and provide a method that utilizes less resources, we moved to a decellularized system. To generate a decellularized scaffold, we first seeded rejuvenated patient cells (iDFU), or controls, on CG scaffolds and allowed them to deposit ECM for 3 weeks (Figure 5). Decellularization was then performed through osmotic shock and freeze-drying, followed by DNAase. Beyond reducing cell numbers, this approach simulates ECM production in 3D, and should leave the folded proteins in place in a more natural configuration than would result from harvesting and blending.

Figure 5.

Schematic of the decellularization approach. Cells (iPS rejuvenated iDFU) are seeded for 3 weeks (with 100μg/mL ascorbic acid (AA)) allowing ECM to be deposited directly in place prior to decellularization. This should leave an ECM enriched scaffold with unaltered coated rejuvenated elements to be tested for DFU seeding. Decellularization takes place in 2 subsequent steps: Freeze thaw in ddH₂O followed by lyophilization (1^st step) and rehydration in ddH₂O and gDNAse (2^nd step) washings. These scaffolds are considered decellularized (D-iDFU or D-DFU). The resulting scaffolds are then crosslinked a second time and sterilized for analysis and seeding.

2.6. Reprogrammed iDFU fibroblasts enrich scaffolds with pro-regenerative ECM

After 3 weeks, DFU and iDFU have penetrated the scaffolds (Figure 6A); while there are higher numbers of DFU, the iDFU cells appear to more uniformly migrate to the core of the scaffold. Quantification confirmed the higher number of DFU versus iDFU cells –consistent with our 2D data – while GAG levels were consistent between the 2 groups (but lower than the CG controls due to the inclusion of GAG in those scaffolds) (Figure 6B–C) ^{[13d, 27]}. The ECM markers (for both protein and mRNA) have higher averages in iDFU seeded scaffolds versus DFU seeded scaffolds – in accordance with what was found in (Figure 1 and 2). Additionally, there was significant enrichment in the ratio of collagen III to collagen I (mRNA and protein in iDFU groups) (Figure 6D–E–F), consistent with previous findings ^{[9a, 10c, 10e, 18]}. These scaffolds were then used for decellularization and tested to see if the iDFU scaffolds maintain there ECM enrichment.

Figure 6.

Biopsy-derived diabetic foot ulcer cells and their post-iPS derived iDFU fibroblasts cell micrographs, 3 weeks post seeding on crosslinked CG scaffolds prior to decellularization. (A) Representative images of cell penetration into scaffolds (n=3; blue=dapi; white=phalloidin 633; scale bar 500 μm). (B) DNA (ng/mL) and (C) GAG content (μg/mL) of CG seeded scaffolds at 3 weeks, before decellularization (DFU – red in graphs; iDFU –green in graphs; empty CG –black in graphs). Graphs express mean±SEM, ANOVA + Tukey’s *=p<0.05 (n=3). (D) 3 weeks RNA expression of CG seeded scaffolds of main structural genes, normalized for control fibroblasts at time 0. Graphs express mean±SEM, one-way ANOVA + Tukey’s, *=p<0.05 (n=3). (E) Representative images of fibroblasts seeded on CG vs empty control CG scaffolds (n=3), blue=dapi, green=collagen I, collagen IV, fibronectin 1, LAMA1 and LAMA5, red=collagen III, scale bar 100 μm. (F) Integrated density of fluorescence. Graphs express mean±SEM, one-way ANOVA + Tukey’s *=p<0.05 (n=3).

2.7. Decellularization of iDFU seeded scaffolds maintains higher collagen III levels

The decellularization technique does not disrupt the overall architecture of the scaffolds (Figure 7A), with scaffolds retaining the same pore size and compression modulus as control CG and consistent with literature (Figure 7B–C) ^{[32a, 32b]}. This is different from the observation for the ECM-functionalized scaffolds formed by the blending technique, where pores are smaller and the Young’s modulus is lower (see Figure 2). Following decellularization, the average dry weight is higher for the DFU decellularized scaffolds (D-DFU) and for the iDFU decellularized scaffolds (D-iDFU) versus the control treated the same way (CG). DNA is more efficiently eliminated in the iDFU scaffolds; however, when normalized for weight, more DNA remains in the iDFU scaffolds (Figure 7D–E). The percentage of decellularization was around 80% following our technique, confirming a combined approach from what described allows an efficient DNA reduction ^[34]. DNA reduction correlates unexpectedly with the same level of reduction in GAGs (Figure 7F–G). This may reduce regenerative activity and could indicate a loss in proteins ^[35]. On the other hand, the amount of GAGs required for regeneration may not be as high as that included in the CG control. Using this decellularization technique, we observed the same ECM proteins as those found for the scaffold formed from iDFU ECM using the blending technique (Figure 2), but now the proteins are deposited in their more natural configuration. Specifically, collagen III, IV and fibronectin 1 had higher averages in D-iDFU than in D-DFU, with a significant difference observed in collagen III levels (Figure 7H–I). We next tested the suitability of the decellularized scaffolds for further cell culture with normal fibroblasts and DFU fibroblasts, and assessed their vascularization potential.

