Mammalian musculoskeletal regeneration is associated with reduced inflammatory cytokines and an influx of T cells

Whether the immune response to injury contributes to tissue regeneration is not well understood. We quantified systemic and local cytokines during ear pinna repair to provide the first comprehensive comparison of the immune response to injury between mammalian regeneration (A. cahirinus and A. percivali) and fibrotic repair (M. musculus). Importantly, by comparing laboratory-reared and wild-caught animals we identified responses specifically associated with healing outcome. Fibrotic repair showed a greater local release of IL-6, CCL2 and CXCL1. Conversely, regeneration showed decreased circulating IL-5, IL-6, IL-17, CCL3 and CXCL1 and increased local IL-12 and IL-17. The differential IL-6 response was substantiated by increased pSTAT3 during the inflammatory phase of fibrotic repair and with blastema formation and tissue morphogenesis in Acomys. COX-2 inhibition was not sufficient to induce regeneration. Interestingly, a unique influx of lymphocytes was coupled with regeneration and RNA-expression analysis suggested they were regulatory T cells. Together, the data support regeneration-specific inflammation and T cell responses in Acomys.


INTRODUCTION
Fibrotic repair is associated with elevated amounts of circulating IL-5, IL-6 and CCL3. 165 Using our cytokine assay, we first compared circulating serum cytokine concentrations 166 from uninjured animals among groups (species and source population) to establish a systemic 167 baseline for each group ( Figure 1A). A total of 13 cytokines were compared as CSF2 was not 168 present in the serum of any species. While many baseline concentrations were similar between 169 groups, immune-challenged animals (i.e., wild) exhibited higher IL-4, IL-6, CCL2, and TNFα 170 compared to laboratory-reared animals ( Figure 1A). Interestingly, the Mm-Kenya animals were 171 a transitional group between Mm-UKY and Mm-Wild for TNFα and IL-4 ( Figure 1A).

172
Heightened concentrations of IL-6, TNFα and IL-4 support previous pathogen exposure and a 173 possibility of current infection [43,44]. Thus, Mm-Kenya, Mm-Wild and Ap had a relatively 174 activated immune system, while Mm-UKY and Ac possessed a more naïve immune system [45]. 175 There were no consistent differences between regenerators and non-regenerators ( Figure 1A).
Next, we quantified the systemic injury response for each cytokine compared to its 177 baseline, beginning 24 hr (D1) after injury and over the next twenty days ( Figure 1B). In most 178 cases (except IL-2, IL-6, IL-17 and CXCL1), there was no effect of day (Supplemental Table 2), 179 indicating that the immediate systemic response persisted for 20 days. Animals with a more 180 naïve immune system showed increased IL-2 and TNFα, and decreased IL-1α compared to a 181 relatively activated immune system ( Figure 188 and a dampened pro-inflammatory cytokine response. 189 Resident cells and infiltrating immune cells secrete cytokines that likely polarize the 190 injury microenvironment to support regeneration or fibrotic repair [12,13,15,16,34,46,47]. 191 Thus, to quantify local cytokine concentrations we assayed tissue lysate collected throughout the 192 healing response. IL-1α could not be compared because baseline concentrations were above the 193 upper limit of quantification in more than 80% of samples, indicating that IL-1α in the ear pinna  Table 3). Supporting an inflammatory response in all groups, 204 several pro-inflammatory cytokines (IL-6, TNFα) and myeloid chemotactic factors (CCL3,CSF2 205 and CXCL1) showed an increase compared to baseline between D1 and D3 that then decreased 206 to baseline or below between D5 and D20 ( Figure 2). There was also an overall decrease 207 compared to baseline for IL-5 and a small but significant change from baseline for IL-2 and IL-4 While there were some differences among groups for the timing of resolution, IL-1β, TNFα, and 223 CCL2 were similar to or below baseline at D10 for each species (Figure 3). IL-6 also followed 224 this pattern; however, there was a differential response where Ac remained elevated through D20 225 while all other species decreased below baseline ( Figure 3). 226 At D20, during tissue morphogenesis, the only cytokines that showed a differential 227 response were IL-12 and IL-17 that were increased during regeneration compared to fibrotic 228 repair ( Figure 3). The anti-inflammatory cytokine IL-4 did not differ over time with respect to 229 regenerative ability, suggesting that the differences in pro-inflammatory cytokine release is likely 230 not an IL-4 mediated response. Our results suggest that subtle differences in how cytokines are 231 deployed in the injury microenvironment can distinguish regeneration or fibrotic repair. These 232 data suggest that strong, acute increases in the pro-inflammatory cytokines IL-6, CCL2 and 233 CXCL1 are associated with fibrosis, while the release of IL-12 and IL-17 during tissue 234 morphogenesis is associated with regeneration.

