Immunosuppressive roles of pericytes in cancer disease: insights from a systematic review and meta-analysis

Abstract

Background

Although the crosstalk between the immune system and tumour endothelial cells has been extensively investigated, interactions with tumour-associated pericytes (TAPs) remain poorly understood. Such issue may be of great interest considering the immunoregulatory roles of pericytes reported in both healthy and tumour tissues. This systematic review integrates current evidence on TAP–immune system interplay, encompassing functional roles, molecular mechanisms, associated biomarkers, and in vitro assessment methodologies. Preclinical investigations on the anticancer potential of TAP-targeted immunotherapies were also critically analysed, complemented by a meta-analysis assessing TAP-mediated immune regulation across distinct malignancies with a larger sample size.

Methods

Following the 2020 PRISMA guidelines, literature searches were conducted in PubMed, Web of Science, and Scopus for studies published up to February 2026. Of 3,557 records, 64 met the inclusion criteria after multi-stage screening (“title and abstract,” “full text” and “reference check”). Risk of bias was assessed using the OHAT tool for in vitro studies and the SYRCLE tool for in vivo research. Meta-analyses were performed using UALCAN and TIMER platforms.

Results

Our findings identified tumour-associated macrophages (TAMs) as the main immune regulators of TAPs, while TAPs modulate a broad range of immune cells—including TAMs, dendritic cells, B and T lymphocytes, regulatory T cells, myeloid-derived suppressor cells, and natural killer cells—thereby promoting an immunosuppressive tumour microenvironment. Indeed, experimental studies demonstrate the capacity of TAPs to reduce the anti-tumour efficacy of traditional cancer immunotherapies like inhibitory immune checkpoint blockade or adoptive immune cells therapies. Evidence remains scarce for certain cancers (e.g., cholangiocarcinoma, hepatic, and head and neck squamous cell carcinoma) and for lymphatic TAPs. Marked heterogeneity among TAPs and the limited availability of commercial pericyte-like cell lines represent major experimental challenges. Immunotherapies targeting TAPs, such as cancer vaccines and CAR-engineered immune cells, showed promising preclinical efficacy, highlighting their potential as translational medicine. Our meta-analysis further supports the strong immunosuppressive potential of TAPs, even in the underexplored malignancies above mentioned.

Conclusions

Collectively, this study highlights the crucial interplay between TAPs and immune cells in tumour biology and underscores the therapeutic promise of TAP-targeted immunotherapies, along with the importance of long-term evaluation of TAP-targeted immunotherapies and the development of standardized pericyte-like cell models.

Trial registration

The section “Immunotherapies Targeting TAPs” was registered in PROSPERO (ID: CRD420251053567).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-08216-9.

Introduction

Cancer remains one of the leading causes of mortality worldwide. The American Cancer Society projects more than two million new cancer cases and over 600,000 related deaths in the United States in 2025 [1]. Over the past two decades, immunotherapy has emerged as a transformative approach in cancer treatment and one of the fastest-growing therapeutic fields [2]. However, several tumour types remain refractory or develop immune resistance, particularly solid malignancies such as advanced melanoma [3], hepatocellular carcinoma [4], and pancreatic ductal adenocarcinoma (PDAC) [5]. Increasing evidence indicates that stromal cells within the tumour microenvironment (TME) play pivotal roles in mediating cancer immune evasion. For instance, a study in non-small cell lung cancer reported that interactions between cancer cells, tumour-associated macrophages (TAMs), and cancer-associated fibroblasts (CAFs) promoted extensive collagen fibers deposition at tumour margins, leading to impaired T-lymphocyte infiltration and resistance to immune checkpoint inhibitors (ICIs) [6].

Tumour aberrant angiogenesis, characterized by both structural (immature, tortuous, blind-ended vessels) and functional (increased permeability, leakiness, and deficient perfusion) abnormalities, has long been associated with tumour growth, progression, and drug resistance [7]. Tumour vasculature and its cellular components also contribute to cancer progression beyond their role in supplying nutrients and oxygen. Several studies have demonstrated that abnormal tumour vasculature can directly impair the migration and infiltration of CD8⁺ T cells within the tumour mass, whereas vascular normalization improves cancer immunotherapy efficacy [8]. However, this may appear counterintuitive given the higher permeability of abnormal tumour vessels. In this context, the phenomenon of “endothelial cell anergy” has been well documented, wherein tumour-associated endothelial cells fail to express key adhesion molecules (e.g., VCAM-1, ICAM-1) in response to proinflammatory stimuli, thereby reducing intratumoural leukocyte recruitment [9]. Similarly, a recent study identified a subset of CXCL12⁺ tumour-associated endothelial cells in hepatocellular carcinoma that promote immune resistance by recruiting myeloid-derived suppressor cells (MDSCs) and inhibiting the differentiation of cytotoxic CD8⁺ T cells [10].

Pericytes, the other major cellular component of small capillaries, also play essential roles in vascular integrity and immune homeostasis. For example, physical interactions between mural cells (including pericytes) and macrophages regulate T-lymphocyte trafficking within the central nervous system [11]. In cancer, colorectal tumour cells have been shown to activate tumour-associated pericytes (TAPs) via TGFβ1 secretion, conferring a more tumorigenic phenotype. These TAPs secrete high levels of IGFBP-3, which enhance malignant cell proliferation, migration, invasiveness, chemoresistance, and stemness [12].

Nevertheless, the crosstalk between immune cells and pericytes remains relatively underexplored in oncology. For instance, in nasopharyngeal carcinoma, TAPs were reported to recruit and polarize TAMs toward an immunosuppressive M2-like phenotype (CD163⁺/CD206⁺), thereby promoting tumour growth and metastasis [13]. Conversely, TAMs have been shown to induce TAP maturation, characterized by increased αSMA expression, and to enhance pericyte-endothelial cell interactions, contributing to vascular stabilization and tumour angiogenesis [14].

Furthermore, the direct targeting of TAPs through immunotherapy-based strategies has shown promising preclinical outcomes, although this therapeutic approach remains in its early stages. For example, Komita et al. identified haemoglobin-β (HBβ) as a TAP-associated antigen capable of eliciting cytotoxic CD8⁺ T-cell responses following intratumoural dendritic cell (DC) vaccination, thereby reducing tumour incidence and inhibiting tumour progression [15]. Both prophylactic (reduced tumour incidence) and therapeutic (suppressed tumour growth) effects were demonstrated in murine models of sarcoma, breast, and colon cancers [15].

Given the growing relevance of cancer immunotherapy, it is essential to understand the complex mechanisms governing immune regulation within the TME. Therefore, the aim of this study was to conduct a qualitative systematic review of studies evaluating tumour immunotherapies directly targeting TAPs (e.g., cancer vaccines or chimeric antigen receptor [CAR]-engineered immune cells). To support this analysis, we also reviewed experimental research describing the interplay between pericytes and immune cells in the context of cancer, discussing current evidence and identifying key knowledge gaps, including methodological challenges related to pericyte isolation and characterization.

In addition, we performed a quantitative meta-analysis to compare the expression of several pericyte-related biomarkers across multiple cancer types and corresponding normal tissues. We further assessed the correlations between these biomarkers and immune infiltration patterns, specifically CD8⁺ T cells, regulatory T cells (Tregs), and macrophages, across diverse malignancies. Overall, our findings suggest that pericytes/TAPs possess significant immunosuppressive potential in several cancers, particularly head and neck carcinoma, hepatocellular carcinoma, and cholangiocarcinoma, highlighting promising avenues for future experimental investigation.

