Hydrogel-evoked FBRs vary and mimic CNS wound responses

We first compared cellular profiles of mild or severe CNS wound responses with FBRs of two structurally similar, synthetic, diblock copolypeptide hydrogels (DCH) that present either strongly cationic (DCH_K) or nonionic (DCH_MO) interfaces to host cells (Fig. 1a). Lysine-based DCH_K and methionine sulfoxide based DCH_MO are both well tolerated in vivo^8,25, but exhibit noticeably different FBRs. For comparison, mild or severe CNS wound responses were induced by injection of innocuous phosphate buffered saline (PBS), or N5-(1-iminoethyl)-l-ornithine (L-NIO) solution that creates a small focal ischemic stroke²⁷. PBS, DCH_MO, DCH_K, or L-NIO were injected into caudate putamen (CP) of mouse forebrain (Fig. 1b). The CP site is easily accessible, allows for consistent and reproducible hydrogel injections that are well tolerated by the mice, and is composed of neural tissue with neuronal cell bodies, myelinated axon bundles and a diversity of neuroglial making it an advantageous location for standardized CNS FBR assessments²⁵. Hydrogel FBRs and wound responses were characterized with immunohistochemistry (IHC) for CD13 to identify non-neural cells including stromal and peripheral myeloid lineage cells recruited as part of the sterile inflammatory response to tissue damage²⁸. CD13 positively identifies the full temporal range of non-neural tissue remodeling from acute inflammation through to chronic fibrotic lesion formation. In the healthy uninjured mouse brain, CD13 is expressed by pericytes and perivascular fibroblasts along cerebrovasculature and meningeal fibroblasts²⁹ (Supplementary Fig. 1a–d). CD13 is not expressed by neuroglia or neurons. Staining for glial fibrillary acidic protein (GFAP) identified reactive and border-forming astrocytes³⁰, P2RY12 identified CNS microglia and NeuN identified viable neurons.

At 1 week after injections, (i) PBS infusions were detectible as small areas of GFAP-positive reactive astrocytes; (ii) nonionic DCH_MO and cationic DCH_K, exhibited similarly sized deposits; and (iii) L-NIO created sharply demarcated areas of infarcted tissue equivalent in size to hydrogel deposits (Fig. 1c). DCH_MO and DCH_K deposits and L-NIO infarcts were all surrounded by similarly appearing borders of GFAP-positive astrocytes that clearly separated neural tissue containing NeuN-positive neurons from non-neural cores with CD13-positive cells and hydrogels (Fig. 1c, d). CD13-positive cells were not elevated at the centers of PBS injections or at interfaces between DCH_MO and host. By contrast, a dense CD13-positive cell capsule, 164 ± 19 µm in thickness, had completely surrounded and infiltrated the margins of DCH_K deposits, whereas CD13-positive cells had entirely filled L-NIO infarcts (Fig. 1c, d). Similar CD13-positive cell accumulation and non-neural/neural cell compartmentalization was observed 1 week after forebrain stab injury through cerebral cortex (Supplementary Fig. 1e). Quantification showed that total CD13 levels across the different compartments of infusion sites and adjacent neural tissue were similar after PBS and DCH_MO and were significantly higher after DCH_K and L-NIO (Fig. 1e; Supplementary Fig. 2). Notably, the significant increases in CD13 levels after DCH_K and L-NIO were confined to central non-neural tissue compartments of hydrogel plus infiltrating cells, and there was no significant difference in CD13 levels in surrounding neural tissue (defined as tissue containing neuroglia and neurons) adjacent to PBS, DCH_MO, DCH_K, or L-NIO (Fig. 1f). Moreover, although GFAP levels were significantly lower adjacent to PBS injections, there was no significant difference in GFAP levels adjacent to DCH_MO, DCH_K, or L-NIO (Fig. 1g).

These findings show that DCH-based hydrogels presenting precisely defined cationic (DCH_K) or nonionic (DCH_MO) interfaces to host cells exhibit FBRs whose cellular profiles and compartmentalization mimic those of normal CNS wound responses that are prominent or minimal, respectively, and that the FBR to different hydrogels can vary significantly. These findings also suggested that hydrogel surface chemistry and properties such as charge may influence hydrogel FBR, which we investigated next.

FBRs vary with definable hydrogel properties

We next characterized FBRs to various cationic, anionic, and nonionic hydrogels. We manufactured DCH that presented different polar side chains to interface with host cells but maintained consistent poly(l-leucine) hydrophobic blocks^8,24,25 (Supplementary Fig. 3a, b). We compared these DCH with various commercially available formulations commonly used in CNS applications: methylcellulose (MC)³¹, a hyaluronic acid/methylcellulose blend (HAMC)³², and a chitosan/β-glycerophosphate (CHIT) system³³. Hydrogels with comparable mechanical properties (see ref., ^8,25,34 and Supplementary Fig. 3c) were injected into mouse forebrain (Fig. 1b), and cellular FBR profiles were first characterized with IHC for CD13, GFAP, and NeuN (Fig. 2; Supplementary Figs. 2–5).

Qualitative and quantitative examination showed that nonionic MC, like nonionic DCH_MO, exhibited barely detectable levels of CD13-positive cells at interfaces with host tissue (Fig. 2a, c; Supplementary Fig. 4a). Moreover, anionic hydrogels HAMC as well as glutamate (E) inclusive DCH_E and DCH_MOE15 also exhibited barely detectable levels of CD13-positive cells at host interfaces (Fig. 2a, c; Supplementary Fig. 4a). However, in striking contrast, cationic hydrogels, DCH_K, DCH_MM, DCH_MOK15, DCH_MOK10, and CHIT all exhibited large rims of CD13-positive cells infiltrating into deposits around their entire host interface and had significantly higher CD13 levels in central non-neural compartments, which did not differ in magnitude across different cationic hydrogels (Fig. 2b, c; Supplementary Fig. 4b). To directly probe the relationship between cationic charge and FBR, we incorporated variable amounts of lysine (K) into nonionic DCH_MO as a statistical copolymer⁸ (Supplementary Fig. 3b). DCH_MOK15 and DCH_MOK10 both exhibited outer rims of infiltrating CD13 cells similar to DCH_K, but with more cells infiltrating into the deposit center, suggesting that small amounts of dispersed cationic charge attracted CD13 cells, whereas dense charge in DCH_K prevented these same cells from distributing throughout the bulk of the material. In addition, we generated cationic, methyl-sulfonium-based DCH_MM (Supplementary Fig. 3b)⁸, which also exhibited a rim of infiltrating CD13 cells similar to DCH_K, but in addition exhibited many DAPI-positive but CD13-negative cells that formed a distinct and separated layer of cells that directly interfaced with the hydrogel deposit surface (Fig. 2b). Similar CD13-negative cells at the material interface were not obviously present in other DCH, but were present within CHIT deposits (Fig. 2b), suggesting a distinct cellular infiltrate requiring further characterization as conducted below.

