An Antimicrobial, Cell‐Adhesive, and Growth Factor‐Producing Living Material Engineered From Bacillus subtilis Biofilms for Wound Healing

ABSTRACT

Chronic wounds represent a burden for healthcare systems and critically impair quality of life of patients. Engineered living materials (ELMs) show great potential as scaffolds and delivery systems in tissue regeneration. As a generally regarded safe biofilm‐producing organism, Bacillus subtilis (B. subtilis) presents interesting properties to develop functional ELMs in biomedicine. In this study, we engineer B. subtilis to design a protein‐based living matrix eluting bioactive sequences for wound healing. Accordingly, we create a biofilm‐deficient B. subtilis strain that expressed the biofilm scaffold protein, TasA in fusion with a cell‐adhesive motif (TasA‐PRb), which results in biofilms with cell‐adhesive properties for fibroblasts and keratinocytes. Besides, B. subtilis is known to elicit antimicrobial properties against strains of E. coli and S. epidermidis, which are not impeded by biofilm engineering. We also incorporate secretion of fibroblast growth factor 2 (FGF2), known to promote wound healing, in the biofilm‐deficient strain. Co‐culture of FGF2‐expressing and TasA‐PRb producing strains resulted in biofilms secreting bioactive FGF2. Chronic wound healing assays in diabetic mice reveal accelerated wound closure in mice treated with the TasA‐PRb/FGF2 biofilms, with absence of observable adverse effects. This underlines the potential of the B. subtilis biofilm for the design of functional ELMs in biomedicine.

Bacillus subtilis is engineered to produce cell‐adhesive biofilms by functionalization of its biofilm scaffold protein TasA with a cell‐adhesion sequence. Living bacteria within the biofilm are also engineered to secrete growth factors and natively produce antimicrobial peptides (AMPs), resulting in a cell‐adhesive, antimicrobial, and growth factor‐producing biofilm whose topical application promotes wound healing in diabetic mice.

1. Introduction

Chronic wounds represent an important burden on healthcare systems around the world, as their global prevalence is estimated at more than 2 per 1000 population [1]. Chronic wounds can lead to amputations and cause physical and psychological trauma, thus impairing the quality of life of affected patients [2]. Wounds usually heal in four stages: hemostasis, inflammatory, proliferative, and remodeling. In chronic wounds, transition from the inflammatory to the proliferative stage is halted, and these wounds are characterized by decreased cell proliferation, angiogenesis, oxygenation and deposition of extracellular matrix (ECM), increased inflammation, elevated proteolysis, and high risk of infection [3, 4, 5]. Several studies have demonstrated the value of adding growth factors to chronic wounds to accelerate their healing [6, 7, 8, 9], as well as antimicrobial molecules to prevent, or treat, infection [10]. However, due to the highly proteolytic environment of chronic wounds, polypeptide growth factors show a diminished half‐life, thus limiting their therapeutic effect. In this context, matrices and hydrogels have emerged as promising treatment options since they can act simultaneously as a scaffold for cell proliferation and as a delivery system extending the activity half‐life of therapeutic molecules [11, 12, 13, 14, 15]. Living materials, which are a type of scaffold that contains living cells, have shown interesting properties for wound healing such as stimuli responsiveness, possibility for sustained secretion and release of therapeutic molecules and self‐regenerative properties [16, 17, 18, 19].

Bacillus subtilis is a candidate of interest for the design of engineered living materials for health‐related applications since it is a biofilm‐forming organism generally regarded as safe (GRAS) [20, 21]. Notably, it does not produce proinflammatory lipopolysaccharides and is known to produce antimicrobial peptides (AMPs) [22, 23]. This well‐characterized organism is also adapted for protein expression and secretion, which enables the development of a living matrix that can regenerate the protein scaffold and secrete therapeutic proteins such as growth factors. A few studies have shown the usage of other probiotics for the delivery of growth factors in the wound [24, 25], but the potential of B. subtilis and its biofilm for biomedical applications remains poorly studied. In this study, we engineered B. subtilis to produce a biofilm‐based cell‐adhesive and growth factor‐eluting matrix for promotion of wound healing. We co‐cultured a strain of B. subtilis engineered to produce its biofilm scaffold protein (TasA) in fusion with a cell adhesion sequence (PRb) [26] with another strain that we engineered to secrete fibroblast growth factor 2 (FGF2), previously shown to promote wound healing upon topical application [8]. Co‐culture of these genetically engineered bacteria resulted in the formation of TasA‐PRb based fiber‐rich matrices that secrete FGF2. These matrices showed cell‐adhesive properties toward fibroblasts and keratinocytes and antimicrobial properties against strains of Escherichia coli and Staphylococcus epidermidis as well as cell‐migration activity toward fibroblasts. Topical application of this matrix to diabetic full‐thickness mice wounds accelerated wound closure and improved re‐epithelialization, angiogenesis, and collagen deposition.

2. Results and Discussion

2.1. Engineering and Characterization of the B. subtilis Biofilm

In this study, we used the 3A38 B. subtilis strain (assigned as wild type (WT) throughout the study), which is a biofilm‐forming and competent strain derived from the ancestral NCIB 3610 strain [27]. The protein TasA constitutes the scaffold of the biofilm and assembles into fibers through N‐terminal donor strand complementation [28]. The native biofilm also contains high levels of exopolysaccharides (EPS), whose presence favors expression and incorporation of the BslA protein, which confers hydrophobicity to the biofilm [29]. In this project, we use TasA fibers as a scaffold that mimics the ECM of the skin to promote cell attachment. Indeed, attachment and proliferation of endogenous cells is a key requirement for materials in the field of tissue regeneration [30]. To promote adhesion of endogenous cells to the biofilm scaffold, we created a fusion TasA protein incorporating the PRb fibronectin‐mimetic cell adhesion sequence at its C‐terminus (TasA‐PRb). PRb contains the RGD motif and its well‐characterized synergistic motif PHSRN and is known to bind integrins, specifically, α5β1 [26]. Binding of this integrin is involved in re‐epithelialization and angiogenesis, two processes critical for wound healing [31, 32]. It has been shown that TasA tolerates the C‐terminal genetic fusion of protein sequences of up to 603 residues without impairment of fiber assembly, further justifying the fusion of the PRb peptide at its C‐terminus [33]. Since the presence of EPS could reduce cell adhesion by increasing the hydrophobicity of the biofilm and shielding TasA‐PRb fibers [29, 34], we created a B. subtilis strain with a deletion of the epsABCDEFGHIJKLMNO (eps) gene cluster, responsible for EPS production. The native tasA gene and its repressor, sinR, which is adjacent to tasA, were subsequently deleted to create a double deleted ΔepsΔst expression system capable of expressing functionalized TasA‐based biofilms. Inducible plasmids containing the tasA or tasA‐PRb genes were then transformed into this expression system. To further promote cell migration and proliferation, and re‐epithelialization, we also transformed the double deleted strain with plasmids encoding for secreted FGF2 growth factor, whose topical application has been shown to favor wound healing [6, 35]. Strains expressing TasA‐PRb and FGF2 were co‐cultured. Additionally, B. subtilis is known to secrete antimicrobial molecules, which could prevent wound infection. Overall, engineering of the B. subtilis biofilm allowed the development of a living matrix that produces a cell‐adhesive protein scaffold and secretes a growth factor and antimicrobial peptides (AMPs) (Figure 1).