236
Regeneration is associated with an early burst of T cell influx to the injury site. 237 The release of IL-12 and IL-17 into the regenerative microenvironment suggested 238 enhanced T cell activation during regeneration [48,49]. Therefore, we quantified T cell influx 239 into uninjured and healing tissue from our laboratory populations of Mus (Mm-UKY) and 240 Acomys (Ac) using flow cytometry with an antibody to the extracellular portion of the T cell 241 marker CD3. We observed significant differences in CD3+ cells in injured tissue between 242 species over time n=57;species: Df=1,F=49.49 P<0.001;day: Df=6,243 F=89.07, P<0.001; species*day: Df=6, F=21.49 P < 0.001) ( Figure 4A). In uninjured tissue,

244
Mm-UKY had 10-times more CD3+ cells compared to Ac (Tukey-Kramer HSD post-hoc test, Df=6, t=6.21, P<0.001) ( Figure 4A). While the total number of CD3+ cells that infiltrated the 246 wound was higher in Mm-UKY compared to Ac, there was a greater fold change relative to D0 247 for CD3+ cells during regeneration compared to fibrotic repair ( Figure 4B). Ac exhibited a 248 monophasic response to injury starting on D1 with a 78-fold influx of T cells that peaked on D3 249 and remained above baseline at D15. Mm-UKY showed a biphasic response with peak influx of 250 10-fold at D7 that returned to baseline at D15 ( Figure 4B). Importantly, at D15, when IL-12 was 251 increased (Figure 2), the influx of CD3+ cells remained high in Ac compared to Mm-UKY 252 ( Figure 4B). 253 We next used immunohistochemistry with an antibody specific to the intracellular portion 254 of the CD3 receptor to assess the spatial distribution of T cells during acute inflammation and 255 morphogenesis ( Figure 4C-E). First, these data confirm the greater influx of CD3+ cells at D5 256 and D15 in Ac compared to Mm-UKY ( Figure 4C-E). Second, in Mm-UKY most CD3+ cells 257 were associated with the epidermis and were rarely observed distal to the amputation plane 258 ( Figure 4C). On the other hand, CD3+ cells in Ac were present in the epidermis and dermis, and 259 regularly observed in healing tissue distal to the amputation plane ( Figure 4E). At D15, CD3+ 260 cells were found in the epidermis and dermis of both species ( Figure 4D, F). Interestingly, 261 CD3+ cells associated with the epidermis in Mm-UKY ( Figure 4G) exhibited a spindle-shape 262 morphology compared to a rounded shape in Ac ( Figure 4H). There also appeared to be more 263 CD3+ cells in the dermis of Ac compared to Mm-UKY ( Figure 4D, F), and the CD3+ cells 264 tended to localize near regenerating hair follicles in Ac ( Figure 4I). Attempts to characterize 265 individual T cell phenotypes during regeneration using flow cytometry and IHC using 19 266 commercially available antibodies, supported significant differences in antibody-epitope binding 267 between species that prevented further T cell phenotyping by receptor subtype in Acomys 268 (Supplemental Table 5). Therefore, we interrogated a comparative injury RNAseq dataset for 269 differential expression of T cell associated transcripts between Mus and Acomys [40]. While 270 expression for genes associated with non-lymphocyte immune cell populations were generally 271 similar between species, several transcripts associated with T cells and natural killer cells were 272 increased in Acomys and decreased in Mus in response to injury ( Figure 4J). Increased 273 expression of Cd8, Ctla4, Il2ra, Foxp3, and Tnfrsf4 specifically suggested an activated cytotoxic 274 and regulatory T cell response during regeneration but not fibrotic repair ( Figure 4J). During 275 fibrotic repair, Cd4 was differentially increased at D5 and D10 suggesting the presence of CD4 276 helper T cells not present during regeneration ( Figure 4J). Together, these data demonstrate that 277 regeneration was associated with a proportionally greater influx of CD3+ cells that accumulate 278 quickly at the injury site and that specific subtypes of activated T cells were preferentially 279 associated with regeneration. 280 281 STAT3 is activated independently from IL-6 during blastema formation 282 We also sought to test our observation that strong induction of the pro-inflammatory 283 cytokine IL-6 was associated with the acute inflammatory phase of fibrotic repair. To do this, 284 we assayed for IL-6 signaling using STAT3 phosphorylation ( Figure 5A-F). STAT3 is 285 phosphorylated in response to the ligand IL-6 binding its membrane receptor, which activates 286 signal transduction in target cells [50]. Corroborating our ELISA quantification for IL-6 in the 287 tissue microenvironment, we found that pSTAT3 increased 8-fold in response to injury in Mm-288 UKY during the acute inflammatory phase ( Figure 5A, B). Similarly, during fibrosis when IL-6 289 concentrations resolved in Mm-UKY, pSTAT3 began to decline toward baseline ( Figure 5A, B).