These findings show that hydrogel FBRs vary with definable hydrogel properties and implicate cationic charge presented at material interfaces with host tissue as a major factor determining FBR severity. At interfaces with host tissue, cationic materials exhibited a significantly greater loss of neural tissue containing NeuN-positive neurons in the FBR non-neural tissue zone that immediately surrounded depots, resulting in overall larger lesions and greater loss of viable neuropil (Fig. 2d; Supplementary Fig. 5). Remarkably, despite significantly elevated CD13 in the non-neural compartments of cationic hydrogels (Fig. 2a–c), there was no significant difference in either CD13 (Fig. 2d; Supplementary Fig. 5) or GFAP (Fig. 2e) levels, and no detectable quantitative difference in NeuN-positive neuron density (Supplementary Fig. 5), in the neuropil adjacent to either cationic, nonionic or anionic hydrogels, suggesting that neural tissue beyond the interface between hydrogel and host was not substantively differentially impacted. The unexpected finding that GFAP levels were indistinguishable in neural tissue adjacent to hydrogels that evoked markedly different FBR at host interfaces indicates that GFAP staining is not a sensitive or sufficient marker with which to evaluate and differentiate FBRs.

Hydrogel FBRs differ in inflammatory cell recruitment/persistence

To discriminate among diverse innate-immune inflammatory cells involved in hydrogel FBRs, we used IHC for: (i) CD45 to identify leukocytes and reactive CNS microglia³⁵; (ii) Iba-1 to identify blood-borne monocytes/macrophages and CNS microglia³⁵; (iii) P2RY12 to identify only microglia^20,36,37; (iv) Ly6B2 to identify neutrophils²¹; and (v) CD13 to identify myeloid lineage peripheral inflammatory cells such as monocytes and macrophages or fibroblast lineage cells (Fig. 3)²⁸. We examined combinations of these markers at 1 week after hydrogel injections (Fig. 1b).

For all hydrogels, the neural tissue immediately adjacent to deposits contained reactive, CNS-derived microglia that were Iba-1-, CD45-, and P2RY12-positive and CD13-negative that intermingled with astrocytes and neurons, remained within the neural tissue compartment and did not detectably infiltrate into any hydrogel (Fig. 3a, b; Supplementary Fig. 6a). P2RY12 expression was reduced in reactive microglia that dispersed directly within forming astroglia borders. Immediately abutting these microglia, hydrogel deposits exhibited variable intensities of inflammatory cell infiltrates that were CD45- and CD13-positive, and P2RY12-negative, and were peripherally derived myeloid lineage leukocytes and not CNS microglia (Fig. 3a, b, Supplementary Fig. 6a, b). A comparable segregation of blood-borne inflammatory cells into a non-neural tissue compartment away from neural tissue was also observed for the L-NIO stroke lesion (Supplementary Fig. 6c).

Different hydrogels exhibited remarkable variation in both the intensity and cellular phenotype of the blood-borne inflammatory response that they attracted. Non-ionic DCH_MO, which displayed a FBR comparable to all the other nonionic and anionic hydrogels evaluated, attracted only rare infiltrating CD45-positive leukocytes at interfaces with host tissue, and these were macrophage lineage (Iba-1-positive and CD13-positive) (Figs. 2a, 3a, b; Supplementary Fig. 6a). In contrast, cationic DCH_K and DCH_MM both attracted thick rims of infiltrating macrophage lineage cells (CD45-positive, Iba-1-positive, and CD13-positive) along their entire host interfaces (Figs. 2b, 3a, b; Supplementary Figs. 4, 6a). Immunohistochemical staining also revealed that the thick layers of DAPI-positive and CD13-negative cells infiltrating into DCH_MM and CHIT (Fig. 2b) were leukocytes (CD45-positive), but were not macrophage lineage (Iba-1-negative) (Fig. 3a), and instead were highly phagocytic and cytotoxic Ly6B2-positive neutrophils (Fig. 3b, c; Supplementary Fig. 6b). In striking contrast, nonionic DCH_MO and weakly cationic DCH_MOK10 or anionic DCH_MOE15, contained no detectable Ly6B2-positive neutrophils, whereas strongly cationic DCH_K attracted a few of these cells (Fig. 3c, d; Supplementary Fig. 6a).

To test whether inflammatory response intensity correlated directly with strength of cationic charge at hydrogel surfaces, we measured zeta potentials of DCH hydrophilic polymer chains. DCH_K, containing homopolymers of lysine, had a significantly greater positive charge at physiological pH than DCH_MM (containing statistical copolymers with 10% uncharged alanine) due to a higher total number of charged side chain groups along each hydrophilic chain (Supplementary Fig. 6d). However, DCH_K had significantly lower levels of neutrophils and equivalent levels of macrophages compared to DCH_MM (Fig. 3d). Non-ionic DCH_MO had no detectable charge, no infiltration of neutrophils and only rare macrophages (Fig. 3a–c; Supplementary Fig. 6d).

Persistence of peripheral inflammatory cells at hydrogel deposits was evaluated at 3 weeks after injection (Supplementary Fig. 7a, b). Notably, while Ly6B2-positive neutrophils were mostly resolved at CHIT deposits by 3 weeks, thick rims of these acute inflammatory cells were sustained at DCH_MM deposits with a distribution that was not obviously different to that at 1 week. DCH_MM deposits also remained largely unresorbed at 3 weeks, whereas CHIT injections were completely infiltrated by cells at this timepoint (Supplementary Fig. 7c).

These findings show that the intensity and duration of the inflammatory cell components of CNS FBRs varied markedly in response to deposits of different hydrogels and that cationic charge at material–host interfaces is an important factor in attracting blood-borne phagocytic leukocytes. Nevertheless, charge magnitude alone does not determine the nature or distribution of phagocytes attracted.

