Abstract
Melanoma, renowned for its aggressive behavior and resistance to conventional treatments, stands as a formidable challenge in the oncology landscape. The dynamic and complex interplay between cancer cells and the tumor microenvironment has gained significant attention, revealing Melanoma-Associated Fibroblasts (MAFs) as central players in disease progression. The heterogeneity of MAFs endows them with a dual role in melanoma. This exhaustive review seeks to not only shed light on the multifaceted roles of MAFs in orchestrating tumor-promoting inflammation but also to explore their involvement in antitumor immunity. By unraveling novel mechanisms underlying MAF functions, this review aims to provide a comprehensive understanding of their impact on melanoma development. Additionally, it delves into the potential of leveraging MAFs for innovative immunotherapeutic strategies, offering new avenues for enhancing treatment outcomes in the challenging realm of melanoma therapeutics.
Introduction
The incidence of melanoma has increased over the past two decades. Each year, melanoma affects more than 325 000 people. Males experience more frequent occurrences, with 174 000 yearly cases compared to females, with 151 000 cases [1, 2]. The intricate dynamics between cancer cells and their microenvironment have come to the forefront of melanoma research, revealing a complex interplay that significantly influences disease progression [3]. Within this dynamic landscape, fibroblasts emerge as pivotal players, particularly cancer-associated fibroblasts (CAFs). CAFs constitute a unique subset of fibroblasts within the tumor stroma, distinguished by their altered phenotype and functional characteristics induced by signals from cancer cells [4]. As orchestrators of the tumor microenvironment, CAFs contribute to various aspects of cancer progression [5]. They actively engage in extracellular matrix (ECM) remodeling, secrete a spectrum of growth factors, cytokines, and chemokines, and create a supportive niche for tumor cells [6]. In addition to inflammatory signals such as IL-1, IL-6, and TNF, stimuli including TGF-β, Notch signaling, reactive oxygen species, ECM remodeling and cytotoxic cancer therapies mediate CAF activation and shape their phenotypes within the TME [7] (Fig. 1). Melanoma-Associated Fibroblasts (MAFs) represent a distinct subset of fibroblasts intricately associated with melanoma [8]. While sharing foundational characteristics with CAFs, MAFs undergo molecular adaptations that tailor their functionality to the unique needs of melanoma cells [9]. Melanoma activates MAFs by secreting inflammatory support cytokines and chemokines, such as IL-6, FGF, and IL-8 [10, 11]. Conversely, the inflammatory cytokines [12], chemokines [13] and growth factors [14–16] secreted by activated MAFs significantly promote melanoma proliferation, invasion, metastasis, and angiogenesis. This bidirectional interaction establishes a complex tumor microenvironment, accelerating the malignant growth of melanoma.
Mechanisms of cancer-associated fibroblasts activation. In addition to inflammatory signals such as IL-1, IL-6, and TNF, stimuli including TGF-β, notch signaling and Eph-ephrins, reactive oxygen species and disrupted metabolism, ECM remodeling, RTK ligands and cytotoxic cancer therapies including chemotheraopy and radiotherapy mediate CAF activation and shape their phenotypes within the TME.
To date, the majority of studies have depicted the tumor-promoting functions of MAFs. Generally, the abundance of MAFs is associated with adverse outcomes in melanoma [17, 18]. However, there is a growing body of research that has revealed MAFs supporting antitumor immunity in melanoma [19]. The inherent ambiguity stems, in part, from the acknowledgment that CAFs do not constitute a singular, homogeneous entity. Instead, they form a heterogeneous group of cells within the TME, showcasing diverse activation patterns [20]. This review seeks to delve into the intricate relationship between MAFs and melanoma, aiming to comprehensively explore their roles in both tumor-promoting inflammation and antitumor immunity. By examining the nuanced mechanisms through which MAFs contribute to melanoma progression, this exploration aims to provide a comprehensive foundation for understanding the complexities of this symbiotic relationship. The subsequent sections will unravel the multifaceted involvement of MAFs, shedding light on novel mechanisms and potential immunotherapeutic strategies that may pave the way for more effective and targeted interventions in the fight against melanoma.