Figure 7.

Decellularization of patient derived fibroblasts seeded scaffolds. (A) Scanning electron microscopy 150X of decellularized scaffolds CG (black), D-DFU (red) and D-iDFU (green). (B) Pore size quantification (150 total pores each scaffolds type), n.s.=p>0.05, one-way ANOVA + Tukey’s post test. (C) Compressive test of scaffolds, n.s.=p>0.05, one-way ANOVA +Tukey’s post test (n=6). (D) Average dry weight after decellularization process average ± SEM one-way ANOVA +Tukey’s post test, *=p<0.05 (n=3). (E) Picogreen after decellularization, DNA values normalized to scaffold dry weight ±SEM, above the clinically acceptable level of 50ng/mL (n=3). (F) DFU and iDFU after decellularization with the % of GAG loss (right). Bars express average ± SEM one-way ANOVA +Tukey’s post test, *=p<0.05 (n=3). (G) Picogreen quantification of residual DNA after decellularization step of DFU (left) and iDFU (middle), with the % of DNA loss (right). Bars express average±SEM one-way ANOVA +Tukey’s post test, *=p<0.05 (n=3). (H) Representative images of seeded CG vs empty control CG scaffolds (n=3), blue=dapi, green=collagen I, collagen IV, fibronectin 1, LAMA1 and LAMA5, red=collagen III, scale bar 100 μm. No residual nuclei shape is seen. (I) Integrated density of fluorescence. Graphs CG background is subtracted (except for collagen I) and express mean±SEM, one-way ANOVA +Tukey’s *=p<0.05 (n=3).

2.8. Decellularized D-iDFU scaffolds induce a regenerative response in DFU patient matched fibroblasts

Decellularized scaffolds were crosslinked with EDAC and seeded with normal fibroblasts (control) and pathological DFU patient fibroblasts. Decellularized scaffolds D-iDFU outperformed the CG control in supporting the growth of healthy fibroblasts (unmatched patient controls) and patient-matched pathological DFU fibroblast growth, while D-DFU impaired cell growth for both cell types (Figure 8A). The same trend is seen for GAGs, where D-iDFU shows high GAG levels compared to controls, and D-DFU had the lowest for both cell types (Figure 8B). While we detected a difference only in collagen I RNA expression in D-iDFU seeded with DFU versus the controls (Figure S7), at the protein level we recorded a higher ECM enrichment as the result of events occurring over the course of 3 weeks. Rejuvenated D-iDFU scaffolds seeded with DFU fibroblasts, induce the highest levels of collagen I and III –which is promising for regeneration – and similar level of collagen IV – which induces epithelization –as controls (Figure 8C). This observation shows that these cells can be induced to be pro-repair if placed in the appropriate ECM environment ^[13d]. Moreover, when compared to the scaffolds where the ECM was blended in with collagen during fabrication (Figures 1 – 4), decellularized scaffolds perform better in allowing fibroblast growth and GAG presence, while keeping ECM protein deposition to the same extent seen for the blended scaffolds (Figure 3). Furthermore, these two approaches both show that the rejuvenated ECM from iDFU is beneficial for stimulating DFU regeneration. The consistency in the observations between the two approaches reinforces the idea of a patient-matched ECM memory present in DFU fibroblasts. Indeed, here we also observed that the DFU cells outperformed healthy cells on the patient-matched scaffold. To further verify the potential benefit of these scaffolds, we tested the vascularization potential of D-iDFU scaffolds. Moreover, the presence of increased collagen III and GAGs in our scaffolds suggested iDFU activated scaffolds can induce regeneration, similar to what was seen for scaffolds consisting of recombinant collagen III ^{[15a, 23a]}. These decellularized scaffolds, in addition to Collagen III enrichment, have the proteins deposited in situ (i.e., not undergoing blending and processing), potentially leaving the native protein conformation which can further help regeneration. Our previous data, showed already an improvement of DFU ECM deposition on blended scaffolds activated with ECM derived from post-iPSF (healthy fibroblasts line of neonatal origin) ^[23a]. These data are similar to what we observed here, where we have DFU seeded on mismatched ECM scaffold from healthy patients, with an average increased production in collagen III (Figure 8C). However the same figure shows increased efficacy when DFU are seeded on matched rejuvenated ECM from iDFU.

