Abstract

   Fibrosis-associated fibroblasts have been identified across various
   fibrotic disorders, but not in the context of biomaterials, fibrotic
   encapsulation, and the foreign body response. In other fibrotic
   disorders, a fibroblast subpopulation defined by Thy-1 loss is strongly
   correlated with fibrosis yet we do not know what promotes Thy-1 loss.
   We have previously shown that Thy-1 is an integrin regulator enabling
   normal fibroblast mechanosensing, and here, leveraging nonfibrotic
   microporous annealed particle (MAP) hydrogels versus classical fibrotic
   bulk hydrogels, we demonstrate that Thy1^−/− mice mount a fibrotic
   response to MAP gels that includes inflammatory signaling. We found
   that a distinct and cryptic α–smooth muscle actin–positive Thy-1^−
   fibroblast population emerges in response to interleuklin-1β (IL-1β)
   and tumor necrosis factor–α (TNFα). Furthermore, IL-1β/TNFα-induced
   Thy-1^− fibroblasts consist of two distinct subpopulations that are
   strongly proinflammatory. These findings illustrate the emergence of a
   unique proinflammatory, profibrotic fibroblast subpopulation that is
   central to fibrotic encapsulation of biomaterials.
     __________________________________________________________________

   A regenerative, nonfibrotic biomaterial helped identify inflammatory
   fibroblast subpopulations central to fibrotic encapsulation.

INTRODUCTION

   Engineered materials have long been used as implants to serve as
   sensors, drug delivery devices, stents, tissue/joint replacement
   components, or surgical meshes ([38]1). Implant success can be hindered
   by the foreign body response, a chronic inflammatory response that
   leads to fibrotic encapsulation and inhibits implant function
   ([39]2–[40]5). This fibrotic response is a critical barrier to success
   for many implants that the biomaterials and regenerative medicine
   fields are focused on overcoming. Recently, a microporous annealed
   particle (MAP) hydrogel system was developed, which entails injectable
   microgel particles that are subsequently annealed together in situ,
   creating an interconnected porous structure. In various animal models,
   MAP gels have been shown to regulate innate and acquired immunity,
   evade fibrosis, and accelerate wound closure and regeneration
   ([41]6–[42]9).

   Given the reduced inflammation and fibrosis associated with MAP gel
   implantation, this biomaterial system has the ability to also serve as
   model system to study fundamental mechanisms of biomaterial-associated
   fibrosis. Fibrotic disorders (e.g., cardiovascular fibrosis, idiopathic
   pulmonary fibrosis, liver cirrhosis, and systemic sclerosis) have been
   associated with 45% of deaths in the developed world, indicating that
   the problem of fibrosis is a major health care burden ([43]10).
   Fibrosis is characterized by chronic inflammation followed by excessive
   myofibroblast accumulation, aberrant extracellular matrix (ECM)
   deposition, and aberrant mechanotransduction. The chronic inflammatory
   component of fibrosis has been well characterized ([44]10–[45]15), and
   even more so the myofibroblast and ECM components ([46]16–[47]18).
   However, there is a dearth of studies investigating how chronic
   inflammation leads to pathogenic myofibroblast and ECM accumulation in
   fibrosis. Bridging the gap between inflammation and dysfunctional
   mechanotransduction is a critical piece to better understand key
   mechanisms behind fibrosis. In addition, understanding how
   immune-stromal cross-talk can inspire or mitigate fibrosis could help
   inform better biomaterial design and identify previously unknown
   avenues for therapeutic approaches for fibrosis ([48]19).

   In human fibrotic disorders, Thy-1–negative fibroblasts emerge as a key
   player in progressive tissue remodeling and fibrosis. Thy-1 is a
   glycosylphosphatidylinositol-anchored membrane protein that is
   expressed on fibroblasts, neurons, and thymocytes, and we have
   previously established that it is essential to normal fibroblast
   mechanosensitivity. Thy-1 directly binds αvβ3 integrin in cis,
   regulates its tonic/baseline activity, and primes the integrin for
   efficient signal transduction. This critical regulatory mechanism for
   Thy-1 prevents aberrant fibroblast contractility, strain stiffening of
   surrounding matrix, and fibrosis ([49]16, [50]20). Thy-1–negative
   fibroblasts, even while on soft substrates, display elevated
   contractility and ECM deposition and are prone to myofibroblastic
   differentiation in fibrosis in the lung and heart ([51]Fig. 1A)
   ([52]20–[53]23). Fibrotic disorders and, in some cases, the very
   fibrotic lesions themselves are characterized by Thy-1–negative
   fibroblasts ([54]21, [55]24). However, what is unknown is how this
   subpopulation loses Thy-1 and whether Thy-1–negative fibroblasts
   contribute to the fibrotic encapsulation that occurs with implanted
   biomaterials.

Fig. 1. Hydrogel fabrication and subcutaneous implantation.

   [56]Fig. 1.
   [57]Open in a new tab

   (A) Schematic describing regulatory mechanism of Thy-1 in fibroblast
   mechanotransduction. TGFBR, TGF-β receptor. (B) Illustration detailing
   subcutaneous injection regimen on the dorsal side of mice and a summary
   of tissue and cellular-level responses to MAP and NP PEG hydrogels. (C)
   H&E, Masson’s trichrome, and picrosirius red micrographs of MAP and NP
   hydrogels subcutaneously implanted into WT and Thy-1^−/− mice for 3
   weeks; brackets indicate the fibrous capsule. (D) Quantification of
   fibrotic capsule thickness with multiple measurements across each
   micrograph, (E) Thy-1–positive fibroblasts, and (F) percent area of
   picrosirius red staining under polarized light. KO, knockout; ns, not
   significant. ****P < 0.0001, **P < 0.01, *P < 0.05, determined by
   two-way ANOVA with Tukey’s multiple comparison test. n = 5 mice per
   group; two sections per animal were analyzed. Scale bars, 100 μm.