Hydrogel FBRs involve stromal cells and fibrosis

To characterize fibrosis associated with hydrogel FBRs, we compared a representative hydrogel with limited detectable CD13 levels (DCH_MO) to a representative hydrogel with a pronounced but focal rim of CD13 cells (DCH_K) using IHC for cell surface and extracellular matrix molecules, collagen-1a1, fibronectin, laminin, and galectin-3 with staining for CD13 at various times after hydrogel injections (Figs. 1b, 4a–e). Collagen-1a1, fibronectin, and laminin demarcate stromal cells³⁸, whereas galectin-3 can demarcate multiple cell types including stromal and inflammatory cells^39,40,41. In uninjected normal neural tissue, (i) CD13-stained stromal cells along blood vessels and at the meninges, (ii) collagen-1a1 and laminin lightly decorated cell surfaces along blood vessels and around neurons, (iii) fibronectin was present at large vessels and at the meninges, and (iv) galectin-3 was essentially undetectable⁴² (Supplementary Fig. 8a).

a, b Detail images of stromal-associated markers, collagen-1a1 (Col1a1), fibronectin, laminin, and galectin-3 (Gal-3) and their relationship to non-neural CD13-positive cells at the tissue interface of DCH_MO a and DCH_K b at 1 week after injection (hydrogels, G). Colocalization of markers with CD13-positive cells is seen as white staining. c Quantification of total Col1a1 and Gal-3 staining at 1 week after hydrogel injection. ***P = 0.0003 and =0.0002 for DCH_MO versus DCH_K for Col1a1 and Gal-3 stromal markers, respectively, two-way ANOVA with Bonferroni. Graphs show mean ± s.e.m with individual data points showing n = 5 and 7 mice for DCH_MO and DCH_K, respectively. d Detail image showing that adjacent to DCH_MO, Gal-3 colocalizes with a narrow band of GFAP-positive astrocytes that border the gel at 1 week. Gal-3 and GFAP colocalization is seen as white staining. e, f Detail images showing the temporal evolution of Gal-3 expression at the interface of DCH_MO e and DCH_K f at acute (48 h, 48 h), subacute (1 week, 1wk) and chronic (6 weeks, 6wk) timepoints after injection. Gal-3 and GFAP colocalization is seen as yellow staining. Gal-3 and CD13 colocalization is seen as white staining.

At 1 week after injection, tissue adjacent at interfaces with nonionic DCH_MO exhibited staining for collagen-1a1 and laminin only along blood vessels in a manner similar to normal neural tissue, whereas fibronectin and galectin-3 were moderately elevated in narrow rims along host–gel interfaces (Fig. 4a, c–e). In contrast, the thick capsule of CD13-positive cells circumscribing cationic DCH_K exhibited significantly greater levels of collagen-1a1, laminin, fibronectin, and galectin-3 (Fig. 4b, c). Around DCH_K, galectin-3 was robustly expressed by essentially all CD13-positive cells, whereas collagen-1a1, laminin, and fibronectin demarcated subsets of stromal cells in outer margins (Fig. 4b) that interfaced directly with the forming astrocyte limitans border. An interior layer of CD13- and galectin-3-positive but collagen-1a1, fibronectin, and laminin-negative cells, likely peripheral inflammatory cells, interacted directly with the DCH_K surface (Fig. 4b). This inflammatory and fibrotic cell stratification around DCH_K is reminiscent of the cellular organization in CNS abscesses and traumatic injury lesions^40,43. CD13- and galectin-3-positive cells persisted chronically in the non-neural tissue compartment of resorbed DCH_K for at least 6 weeks (Fig. 4f). Interestingly, the GFAP-positive astrocyte limitans border that formed the direct host interface with nonionic DCH_MO also extensively expressed galectin-3 and this expression persisted chronically for at least 6 weeks (Fig. 4d, e; Supplementary Fig. 8b).

These findings show that DCH_MO evokes minimal levels of fibrosis, whereas cationic DCH_K evokes a stronger fibrotic FBR with elevated levels of multiple matrix molecules at the interface with preserved neural tissue. In addition, we identified a possible conserved function for galectin-3 in isolating foreign bodies from parenchymal neural tissue regardless of the severity of the FBR, with its expression occurring rapidly following hydrogel injection and present only in cells forming the material–tissue interface, regardless of cell type (inflammatory cells, stromal cells, or astrocytes). Galectin-3 is known to be associated with fibrosis and wound healing in various tissue injuries including kidney, liver, and heart⁴⁴ and is significantly upregulated in macrophages/microglia and reactive astrocytes after various traumatic CNS injuries^42,45,46,47. Here, we identify galectin-3 to also be an important constituent of CNS FBRs to materials.

Tissue damage drives FBR and causes acute amyloid formation

We next looked for associations between FBR, inflammation, fibrosis and host tissue loss, and for other potential molecular features shared across FBR and CNS wound response. To do so, we first used mice expressing reporter protein transgenically targeted to host astrocytes via Aldh1l1-Cre-ERT2 and evaluated host neural tissue damage and changes in various molecular markers, including accumulation of amyloid precursor protein (APP), a robust marker of axonal injury⁴⁸ and one of its breakdown products, beta-amyloid (Aβ)⁴⁹. We compared representative hydrogels displaying minimal or pronounced FBRs with the CNS wound response to L-NIO stroke lesions at various timepoints.

