Contribution to the peripheral vasculopathy and endothelial cell dysfunction by CXCL4 in Systemic Sclerosis

Objective CXCL4, a chemokine with antiangiogenic property, is reported to be involved in systemic sclerosis (SSc) related pulmonary arterial hypertension (PAH). We investigated the contribution of CXCL4 to SSc development by focusing on the correlation of circulatory CXCL4 levels with their peripheral vasculopathy, as well as the effect of CXCL4 on endothelial cell dysfunction and angiogenesis disturbance in SSc and the potential signaling. the levels of CXCL4 and the peripheral vasculopathy in SSc evaluated clinically or by nailfold videocapillaroscopy (NVC) was also studied. Furthermore, we explored the potential signaling mechanism mediated by CXCL4, which could be functionally contribute to the defective angiogenic process characteristic of this disease spectrum. Ethics Committee of Serum was collected, centrifuged, aliquoted, and stored at − 80 °C. The serum samples had undergone only 1 freeze/thaw cycles before protein measurements. All samples were measured in duplicates, in accordance with the approved guidelines and regulations.


Introduction
Systemic sclerosis (SSc) is an autoimmune disease with skin and multiple viscera involvement, which characterized by a complex interplay of vasculopathy, immune system activation, and persistent tissue brosis [1]. Progressive and marked reduction in capillaries is a hallmark nding in early stage SSc with avascularity increasing with disease progression. Patients with SSc may develop a spectrum of vascular disease, mainly including Reynaud's phenomenon (RP), digital ulcer (DU), pulmonary arterial hypertension (PAH), and scleroderma renal crisis (SRC). In SSc patients, the universal prevalence of RP demonstrated that increased vascular dysfunction exists in the process of SSc [2]. As the rst onset symptom, RP leads to continuous digital ischemia, which may develop into digital ulcers (DU) or severe digital ischemia with gangrene in some extreme cases [3]. These early events induce increasing vascular tone, decreasing capillary blood ow, and chronic tissue hypoxia, ultimately leading to the extracellular matrix (ECM) [4] accumulation and tissue brosis. Therefore, it is necessary to identify biomarkers that could predict the very early SSc, before inevitable organ injury and brosis occurred.
The alteration of both pro-angiogenic mediators and inhibitors of angiogenesis have been recently implicated in vascular dysfunction in SSc [5]. In addition, impairment of angiogenic signal transduction pathways in endothelial cells contributed to disturbing angiogenesis in SSc [6]. Previous studies have focused on CXCL4 and its anti-angiogenic effects in the pathogenesis of atherosclerosis [7], and cancer [8]. Proteome-wide analysis revealed CXCL4 as a biomarker in SSc, and subsequently closely correlated with interstitial lung disease (ILD) and PAH [9]. CXCL4, unlike other chemokines that bind to their speci c receptors, exerts its effects through its high a nity for proteoglycans and other negatively charged molecules. It was also reported that CXCL4 can directly bind to the CXCR3B chemokine receptor isoform or lipoprotein-related protein-1(LRP1) [11]. In patients with SSc, CXCL4 binds DNA into speci c immune complexes that magnify plasmacytoid dendritic cell (pDC) -hyperactivation and produce in ammatory cytokines, which independent of CXCR3 [12]. These data suggest the involvement of CXCL4 in SSc pathology.
Therefore, the aims of this study were to investigate whether the levels of CXCL4 could be altered in the very early diagnosis of SSc (VEDOSS), which manifesting typical vasculopathy but not brosis yet. Then, the association between the levels of CXCL4 and the peripheral vasculopathy in SSc evaluated clinically or by nailfold videocapillaroscopy (NVC) was also studied. Furthermore, we explored the potential signaling mechanism mediated by CXCL4, which could be functionally contribute to the defective angiogenic process characteristic of this disease spectrum.

Method
Patients, controls and serum samples.