Figure 8.

Decellularized scaffolds of crosslinked CG (black), D-DFU (red) and D-iDFU (green) seeded with patient unmatched control fibroblasts and pathological patient matched DFU fibroblasts. (A) DNA and GAG quantification 3 weeks after scaffold seeding. One-way ANOVA + Tukeys *=P<0.05 (n=3). (B) Immunofluorescence and ECM protein Scale bar 100μm. Collagen I, IV Fibronectin 1 LAMA1 and LAMA5 (green), Collagen III (red), Blue=dapi nuclei. (C) Immunofluorescence quantification of ECM protein deposited on decellularized scaffolds, one-way ANOVA+Tukeys *=p<0.05 (n=3).

2.9. Decellularized D-iDFU scaffolds have increased vascularization versus D-DFU ones

Decellularized D-iDFU scaffolds had on average more vessels than the control CG (not statistically significant), while D-DFU decellularized scaffolds performed significantly worse that iDFU and CG (Figure 9A–B). This effect was expected, as D-DFU includes impaired DFU-like ECM, which normally does not lead to regeneration nor to vascularization ^{[9b, 13a–c, 16b, 26]}. The decellularized scaffold containing iDFU ECM (D-iDFU), performs as good as the control CG (based on a clinically-used product ^[4a]) if not slightly better, and we speculate this may due to its high content in rejuvenated ECM elements deposited in the correct conformation onto the CG scaffold backbone ^{[9b, 15a, 32d]}. As none of the scaffolds induced VEGF release that would be high enough to stimulate vascularization (10 ng/ml; Figure 9C–D–E), the D-DFU results suggest an underlying problem in the ECM deposited contributes to DFU revascularization ^[36]. In a tubule formation assay, conditioned media from fibroblast-seeded scaffolds demonstrated that both D-DFU and D-iDFU groups outperformed CG (Figure 9F–G). The individual length was shorter than the control (Figure 9H), but overall there was an increase in total length for D-DFU and D-iDFU groups (Figure 9I). These results are in line with the known ability of iPSC cycled fibroblasts to induce vascularization in different systems ^{[9b, 13d]}. Compared to the blended scaffolds, decellularized D-iDFU scaffolds increase the amount of vessel like structures (Figure 4).

Figure 9.

In-vitro vascularization on decellularized and crosslinked CG (Black), D-DFU (red) and D-iDFU (green) scaffolds. (A) number of vessel-like structures formed by HUVEC in co-culture with MSC, one-way ANOVA+Tukey’s post test, n.s.=p>0.05 (n=6), mean±SEM. (B) Immunofluorescence of vascularized scaffolds, blue=DAPI and red=phalloidin are counted as vessel-like structure from HUVEC, scale bar 300μm. Mean±SEM of VEGF ELISA quantification: respectively (C) D-DFU and D-iDFU passive release 48h (n=3), (D) decellularized scaffolds seeded with control fibroblasts or (E) DFU for 3 weeks. *=p<0.05 ANOVA (n=3). (F) Tubule formation assay of HUVEC in matrigel in contact with conditioned media of control fibroblasts or DFU seeded scaffolds for 6h, representative images. (G) Quantification of F) in number of vessels formed, (H) average length and (I) total length of conditioned media derived from control fibroblasts or DFU 3 weeks seeded scaffolds, *=p<0.05 one-way ANOVA + Tukey’s (n=3).

When comparing the two approaches developed here, scaffolds with ECM incorporated during fabrication (Coll-DFU and Coll-iDFU) can additionally allow for other scaffold modifications during fabrication (e.g., controlled drug delivery of an antibiotic). However, from a pure regeneration standpoint, we feel that the decellularization technique has a higher potential for success (despite future improvement to prevent the material loss seen): it reduces the amount of cells required; it allows growth in a 3D environment; and deposits the rejuvenated ECM from iPS-reprogrammed fibroblasts in a more natural configuration. Given that primary patient-derived cells might not be as efficient as a cell line in producing high amounts of material, this approach is preferred for a personalized medicine setup. In addition, the decellularization technique, produces scaffolds with similar structure, pore size and Young’s modulus to the control CG scaffold, which are parameters that have previously been optimized and translated to clinical products ^{[11b, 11c, 25, 32c, 37]}. By having the rejuvenated matrix coated onto the decellularized scaffolds here, there appears to be advantages to cell growth, ECM deposition (within 3 weeks) and vascularization. These standard parameters can be further monitored in an in vivo model and compared to previous studies ^{[15a, 23a, 24, 27, 33]}. The promising in vitro readouts from our study are consistent with previous work that implanted reprogrammed patient cells –which had been cultured in advance so that cells were embedded in their own matrix –in a diabetic mouse model. Those data demonstrated that the iPS-reprogrammed group outperformed the diabetic source cell group^[13d]. Specifically, we saw increased efficiency in ECM deposition (evenly distributed) in using patient matched matrix activated scaffolds versus a previously studied scaffolds, using a “standardized” fibroblasts cell line, speaking in favor of this personalized approach ^[23a].