At 48 h after injection, DCH_MO exhibited a minimal zone of host neural tissue loss averaging only 25 µm, the equivalent of one or two cells in thickness, whereas DCH_K exhibited a significantly greater, but still relatively small zone of host neural tissue loss of ~150 µm in thickness, as evidenced by: (i) astrocyte and neuron loss but persistence of vasculature, (ii) presence and phagocytosis of reporter protein debris, (iii) extravasation of blood-borne fibronectin⁵⁰, (iv) upregulation of CD13 expression along thin PECAM-1-positive blood vessels within the damaged tissue zone, and (v) infiltration of blood-borne phagocytic leukocytes (Fig. 5a, b; Supplementary Fig. 9a–e).

a Detail images of the material–tissue interface for DCH_MO and DCH_K at 48 h after injection. ALDH1L1-tdT reporter and GFAP identify host astrocytes and the loss of these cells demarcates regions of host tissue damage. At 48 h, CD13-positive cells in this damaged tissue region predominately label actively remodeling vasculature. b Quantification of radial thickness of tissue damage around DCH_MO and DCH_K (n = 9 mice per group). ***P = 0.0001 Welch’s (unequal variance) two-tailed t test (t = 6.56, df = 8.74). c, d Detail images for DCH_MO, DCH_K, and L-NIO-induced stroke at acute (48 h, 48 h), subacute (1 week, 1wk), and chronic (6 weeks, 6wk) timepoints after injection, comparing staining for amyloid precursor protein (APP) c or amyloid-beta (Aβ) d with GFAP and CD13. e Images show Aβ, Iba-1, and CD13 at the interface of DCH_MO, DCH_K, and DCH_MM with host tissue at 1 week after injection. f Quantification of Aβ at 1 week after DCH_MO, DCH_K and DCH_MM. NS or *P = 0.0153 and =0.0227 for DCH_MO versus DCH_K and DCH_MM, respectively, one-way ANOVA with Bonferroni, n = 4 mice per group. All graphs shows mean ± s.e.m with individual data points superimposed. G hydrogels, ax axons.

Further characterization of the extent of axonal injury was provided by APP staining. APP is low or undetectable by IHC in healthy CNS tissue, but increases markedly after axonal injury^48,51. After 48 h, DCH_MO displayed only small amounts of APP accumulation at the hydrogel-host interface (Fig. 5c). In contrast, APP was present throughout the narrow zone of neural tissue damage induced by DCH_K and APP accumulation was prominent in damaged axon bundles close to DCH_K deposits in a manner comparable with L-NIO stroke lesions (Fig. 5c). By one week, APP staining had decreased substantially and was detectable only at the interface of neural to non-neural tissue, and APP had returned to baseline at 6 weeks across all conditions (Fig. 5c).

The formation of Aβ via cleavage of APP by β- and γ-secretase enzymes is increasingly recognized as a normal occurrence during CNS wound responses^52,53. We evaluated the progression of APP to Aβ (Fig. 5d). At 48 h, Aβ was sparse at the hydrogel-tissue interface for both DCH_MO and DCH_K, whereas L-NIO stroke lesions already showed clear Aβ accumulation at the borders of infarcted and spared neural tissue (Fig. 5d). By one week, Aβ was barely detectable around DCH_MO and was comparable to PBS injections where Aβ was constrained to the track of tissue disrupted by the injection pipette (Supplementary Fig. 10a, b). At 1 week after DCH_K, DCH_MM injections and L-NIO stroke lesions, Aβ accumulation was most prominent within the layers of CD13-positive inflammatory cells that infiltrated areas of neural tissue damage (Fig. 5d, e). In addition, in all conditions, there was prominent colocalization of Aβ within Iba-1-positive reactive microglia in adjacent spared neural tissue, suggesting phagocytosis by these cells as well (Fig. 5e, Supplementary Fig. 10c). Quantification confirmed significantly higher levels of Aβ at one week for DCH_K and DCH_MM compared with DCH_MO (Fig. 5e, f). Increased neutrophil accumulation and persistence in DCH_MM compared with DCH_K (Fig. 3c, d) was not associated with any increase in Aβ formation (Fig. 5e, f), suggesting that Aβ is formed as a result of acute neural tissue injury and not by ongoing chronic inflammation. At 6 weeks, Aβ levels had reduced markedly and were evident only within the persistent non-neural tissue lesions for DCH_K and L-NIO stroke (Fig. 5d).

These finds show that the inflammatory and fibrotic FBRs evoked by cationic hydrogels were driven by narrow but measurable zones of host neural tissue damage occurring along hydrogel-host interfaces soon after injection and that this was essentially absent with the nonionic DCH_MO. These findings also show that APP accumulation and Aβ formation around hydrogel deposits occurred as a result of damage to neural tissue and axonal injury. These observations are consistent with growing evidence that APP accumulation and subsequent Aβ formation are part of a conserved innate wound and FBR. Although the specific functions of Aβ production in this context are not yet defined, recent evidence suggests that Aβ may exert antimicrobial activities^52,53.

Astrocyte limitans borders isolate FBRs from neural tissue

Neural tissue contains several types of glial cells that become reactive around damaged CNS tissue and participate in FBRs. Astrocytes are well known to form “scar borders” that serve as limitans borders to isolate CNS lesions¹⁹. Microglia contribute to border formation and exert phagocytic functions²⁰, whereas oligodendrocyte progenitors (OPC) proliferate and repair myelin²². We identified astrocytes and microglia by IHC for GFAP and P2RY12, respectively, and OPC by staining for OLIG2 and the reporter protein, td-Tomato (tdT), which had been transgenically targeted by NG2(CSPG4)-Cre-ERT2⁵⁴. We compared combinations of markers and probed blood–brain barrier (BBB) integrity at 1 and 6 weeks after hydrogel injection or L-NIO stroke (Figs. 6 and 7).

a, b Survey and detail images show qualitative reductions in size of DCH_K deposit and L-NIO-induced infarct, but not of DCH_MO deposit at 6 weeks after injection c, d. Survey and detail images show DCH_MO deposit remains unresorbed up to 12 weeks after injection. e Quantification of GFAP-positive cell border thickness at lateral (gray matter) or medial (internal capsule) borders at 6 weeks. *P = 0.0469 and **P = 0.0021 (lateral), = 0.0012 (medial), and ***P = 0.0001 versus DCH_MO on same side; not significant (NS) between DCH_MO on either side, **P = 0.0012 for border location effect across all samples, two-way ANOVA with Tukey (n = 5, 4, 4 mice for DCH_MO, DCH_K, and L-NIO, respectively). f Quantification of change in hydrogel radius for DCH_MO and DCH_K from 2 to 42 days. NS, ***P = 0.0001 and ***P < 0.0001 for DCH_K versus DCH_MO at 7 and 42 days, respectively; NS for DCH_MO at 7 versus 42 days and for DCH_MO versus DCH_K at 2 days, two-way ANOVA with Bonferroni. g Quantification of change in total CD13 for DCH_MO and DCH_K from 2 to 42 days. **P = 0.0067 and ***P < 0.0001 for DCH_K versus DCH_MO at the same timepoints. NS between the various timepoints for DCH_MO, two-way ANOVA with Bonferroni. For both f and g, n = 9, 10, 5 mice for DCH_MO at 2, 7, and 42 days; n = 9, 12, 5, 4 mice for DCH_K at 2, 7, 21, 42 days. h Schematic of microgel particle (MP) synthesis involving inverse emulsion of polyethylene glycol (PEG)-based thiol and acrylate functionalized oligomers that react via Michael addition. MP can be readily suspended in hydrogels. i, j Survey images of FBR to DCH_MO loaded with MP at 1 week and 6 weeks show that incorporation of MP into DCH_MO leads to recruitment of CD13-positive cells and hydrogel resorption. k Quantification of effect of incorporation of MP(+) on total CD13-positive cell response for DCH_MO and DCH_K at 1 week. NS and ***P < 0.0001, two-way ANOVA with Bonferroni, n = 10, 6, 12, and 9 mice for DCH_MO, DCH_MO + MP, DCH_K and DCH_K + MP, respectively. All graphs show mean ± s.e.m with superimposed individual data points.