MAFs in tumor-promoting inflammation
Decades ago, the involvement of fibroblasts in inflammation was proposed [21]. Since then, extensive efforts have been made to decipher the specific mechanisms through which CAFs mediate inflammation-driven tumor development. In this context, this article will delve into the attraction exerted by MAFs on pro-tumorigenic and immunosuppressive cell populations. This attraction is achieved through the secretion of pro-inflammatory cytokines, directly stimulating the growth and malignancy of tumor cells, as well as regulating the anti-tumor T-cell response (Fig. 2).
MAFs in tumor-promoting inflammation. MAFs promote tumor progression by inducing the tumor-promoting phenotype of myeloid cells (macrophages and neutrophils) via cyclooxygenase/indoleamine 2,3-dioxygenase-dependent manner, secreting pro-inflammatory cytokines including IL-6, IL-11 and LIF, and suppressing adaptive immune responses involves inhibiting the activity of cytotoxic CD8+ T cells through the recruitment of Tregs and MDSCs, as well as the secretion of cytokines.
MAFs induce the tumor-promoting phenotype of myeloid cells
Myeloid cells, including macrophages, granulocytes, dendritic cells (DC), and myeloid-derived suppressor cells (MDSC), infiltrate tumor sites, actively participating in tumorigenesis. Their roles encompass facilitating tumor cell invasion and metastasis, promoting angiogenesis, and dampening adaptive immune responses, as indicated in the literature [22]. It is noteworthy that among these myeloid cell subpopulations, conventional DCs (cDCs) have been identified as potential contributors to antitumor immunity [23]. Nonetheless, the prevailing consensus is that the presence of myeloid cells in the tumor microenvironment is generally associated with an unfavorable prognosis [22].
Erez et al. have presented groundbreaking evidence elucidating the functional and molecular mechanisms through which CAFs facilitate the growth of squamous cell carcinoma. Their findings highlight the pivotal role of CAFs in attracting macrophages through the release of pro-inflammatory cytokines, including IL-1β and IL-6, as well as chemotactic factors such as CXCL1 and CXCL2, creating a conducive environment for tumor formation [13]. Furthermore, in conditions of low oxygen levels, hypoxia-inducible factor-2α orchestrates the migration of macrophages within CAFs towards pancreatic tumors, guiding their polarization into a pro-tumorigenic phenotype [24]. The above-mentioned study reveals the inducing role of CAFs in promoting the oncogenic phenotype of myeloid cells in cancer.
Similarly, we have observed a similar phenomenon in melanoma. A series of studies have found that MAFs intimately interact with macrophages in vivo, enhancing the secretion of IL-10 from macrophages in a cyclooxygenase/indoleamine 2,3-dioxygenase-dependent manner [25]. Notably, IL-10 is mainly secreted by M2-like macrophages (immunosuppressive type) [26]. Additionally, MAFs, through the secretion of amyloid-like β-peptides (associated with neurodegenerative and inflammatory diseases [27]), employ a reactive oxygen species (ROS)-dependent mechanism. This mechanism, driven by CD11b, induces neutrophils to release chromatin-bound nuclear DNA and cytotoxic granules, forming extracellular traps known as NETs (Neutrophil Extracellular Traps). Reciprocally, NETs enhance MAF activation. This process promotes the progression of melanoma, while inhibiting NETosis in murine tumors shifts neutrophils towards an anti-tumor phenotype, thereby preventing tumor growth [28]. Future research should focus on elucidating the specific molecular mechanisms governing MAF-induced myeloid cell phenotypes, exploring therapeutic strategies to modulate these interactions, and assessing their implications for designing targeted interventions in cancer therapy.