Interestingly, the data in this manuscript suggests that pathological DFU fibroblasts have a higher ECM deposition response when we matched them with their own iPS-reprogrammed ECM. This leads us to speculate a sort of signature or an ‘ECM memory’ is somehow present in these fibroblasts and that diabetic fibroblasts may perform better in tissue from a diabetic background ^[31]. We note, however, that the feasibility and potential demonstrated in this manuscript, along with the fabrication techniques should be further tested in additional source and reprogrammed cells to confirm the universality of this approach. We tested whether this was a due to an immune response ^[30]; however, we detected only cytokine variation responses in a minimal range, suggesting no cytokine storm nor adverse reactions (Figure S8–S9). This point is particularly important as it adds further credence in favor of personalized medicine approaches having potential in DFU treatment and it opens up exciting new avenues for investigation.

3. Conclusion

Here we have demonstrated for the first time the ability to form pro-regenerative personalized porous scaffolds using the ECM from pathological fibroblasts cycled through iPS reprogramming. We have shown the feasibility of using iDFU rejuvenated fibroblasts (from a DFU patient origin), using them as source of rejuvenated material.

Two engineering approaches were successfully developed to do so: 1) direct ECM blending into scaffold fabrication; and 2) seeding scaffolds with reprogrammed fibroblasts for ECM deposition followed by decellularization. Monitoring ECM deposition and in vitro vascularization as readouts for improved regeneration, we observed the decellularization approach is appealing for cases in which low numbers of primary cell are available and we have shown that patient-matched ECM deposited directly by iDFU cells improves ECM deposition by DFU and enhances vascularization. Interestingly, DFU cells outperformed healthy cells in several assays, suggesting some ECM memory of the diabetic signature may be advantageous.

This work demonstrates the feasibility of using patient pathological DFU fibroblasts can indeed be stimulated to regenerate tissue when provided with the right engineered ECM-scaffold environment, motivating this work and others for the identification of the optimum acellular biomaterial for further DFU repair.

4. Experimental Section

Cell culture expansion:

All biopsy-derived samples and cells were isolated from deidentified discarded skin specimens collected under a protocol approved by the Beth Israel Deaconess Medical Centre Institutional Review Board (#2011P000226). Patient selection and skin biopsies were obtained after patients provided informed consent, and subsequently fibroblasts were isolated from biopsies. The source diabetic foot ulcer cells were selected based on the inability of these fibroblasts to generate a complete skin layer, as previously described ^{[13a, 13d, 16b]}. Biopsy-derived non-diabetic patient foot fibroblasts (control fibroblasts), and diabetic foot ulcer fibroblasts (DFU) were used between P4 and P8. The same diabetic foot ulcer cells were cycled through induced pluripotent stem cell-reprogramming and re-differentiated back into fibroblasts (iDFU); these cells were used between P12 and P14 ^[13d]. Control fibroblasts and DFU were expanded in expansion media: DMEM (life technologies) comprising 10% FBS, 1% P/S (Sigma) and 1% HEPES (Life Technologies) ^{[9a, 13d, 23a]}. iDFU were expanded and cultured in differentiation media on recombinant human collagen I (Sigma), diluted 1:100, coated flasks, 3:1 DMEM:F12 (Life Technologies) with 5% Fetal Clone II serum (Hyclone), 0.18 mM adenine (Sigma), 8 mM HEPES (Sigma), 0.5 μg/mL hydrocortisone (Sigma), 10⁻¹⁰M cholera toxin (MP Biological), 10 ng/mL epidermal growth factor (Austral Biological), and 5 μg/mL insulin (Millipore) ^{[9a, 13d, 27]}. All cell types were at 80% confluence prior to use, and were regularly checked for mycoplasma infection (Mycoalert kit^™, Lonza). Expansion and differentiation media were changed every 72h, with ascorbic acid supplementation at 100 μg/mL.

