Comparative Proteomic Analysis in Scar-Free Skin Regeneration in Acomys cahirinus and Scarring Mus musculus

The spiny mouse, Acomys cahirinus displays a unique wound healing ability with regeneration of all skin components in a scar-free manner. To identify orchestrators of this regenerative response we have performed proteomic analyses of skin from Acomys and Mus musculus before and after wounding. Of the ~2000 proteins identified many are expressed at similar levels in Acomys and Mus, but there are significant differences. Following wounding in Mus the complement and coagulation cascades, PPAR signaling pathway and ECM-receptor interactions predominate. In Acomys, other pathways predominate including the Wnt, MAPK, the ribosome, proteasome, endocytosis and tight junction pathways. Notable among Acomys specific proteins are several ubiquitin-associated enzymes and kinases, whereas in Mus immuno-modulation proteins characteristic of inflammatory response are unique or more prominent. ECM proteins such as collagens are more highly expressed in Mus, but likely more important is the higher expression of matrix remodeling proteases in Acomys. Another distinctive difference between Acomys and Mus lies in the macrophage-produced arginase 1 is found in Mus whereas arginase 2 is found in Acomys. Thus, we have identified several avenues for experimental approaches whose aim is to reduce the fibrotic response that the typical mammal displays in response to wounding.

proteomic analysis of normal Acomys and Mus skin. To elucidate whether or not the protein profiles would reveal intrinsic biological differences between Acomys and Mus before wounding we performed Gene ontology (GO) enrichment analyses with total proteins detected from both species, according to their location in the cell components ( Fig. 2A) and related biological functions (Fig. 2B) at day 0. The cellular locations of the identified proteins were highest for the cytoskeleton and mitochondrion but showed a similar distribution between the two species. Likewise, the biological functions of the identified proteins were highest for protein localization, protein transport and oxidation reduction, but showed a similar distribution between species.
A list of common and unique proteins is shown in Table S2 revealing that there were very similar protein profiles in Acomys and Mus with regard to the presence of the most abundantly reported mouse skin proteins such Global protein profiling between Acomys and Mus skin over 14 days after full thickness skin wounding. We first compared the protein expression profiles to assess the trends between the two species during wound healing by a principal component analysis (PCA) of the common proteins (a total of 1545). This showed differences in the overall expression profiles between Acomys and Mus through 14 days (Fig. 3). Interestingly, the protein expression levels in Acomys from 0 to 3 days were clearly separated but, at 5, 7 and 14 days were clustered closely. Correspondingly, Mus at 0 day was well separated, at 3, 5, and 7 days clustering was apparent but not at 14 days which unexpectedly showed a return to the day 0 profile. This suggests that by day 14 the Mus skin has effectively completed its scarring process, but the Acomys skin is still in the process of new protein production and regeneration. To assess the statistical significance associated with biological variation from three biological replicates at the different time points, the coefficient of variations (CV) were determined and presented in the supplemental Fig. S3. This showed the high reproducibility across all samples.
We next compared the global changes of the common proteins between Acomys and Mus over 14 days (see Fig. 4A-D). This showed no clear changes over the time periods, but common proteins were highly biased in Mus towards negative fold changes. This was not the case in Acomys where the fold changes were more evenly distributed than Mus between positive and negative.
KEGG analysis of the differentially expressed proteins in Acomys and Mus skin during wound healing. Proteins were considered as differentially expressed in Acomys and Mus if they exhibited a fold change during the four time periods of >1.6 or <-2 with p-value < 0.01 among the biological replicates. When we analyzed the counts of differentially expressed proteins (DEPs) in the two species in terms of their cell locations, despite being similar in the normal skin at day 0 ( Fig. 2) it was apparent that during regeneration/scarring differences appeared (Fig. 5). Whereas the DEPs in Acomys localized to cytosol, extracellular matrix, ER and cytoskeletal components, DEPs in Mus were far more strongly localized to the extracellular matrix and mitochondrial  components. This suggests that these two cellular locations characterize Mus scarring and the more regenerative phenotype of Acomys is characterized by the ER and higher cytoskeletal and cytosolic representations.