   With the advent of single-cell transcriptomics, fibroblast
   heterogeneity is becoming better characterized with each new study
   ([58]25–[59]31). In addition to the heterogeneity of their cell state,
   there is also diversity in the origin of fibroblasts and
   myofibroblasts. Some critical fibroblast populations in dermal repair
   have been shown to mobilize from the underlying fascia ([60]32), while
   other studies have demonstrated that adipocytes and adipogenic
   progenitors are a source for myofibroblasts and are required for skin
   repair ([61]29, [62]33). As for the variety of cell subpopulations that
   can emerge from these different origins, the presence of
   immunologically active fibroblasts in both health and disease is
   becoming more clear, especially in the cancer field ([63]34–[64]38).
   However, we do not know whether immunologically active/inflammatory
   fibroblasts contribute to fibrosis. In this study, by using
   regenerative and nonfibrotic MAP gels compared to traditional fibrotic
   bulk nanoporous (NP) hydrogels and implanting them in wild-type (WT)
   versus Thy1^−/− mice, we demonstrate that, in the absence of Thy-1, MAP
   gels become as fibrotic as the NP gel. The fibrotic capsule around the
   NP gel in WT mice displayed increased interleukin-1 receptor I (IL-1RI)
   expression, α–smooth muscle actin (α-SMA) expression, and nuclear
   factor κB (NF-κB) p65 phosphorylation when compared to MAP gels.
   However, among Thy1^−/− mice, both MAP and NP gels displayed elevated
   IL-1RI expression, α-SMA expression, and NF-κB p65 phosphorylation,
   indicating that the absence of Thy-1 predisposed biomaterial implants
   to both inflammation and fibrosis and implicates Thy-1–negative
   fibroblasts at the center. We demonstrated that, while fibroblasts were
   cultured on a soft hydrogel substrate conjugated with fibronectin to
   recapitulate their native biophysical and biochemical microenvironment,
   interleuklin-1β (IL-1β) and tumor necrosis factor–α (TNFα) were able to
   promote a Thy-1–negative subpopulation that matched the previously
   reported phenotype of Thy-1–negative fibroblasts and that some cells
   were predisposed to Thy-1 loss while other cells were resistant to
   Thy-1 loss. This noted heterogeneity led us to use single-cell RNA
   sequencing (scRNA-Seq) to identify the subpopulations present with and
   without IL-1β and TNFα treatment. This demonstrated that two
   subpopulations emerged, which were defined by inflammatory cytokines or
   receptors among their differentially expressed genes, inflammatory
   pathways up-regulated via Kyoto Encyclopedia of Genes and Genomes
   (KEGG) pathways analysis, and low transcriptional expression of Thy-1.
   Together, these findings demonstrate that Thy-1–negative fibroblasts
   help bridge the gap between inflammation and dysfunctional
   mechanotransduction via emerging in response to inflammatory cytokine
   signaling and demonstrating a proinflammatory, profibrotic phenotype
   that is central to material-associated fibrosis.

RESULTS

Absence of Thy-1 promotes fibrosis in traditionally regenerative,
immunoregulatory biomaterials

   To determine if expression of the αvβ3 integrin regulator Thy-1
   ([65]Fig. 1A) contributes to biomaterial-mediated fibrosis, we
   specifically asked if Thy-1 loss is sufficient to reverse the
   nonfibrotic nature of the host response to MAP hydrogels ([66]6). While
   Thy-1 expression is present among both fibroblasts and lymphocytes, due
   to the absence of a reliable pan fibroblast marker and a previous work
   demonstrating that the loss of Thy-1 among lymphocytes has no effect on
   fibrosis ([67]39), we used Thy1^−/− mice to interrogate Thy-1 loss in
   biomaterial-mediated fibrosis. We leveraged Thy1^−/− mice and
   controlled the experiment with a standard fibrogenic NP bulk hydrogel
   with the same polyethylene glycol (PEG)–based chemistry as the
   implanted MAP gels ([68]Fig. 1B). As expected, when implanted
   subcutaneously into WT mice, the NP hydrogel promoted a fibrous capsule
   that was ~220 μm in thickness after 3 weeks, whereas the MAP hydrogel
   displayed an almost undetectable fibrous capsule that was 20 μm in
   thickness and retained its porous structure, as previously described
   (6). In stark contrast, when implanted into Thy1^−/− mice, both
   hydrogels mounted a robust fibrous encapsulation with no discernable or
   statistical difference between NP and MAP hydrogels ([69]Fig. 1, C and
   D, and fig. S1). Furthermore, we harvested and digested NP and MAP
   implants to determine the presence of Thy-1–positive fibroblasts at the
   implant site and there was significantly higher Thy-1–positive
   fibroblasts collected from MAP gels compared to NP gels ([70]Fig. 1E
   and fig. S2). Using picrosirius red staining imaged under polarized
   light, we were able to determine collagen deposition among each group.
   This captured the collagenous fibrotic capsule present around the NP
   hydrogel and both gels implanted into Thy1^−/− mice. The most amount of
   collagen was detected with the MAP gel, which can be attributed to
   tissue ingrowth into the MAP gel based on the depth of the collagen
   network within the hydrogel ([71]Fig. 1F). These findings are in
   keeping with prior evidence that loss of Thy-1 expression leads to a
   highly activated, ECM-synthetic fibroblast, even in normally
   quiescence-inducing soft environments. Thy-1–negative fibroblasts
   display high baseline levels of active αvβ3, are far more adhesive, and
   display rapid turnover of integrin adhesion clusters, which promotes
   elevated matrix assembly and contraction, both hallmarks of tissue
   fibrosis ([72]16, [73]20, [74]23).