294
To determine the cellular localization of STAT3 phosphorylation, we assayed for 295 pSTAT3 using immunohistochemistry during the acute inflammatory phase (D2) and new tissue  Figure 5A, B), we found that nearly every epidermal cell in Mm-UKY appeared 302 positive for pSTAT3 whereas less than half of the epidermal cells were positive in Ac ( Figure   303 5C', D'). The internal tissue compartments (e.g., dermis, cartilage, muscle and adipose) at D2 304 were similar between species with approximately half of the total cells positive for pSTAT3. At 305 D15, only a few pSTAT3 positive cells were present in Mm-UKY and they were isolated to the 306 epidermis distal to the amputation plane ( Figure 5E, E'). In contrast, pSTAT3 positive cells 307 were widespread throughout the blastema in Ac ( Figure 5F, F'). Together, these data support 308 stronger IL-6 mediated STAT3 activation in Mm-UKY compared to Ac during the acute 309 inflammatory phase and increased STAT3 activation during blastema formation.

310
Greater increases in IL-6 and CXCL1 during the acute inflammatory phase of fibrotic 311 repair in M. musculus suggested that these molecules might antagonize a potential regenerative 312 response. Previous studies have shown that a balance in these molecules regulate wound healing 313 as IL-6 and CXCL1 are potent pro-inflammatory molecules and hyper-elevated concentrations 314 after injury are attributed to aberrant healing and chronic inflammation [51][52][53]. Additionally, 315 genetic ablation of IL-6, the IL-6 receptor, or the CXCL1 receptor (CXCR2), causes severely 316 delayed re-epithelialization, scab formation and abhorrent wound healing in cutaneous and 317 incisional wounds [54][55][56][57]. IL-6 signaling activates several downstream mediators of 318 inflammation including cyclooxegenase-2 (COX-2), and its enzymatic products can amplify the 319 inflammatory response [58]. To test if COX-2 activity promotes fibrosis in the ear pinna, we 320 used our ear punch assay in Mm-UKY treated healing tissue with Celecoxib, a specific and

334
In this study, we performed a comprehensive cytokine characterization of the immune 335 response to injury where identical injuries in closely-related species undergo two different 336 healing responses: regeneration or fibrotic repair. Importantly, our experimental design 337 leveraged a comparison of animals with an activated and naïve immune system in order to 338 identify species-specific cytokine changes that were associated with regeneration and not due to 339 an environment-immunity interaction. Our analyses showed that regardless of healing outcome, 340 injury induced a common set of pro-inflammatory factors (IL-6, and TNFα) and chemokines 341 (CCL3, CSF2 and CXCL1) during the acute inflammatory phase of fibrosis and regeneration.