To obtain further insight into the functional roles for DEPs associated with wound healing in Acomys and Mus, we carried out GO-BP and KEGG pathway analyses. We first made pathway annotations based on the KEGG database for both positive and negative DEPs. Twenty-five categories were enriched (Fig. 6), among which fourteen were particularly prominent, in Acomys. On the other hand, nine were enriched in Mus with only three prominent, the hub proteins enriched in pathways in our study are given in Table S3. The most notably enriched categories in Acomys were those involved in tight junction formation, endocytosis, the ribosome pathway, the proteasome pathway, Wnt signaling, MAPK signaling and vasopressin-regulated water reabsorption. In contrast, there was little enrichment of pathways in Mus overall, except for the complement and coagulation cascades, PPAR signaling, (Table S3) ECM-receptor interactions and metabolic pathways (Fig. 6).
The fold changes of levels of proteins representative of five different pathways during different times from day 0 for Acomys and Mus are shown as volcano plots in Fig. 7A-D to show the changes over time rather than a static plot. These figures showed the trends of DEPs changes at different time points.
Ribosomal proteins showed clearly different behavior, being elevated in Acomys but depressed in Mus. Spliceosome proteins were elevated in Acomys through 14 days, but in Mus were initially (0-7d) little changed and then elevated after 14 days. The ubiquitin/proteasome pathways were elevated in Acomys throughout all time points. The complement proteins are elevated in both species but there are more representatives in Mus and the ER components show changes in both Acomys and Mus but at different time points.

Profiles of intermediate filaments (IFs) in Acomys and Mus. Keratins were highly abundant in both
species and twenty-three are listed in Table S4. The amounts of various cytoskeleton keratins showed little change through 14 days with a couple of exceptions, but cuticular keratins (K31, K32, K33b, K35 and K36) were significantly higher in Acomys compared to Mus at day 0, subsequently showing fairly constant values over 14 days (Table 1). Vimentin, a fibroblast marker prominent in both species, was 1.5-fold elevated in Acomys from day 0 to day 7, and 2.5-fold upregulated at day 14 whereas Mus showed almost unchanged levels at the same time period (Table 1). Interestingly the only Wnt detected was Wnt 9b in Acomys. The small GTPases RhoA and Rac1 were detected at significantly higher levels in Acomys at most time points (Table 1).

Regulatory pathways of immune/inflammatory responses in Mus.
Innate immune response-related proteins, especially those related to the complement pathway, were up-regulated in Mus. 16 complement and coagulation factors and 6 immuno-modulatory proteins were detected in Mus, with many showing temporary increases over 3-7 days before decreasing at day 14. Exclusive detection of complement and coagulation factors (C1S, C4BP, C8A, C8G, C9, CFD, and F12) as well as significantly elevated levels of C2, C3, CFI, FGA, FGB, FGG, F2, F10 and PLG were observed in Mus. Of these 22 proteins 10 were absent from Acomys extracts and the temporary increases in levels was again apparent but less pronounced (Table 1). Myeloperoxidase as a neutrophil marker was highly increased in both species. However, arginase 1 was only identified in Mus and arginase 2 only in Acomys (Table 1). 10 proteases were identified and showed very different www.nature.com/scientificreports www.nature.com/scientificreports/ levels of expression between Acomys and Mus. Cathepsin G, mast cell protease 4 and tryptase beta 2 were only expressed in Acomys, whereas matrix metalloprotease 9, plasma kallikrein and pro-cathepsin H and cathepsin L1 were solely detected in Mus (Table 1). Cathepsins B, D, and Z were expressed in both species, they were at constant high levels in Mus but increased through 14 days in Acomys. A total of 17 serine protease inhibitors (SERPINs) were quantified in our study (Table 1 and S5). Strikingly, the vast majority were not detected in Acomys at any stage, for example only 2 of the 13 listed in Table S5 were identified in Acomys

Collagenous composition in extra cellular matrix (ECM).