GFAP-positive astrocytes had begun to form distinct borders around all hydrogel deposits by 1 week after injections, and these borders persisted and consolidated by 6 weeks. Astrocyte borders around hydrogels were similar in appearance to borders formed around L-NIO infarcts and clearly segregated persisting hydrogel material and CD13-positive stromal and inflammatory cells from neural tissue containing NeuN-positive neurons (Figs. 1c, d; 2a, b; 6a, b; 7a, b). Notably, astrocyte borders adjacent to DCH_MO and other nonionic hydrogels interfaced directly with gel surfaces with limited intervening fibrosis or inflammation, whereas astrocyte borders adjacent to DCH_K and other cationic gels or L-NIO interfaced with stromal and inflammatory cells (Figs. 2a, b, 4, 6a, b). Astrocyte borders around hydrogels and L-NIO infarcts were similar in appearance to astrocyte limitans borders that separate healthy neural tissue from non-neural stromal cells along all interfaces of normal CNS with meninges (Supplementary Fig. 1c)⁵⁵. Astrocyte border thickness at 6 weeks was proportionally greater with increasing severity of FBR, such that DCH_MO displayed the thinnest astrocyte borders while DCH_K and L-NIO both had similar sized borders of more than double the thickness (Fig. 7e). Across all conditions at 6 weeks, lateral astrocyte borders formed by gray matter astrocytes, were thinner compared with medial borders that recruited white matter astrocytes from the internal capsule (Figs. 1b, 7a, b, e).

Neural parenchyma extending from astrocyte borders exhibited initially moderate astrocyte reactivity indicated by elevated GFAP, which declined significantly over time and by 6 weeks was minimal in all cases (Figs. 6a, b; 7a, b). Reactive microglia and OPC intermingled with astrocytes along borders and adjacent neural tissue, but did not infiltrate into hydrogels or contribute to volumes of CD13-positive fibrosis and inflammation at any timepoint examined (Fig. 6a–d and Supplementary Fig. 11a). At 6 weeks a thin, single cell layer of non-neural tissue (CD13-positive) that expressed stromal cell markers PDGFRβ and fibronectin interfaced directly with DCH_MO and the astrocyte limitans border (Supplementary Fig. 11b), whereas these cells contributed to large volumes of non-neural fibrotic tissue in DCH_K and L-NIO strokes (Supplementary Fig. 11b, c).

Astrocyte border formation is essential for re-establishing BBB integrity after CNS injuries³⁰. To probe BBB integrity adjacent to hydrogel deposits, we stained for mouse immunoglobulin (IgG) and albumin, the two most abundant serum proteins³⁰. As expected, after 1 week and prior to border formation³⁰, IgG, and albumin staining were increased in neural parenchyma around hydrogel injections and L-NIO stroke lesions, and were significantly higher around DCH_K, compared with DCH_MO (Fig. 6e, f; Supplementary Fig. 12a–c). By 6 weeks, serum protein levels around DCH_MO were indistinguishable from those in uninjured tissue and were restricted to a thin layer of non-neural tissue that interfaced with the astrocyte limitans border. In contrast, around L-NIO stroke lesions and cationic DCH_K, serum proteins remained significantly elevated in neural tissue at 6 weeks (Fig. 6e, f, Supplementary Fig. 12a–c).

These findings show that astrocytes rapidly form limitans borders around hydrogels in a manner similar to borders formed around ischemic or traumatic tissue damage, or that exist along meningeal non-neural stromal tissue around healthy CNS. Astrocytes, microglia, and OPCs become reactive in neural tissue adjacent to hydrogel deposits or stroke lesions, but do not migrate into the non-neural fibrotic tissue compartments that persist chronically. Glial reactivity and BBB leakiness into neural tissue adjacent to astrocyte borders persist longer adjacent to cationic materials that generate substantial inflammation and fibrosis compared with nonionic materials that do not.

FBR determines hydrogel resorption or persistence

To evaluate the relationship between FBR and hydrogel resorption or persistence, we compared nonionic DCH_MO and cationic DCH_K deposits at various times after injection (Figs. 1 and 7). Quantification showed that after 48 h, DCH_MO, and DCH_K exhibited deposits of similar size. After 1 week, DCH_MO deposits had decreased in size by 30%, but remained at this size after 6 weeks and persisted essentially unchanged after 12 weeks (Fig. 7a–d, f). In contrast, after 1 week DCH_K deposits had decreased in size significantly by over 50% and continued to steadily decrease in size until there was no detectable deposit remaining at 6 weeks (Fig. 7a, b, f). Over time, DCH_MO exhibited no increase in CD13 levels above baseline, whereas DCH_K evoked a steadily increasing infiltration of CD13-positive inflammatory and stromal cells that proceeded in a concentric fashion from hydrogel-tissue interfaces inwards until the entire deposit was consumed (Figs. 1c, d, 3a, 4b, 7a, b, g; Supplementary Fig. 7a). By comparison, the wound response to L-NIO-induced ischemia attracted a pronounced infiltration of CD13-positive cells that rapidly filled the entire volume of damaged tissue and then persisted (Figs. 1c, d, 7a, b).