342
Additionally, we observed similar responses between healing outcomes for IL-2, IL-4 and IL-5 343 in the tissue microenvironment. While this supports that fibrotic repair and regeneration share 344 inflammation during the healing process, we also found significantly greater responses for IL-6, 345 CCL2 and CXCL1 during fibrotic repair compared to regeneration. In contrast, regeneration was 346 uniquely associated with local increases in IL-12 and IL-17. Supporting our cytokine analysis, 347 we found that regeneration was associated with a strong influx of T cells during acute 348 inflammation compared to fibrotic repair and that regeneration-competent T cells were closely 349 associated with the dermis during blastema formation. This latter point suggests that T cells may  Recent studies comparing immune profiles between laboratory-reared and pet-store or 353 wild-caught M. musculus demonstrate that non-laboratory strains have more CD44 + effector T 354 cells, memory T cells and circulating neutrophils [60,61]. Although neither group directly 355 measured serum cytokines from the different populations, the elevated baseline concentrations of 356 IL-4, IL-6, CCL2 and TNFα that we measured in circulation from immune-challenged animals 357 support larger active populations of effector and memory T cells. These data support that our 358 immune-challenged group have been exposed to more pathogens than the laboratory-reared mice 359 which is undoubtedly the case. In addition to the increased baseline concentrations of these 360 cytokines, we also found significant differences in the response to injury for IL-1α, IL-2 and 361 TNFα between animals with an activated or naïve immune system. Studying wild-caught 362 populations enabled us to identify responses that accurately reflected phenotypic differences 363 between species, rather than differences that could be explained by immune status. Of particular 364 importance was our inclusion of wild-caught A. percivali that indicate increases in TNFα, CCL2 365 and CXCL1 are not inhibitory to regeneration. Additionally, we observed high variation in 366 cytokine concentrations across our dataset, indicating that the immune response to injury could 367 be confounded by individual variation. In other words, researchers should not expect to identify 368 clear transition phases based on time after injury with small sample sizes. Ultimately these data 369 support that injury elicits a local cytokine response that is dependent of baseline immune status 370 with respect to release of cytokines that in turn effects the timing of events but does not change 371 healing outcome.

372
Acute inflammation is a necessary component of the innate immune response designed to 373 fight invading microbes by recruiting leukocytes from circulation and activating local myeloid 374 and lymphoid cells. Our analyses demonstrate that injury induces an acute inflammatory 375 response regardless of healing outcome that resolves within ~10 days; a timeframe in line with 376 human and rodent wound healing studies [7,62]. In particular, we found the local release of 377 CCL3, CSF2 and CXCL1 in all groups, which are known to be potent chemokines for 378 monocytes, macrophages and neutrophils. Moreover, the local release of IL-6 and TNFα 379 supports the presence of activated macrophages and neutrophils as a common injury response.

380
Supporting our previous work that neutrophils infiltrate injured spiny mouse tissue slower 381 compared to laboratory mice [15], regeneration was associated with delayed release and a 382 reduced maximal fold change in IL-6 and CXCL1 compared to fibrotic repair. Additionally, we 383 did not detect a CCL2 response during regeneration, and IL-6, CXCL1 and CCL2 are known to 384 positively regulate the speed of re-epithelialization [54-56], which we find to be delayed at most 385 five days in A. cahirinus compared to M. musculus [40]. Thus, while acute inflammation is a 386 component of regeneration and fibrotic repair, the cellular differences are likely attributed to 387 reduced pro-inflammatory cytokines released into the microenvironment.

388
Corroborating our observation that the IL-6 response was weaker during regeneration 389 compared to fibrotic repair, we found diminished activation of STAT3 during acute 390 inflammation (D1-10) in spiny mouse epidermis compared to mouse. Interestingly, we observed 391 an increase in pSTAT3 during blastema formation, whereas pSTAT3 levels declined during 392 fibrotic repair. Furthermore, during tissue morphogenesis at D15 many blastemal cells were 393 STAT3 positive. Given that IL-6 concentrations did not appreciably increase during blastema 394 formation or tissue morphogenesis the increase in STAT3 activity is likely independent of IL-6.

395
STAT3 is activated through multiple pathways (e.g. leukemia inhibitory factor, epidermal 396 growth factor, palette derived growth factor, IL-10, IL-17, etc.). Although IL-17 increased in A. 397 percivali after D10, it did not increase in A. cahirinus suggesting it is not responsible for the late 398 phase of STAT3 phosphorylation. Given that STAT3 signaling is multifaceted, one potential 399 biological link is that STAT3 activity is necessary for satellite-cell activation and axon 400 regeneration in mammals [63][64][65]. Interestingly, the expression of Sal4-a factor necessary for 401 blastema maintenance in Xenopus and Ambystoma-is regulated by pSTAT3 [66][67][68]. While Sal4 does not have a mammalian homolog, this data supports that activation of STAT3 in 403 regenerating tissue is an evolutionary conserved mechanism and interrogating unique STAT3 404 targets in spiny mice may uncover mechanisms that regulate blastema formation in mammals.