Next, we investigated collagen levels in the two species and 10 collagens, all alpha isoforms, were identified. Only 1 could not be detected in Acomys, COL 17A1, and the remainder increased throughout the 14 days period in both species, but with a greater increase seen in Mus. The largest increase in Mus over Acomys was seen in COL12A1. The collagen triple helix repeat-containing protein 1 which is involved in collagen remodeling showed far higher levels in Acomys (Table 1).

Verification of fibrotic/non-fibrotic related proteins by western blot analysis.
Since they are of particular interest due to their involvement in immune defense and ECM synthesis, we verified the expression of COL1, COL3, COL12, S1008, MMP9, TIMP and 14-3-3 by western blot analysis (Fig. 8). Wound collagens at day 14 revealed significant changes compared to day 0. COL1, COL3 and COL12 were 7.6 times, 5.5 times and 58 times increased in Mus and 5.7 times, 7.8 times and 21 times increased in Acomys respectively, confirming the generally higher levels in Mus that the proteomic analysis recorded (Table 1). S100A8 and MMP9 at day 5 and TIMP at day 7 were detected only in Mus. 14-3-3 δ at day 7 was increased 2.8 times in Acomys and less than 1 in Mus.

protein -protein interaction analysis. To examine the interactions between DEPs detected in response
to wound healing in Acomys and Mus, we constructed network models using DEPs from the STRING database (http://string-db.org) (Fig. 9A,B). From this map, we searched the key proteins and their interactions involved in biological processes that may influence the wound healing outcomes. Overall, several hub proteins exhibiting physical and co-expression interactions with multiple proteins in diverse pathways were identified in the two species. For Acomys (Fig. 9A) the number of smaller interaction groups were identified and consisted of cell signaling and protein degradation and synthesis pathway components. These included protein tyrosine kinase activities such as MAP2K1 and FGFR4, GTPase activity (RAC1, RHOA), cell adhesion (CTNNB1, Calr), protease binding (UBA52, S27A), RNA binding (HAPA8, EFTUD2), cytoskeletal structure (ARPC3, ARPC1B), and ribosome structure (L27A, S13) (Fig. 9A). In Mus, the interacting groups included integrin binding which would be associated with the higher levels of collagens that we detected, catalytic activity (ACAA2, UQCRC1), oxidoreductase activity (SDHA, GSR), NADH dehydrogenase (NDUFS2 and NDUFS3), serine peptidase inhibitor (KLKB1, F2), growth factor binding (COL6A1, ITGAV), peptidase inhibitor activity (KNG1, SERPIND1), protease binding (SERPINC1, SERPINF2) and transferase activity (TKT, TALDO1) (Fig. 9B). Additional DEPs not exhibiting interactions with other proteins found in this analysis also likely play indirect roles in wound healing for Acomys and Mus.

Discussion
Acomys cahirinus retains a remarkable capacity for regenerative healing potential in a scar-free manner across multiple adult tissues. To reveal the mechanisms that support regeneration of Acomys on a molecular level we have undertaken a qualitative and quantitative proteomic analysis of Acomys compared to the scarring Mus using LC-MS/MS approaches. We identified proteins and their expression levels from skin tissue extracts with label free quantification through 0, 3, 5, 7 and 14 days as the wounds are undergoing re-epithelialization, the inflammatory phase and establishing granulation tissue. These stages were chosen as key time points for identifying protein differences potentially responsible for regeneration vs scarring and to correlate with our previous cellular and molecular studies 17,18 . The quantitative results of several proteins from skin wounds in this data were also validated by western blot analysis and demonstrate a good correlation of protein levels between our mass spectrometric based methodology and immunoblotting.