MAFs secrete pro-inflammatory cytokines to promote tumor growth
CAFs exert an influence on tumor growth through an additional mechanism, involving the release of pro-inflammatory cytokines. These cytokines act directly on tumor cells, triggering processes such as cell proliferation, resistance to cell death, and epithelial-to-mesenchymal transition (EMT). Notably, members of the IL-6 family of pro-inflammatory cytokines, including IL-6, IL-11, and leukemia inhibitory factor (LIF), play pivotal roles as essential mediators in the context of inflammation-driven tumorigenesis [29, 30]. In response to diverse stimuli such as IL-1β and TGF-β, CAFs actively secrete IL-6, IL-11, and LIF [31–33]. Significantly, IL-6 secreted by CAFs exhibits pro-proliferative effects across various cancers, including lung cancer [34], colon cancer [35], breast cancer [36], and head and neck squamous cell carcinoma (HNSCC) [37]. Moreover, IL-6 derived from MAFs has been identified as a factor inducing invasiveness in human melanoma cells [12]. Importantly, CAF-induced elevation of invasiveness could be fully inhibited by blocking of IL-8 and IL-6 signalling [12]. Despite the absence of validation through an animal model, the utilization of a 3D co-culture model in this study enhances the credibility of the results. However, the study did not elucidate how the IL-6 released by MAFs regulates the invasion of melanoma, which remains a direction for further investigation in the future. Notably, it has been observed that the hypoxic microenvironment stimulates the secretion of IL-6 by MAFs [38], providing a potential avenue for the development of novel therapeutic strategies targeting the hypoxic microenvironment. This may play a crucial role in the treatment of melanoma. The current research has not investigated whether IL-11 and LIF play a role in mediating the regulation of melanoma growth by MAFs. Future research directions should include an exploration of the involvement of IL-11 and LIF in MAFs’ control of melanoma growth. Further studies should focus on unraveling the specific mechanisms of these two factors in melanoma invasion and growth, aiming to enhance a comprehensive understanding of the role of MAFs in melanoma development.
In addition, other cytokines, such as CXCL5, have been found to be secreted by MAFs and act on melanoma cells, activating the PI3K/AKT signaling pathway in melanoma. This activation promotes the expression of PD-L1, thereby mediating immune escape in melanoma [39]. These findings reveal the intricate role of the CXCL5-CXCR2 axis in shaping the complex tumor-promoting microenvironment, emphasizing that CAFs are a promising target for inhibiting immune evasion.
MAFs suppress adaptive immune responses
Additionally, CAFs contribute to adaptive immune suppression through the mediation of regulatory T cells (Tregs) and MDSCs. These cells employ various inhibitory mechanisms to hinder the function of cytotoxic T cells, particularly CD8+T cells, serving as key facilitators of the suppressive TME [40, 41].
Regarding MAFs, the activation of Snail1 within MAFs promotes the development of an immunosuppressive microenvironment shaped by Tregs, thereby inhibiting the activity of CD8+T cells. Mechanistically, the up-regulation of Snail1 in MAFs activates the expression of fibroblast activating protein (FAP) [42]. FAP is prominently expressed in CAFs in more than 90% of human cancers, including melanoma, and has been implicated in cancer proliferation [43, 44]. It is intriguing that IL-6 derived from CAFs can induce the formation of a subset of Tregs, specifically CD73+γδTregs. However, current research has only revealed its inhibitory effect on the proliferation of CD4+ T cells in breast cancer. It is worth exploring whether MAFs can induce the formation of CD73+γδTregs in melanoma and whether CD73+γδTregs can reshape the immune microenvironment of melanoma [45].
Moreover, MAF-generated chemokines, including CXCL1, CXCL2, CXCL5, CCL2, and CCL3, play a role in attracting pro-tumor MDSCs to inhibit CD8+T cells, contributing to tumor-promoting effects [46]. In addition to inducing MDSCs chemotaxis, studies have suggested that CAFs in pancreatic cancer can promote the differentiation of monocytes into MDSCs by secreting IL-6 [47]. Future research should focus on whether MAFs can induce the differentiation of MDSCs in melanoma and delve into the specific mechanisms involved. In-depth exploration in this field will contribute to a comprehensive understanding of the role of MAFs in melanoma immune modulation, laying a theoretical foundation for more effective therapeutic strategies.
In addition to these indirect influences on the functionality of CD8+T cells, the direct interplay between MAFs and CD8+T cells contributes to the immunosuppressive effects mediated by MAFs. For example, MAFs interfere with intracellular signaling in CD8+T cells through the release of soluble mediators, leading to CD8+T cell anergy. Additionally, they increase the expression of immune checkpoint receptors, such as TIGIT and B and T lymphocyte attenuator (BTLA), by enhancing arginase activity [48]. It is well-known that the expression of TIGIT and BTLA on CD8+T cells can induce CD8+T cell exhaustion [49, 50]. Similarly, research by Jenkins et al. has revealed that MAFs inhibit the infiltration of CD8+ T cells and confer resistance to immune checkpoint blockade (ICB). However, the inhibitory effect of MAFs was reversed, and the sensitivity of melanoma to ICB was enhanced upon downregulation of the MAF receptor Endo180, highlighting the potential benefits of therapeutically targeting MAFs [51]. Moreover, hypoxia enhances the expression and/or secretion of various immunosuppressive factors (such as TGF-β, IL6, IL10, VEGF, and PD-L1) by melanoma-associated fibroblasts, exerting a more pronounced impact on CD8+T cell-mediated cytotoxicity [38].