405
Inhibition of downstream signaling induced by IL-6 / CXCL1, such as arachidonic acid 406 metabolism by COX-2, has been shown to reduce fibrosis post epidermal injury (e.g. incisional, 407 cutaneous and chronic pressure wounds) [69][70][71]. Celecoxib treatment to inhibit COX-2 in the 408 present study may have slowed re-epithelialization. Additionally, while the total area of fibrosis 409 was not different between celecoxib-and vehicle-treated animals there appeared to be a small 410 reduction in the total amount of collagen produced in celecoxib-treated animals from reduced 411 intensity of picrosirius staining. However, similar to previous reports, reduction in COX-2 412 activity did not induce regeneration, supporting that inflammation is not the ultimate inhibitory 413 barrier.

414
In addition to the magnitude increase in IL-6 and CXCL1, our analyses found that 415 increased local CCL2 was specific to fibrotic repair. CCL2 was first identified as a monocyte-416 specific chemoattractant to sites of injury and infection, and activates macrophages [72,73].

417
CCL2 also attracts neutrophils and supports neutrophil-dependent tissue damage [72]. As such, 418 the amount of CCL2 that is released into an injury microenvironment regulates the healing 419 response and studies support there is a positive relationship between CCL2 and the amount of 420 fibrosis during fibrotic repair [74][75][76]. However, a careful balance must be maintained as CCL2 421 knockout mice do not heal wounds [77]. Thus, it is possible that the reduced IL-6, CCL2 and 422 CXCL1 responses are responsible for reduced fibrosis in spiny mice. Although these key factors 423 appear to interact in the hierarchy of the progression of fibrotic repair, the paracrine mechanism 424 of how they would activate dermal fibroblasts remains unknown. It is likely another cell-type, 425 such as a macrophage or T cell, is mediating the signal.

426
In addition to this study, two studies have quantified cytokines during regeneration-one 427 in axolotl limbs [12] and the other in spiny mouse dorsal skin wounds [34]. Godwin et al. (2013) 428 used a mouse cytokine array to analyze regenerating salamander limbs and found that all but two 429 cytokines detected reached peak amounts within 48hrs of injury and that every cytokine returned studies support that release of CCL3 and TNFα in tandem with a differential inflammatory 435 response occurs prior to tissue regeneration. However, our comparative analyses also suggested 436 that the magnitude of the increase in IL-6 and CXCL1 might serve as early indicators of a 437 fibrotic repair trajectory. For example, the IL-6 response to injury, although present, was small 438 and CXCL1 did not respond during both axolotl limb and spiny mouse skin regeneration.

439
Finally, our cellular analysis uncovered a surprisingly rapid adaptive immune response 440 measured as an early influx of T cells in regenerating compared to non-regenerating species.

441
Importantly, our findings support that the arrival of T cells in spiny mice is concurrent with the 442 arrival and proliferation of monocytes [15], which suggests there is a regenerative-competent T 443 cell response that is different from a fibrotic T cell response. Contrary to hypotheses suggesting will create a framework to begin testing how the immune response functions during complex 466 tissue regeneration in a mammalian model. We believe that modulating the immune response at 467 the injury microenvironment will be an essential piece to inducing epimorphic regeneration in 468 tissues that naturally heal by fibrotic repair.      Inc., Richmond, VA) and given autoclaved water and a 3:1 mixture by volume of 14% protein 598 mouse chow (Teklad Global 2014, Envigo) and black-oil sunflower seeds (Pennington Seed Inc.,

599
Madison, GA) [91]. Additionally, the air within facility was filtered, and the animals were wire cages with pelleted pine bedding and given autoclaved water and 18% protein mouse chow.

607
The animals acclimated to captivity for at least twenty-one days before any experiments were 608 started. The air within the facility was filtered and the animals were exposed to a 12:12h L:D 630 631

Sample collection and preparation
We used a 4 mm biopsy punch to create a hole through the ear pinna, as previously 633 described [40]. Healing ear tissue was collected on D0, 1, 2, 3, 5, 10, 15 and 20. To minimize 634 circadian effects, animals were injured between 10:00 and 12:00, and samples were collected 635 between 11:00 and 15:00. Animals were deeply anesthetized with 5% (v/v) isoflurane and a 636 maximal amount of blood was collected by cardiac puncture using a 25-guage needle. An 8 mm 637 biopsy punch was used to harvest healing ear tissue.