Re-epithelialization and histolysis are key early events required for the ability to regenerate complex tissues. Our proteomics studies revealed that the most enriched pathway and proteins among the DEPs were categorized   0-14d (D). The proteins highlighted refer to processes/pathways related to protein degradation/ synthesis and inflammatory response. viz. red dots for ribosome, blue dots for spliceosome, green triangle for protein processing in E.R, black diamond for ubiquitin/proteasome and black cross for complement and coagulation cascades. (2020) 10:166 | https://doi.org/10.1038/s41598-019-56823-y www.nature.com/scientificreports www.nature.com/scientificreports/ recent study demonstrates highly enriched ubiquitination/proteasomal degradation in liver regeneration, especially NEDD4 as an essential regulator by performing EPS15 ubiquitination which promotes EGFR internalization and efficient signaling in hepatocytes 21 . We were able to detect EGFR and EPS15 after wounding with a 2.7-fold increased NEDD4 at day7 in Acomys but EGFR and NEDD4 were either constant or significantly decreased in Mus, with little EPS15.
Although both Mus and Acomys can re-epithelialize their skin wounds the former does not normally replace the hairs and so the repaired skin remains hairless unless large wounds are made in some Mus strains, so called wound induced hair follicle neogenesis 13 . Acomys on the other hand regenerates all the hairs beginning as early as day 11 after wounding. The Wnt pathway plays a role in controlling epidermal stem cell renewal, in reciprocal interactions with the dermal papilla requirement and activation of the migrating wound epidermis [22][23][24][25] . Hair follicle induction specifically involves Wnt7a as ectopic expression of this gene in Mus induced hair follicle neogenesis 26 and Wnt7a expression is specifically up-regulated in Acomys wounds 17 . The only Wnt that we detected here was Wnt9, only in Acomys (Table 1) and its levels increased 2-fold on day 7. However, many members of the Wnt pathway were identified as up-regulated in Acomys following a KEGG analysis confirming the role of this pathway in hair follicle regeneration.
Another feature of Acomys re-epithelialization is a 2x faster rate of cell migration 11,27 . We identified many keratins in this proteomic analysis, the majority of which showed no differences between Acomys and Mus except for some cuticular keratins such as K31, K32, K35 and K36 which were elevated in Acomys. We have also observed similar expression levels of K17, the embryonic keratin involved in follicle neogenesis but a 2 times higher expression of 14-3-3δ as well as strongly increased vimentin levels in Acomys. Since an interaction of keratin 17 with 14-3-3δ and with signaling proteins for the regulation of cell growth in wounds 28 and a vimentin interaction with ribosomal protein S17 and 4-3-3δ for stimulating protein synthesis 29 have been demonstrated, it is perfectly feasible that increased levels of intermediates such as 14-3-3 δ in Acomys could have significant stimulatory effects on regeneration. Furthermore, keratins can profoundly influence cytoarchitecture by regulating signaling pathways and modulating protein synthesis and cell growth during the process of tissue repair 30 so these particular proteins may also be having some influence on the rate of re-epithelialization if not more profound effects on regeneration after injury in Acomys.