Methodology

This article comprises three main components: (i) a systematic review of in vitro and in vivo experimental studies assessing the direct interactions between pericytes and immune cells in the context of cancer; (ii) a systematic review of preclinical studies evaluating the antitumour potential of various immunotherapy-based strategies targeting pericytes; and (iii) a meta-analysis of pericyte-associated genes identified in parts I and II. Part I and III were not pre-registered, whereas Part II was registered in PROSPERO (ID: CRD420251053567). The entire study was conducted in accordance with the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [16].

Search strategy and data extraction

Literature review was carried out using public databases including PubMed, Web of Science and Scopus, from inception until February 2026, employing the following search string: (“cancer*” OR “tumor*” OR “tumour*”) and (“pericyte*” OR “mural cell*”) and (“immune cell*” OR “immune system*” OR “lymphocyte*” OR “t cell*” OR “b cell*” OR “macrophage*” OR “natural killer*” OR “neutrophil*” OR “dendritic cell*” OR “mast cell*” OR “immunotherapy*” OR “immune checkpoint blockade*” OR “immune checkpoint inhibitor*” OR “cancer vaccine*” OR “adoptive t cell*” OR “car-t” OR “car T” OR “TCR”).

The screening protocol and outcomes are illustrated in Fig. 1. Articles included in Part I were evaluated for eligibility based on specific inclusion and exclusion criteria (Table 1), while those in Part II were selected according to the criteria defined in the PROSPERO registry (ID: CRD420251053567). In total, 3557 articles were initially obtained, and 64 papers were ultimately selected for data analysis. The screening process was conducted independently by two investigators, and any discrepancies were resolved by consultation with a third reviewer. Data extraction was performed by two independent authors using predefined variables, with disagreements similarly resolved through adjudication by a third investigator.

Fig. 1.

PRISMA flow diagram of study selection. Flowchart illustrating the study identification, screening, elegibility and inclusion process. A total of 3,557 records were identified through database searching (PubMed, Web of Science and Scopus). After removal of 1,222 duplicates, 2,335 records were screened by title, abstract and keywords. Of these, 2,045 were excluded based on predefined criteria. A total of 290 full-text articles were assessed for elegibility, of which 227 were excluded for reasons including lack of relevance to the research topic, absence of a relationship between pericytes and immune cells, computational-only studies or unavailable full texts. Finally, 64 studies were included in the systematic review

Table 1.

Inclusion and exclusion criteria used in the literature search process

Category	Inclusion criteria	Exclusion criteria
Article type	Original in vitro, in vivo, or analyses using patient samples	In silico studies, Systematic reviews, reviews, case report/case series, comments/notes, book chapters, thesis, conference papers/abstracts, editorial letters, observational studies, comparative studies, clinical trials. Articles with full text not available
Experimental context	Original in vitro and in vivo studies (using mouse models) showing a direct link between pericytes and immune cells in the context of cancer disease. Preclinical studies testing a particular anticancer immunotherapy targeting pericytes	Articles not focusing on cancer. Studies not reporting a direct/causative link between pericytes and immune cells regulation (e.g., vascular normalizing drugs that, although exhibiting immunomodulatory properties and altering pericyte coverage of blood vessels, they mainly target other cell types like endothelial cells).
Publication date	From inception until February 2026	Articles published after February 2026
Language	Publications in English language	Different from English language.

Risk of bias

For in vitro works, risk of bias was assessed employing the Office of Health Assessment and Translation (OHAT) tool. The procedure was carried out based on the following fields: identical experimental conditions, blinding of researchers, completeness of outcome data, exposure characterization, outcome assessment, complete reporting of outcomes, and statistical precision [17]. Quality evaluation of studies using animal models was performed with the Systematic Review Center for Laboratory Animal Experimentation (SYRCLE) tool considering the following criteria: sequence generation (selection bias), baseline characteristics (selection bias), allocation concealment (selection bias), random housing (performance bias), blinding (performance bias), random outcome assessment (detection bias), blinding (detection bias), incomplete outcome data (attrition bias) and selective outcome reporting (reporting bias) [18]. Risk of bias assessments were conducted by two independent researchers, resolving any discrepancies according to the criteria of a third author. All eligible studies were included in the review regardless of their risk of bias rating. The Robins tool was utilized to visualize such quality assessments [19].

Meta-analysis

Analysis of gene expression levels (UALCAN)

For this meta-analysis, TCGA datasets corresponding to different cancer types were queried through UALCAN (http://ualcan.path.uab.edu). For each cancer type, a predefined set of pericyte-related biomarkers was evaluated in tumour and corresponding normal tissue cohorts. Cancer types were included only when both tumour and matched control samples were available and cohort definitions were clearly established. The number of patients per group (tumour vs. control) for each cohort is summarized in Table 2.

Table 2.

Patient distribution by cancer type (Tumour vs. Control)

Cancer Type	Control (n)	Tumour (n)
Breast Invasive Cancer	114	1097
Bladder urothelial carcinoma	19	408
Cervical squamous cell carcinoma	3	305
Cholangiocarcinoma	9	36
Colon adenocarcinoma	41	286
Esophageal Carcinoma	11	184
Glioblastoma multiforme	5	156
Head and neck squamous cell carcinoma	44	520
Kidney renal clear cell carcinoma	72	533
Liver hepatocellular carcinoma	50	371
Lung adenocarcinoma	59	515
Lung squamous cell carcinoma	52	503
Pancreatic Adenocarcinoma	4	178
Prostate adenocarcinoma	52	497
Rectum Adenocarcinoma	10	166
Sarcoma	2	260
Stomach adenocarcinoma	34	415
Thyroid carcinoma	59	505
Uterine Corpus Endometrial Carcinoma	35	546

For every cancer type, the median expression level was used as the summary statistic to reduce the influence of outliers and skewed distributions typical of patient data. A heat-map representation was generated from log10(fold-change) values computed between tumour and control medians. Statistical comparisons between normal and tumour groups were performed using the tools provided by UALCAN; specifically, differences in medians were assessed with the t-test. Statistical significance thresholds were set at p < 0.05, p < 0.01, and p < 0.001.

Immune infiltration association analysis (TIMER3)

To assess how pericyte biomarkers relate to the immune contexture of tumours, we used TIMER3 (https://compbio.cn/timer3/). TIMER3 integrates TCGA RNA-seq data and estimates immune cell abundances via multiple deconvolution algorithms (TIMER, CIBERSORT/CIBERSORT-ABS, quanTIseq, xCell, EPIC, and MCP-counter). For each TCGA cancer type with available data, we queried the Immune Association/Gene module using the following biomarker panel: CSPG4, PDGFRB, MCAM, RGS5 and CD248.

For each gene–cancer–method combination, TIMER3 reports partial Spearman correlations between gene expression and inferred immune populations (e.g., macrophages, CD8 + T cells, Treg), adjusting for tumour purity. We extracted the correlation coefficient and p-value as returned by the platform. Heat-maps were constructed from these partial correlations to visualize cross-tumour patterns.

Significance was defined at p < 0.05, p < 0.01, and p < 0.001. When synthesizing results across algorithms within a cancer type, we highlighted associations that were directionally consistent in ≥ 2 methods.

Results

Article examination procedure

The search strategy applied across the three databases yielded a total of 3,557 publications, of which 1,222 were removed as duplicates. Titles, abstracts, and keywords of the remaining studies were then screened according to the predefined inclusion and exclusion criteria (Table 1; PROSPERO ID: CRD420251053567), resulting in 290 articles selected for full-text review. Based on the same criteria, 63 studies were ultimately retained for data extraction and analysis. In addition, the reference lists of these articles were examined to identify further eligible studies not captured by the initial search, yielding one additional publication. Consequently, a total of 64 studies were included in the systematic review. Details of the screening methodology are illustrated in Fig. 1, and the complete list of selected articles is provided in Appendix A.