638
To isolate serum, blood was collected into a serum separator tube (#454243, Greiner bio-639 one, Kremsmünster, Austria) and allowed to clot for at least 45 minutes, followed by Inc., Troy, NY), centrifuged at 10,000 x g for 15 minutes to pellet insoluble protein, and the 651 soluble protein was separated into a new tube. The total protein was quantified by bicinchoninic 652 acid assay (#23225, Thermo Scientific) with a standard curve created from the same stock of 653 bovine serum albumin, and then the protein lysate was stored at -80°C or on dry ice until 654 analysis.

655
Cytokine assay 657 To assess the immune response to injury in multiple species, we evaluated methods that: 658 1) used minimal sample, 2) measured local (tissue lysate) and systemic (serum) samples, 3) 659 measured several cytokines at once, 4) differentiated the magnitude and type of immune 660 response during an ear punch assay, and 5) exhibited cross-reactivity among the study species.

661
We used a custom-designed, multiplexed, sandwich ELISA array (Quansys Biosciences, Logan subsequent analyses. If the average value was above the lower limit of detection and the pixel 705 intensity co-efficient of variation between duplicates was greater than 15%, the sample was re-706 assayed on another plate and a new average calculated. Initially, we re-assayed tissues samples 707 below the limit of detection with a greater amount of total protein, but in most cases, additional 708 protein did not equate to quantifiable antigen, suggesting that there was a minimal amount of 709 antigen in those samples. Thus, to maximize use of the plates, we opted to quantify a greater 710 total number of samples and assayed each sample at one dilution. Antigens below the lower 711 limit of detection were recorded as "not present", and to calculate ratios they were assigned the 712 largest value of the lower limit of detection for that antigen across all plates assayed [97].      Table 1: One-way ANOVA analyses of log-transformed uninjured serum data.

808
Groups listed in comparison column represent groups for which data was quantified and could be 809 compared. P-values indicate where at least one group is significantly different from another 810 group. Tukey-Kramer HSD post-hoc tests were used for pairwise comparisons are summarized 811 in Figure 1A (See Supplemental File).  Table 2: Two-way ANOVA analyses of log-transformed serum time series data.

815
Groups listed in comparison column represent groups for which data was quantified throughout  None -all undetectable 0 n/a n/a n/a n/a n/a n/a CCL2  the results are summarized in Figure 2 (See Supplemental File).
There was a stronger increase in CCL3 a muted increase for IL-12 and CXCL1 in immune-naïve (solid lines) compared to immune-primed animals (intermittent lines). Non-regenerators (black) had stronger increases for IL-6, CCL2 and CXCL1 compared to regenerators (red). Additionally, IL-17 decreased in non-regenerators and increased in regenerators. Data represent mean and S.E.M. for at least n=5 animals per species per timepoint. The dashed line at Y=1 represents no change compared to D0 and the yellow boxes represent the inflammation resolution window.

Figure 4 The regeneration microenvironment is primed by greater T cell influx and TREG signature.
(A) Comparison of total CD3+ cells quantified by flow cytometry from disassociated ear pinna and (B) the ratio relative to uninjured tissue for M. musculus (black) and A. cahirinus (red). Data represent mean and S.E.M. and n=4 or 5. An * denotes P<0.05 for pairwise comparison within the day between species for Holm-Sidak posthoc test. (C-I) Representative immunohistochemistry for CD3 (red) counterstained with DAPI (gray) at the proximal wound margin (amputation plane can be determined from the end of the cartilage-indicated by the dotted line) from D5 and D15 after injury of M. musculus (C, D) and A. cahirinus (E, F). More T cells (yellow arrowhead) were present throughout the wound bed and were mainly found in the dermis of A. cahirinus compared to M. musculus. The T cells associated with epidermis (boundaries indicated by the dotted line) tended to be spindle-like in M. musculus (G), while rounded in A. cahirinus (H). The dermal T cells in A. cahirinus also tended to be in close proximity to regenerating epidermal appendages (I). N=4 and bar equals 200 μm (C-F) or 20 μm (G-I). (J) Heatmap of differential gene expression compared to uninjured tissue suggests that the regeneration microenvironment contains a substantial NK, CD8+ and TREG cell response while fibrotic repair has a CD4+ cell response. Data comes from a previously published analysis [34].