We know from previous molecular and cellular analyses that the immune and macrophage responses to injury are quite different between Acomys and Mus. In skin regeneration there is a huge immune cytokine response in Mus and M1 macrophages (F4/80) are present throughout the resolving granulation tissue after wounding whereas in Acomys the immune response is blunted and M1 macrophages are absent or deficient from an equivalent region 18,31 . The regenerating Mus ear cytokines are expressed at higher levels than in Acomys and www.nature.com/scientificreports www.nature.com/scientificreports/ similarly, the ear is nearly devoid of classically activated macrophages (CD86) but shows plenty of M2, CD206 macrophages 19 . The same is seen during skeletal muscle regeneration and kidney regeneration in Acomys 32,33 . Nevertheless macrophages are necessary for epimorphic regeneration 19 so the M2 phenotype must provide these pro-regenerative cytokines. We have found that to be the case here using proteomics. Thus, several immune modulators including S100A6, S100A8 and S100A9 were detected only in Mus. Several complement and coagulation factors were detected only in Mus and many serine proteases (serpins), proteins associated with inflammation and fibrosis, were present exclusively in Mus (Table S5). With regards to macrophage markers, arginase 1 was observed only in Mus whereas arginase 2 was solely identified in Acomys. Previous studies showed that arginases 1 and 2 as macrophage phenotypic markers are differentially related to inflammatory responses 34 with arginase 1 being widely expressed in either M1 or M2 polarized environments 35 and upregulated Arg1 by some M1 macrophages 36 . Yang showed that iNos-independent pro-inflammatory responses mediated by Arg2 in macrophages are due to enhanced mitochondrial ROS, but how Arg2 affects mitochondrial function leading to ROS production in macrophages remains unclear 34 . Notably, an acute inflammatory response was characterized by a higher myeloperoxidase activity in Mus and elevated ROS production in Acomys 19 , just as we have seen here and ROS production has been suggested as an essential early signal for regeneration based on studies in Xenopus www.nature.com/scientificreports www.nature.com/scientificreports/ and zebrafish tail models of regeneration 37,38 . Although we have not directly shown that high ROS production is generated specifically by Arg2, its presence identified here in Acomys and high ROS production in Acomys ear regeneration 19 may be no coincidence.
On the other hand, macrophage mannose receptor isomers, MRC1 (CD 206) and MRC2 (CD280) were present in both species. Several studies have demonstrated the critical role of the innate immune system in regulating regeneration and in the absence of macrophages Acomys ear regeneration and salamander heart regeneration is inhibited 19,39 . Presumably in these classes of macrophages are those that are crucial for non-fibrotic regenerative events across the vertebrates.
The profiles of extracellular matrix (ECM) proteins were different between Acomys and Mus. 11 members of the alpha collagen family were detected and, after wounding there is an increase in levels in all but one case with generally higher levels in Mus, the highest being for COLXII and COLXIV. Gawriluk et al. by transcriptomic studies of ear hole closure showed increased expression of collagen subunits in Mus relative to Acomys 14 . But in addition to matrix molecules there were several proteases and collagen remodeling proteins which were more highly or exclusively expressed in Acomys suggesting that despite the presence of relatively high levels of collagens a less fibrotic matrix could be generated which is certainly seen in histological analyses 18 . For example, CTHRC1 was more highly expressed in Acomys which can reduce collagen I mRNA and protein expression, inhibit TGF-β and promote cell migration 40 . The mast cells-related proteases cathepsin G, mast cell protease 4 and tryptase β2 were exclusively detected in Acomys and on the other hand the MMP inhibitors TIMP1 and cystatin were solely detected in Mus. Thus, a more fibrotic and rigid matrix is generated in Mus which may be why we saw integrin binding appearing as an interacting group in the network analysis of Mus proteins (Fig. 9B).
It is clear that many differences in protein expression between regenerative repair in Acomys and fibrotic repair in Mus can be identified. It will be important to design studies to determine if there are particular genetic or metabolic components that can trigger regeneration repair in favor the fibrotic/inflammatory pathways that are normally seen after wounding in mammals.

Animals.
All experiments were performed following guidelines of the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Florida (# 201203505 (Mus) and 201207707 (Acomys)) and animals were housed under the care of the University of Florida's Animal Care Services. A. cahirinus were obtained from a breeding colony house at University of Florida and M. musculus of the CD-1 outbred strain was purchased from Charles River (Wilmington, MA). Animals were 6 months of age at time of experiments. Animals were anaesthetized with iso-fluorane, the hair on the dorsum was shaved and two 8 mm biopsy punch wounds made through the mid-dorsal full thickness skin. At various times after wounding (3, 5, 7, 14 days) the animals were sacrificed, and the wound tissue dissected out excluding the surrounding normal skin.