MAFs exert a profound impact on adaptive immune responses, and understanding their role holds significant clinical relevance and potential applications in melanoma treatment. The immunosuppressive effects mediated by MAFs, particularly through Tregs and MDSCs, hindering the function of cytotoxic T cells, especially CD8+T cells, highlight their crucial role in shaping the suppressive TME. Recognizing the clinical implications of MAF-mediated immune suppression opens avenues for innovative therapeutic interventions and personalized approaches in melanoma treatment. Future research endeavors should focus on unraveling the intricate mechanisms, identifying biomarkers, and developing targeted therapies to enhance the efficacy of existing treatments and pave the way for novel therapeutic strategies.
MAFs in antitumor immunity
CAFs are generally seen as tumor-promoting entity. However, particular CAF subsets can also stimulate and support antitumor immunity and thus restrain tumor growth [31, 52]. For example, it has been demonstrated that CD105-negative CAF subtypes can inhibit tumor growth in pancreatic cancer. This inhibitory phenotype is dependent on an intact adaptive immune system, suggesting that CD105-negative CAFs may promote anti-tumor immunity [53]. Similarly, in melanoma, a subset of MAFs expressing Cxcl13 with a pronounced immunostimulatory signature can reduce local CD8+T cell exhaustion, thereby sustaining control of tumor growth [19].
Tertiary lymphoid structures (TLS), also recognized as “Ectopic lymphoid-like structures” (ELS) or “Tertiary lymphoid organs” (TLO), represent aggregates of B and T cells that form within inflamed and cancerous tissues, mirroring the essential characteristics of secondary lymphoid organs [54]. A noteworthy observation lies in the significant correlation between the presence of tumor-associated TLS (TA-TLS) and improved prognosis, as well as heightened responsiveness to ICB across diverse cancer types including melanoma [55–58]. The increased expression of FAP in CAFs exhibits an inverse correlation with patient survival, and the depletion of these FAP(+) CAFs leads to reduced murine tumor outgrowth [59]. The immunosuppressive role of FAP(+) CAFs involves CXCL12, a chemokine that binds to cancer cells and employs a mechanism dependent on CXCR4 signaling to exclude T cells [60]. The latest research has identified a previously unknown function of a subset of FAP-negative MAFs as organizers of TA-TLS. FAP-negative MAFs stimulate the gathering of B cells in melanomas by activating the CXCL13-CXCR5 axis, thereby facilitating the formation of TA-TLS [61]. Interestingly, this finding underscores the potential diversity of fibroblast functions, as exemplified by a mouse model of Sjögren’s syndrome where FAP + fibroblasts predominantly participate in TLS assembly, indicating a context-dependent role for different fibroblast subsets based on their localization and the specific pathology involved [62]. The evolving landscape of evidence suggests that MAFs extend beyond their traditional understanding, assuming roles in antitumor immune functions. These encompass intricate interactions with Tregs, MDSCs, and the B-cell compartment. As our comprehension of these dynamic interactions deepens, it becomes increasingly evident that the immune-modulatory functions of MAFs contribute significantly to the complex interplay within the tumor microenvironment, thereby influencing therapeutic responses.
MAFs as targets in cancer immunotherapy
Immunotherapy has demonstrated significant potential in cancer treatment, particularly by activating the patient’s own immune system to combat malignant tumors. Innovative treatment approaches, such as immune checkpoint inhibitors, CAR-T cell therapy, and cancer vaccines, have achieved notable successes [63–65]. However, immunotherapy also faces limitations, with one of them being considerable variations in treatment outcomes among different patients. The heterogeneity of individual immune systems and the diversity of tumors make it challenging to apply a single treatment approach to all cancer patients. Additionally, some patients may develop resistance to immunotherapy after initial treatment, limiting its long-term effectiveness [66]. Hence, targeting MAFs emerges as a hopeful strategy to overcome resistance to immunotherapy for melanoma. However, it is crucial to exercise caution when depleting MAFs from the tumor microenvironment, as indicated by the potential adverse impacts observed in preclinical models, as mentioned earlier.