IL-1RI expression and NF-κB signaling correspond to α-SMA expression and
fibrotic encapsulation

   Given the role of macrophage-derived foreign body giant cells in
   biomaterial responses, it is critical to understand how local
   fibroblasts contribute to the foreign body response. More specifically,
   Thy-1 loss is not predicted to affect inflammatory responses as Thy-1
   loss in models of fibrosis is historically tied to mechanotransduction
   and integrin regulation ([75]16, [76]20, [77]21, [78]40). Although
   there is strong evidence to suggest that Thy1^−/− fibroblasts are prone
   to myofibroblastic differentiation and generation of fibrosis, an
   elevation in inflammation due to a fibroblast defect would be
   unexpected. This highlights a gap in our understanding of the nuanced
   relationship between inflammation and fibrosis. Certainly, for over a
   decade, it has been appreciated that skewing macrophage responses can
   lead to better tissue remodeling outcomes, and recently, it has been
   shown that recruitment of interleukin-33 type 2 myeloid cells and
   activation of adaptive immunity can drive biomaterial-associated
   regeneration ([79]9). However, these results further corroborate that
   immune signaling may not be unidirectional, from immune cells to
   stromal cells, but rather bidirectional ([80]41–[81]45). We further
   corroborated this within the biomaterial context by using a publicly
   available scRNA-Seq dataset that identifies immune and stromal cell
   subpopulations ([82]GSE175890) ([83]46) and used the Connectome package
   to infer cell-cell signaling based on ligand and receptor expression
   ([84]47). This analysis revealed that, within the biomaterial
   microenvironment, macrophages and fibroblasts display bidirectional
   signaling across multiple pathways (fig. S3). Among the pathways that
   came out of this analysis, the IL-1β–IL-1R signaling axis emerged
   between multiple subpopulations of macrophages and fibroblasts.

   We began unraveling the immune-stromal signaling axis by exploring
   IL-1R signaling. Chronic IL-1R signaling is elevated in fibrosis across
   various tissues ([85]48–[86]50). IL-1RI expression was clearly evident
   within the fibrotic capsule of the NP hydrogel and grossly absent in
   and around the MAP hydrogel within WT mice. In juxtaposition, IL-1RI
   expression was present and prominent in the fibrotic capsule of both NP
   and MAP hydrogels implanted in Thy1^−/− mice ([87]Fig. 2A). IL-1RI
   expression within fibrotic capsules corresponded with downstream NF-κB
   signaling, as shown by phospho-p65, and α-SMA expression, a marker of
   myofibroblast differentiation. Specifically, in NP hydrogels, there was
   notable coexpression of α-SMA and phosphorylated p65 within the fibrous
   capsule among WT mice, indicating a potential role for inflammatory
   signaling in the tissue remodeling phase in addition to the
   inflammatory phase in response to biomaterials. MAP hydrogels implanted
   in WT mice demonstrated very little detectable α-SMA expression or p65
   phosphorylation at the biomaterial surface or among cells migrating
   into the implant. These differences are also reflected in regions with
   less fibrotic encapsulation of the NP hydrogel; there is less α-SMA
   expression and p65 phosphorylation (fig. S4, A and B). Consistent with
   our analysis of fibrotic outcomes, NP and MAP hydrogels both display
   robust α-SMA expression and p65 phosphorylation within their fibrous
   capsule when implanted in Thy1^−/− mice ([88]Fig. 2, B to D). The data
   clearly demonstrate that genetic knockout of Thy-1 is sufficient to
   overcome the regenerative and nonfibrotic nature of MAP hydrogels and
   illustrate a potential fibroblast cytokine signaling mechanism.

Fig. 2. Inflammatory signaling among hydrogels in Thy-1^−/− mice.

   [89]Fig. 2.
   [90]Open in a new tab

   (A) Fluorescence images of MAP and NP hydrogel subcutaneous implants
   stained for IL-1RI (green) and DAPI (blue). (B) Fluorescence images
   staining for α-SMA (green), Thy-1 (red), phospho-p65 NF-κB (cyan), and
   DAPI (blue) with quantification for the sum of (C) α-SMA pixel values
   and (D) phospho-p65 NF-κB^+ cells per field of view (FOV). ***P <
   0.001, **P < 0.01, *P < 0.05, determined with two-way ANOVA with
   Tukey’s multiple comparison test. n = 5 mice per group; two sections
   per animal were analyzed. Scale bars, 100 μm.

IL-1β and TNFα promote fibroblast Thy-1 loss

   The evident differences in immune-stromal signaling between NP and MAP
   hydrogels and the ability of Thy1^−/− mice to effectively normalize the
   fibrotic responses between these two systems led us to hypothesize that
   inflammatory cytokines promote the loss of Thy-1 and a coincident
   activated myofibroblastic phenotype among fibroblasts. Using an
   unbiased approach with a publicly available RNA sequencing (RNA-Seq)
   dataset ([91]GSM2072332) ([92]51), we identified the cytokine receptors
   expressed on fibroblasts and determined the T helper 1 (T[H]1)
   [interleukin-1α (IL-1α), IL-1β, TNFα, and interferon-γ (IFN-γ)], T
   helper 2 (T[H]2) [interleukin-4, interleukin-6 (IL-6), and
   interleukin-13], and T helper 17 [interleukin-17 (IL-17)] cytokines
   that are capable of signaling within fibroblasts. Fibroblasts (CCL-210)
   treated with T[H]1 cytokines (IL-1α, IL-1β, IFN-γ, and TNFα) all
   display elevation in expression of the myofibroblastic marker α-SMA
   (fig. S5). We additionally observed significantly elevated fibroblast
   spreading on soft (2 kPa) fibronectin-coated substrates when treated
   with IL-1β and TNFα and elevated but not significantly increased spread
   area when treated with IL-17, IFN-γ, and transforming growth factor–β
   (TGF-β). Only IL-1β and TNFα, and to a lesser extent IFN-γ, induced a
   loss of Thy-1 surface expression, and only IL-1β and TNFα show
   consistently promoted α-SMA expression. These short-term, acute
   cytokine stimulations indicate immune-stromal signaling that results in
   alteration of fibroblast phenotype but fall short of establishing the
   emergence of a persistent Thy-1–negative fibroblast subpopulation.

