Quantitative proteomics approach reveals novel biomarkers and pathological mechanism of keloid

Abstract Background Keloid is a pathological skin scar formation with complex and unclear molecular pathology mechanism. Novel biomarkers and associated mechanisms are needed to improve current therapies. Objectives To identify novel biomarkers and underlying pathological mechanisms of keloids. Methods Six pairs of keloid scar tissues and corresponding normal skin tissues were quantitatively analyzed by a high‐resolution label‐free mass spectrometry‐based proteomics approach. Differential protein expression data was further analyzed by a comprehensive bioinformatics approach to identify novel biomarkers and mechanistic pathways for keloid formation. Candidate biomarkers were validated experimentally. Results In total, 1359 proteins were identified by proteomic analysis. Of these, 206 proteins exhibited a significant difference in expression between keloid scar and normal skin tissues. RCN3 and CALU were significantly upregulated in keloids. RCN1 and PDGFRL were uniquely expressed in keloids. Pathway analysis suggested that the XBP1‐mediated unfolded protein response (UPR) pathway was involved in keloid formation. Moreover, a PDGFRL centric gene coexpression network was constructed to illustrate its function in skin. Conclusions and Clinical Relevance Our study proposed four novel biomarkers and highlighted the role of XBP1‐mediated UPR pathway in the pathology of keloids. It provided novel biological insights that contribute to develop novel therapeutic strategies for keloids.


INTRODUCTION
Keloid scars are pathological fibro-proliferative disorders of the skin that exhibit abnormal phenotypes including fibroblast proliferation and collagen deposition [1]. There are several treatments for keloids including conventional surgical therapies and adjuvant therapies; however, a high recurrence rate of keloids is routinely observed after treatment [2]. Therefore, an improved understanding of the pathogenesis leading to keloid formation is needed to be able to develop novel targeted therapies. Cytokines and various growth factors are reported to be involved in keloid pathogenesis [1,3]. For example, platelet derived growth factor (PDGF) regulates the production of collagen and fibronectin (FN), thus contributing to wound healing [4][5][6]. The abnormal activation of PDGF receptors and insulin-like growth factor-1 (IGF-1) receptors induced by transforming growth factor-β (TGF-β) is known to be associated with the regulation of numerous processes in the skin regeneration pathway [7][8][9]. However, the overall pathogenic process underlying the development of keloids is still unclear and novel markers for treatment are urgently required.
Quantitative proteomics approaches have been proved to be an efficient approach for the investigation of pathological mechanisms and novel biomarkers. To date, a few proteomics studies have been conducted on keloid scars and some biomarkers of keloid scars have been suggested [10][11][12]. However, a more comprehensive and in-depth analysis of keloid scars is still needed to reveal further details relating to the pathological mechanisms associated with keloids. The development of high-resolution mass spectrometry and its combination with a stateof-the-art bioinformatics approach provide us with an exciting option for investigative research. We have previously proposed several novel biomarkers and pathways for a variety of tumors through our proteomics platform [13][14][15][16]. In this study, we present a label-free quantitative proteomics analysis to explore differential protein expression profiles in normal skin and keloid scar tissues based on nano-liquid chromatography and tandem mass spectrometry (nano-LC-MS/MS).
The expression profiling was further bioinformatically analyzed to identify the key regulators and pathways involved in keloid formation.
Our findings provide a more comprehensive expression landscape of keloid proteins and yield novel pathological insights into the formation of keloid scars.

Protein extraction
The specimen was homogenized in lysis buffer (20 mM HEPES, 9 M Urea, EDTA-free Protease Inhibitor Cocktail) and incubated for 30 min on ice. Then, the sample was sonicated with an ultrasonication probe (10 × 1 s) and centrifuged at 13,400 rpm for 10 min at 4 • C. Supernatant was collected and the protein concentration of the protein extract was determined by Bradford assay (TaKaRa, Dalian, China) according to the manufacturer's instructions.

Protein digestion and peptide purification
In-solution digestion was performed before MS analysis. Twenty μg proteins were diluted to a final volume of 20 μL with digestion buffer Finally, the sample was purified by using C18 spin columns (Thermo Fisher Scientific, Rockford, USA) according to the manufacturer's instructions and dried with a vacuum centrifuge (Thermo Fisher Scientific, Asheville, USA).

Liquid chromatography (LC)-mass spectrometry (MS)/MS analysis
The peptides were separately analyzed by using Q-Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with an EASY-nLC 1200 (Thermo Fisher Scientific, Bremen, Germany). Data dependent acquisition (DDA) mode was applied. Precolumn (2 cm, 100 μm inner diameter, 5 μm C18 filler; Thermo Fisher Scientific, Bellefonte, USA) and analytical columns (10 cm, 75 μm inner diameter, 3 μm C18 filler; Thermo Fisher Scientific, Bellefonte, USA) were used for analysis. Peptides were eluted with a 90 min HPLC gradient from 0% to 100% in buffer (80% acetonitrile, 1% formic acid) at a flowrate of 250 nL/min. The scan range was set to 400-1700 m/z, and 70,000 resolution (at m/z 200) was used. The ten most-abundant MS1 features were selected for high-energy and MS/MS scans. Raw data were processed with Xcalibur software (Thermo Fisher Scientific, Bremen, Germany). The mass spectrometry proteomics data were deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD029631 [17].