FIGURE 1.

Design of functional Bacillus subtilis biofilms for wound healing. Strains with deletion of the biofilm‐associated gene clusters epsA‐O and tasA‐sinR were transformed with plasmids encoding for a TasA‐PRb fusion or secreted FGF2, and co‐cultured, resulting in a PRb‐functionalized biofilm eluting a growth factor and AMPs.

Mesoscopic observation of biofilms cultivated in monosodium‐glutamate‐glycerol (MSgg) liquid medium and on MSgg agar (Figure 2A) showed that deletion of eps resulted in the loss of the characteristic wrinkled morphology of B. subtilis biofilms, as previously described [36]. As expected, the Δeps strain was still able to form a biofilm, mainly composed of TasA fibers, as observed by scanning electron microscopy (SEM) and SDS‐PAGE, where the most abundant protein had a molecular weight slightly above 25 kDa, corresponding to the 25.7 kDa molecular weight of tasA in B. subtilis (Figure 2B). Accordingly, the total quantity of the Δeps biofilm cultured in 12‐wells plates (21.6 ± 4.1 mg) was lower than that of the WT biofilm (43.3 ± 4.1 mg), but the proportional protein quantity within the biofilm was higher for the Δeps strain (14.8 ± 1.8 µg/mg) than the WT strain (6.0 ± 2.3 µg/mg), accounting for the loss of EPS in the biofilm (Figure 2C). As expected, deletion of eps also decreased the hydrophobicity of the biofilm, due to the absence of the BslA hydrophobin, known to interact with EPS [29]. Further deletion of sinR and tasA resulted in a biofilm‐deficient strain for which no fibers and no TasA production were observed by SEM and SDS‐PAGE, respectively. In this strain, the most abundant protein shows a molecular weight slightly under 37 kDa, which could correspond to B. subtilis flagellin, Hag [37]. The total quantity of material collected after growth in the same conditions as the biofilms (13.2 ± 1.9 mg) was even lower than that of the Δeps strain (21.6 ± 4.1 mg), likely due to the absence of TasA itself, but also to the absence of biofilm resulting in lower retention of other molecules and bacteria upon collection by centrifugation. The proportional protein quantity for this ΔepsΔst strain (10.4 ± 2.4 µg/mg) was lower than for the Δeps strain, which still expresses the tasA protein, but higher than for the WT strain, in which the EPS account for a large proportion of the material. CFU counting also showed a decrease in the number of living cells in the ΔepsΔst culture compared to the WT and Δeps strains. Indeed, bacterial biofilms are known to favor the survival of resident cells in high density by facilitating nutrient trapping and distribution, which could explain decreased growth and viability of the ΔepsΔst cells, especially in a defined culture medium [38].

FIGURE 2.

Characterization of engineered Bacillus subtilis biofilms. (A) Images of biofilms cultivated in liquid MSgg medium (top) and MSgg agar (second row), scanning electron micrographs of biofilms cultivated in liquid MSgg (third row) and water contact angle (WCA) with biofilms cultivated on MSgg agar (bottom). Scale bar for SEM is 1 µm. (B) Detection of TasA and TasA‐PRb proteins in biofilms cultivated in liquid MSgg medium and MSgg agar by SDS‐PAGE. (C) Composition of biofilms cultivated in liquid MSgg in terms of total biofilm mass per 2 mL of culture (left), relative quantity of proteins (middle) and B. subtilis colony forming units (right) per biofilm mass. The data are presented as mean ± SD, n = 3 for WCA and n = 4–5 for composition analyses in (C), ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001 as determined by one‐way ANOVA with Tukey's post‐hoc test. All biofilms were cultivated for 4 days at 30°C without agitation.

Re‐introduction of the tasA gene in the ΔepsΔst strain using the IPTG‐inducible pHT01 plasmid restored a phenotype similar to the Δeps strain upon induction. Indeed, the ΔepsΔst + tasA strain, identified as tasA, was able to produce a defined biofilm, as confirmed by SEM. The total quantity of biofilm (20.9 ± 3.4 mg) and proportional protein quantity (14.8 ± 3.3 µg/mg) were also restored to levels equivalent to the biofilm produced with the Δeps strain. The same expression system transformed with the tasA‐PRb containing plasmid showed a similar mesoscopic morphology of the biofilm. SDS‐PAGE analysis showed a protein with a molecular weight slightly higher than TasA and corresponding to TasA‐PRb (28.4 kDa). As expected, no biofilm production and little to no TasA and TasA‐PRb expression was observed in liquid MSgg medium for non‐induced ΔepsΔst + tasA and ΔepsΔst + tasA‐PRb (tasA‐PRb) strains (Figure S1), confirming that the SDS‐PAGE bands attributed to TasA and TasA‐PRb are indeed overexpressed upon induction. This also indicates very low leaky expression of the pHT01 plasmid in this expression system. Total material quantities for the non‐induced strains are slightly lower than for ΔepsΔst at 12.4 ± 2.9 mg and 10.6 ± 0.8 mg for the tasA and tasA‐PRb strains respectively, likely due to the presence of chloramphenicol in the medium and replication of the pHT01 plasmids slowing down bacterial growth. Additionally, culture of both strains in biofilm‐promoting conditions showed higher number of living cells than the biofilm‐deficient strain, highlighting the benefits of growing B. subtilis as a biofilm for increased growth and/or viability of this living material.

