Serological assessment of collagen fragments and tumor fibrosis may guide immune checkpoint inhibitor therapy

Detection of PD-L1 tumor tissue expression with immunohistochemistry emerged as one of the first and most studied biomarkers for anti-PD-1/PD-L1 therapy based on the assumption that PD-L1 should be expressed for anti-PD-1/PD-L1 to induce a response [38]. But, across different solid tumor types treated with anti-PD-1/PD-L1 drugs, PD-L1 has only been shown to be predictive in around 30% of cases [30]. PD-L1 has been FDA approved as a biomarker for use in NSCLC, head and neck squamous cell carcinoma, bladder, breast, cervical and gastric cancer, but though extensively studied in melanoma, it has not been FDA approved for this indication [29, 30]. When assessing the PD-L1 biomarker in several melanoma studies, a correlation between PD-L1 expression and response to anti-PD-1 therapy was only seen in five out of eight studies [20]. One of the main concerns is that a subset of patients that have low PD-L1 levels will respond to anti-PD-1 therapy.

There are a variety of limitations concerning PD-L1 expression measured in a tissue biopsy including differences in PD-L1 thresholds, differences in diagnostic antibodies applied, different detection assays, the heterogenicity of PD-L1 expression between serial sections of a primary tumor and metastatic sites, the dynamic PD-L1 expression over time, and that it is a semiquantitative approach based on visual assessments [29, 39].

Although PD-L1 expression is the most investigated and best-validated biomarker of response to ICI therapy, at least more refinement, optimization, and validation are needed for PD-L1 expression to be a useful and reliable predictive biomarker [29, 40].

Modern high-throughput technologies including mass spectrometry, whole-exome, and RNA sequencing allow comprehensive profiling of individual patients and assessment of thousands of genes and proteins [41, 42]. High TMB, which may induce more tumor-specific neoantigens that can enhance T cell responses against tumors, has been widely investigated as a predictive biomarker for ICIs in solid tumor types such as melanoma, breast cancer, renal cancer, and NSCLC [20, 28, 31, 32, 43]. Recently, Litchfield et al. have assessed whole-exome and transcriptomic data for > 1000 ICI-treated patients across seven tumor types and shown that clonal TMB is a strong predictor of response followed by TMB and CXCL9 expression [44]. Moreover, based on the successful phase 2 KEYNOTE-158 study, the FDA has granted accelerated approval of pembrolizumab (anti-PD-1) for the treatment of patients with unresectable or metastatic TMB-high solid tumors whose cancer has progressed after previous therapy [45]. However, a new TCGA analysis with over 10,000 patient tumors failed to support the application of TMB-high as a biomarker for ICI therapy in all solid tumor types and suggests that further tumor-specific studies are needed [46].

Despite encouraging findings of TMB in some tumor types, such as NSCLC [28, 47], and advantages with high-throughput technologies, the broad applicability of TMB as a biomarker across all solid tumors is unclear. It is challenging to detect the exact neoantigens that induce T cell responses, and TMB assessment in a tissue biopsy is limited by patient surgical risk, intratumor heterogeneity, different cut-off values, detection methods, and overlap between responding and non-responding patients [20, 31, 32, 48, 49].

In the pursuit of overcoming the tumor heterogeneity detected in tumor biopsies, and the fact that tumor biopsies are challenging and sometimes impossible to obtain [35], assessment of circulating biomarkers originating from tumors including circulating tumor cells, proteins, tumor DNA, and tumor RNA have been investigated as biomarkers [50]. A decrease in circulating tumor DNA (ctDNA) assessed in peripheral blood during ICI therapy has been found to correlate with response to combined anti-CTLA-4 and anti-PD-1 therapy in metastatic melanoma patients [51]. Exosomal PD-L1 from tumors has been identified as an alternative mechanism of PD-L1 activity, which suppresses T cell activation suggesting that exosomal PD-L1 represents a potential therapeutic target and novel predictive biomarker for anti-PD-1 therapy [52, 53]. Low soluble PD-L1 plasma levels have also been discovered as a potential biomarker for the prediction of response to anti-PD-1 therapy in NSCLC patients [54].

However, similar to PD-L1 expression assessed with immunohistochemistry, the clinical application of these tumor-derived liquid biopsies is also challenged by limited standardized methods, discordant results, and false negatives [50].

The biomarkers reviewed in this section, such as PD-L1 expression, TMB, and ctDNA have some technical limitations. Moreover, though the biomarkers are related to T cell activity, their predictive value for ICI response varies across different solid tumor types suggesting that a broader view of regulating factors in the tumor microenvironment is needed.

Increasing evidence suggests that the ECM, the non-cellular component of the tumor microenvironment, influences response to immunotherapy [33, 55]. The ECM is composed of the basement membrane and the interstitial matrix. The basement membrane is a sheet-like ECM layer located beneath the endothelial and epithelial cells, serving as a barrier to the underlying stroma, and is primary composed of type IV and VIII collagen, laminin, nidogen, and perlecan [56]. The interstitial matrix is responsible for tissue structure and is mainly composed of type I, III, V, VI, and XI collagen, which are primarily produced by fibroblasts [57].