In the discovery cohort, serum samples from 58 patients with systemic sclerosis (SSc) [13]

Statistical analysis
GraphPad Prism 5 was used to conduct statistical analysis. Normally distributed data are presented as mean ± SD, unless otherwise indicated. Inter-group differences were assessed for signi cance using an independent two-group t-test or one-way ANOVA with the use of a Bonferroni correction. The chi-square test was used to compare clinical characteristics and CXCL4 levels between different clinical groups, as appropriate. To assess the most effective cutoff value for CXCL4, we used DeLong's method to compute a receiver-operating-characteristic (ROC) curve. The area (± SE) under the ROC curve was 0.8051 ± 0.02879 (95% con dence interval [CI], 0.7487 to 0.8616; P < 0.0001). The cutoff value for CXCL4 of 2797 ng per milliliter was determined using the Youden Index (Youden index = sensitivity + speci city -1), the sensitivity was 60%, and the speci city was 95.7% at this cutoff value. Two-sided signi cance thresholds were used: *p < 0.05, **p < 0.01.
Other methods and any associated references are available in the online supplement.

Result
Circulating levels of the CXCL4 are raised in the patients with SSc and the patients with a very early diagnosis of SSc (VEDOSS) We rst examined serum CXCL4 in SSc and VEDOSS patients in the identi cation cohort, and evaluated the possible correlation with speci c clinical features of the disease. A total of 58 patients with SSc, 10 patients not ful lling the 2013 ACR/EULAR classi cation criteria for SSc (total score < 9) and identi ed as VEDOSS (presented with Raynaud's phenomenon, puffy gures, and positivity of antinuclear antibodies, together with SSc speci c antibodies and / or pathognomonic microvascular alteration at capillaroscopy), and 80 healthy controls (HC) were recruited and strati ed as described in Methods. In this identi cation cohort, serum CXCL4 levels were 60.59% higher in patients with SSc and 132.95 % higher in patients with VEDOSS than matched HC (HC: 1751 ± 917.1; SSc: 2812 ± 1445; VEDOSS: 4079 ± 1978 ng / ml; p < 0.0001 compared with HC respectively; Fig. 1A). Notably, signi cantly higher serum CXCL4 levels were detected in the patients with VEDOSS than the patients with SSc (p = 0.0089, Fig. 1A). These results were subsequently validated in the replication cohort, where circulating CXCL4 resulted signi cantly increased in SSc (n=50) and in VEDOSS (n=12) serum compared with control serum (n=80) (HC: 1288 ± 839.2; SSc: 3423 ± 1617; VEDOSS: 5003 ± 1814 ng / ml; p < 0.0001 compared with HC respectively; Fig. 1B). Similarly, further elevation of CXCL4 was detected in the patients with VEDOSS compared to the patients with SSc (p = 0.0004, Fig. 1B).
We performed the pooled analysis by combining patients and controls from the identi cation and replication cohorts. Circulating CXCL4 levels were 103.62 % higher in patients with SSc and 201.51 % higher in the patients with VEDOSS than in healthy controls, also signi cantly higher in SSc patients relative to VEDOSS patients (HC: 1520 ± 906.4; SSc: 3095 ± 1551; VEDOSS: 4583 ± 1903 ng / ml; p < 0.0001 compared with each other; Fig. 1C).
Increased serum CXCL4 levels positively correlate to the much severer skin brosis and peripheral vasculopathy in SSc patients We next assessed the association between CXCL4 levels and the clinical phenotype in the combined cohort, and the patients with SSc were strati ed as described in Methods. The levels of CXCL4 in SSc were positively correlated with their mRSS score ( Fig. 2A), demonstrating that the increased CXCL4 level may implicate much severer skin brosis. Notably, signi cantly higher serum CXCL4 levels were detected in the SSc patients with digital ulcer (DU) (n=41) than those without DU (n=67) (3686±1769 vs 2734±1285 ng/ml, p = 0.0016; Fig. 2B), and the levels of CXCL4 in SSc were positively correlated with the number of digital ulcers (Fig. 2C), indicating the close association of CXCL4 with peripheral microvascular involvement severity in SSc.