Cytokine-mediated loss of Thy-1 recapitulates the traditional Thy-1–negative
phenotype

   We therefore treated fibroblasts, seeded on soft (~2 kPa)
   fibronectin-coated hydrogels longitudinally with IL-1β and TNFα
   individually or in combination for 3 to 7 days. As in short
   stimulations, IL-1β or TNFα treatment for 72 hours resulted in a
   significant increase in the Thy-1–negative population (~20% of the
   total population) compared to fibroblasts in media. The prominence of
   the Thy-1–negative population was enhanced when stimulated with a
   cocktail of IL-1β and TNFα ([93]Fig. 3A). Stimulation over a 7-day
   period further intensified the proportion of Thy-1–negative
   fibroblasts, reaching ~30% in IL-1β and TNFα treatments and the
   combination yielding fibroblasts that were 40% Thy-1–negative (fig.
   S6). Concomitant with Thy-1 loss, fibroblasts displayed significantly
   greater cell spreading on soft hydrogels stimulated with IL-1β and TNFα
   for 72 hours despite showing no discernable differences in focal
   adhesion size ([94]Fig. 3, B and C). Fibroblast spreading on soft
   hydrogels suggests altered mechanosensory signaling, and our
   observations here are consistent with our prior discovery of Thy-1’s
   role in enabling proper mechanotransduction and stiffness sensing in
   fibroblasts ([95]20). Only IL-1β stimulation leads to a significant
   increase in α-SMA expression ([96]Fig. 3D), suggesting a transition to
   a myofibroblast phenotype. This suggests cooperative, but not
   overlapping, roles between IL-1β and TNFα signaling in fibroblasts.

Fig. 3. Cytokine-mediated Thy-1 loss.

   [97]Fig. 3.
   [98]Open in a new tab

   (A) Flow cytometric analysis of cytokine-treated CCL-210 human lung
   fibroblasts measuring the emergence of a Thy-1–negative fibroblast
   subpopulation. Representative histogram of six independent experiments.
   (B) Fluorescence confocal images of CCL-210 human lung fibroblasts on
   fibronectin-coated 2-kPa polyacrylamide hydrogels for 48 hours.
   Staining is for vinculin (green), F-actin (red), and DAPI (blue). (C)
   F-actin images were used to calculate cell area, and vinculin images
   were used to measure focal adhesion area. (D) Flow cytometric analysis
   of cytokine-treated CCL-210 human lung fibroblasts measuring α-SMA
   expression depicted as geometric mean fluorescence intensity (gMFI).
   (E) Measuring Thy-1 expression using flow cytometry in CCL-210 human
   lung fibroblasts that have been sorted based on Thy-1 expression to
   identify if Thy-1 loss is deterministic. FITC, fluorescein
   isothiocyanate. ****P < 0.0001, ***P < 0.001, *P < 0.05, determined
   with one-way ANOVA with Tukey’s multiple comparison test. n = 5
   biological replicates. Scale bars, 100 μm.

Thy-1^+ and Thy-1^− fibroblast subpopulations were stable and deterministic

   The existence of a true subpopulation implies both a deterministic
   emergence and population persistence. Given the grossly Thy-1^+ nature
   of naïve fibroblasts (only 5 to 10% are inherently Thy-1^−), this would
   imply that there is a hidden or cryptic fibroblast subpopulation
   predetermined to lose Thy-1 in response to IL-1β and/or TNFα that then
   persists in the absence of the stimuli and/or is resistant to
   additional stimulation. To establish the presence of a deterministic
   subpopulation, we treated fibroblasts for 72 hours with either IL-1β or
   TNFα. Fibroblasts were then sorted based on Thy-1 expression into
   negative and positive populations. Each population was then
   subsequently stimulated with either media alone or a second dose of the
   initial cytokine for another 48 hours. Thy-1^+ fibroblasts emerging
   from the initial cytokine stimulation were both stable, i.e., remained
   Thy-1^+ in subsequent culture in media alone, and deterministic, i.e.,
   resistant to Thy-1 loss upon a second stimulation. Similarly, Thy-1^−
   fibroblasts resulting from IL-1β or TNFα stimulation were also stable
   and deterministic ([99]Fig. 3E). These data indicate that there are
   subpopulations within the native fibroblast population that are
   resistant to cytokine-induced loss of Thy-1 and others predisposed to
   lose Thy-1 and remain Thy-1–negative. The emergent Thy-1^− fibroblasts
   represent a cryptic subpopulation that emerge specifically in response
   to IL-1β and/or TNFα.