Collagens are the most abundant ECM proteins in the tumor microenvironment across different solid tumor types, and in addition to their well-known role in tumor progression and metastasis, several recent studies suggest that collagens have a direct role in resistance to ICI therapy [11‐16, 58‐60]. It has been shown that an immune-excluded tumor phenotype is characteristic of having CAF activity, and a collagen-rich stroma (tumor fibrosis) that influences the efficacy of ICIs by acting as a protective shield for the tumor by trapping the T cells [11, 16, 61, 62]. CAFs are one of the most abundant cell types in the tumor microenvironment, and in addition to being the main producer of fibrotic ECM proteins and collagens, CAFs have multiple immunosuppressive functions and secrete numerous chemokines, cytokines, and proteases, such as TGF-β, VEGF, CXCL12/CXCR4, interleukin-6, and matrix metalloproteinases (MMPs) [58, 63]. In metastatic urothelial cancer patients, lack of response to anti-PD-L1/PD-1 therapy was associated with TGF-β signaling in fibroblasts and CD8+ T cell excluded tumors, where T cells were retained in a fibroblast-and collagen-rich peritumoral stroma [11, 12]. Moreover, co-administration of anti-PD-L1 and anti-TGF-β antibodies in both a mammary carcinoma mouse model and in colon carcinoma mouse models have been shown to induce T cell infiltration into the tumor, which was associated with an anti-tumor response [11, 64, 65]. The association between upregulated ECM and TGF-β genes, and T cell excluded tumors with a peri-tumoral location of CD8+ T cells has also been detected in hepatocellular carcinoma patients supporting that tumor fibrosis affects the anti-tumor immune response across different tumor types [16]. Moreover, a large pan-cancer analysis has shown that an ‘ECM-up’ signature is activated by TGF-β signaling in CAFs and that it is the ‘ECM-up’ signature per se that is associated with anti-PD-1 resistance, and not just CAFs or TGF-β activation [13]. Interestingly, the predictive performance of the ‘ECM-up’ signature outperformed a T cell inflamed signature and mutational load alone [13]. TGF-β has a highly pleiotropic nature with both immune regulatory mechanisms and a critical role in generating the fibrotic tumor microenvironment [66]. TGF-β family members including TGF-β1 signal through TGF-βR to promote the expression of ECM proteins including collagen, fibronectin, tenascin C, and laminin [67, 68]. Cancer cells secrete TGF-β that promotes CAF contractility and secretion of TGF-β and MMPs leading to collagen production and degradation, respectively [69]. TGF-β signals by activating non-canonical signaling pathways such as RhoA/ROCK, and by activating canonical signaling pathways to induce SNAIL1 and TWIST1 gene transcription [66, 70‐72].

Tumor fibrosis and collagens also have direct immunosuppressive functions. A high-density collagen matrix induces the downregulation of cytotoxic activity markers and upregulation of regulatory T cell markers [15]. Increased collagen levels correlate with an increased amount of exhausted CD8+ T cells and resistance to PD-1/PD-L1 blockade in lung tumors [14]. This CD8+ T cell exhaustion is induced by collagen binding to the collagen receptor leukocyte-associated immunoglobulin-like receptor 1 (LAIR-1), which suppresses T cell activity through SHP-1 signaling supporting a direct immune regulating role of collagen [14]. A recent study also shows the relevance of blocking the LAIR-1-collagen interaction as a novel checkpoint inhibitor approach in cancer [73]. Cleavage of collagen and ECM proteins by MMPs and other remodeling enzymes generate a variety of bioactive peptide fragments that may have signaling properties and act as chemokines, cytokines, or interact with immune regulators such as LAIR-1 [25, 74].

Together, the studies highlighted in this section reveal the central roles of upregulated TGF-β-signaling in CAFs and the increased production of collagens in inducing immune evasion, immune exclusion, and resistance to ICIs. In contrast, it has also been shown that depletion of fibroblasts and reduction of collagens/tumor fibrosis can lead to immune suppression and poor survival [75, 76]. Chen et al. have recently shown in a mouse model with spontaneous pancreatic cancer that a significant reduction in total stromal type I collagen can accelerate pancreatic tumor progression and augment suppression of CD8+ T cells leading to decreased overall survival [76]. This proposes that optimal T cell immunity and response to immunotherapy are associated with a balanced degree of ECM/collagen formation and degradation and that both excessive collagen production and excessive collagen degradation associate with resistance and poor survival outcomes (Fig. 1). Together, this highlights a need and a potential for collagen-based biomarkers that assess the activity of TGF-β, CAFs, collagen formation, and collagen degradation.