Risk of bias assessment

Risk of bias assessment for in vitro studies indicated that most exhibited a low risk of bias, as they employed well-defined methodologies and provided comprehensive reporting. However, the domain related to researcher blinding could not be assessed in any study due to insufficient information. Eleven studies were classified as having a moderate risk of bias, while four were categorized as high risk (Fig. 2A and C).

Fig. 2.

Risk of bias assessment for in vitro and in vivo studies. (A) Summary of risk of bias in studies containing in vitro experiments, evaluated using the OHAT tool. (B) Risk of bias in studies involving in vivo animal models, assessed with the SYRCLE tool. Bar plots represent the proportion of studies classified as low (green), moderate/unclear (yellow) or high (red) risk of bias across each methodological domain. (C–D) Individual risk-of-bias assessments for each study (in vitro and in vivo, respectively). Each row represents a single study and each column corresponds to a specific bias domain. Colored circles indicate the level of risk of bias (green = low, yellow = moderate/unclear, red = high). A low risk of bias indicates that the study adequately addressed the methodological domain, a high risk reflects potential sources of bias that may affect the validity of the findings, and a moderate/unclear risk denotes insufficient information or partial adherence to methodological standards

For studies involving animal models, the majority were rated as high risk. In particular, elevated risk was frequently observed in the domains of attrition bias (due to animal loss during experiments) and selective outcome reporting (associated with omission of statistically non-significant or inconsistent results). Nonetheless, a substantial proportion of studies were rated as low risk in other areas, including sequence generation, baseline characteristics, allocation concealment, and random housing (Fig. 2B and D).

Preliminary analyses and in vitro assessment of TAPs

Among the analysed studies, glioblastoma and lung cancer were the most frequently investigated tumour types. For example, glioblastoma cells have been shown to induce a molecular process known as “chaperone-mediated autophagy” in TAPs, promoting their transition from an immunogenic to an immunosuppressive phenotype and thereby supporting tumour growth [20]. Similarly, stimulation of CD11b⁺ bone marrow–derived myeloid cells with TNFSF15 in lung cancer models has been reported to induce their differentiation into TAPs, enhancing the formation and stabilization of tumour-associated vasculature [21].

In contrast, other solid tumours have been less extensively studied. A single investigation in thyroid carcinoma demonstrated that activation of the TGFβ pathway in TAMs promotes TAP activation—marked by NG2 upregulation—and contributes to vascular abnormalities such as deficient pericyte coverage [22].

Likewise, few studies have explored hematologic malignancies. In one report, brain pericytes (PDGFRβ⁺/NG2⁺)—but not endothelial cells, astrocytes, or fibroblasts—were shown to support the long-term proliferation and survival of central nervous system lymphoma cells in vitro through HGF secretion and upregulation of PIN1 expression in malignant cells [23]. Consistently, another study in diffuse large B-cell lymphoma found that inhibition of PDGFβ signaling in pericytes using imatinib induced pericyte apoptosis, impaired pericyte network formation, and suppressed tumour growth in xenograft models [24]. Additionally, malignant T cells from acute lymphoblastic leukaemia have been observed to secrete pro-inflammatory cytokines which induce the secretion of CXCL10 by meningeal pericytes. By this way, these TAPs can foster the migration and meningeal invasion of malignant T lymphocytes [25]. No eligible studies were identified for other malignancies such as cholangiocarcinoma or head and neck carcinoma, highlighting a major gap in the current literature.

In vitro investigations addressing TAPs are limited by the scarcity of commercially available pericyte-like cell lines. Among these, human brain vascular pericytes (HBVPs) are the most frequently employed, particularly in studies involving brain malignancies such as glioblastoma [26]. In this context, glioblastoma cells have been shown to induce a mesenchymal stem cell (MSC)-like phenotype in pericytes (CD90⁺CD44⁺CD29⁺PDGFRβ⁺), accompanied by increased secretion of immunosuppressive mediators including TGFβ, PGE₂, INOS, and HGF. These changes enhance the capacity of TAPs to inhibit the proliferation and intratumoural infiltration of cytotoxic CD8⁺ T cells [26]. Conversely, the secretome of diffuse high-grade glioma cells (containing high concentration of osteopontin) has been observed to induce the transformation of stromal MSCs into TAPs expressing NG2, PDGFRβ and the immune checkpoint B7-H3 [27]. Additional studies have utilized alternative pericyte-like models such as human retinal microvascular pericytes (HRMVPs) in clear cell renal cell carcinoma [28], HUM-iCell-N011 human microvascular pericytes in glioblastoma [29], and human aortic smooth muscle cells (HASMCs) in lung cancer [30]. Other research groups have employed 10T1/2 mesenchymal precursor cells, which can be differentiated into pericyte-like cells following stimulation with recombinant TGFβ, leading to the expression of canonical pericyte markers (PDGFRβ, NG2, RGS5) [31]. In parallel, direct pericyte isolation from murine or human tissues has been achieved through fluorescence-activated cell sorting (FACS) using specific biomarker panels, as demonstrated in studies of glioblastoma and brain metastases from melanoma, lung, and breast cancers [32].

A notable limitation in this research area is the lack of consensus regarding pericyte-specific biomarkers, which contributes to heterogeneity and variability in experimental outcomes across studies. Some investigators have identified NG2, RGS5, and CD248 as TAP-related markers [33], whereas others have used combinations of NG2, αSMA, and/or PDGFRβ [34, 35]. Additionally, another study employed the combination of CD146 (MCAM) and PDGFRβ biomarkers along with their location adjacent to endothelial cells [36]. This study also noted that the use of NG2 may not be recommended to identify brain pericytes as a multitude of other cellular type in brain tissue can also express this marker. Furthermore, pericyte heterogeneity between different tissues and tumour types has been observed at both transcriptional and protein levels. For example, pericytes derived from healthy brain and glioblastoma tissue display moderate to high CD19 expression, while those from healthy lung lack CD19, rendering glioblastoma and normal brain potentially amenable to CD19-targeted immunotherapies (e.g., CAR–natural killer [NK] cells) without off-target pulmonary effects [37].

The high transcriptomic and functional diversity between specific subpopulations of healthy pericytes and TAPs has also been documented. A recent study of melanoma, breast, and lung cancer brain metastases identified two main TAP subtypes, “interferon-like” and “extracellular matrix–secreting, characterized by expression of the inhibitory immune checkpoint CD276 (B7-H3), which suppresses infiltration and activation of cytotoxic and memory CD8⁺ T cells. In contrast, pericytes in the healthy brain were predominantly classified as “transport-like” or “proliferative-like” and lacked CD276 expression [38]. Even more complex, it has been detected TAPs functional heterogeneity depending on glioblastoma regions including tumour bulk, tumour borders or contralateral region (spatial-specific TAPs subclones distribution; with a more immunogenic phenotype in the tumour bulk) [39]. Collectively, these findings underscore the methodological challenges associated with TAP research, particularly the limited availability of standardized pericyte models and the absence of universally accepted molecular markers, both of which hinder reproducibility and cross-study comparisons in in vitro investigations.

Pericyte-immune cell interactions in tumour progression

The reported associations between pericytes and immune cells in cancer were next evaluated. Several studies initially appeared to meet the inclusion criteria but were ultimately excluded because they did not demonstrate a direct or causative relationship between immune cells and TAPs. For example, a study on lung and colon cancers reported strong statistical correlations between alterations in pericyte coverage of tumour vasculature, modulation of the immune landscape, and enhanced immunotherapy efficacy following anti–PD-1 treatment. However, these correlations were observed after administration of an anti-VEGFR2 antibody, a classical anti-angiogenic therapy that primarily targets endothelial cells [40]. Consequently, the observed changes in pericyte coverage and immune cell homeostasis are more likely secondary outcomes of endothelial cell normalization rather than evidence of a direct TAP-immune cell interaction. Notably, this study did not provide any indication of direct TAP targeting by the anti-angiogenic compound.