scRNA-Seq reveals Thy-1^− fibroblasts possess a distinct inflammatory
signature

   While Thy-1^+/− subpopulations help define the heterogeneity of
   fibroblasts, they certainly do not fully encompass the true
   heterogeneity within the population. Recently, scRNA-Seq analysis of
   fibroblasts from various tissues and diseases has elucidated as many as
   four to seven transcriptionally distinct fibroblast subpopulations
   ([100]30, [101]31, [102]52–[103]54). We similarly took this approach to
   uncover both the endogenous heterogeneity of our fibroblast population
   and how those subpopulations change and new subpopulations emerge (such
   as Thy-1^−) in response to inflammatory signaling. Our scRNA-Seq
   analysis (figs. S7 to S10) revealed five transcriptionally distinct
   clusters of cells among untreated fibroblasts cultured on soft (2 kPa)
   hydrogels, a degree of heterogeneity that matches what has been
   observed among fibroblasts in mouse and human tissues. Following
   72-hour stimulation with IL-1β and TNF, two new clusters emerge
   (clusters 2 and 5) along with the expansion of an existing cluster
   (cluster 0; [104]Fig. 4, A and B), all of which possess a distinct
   transcriptional signature based on differential expression analysis
   ([105]Fig. 4C). At the transcriptional level, subpopulations 0, 2, and
   5 also displayed reduced Thy-1 expression ([106]Fig. 4B). According to
   Slingshot pseudotime analysis that predicts cell trajectory ([107]55),
   cluster 0 represents an endogenous, native Thy-1 low population,
   whereas clusters 2 and 5 are the emergent, cryptic Thy-1 low
   subpopulations observed in prior experiments ([108]Fig. 4D). The top
   markers of clusters 0, 2, and 5 are inflammation related: IL-1RI, IL-6,
   IL-1β, interleukin-8 (IL-8), interleukin-33, and interleukin-11 (IL-11)
   (fig. S11 and [109]Fig. 4C). These transcriptional signatures were
   further confirmed by KEGG pathway enrichment analysis that identified
   overexpression of transcripts related to phosphatidylinositol 3-kinase
   (PI3K)–Akt, TNF, IL-17, and NF-κB signaling pathways ([110]Fig. 4E).
   Intriguingly, the top gene identifying cluster 5 is IL-1β, the
   inflammatory cytokine that induces Thy-1 loss and the emergence of the
   profibrotic fibroblast subpopulations. The top markers and KEGG pathway
   analysis for the Thy-1 low subpopulations indicates a clear
   inflammatory transcriptional signature. This inflammatory signature was
   further corroborated by analyzing a publicly available dataset
   ([111]GSE211834) from an in vivo bleomycin-induced skin fibrosis model
   where WT and Thy1^−/− dermal fibroblasts underwent RNA-Seq analysis.
   Using KEGG pathway analysis, we identified that Thy1^−/− dermal
   fibroblasts in this model displayed an enrichment for more inflammatory
   pathways (fig. S12), many of which were the same pathways that emerged
   in our scRNA-Seq dataset. Together, these data provide a compelling
   argument that initial signaling by T[H]1 cytokines associated with the
   acute inflammatory response to biomaterials induces a profibrotic
   immunofibroblast capable of driving a positive feedback autocrine
   signaling loop capable of sustaining fibrotic progression.

Fig. 4. Cytokine-mediated changes in fibroblast heterogeneity.

   [112]Fig. 4.
   [113]Open in a new tab

   CCL-210 fibroblasts were treated with inflammatory cytokines on 2-kPa
   fibronectin-coated hydrogels for 72 hours. Cells were collected for
   scRNA-Seq. (A) Uniform Manifold Approximation and Projection (UMAP)
   clustering projection showing subpopulations of CCL-210 fibroblasts in
   media alone or treated with IL-1β and TNFα. (B) Violin plot
   demonstrating Thy-1 expression across subpopulations and fraction of
   cells coming from each subpopulation based on treatment group. (C)
   Violin plots of top 2 differentially expressed genes that serve as
   markers identified for each subpopulation. GJA1, gap junction protein
   alpha 1; TAGLN, transgelin; VIM, vimentin; COL8A1, collagen 8 alpha 1;
   COL1A1, collagen 1 alpha 1; COL1A2, collagen 1 alpha 2; CENPF,
   centromere protein F; TOP2A, topoisomerase II alpha; MGP, matrix Gla
   Protein. (D) Slingshot pseudotime analysis demonstrating cell
   trajectory throughout subpopulations. (E) KEGG pathway enrichment
   analysis identifying up-regulated pathways corresponding to
   differentially expressed genes for each cluster. ER, endoplasmic
   reticulum.