Multiple biofilm culture conditions were tested (Figure S2), but cultivation in standard MSgg medium over 4 days was kept as the optimal conditions, as it allowed robust production of biofilm composed of TasA‐PRb fibers. MSgg is indeed known to promote biofilm formation [39]. Although cultivation in LB medium should allow inducible TasA expression, it did not allow biofilm formation and no TasA was detected in the material collected by centrifugation. This could be due to a decreased expression of biofilm‐associated protein machinery, including TapA, which anchors TasA to the bacterial membrane and nucleates its assembly into fibers [38, 40]. Cultivation over longer periods of time such as 7 days in MSgg seemingly increased TasA expression, although it resulted in biofilm breakdown. Biofilm disassembly is the final stage of the biofilm cycle and is associated with higher production of proteases during the late stages of biofilm formation [38, 41].

Rheological analysis of the biofilms cultivated in liquid MSgg medium was conducted. The linear viscoelastic (LVE) range of each biofilm was determined by a strain sweep experiment (Figure S3). For the WT biofilm, the LVE was observed between approximately 0.05% and 10% strain, whereas the other biofilms, lacking EPS, all showed LVEs from 0.05% to 1% strain. The LVE could not be determined for the ΔepsΔst strain because it does not form a biofilm and is therefore an easily deformable cell suspension. Frequency sweeps were accomplished at 0.1% strain, which is within the LVE for all biofilms (Figure 3A). For all biofilms, the storage modulus (G’) is higher than the loss modulus (G″), indicative of their solid‐like behavior, likely conferred by the protein fiber network. With a G’ around 200 Pa, WT biofilms show higher stiffness than Δeps biofilms, with a G’ of around 25 Pa, which is in accordance with the more organized structure and fiber network observed macro‐ and microscopically for the WT biofilm. Surprisingly, although the Δeps and tasA biofilms share similarities (Figure 2), the tasA biofilm shows a higher stiffness, which could be due to the presence of a higher number of living bacteria within the biofilm affecting organization of the fiber network. As the tasA‐PRb biofilm shows the weakest G’ of the tested materials, the addition of the PRb sequence at the C‐terminus, although it does not impair tasA fiber assembly, disrupts higher organization of the biofilm. Cyclic strain time sweep experiments showed transition from a solid‐like material to a flowing state at higher strain (100% strain), with rapid restoration to their initial solid‐like properties upon decrease of the strain (Figure 3B). As previously reported for B. subtilis biofilms cultivated on MSgg agar, the WT biofilms cultivated in liquid medium showed better reversible deformability as observed by their instantaneous recovery, while the other biofilms showed recovery after 8–30 s. According to these previous reports, this slower recovery is associated with better suitability for 3D printability, which could be of interest in the field of tissue engineering [33].

FIGURE 3.

Rheological properties of the living materials. (A) Frequency sweep and (B) cyclic strain time sweep. All biofilms were cultivated for 4 days in liquid MSgg medium at 30°C without agitation.

2.2. Modified Bacillus subtilis Strains Retain Their Antibacterial Properties

Chronic wounds are associated with higher risks of infection by bacterial pathogens, which further impair wound healing [42]. This emphasizes the importance of developing wound dressings with antimicrobial properties. B. subtilis is known to naturally secrete antimicrobial molecules that allow it to compete with other microorganisms [23, 43]. We therefore evaluated whether the modified biofilms retain B. subtilis’ natural antimicrobial properties against strains of Gram‐negative (E. coli) and Gram‐positive (S. epidermidis) bacteria. The double‐deleted strain did show a slight but statistically significant decrease in antimicrobial activity against E. coli compared to the WT strain when grown on Mueller–Hinton agar, as observed by cross‐streaking assay, and when grown as a biofilm, as observed by well diffusion assay. However, in both cases, the antimicrobial activity for induced tasA and tasA‐PRb strains was similar to the WT strain (Figure 4A,C). The antimicrobial compounds produced by Bacillus subtilis include mostly antimicrobial peptides such as subtilin and surfactin [23, 44], whose production should not be affected by modifications in the biofilm‐associated genes. This explains why most biofilms expressed in this study retain a similar antimicrobial activity to the WT strain. The slight decrease in antimicrobial activity observed for the ΔepsΔst strain could be due to the decreased number of living cells in the ΔepsΔst culture (Figure 2C) or to the absence of TasA, which has been reported to elicit antimicrobial activity against E. coli strains [45]. Mueller‐Hinton medium appeared to induce biofilm formation, and TasA expression, as indicated by the wrinkled morphology of the WT streak on agar, even though this medium is not identified in the literature as biofilm‐promoting for Bacillus (Figure S4). No significant differences in antimicrobial activity against S. epidermidis were observed for the modified strains compared to the WT (Figure 4B,D), further confirming that modification of B. subtilis biofilm‐associated genes did not impair its antimicrobial activity.

FIGURE 4.

Antibacterial properties of engineered biofilms. Growth inhibition of (A) and (C) E. coli and (B) and (D) S. epidermidis by B. subtilis biofilms observed by (A) and (B) Cross‐streak assay and (C) and (D) wells diffusion assay. The data are presented as mean ± SD, n = 4, * = p < 0.05, ** = p < 0.01, **** = p < 0.0001 as determined by one‐way ANOVA with Tukey's post‐hoc test.

2.3. Engineered Biofilms Exhibit Cell‐Adhesive Properties

Biofilm fibers constitute an interesting scaffold candidate for skin cells in tissue regeneration applications owing to the fact that, as protein unbranched filaments, they are mimicking the ECM of the skin and they are biodegradable. In biomaterial scaffolds for wound healing, full biocompatibility and cell adhesion are also critical. This is why the PRb cell‐adhesion moiety was genetically added to the biofilm scaffold protein TasA. Cell adhesion of HaCaT human keratinocytes at the surface of biofilms was assessed qualitatively using a LIVE/DEAD assay, which also gives indications on the biocompatibility of the living matrix (Figure 5A). This assay revealed high biocompatibility of the biofilms with adhered cells, as no red cells, indicative of damaged plasma membranes, were observed. Furthermore, keratinocytes adhered preferentially on PRb‐functionalized surfaces, including the TasA‐PRb expressing biofilms. This was confirmed by indirect quantification using the resazurin reduction assay on adhered cells after 1 h of culture on the surfaces (Figure 5B). HaCaT cells showed higher adhesion on the double‐deleted strain than the WT biofilms, likely due to the decreased hydrophobicity and increased wettability of the double‐deleted strain surface. Surface hydrophobicity is well‐known to reduce protein adsorption and cell adhesion [34].