Among the studies included in this review, TAMs were identified as the primary immune cell population influencing pericyte behaviour in cancer (Fig. 3). For instance, malignant cells derived from melanoma, lung, and breast cancers were shown to polarize TAMs toward an M2 anti-inflammatory phenotype characterized by PDGF-CC secretion, which supported the proliferation and maintenance of αSMA⁺ TAPs [41]. These findings suggest a strong potential for direct interaction between M2-polarized TAMs and TAPs in the formation and stabilization of the tumour perivascular niche. In a similar manner, it has been demonstrated that the pharmacological targeting of TAMs can modulate TAP’s behaviour. A study using murine models of breast and pancreatic cancer described that intratumoural delivery of the cytokine LIGHT promoted activation of the Rho kinase pathway and TGFβ secretion by perivascular TAMs. This alteration enhanced the contractile capacity of immature TAPs, leading to tumour vascular normalization [42]. Likewise, studies in glioblastoma showed that TAM reprogramming through inhibition of CYP4A and CYP4 × 1 by flavonoid compounds induced their transition from an M2- to an M1-like phenotype, thereby promoting TAP’s recruitment to the perivascular niche and contributing to vessel stabilization [43, 44]. Moreover, other reports have indicated that pharmacological modulation of TAMs, either by reducing VEGF secretion [45] or by increasing the release of soluble VEGFR1 acting as a “VEGF trap” [46], decreases activation of the VEGF/VEGFR axis in TAPs, improving their maintenance within the perivascular niche, enhancing pericyte coverage and promoting tumour vascular normalization.

Fig. 3.

Schematic representation of the interactions between tumour cells, TAPs and immune cells within TME. Tumour cells release soluble factors that modulate immune cell recruitment and polarization, leading to the polarization of TAMs. Both TAMs and cancer cells contribute to the remodelling of pericytes into TAPs. In turn, TAPs can regulate the infiltration, behaviour and suppression status of a higher variety of immune cell subsets, thereby shaping the “immunologically cold” landscape within the tumour mass. Main biomarkers related to TAPs (CSPG4, RGS5, MCAM, PDGFRβ, CD248) have also been included in the figure

Conversely, increased CD248 expression in TAPs from renal cell carcinoma has been correlated with advanced tumour stage, poor overall survival, and M2 polarization of TAMs [47]. In melanoma, TAPs conditioned by tumour cells specifically expressed MFG-E8, which promoted melanoma growth in vivo and induced TAMs to adopt an M2 phenotype [48]. Similarly, another study demonstrated that lung cancer cells secrete PDGF-BB, which activates the PDGFRβ/SOX7 signaling axis in TAPs, leading to IL-33 production. TAP-derived IL-33 subsequently recruits and reprograms TAMs toward an M2-like state through ST2 receptor–mediated activation and phosphorylation of MAPK/p38/IκBα, ultimately facilitating tumour metastasis [49].

The close bidirectional relationship between TAPs and TAMs is further supported by studies describing a reciprocal feedback loop between these two cell types. In glioblastoma, M2 TAMs secrete PDGFβ, promoting TAP migration and recruitment, while TAPs release periostin, which in turn drives TAM recruitment and M2 polarization [50]. Notably, it has also been reported that normal brain pericytes express macrophage- and microglia-associated markers such as CD68 and Iba-1, both of which are lost during the chaperone-mediated autophagy process induced by glioblastoma cells [51].

TAPs are also capable of modulating the activity of a broad range of immune cells beyond TAMs, including T lymphocytes. In murine cancer models, specific ablation of pericytes has been shown to inhibit CD8⁺ T-cell infiltration into tumour tissue [52]. Consistently, a recent study reported a positive correlation between the abundance of NG2⁺ pericytes in tumour-associated blood vessels and increased intratumoural infiltration of B cells, CD4⁺, and CD8⁺ T lymphocytes in colorectal cancer models [53]. Supporting these observations, single-cell RNA sequencing analyses in lung cancer have suggested a pivotal role for TAPs in leukocyte chemotaxis and highlighted their potential function as antigen-presenting cells [54].

Several studies in melanoma and glioblastoma have further demonstrated that disruption of TAP antigen-presenting capabilities, driven by tumour-induced downregulation of MHC-I and MHC-II expression, contributes to CD4⁺ and CD8⁺ T-cell anergy and immune evasion [31, 55–57]. In addition, the immunomodulatory capacity of TAPs may depend on their ability to recruit MDSCs into the TME. In PDAC, a distinct pericyte–stem cell subpopulation (CD106⁺CD44⁺Nanog⁺OCT4⁺αSMA⁺PDGFRβ⁺RGS5⁺CD24low), absent in healthy pancreatic tissue, was shown to promote MDSC accumulation, enhance M2 TAMs prevalence, reduce DCs infiltration, and suppress cytotoxic CD8⁺ T-cell activation [58].

Furthermore, differences in immune profiles between primary and metastatic tumours may also be partly mediated by TAPs. In colorectal cancer, TAPs isolated from liver metastases, but not from primary tumours, induced a stress-like phenotype in CD8⁺ T cells and NK cells via secretion of APOE, THBS1, LAMB1, TGFβ2, and FGF2, resulting in immune exhaustion [59]. Consequently, the immune landscape of liver metastases exhibits a more suppressed and dysfunctional phenotype compared with that of the primary tumour, contributing to the enhanced aggressiveness of metastatic disease [59].

Strikingly, we only detected a single study reporting a link between TAPs and neutrophils. In fact, it has been documented the significant contribution of TAPs in the recruitment of neutrophils near to the tumour mass, thereby facilitating the formation of circulating tumour cells-neutrophils clusters that enhance the survival of these malignant cells within the bloodstream and induce metastatic spread of colorectal cancer [60]. The crosstalk between TAPs and other stromal cell populations within the TME represents a critical mechanism underpinning tumour immune evasion. In lung cancer, elevated expression of soluble guanylate cyclase (SGC) in TAPs has been shown to enhance TGFβ1 secretion, thereby promoting the expansion of immunosuppressive CAFs and activating both TGFβ and Notch signaling pathways in tumour-associated endothelial cells [61]. Collectively, these stromal interactions drive the polarization of TAMs from an M1 anti-tumour phenotype toward an M2 immune-suppressive state, ultimately fostering tumour immune tolerance [61].

Moreover, the phenomenon known as “pericyte-to-fibroblast transition” (PFT), whereby TAPs transdifferentiate into CAFs, has been implicated in the suppression of the abscopal effect in murine models of melanoma, breast, and lung cancers. Tumour cells from an irradiated primary lesion can secrete PAI1, which activates TAPs within the TME of a secondary, non-irradiated tumour. This activation leads to the emergence of a CAF subpopulation capable of promoting M2 TAM polarization, enhancing Treg infiltration, and reducing CD8⁺ T-cell infiltration and activation, thereby inducing resistance to ICIs and impairing systemic anti-tumour immune responses [62]. Similarly, it has been suggested that profibrotic TAPs of colorectal cancer could transdifferentiate into myo-CAFs expressing B7-H3, which can correlate with poor prognosis and effector T cells inhibition [63]. Reinforcing the importance of pericyte-like phenotype plasticity, it has been observed that the mutation of EGFR gene in glioma cells can drive their transdifferentiation into TAPs-like cells with a strong activation (phosphorylation) of PDGFRβ, which may enhance the intra-tumour infiltration of macrophages and leukocytes [64]. Particularly, this same work exposed that the mutation of EGFR promoted its constitutive activation/phosphorylation, driving the phosphorylation of BMX and, subsequently, inducing the accumulation of SOX9 protein and enabling the glioma-to-pericyte conversion [64].