We also investigated whether CXCL4 could serve as a biomarker, which performed on the combined group using ROC analysis to select a threshold (Fig. 2D). Patients who had a higher baseline level of CXCL4 (over 2797 ng/ml) had a signi cantly increased prevalence of newly-onset of digital ulcer in 6 months (34.78% vs 65.22%, odds ratio 6.133, P = 0.0335; Fig. 2E), with unbiased background treatments (data not shown).
Next, we strati ed SSc patients according to their nailfold videocapillaroscopy (NVC) pattern and compared the CXCL4 levels between subsets, we also evaluated the possible correlation of CXCL4 levels with the mean number of nailfold capillary in the patients with SSc (n = 58). SSc patients with early NVC pattern demonstrated elevated serum CXCL4 levels rather than those with active or late NVC pattern (p=0.0159 and p=0.0063 for each comparison) (Fig. 2F). The levels of CXCL4 were negatively correlated with the mean number of nailfold capillary in patients with SSc (Fig. 2G).

SSc derived CXCL4 disturbed angiogenesis
It is widely reported that the stimulation with SSc sera disturbed angiogenic performance of HUVECs [16][17][18], and, as we showed above, circulatory CXCL4 levels were raised in the patients with SSc and correlated with their peripheral vasculopathy. Thereby, we treated HUVECs using recombinant human CXCL4 or SSc sera, with or without CXCL4 neutralized antibody, in order to testify the contribution of SSc sera derived CXCL4 to the angiogenesis of endothelial cell line HUVECs, including viability, migration and tube formation.
Firstly, the addition of CXCL4 or the SSc sera to the medium of HUVECs inhibited endothelial cell proliferation determined using CCK-8 assay, which could be signi cantly ameliorated by antibody-mediated neutralization (***P < 0.001, Fig. 3A and 3B).
Furthermore, the stimulation with CXCL4 signi cantly decreased the ability of the tube formation and migration of HUVECs, which were dramatically improved by treating with anti-CXCL4 antibody (*P < 0.05, ***P < 0.001, Fig. 3C and 3D). Similarly, both of tube formation and migration was signi cantly impaired after challenging with 10% SSc sera, and this inhibitory effects on HUVECs were signi cantly reversed by pretreatment with an anti-CXCL4 antibody ( Fig. 3E and 3F). Therefore, these data revealed the antiangiogenic effects of SSc derived CXCL4.
As CXCR3 mediated Ca 2+ mobilization and chemotaxis in response to C-X-C chemokines, including CXCL4, we also con rmed the expression of CXCR3 in HUVECs (Supplemental Fig. 1A), we thereby wondered if CXCL4 inhibited endothelial cell proliferation via its receptor CXCR3. However, the addition of AMG487, a speci c antagonist of CXCR3, did not show to reverse the inhibition of cell viability, tube formation and migration induced by CXCL4 or SSc sera (P > 0.05 against the CXCL4 group or SSc sera group; Supplemental Fig. 1B-G). Therefore, the data showed that CXCL4 exerted its anti-angiogenic effect on HUVECs not through CXCR3.
Since the activation of c-Abl pathway is a negative regulating Fli-1 de ciency [22,23], we initially looked at the effect of CXCL4 on the c-Abl signaling. As shown in Fig.5A, CXCL4 increased the expression of c-Abl in a dose-dependent manner, and CXCL4 neutralizing antibody inhibited the overexpression of c-Abl induced by CXCL4 (P < 0.001 for each comparison). Next, we showed the time course of c-Abl induction by CXCL4 treatment in HUVECs (P < 0.001 for each comparison; Fig. 5B). As ponatinib [24] could inhibit c-Abl pathway (Fig. 5C), we also investigated that ponatinib normalized the reduced levels of Fli-1(*P < 0.05, **P < 0.01, ***P < 0.001; Fig. 5D and 5E). Collectively, these data suggest that CXCL4 regulates Fli-1 via c-Abl pathway in SSc vasculopathy.