2.5
Data analysis and quantification of proteomic raw files MS raw files were analyzed by MaxQuant (version 1.6.12.0) with the UniProt reference protein database (Homo sapiens, November, 2019).
The label-free quantification (LFQ) algorithm was used for protein quantification. The option "match between runs" was applied to increase the number of peptide searches. Other options were set as default settings. Data arising from MaxQuant analysis was further processed with Microsoft excel.

RNA isolation and quantitative real-time PCR (qRT-PCR)
Total RNAs were extracted with the Trizol Reagent. Reverse transcription was performed with HiScript III 1st Strand cDNA Synthesis Kit (Vazyme, Nanjing, China). qRT-PCR was carried out with the SYBR Green PCR Master Mix (Takara, Dalian, China). All processes were carried out in accordance with the manufacturer's protocol.

Western blot (WB)
Tissue sample was homogenized in RIPA lysis buffer with cocktail and centrifuged at 12,000 g for 10 min at 4 • C. Supernatant was collected and the protein concentration was determined by Bradford assay.
Twenty μg of proteins were used for WB analysis. Anti-SPARC, anti-RCN1, and anti-CALU antibodies were obtained from Abcam. Anti-PDGFRL and anti-XBP1 antibodies were purchased from Cusabio and Sino Bioloqical separately.
Principal components were calculated by the default method in the R package using data that contained missing values. Heatmap analysis with a clustering tree was obtained via the heatmap tool in the R package.

Gene set enrichment analysis
The online Gene SeT AnaLysis Toolkit (WebGestalt) was used for gene set enrichment analysis by using the human genome reference gene set as a background which is freely available at http://www.webgestalt.

org. Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes
(KEGG) analysis were performed by using the companion tool GOView.
GO analysis was used to cluster the differently expressed proteins (p < 0.05) into three key biological aspects, including biological process, cellular component, and molecular function. The over-representation of both GO and KEGG pathways was determined by the hypergeometric test. The p-value was adjusted and calculated by Benjamini and Hochberg methods. Ingenuity pathway analysis (IPA) was performed to identify pathways involved in genes related to keloid formation.

Protein-protein interaction (PPI) network construction
The interaction between differently expressed proteins was illustrated by using the online tool for the Retrieval of Interacting Genes (STRING) online database (http://string-db.org). The differential gene lists were submitted and searched by selecting "Organism" and "Homo sapiens" as key parameters. PPI networks were constructed based on the STRING database.

Pathway analysis
To address the differentially expressed proteins enriched pathways, we performed IPA analysis. The top 10 related pathways are presented in Figure 4A. Pathway related to the unfolded protein response (UPR) function was significantly activated in keloids (Z score > 1.5, p = 1.20 × 10 −5 ). A detailed gene list related to this pathway is given in Supporting Information Table S4. Further details of the UPR pathway are shown in Figure 4B, in which PDI, BIP, and CALR were upregulated and induced endoplasmic reticulum stress. Although the overexpression of XBP1, the regulator of the UPR pathway, was not detected in our proteomic analysis, our pathway analysis showed that XBP1 might be a potential therapeutic target for keloids because most proteins that were overexpressed in keloids can be regulated by XBP1 ( Figure 4C).
The upregulated expression level of XBP1 in keloids was further verified by WB ( Figure 4D). NetworkAnalyst software (www.networkanalyst.ca) so that we could analyze the PPI network with regards to curated and nonredundant sets of protein interactions in the IMEx consortium database [19]. As shown in Figure