Cytofluorimetric analysis:

All cell types were fully described and characterized previously [9a, 9b, 13a, 13c, 15, 17]. Cytofluorimetric analysis was conducted to check no phenotypic drifting had occurred, as follows: cells were trypsinized, counted, washed with PBS, and blocked with 2% heat-inactivated FBS in PBS. Physical gating parameters were selected on unstained fibroblasts. The samples were then incubated with the following phycoerythrin-conjugated primary antibodies: anti-CD10, anti-CD31, anti-CD34, anti-CD73, anti-CD90, anti-CD105, and anti-CD140b (PE or FITC conjugate Biolegend). The data were collected with AttuneNXT (Life Sciences); 10,000 events were gated per sample and analyzed with Flowjo (v.10)

Blended scaffold fabrication:

ECM-activated scaffolds were fabricated according to methods previously described ^{[23a, 33b, 38]}. Briefly, control collagen-glycosaminoglycan (CG) slurry was produced by blending bovine tendon collagen I (5mg/mL with chondroitin-6-sulphate (GAG=0.04%) (Sigma-Aldrich, Germany) in 0.05 M acetic acid ^[39]. DFU and iDFU ECM was generated in T175 flasks, by seeding the cells at a concentration of 64k cells/cm² in differentiation media containing 100 μg/mL ascorbic acid (changed every 72 hours) for 3 weeks. After washing in PBS, the 2D cell monolayer was scraped off the flask, centrifuged, and after supernatant removal, snap frozen in liquid nitrogen. Lyophilization occurred overnight at −56°C and 0.0180 mBar pressure. Samples were collected and weighed prior to blending in acetic acid ^[23a]. Coll-DFU and Coll-iDFU scaffolds were generated by blending 2.5mg/ml collagen I with 2.5mg/ml of matrix (Coll-DFU from DFU ECM, Coll-iDFU from iDFU ECM) in acetic acid. After 3 cycles of degassing at 4Torr, the slurries were freeze-dried in stainless steel moulds (6 mm diam.) at −10°C, (60 min ramp) and sublimated at 100 mTorr at 0°C for 17 hr ^[23a]. All scaffolds were cross-linked and sterilized using physical dehydrothermal crosslinking and chemical crosslinking using 1-ethyl-2-(3-dimethylaminopropyl)carbodiimide (EDAC) (6mM per gram of collagen). N-Hydroxysuccinimide (NHS) was used as a catalyst with EDAC for the crosslinking reaction at a molar ratio of 5:2 EDAC:NHS ^[38]. Scaffolds were stored sterile at 4°C and used within 5 days of crosslinking ^[24].

Decellularization of scaffolds with direct ECM deposition:

To generate decellularized scaffolds, control CG scaffolds (physically and chemically crosslinked) were seeded for 3 weeks with DFU or iDFU in differentiation media +100μg/mL ascorbic acid. Decellularization was adapted from combining a physical (1^st step) and an enzymatic (2^nd step) techniques described in this review ^[40]. Method combination results in a powerful decellularization with structure preservation (see description of decellularization protocols in ^[34a]). Seeded scaffolds have been placed in ddH20 to induce osmotic shock and rotated for 1h at RT before snap freezing in liquid N2. Scaffolds were lyophilized overnight (0.018 mmBar and −57°C) before rehydrating (freeze-thaw and osmotic shock method extracted from ^[34]). Subsequently, scaffold batches were rehydrated and washed in ddH2O for 24h, rotating at RT. However, effective, this is not sufficient to eliminate genomic DNA, which require an additional enzymatic step, which includes the use of a gDNAse at 37°C ^{[34a, 40]}. Other methods, such as the use of detergents (SDS, TX100), were not considered as reported to disrupt protein folding and GAGs ^[41]. Scaffolds were placed in 500μL solution containing gDNAse (Qiagen) 1:7 in ddH2O for 8h in a thermomixer (Eppendorf) at 42°C, 1500 rpm. Samples were then collected and washed for 16h, rotating in ddH2O at RT, before sterilizing in EtOH 70% (enzymatic method isolated from ^[41a]). The final decellularized products were lyophilized for EM, tested for GAG content, and cryosectioned to evaluate extracellular matrix constituents after the decellularization process. Following confirmation of the decellularization process, these scaffolds were prepared for recellularization by cross-linking and sterilizing them (EtOH 70%). Chemical crosslinking occurred using 1-ethyl-2-(3-dimethylaminopropyl)carbodiimide (EDAC) (6mM per gram of collagen). N-Hydroxysuccinimide (NHS) was used as a catalyst with EDAC for the crosslinking reaction at a molar ratio of 5:2 EDAC:NHS ^[38]. Scaffolds were stored at 4°C and used within 5 days of crosslinking ^[24].