FIGURE 5.

Cell‐adhesive properties of engineered biofilms. (A) LIVE/DEAD assay on HaCaT cells following 1 h adhesion on biofilms. Representative images with merge of the Calcein‐AM (Live, green) and Ethidium bromide (Dead, red) are shown. Scale bar is 100 µm (B) Adhesion of L929 mouse fibroblasts (left), 1BR3 human fibroblasts (middle), and HaCaT human keratinocytes (right) to biofilms after 1 h measured using the resazurin reduction assay. Results are expressed as relative adhesion compared to the tissue culture treated (TCT) surface, represented by a dashed line. The data are presented as mean ± SD, n = 4, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001 as determined by one‐way ANOVA with Tukey's post‐hoc test. For L929 and 1BR3 cells, statistics are presented for comparison with the negative control, BSA, as no other comparisons yield statistically significant results.

The engineered biofilms demonstrated robust adhesive properties for HaCaT cells, with the TasA‐PRb‐based biofilm leading to higher cell adhesion than the TasA‐based biofilm. The presence of the TasA fibers‐based scaffold likely favors adhesion by creating a more hydrophilic and densely charged surface. Indeed, engineered biofilms show higher levels of proteins, of which the most abundant is TasA or TasA‐PRb (Figure 2B). These proteins are highly hydrophilic, with gravy scores of −0.751 and −0.814, and 69/234 and 72/265 ionizable residues respectively (Table S4) [46], which should favor protein adsorption and cell adhesion. Higher adhesion of HaCaT cells to TasA‐PRb would be due to the presence of the PRb sequence, which was initially engineered to promote integrin binding and cell adhesion in a fibronectin‐mimetic manner [26]. As observed with HaCaT keratinocytes, all the tested biofilms and the double‐deleted strain showed adhesive properties for mouse (L929) and human dermal (1BR3) fibroblasts (Figure 5B). However, no significant differences in fibroblast adhesion were observed between the different living materials. As keratinocytes and fibroblasts interact during wound healing, cell‐adhesive properties toward both are an interesting feature of the engineered biofilm scaffolds.

2.4. Modified B. subtilis can be Engineered to Secrete Bioactive FGF2

Application of polypeptide growth factors to wounds has been proposed as a strategy to promote chronic wound healing, but the efficacy remains limited by their rapid degradation in the highly proteolytic environment of chronic wounds [5]. Matrix encapsulation [11] and sustained secretion by living cells in situ constitute promising strategies to circumvent this limitation [24]. We therefore aimed at using B. subtilis cells within biofilms as a delivery system for FGF2, which is known to promote wound healing [8]. As the combination of signal peptide with the protein to be secreted significantly affects secretion levels [47], three sequences encoding different signal peptides were evaluated in combination with the encoding sequence of secreted murine FGF2 (Table S3): (i) AmyQ (spA), which has helped with high level secretion of other proteins in B. subtilis [48]; (ii) CnBP, the secretion signal of L. reuteri collagen‐binding protein (spC), which has been used for secretion of other murine growth factors in Gram‐positive bacteria [24]; (iii) the secretion signal of B. subtilis TasA (spT), already used in this study. Sequences encoding for a (GS)_n linker were added between the coding sequences for the secretion signal and the protein to be expressed to favor effective cleavage of the signal sequence and minimal cleavage of FGF2 N‐terminal residues as predicted using SignalP 5.0 [49].

Western blot analyses of double‐deleted B. subtilis strains transformed with inducible plasmids encoding for these fusions indicate that signal peptide CnBP allows the highest level of FGF2 secretion upon cultivation and induction in LB medium as well as in biofilm‐promoting conditions (Figure 6A), as evidenced by the band between 15 and 20 kDa detected in the supernatants of the cultures, which corresponds to 16.3 kDa FGF2. Analysis of the bacterial culture pellets also shows that some cleaved FGF2 remains at the surface of the cells, while a strong band around 20 kDa indicates high levels of un‐cleaved spCnBP‐FGF2 (19.5 kDa) at the surface of or inside the cells.

FIGURE 6.

Secretion and bioactivity of FGF2 from B. subtilis. (A) Western blot analysis of FGF2 secreted by B. subtilis ΔepsΔst using signal peptides AmyQ (spA), CnBP (spC) or TasA (spT) after induction for 5 h in LB medium at 37°C with agitation (top) or 3 days in MSgg liquid medium at 30°C without agitation (bottom). (B) Morphology (top) and Western blot analysis of biofilms resulting from co‐cultures of the spCnBP‐FGF2 strain with inducible biofilm‐forming strains. (C) Migration of L929 fibroblasts in presence of cell‐free supernatants obtained from biofilm cultures after 24 h measured by trans‐well assay. Results are normalized to migration of cells treated with PBS, represented by a dashed line. The data are presented as mean ± SD, n = 4, *** = p < 0.001, **** = p < 0.0001 as determined by one‐way ANOVA with Tukey's post‐hoc test.

As we aimed to produce a cell‐adhesive and growth‐factor secreting living material, we cocultured this FGF2‐producing strain with TasA and TasA‐PRb‐producing strains (Figure 6B). This resulted in biofilms with similar morphologies to the respective single biofilm cultures (Figure 2A). Western blot analysis confirmed that TasA‐PRb/FGF2 cocultures do secrete and encapsulate FGF2, but at lower levels than FGF2 single cultures. As both co‐cultured strains use the same medium, it is likely that each strain grows less than on its own, explaining the reduction in FGF2 production when co‐cultured. Design of a strain expressing both proteins could be considered to address this phenomenon but remains non‐essential to this study aiming at validating the usage of B. subtilis living materials for wound healing. It is to be noted that, although the FGF2‐producing strain does not express TasA, it is still likely attached to the biofilm due to complementation. This phenomenon allows exogenous TasA accumulation and fiber formation at the surface of non‐TasA‐producing cells that still express the TapA anchoring protein, which has not been deleted from any strain in this study [36]. Re‐culture of the FGF2‐producing cultures and cocultures without induction over 3 days still yielded some encapsulated or biofilm‐bound FGF2.