Importantly, it has also been highlighted the tumour suppressive roles of healthy brain pericytes which can attenuate glioblastoma growth, invasiveness, aggressiveness and restrict the immunosuppressive TME by inhibiting M2 TAMs polarization [65]. These findings reinforce the tumour-suppressive toward tumour-supportive transition experienced when healthy pericytes become TAPs. According to their immunomodulatory properties, TAPs have been implicated in the reduced efficacy of cancer immunotherapies. Single-cell RNA sequencing studies in lung cancer have identified a statistical association between TAP-specific transcriptional profiles (characterized by MCAM⁺ACTA2⁺PDGFRβ⁺ phenotypes and Notch pathway hyperactivation) and resistance to ICIs, particularly PD-1 blockade, which correlated with poor overall survival [66, 67]. Similarly, RGS5⁺ TAPs in insulinoma and fibrosarcoma murine models were shown to hinder the intratumoural infiltration of CD4⁺ and CD8⁺ T cells, thereby promoting tumour tolerance to cancer vaccines and adoptive T-cell transfer therapies [68].

In prostate cancer, activation of the PDGFRβ signaling pathway in TAPs has been linked to increased infiltration of Tregs and MDSCs, a higher M2/M1 TAM ratio, and reduced presence of CD8⁺ and follicular helper T cells, ultimately contributing to resistance to PD-1 blockade [69]. Likewise, in non-small cell lung cancer and hepatocellular carcinoma, clinical analyses established statistical correlations between TAP-specific gene expression signatures and the infiltration or activation of multiple immune cell subsets, including TAMs, Tregs, T helper cells, dendritic cells, mast cells, and NK cells, as well as with ICI (anti-PD-1, anti-CTLA-4) responsiveness [70]. Other research about esophageal squamous cell carcinoma exposed the existence of a subpopulation of TAPs (secreting EGFL6) with the potential to enhance cancer invasiveness, lung and lymph nodes metastases and exhibiting immunosuppressive roles through the promotion of Tregs infiltration, T cells exhaustion and resistance to anti-PD1 therapy [71]. Furthermore, recent studies in melanoma, pancreatic, and breast cancers have demonstrated that pharmacological activation of ROCK1 in TAPs using Combretastatin or Eribulin induces a transition from a proliferative-like to a mature/quiescent phenotype [72]. This switch was associated with enhanced infiltration and activation of CD4⁺, CD8⁺, and tissue-resident memory T cells, an increased M1/M2 TAM ratio, reduced Treg infiltration, and improved efficacy of PD-1 blockade and adoptive T-cell transfer. Although these effects were initially attributed to vascular normalization and improved perfusion, the intrinsic immunoregulatory functions of TAPs described in the present study suggest that TAP phenotypic plasticity itself may directly modulate these immune responses.

A schematic overview of the complex interactions between TAPs and immune cells within the TME is shown in Fig. 3.

Cancer immunotherapy targeting TAPs

According to the current systematic analysis, several studies have preclinically evaluated the antitumour efficacy of immunotherapy-based strategies targeting TAPs. Of the 64 studies included in this review, 11 met the criteria for this section but some other publications identified during articles screening were excluded. For example, a study investigating an oncolytic virus targeting FAPα-expressing TAPs in glioblastoma was not incorporated into the final dataset because, although oncolytic virotherapy can be considered a form of immunotherapy, it did not provide evidence of immune modulation or immune-TAP interaction [73].

Among the studies included in this analysis, most of them focused on cancer vaccine development, demonstrating significant antitumour and immunomodulatory effects. For instance, the combined use of two cancer vaccines targeting the TAP-associated molecules DLK1 and DLK2 resulted in marked tumour growth inhibition in melanoma and renal cancer models [74]. This dual-target approach also promoted tumour vascular normalization, enhanced intratumoural infiltration of activated CD8⁺ T lymphocytes, and reduced the accumulation of MDSCs and Tregs, collectively contributing to an improved antitumour immune response.

In this context, several vaccine formulations have demonstrated promising preclinical efficacy. Notably, Listeria monocytogenes (LM)-based vaccines targeting the pericyte-associated antigen RGS5 have been shown to elicit a robust Th1-type immune response directed against TAPs, resulting in a significant reduction of colorectal tumour growth under both prophylactic and therapeutic regimens [75]. Comparable prophylactic and therapeutic benefits were also observed with an LM-based vaccine targeting the high molecular weight melanoma-associated antigen (HMW-MAA), the human homolog of the murine pericyte marker NG2, in murine models of melanoma, renal, and breast cancers [76].

In renal cell carcinoma, a combined approach using both lentiviral gene-based and DCs-based vaccines engineered to express peptides derived from the TAP-associated antigen DLK1 achieved notable tumour growth inhibition and vascular normalization [77]. Similarly, a strategy employing DCs pulsed with synthetic peptides corresponding to multiple TAP-related antigens, including DLK1, RGS5, PDGFRβ, HBβ, and TEM1 (CD248/endosialin), demonstrated dual prophylactic and therapeutic efficacy. Specifically, this approach reduced the incidence of primary melanoma and colon cancer, while also inducing tumour regression, metastasis prevention, and prolonged overall survival in murine models [78]. Consistently, an independent study in melanoma corroborated these findings, showing that DCs-based vaccines targeting HBβ, PDGFRβ, RGS5, TEM1, and NG2 elicited strong adaptive immune responses against pericyte-associated antigens, culminating in significant tumour reduction [79].

On the other hand, only two studies within the current systematic analysis investigated CAR-engineered immune cell therapies directed against TAPs. One study in breast and lung cancer models demonstrated that second-generation CAR-T cells targeting endosialin (CD248) expressed by TAPs, effectively reduced primary tumour growth, metastatic burden, and number of metastases [80]. Notably, treatment-related adverse events, such as weight loss, cytokine release syndrome, and macrophage activation syndrome, were observed in the same study exclusively in BALB/c mice bearing 4T1 breast cancer xenografts, but not in other strains (C57BL/6 or FVB/N) or tumour types (HRM1 or LLC) [80]. These findings highlight the importance of personalized therapeutic strategies tailored to the specific molecular and immunological context of each tumour and host. Additionally, another study reported that CD19 CAR-expressing NK cells targeting brain TAPs in glioblastoma models significantly suppressed tumour growth [37].

Among the studies included in this section, considerable variability was observed in treatment administration routes, including oral gavage, intradermal, intraperitoneal, intravenous, and intratumoral delivery, as well as in dosage and scheduling, typically weekly administrations for vaccine-based therapies and a single dose for CAR-engineered cell treatments. Notably, all selected studies were designed as short-term evaluations of therapeutic efficacy. For example, the antitumour potential of a PDGFRβ-based DNA vaccine in reducing primary tumour incidence, growth, angiogenesis, and metastatic dissemination in breast, colon, and lung cancer models was assessed over a period of only three weeks [81]. The longest survival follow-up reported among the reviewed studies reached 75 days, documented in a melanoma model, where DC-based vaccines targeting multiple pericyte-associated antigens, including HBβ, DLK1, NG2, PDGFRβ (CD140b), RGS5, and TEM1, resulted in significant tumour reduction and extended overall survival in treated mice [82].