Scaffolds seeding:

For the 3D studies, CG control scaffolds, ECM activated scaffolds (Coll-DFU and Coll-iDFU), decellularized patient in situ scaffolds (D-DFU and D-iDFU) were seeded with patient matched DFU or control fibroblasts; the control fibroblasts were not from the same patient. All cells were seeded at 250,000 cells per scaffold (6mm diam.×4mm) ^[23a] for 24h in expansion media. Thereafter, from t = 0 (i.e., 24h after seeding) to 3 weeks – when samples were ready for collection – media was changed every 72h.

Immuno-fluorescence staining:

The methods used to assess regeneration include quantification of deposited proteins using immunofluorescent staining and in vitro vascularization assays developed previously, and translatable to other scaffolds and future in vivo efficacy evaluation ^{[7–8, 15a, 27]}. Seeded scaffolds were collected after 3 weeks of culture and cryo-embedded in Optimal Cutting Temperature (OCT) compound (TissueTek). Samples were sectioned into 7μm slices using a Leica microtome. Sections were then fixed in PFA 4%, blocked with 3% BSA (sigma) and stained for Collagen I, Collagen III, Collagen IV, Fibronectin 1, LAMA1, LAMA5 following previously published protocol and conditions per antibody ^{[23a, 27]}. Images were captured using an epifluorescence microscope (Nikon Eclipse 90i & Plan Fluor 20x Objective) keeping the acquisition/microscope setting constant for each stain to ensure the intensity results are comparable, merged and assembled (n=3). The fluorescence intensity of each marker was analyzed using an ImageJ intensity macro and the results were graphed in Graphpad Prism. Integrated density has been reported in all graphs ^{[23a, 33a, 42]}. Briefly, background pixels were eliminated and only illuminated pixels were counted towards the average. This allowed a more normalized quantification between different areas. Immunofluorescence quantification was preferred over western blotting to allow us to test whether proteins are uniformly coated on the backbone of the scaffolds, even if there are relatively small amounts. Additionally, the assay is not disruptive (i.e., degeneration during processing can make the small amounts protein undetectable) and this approach allows us to compare it with our previous results on cell line and other studies ^{[8a, 15a]}. CG background subtraction has been applied only to empty controls after decellularization to quantify remaining proteins after the process; this was not applied to collagen I as scaffolds are blended with collagen I. Note this subtraction was not done for any of the seeded scaffolds as cells can add or remove the other matrix elements (e.g., collagen III, IV, fibronectin) to the scaffolds. After analysis, images intensity was increased to ensure the staining was visible in printout ^{[23a, 27]}

DNA and GAG analysis:

For DNA and GAG analysis, samples were collected from the well or scaffold in 1mL digestion buffer (Papain 1mg/mL, Sigma) according to manufacturer’s instructions, and left overnight at 67°C. Samples were evaluated for DNA (Picogreen, Invitrogen) and GAG content (Biocolor) at time 0, week 1 and week 3 (n=3).^[23a].

RNA analysis:

RNA was extracted by lysing the scaffold with TissueLyser II (Qiagen) in Qiazol or by scraping the tissue culture well in Qiazol, and subsequently following Qiagen column extraction procedure for both sample types. Purified RNA was then quantified (Nanodrop 2000) and retrotranscribed at an initial concentration of 500ng/mL. The resulting cDNA was analyzed for Collagen I, Collagen III, Collagen IV, Fibronectin 1, LAMA1, LAMA5, bFGF, MMP3, VEGF and GAPDH at a final concentration of 0.67ng/μL per well (96 well plate) using Realplex realtime PCR as per the manufacturer’s instructions. Real-time cycles were analyzed in excel using the 2−ΔΔCT method, compared to normal fibroblasts, normalized according to GAPDH expression. (n=3; time points t=0 seeding day and 3 weeks after seeding). All primers were optimized (Primertech Qiagen) and used as previously described ^[23a].