Transwell migration assays with L929 fibroblasts indicated that supernatants of the initial FGF2 culture and co‐culture with the tasA‐PRb strain induce cell migration, confirming the bioactivity of secreted FGF2 (Figure 6C). This activity is similar to the activity of a 2 nM solution of purified FGF2, indicative of concentrations of secreted FGF2 in the low nanomolar range for B. subtilis cultures. The double‐deleted and TasA‐PRb producing strains did not secrete any compounds that induce fibroblast migration. Supernatants from the co‐culture showed a slightly decreased cell migratory activity compared to the FGF2 culture, which would be in accordance with the levels of secreted FGF2 observed by Western Blot, but this difference is not statistically significant. Overall, these results confirm the ability of this B. subtilis expression system to produce living matrices that secrete bioactive growth factors and highlight the importance of selecting optimal signal peptide for efficient secretion.

2.5. Engineered Biofilms Promote Wound Closure in Diabetic Mice

As the modified biofilms showed bioactivity in vitro, their wound healing properties were then evaluated in vivo, in a chronic wound model. Full‐thickness excisional splinted wounds were created on the back of db/db mice, who develop type 2 diabetes and show impairment of cutaneous wound healing [50, 51, 52]. Engineered matrices were topically applied to the wounds three times at 3‐day intervals, and wound area was measured every 3 days for 3 weeks (Figure 7A). Mice treated with the TasA‐based living biofilm or the TasA‐PRb‐based biofilm secreting FGF2 did not exhibit significant adverse effects, as they showed a weight loss of around 5% on average over the first 3 days following wound creation and the first topical application, which was comparable to the weight loss observed for the vehicle‐treated group (Figure 7B). For all groups, the average weight stabilized around 95%–100% of the initial weight for the remainder of the experiment. Very few signs of discomfort were observed for all groups over the first week of the experiment, and biofilm‐treated groups showed a similar average score of discomfort to the vehicle‐treated group. These results indicate that the engineered biofilms cause no notable adverse effects upon topical application to diabetic wounds, which is in accordance with what has been previously reported for incorporation of B. subtilis in wound dressings [53, 54]. In addition, no visual signs of increased inflammation or infection (wound border redness, swelling, suppuration) were detected for wounds treated with biofilms.

FIGURE 7.

Wound healing assay with engineered biofilms in diabetic mice. (A) Experiment timeline. Full‐thickness splinted wounds with a 6 nm diameter were created at day 0, biofilms or controls were topically applied at days 0, 3 and 6, and the wounds were photographed every 3 days until day 21. (B) Weight (left) and signs of discomfort (right) of the mice in the week following wound creation and first topical application. (C) Profile of the wound area reduction over 21 days and statistical significance for comparison of the areas under the curves. (D) Detailed observations of the wound area. Photographs of representative wounds of each group over 21 days (top) and histograms of the wound areas over 15 days. The data are presented as mean ± 95% CI, n = 6 mice for (B), n = 9–12 wounds for (C) and (D), * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001 as determined by one‐way ANOVA with Tukey's post‐hoc test.

The decrease in relative wound area over the course of the experiment was used as an indicator of wound healing. Reduction of the wound area to below 5% happened over 21 days for the vehicle‐treated group, and over only 15 days for the biofilm‐treated groups (Figure 7C). This, as well as significant differences in areas under the curves of relative wound area over time between the three treatment groups, indicate accelerated wound closure following treatment with engineered biofilms compared to treatment with the buffer vehicle. When looking at the detailed kinetics of wound closure, we observed that the TasA‐PRb/FGF2 matrix significantly promotes wound area reduction over the TasA‐based matrix and the vehicle control as early as 3 days after the first application (Figure 7D). Indeed, 3 days after wounding, the TasA‐PRb/FGF2 treated mice showed an average relative wound area of 85%, compared to 97% and 95% for vehicle‐treated and TasA biofilm treated groups respectively. At day 9 post‐wounding, that is, 3 days after the last topical application, both biofilm matrices showed statistically significant promotion of wound area reduction compared to the control, with average relative wound areas of 35% and 23% for TasA and TasA‐PRb/FGF2 biofilms respectively, compared to 57% for the control group. At day 12, wound closure with TasA‐PRb/FGF2 biofilms was almost completed, with most wounds showing a relative area between 0% and 10%. At this same time point, all vehicle‐treated wounds show a relative area above 10%, and wounds treated with TasA biofilms show a dispersion of wound areas mostly between 0% and 20%, with an average of 11%. After 15 days, treatment with TasA‐based biofilms also resulted in almost complete wound closure, with an average relative wound area of 3% and most wounds presenting an area of 0%, while the control group still showed an average relative wound area of 12%. The vehicle‐treated wound only reached an average wound area below 10%, similar to the one reached at day 12 for the groups treated with TasA‐PRb/FGF2 biofilms, after 18 days. Wound areas reached for the control group after 21 days were similar to wound areas for biofilm‐treated groups after 15 days.

The kinetics of reduction of the wound area over time for the control group is similar to what has been reported in the literature for this diabetic splinted wound model [50]. As previously mentioned topical application of biofilm matrices does accelerate wound closure, although the engineered TasA‐PRb/FGF2 biofilm promotes an acceleration of this phenomenon over TasA‐based biofilm. Histological analysis at 21 days post‐wounding also indicates that both biofilms accelerate re‐epithelialization, as a mature epidermis with stratum corneum is present for the biofilm‐treated mice as opposed to the vehicle‐treated mice. (Figure 8) The regenerated skin appears more mature in the biofilm‐treated mice than in the control mice, as indicated by the denser presence of hair follicles and sebaceous glands. The higher number of capillaries and small blood vessels in the biofilm‐treated tissues also point toward angiogenetic properties for both TasA‐PRb/FGF2 and TasA‐based biofilms, which is crucial in healing of chronic wounds, often characterized by decreased vascularization and oxygenation [4, 5]. Furthermore, Masson's trichrome (MT) coloration shows enhanced collagen deposition in the dermis of the biofilm‐treated tissues, in which a thicker and denser collagen layer is observed, as opposed to a sparse and thinner collagen layer for vehicle‐treated tissues (Figure 8). No major difference in tissue architecture is observed between the mice treated with the TasA or TasA‐PRb/FGF2 biofilms at day 21, likely due to the advanced stage of healing. Histological analysis at earlier timepoints would potentially uncover differences in the healing processes favored by each material.

FIGURE 8.