DISCUSSION

   Biomaterial-mediated fibrosis and the foreign body response are the
   critical obstacles to implant success. There has been extensive work
   demonstrating the role chronic inflammatory signaling has in promoting
   implant failure and efforts to modulate the immune system to attenuate
   fibrosis. Hu et al. ([114]56) demonstrated that Toll-like receptor
   2–modulating biomaterials demonstrated the ability to promote
   islet-xenograft survival. Chung et al. ([115]57) showed that reducing
   IL-17 signaling or treatment with a senolytic agent reduces the
   fibrotic response to implanted biomaterials. Another study demonstrated
   that modulating macrophage activation state is critical to promoting
   implant vascularization, a key step in promoting tissue ingrowth as
   opposed to fibrotic encapsulation ([116]58). Sadtler et al. ([117]59)
   showed that proregenerative biomaterials require T[H]2 cells to guide
   macrophage activation to guide tissue repair. While inflammatory
   signaling and various immune cell populations are critical to promoting
   or avoiding fibrotic encapsulation, it is the fibroblasts and not
   immune cells that are the primary contributors to scar formation.
   Therefore, it is critical to understand how immune-fibroblast
   cross-talk contributes to biomaterial-mediated fibrosis. Here, we
   demonstrate that implants that undergo fibrotic encapsulation display
   robust NF-κB signaling as well as IL-1RI^+ fibroblasts within the
   fibrous capsule. We demonstrate that IL-1β treatment promoted α-SMA
   expression, increased cell area indicating increased cell
   contractility, and loss of Thy-1 expression. Thy-1 loss has been shown
   to be critical to various forms of tissue fibrosis and promotes
   dysfunctional mechanotransduction among fibroblast subpopulations that
   undergo myofibroblastic differentiation independent of the mechanical
   cues from the surrounding microenvironment ([118]16, [119]20, [120]21,
   [121]60). When nonfibrotic NP gels and regenerative MAP gels were
   implanted in Thy1^−/− mice, these MAP gels induced a fibrotic response
   that resembled the encapsulation seen in response to bulk NP gels. Our
   data show that the ability to evade fibrotic remodeling by MAP gels is
   not necessarily a salient feature of the material per se and can be
   reversed by simply tuning the fibroblast cell milieu within the
   material microenvironment. scRNA-Seq determined that IL-1β and TNFα
   treatment altered the fibroblast subpopulations and promoted the
   emergence of a Thy-1–negative “immunofibroblast” defined by cytokine,
   chemokine, and cytokine receptor expression. These findings provide a
   critical step forward in our understanding of responses to implanted
   biomaterials by identifying a critical subpopulation of fibroblasts
   that can promote fibrosis in traditionally nonfibrotic materials.
   Furthermore, we identify a signaling axis that promotes the emergence
   of this potent fibrotic and immune-responsive fibroblast subpopulation.
   These findings might inform biomaterial design (i.e., targeting IL-1R
   signaling or Thy-1–negative fibroblasts) and present an opportunity for
   a previously unidentified in vitro approach for biomaterial
   diagnostics.

   One key limitation of this study is the use of global Thy-1 knockout
   mice as opposed to fibroblast-specific Thy-1 knockout mice. As the
   fibroblast field narrows in on a reliable pan-fibroblast marker, our
   future studies investigating immune-stromal cross-talk in
   biomaterial-mediated fibrosis will certainly use them. As is, our
   findings paint an interesting picture because they imply that if Thy-1
   expression is restored, then perhaps traditionally fibrotic implants
   can become nonfibrotic. This hypothesis was validated in a recent study
   where soluble Thy-1 was delivered intratracheally and shown to reverse
   fibrosis ([122]61). Designing or modifying biomaterials to promote
   Thy-1 expression in an effort to avoid fibrotic encapsulation would be
   a critical next step in applying these findings.

   Macrophage/monocyte-to-myofibroblast transitions ([123]62), so-called
   “fibrocytes,” have been well described in fibrotic responses. Here, in
   an unexpected twist, elements of the transcriptional signature of these
   unique fibroblasts include traditional macrophage markers, possibly
   implicating a form of myofibroblast-to-macrophage transition such that
   fibroblasts are capable of sustaining a macrophage-like response well
   beyond the termination of the acute inflammatory phase. Immune-stromal
   communication has been previously reported in disease ([124]63) and
   endogenous repair ([125]29) where inflammation instructs fibrosis, yet
   our data suggest that fibroblasts can “talk back,” driving chronic
   inflammation. This inflammatory phenotype of fibroblasts has largely
   been observed in the context of cancer but has yet to be discussed
   within the biomaterials field ([126]34–[127]38). Tackling these
   unexpected and cryptic contributors to inflammation and fibrosis will
   be the key to future success in biomaterial development and the
   longevity of implant function.

MATERIALS AND METHODS

Hydrogel fabrication

   Four-arm PEG-maleimide (PEG-Mal; 10 and 20 kDa) was purchased from
   Nippon Oil Foundry Inc. (Japan) and used as the backbone. RGD peptide
   (Ac-RGDSPGGC-NH[2]) and the matrix metalloproteinase 2 (MMP-2)
   degradable cross-linker (Ac-GCGPQGIAGQDGG-NH[2]) were purchased from
   WatsonBio Sciences (United States). All materials were dissolved in
   either ultrapure water or 0.1% trifluoroacetic acid solutions (to
   prevent disulfide bond formation) and aliquoted at specific volumes to
   ensure precision. Aliquots were lyophilized and stored at −20°C until
   use. A heterofunctional maleimide/methacrylamide four-arm PEG macromer
   (MethMal) was synthesized as previously reported ([128]64). A 6.5 wt %
   gel was prepared by dissolving PEG-Mal, MethMal, and RGD in 10X
   phosphate-buffered saline (PBS) (pH = 2). This solution was combined
   with an MMP cross-linker and 5 μM biotin-maleimide, which were
   dissolved in ultrapure water. The final concentrations in the hydrogel
   solution were 81.31 mg/ml PEG-Mal, 14.90 mg/ml MMP, 0.80 mg/ml RGD, and
   8.02 mg/ml MethMal. Microgels were produced using a microfluidic
   polydimethylsiloxane mold created using a previously published design
   ([129]65) in a dust-free hood. Briefly, a 1% Pico-Surf surfactant
   (Sphere Fluidics) solution diluted in NOVEC 7500 oil (3 M) was run
   through the oil channel and the gel formulations described above were
   in the aqueous channel. Using a syringe pump, the surfactant and gel
   solutions were run at 5 ml/hour in the device and collected in a
   conical tube. The resultant particles were mixed with a triethylamine
   (TEA) solution (20 μl of TEA/ml of gel) to increase the pH and
   accelerate gelation. Microgels were washed three times with NOVEC 7500
   oil (1X volume of gel). Next, microgels were combined with PBS (5X
   volume of gel) and washed three times with NOVEC 7500 oil (1X volume of
   gel), allowing separation of the oil and aqueous solution by settling.
   Last, the oil was removed and the microgels were washed three times
   with PBS (3X volume of gel) and hexanes (3X volume of gel). After the
   final wash, microgels reacted overnight at 37°C with a 100 mM
   N-acetyl-l-cysteine solution in 1X PBS to cap any excess maleimides.
   Microgels were removed from solution via centrifugation (4696g x 5min),
   and all further steps were performed in a biosafety cabinet. Microgels
   were washed three times with 70% isopropyl alcohol followed by four
   washes with sterile 1X PBS (pH = 7.4) before being mixed with a
   photoinitiator solution.