Western blot analysis:

Western blotting analysis was carried out as previously described on lysates of 2D culture only ^{[23a, 27]}. Briefly, the lysing solution consisted of: 4% sodium deoxycholate (Sigma) in 20 mM Tris, pH 8.8, and 1% SDS + protease inhibitor cocktail (Sigma). Protein concentration was determined using bicinchoninic acid (Pierce BCA, Thermo Fisher Scientific). Protein samples were loaded at the standard quantity of 5μg in 15μL volume (7μL 2x Laemmli buffer-β-ME, Biorad, + 7 μL protein suspension) resolved on 4–20% TGX Stain-Free Protein Gels (Bio-Rad Laboratories), under denaturing and reducing conditions, and transferred to the blot membrane (Bio-Rad Laboratories). The membranes were blocked with 3% fat-free milk in Tris-buffered saline and Tween 20 (TBS-T) for 30 minutes at room temperature and incubated with primary antibodies against Collagen I (COL1A1 Antibody mouse anti-human primary antibody, Santa Cruz SC293182) 1:250 in TBS-T 3% BSA, 4°C overnight; Fibronectin 1 (FN1 Antibody mouse anti-human primary antibody, Santa Cruz SC-8422) 1:500 in TBS-T 3% BSA, RT 2h; Collagen III (COL3A1 Antibody rabbit anti-human primary antibody, SIGMA-HPA007583) 1:1000 in TBS-T 3% BSA, 4°C overnight; β-Actin (Antibody rabbit anti-human primary antibody, SIGMA-A2066) 1:10000 in TBS-T 3% BSA, RT 1h; Secondary Goat-HRP conjugated (anti-rabbit SIGMA-A0545), (anti-mouse SIGMA- 12–349) 1:10000 in TBS-T 3% BSA, RT 1h. The membranes were then washed with TBS-T and incubated for 1 hour at room temperature, 1% Milk-TBS-T in respective goat anti-mouse antibody. The membranes were developed using ECL (Bio-Rad Laboratories) and digital images acquired at Hamersham 600 imager according to the saturation time. Digitalized TIFF images were then analysed for intensity signal vs control and normalized to actin, in ImageLab (BioRad).

Mechanical testing:

Uni-axial compression testing was performed as previously on a Zwick mechanical tester (Z050, Zwick-Roell, Germany; 5N load cell) to evaluate the bulk mechanical properties of the decellularized scaffolds as previously described ^[23a]. All samples were tested in unconfined compression and were hydrated before and throughout testing with PBS. The compressive moduli of the scaffolds were obtained by using a linear fit of the stress/strain curve, between 2 and 5 percent strain. For each scaffold type, 6 samples were tested (n=6).

Scaffold imaging using Scanning Electron Microscopy:

The ultrastructure of the freeze-dried and decellularized scaffolds was observed using Scanning Electron Microscopy (EM) as previously described ^[23a]. Dry samples were sputter coated using a polaron sputter coater (Quorum Tech., UK) with a gold/palladium target and observed from a secondary electron detector using a Mira Variable Pressure Field Emission Scanning Electron Microscope (Tescan, USA) at an accelerating voltage of 5kV. Column width and the pore size of the scaffold micro structure was quantified by manually assessing SEM images (n=3 scaffolds) using an ImageJ quantification tool. ^[27]

Assessment of VEGF expression:

Supernatants from cultured scaffolds were collected at the 3-week time-point (48 hours after the last media change) and centrifuged to remove any cellular debris. Samples were stored at −80°C until analysis. Samples were tested for the presence of VEGF using sandwich ELISA following the manufacturer’s instructions (Biotechne Ltd., UK).

Assessment of scaffold vascularization:

An assessment of in vitro vascularization in the scaffold was performed as previously described ^{[33b, 43]}. Briefly, human umbilical vein endothelial cells (HUVEC) (CC-2519 Lonza, Switzerland) and human bone marrow-derived (iliac crest) mesenchymal stromal cells (hMSC; Lonza Biologics PLC) were co-cultured at a ratio of 4:1. hMSC were only seeded 3 days after initial seeding with HUVEC. Samples were cultured with EGM-2 medium (Lonza, Switzerland) supplemented with 2% FBS for the first 3 days (HUVEC only), later changed to 20% FBS supplementation once hMSc were added to the system. Media was replaced twice a week. After a total of 10 days culture, samples were fixed in buffered formalin, sliced in half and stained with Phalloidin-Atto 488 (Sigma) and DAPI readymade solution (Sigma) according to the manufacturer’s instructions. Transversal images were acquired after visualization using a Carl Zeiss LSM 710 confocal microscope equipped with a W N-Achroplan 10× (numerical aperture 0.3) objective lens. One image was taken from the center of each sample in addition to two images from the periphery of the same scaffold. Vessel-like structures were considered positive when a tubular structure was stained with phalloidin. These structures derived from HUVEC were previously shown to be positive for the endothelial cell marker CD31 ^[33]. Vessel-like structures were manually counted using Fiji software ^[42]. Numbers (n=6) and time-point (d10) were evaluated based on literature and on our previous published protocol ^{[23a, 32d, 33, 44]}.