CXCL4 blocked the pro-angiogenic effect of TGF-β and PDGF in HUVECs
Numerous pro-angiogenic mediators like transforming growth factor (TGF)-β and platelet-derived growth factor (PDGF) activated in SSc [25,26]. Since CXCL4 could bind to growth factors directly to exert its effect, we examine whether TGF-β and PDGF are involved in the progressive vasculopathy of CXCL4 induced. As shown in Fig.6 A and 6B, TGF-β and PDGF signi cantly induced the proliferation of HUVECs as previously reported [27][28][29][30] and the inhibitors against TGF-β or PDGF blocked their pro-angiogenic effect respectively (P < 0.0001 for each comparison). Interestingly, the addition of CXCL4 reduced the cell proliferation of HUVECs induced by TGF-β and PDGF, which were completely reversed by the pretreatment of the CXCL4 neutralizing antibody. These data suggested that CXCL4 exerted its antiangiogenic effect by antagonizing TGF-β and PDGF signaling.

Discussion
In this article, we provide the evidence that CXCL4 plays an important role in peripheral vasculopathy of SSc. First, we showed here that serum CXCL4 levels are signi cantly increased in patients with SSc and VEDOSS patients. Higher circulating CXCL4 levels associated with the microvascular abnormalities in SSc. Second, CXCL4 was found to exerts its anti-angiogenesis function in SSc, including inhibiting endothelial cells proliferation, migration, and tube formation with c-Abl/Fli-1 pathway involvement. Third, TGF-β and PDGF signaling may participate in the anti-angiogenesis effects of CXCL4.
CXCL4, as one of the recognized anti-angiogenesis chemokines, affects anti-angiogenesis via various ways. CXCL4/PF-4 inhibits endothelial cell proliferation [31] and migration [32], and thus displays antitumoral activity by inhibiting tumor growth [33] and suppressing the formation of metastasis [34]. In SSc patients, the serum level of CXCL4 is positively correlated with PAH [35], suggesting that CXCL4 may be involved in SSc vasculopathy. Similar to previous studies, we found that exogenous CXCL4 can effectively inhibit angiogenesis on HUVECs, including cell viability, migration, and tube formation. Notably, we demonstrated that the ability of serum from SSc patients to inhibit angiogenesis was largely dependent on CXCL4 in HUVECs, as the addition of anti-CXCL4 almost completely reverted antiangiogenesis following serum stimulation.
As the mechanism of CXCL4 in SSc vascular lesions is not yet clear, we selected two related factors of vascular mediators, i.e. ET-1 and Fli-1, and further explored the usability of anti-CXCL4 as a possible therapeutic target in SSc vasculopathy. In vitro, we found that CXCL4 and SSc patient serum could increase the expression of ET-1 and decrease Fli-1, and anti-CXCL4 could signi cantly reverse these effects. In this respect, the downregulation of Fli-1 is concentrated, since the elective endothelial cell deletion of Fli-1 in mice models leads to a disorganized dermal vascular network with greatly compromised vessel integrity and markedly increased vessel permeability [36]. Fli-1 de ciency is also related with the vasculature dysfunction in SSc [37]. Importantly, in HDMECs gene silencing of Fli-1 modulates the expression of various genes, including VE-Cadherin, PECAM1, MMP9, cathepsin B, and cathepsin V, towards an SSc EC phenotype [38][39][40]. These ndings have explained the existence of high levels of pro-angiogenic factors in SSc patients, as well as the presence of anti-angiogenic factors like CXCL4. The imbalance of pro-and antiangiogenic factors might explain the pathogenetic mechanisms of SSc vasculopathy.