Meta-analysis relying on pericyte-associated markers

The meta-analysis first assessed the expression patterns of pericyte-associated markers identified during the systematic review across multiple tumour types compared with their corresponding healthy tissues (Fig. 4, Supplementary Table S1). Notable variability was observed in the expression profiles of PDGFRβ and CD248. PDGFRβ was downregulated in cervical squamous cell carcinoma, glioblastoma multiforme, lung squamous cell carcinoma, and uterine corpus endometrial carcinoma, whereas it was upregulated in cholangiocarcinoma, colon adenocarcinoma, head and neck squamous cell carcinoma, kidney renal clear cell carcinoma, liver hepatocellular carcinoma, rectum adenocarcinoma, and stomach adenocarcinoma. Similarly, CD248 expression was decreased in breast invasive carcinoma, bladder urothelial carcinoma, cervical squamous cell carcinoma, prostate adenocarcinoma, and uterine corpus endometrial carcinoma, while elevated levels were detected in cholangiocarcinoma, glioblastoma multiforme, head and neck squamous cell carcinoma, kidney renal clear cell carcinoma, liver hepatocellular carcinoma, and stomach adenocarcinoma.

Fig. 4.

Heatmap of differential pericyte-related gene expression. Each cell shows the log2(fold change) in expression for different genes across cancer types, computed as log2(Tumour/Normal) using expression values from UALCAN (TCGA cohorts). Positive (red) indicates over-expression in tumour, negative (blue) indicates under-expression, and values near 0 (white) reflect minimal change. The colour bar indicates magnitude in log2 units

We focused primarily on the five biomarkers most frequently used to characterize pericytes: CSPG4 (also known as NG2), PDGFRβ, MCAM, RGS5, and CD248, excluding ACTA2 due to its frequent expression in CAFs. Notably, several cancer types could be highlighted, including cholangiocarcinoma, head and neck squamous cell carcinoma, kidney renal clear cell carcinoma, and liver hepatocellular carcinoma, due to the concurrent overexpression of at least three of these markers, rather than relying on isolated gene upregulation (given the lack of pericyte-specific biomarkers). These markers exhibited fold changes greater than 1.5 and p-values below 0.05 (Table 3). Based on these findings, it was hypothesized that alterations of TAPs may play pivotal roles in the pathogenesis of these malignancies.

Table 3.

Tumour types showing significant pericyte-associated biomarker upregulation

Tumour type	Parameter	CSPG4	PDGFRb	MCAM	RGS5	CD248
Cholangiocarcinoma	Log2(Fold Change)	3,340897331	2,267321131	3,584987929		3,50311278
Control (n = 9) and Tumor (n = 36)	Statistical Significance (p values)	6,3E-08	5,3E-09	3,4E-03		4,2E-11
Head and neck squamous cell carcinoma	Log2(Fold Change)	2,277337944	2,553126414	1,59844009		1,63969057
Control (n = 44) and Tumour (n = 520)	Statistical Significance (p values)	1,0E-12	2,0E-12	2,4E-05		3,7E-02
Kidney renal clear cell carcinoma	Log2(Fold Change)	2,927271945		2,292471247	2,213044148	1,920093421
Control (n = 72) and Tumour (n = 533)	Statistical Significance (p values)	1,0E-12		1,6E-12	1,6E-12	1,0E-12
Liver hepatocellular carcinoma	Log2(Fold Change)	2,927875621	1,631107945	1,727904325	1,940583877
Control (n = 50) and Tumour (n = 371)	Statistical Significance (p values)	1,0E-12	1,0E-12	1,6E-12	1,6E-12

Next, the TIMER3 online platform was used to evaluate the associations between the five principal pericyte-related biomarkers and the infiltration of key immune cell subsets within tumour tissues, including macrophages, cytotoxic CD8⁺ T cells, and Tregs. Analysis of CSPG4 revealed a positive correlation between its expression and increased infiltration of macrophages/monocytes displaying a pro-tumorigenic M2-like phenotype, as well as immunosuppressive Tregs, particularly in stroma-rich malignancies such as pancreatic adenocarcinoma (PAAD), stomach and oesophageal carcinoma (STAD/ESCA), liver hepatocellular carcinoma (LIHC), cholangiocarcinoma (CHOL), sarcomas (SARC), and head and neck squamous cell carcinoma (HNSC), especially within the HPV-negative subgroup (HNSC-HPV⁻). Conversely, CSPG4 expression was negatively correlated with the infiltration of cytotoxic CD8⁺ T lymphocytes, indicating immune-exclusion signatures. Similar patterns, though with greater variability, were observed in gliomas (GBM/LGG) and melanoma (SKCM), whereas correlations were less consistent or statistically non-significant in lung (LUAD/LUSC) and renal (KIRC/KIRP) cancers (Fig. 5A).

Fig. 5.

Partial-correlation heatmaps showing the association between gene expression and immune infiltration across TCGA cancer types (A–E). Heatmaps correspond to CSPG4 (A), PDGFRβ (B), RGS5 (C), MCAM (D), and CD248 (E). Rows represent TCGA cancer types, and columns represent immune cell signatures estimated using multiple deconvolution methods (EPIC, TIMER, CIBERSORT-ABS, quanTIseq, xCell, MCP-counter, ImmuCellAI, and CONSENSUS_TME), together with cytotoxic T lymphocyte (CTL) scores and TIDE Exclusion scores. Within each cancer type, associations were computed as partial Spearman correlations adjusted for tumour purity (TIMER/ABSOLUTE estimates). Colour intensity indicates the strength and direction of the correlation (red= positive correlation; blue= negative correlation). Crossed cells () denote non-significant associations (p > 0.05), whereas uncrossed cells () indicate statistical significance (p ≤ 0.05, two-sided)

The strongest and most consistent correlations were observed for PDGFRβ (Fig. 5B) and RGS5 (Fig. 5C). Both markers exhibited a clear association between their overexpression and increased infiltration of M2-like macrophages and Tregs, accompanied by the exclusion of cytotoxic CD8⁺ T lymphocytes. These patterns were particularly evident across several stroma-rich malignancies, including PAAD, CHOL, HNSC-HPV-, STAD/ESCA, LIHC, SARC, and GBM/LGG. Greater variability was detected in LUAD/LUSC, KIRC/KIRP, and SKCM datasets, likely reflecting histological heterogeneity and differences in stromal and vascular architecture.

A comparable immunosuppressive profile, characterized by elevated M2 macrophage infiltration and reduced CD8⁺ T-cell presence, was also associated with MCAM expression. Moreover, increased Treg infiltration was frequently observed concomitant with cytotoxic T-cell exclusion (Fig. 5D), particularly in stroma-rich tumours such as PAAD, STAD/ESCA, LIHC, CHOL, SARC, and HNSC-HPV⁻, although with a higher proportion of non-significant associations. Again, pronounced heterogeneity was evident in LUAD/LUSC and SKCM.

Results for CD248 are presented in Fig. 5E, showing a general positive correlation between its upregulation and infiltration of pro-tumorigenic macrophages, reflected by an increased M2/M1 ratio, particularly in PAAD, STAD/ESCA, LIHC, CHOL, SARC, GBM/LGG, and HNSC-HPV⁻. A similar trend was observed for Treg infiltration, especially in stroma-rich cancers. In contrast, CD248 expression correlated negatively with effector CD8⁺ T-cell infiltration, further reinforcing the immunosuppressive potential of this biomarker.

Overall, the immunosuppressive role of TAPs identified through our systematic review is further supported by the present meta-analysis, which incorporated a larger sample size. The analysis revealed strong correlations between the overexpression of pericyte-associated markers and the increased infiltration of immunosuppressive M2-like macrophages and Tregs, accompanied by exclusion patterns of cytotoxic CD8⁺ T lymphocytes. These findings appear particularly relevant for head and neck squamous cell carcinoma, liver hepatocellular carcinoma, and cholangiocarcinoma, where upregulation of key pericyte-related biomarkers was detected and coincided with the development of an immunosuppressive TME.