Depleting the immunosuppressive FAP + MAF population
In a recent study, Hajdara et al. demonstrated in an in vitro model that zoledronic acid activated γδT cells targeted MAFs, leading to apoptosis of melanoma cells. Given the safe and tolerable administration of zoledronic acid in humans, the results of this study contribute valuable data for potential future clinical studies exploring its efficacy in melanoma treatment [67]. However, as mentioned in the above study, we have already proposed that FAP-negative MAFs support anti-tumor immunity. Therefore, non-specific targeting of FAP + MAFs may lead to the elimination of immunostimulatory MAFs. It is noteworthy that an early study has already focused on the potential therapeutic value of targeting FAP + MAFs. Sorrentino et al. found that the activation of the A2B adenosine receptor in B16 melanomas induces CXCL12 expression in FAP + MAFs, thereby enhancing tumor progression. However, the pharmacological inhibition of A2BR with PSB1115, known to inhibit tumor growth, led to a decrease in the number of FAP + MAFs in melanoma [68]. This suggests that targeting FAP+ MAFs with drugs could benefit the clinical treatment of melanoma, emphasizing the pressing need for the development of more specific and safer targeted therapeutic strategies against FAP+ MAFs.
Chimeric Antigen Receptor (CAR)-T cell therapy is an innovative cancer treatment method known for its advantage of enhancing the anti-tumor capabilities of a patient’s T cells through their redesign. Grounded in the patient’s own T cells, CAR-T cell therapy involves modifying these cells to target specific cancer cells within the patient’s body. This personalized treatment approach not only enhances efficacy but also reduces side effects. Additionally, CAR-T cells are engineered to precisely recognize and attack specific antigens on the surface of cancer cells, minimizing damage to normal cells. This provides a more accurate and targeted therapeutic option [69, 70]. Multiple CAR-T therapies have been found to inhibit melanoma progression in preclinical studies [71, 72]. In addition, several clinical trials have been conducted to investigate the efficacy and safety of CAR-T therapy in melanoma (NCT03638206, NCT03060356, NCT04119024). Remarkably, findings from ongoing clinical trials suggest that intravenous cMET-directed CAR-T cell therapy is both safe and viable for patients dealing with metastatic melanoma [73]. Nevertheless, there is currently a dearth of clinical or preclinical studies investigating CAR-T targeting FAP in melanoma. Noteworthy research has shown that CAR-T therapy directed at FAP+ cells significantly reduces tumor burden in a lung cancer model [74]. Similarly, in vivo animal models conducted by Wang et al. demonstrated that FAP-specific CAR-T therapy enhances the response of endogenous CD8+ T cells to tumors [75]. However, it is crucial to acknowledge that FAP-specific CAR-T therapy has been linked to severe bone toxicity and cachexia in various subcutaneous tumor models, emphasizing the need for caution when extrapolating these findings to clinical applications. The aforementioned results underscore the potential of CAR-T therapy targeting FAP+ cells for cancer treatment. Subsequent research could involve the design of CAR-T cells targeting FAP and an exploration of their efficacy and safety in melanoma. However, careful consideration must be given to design strategies, with a focus on mitigating potential bone toxicity and cachexia.
Combination therapy of immunotherapy targeting MAFs with chemotherapy
An examination of immune infiltration and the clinical relevance of features associated with MAFs in cutaneous melanoma has indicated that patients with elevated MAF scores are less likely to respond to immune checkpoint inhibitors. However, they may demonstrate heightened sensitivity to specific chemotherapy drugs. This points to the possibility of implementing a combined approach involving chemotherapy and anti-MAF treatments to improve the responsiveness of skin cutaneous melanoma patients to existing immunotherapy [76].