NP gel fabrication

   In a biosafety cabinet, an identical 6.5 wt % gel with a PEG-Mal
   backbone was prepared by dissolving PEG-Mal, MethMal, and RGD in
   sterile-filtered 2X PBS (pH = 6.74). The MMP cross-linker and 5 μM
   biotin-maleimide were dissolved in ultrapure water. The solutions were
   combined and loaded into a syringe for injection. The final
   concentrations in the NP gel solution were 81.31 mg/ml PEG-Mal, 14.90
   mg/ml MMP, 0.80 mg/ml RGD, and 8.02 mg/ml MethMal.

Mice and subcutaneous implantation of hydrogels

   All animal care and experiments followed guidelines by the University
   of Virginia’s Institutional Animal Care and Use Committee. C57BL/6 (WT)
   and Thy1^−/− mice were purchased from the Jackson Laboratory.
   Ten-week-old male and female mice were used for all studies. For
   subcutaneous implantation of hydrogels, 10-week-old C57BL/6 mice were
   anesthetized with aerosolized isoflurane and received three 80-μl
   subcutaneous dorsal-side injections of MAP hydrogel solution and three
   80-μl injections of NP hydrogel solution as a contralateral control.
   After 3 weeks, mice were euthanized via carbon dioxide asphyxiation and
   the hydrogels and surrounding subcutaneous tissue were collected and
   frozen in optimal cutting temperature compound over dry ice. Sections
   (10 μm) were cut using a CryoStar NX50 cryostat and mounted on
   SuperFrost Plus slides for downstream histology and immunofluorescence.

Histology and analysis

   Slide-mounted 10-μm sections of hydrogel samples were fixed with 4%
   paraformaldehyde and then stained with hematoxylin and eosin (H&E),
   Masson’s trichrome, and picrosirius red using standard procedures. The
   subsequent histology images were blinded and analyzed for fibrous
   capsule thickness, which we defined as the collagenous tissue
   surrounding the circumference of the hydrogel implant below the skin,
   and foreign body giant cell presence, which we defined as
   multinucleated cells within the fibrotic capsule with a large bundle or
   ring of nuclei ([130]66). Picrosirius red staining was imaged under
   polarized light using a Zeiss Axioscan 7 slide scanner, and collagen
   deposition was quantified as percent area of positive picrosirius red
   staining. Masson’s trichrome and picrosirius red staining was done with
   the help of the Research Histology core at University of Virginia
   (UVA).

Tissue immunofluorescence and quantification

   Slide-mounted 10-μm sections were fixed with 4% paraformaldehyde and
   blocked with 3% normal goat serum (NGS) + 0.1% Triton in 1X PBS for
   intracellular antigens. Primary antibody dilutions were prepared in 3%
   NGS in 1X PBS as follows: anti-mouse α-SMA (1A4, Thermo Fisher
   Scientific) at 1:200, rat anti-mouse Thy-1.2 (53-2.1, BD Biosciences)
   at 1:100, rat anti-mouse phospho–NF-κB p65 (Ser^536; 93H1, Cell
   Signaling Technologies) at 1:100, and biotinylated anti-mouse IL-1RI
   (JAMA-147, BioLegend) at 1:100. The primary antibody incubation was
   done overnight at 4°C, then washed three times with 1X PBS, and then
   stained with a secondary antibody for 1 hour at room temperature. The
   secondary antibodies (goat anti-mouse Alexa Fluor 488, goat anti-rat
   Alexa Fluor 555, and goat anti-rabbit Alexa Fluor 647; Invitrogen) were
   all prepared in 3% NGS in 1X PBS at a dilution of 1:1000.
   4′,6-Diamidino-2-phenylindole (DAPI) was used as a counterstain at 300
   nM in 1x PBS for 5 min at room temperature and then mounted in mounting
   medium ProLong Diamond Antifade (Thermo Fisher Scientific). Slides were
   imaged on a Keyence BZ-X810 microscope using the 10X objective. To
   account for autofluorescence, each sample was treated with the Vector
   TruView autofluorescence quenching kit to diminish autofluorescence
   (Vector Laboratories). α-SMA expression was quantified by defining a
   region of interest in Fiji around the fibrotic capsule within the field
   of view and measuring the sum of the α-SMA signal pixel intensity as
   defined by raw integrated density in Fiji. Phospho-p65^+ cells were
   measured by recording the number of cells that expressed phosphop-p65
   in thresholded images per field of view using the “Analyze Particles”
   function in Fiji ([131]67). Exposure settings and thresholding approach
   were the same across all images and incorporated the use of primary
   delete images (fig. S13). Antibodies were validated via
   immunofluorescence and flow cytometry on lung tissue, which is known to
   be high in all the observed markers.