DISCUSSION
In this study, we presented the most comprehensive proteomics study  [20]. In both studies, the accumulation of collagens showed molecule-specific (e.g., upregulated COL12A1 vs. downregulated COL6A3 in keloids). This means more special proteases for the collagen degradation are needed for the treatment of keloids. The fact that some proteoglycans related to skin mechanical forces (e.g., ASPN, VCAN, and OMD) were also overexpressed in keloids proved the interaction between skin mechanical disorder and keloid formation again.
Bioinformatics analysis indicated that pathways related to ER stress are enriched in the keloids. ER homeostasis is constantly challenged by a range of physiological activities, including Ca 2+ reservoir regulation and the biosynthesis of lipids and sterols during physiological and pathophysiological perturbations. There are three major UPR response proteins: PERK, eIF2α, and IRE1α; these all bind to BiP, also known as HSPA5, in an inactive form under non-ER stress conditions. In our research, we detected upregulated BiP and PDI, which help to correct the folding of proteins, including P4HB, PDIA3, and PDIA6 in keloids ( Figure 4B). It seems that XBP1 plays an important role in keloid formation because most of the upregulated proteins were shown to be related with XBP1 ( Figure 4C). For example, SPARC, a Ca 2+ -binding glycoprotein, which exhibits increased level of expression in hypertrophic scars, has been reported to be upregulated by XBP-1 in hepatocellular carcinoma cells [21,22]. P3H1 and PPIB, which are essential for prolyl 3-hydroxylation and folding of procollagens in the ER, have been shown to be upregulated by XBP1 in NIH-3T3 fibroblasts [23]. PDI, chaperones, and COPI vesicles, which help protein folding and transport, have also been considered to be upregulated by XBP1, including P4HB, PDIA3, PDIA6, CALR, HSPA5, HSP90B1, FKBP7, FKBP10, COPB1, and ARCN1 [23]. Excisional wound healing is also related to increased level of the active form of XBP1 when compared with normal fibroblasts. Moreover, XBP1 along with the UPR signal pathway were activated in keloid fibroblasts when exposed to a hypoxic environment [24]. In another study, the inhibition of IRE1α also decreased scar formation and decreased XBP1 expression [25]. In a recent study, higher ER stress signaling was demonstrated in keloids than in normal tissues; this was consistent with our current results, and the inhibition of ER stress significantly decreased scar formation in a rabbit model [26].
These results suggested that XBP1, as a key regulator of UPR pathway, is closely related with keloid formation and may represent a potential therapeutic target for keloids.
Abnormal calcium homeostasis is associated with ER stress and has been reported to exist in keloid fibroblasts [27]. In our study, we found CALU, RCN3, and RCN1, the members of CREC protein family which carries out a number of functional activities, including calcium homeostasis and secretory cargo sorting, were significantly upregulated in keloids [28]. Of these, CALU and RCN3 showed abnormally high expression levels with lower p values in keloids ( Figure 1A); while RCN1 was uniquely present in keloids (Supporting Information Table   S2a). RCN1 is known to be able to suppress ER stress-induced apoptosis and is related with tumorigenesis [29][30][31][32]. CALU localizes to the entire secretory pathway, including the ER, Golgi apparatus, and the extracellular matrix. It appears that extracellular CALU inhibits cell migration, whereas nuclear isoform calumenin-15 promotes filopodia formation and cell migration, which suggests that this protein exerts different functions when localized in different sites [33,34]. RCN3 has been reported to be associated with the maturation of alveolar epithelial type II (AECII) cells during alveogenesis [35]. The expression of RCN3 in AECIIs appears to contribute to cell survival and wound healing [36]. The overexpression of RCN3 in keloids was also highlighted in a study of familial keloid [37]. These results indicated that CALU, RCN3, and RCN1 may be associated with keloid formation and could be novel potential biomarkers of keloids. and DBN1. In these proteins, TFAP2C is a transcription factor that is known to help organize ECM fibers [39]. LMO2 is known to regulate hematopoiesis and vascular development and has been reported to play an import role in tissue regeneration [40]. GBR2 is involved in the Ras signaling pathway and is linked to mitogenesis and cytoskeletal reorganization by EGF and PDGF [41]. DBN1, an actin-binding protein, is involved in many cell-cell communication systems, including gap junctions and adherens junctions [42]. CLU, which performs a number of different functions and has also been associated with neurodegenerative diseases and cancer, also acts as a chaperone outside the cell and facilitates the extracellular clearance of misfolded proteins [43].
APP is mainly associated with Alzheimer's disease and participates in the control of epidermal wound repair [44]. As important components of the ECM, FBLN1, COL1A1, and FN1 are involved in ECM assembly, cell migration, and wound healing. They are also related to the proliferation and differentiation of osteoblasts [45][46][47][48]. Considering that PDGFRL regulates the proliferation of chondrocytes, which share the same origin as osteoblasts, it follows that there may be potential interactions between FBLN1, COL1A1, FN1, and PDGFRL. The fact that we observed the upregulation of FBLN1, FN1, and PDGFRL in keloids in our proteomic profiles enhances the possibility of this hypothesis. We also identified PDGFRB in our network. However, the overexpression of PDGFRB was not detected in our study. PDGFRL may show a redundant function or a differential function with PDGFRB; thus, PDGFRL may also be a potential therapeutic target of keloids.

CONCLUDING REMARKS
In summary, we provided a comprehensive proteome profiling of keloid scars and normal skin tissues. We identified 206 proteins that showed significant differences in expression between keloid scars and normal skin tissues, including RCN3, RCN1, CALU, and PDGFRL which may play an import role in keloid formation. We proposed that the XBP1mediated UPR pathway plays a role in keloid formation. We also created a PDGFRL coexpression gene network to identify its potential functions. In summary, our study provides novel information relating to the pathological process of keloid formation and could contribute to the development of new therapeutic strategies for keloid scars.

ACKNOWLEDGMENTS
The project was supported by Taishan