Tubule formation assay on soluble factors released by scaffolds:

120 μL/cm² of Matrigel R (corning) was solidified at 37°C for 30 min. 30k HUVECs were seeded per cm². A tubule formation assay was performed for 6hwith 500 μL conditioned medium collected from seeded scaffolds and tested on HUVEC seeded on plastic.. EGM-2 supplemented with 2% FBS were used as a positive controls. Images were taken after 6h. The presence of intersection points, number of tubes, total tube length and average tube length were manually measured using an Image J measuring tool ^[33a].

Macrophage polarization assay:

Experiments and data analysis/presentation were conducted as previously published ^{[33b, 45]}. Briefly, THP-1 monocytes were expanded in THP1 cell culture medium [RPMI-1640 R8758 containing 10% Fetal Bovine Serum (FBS) (South American Origin, LabTech, United Kingdom), 1% Penicillin/Streptomycin, 10mM hydroxyethyl piperazineethanesulfonic acid (HEPES), 1mM Sodium Pyruvate, 2.5g/L Glucose and 0.05mM Mercaptoethanol]. Cells were routinely tested for mycoplasma contamination. 500k THP1 monocytes were seeded on scaffolds following a protocol developed in our laboratory. Briefly, each side was seeded with 25μl of 250,000 cells and 20ng/ml phorbol 12-myristate 13-acetate (PMA) to promote monocyte to macrophage differentiation, with a 15 minute incubation between the seeding of each side. This was followed by the addition of THP1 growth medium with 20ng/ml PMA following which scaffolds with cells were cultured overnight at 37°C and 5% CO₂. Fresh growth media was added after overnight incubation and removal of the PMA medium for 6 hours. This was followed by addition of growth medium (M0), plus 5ng/ml Interferon-γ (IFNγ) and 100ng/ml lipopolysaccharide (LPS) for M1 induction or 20ng/ml of Interleukin-4 and 13 for M2 induction. Cytokines were measured via ELISA (R&D Systems) for VEGF (DY293B), IL1β (DY401), IL1ra (DY280), TNFα (DY210), MIP1α (DY270), MCP1 (DY279) according to manufacturer’s instructions.

Data collection and statistics:

All data were collected with RAW (including images, excel tables and hard samples) and files were analyzed in excel. Statistical analysis was conducted using Graphpad Prism on n=3 samples. The graphs generated express averages ± standard error mean. Where 3 groups were analyzed, one-way ANOVA and Tukey’s post-test were used. Where more than three groups were analyzed, the more stringent Bonferroni correction was used. Results were considered significant when p<0.05. This was indicated on the graphs as *. For other datasets, the statistical analysis used was reported in the figure legend. Figures were assembled in vector form using Inkscape software ^{[23a, 27]}.

Table 1.

Cells and scaffolds used in the study (with relevant sections in paper highlighted)

Name	Description of Scaffold or Cell	Notes	Paper sections
CG	Collagen-glycosaminoglycan scaffold	Acellular	3.1 to 3.9
Control Fibroblasts	Control cells derived from biopsies from non-diabetic patients	-	3.1 to 3.9
DFU	Patient-derived diabetic foot ulcers fibroblasts isolated from debridement	Expanded in 2D plastic to produce ECM for blending; or seeded on CG	3.1 to 3.9
iDFU	DFU cycled through inducible pluripotent stem cell reprogramming (iPS) and re-differentiated into fibroblasts	Expanded in 2D plastic to produce ECM for blending; or seeded on CG	3.1 to 3.9
Coll-DFU	Freeze-dried scaffold using a blend of Collagen I and ECM from DFU	Results in an acellular scaffold that is subsequently seeded (control or DFU fibroblasts)	3.1 to 3.4
Coll-iDFU	Freeze-dried scaffold using a blend of Collagen I and ECM from iDFU	Results in an acellular scaffold that is subsequently seeded (control or DFU fibroblasts)	3.1 to 3.4
D-DFU	Decellularized scaffold formed by seeding a CG scaffold with DFU and decellularizing	Decellularized scaffold is subsequently seeded (control or DFU fibroblasts)	3.5 to 3.9
D-iDFU	CG scaffold seeded with iDFU and decellularized as per section 3.5;	Decellularized scaffold is subsequently seeded (control or DFU fibroblasts)	3.5 to 3.9