Discussion

This study primarily explored the interplay between pericytes and immune cells in cancer, emphasizing their immunosuppressive potential. Such interactions have been scarcely documented in liquid malignancies and in certain solid tumours, including thyroid carcinoma or head and neck cancer. Further research into these tumour types may be valuable, considering the role of immunosuppressive TAMs in the progression of differentiated thyroid carcinoma into the highly aggressive anaplastic subtype [83]. Similarly, TAM polarization and T-cell inhibition have been shown to contribute critically to head and neck squamous cell carcinoma progression and metastasis [84].

Unexpectedly, none of the included studies investigated pericytes residing in blood vessels of lymph nodes adjacent to the tumour mass. Exploring this aspect could yield important insights, given the relevance of lymph node immunosuppression, particularly CD4 + T-cell exhaustion, during lymph node metastasis of oral squamous cell carcinoma [85]. Indeed, employing malignant cell lines belonging to lymph nodes metastases, like CCL227 of colon cancer [86], or lymph nodes metastasis murine models [87] may represent valuable tools to assess the immunosuppressive roles of TAPs in in vitro or in vivo-based future studies, respectively. Additionally, blood vessels within lymph nodes invaded by melanoma cells exhibit distinct pericyte subpopulations, such as CD146 + CD36+ and CD146 + NG2+ cells, suggesting potential functional divergence of TAPs in modulating immune cell behaviour [88]. Importantly, it is necessary to note that endothelial cells from lymphatic vessels are also surrounded by pericytes, which can play key roles in the regulation of lymphatic microcirculation [89]. Therefore, it seems reasonable to suggest that this fact could significantly contribute to TAPs heterogeneity within lymph nodes, alluding to the relevance of stablishing reliable biomarkers to differentiate between blood and lymphatic vessels TAPs. Furthermore, specifying from which lymph node TAPs are isolated may be critical to mitigate discrepancies between studies considering TAPs heterogeneity and the preference of each cancer type to metastasize in lymph nodes belonging to particular anatomical regions, for instance mediastinal and hilar lymph nodes in lung cancer [90].

Marked TAP heterogeneity was observed across tumour types in both phenotype and biomarker expression. Supporting this, in silico analyses have identified two principal TAP subpopulations across several malignancies, including PDAC, renal clear cell carcinoma, colorectal, hepatocellular, lung and ovarian cancers: (i) “fibrogenic” TAPs associated with extracellular matrix deposition and (ii) “vascular” TAPs involved in tumour angiogenesis [91]. This diversity parallels that described for endothelial cells in prior reviews on endothelial-to-mesenchymal transition (EndMT) in solid tumours [92]. As noted for endothelial cells in such previous article, careful selection of appropriate pericyte-like cell lines may also be crucial to minimize experimental variability. Given the intimate functional/physical relationship between pericytes and endothelial cells, their heterogeneity likely influences each other reciprocally.

The in vitro assessment of TAPs presents additional challenges compared to endothelial cells, particularly the limited availability of commercial pericyte-like cell lines. Although HBVPs are widely used in cancer research [93], their derivation from a single tissue (brain) may not represent the organ-specific diversity of pericytes seen under physiological conditions [94]. Expanding the repertoire of human pericyte-like cell lines from distinct tissues is therefore needed. Alternatively, direct isolation of pericytes from human or murine tissues may be feasible, although human-derived TAPs would provide higher translational value given species-specific transcriptional differences [95]. However, the lack of exclusive pericyte biomarkers complicates their isolation, often requiring co-expression of multiple markers and localization adjacent to endothelial cells [96]. Moreover, other authors have emphasized the urgent need for more reproducible and high-yield protocols for pericyte isolation [97]. In this regard, a recent study described a protocol for isolating and culturing primary murine pericytes from various tissues, including bone, brain, lung, and liver, highlighting the critical importance of maintaining physiological oxygen tension during pericyte ex vivo culture [98].

Among immune populations, TAMs emerge as the primary regulators of TAPs during tumour progression. Interestingly, healthy pericytes share phenotypic traits with macrophages, including phagocytic capabilities [99]. Glioblastoma cells can induce TAPs to switch from a pro-inflammatory to an immunosuppressive phenotype [56], a process reminiscent of TAM polarization from M1 to M2 states observed in several tumours such as melanoma [100]. Whether a certain subpopulation of TAPs may transdifferentiate into TAMs remains an intriguing hypothesis, supported by their stem-like properties and capacity to differentiate into mesenchymal lineages and, subsequently, other multiple cell lines [101]. Indeed, PDGF-BB stimulation can endow TAPs with fibroblast-like traits, enhancing tumour growth and metastasis [102].

Conversely, our review revealed that TAPs influence a broader spectrum of immune cells, including TAMs, NK cells, CD4 + and CD8 + T cells, tissue-resident memory T cells, Tregs, DCs and MDSCs, than the reverse. Beyond cancer, pericytes and neutrophils can modulate each other, as formation of neutrophil extracellular traps (NETs) has been shown to induce CD11b+ pericytes in brain injury models [103], while perivascular pericytes expressing ICAM1 facilitate neutrophil transmigration during inflammation [104]. Investigating such interactions in cancer may be informative, considering the role of neutrophil polarization in the progression and therapeutic resistance of hepatocellular [105], pancreatic [106], and breast cancers [107]. Interestingly, tumour-associated neutrophils tend to exhibit a similar phenotypic plasticity, with a pro-inflammatory, tumour-suppressive N1-like phenotype and an anti-inflammatory, immunosuppressive and tumour-supportive N2-like one [108]. Moreover, it has been reported the reciprocal communication between NETs and M2 TAMs in hepatocellular carcinoma to promote tumour progression and aggressiveness [109]. Considering the close relationship between TAPs and M2 TAMs, it seems tempting to experimentally explore in future studies whether TAPs are able to directly or indirectly induce the N2 polarization in tumour-associated neutrophils and, thus, enhance their immunosuppressive behaviour. Similarly, it would be interesting to check whether TAPs can directly or indirectly induce the formation of NETs and, thereby, enhance tumour progression.

This study also reviewed preclinical immunotherapeutic strategies targeting TAPs. Cancer vaccines were the most common approach, aiming to elicit adaptive immune responses against TAP-associated antigens. However, TAPs’ intrinsic immunosuppressive activity may limit the long-term success of such therapies in the clinic. Combining TAP-targeted vaccines with immunomodulators, such as immune checkpoint agonists or antagonists (e.g., antibodies against CD40, OX40, or CD127), could potentiate host immune activation [110]. Thus, it seems reasonable to propose a potential synergistic effect between vaccines targeting TAPs and conventional ICIs, as previously noted, given that TAP eradication or functional reprogramming appears to enhance ICI efficacy.

Most of the selected studies used short-term follow-up protocols, with the longest extending to 75 days. Such limited monitoring likely underestimates resistance mechanisms arising from own TAPs’ immunosuppressive functions. Longer-term studies, as demonstrated by extended (65-weeks) vaccine monitoring models, may provide more reliable preclinical insights [111], thus facilitating the clinical translation of this therapeutic strategy.

CAR-engineered cell therapies have also shown promising results, though remain scarcely explored. Development of CAR-macrophages targeting TAPs may represent a promising avenue, considering the intimate cross-talk between both cell types. Indeed, CAR-macrophages directed against FAPα effectively reduced fibrosis in PDAC, enhancing both chemo- and immunotherapy responses [112]. Moreover, engineering CAR cells to resist TAP-mediated immunosuppression could further strengthen their efficacy.