The findings that collagens seem to have a direct role in resistance to ICI therapy are mainly based on collagen mRNA expression and trichrome staining for total collagen in tissue biopsies [11‐14, 16]. Such tissue-based measures only provide a snapshot of the status of the total collagen content which makes it difficult to distinguish increased collagen deposition and degradation. Moreover, it is difficult to monitor patients as serial biopsies are rarely an option. Tumor fibrosis is an active process that includes the activation of fibroblasts and CAFs, the release of TGF-β, the production of collagen, and degradation of collagens by proteases such as MMPs [59]. Serological assessment of collagen fragments and fibrotic activity may be ideally suited as a personalized health care tool that can provide additional information to the static measure of tumor fibrosis assessed by tissue-based technologies such as immunohistochemistry or similar.

Focusing on specific subdomains of proteins, such as the collagen pro-peptides, rather than quantification of the total proteins with standard assay technologies, may provide an advantage. Collagen contains pro-peptides which are released during collagen synthesis. These pro-peptides are thus an indirect measure of CAF activity, collagen production, and fibrogenesis [57, 77]. In direct alignment, quantification of special degradation epitopes, provides a completely different type of information, on tissue degradation by specific cell types [78]. Par example, quantification of neo-epitopes which are the product of a signature ECM protein and a specific protease makes it possible to assess separate processes such as MMP-mediated collagen degradation and collagen formation of specific collagens [79, 80]. During tissue remodeling, collagens are remodeled and these specific protein fragments are released into the circulation and can be used as non-invasive biomarkers assessed in a liquid biopsy (serum or plasma) by quantitative immunoassays compared to a semi-quantitative approach such as trichrome staining of tissue (Fig. 2) [79, 80].

Two examples of neo-epitope assays are related to the two most abundant collagens in the ECM, type III (interstitial matrix) and type IV collagen (basement membrane) that have different positions and roles in the tumor microenvironment (Fig. 2) [59]. The PRO-C3 biomarker measures the pro-peptide released during type III collagen formation, which makes it feasible to assess the production and deposition of type III collagen [81], whereas type III collagen degraded by MMP can be quantified by the C3M biomarker [82]. PRO-C3 is found to be released by TGF-β stimulated CAFs in vitro [83], and high PRO-C3 in pre-treatment serum has been shown to predict poor overall survival in metastatic melanoma patients treated with anti-CTLA-4 or anti-PD-1 therapy [83‐85]. These studies suggest that PRO-C3 can be used to assess CAF activity and the deposition of the collagen-rich peritumoral stroma that is associated with resistance to ICI therapy [11‐16].

In the pursuit of identifying patients with high T cell infiltration, a biomarker measuring granzyme B degraded type IV collagen products (C4G) in serum has been developed [86]. The discovery of C4G was based on a study from Prakash et al. showing that granzyme B promotes cytotoxic lymphocyte transmigration over the basement membrane by degrading components such as type IV collagen [87]. High levels of this C4G biomarker at baseline could identify melanoma patients responding to anti-CTLA-4 treatment and when combined with low PRO-C3, this biomarker combination could identify additional responding patients [86]. Another study shows in vivo how C4G levels are increased in serum after inducing T cell activity with antibodies that block the LAIR-1-collagen interaction [73]. Here, the increase in C4G was observed at the time of tumor eradication supporting that the fragments were derived from the tumor microenvironment as a result of T cell activation and effector function. Interestingly, while the granzyme B degraded type IV collagen (C4G) fragments were increased in melanoma patients responding to anti-CTLA-4 treatment and modulated by inducing T cell activity in vivo, another MMP-degraded type IV collagen (C4M) fragment showed the opposite prognostic potential in the melanoma patients treated with anti-CTLA-4 and was not modulated by inducing T cell activity in vivo [73, 86]. This suggests that different protease-generated neo-epitopes on the same collagen reflect different biological processes (Fig. 2) [73, 84, 86]. Furthermore, these studies support the value of measuring specific neo-epitopes and not just the total protein.

The ECM is highly dynamic and constantly remodeled, and with these neo-epitope activity biomarkers, it is possible to assess processes such as fibrotic activity and T cell activity. Interestingly, it has been shown that high type III collagen turnover (C3M/PRO-C3 ratio) measured in the circulation is superior to hyaluronan assessed in a tissue biopsy from matched pancreatic cancer patients in predicting response to a stromal modifier in combination with chemotherapy [88]. Despite concomitant overexpression of type III collagen and hyaluronan in pancreatic tumors, high type III collagen turnover (C3M/PRO-C3 ratio) has predictive value both in hyaluronan low and hyaluronan high patients suggesting that the tissue biopsy and liquid biopsy do not identify the same patients [88, 89]. This study highlights the unique value of measuring the fibrotic activity non-invasively compared to assessing the static fibrosis in a tumor biopsy, where it is difficult to distinguish fibrolysis and fibrogenesis.