Architecture of healed skin tissues in diabetic mice after 21 days. In Hematoxylin & Eosin coloration (top), the stratum corneum (SC) and epidermis (E) are identified. Black arrows indicate hair follicles, yellow arrows indicate sebaceous glands, and blue arrows represent capillaries or small blood vessels. In Masson's trichrome coloration (below), the stratum corneum and keratinocytes in the epidermis are colored in red, and collagen, in blue. Two representative images from different wounds are presented per group. Scale bar is 50 µm.

Various studies have reported that topical application of diverse non‐engineered probiotics, including Lactobacillus and Lactococcus species, does accelerate wound closure [24, 55]. Inclusion of B. subtilis strains in wound dressings in previous studies has also shown their ability to promote wound healing, likely via antimicrobial compound and protease release in the wound, as well as scavenging of reactive species by lipopeptides and surfactants produced by the bacteria [53, 54]. Although B. subtilis has already been incorporated in exogenous and synthetic matrices for wound healing, to our knowledge, this is the first study reporting usage of B. subtilis in a self‐produced matrix for biomedical applications. This offers the possibility for self‐regeneration of the material. Usage of the engineered TasA‐PRb/FGF2 biofilm showed even more potent promotion of wound healing than the TasA biofilm, likely due to the presence of the cell‐adhesion sequence PRb, as well as the secretion of FGF2. Various studies have indeed demonstrated the benefits of matrices conjugated with cell adhesion sequences in promoting cell adhesion and wound healing [56, 57]. Furthermore, topical application of FGF2 has already proven to be effective for wound healing [8], and its effect would likely be potentiated by slow release from a matrix and/or bacterial secretion, as previously shown for other growth factors [11, 24]. Although this study reports on the cell‐adhesive properties of the PRb‐functionalized biofilms and the cell‐migratory activity of FGF2‐secreting biofilms in vitro, the individual effects on wound healing of each of these bioactive moieties within the biofilm has not been evaluated in vivo.

3. Conclusion

In this study, we showed the development of a self‐produced living biomatrix from the B. subtilis biofilm for wound healing. We engineered a biofilm‐deficient strain with deletions of genes responsible for production of EPS and the TasA protein scaffold for fabrication of functionalized biofilms. Expression of a genetic fusion of the fibronectin‐mimetic cell‐adhesion sequence PRb to the biofilm scaffold protein TasA in this strain allowed the production of a living biofilm with enhanced cell‐adhesive properties toward skin cells, including fibroblasts and keratinocytes. EPS removal from the biofilm also increased cell adhesion, likely by mitigation of surface hydrophobicity. These results support the usage of engineered biofilms in the field of tissue regeneration, as cell‐adhesiveness and biocompatibility are considered key properties of scaffolds in this field. Moreover, modifications of the biofilm‐associated genes did not significantly affect the antimicrobial properties of B. subtilis, which are of interest for biomedical applications. Selection of a signal peptide allowing optimal FGF2 secretion also enabled the design of a biofilm‐deficient B.subtilis strain eluting growth factors. Co‐cultivation of this strain with the one expressing the functionalized TasA allowed incorporation of the FGF2‐producting strain in the biofilm, which resulted in a growth factor‐eluting biofilm. Wound healing experiments in diabetic mice showed the ability of this cell‐adhesive growth factor‐eluting living biofilm to promote wound healing in a chronic wound model without causing observable adverse effects. Biofilms accelerated wound closure in association with improvement of collagen deposition as well as re‐epithelialization and angiogenesis in the chronic wounds. Exact mechanisms underlying these results could be uncovered by further immuno‐histological analyses at multiple stages of the wound closure. Although B. subtilis has previously been studied in a wound healing context, to our knowledge, this is the first study reporting engineering and usage of its endogenously produced matrix, its biofilm, for biomedical applications. This, as well as the living nature of the material would enable further engineering of B. subtilis biofilms into self‐regenerative materials with programmable responses to stimuli within the wound environment. In vivo biosafety evaluation will be needed to ensure the bacteria do not diffuse in the organism or cause adverse localized or systemic inflammation. As the materials possess antimicrobial properties, healing assays on infected wounds would also be required to assess how the antimicrobial properties of the biofilms observed in vitro translate in vivo.

4. Experimental Section

4.1. Plasmids and Strains Construction

All plasmids and bacterial strains used in this study are listed in Table S1. Oligonucleotide primers are listed in Table S2. The undomesticated and transformable strain Bacillus subtilis 3A38, obtained from the Bacillus Genetic Stock Center, was used to create the markerless ΔepsA‐O (Δeps) and ΔepsA‐OΔsinR‐tasA (ΔepsΔst) biofilm‐deficient strains. Allele deletion plasmids pHoss‐ΔepsA‐O and pHoss‐ΔsinR‐tasA were created as described elsewhere [58]. For construction of pHoss‐ΔepsA‐O, a 1‐kb fragment upstream of the epsA‐O gene cluster was amplified from the genomic DNA of Bacillus subtilis subsp. subtilis (Ehrenberg) Cohn (ATCC 23857D‐5) using Q5 High‐Fidelity DNA Polymerase (NEB) and primers EcoRI‐deps‐upF and cdwF‐deps‐upR. The 1‐kb genome fragment downstream of epsA‐O was amplified using primers cupR‐deps‐dwF and KpnI‐deps‐dwR. Both fragments were joined by fusion PCR, resulting in a 2‐kb fragment that was further amplified by PCR using EcoRI‐deps‐upF and KpnI‐deps‐dwR. This 2‐kb fragment was inserted into the pHoss plasmid by digestion with EcoRI and KpnI followed by ligation with T4 DNA ligase. E. coli DH5α competent cells were transformed with the ligation products and plated on ampicillin‐containing LB agar plates. Presence of the 2‐kb fragment was confirmed by colony PCR with Taq DNA polymerase (BioBasic) followed by 1% agarose gel migration and Sanger sequencing using pHoss‐MCS‐F and pHoss‐MCS‐R primers.

For construction of pHoss‐ΔsinR‐tasA, the same procedure was used, using primers EcoRI‐dst‐upF and cdwF‐dst‐upR for upstream fragment amplification and primers cupR‐dst‐dwF and KpnI‐dst‐dwR for downstream fragment amplification.