The selection of TAP-targeting biomarkers must be carefully optimized to minimize off-target effects in healthy tissues and to potentiate the translation towards the clinical practice of immunotherapies against TAPs. In this sense, it has been described that TAPs are the main cell type within glioblastoma expressing FAPα, making it a valuable TAPs-specific biomarker against which immunotherapy-based strategies could be directed [73]. Additionally, endosialin (CD248) has been documented to be almost exclusively expressed by TAPs and perivascular CAFs (but not by cancerous cells, endothelial cells, or healthy pericytes) in several malignancies such as ovarian, breast, and colorectal cancer [80]. Similarly, it has been reported that HMW-MAA antigen is preferentially expressed by TAPs in murine models of melanoma compared to healthy pericytes [76].

Complete pericyte ablation, however, may act as a “double-edged sword”: while reducing TAMs recruitment, it could simultaneously facilitate tumour cell intravasation due to vascular leakage [13]. Consequently, reprogramming or phenotypic modulation of TAPs may be preferable to their eradication, analogous to the therapeutic reprogramming of CAFs [113] and TAMs [114].

On the other hand, although cancer immunotherapy has revolutionized cancer treatment over the past decades, the mentioned immunotherapeutic approaches targeting TAPs face other challenges, which may limit their translational potential. For example, it could be highlighted the cross-reactive toxicities and the complex manufacturing procedures associated with CAR-engineered cell therapies or the general limited efficacy and the variable clinical outcomes (depending on the specific immune system status of each patient along with the particular immune landscape within the tumour mass) related to cancer vaccines [115]. Nevertheless, new promising therapeutic strategies are being explored to overcome such handicaps. Regarding CAR-engineered cells, it has been proposed alternative approaches (e.g., the use of non-viral vectors or employing genetically edited cells from healthy donors to create universal CAR-engineering cell products suitable for treating multiple patients, to mitigate the high manufacturing costs linked to the need for transportation to specialized infrastructures, the requirement of personalized and autologous treatments, the lengthy cell expansion protocols or the reliance on viral vectors requiring advanced laboratories [116]. Moreover, the combination of nanoparticles-based approaches, with the objective of remodelling the immunosuppressive TME by reprogramming TAMs toward an M1 phenotype, with cancer vaccines has been reported to enhance the efficacy of such immunotherapy [117]. Interestingly, this combination of therapeutic strategies could even synergize with specific vaccines targeting TAPs considering the immunosuppressive roles of such cellular type.

Our meta-analysis revealed substantial variability in the expression of pericyte-related markers, particularly PDGFRβ and CD248, across tumour types. Previous computational analyses corroborate this heterogeneity and link their overexpression to poor prognosis and immunosuppression [118, 119]. Notably, our findings highlight the potential immunosuppressive role of TAPs in head and neck, hepatocellular, and cholangiocarcinoma—tumours for which available data are limited. Supporting this, a study in hepatocellular carcinoma identified a TAP-related gene signature (ARHGEF26, GPRIN1, AKAP12, ASPM, and RRA) correlated with disease progression, immune infiltration, and response to immunotherapy [70].

Conclusions

The present study demonstrates the link between TAPs and immune cell regulation, highlighting their immunosuppressive potential and, consequently, their contribution to tumour progression. In particular, we emphasize the close interplay between TAPs and TAMs. Our findings also reveal the limited information available in key areas where further research could provide crucial insights, such as the role of TAPs in specific malignancies like head and neck squamous cell carcinoma, their interaction with neutrophils, their immunomodulatory functions within lymph nodes, and the urgent need to develop a broader range of pericyte-like cell lines for in vitro studies.

Furthermore, the marked heterogeneity of TAPs in terms of biomarkers and functionality, both across tumour types and within individual tumours, suggests the involvement of multiple regulatory factors, such as the intrinsic variability of endothelial cells. This complexity hinders the establishment of reliable pericyte-like cell lines and the precise characterization of TAPs’ immunomodulatory roles in relation to specific immune cell populations.

We also underscore the therapeutic potential of targeting TAPs through various immunotherapy-based approaches, including cancer vaccines and CAR-engineered immune cells, with particular promise in the development of CAR-macrophages directed against TAPs. In the same vein, we highlight the importance of long-term evaluation of TAP-targeted immunotherapies to anticipate and overcome potential drug resistance driven by their immunosuppressive activity.

Finally, our meta-analysis reinforces the immunomodulatory roles of TAPs observed in our systematic review. Notably, several key pericyte-associated biomarkers were found to be significantly upregulated in multiple malignancies, including head and neck squamous cell carcinoma, liver cancer, and cholangiocarcinoma. This finding is particularly relevant given the strong correlation between these markers and the infiltration of immunosuppressive cells such as M2 macrophages and regulatory T cells, alongside the exclusion of immunologically active T cells. Altogether, these results underscore the significance of TAPs in cancer immune biology and their promise as novel targets for therapeutic intervention.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Abbreviations

CAFs

Cancer-associated Fibroblasts

CAR

Chimeric Antigen Receptor

CHOL

Cholangiocarcinoma

DCs

Dendritic Cells

EndMT

Endothelial-to-mesenchymal Transition

GBM

Glioma

HASMCs

Human Aortic Smooth Muscle Cells

HBβ

Haemoglobinβ

HBVPs

Human Brain Vascular Pericytes

HMW-MAA

High Molecular Weight Melanoma-associated Antigen

HNSC

Head and Neck Squamous Cell Carcinoma

HRMVPs

Human Retinal Microvascular Pericytes

ICIs

Immune Checkpoint Inhibitors

LIHC

Liver Hepatocellular Carcinoma

LM

Listeria Monocytogenes

MDSCs

Myeloid-derived Suppressor Cells

MSCs

Mesenchymal-stem Cells

NETs

Neutrophil Extracellular Traps

NK

Natural Killer

OHAT

Office of Health Assessment and Translation

PC

Pancreatic Cancer

PDAC

Pancreatic Ductal Adenocarcinoma

PFT

Pericyte-to-fibroblast Transition

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

SARC

Sarcoma

SGC

Soluble Guanylate Cyclase

STAD/ESCA

Stomach and Oesophageal Carcinoma

SYRCLE

Systematic Review Center for Laboratory Animal Experimentation

TAMs

Tumour-associated Macrophages

TAPs

Tumour-associated Pericytes

TME

Tumour Microenvironment

Tregs

Regulatory T Cells

Author contributions

PH: conceptualization of the work, design of tables and figures, writing and original draft preparation, article screening, data extraction, risk of bias; BT: article screening, data extraction, creation of figures and tables, writing, risk of bias; AB-D: meta-analysis, creation of figures, writing; MP-R: Data extraction; AG-T: creation of figures. MP: Data extraction, article review, article editing; RM: article review, editing; JAM: article review, editing. All authors contributed to article review.

Funding

This work was supported by grant QUAL21-11 funded by Consejería de Universidad, Investigación e Innovación, Junta de Andalucía, by grant from Jaen University, Spain (“Plan de Apoyo a la Investigación, Desarrollo Tecnológico e Innovación. IV Programa de Ayudas a los Investigadores”) (MP); by Pancreatic Cancer Europe (PCE) Organization “2025 Early Career Investigator (ECI) Short-term Scientific Stay Award (S3A)” and by COST Action “Identification of biological markers for prevention and translational medicine in pancreatic cancer (TRANSPAN),” CA21116, supported by COST (European Cooperation in Science andTechnology) (BT); by grant PID2022-140151OB-C22 funded by MICIU/AEI /10.13039/501100011033, by the Chair “Doctors Galera-Requena in cancer stem cell research” (CMC-CTS963) (JAM); by grants from the Instituto de Salud Carlos III and co-funded by the European Union (PI24/01797, M.P-R), by the State Research Agency from the Spanish Ministry of Science, Innovation and Universities (funded by MICIU/AEI/10.13039/501100011033; FEDER, ERDF, EU; and FSE) (RYC2020-030808-I, M.P-R).

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors have declared that no competing interests exist.