These findings show that DCH_MO and other nonionic hydrogels persist and are not efficiently resorbed in vivo because they do not attract sufficient CD13-positive phagocytes. In contrast, DCH_K and other cationic hydrogels attract these phagocytic leukocytes to their interfaces with host tissue and are gradually resorbed from the outside in and become replaced by fibrosis. DCH_MM was an exception to this general cationic hydrogel resorption trend and instead showed minimal hydrogel resorption after 3 weeks (Supplementary Fig. 7c). This chronic hydrogel persistence coupled with the unresolved Lys6B2-positive neutrophils at the material–tissue interface suggests that DCH_MM is unable to be readily cleared due to frustrated phagocytosis, possibly owing to toxicity towards or repulsion of phagocytosing cells. Thus, hydrogel properties determine FBR features, which in turn determine hydrogel resorption or persistence.

Microparticles alter FBR and hydrogel resorption

Incorporating nano/microparticles (MPs) into hydrogels may be a useful tool to enhance the control of delivery of multiple and diverse molecular cargos independently from a single construct to the CNS^18,56,57. As introduction of particles may alter the physiochemical properties of hydrogels into which they are loaded, we examined the effects on FBR and resorption of hydrogels laden with polyethylene glycol (PEG)-based nonionic MP formulated via a standard inverse emulsion thiol-ene Michael addition process (Fig. 7h)^58,59,60. MP, with an average diameter of 3.3 µm, imparted a modest increase in mechanical properties to DCH but did not alter injectability (Supplementary Fig. 13a–f). MP at high concentrations were non-toxic to neural progenitor cells in vitro (Supplementary Fig. 13d). As described above, nonionic DCH_MO and HAMC on their own evoked minimal fibrotic or inflammatory responses and formed long persisting deposits. Addition of nonionic MP to these hydrogels caused pronounced recruitment of peripheral inflammatory cell infiltration in the form of both CD13-positive macrophages and Lys6B2-positive neutrophils, as well as rapid hydrogel resorption and fibrotic replacement (Fig. 7i–k; Supplementary Fig. 13g–k). Nevertheless, the FBR of MP loaded DCH_MO did not detectably increase acute neural tissue loss or axonal injury (Supplementary Fig. 13g, h). Similar sized non-toxic MP as the ones evaluated here (~3 µm diameter) have previously demonstrated a high susceptibility to phagocytosis in vitro by macrophages⁶¹. Incorporating MP into nonionic hydrogels may stimulate similar size recognition programs in vivo thus leading to the recruitment of phagocytes by mechanisms other than those associated with host neural tissue damage at the hydrogel-host interface as was the case for cationic hydrogels. These findings show that adding such MP to hydrogels that would otherwise be ignored by phagocytes can induce a phagocytic FBR that steadily resorbs the material and replaces it with fibrosis.

FBR alters hydrogel molecular delivery to CNS parenchyma

To characterize FBR effects on delivery of molecular cargos, we first examined the biodistribution of model non-bioactive molecules released into neural tissue adjacent to hydrogel deposits, and second evaluated the efficacy of bioactive growth factor delivery. As non-bioactive molecules, we used biotinylated dextran amines (BDA) of different molecular weights because they are non-immunogenic, fixable, easily detected by IHC, and because 10 kDa (BDA-10) and 70 kDa (BDA-70) BDAs exhibit hydrodynamic radii that approximate bioactive protein growth factors and therapeutic monoclonal antibodies, respectively^62,63.

BDAs injected in PBS are well documented to diffuse locally throughout CNS neural parenchyma and along perivascular spaces, and are taken up by neurons as well as by microglial and perivascular phagocytic cells in a size dependent manner⁶⁴. A detailed comparison of the biodistributions of BDA-10 and BDA-70 released from PBS, DCH_MO, or DCH_K is presented in Supplementary Figs. 14–16. In brief, under all conditions, BDAs were detectable only in the ipsilateral hemisphere. Notably, delivery of BDAs of either size via DCH_MO or DCH_K resulted in significant differences both in the depth of penetration of BDAs into neural tissue and in cellular uptake such that BDAs of either size released from DCH_MO were found to be more concentrated locally near hydrogel deposits, whereas BDAs released from DCH_K showed increased radial diffusion (Supplementary Figs. 14–16), (Supplementary Fig. 16). BDAs of different sizes showed dissimilar cellular uptake profiles such that BDA-70 was consumed preferentially by phagocytes in neural tissue while BDA-10 showed enhanced neuron uptake. Nevertheless, the hydrogel used for delivery also had substantial effects on cellular uptake of each BDA (Supplementary Fig. 16). BDA-70 showed low but clearly detectable uptake by neurons when released from DCH_MO but no neuron uptake when released from DCH_K. BDA-10 released from DCH_MO showed extensive uptake by neurons as well as axon labeling throughout the adjacent neuropil and low accumulation in phagocytes. By contrast, BDA-10 released from DCH_K showed low uptake by neurons, no obvious axon labeling and pronounced uptake by phagocytes (Supplementary Figure 16). These differences may relate to the increased activation and number of microglia in neural tissue adjacent to DCH_K deposits, which could either stochastically or actively favor greater uptake of BDA-10 by phagocytes over neurons. These findings indicate that hydrogels with different FBRs exhibit differences not only in the distribution of molecular delivery into neural tissue, but also in the type of cells that may be primarily targeted.