B. subtilis 3A38 was made naturally competent and first transformed with pHoss‐ΔepsA‐O using an adapted Spizizen salts method [59, 60]. As described elsewhere [58], for markerless deletion, transformed cells underwent multiple selections as they were first cultivated on LB agar plates with 10 µg/mL erythromycin at 30°C, followed by cultivation with erythromycin at 42°C, cultivation in plain LB media at 30°C and incubation with 1.5 µg/mL tetracycline at 30°C. Deletion of the epsA‐O gene cluster was verified by colony PCR using deps‐BS‐F and deps‐BS‐R primers and by observation of the morphology of resulting biofilms in the appropriate culture media. The ΔepsA‐O strain was then made competent and transformed with pHoss‐ΔsinR‐tasA. The deletion and selection were achieved as described for epsA‐O deletion. Colony PCR was done using primers dst‐BS‐F and dst‐BS‐R. The B. subtilis 3A38, the Δeps and the ΔepsΔst strains were all analyzed by whole genome sequencing to further confirm the deletions.

Genes encoding for TasA, TasA‐PRb, and murine FGF2_10‐154 with different signal peptides (AmyQ, CnBP or TasA) were synthesized and cloned into the Bacillus expression vector pHT01 (MoBiTech) between the BamHI and XmaI sites by Genscript. In TasA‐PRb, TasA is connected at its C‐terminus to the PRb sequence using a GGGSGGGS linker. In FGF2 expression plasmids, a (GS)_n (n = 2‐3) spacer was added to prevent cleavage around alanine 11 of FGF2 during signal peptide removal. All expressed protein sequences are listed in Table S3. The expression plasmids were transformed and amplified in E. coli DH5α in presence of ampicillin. The ΔepsΔst strain was transformed using the adapted Spizizen method mentioned above.

4.2. Biofilm Culture

Overnight culture of Bacillus subtilis strains was achieved in LB medium, with (if transformed with pHT01 plasmids) or without 10 µg/mL chloramphenicol at 30°C, 225 rpm, and diluted 1/50 for further culture in the same medium and conditions to obtain a seed culture with OD_600nm = 1–2. For liquid cultures, biofilms were cultivated in 12‐wells flat bottom plates by diluting 5 µL of seed culture in 2 mL MSgg medium (5 mM potassium phosphate pH 7, 100 mM morpholinopropanesulfonic acid pH 7 (MOPS), 2 mM MgCl₂, 700 µM CaCl₂, 50 µM MnCl₂, 50 µM FeCl₃, 1 µM ZnCl₂, 2 µM thiamine, 0.5% glycerol, 0.5% glutamate, 50 µg/ml tryptophan, and 50 µg/ml phenylalanine) with or without chloramphenicol and with or without 1 µM isopropyl β‐d‐1‐thiogalactopyranoside (IPTG) for induction and incubating at 30°C for 4 days without agitation. For co‐cultures in liquid medium, 5 µL of each strain was added. For solid cultures, 5 µL of seed culture was added to MSgg agar plates (125 µM FeCl₃, 1.5% agar) with or without chloramphenicol and IPTG and incubated in the same conditions as liquid cultures.

4.3. Bacillus Cell Counting in Biofilm

Biofilms were collected from liquid cultures in MSgg by centrifugation and resuspended at 50 g/mL in phosphate buffered saline (PBS) and serially diluted 10‐fold to 0.05 ng/mL. LB agar plates were inoculated with 10 µL of each of the four dilutions, incubated at 37°C for 24 h, and colony forming units (CFU) were counted.

4.4. Protein Quantification and Analysis

Biofilms were collected, resuspended at 80 mg/mL in PBS and sonicated for 20 min in a sonicator bath. Proteins were then quantified using Bradford reagent and analyzed by SDS‐12% polyacrylamide gel electrophoresis (SDS‐PAGE).

4.5. Scanning Electron Microscopy

One milligram of biofilm was washed twice with PBS and fixation was achieved by overnight incubation in 2.5% glutaraldehyde at 4°C. Biofilms were then washed twice in PBS and applied on poly‐l‐lysine coated lamella. Lamella were washed with water and dehydrated in a series of ethanol washes before critical point drying with a Leica EM CPD300 instrument. Samples were coated with 4–5 nm platinum using a Leica EM ACE600 and imaged using a FEI Quanta 450 scanning electron microscope equipped with a EDAX octane super 60 mm² camera.

4.6. Rheological Characterization

Biofilms cultivated in liquid medium were collected and analyzed using an Anton Paar rheometer (MCR 302), equipped with a 15 mm cone‐plate accessory. The linear viscoelastic range (LVE) was determined by an amplitude sweep with strain ranging from 0.01% to 100%. Frequency sweeps between 0.1 and 100 rad/s were done at 0.1% strain. Cyclic strain time sweeps consisted of alternating 60 s cycles at 100% (high strain) and 0.5% (recovery) strain.

4.7. Analysis of FGF2 Expression

For screening of FGF2 expression, ΔepsΔst transformed strains were cultivated in LB medium containing 10 µg/mL chloramphenicol with 225 rpm agitation to an OD_600nm of 0.8 and expression was induced for 5 h with IPTG. Strains were also cultivated in MSgg liquid medium as described above. In both cases, after centrifugation, supernatants were analyzed without further treatment, and cells were softly disrupted in PBS for 20 min in a sonicator bath before analysis by 15% polyacrylamide SDS‐PAGE and Western blot. For re‐culture analysis, biofilms from MSgg liquid medium were collected and half of the biofilm was re‐inoculated in MSgg liquid medium without IPTG for 3 days. For Western blot, proteins were transferred to nitrocellulose membranes, which were blocked with 5% w/v non‐fat dry milk in Tris‐buffered saline (pH 7.4) + 0.05 % Tween 20 (TBST). Membranes were then incubated with rabbit anti‐FGF2 antibody (Invitrogen #PA5‐95596) 1:1000 in 1 % w/v milk in TBST, followed by incubation with goat anti‐rabbit IgG antibody conjugated with horseradish peroxidase (HRP) 1:20 000 in 1% milk in TBST. Bands were detected by chemiluminescence.

4.8. Antimicrobial Assays

For cross‐streak antimicrobial assays, 25 µL of an overnight culture of B. subtilis diluted to 0.5 McFarland density in PBS was streaked in a single line on Mueller–Hinton agar (MHA) without chloramphenicol and with or without 1 mM IPTG. After 24 or 48 h incubation at 37°C, 5 µL of an overnight culture of E. coli ATCC11303 or S. epidermidis ATCC12228 diluted to 0.5 McFarland in PBS was streaked in a line perpendicular to B. subtilis. After further incubation at 37°C for 24 h, the distance between the edge of the B. subtilis culture and the concave part of the inhibited E. coli or S. epidermidis culture was measured.