Since c-Abl pathway negatively regulates the expression of Fli-1 [22], we then put forward to explore the mechanism of Fli-1 de ciency via c-Abl after CXCL4 treatment. We found that the c-Abl pathway was activated under CXCL4 treatment. Furthermore, inhibited c-Abl kinase activity could reverse the decreased expression of Fli-1 in HUVECs. These ndings are also supported by a recent study, which shows that blocking c-Abl kinase activity or decreasing the expression of c-Abl upregulates Fli-1 on SSc broblasts [41]. In fact, c-Abl kinase inhibition like imatinib mesylate is enrolled in dcSSc clinical trials. 1 year of treatment with imatinib signi cantly improved forced vital capacity (FVC) and mRSS in patients with dcSSc [42]. However, imatinib is poorly tolerated and thus limits its application in SSc [43]. In this respect, it is critical to nd other small-molecules as targeted drugs for SSc. Thus, our study provides direct evidence that CXCL4 participates in SSc pathogenesis, and suggests targeting this cytokine as a potential therapeutic strategy for SSc.
A previous study demonstrated that CXCL4 could exert its effect through multiple ways, including binding to its speci c receptors or other growth factors [44]. Our data suggest that although CXCR3 is expressed on HUVECs, blocking CXCR3 did not affect the anti-angiogenic effects of CXCL4. Furthermore, a quantity of evidence suggested that growth factors TGF-β and PDGF play important roles in vasculopathy of SSc [25,26]. Thus, we hypothesis that CXCL4 may directly bind to these growth factors and then exerts its effects in SSc vasculopathy. As shown in Fig. 6, TGF-β and PDGF can effectively promote endothelial cell proliferation as previously reported [25,26]. However, the angiogenic effects of TGF-β and PDGF were signi cantly reversed after CXCL4 pretreatment. Taken together, our data suggested that CXCL4 may directly binding to TGF-β and PDGF in SSc.
CXCL4 not only exerts its anti-angiogenesis ability in SSc, but also activates immune cells, including dendritic cells [45], T cell [46], neutrophil [47], mast cell [48], and macrophage [49] according to many studies, thus widely participate in the pathogenesis of various diseases. For example, CXCL4 induces monocyte differentiation resulting in a macrophage phenotype called 'M4', which especially expresses MMP7 and S100A8. 'M4' is distinct from M1 and M2 macrophages, presenting proin ammatory and cytotoxicity, reduce LDL-uptake in atherosclerosis [7,50,51]. In SSc, CXCL4 can drive broblast activation indirectly via PDGF-BB production by macrophage. However, in vascular endothelial cells, it is still unclear whether CXCL4 can regulate angiogenesis by inducing macrophages. In this study, although M4 macrophages have been induced at 7 µg/ml of CXCL4 as previously reported [51], neither the 'M4' macrophageendothelial cell co-culture system nor the endothelial cells cultured with 'M4' macrophage CM have an effect on the related vascular mediators of SSc. Concerning the fact that we still need further experiments to con rm that 'M4' macrophages do exist in SSc patients, we cautiously conclude that CXCL4's involvement in SSc vasculopathy may not depend on the activation of macrophages. Since the SSc CXCL4 serum level reaches 2000-5000 ng/ml in our current study, it can be considered that the release amount in platelet aggregates is much higher. In view of this, the relatively high ~ 20 ng/ml CXCL4 threshold for inducing anti-angiogenic effects can be used to prevent the release of platelets from low concentrations of chemokines, thereby preventing accidental activation of macrophages [52]. In our study, 20 ng/ml CXCL4 may inhibit the endothelial cell proliferation, but 7 µg/ml CXCL4 was needed in M4 macrophage polarization. This discrepancy may be attributable to the different sensitivity of these two different cells to CXCL4. A higher concentration is required to exert the effect when CXCL4 mediates rapid processes such as macrophage activation, while this concentration became lower when CXCL4 is involved in multistep differentiation processes such as angiogenesis [53]. However, the mechanism of these phenomena still remains to be further investigated.