We therefore next compared the biodistribution of BDA-10 released from different hydrogels that displayed escalating and unique severities of FBR: DCH_MO, DCH_K, DCH_MM, and CHIT (Fig. 8a). We quantified the proportion of BDA present in the non-neural compartment defined as the area of hydrogel deposit and its non-neural surrounding tissue, versus the neural compartment defined as the neural tissue surrounding the deposits containing viable neurons and different neuroglia (Fig. 8b–d, Supplementary Fig. 17). At 2 weeks after injection, DCH_MO, with the least noxious FBR showed the highest total delivery of BDA into the neural compartment, and yielded greater uptake in neurons, which was highest local to the hydrogel-tissue interface (Fig. 8b–d). DCH_K exhibited less neural parenchyma delivery than DCH_MO but more than DCH_MM and CHIT, and BDA was preferentially taken up by microglia rather than neurons (Fig. 8e, f). DCH_MM and CHIT, which show the most severe FBR (Fig. 3b, d), displayed: (i) significantly increased accumulation of BDA in non-neural compartments associated with the FBR (Fig. 8b); (ii) reduced uptake of BDA by local neurons (Fig. 8c); and (iii) greater BDA accumulation in macrophages (CD13-positive/CD68-positive cells) (Fig. 8e). Notably, DCH_K, DCH_MM and CHIT exhibited pronounced BBB leakage with extended distribution of serum albumin through neural parenchyma compared with DCH_MO (Supplementary Fig. 17c, d) and degree of BBB leak correlated with increased BDA biodistribution throughout the brain and increased inflammatory cell phagocytosis of BDA (Fig. 8d, e, Supplementary Fig. 17a). Serum proteins such as albumin bind systemically delivered biological materials such as bioactive proteins, small molecule drugs and nanoparticles, and alter their biodistribution⁶⁵. In addition, serum proteins are proinflammatory in neural tissue^19,22. We also evaluated the biodistribution of BDA-10 released from MP that had been loaded into DCH_MO. The increased non-neural and phagocyte dominated FBR associated with MP inclusion into nonionic DCH_MO correlated with significantly decreased BDA-10 accumulation in neural tissue at 1 week and extensive consumption of BDA-10 by CD13-positive cells (Supplementary Fig. 18). Together, these data show that hydrogels with pronounced non-neural FBRs have reduced molecular delivery efficacy to neural tissue and that this reduction scales with the severity of the FBR.

a Detail images of BDA (10 kDa) biodistribution released from DCH_MO, DCH_K, DCH_MM, and CHIT at 2 weeks. b Quantification of BDA in tissue compartments adjacent to hydrogels. ***P < 0.0001 versus DCH_MO; not significant (NS) for DCH_MM versus CHIT c Quantification of the normalized number of BDA-positive neurons at 2 weeks. *P = 0.0375 (DCH_MM) and = 0.0413 (CHIT) and NS (DCH_K) versus DCH_MO. d Quantification of the radial distance to the 50% percentile BDA-positive neuron from the hydrogel-tissue interface. ***P < 0.0001 versus DCH_MO. e Quantification of BDA-positive inflammatory cells at 2 weeks. *P = 0.0036, **P = 0.0005 (CD13), = 0.0008 (CD68), ***P < 0.0001 versus DCH_MO. *P = 0.0074 (CD13), = 0.003 (CD13 + CD68) and ***P < 0.0001 (CD68) for DCH_K versus DCH_MM. For all graphs in b–e, n = 6 mice per group. f Detail images showing phenotype of BDA-positive cells adjacent to hydrogel deposits. g Detail images showing increase in cholinergic (ChAT-positive) neuron size in striatum by NGF delivered from DCH_MO. h Quantification of increase in cholinergic neuron size normalized to contralateral side at 1 week after injection of 1 µL NGF (1 µg/µl) releasing hydrogels. ***P < 0.0001 and **P = 0.0054 (DCH_MM) and = 0.0029 (CHIT) versus DCH_MO, NS for DCH_MM, and CHIT versus untreated, *P = 0.0429 for DCH_MM and = 0.0233 for CHIT versus DCH_K respectively, NS between DCH_MO and DCH_K (n = 5, 6, 5, 5, and 6 mice for DCH_MO, DCH_K, DCH_MM, CHIT, and untreated, respectively). i Principal component Analysis (PCA) for DCH_MO, DCH_K, DCH_MM, and CHIT hydrogels incorporating data from molecular delivery and immunohistochemistry evaluations. j Graphical representation of the positions of hydrogels along PC1 axis (accounting for 64.22% of total variance) with the positive direction representing effective neural molecular delivery while the negative direction represents molecular delivery to inflammatory cells, ***P < 0.0001 for all versus DCH_MO, ***P = 0.0007 for DCH_K versus DCH_MM and ***P < 0.0001 for DCH_K versus CHIT, NS for DCH_MM versus CHIT (n = 6 mice per group). For all graphs, data are mean ± s.e.m. Statistical analysis using one (c, d, h, j) or two-way (b, e) ANOVA with Bonferroni.

At last, we evaluated the effect of FBR on hydrogel-mediated delivery of bioactive molecules in vivo. Basal forebrain cholinergic neurons are exquisitely sensitive to nerve growth factor (NGF) levels and atrophy when deprived of NGF and hypertrophy when exposed to exogenous NGF (Fig. 8g)^34,66,67. To compare the efficacy of NGF delivery by various hydrogels, we quantified the size of local striatal cholinergic neurons. NGF delivered from DCH_MO and DCH_K stimulated significant 40% increases in ipsilateral cholinergic neuron size relative to uninjected controls, whereas NGF delivered from DCH_MM and CHIT, which attract a severe non-neural FBR, showed highly variable effects with no overall significant increase (Fig. 8h). Incorporating data from across the molecular delivery and FBR phenotype characterizations into a principal component analysis (PCA) showed a significant inverse correlation between neural tissue molecular delivery efficacy and the severity and intensity of hydrogel FBR-associated inflammation (Fig. 8i, j, Supplementary Fig. 19).

These findings show that delivery of molecular cargo from hydrogels to CNS tissue is influenced by the nature and severity of the material-evoked FBR. In particular, an increased recruitment of peripherally derived phagocytes into hydrogel deposits results in increased consumption of cargo molecules and reduced neural tissue delivery. Further, a minimal hydrogel FBR with little or no BBB leakage results in very local delivery with a higher proportion of targeting of cargo molecules to neurons. In contrast, more severe FBRs with pronounced BBB leakage contributes to increased dispersal of delivered molecules throughout larger volumes of neural parenchyma, and to greater targeting of those molecules to phagocytosis by microglia and perivascular cells.