For well‐diffusion antimicrobial assays, MHA was evenly inoculated with 100 µL of an overnight culture of E. coli or S. epidermidis diluted to 0.5 McFarland. Wells with a diameter of 7 nm were created, to which 20 mg of biofilms were added. After incubation at 37°C for 24 h, the inhibition diameter was measured.

4.9. Cell Adhesion Assays

HaCaT human keratinocytes and L929 mice fibroblasts were cultured in DMEM supplemented with 4.5 mg/mL glucose, 2 mM glutamine and 10% FBS. 1BR3 human fibroblasts were cultured in EMEM supplemented with 2 mM glutamine, 1% non‐essential amino acids, and 15% FBS. The control PRb peptide was synthesized by in‐house Fmoc solid phase peptide synthesis and purified by reverse phase liquid chromatography.

High binding 96‐wells plates were coated overnight with 100 µL BSA 3% in PBS, 100 µL of 1 µM PRb peptide or a quantity of biofilm containing 8 ng of proteins as determined by Bradford quantification. Wells were then washed twice with PBS and blocked for 1 h with 3% BSA in PBS. After 2 washes with PBS, 5 000 1BR3 cells or 10 000 HaCaT or L929 cells or 100 µL culture medium per well were added and incubated at 37°C for 1 h with 5% CO₂. Cell culture medium was removed, and wells were washed twice with PBS.

For the resazurin metabolic assay, 150 µL of complete culture medium and 15 µL of 500 µM resazurin in HBSS buffer were added. Cells were further incubated for 3 h, and fluorescence of the wells was measured at 530 nm excitation, 590 nm emission. Each signal produced by the cells was subtracted by the signal of a control well coated with the same biofilm or substance but without plated cells. Adhesion percentage was calculated by dividing this corrected signal by the corrected signal for cells grown in uncoated and unblocked tissue culture‐treated plates.

For fluorescent labeling, after the last wash in PBS, cells were incubated in PBS containing 2 µM calcein‐AM and 4 µM ethidium homodimer‐1 for 20 min. Images were acquired with a 10X objective using an inverted Nikon A1R confocal microscope.

4.10. Transwell Migration Assay

For migration assays, cell‐free biofilm supernatants were obtained by centrifugation of biofilms cultivated in liquid MSgg medium and subsequent filtration with 0.2 µm pore size filters.

Tissue culture inserts with 8 µm pores in 24‐well plates were plated with 2 × 10⁴ L929 cells in 100 µL of complete culture medium. Serum‐free culture medium (DMEM, 4.5 mg/mL glucose, 2 mM glutamine) was added at the bottom of the wells and cells were incubated at 37°C for 24 h with 5% CO₂. The cell‐free supernatants, mouse recombinant FGF2 protein (PeproTech) or PBS was added to the bottom of the well. After further incubation for 24 h, transwells were removed and an area of 71 mm² at the bottom of each well was imaged using an inverted Nikon eclipse Ti microscope equipped with a DS‐Fi3 camera. Cells were counted using the FIJI software (NIH). Results are expressed as fold increase in migrated cells/mm² compared to wells treated with PBS.

4.11. Diabetic Full‐Thickness Splinted Wound Model

The Université du Québec à Montréal animal protection committee approved animal experiments (protocol number 0925‐R1‐050–0926). Ten weeks‐old male diabetic BKS.Cg‐Dock7m/Leprdb/J mice were used for creation of a full‐thickness splinted wound [50]. Mice were anesthetized with 1%–3% isoflurane and hair from the back was removed by shaving followed by application of depilatory cream (Veet). Buprenorphine (Vetergesic, Ceva) 0.1 mg/kg was given as an analgesic right before wound induction and at 12 h post‐wounding. Two circular 6 mm‐diameter full‐thickness wounds were created on the back of the mice using a biopsy punch. 0.5 mm‐thick ring‐shaped silicone splints with an outside diameter of 15 mm and an inside diameter of 8 mm mere glued around the wounds using a cyanoacrylate‐based glue and secured using four simple interrupted 6‐0 nylon sutures. Wounds were photographed and 8 mg of biofilm or 40 µL PBS was applied to each wound, which were then covered with a transparent film dressing (Tegaderm, 3 M). At days 3 and 6 post‐wounding, wounds were photographed, biofilm or PBS was re‐applied and the film dressing was replaced. Following this step, photographs were taken as the film dressing was being replaced every 3 days until day 21. Wound area was measured using FIJI and expressed as percentage of the initial wound area. Wounds for which the splint was removed by the mouse before day 10 were excluded from further analyses. Regenerated skin tissues were collected at day 21 post‐wounding and fixed in 10% formalin for 24 h. Tissues were then dehydrated in ethanol washes and cleared in toluene before embedding in paraffin. Slices with 5 µm thickness were used for H&E, and MT staining.

4.12. Statistical Analysis

For in vitro analyses, data is presented as mean ± SD, n = 3–6 as described in figure captions. For in vivo wound healing experiments, data is presented as mean ± 95% CI, n = 6 mice per group (for weight and adverse effect scoring) and n = 9–12 wounds per group, as wounds for which the splint had been displaced by the mice before day 10 were excluded from further analyses. Wound area was measured using FIJI and expressed as percentage of the initial wound area, and areas under the curves for wound area over time were calculated using GraphPad Prism Version 8. All statistical analyses were conducted with GraphPad Prism using a one‐way ANOVA with Tukey's multiple comparison test. The threshold for statistical significance was established at p < 0.05% with * = p < 0.05, ** = p < 0.01, **** = p < 0.001 and **** = p < 0.0001.

Author Contributions

Mélanie Côté‐Cyr: conceptualization, investigation, writing – original draft, methodology, visualization, writing – review and editing, formal analysis, project administration, and data curation. Marie‐Jeanne Archambault: investigation, methodology. Caroline Dupuis: investigation, methodology. Steve Bourgault: funding acquisition, conceptualization, supervision, project administration, and writing – review and editing.

Funding

This work was supported by funding from a Natural Sciences and Engineering Research Council of Canada (NSERC) grant (RGPIN‐2025‐06834) and from a Tier II Canada Research Chair (CRC‐101207).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adhm71823‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.