Polyacrylamide hydrogels for cell culture

   Human lung fibroblasts (CCL-210) were purchased from the American Type
   Culture Collection were cultured [Dulbecco’s modified Eagle’s medium,
   10% fetal bovine serum (FBS), and 1% penicillin/streptomycin] and used
   between passages 2 and 10. To recapitulate the biophysical
   microenvironment and remove substrate stiffness as a conflating factor
   in this study, fibroblasts were cultured on 2-kPa polyacrylamide
   hydrogels that were purchased from Matrigen and came chemically
   activated ready to bind matrix proteins and were subsequently coated
   with fibronectin (10 μg/ml). For immunocytochemistry of fibroblasts on
   hydrogels, cells were fixed with 4% paraformaldehyde and blocked with
   3% NGS + 0.1% Triton in 1X PBS for intracellular antigens. Primary
   antibody dilutions were prepared in 3% NGS in 1X PBS as follows: mouse
   anti-human vinculin (VIN-54, Abcam) at 1:200 and phalloidin conjugated
   with Alexa Fluor 546 (Invitrogen). The primary antibody incubation was
   done for 1 hour at room temperature, then washed three times with 1X
   PBS, and then stained with a secondary antibody for 1 hour at room
   temperature. The secondary antibody (goat anti-mouse Alexa Fluor 488,
   Invitrogen) was prepared in 3% NGS in 1X PBS at a dilution of 1:1000.
   DAPI was used as a counterstain at 300 nM in 1x PBS for 5 min at room
   temperature and then mounted in mounting medium ProLong Diamond
   Antifade (Thermo Fisher Scientific). Imaging was done at room
   temperature on a Nikon Eclipse Ti microscope with an UltraView VoX
   imaging system (PerkinElmer) using a Nikon N Apo LWD 40X water
   objective (numerical aperature: 1.15). Cell and focal adhesion area
   were measured using Fiji software.

Flow cytometry and FACS

   Fibroblasts were lifted from hydrogels for flow cytometry using TrypLE
   Express enzyme (Thermo Fisher Scientific) and subsequently stained with
   the Zombie Near-Infrared Fixable Viability dye (BioLegend) then fixed
   and permeabilized using the FIX & PERM Cell Permeabilization Kit
   (Thermo Fisher Scientific). After fixation, fibroblasts were surface
   stained for Thy-1 using the allophycocyanin (APC) anti–Thy-1 antibody
   (1:100, BioLegend). For α-SMA staining, after surface staining and
   washing, fibroblasts were incubated in the permeabilization buffer with
   the phycoerythrin anti–α-SMA antibody (1A4, R&D) at a 1:25 dilution.
   For analyzing cells from implanted biomaterials, implants were digested
   with Liberase TM (Roche) and stained with BV711 anti-CD45 antibody
   (1:100, BioLegend), BV605 anti-EpCAM antibody (1:100, BioLegend), and
   BV421 anti-CD31 antibody (1:100, BioLegend) to yield a fibroblast
   population that was defined as CD45^−, EpCAM^−, and CD31^− and
   subsequently stained with APC anti–Thy-1 antibody. All staining was
   done using fluorescence-activated cell sorting (FACS) buffer (1X PBS
   with 10% FBS and 0.05% sodium azide) and staining was done in the
   presence of Fc block (1:100, BioLegend). Flow cytometry was done on a
   BD LSR Fortessa, and cell sorting was done using an Influx Cell Sorter
   through the UVA Flow Cytometry Core Facility (RRID: SCR_017829).

scRNA-Seq and analysis

   Fibroblasts were lifted from hydrogels using TrypLE Express enzyme, and
   viability was confirmed to be greater than 80% using Trypan blue
   staining with the Countess II FL Automated Cell Counter (Thermo Fisher
   Scientific). Cells were resuspended in 0.04% UltraPure BSA in PBS
   (Thermo Fisher Scientific). A total of 1500 fibroblasts in each group
   were targeted. After 72 hours of cytokine treatment, cells will be
   collected and then we used the 10X Genomics Chromium platform for
   automated single-cell barcoding and library preparation on an
   eight-channel microfluidic chip. Then, we used the Illumina NextSeq 500
   Sequencing System for high-throughput sequencing that allows for 400M
   reads, permitting us to have broad coverage of each cell’s
   transcriptome. The Genome Analysis and Technology Core at UVA (RRID:
   SCR_018883) assisted in single-cell barcoding, library construction,
   and sequencing. Gene-barcode matrices were analyzed in R using Seurat
   v3 ([132]68). Cells were filtered for 2500 to 9000 reads per unique
   molecular identifiers and less than 10% of mitochondrial gene content.
   Significant principal components analysis of variation was calculated
   using JackStraw ([133]69) test with 100 repetitions, and clusters were
   defined using 20 principal components of variation. Our analysis was
   done combining together the untreated and treated groups so that we
   might be able to simultaneously analyze all subpopulations. Slingshot
   pseudotime trajectory analysis was used to determine the trajectory of
   fibroblasts across subpopulations ([134]55). KEGG pathway analysis uses
   transcript information to understand higher-order changes in pathways
   or biological systems, and this analysis was done for both the
   scRNA-Seq dataset and the publicly available dataset comparing WT
   versus Thy1^−/− dermal fibroblasts ([135]GSE211834). We used the
   Connectome package ([136]47) that helps explore cell-cell signaling
   networks within scRNA-Seq datasets and applied it to a scRNA-Seq
   dataset ([137]GSE175890) that compares different implanted biomaterials
   in a volumetric muscle loss model to identify bidirectional signaling
   between macrophages and fibroblasts in the presence of biomaterials
   ([138]46). Our analysis for differential expression analysis was
   determined using the nonparametric Wilcoxon rank sum test. Code is
   available on request.

Statistical analysis

   Statistics were performed using GraphPad Prism 9 (GraphPad, San Diego,
   CA). For multiple comparisons along one variable, we performed one-way
   analysis of variance (ANOVA) followed by Tukey’s post hoc test. For
   multiple variable comparisons, we performed two-way ANOVA followed by
   Tukey’s post hoc test. The number of mice used for each analysis was
   determined by power analysis (power = 0.80, α = 0.05). Statistical
   significance was defined at P values < 0.05. Data were visualized, and
   statistical analysis was done using GraphPad Prism (version 9).

Acknowledgments