protein sample preparation. The tissue (100 mg) was homogenized on ice for 30 seconds using a rotor stator type tissue homogenizer (ProScientific Bio-Gen PRO200 Homogenizer; Multi-Gen 7XL Generator Probes) in a protein extraction buffer (1 mL) containing Tris-Cl (50 mM, pH 7.4), NaCl (100 mM), ethylenediaminetetraacetic acid (EDTA, 1 mM), protease inhibitors (1:25 cOmplete UTLRA, Roche), phosphatase inhibitors (1:10 PhosSTOP, Roche) and kept at 4 °C for 1 h. Soluble proteins were separated from undissolved tissue by centrifugation at 80000 g for 30 min at 4 °C. Protein levels were measured using a BCA kit (Pierce) and protein amounts normalized (2 mg/mL) by dilution in protein extraction buffer. For each experiment and each experimental time point, a group size of 3 animals was used.
protein digestion and peptide fractions. Protein samples (100 µg) were loaded on a gel (Novex, 8%, Bis-tris), All steps were carried out as described previously 41 . Briefly, electrophoretic migration was performed to fractionate the protein into 10 gel bands. Incised gel bands were reduced by adding 500 µl of TCEP (10 mM) in NH 4 HCO 3 (100 mM) at 37 °C for 30 min, and then treated with 500 µl of IAA (55 mM) in NH 4 HCO 3 (100 mM) at RT for 1 h in the dark. After removing the excess regents in-gel tryptic digestion was performed with trypsin (total protein: trypsin (50:1, in NH 4 HCO 3 (50 mM) at 37 °C overnight). The following day, trypsinization was quenched by formic acid (10 µL). Digested samples were dried using a Speed Vac and were stored at -80° until used.
Lc-MS/MS analysis. The tryptic digests were analyzed using an LTQ Velos Orbitrap mass spectrometer www.nature.com/scientificreports www.nature.com/scientificreports/ linear trap quadrupole (LTQ) with 10 ms activation time and 35% normalized collision energy (NCE) in CID; and a 2.0 Da isolation window. Previously fragmented ions were excluded for 60 s for all MS/MS scans. The MS1 mass scan range was 400−2000 m/z. The electrospray voltage was 2.2 kV and the capillary temperature was set at 250 °C.
Database search and data validation. MS/MS spectra were extracted by the MM File Conversion Tool (Version3.9, http://www.massmatrix.net/mm-cgi/downloads.py) and sent to a database search using SEQUEST 42 . They were searched against the integrated proteomics pipeline (IP2): SEQUEST with modified parameters (precursor ion tolerance = 50 ppm, fragment ion tolerance 0 0.8 Da, missed cleavage ag2, modification , carbamidomethyl cysteine (fixed), methionine oxidation (variable), and enzyme (trypsin)). For peptide validation, a 1% false discovery rate (FDR) at the peptide spectral match (PSM) and/ or peptide level was used. We used the ProteinInferencer 43 (Scripps Research Institute, La Jolla, CA; http://proteomicwiki.com/wiki/index.php/ ProteinInferencer) for an integration of all data generated from the three search engines using an FDR <1.0% at protein level. The detailed calculation method and the search conditions using this program have been reported 41 . Data were searched against a target-decoy Swiss-Prot database, version 2016_08 from mouse (http://www.uniprot.org). We applied a label-free quantitation of the identified peptides to a protein with manual validation. Protein quantification based on the extracted ion chromatogram (XIC) was obtained by extracting the intensity corresponding to the m/z of the selected peptides along the LC-MS run and by integrating the peak area at their respective retention time (RT). Normalized protein quantitative values were calculated as log2 (protein area/total protein area) x 10 6 . Reverse decoy matches were removed from the protein identification list. At least 2 unique peptides per protein were required for protein identification. Only proteins that were identified and quantifiable in at least two technical of at least three biological replicates in each group were used for relative quantification. The criteria for identifying differentially expressed proteins (DEPs) were at least a 2 fold change in levels in either direction with a p-value for significance of ≤ 0.01 and coefficients of variation (CV) <20%. Experiments were repeated at least three times and the results analyzed using the unpaired t-test assuming equal variance on the normalized, scaled dataset. and adjusted for the false discovery rate (≤ 0.01) using the Benjamini-Hochberg method (BH) 44 .