In a specific practical study, Morales et al. discovered that, through the construction of a 3D culture model, MAFs influence the sensitivity of metastatic melanoma cells to the combination of BRAF inhibitor (vemurafenib) and MEK inhibitor (cobimetinib) [77]. Moreover, Ocoxin (a natural compound-based antioxidant and anti-inflammatory nutritional supplement) has been shown to enhance the anti-tumor effects of vemurafenib and reduce the chemoresistance mediated by MAFs in metastatic melanoma [78]. Crucially, Ocoxin inhibits the migratory ability of MAFs, contributing to the suppression of MAF-mediated immune inhibition [79]. This makes the combined treatment of immunotherapy targeting MAFs and chemotherapy a promising therapeutic strategy for melanoma. Similarly, a TLR3 agonist not only enhances the death of melanoma cells induced by low-dose cisplatin but also inhibits immunosuppressive MAFs. This reinforces cisplatin-based chemoimmunotherapy, and the strategy can alleviate side effects [80].
The clinical significance lies in the prospect of overcoming the limitations associated with immune checkpoint inhibitors and enhancing the efficacy of existing immunotherapies for cutaneous melanoma patients. The combined approach holds promise for more effective and personalized treatment strategies, paving the way for future investigations into the intricate interactions within the tumor microenvironment and the development of targeted therapeutic interventions.
Combination therapy of targeted MAFs with ICB therapy
The emergence of resistance to PD-1/PD-L1 inhibitors has posed a significant obstacle to their application in melanoma treatment, emphasizing the critical need to overcome resistance [81].
Nintedanib elicited a concentration-dependent decrease in FAP expression within TGF-β1-stimulated mouse fibroblast cells. This reduction in FAP expression resulted in the inhibition of fibroblast proliferation and activation, thereby mitigating the immunosuppressive effects of the TME. Consequently, there was an augmentation in the infiltration and activation of CD8+T cells [82]. Importantly, further research revealed that nintedanib increased the sensitivity of tumors to PD-1 blockade therapy by promoting immune activation [82], as shown in Table 1.
MAFs as targets in cancer immunotherapy.
MAFs as targets in cancer immunotherapy.
This study proposes a series of in-depth research directions. Firstly, clinical trials for the combination therapy strategy are crucial and should be advanced to a larger patient population to comprehensively assess its efficacy and safety in treating melanoma. Secondly, deeper mechanistic studies are needed, particularly regarding the impact of Nintedanib on other immune cell types within the TME, for a more comprehensive understanding of its mechanisms in improving the immune environment. Simultaneously, in-depth research on the mechanisms of PD-1 inhibitor resistance is necessary to identify more targets suitable for combination therapy, with a judicious introduction of Nintedanib during the resistance stage. Lastly, to achieve more personalized treatment plans, researchers should explore predictive biomarkers for individualized treatment responses. These research directions hold the promise of providing more effective and precise strategies for future cancer treatments.
Conclusions and future challenges
In conclusion, this review extensively examined the intricate interplay between MAFs and melanoma, highlighting their dual role in tumor-promoting inflammation and antitumor immunity. The heterogeneity of MAFs and their influence on myeloid cell phenotypes, pro-inflammatory cytokine secretion, and adaptive immune responses were thoroughly explored. The bidirectional interaction between MAFs and melanoma contributes to the complexity of the tumor microenvironment, influencing melanoma proliferation, invasion, and immune evasion. Despite the predominantly tumor-promoting functions attributed to MAFs, emerging evidence suggests their involvement in supporting antitumor immunity, revealing a nuanced and context-dependent role.
Future challenges in this field include unraveling the specific molecular mechanisms governing MAF-induced myeloid cell phenotypes, developing targeted therapeutic strategies against immunosuppressive MAFs, and exploring the implications of MAF-mediated immune modulation. Additionally, understanding the potential of combining immunotherapy targeting MAFs with chemotherapy or immune checkpoint blockade therapy presents opportunities for more effective and personalized melanoma treatment. Efforts should focus on advancing clinical trials, elucidating predictive biomarkers, and deepening mechanistic studies to enhance the efficacy and safety of interventions targeting MAFs. The evolving landscape of MAFs’ roles in antitumor immunity calls for continued exploration, offering promise for innovative therapeutic interventions in the challenging realm of melanoma therapeutics.
Acknowledgements
None.
Conflict of interest statement: The authors declare that they have no competing interests.
Funding
This work was supported by Cultivation of Postgraduate Topnotch Innovative Talents in 2023 (Zhou Qiujun) [721100G00736] and Program of Zhejiang Provincial TCM Sci-tech Plan (2024ZR071).
References
Author notes
Qiujun Zhou, Xiaoliang Jin, Ying Zhao and Yueping Wang contributed equally to the manuscript.