FBR alters hydrogel molecular delivery and wound healing in CNS stroke injury

Injectable hydrogels are being tested extensively to deliver drugs, including growth factors, to CNS injuries. The influence of hydrogel FBRs in determining effectiveness of such treatments on neural repair outcomes remains largely uncharacterized. Unfavorable FBR effects may mask or dilute the efficacy of treatments delivered by hydrogels. To determine hydrogel FBR effects on molecular delivery to CNS injuries we injected hydrogels loaded with NGF into mouse CP at 48 h after initiating a large focal stroke lesion via a 2 µL injection of L-NIO (Fig. 9a). We evaluated local striatal cholinergic neuron survival and size, as well as neuropil changes, as measures of NGF-responsiveness at one week after hydrogel injections into stroke lesions (Fig. 9b, c). NGF delivery from all hydrogels stimulated a significant and ~50% increase in size of surviving, ipsilateral ChAT-positive neurons compared with untreated L-NIO only (Fig. 9d). NGF delivered from DCH_MO but not DCH_K or DCH_MM also stimulated a small but significant increase in the size of contralateral ChAT-positive neurons (Fig. 9d). L-NIO stroke injury triggered a significant loss of ChAT-positive neurons in the ipsilateral striatum; and NGF delivered from DCH_MO, but not from DCH_K or DCH_MM, rescued the number of ChAT-positive neurons such that they were comparable to uninjured contralateral striatum (Fig. 9e). To investigate effects of NGF delivery on density of cholinergic axon networks we assessed ChAT staining intensity averaged across the neuropil. NGF delivered from DCH_MO but not DCH_K or DCH_MM stimulated an increase in the mean ChAT intensity in the ipsilateral neuropil compared to the untreated stroke (Fig. 9f). Applying a PCA that combined all ChAT related parameters for individual animals showed a dominant PC1 that described 73.70% of total variation across the cohort (Fig. 9g). PC1 results for different hydrogels showed a remarkably similar hydrogel efficacy correlation to that seen for NGF delivery to uninjured healthy striatum. Animals treated with NGF using DCH_MO showed the most effective NGF delivery outcomes, observed as higher positive PC1 values (Fig. 9g). DCH_K displayed significantly reduced delivery efficacy compared to DCH_MO but still provided benefits over the untreated L-NIO only cohort. Delivery via DCH_MM failed to significantly alter outcome compared with untreated stroke.

a Schematic summarizing experimental paradigm. b, c Images showing cholinergic (ChAT-positive) neurons in striatal stroke at 1 week after hydrogel (G) injection. d Quantification of mean cholinergic neuron size ipsilateral and contralateral to stroke and hydrogel injection. ***P < 0.0001 for hydrogels versus L-NIO only and not significant (NS) between hydrogels ipsilaterally, *P = 0.0179 (L-NIO), = 0.0317 (DCH_K) and = 0.004 (DCH_MM) versus DCH_MO contralaterally. e Quantification of cholinergic neuron number in striatum ipsilateral and contralateral to stroke and hydrogel injection at 1 week. *P = 0.0192, **P = 0.0027, ***P = 0.0002 versus DCH_MO ipsilaterally, NS between all samples in contralateral striatum. f Quantification of mean ChAT intensity in neuropil ipsilateral and contralateral to stroke and hydrogel injection, *P = 0.0442 (DCH_K), = 0.0203 (DCH_MM), ***P < 0.0001 (L-NIO) versus DCH_MO ipsilaterally, NS between all samples in contralateral striatum. g PC1 score for L-NIO only and hydrogels for PCA comparing all ChAT-neuron parameters. Positive direction represents increased ChAT responsiveness to NGF delivery, ***P < 0.0001, **P = 0.0011, NS versus L-NIO only_, *P = 0.0404 and ***P = 0.0004 for DCH_K and DCH_MM versus DCH_MO, respectively. h, i Images showing altered phenotypes of CD13-positive lesion cores (LC), astrocyte border, and preserved neural tissue of striatal strokes at 1 week after hydrogel injections. j CD13-positive cell intensity plots measured radially from the center of the stroke lesion. Data are mean ± s.e.m, with s.e.m. represented as light shaded banded areas. Location of astrocyte (GFAP) border is maximum of mean GFAP intensity plot for L-NIO only. The integral of plot to left and right of GFAP border is total CD13 in lesion and hydrogel FBR respectively. k, l Quantification of total CD13-positive cells in lesion k and FBR l, NS, and ***P < 0.0001 versus L-NIO only, *P = 0.0173 for DCH_MO versus DCH_K. Baseline in l. is mean total CD13 in L-NIO only. m Quantification of percentage neuropil remaining following stroke and hydrogel treatment in immunohistochemistry sections at center of lesion, ***P < 0.0001, *P = 0.0385 and NS versus L-NIO only. For all graphs, data are mean ± s.e.m. with superimposed individual data points showing n = 8 mice per group. Statistical analysis using one (g, k, l, m) or two-way d–f ANOVA with Bonferroni.

The natural wound healing response stimulated following stroke may be altered by hydrogel FBRs. To characterize the perturbation of the sterile inflammatory response after stroke caused by hydrogel FBRs, we used CD13 to identify non-neural cells, GFAP to define the astrocyte border and neurofilament (NFM) to demarcate neural tissue (Fig. 9h, i). In untreated strokes, distinct lesion compartmentalization was observed as described earlier with CD13-positive cells separated from NFM-positive tissue by discrete GFAP-positive astrocyte borders. Wound healing progression, defined by the distribution and total numbers of CD13-positive cells in lesion cores, was unaltered by the injection of DCH_MO (Fig. 9j, k). By contrast, depots of both DCH_K and DCH_MM substantially modified the lesion by reducing the total number of total CD13 cells in the lesion core and stimulating a renewed increase in infiltration and persistence of CD13-negative, DAPI-positive inflammatory cells, likely neutrophils, into this tissue volume. Surrounding the non-neural lesion core, DCH_MO caused no significant increase in baseline CD13 compared with the L-NIO only control. By contrast, the two cationic materials, DCH_K and DCH_MM, stimulated significant increases in CD13-positive cell infiltrates that formed rims around the entire host interface of material deposits that persisted immediately adjacent to lesion cores at 1 week (Fig. 9l). These cationic materials also caused additional damage to ipsilateral neuropil beyond that attributed to the stroke, whereas no significant increase in neuropil loss was detected for DCH_MO (Fig. 9m).

These data show that effectiveness of molecular delivery from hydrogels in stroke lesions is governed by FBR severity in a manner similar to observations in uninjured tissue. Notably, introduction into stroke lesions of hydrogels that evoked FBRs of increasing severities caused escalating destruction not only of neural tissue beyond focal lesion cores but also extensive destruction of peripherally derived cells that infiltrate into lesion cores to initiate wound healing (Fig. 9h–m). This effective re-injuring of CNS lesions caused by hydrogels with severe FBRs may prolong the duration of acute inflammation, prevent re-establishment of BBB and leave the recipient with increased susceptibility to infection and further neural tissue damage thus potentially negating therapeutic effects derived from local molecular delivery.