The possible pathobiological role of CXCL4 in brosis of SSc was highly proved in the previous studies [12,54,55]. For example, serum CXCL4 levels were signi cantly increased in SSc and served as a marker of lung brosis [10]. Consistent with these works, we observed mRSS scores positively correlated with CXCL4 levels, suggesting CXCL4 may indicate much severer skin brosis in SSc.
The association of circulating CXCL4 levels with peripheral vasculopathy in patients with VEDOSS also deserves discussion. In fact, we strongly put forward that CXCL4 could serve as a biomarker in VEDOSS. Firstly, we found that serum CXCL4 levels were signi cantly increased in patients with VEDOSS compared with all subsets of SSc, which is similar to other researches [15,56,57]. Secondly, our data suggested that the abnormal NVC pattern and DUs were positively correlated with CXCL4 levels. Finally, CXCL4 could predict new DU in an expected cohort for 6 months. Given that the abnormal NVC pattern and DUs are the most prominent clinical symptoms re ecting pathologically activated angiogenesis in VEDOSS, it suggested that CXCL4 may be a helpful biomarker in VEDOSS patients.
Beyond the abovementioned achievements, there is still a limitation in this study. Although we revealed that CXCL4 can regulate the c-Abl pathway to reduce the expression of Fli-1, more experiments are required to further validate our ndings in vivo and in vitro. In fact, c-Abl activation sequentially decreased the nuclear localization of PKC-δ and Fli-1 phosphorylation at threonine 312, and eventually leading to Fli-1 de ciency. Considering the commercial factors and intellectual property issues in obtaining Fli-1 phosphorylated protein, it is not available for us to further explore the downstream activation of the c-Abl pathway. However, Akamata et al. demonstrated that activation of the c-Abl/ /PKC-δ/pFli-1 pathway reduced the level of Fli-1 protein in broblasts, epithelial cells, endothelial cells, and in SSc animal models [36,41,58]. An alternative research perspective, which only explores the activation of the c-Abl pathway, could be adopted in investigating Fli-1 de ciency in SSc.

Page 12/25
We documented a series of data con rming a possible contribution of CXCL4 to the development of vasculopathy in SSc. CXCL4 may serve as a biomarker in VEDOSS patients. Moreover, CXCL4 is likely to be involved in endothelial cell dysfunction, leading to the development of vasculopathy in SSc. The involvement of the Fli-1 pathway further supports the notion that CXCL4 is a critical anti-angiogenetic factor of SSc. Availability of data and material Our supporting data are available.

Competing interests
The data presented in this manuscript are original and have not been published or submitted elsewhere.
All listed authors have approved the manuscript and agreed with the submission. The authors declare that they have no con ict of interest. Author's Contributions ZJ, CC and SY carried out the in vitro studies. XZ and ML were responsible for the pathological assessment. ML and HH performed the statistical analysis and interpreted the data. ZJ and ML wrote the paper. HH, XZ and ML conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors have read and approved the nal manuscript.  (*P < 0.05, **P < 0.01, ***P < 0.001) Figure 5 The involvement of transforming growth factor (TGF)-β and platelet-derived growth factor (PDGF) signaling in the anti-angiogenetic effects of CXCL4. (A, B) Cell viability was evaluated by CCK8 assay after 24 hours. (A) HUVECs were treated with TGF-β (10ng/ml) for 24 hours, and some cultures were pretreated with SB525334 (1μM), a selective inhibitor of TGF-βR1, or with CXCL4 (20ng/ml) with or without an anti-CXCL4 antibody (3 µg/ml). (B) HUVECs were treated with PDGF (10ng/ml) for 24 hours, and some cultures were pretreated with blocking PDGFR imatinib (1μM), or with CXCL4 (20ng/ml) with or without an anti-CXCL4 antibody (3 µg/ml). (***P < 0.001) Supplementary Files