Peritoneal surface malignancies

Peritoneal Surface Malignancies (PSMs) were for a long time considered diseases with limited therapeutic options and poor prognosis. Over the last two decades, the combination of perioperative systemic chemotherapy with cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC) has dramatically improved outcomes [1, 2]. The overall prognosis of PSM is today nearly equivalent to that of patients with metastatic disease at other sites.

The majority of PSMs represent peritoneal metastases that have spread secondarily from a distant primary tumour, and only 3% originate from the peritoneum itself [3]. About 22% derive from colorectal cancer (including appendix), 16% from ovarian cancer, 13% from gastric cancer, 7% from pancreatic cancer, and 9% have an extra-abdominal primary malignancy (mainly breast and lung cancer) [3]. In about one of four patients with peritoneal metastases there are no other clinically apparent sites of metastatic spread [4]. Different types of tumours metastasise to the peritoneum at different rates. The reported relative incidence of peritoneal metastases from ovarian cancer is 60–70%, whereas it is less than 10% for other gynaecological malignancies [5]. Of the gastrointestinal cancers, 43% of patients with gastric cancer and 25% of patients with recurrent colorectal cancer develop metastases that are confined to the peritoneal surface [6, 7]. Pathologic features associated with an increased frequency of peritoneal disease in colorectal cancer are mucinous histology, poor differentiation, pT4 status, nodal metastases and a consensus molecular subtype 4 [4, 8]. The consensus molecular subtype (CMS) is a gene expression-based classification of colorectal cancer, which has been introduced in 2015 [9]. It has contributed to a better understanding of disease heterogeneity and prognosis. CMS4 is the mesenchymal type of colorectal cancer, which is characterised by the activation of transforming growth factor-β, stromal invasion, angiogenesis, and fibrosis. It has been recently identified as the predominant subtype in colorectal cancer-derived peritoneal metastasis [8]. Even though appendiceal cancer is anatomically closely linked to the colon, it shows a different gene expression pattern than colorectal cancer [10]. It metastasises more frequently to the peritoneum, resulting mainly in low-grade appendiceal mucinous neoplasm (LAMN), mucinous appendiceal adenocarcinoma (MACA), and mucinous appendiceal adenocarcinoma of intermediate subtype (MACA-Int) [11]. This suggests that the probability of an individual tumour type forming peritoneal metastases depends very much on the primary site of cancer as well as on its gene expression profile.

Although the primary tumours differ considerably in respect to their cellular and genetic background, there are strong similarities in the progression of PSMs resulting from the challenges a tumour cell must overcome to survive and grow in the peritoneal cavity (Fig. 1). Tumour cells must detach from their primary tumour, evade anoikis and gain motility. Once a viable, free cancer cell is floating in the peritoneal cavity, adherence to the peritoneal surface is required to ultimately invade the peritoneum, and form metastases. In order to survive and disseminate better in the peritoneal cavity, metastatic tumours form multicellular clusters. They consist of tumorous as well as non-tumorous cells which aggregate in the peritoneal fluid in a fibrin-mediated manner [12]. Fibroblasts have been frequently found in such ascitic multicellular aggregates in gastric cancer [13], colorectal cancer [14], and ovarian cancer [15]. High-grade serous ovarian cancer cells in the peritoneal fluid are characterised by high expression of integrin α5 which promotes the recruitment of fibroblasts [16]. Furthermore, ovarian cancer cells stimulate their proliferation via transforming growth factor-β1 (TGF-β1) [17]. TGF-β1 was also found in tumour cell-derived exosomes of the malignant ascites which are supposed to additionally enhance peritoneal fibrosis [18]. Fibroblasts which have been recruited, in turn, facilitate the aggregation and compaction of the multicellular clusters by producing collagens and other extracellular matrix (ECM) proteins [17]. They support tumour cell survival in the peritoneal fluid by secreting epidermal growth factor (EGF) and hepatocyte growth factor (HGF) [16, 17]. Moreover, they promote adhesion and invasion of tumour cells by producing matrix metalloproteinase-2 (MMP-2), before becoming essential components of the tumour stroma in newly-formed metastases [16, 17]. Ascitic multicellular aggregates may also contain neutrophils which then form neutrophil extracellular traps (NETs) [19]. Intraperitoneal administration of DNase I in mice showed an 88% decrease in the number of peritoneal metastases confirming that NETs promote survival and adhesion of floating tumour cell aggregates [20].

Fig. 1: The sequence of events leading to development and establishment of peritoneal carcinomatosis.
figure 1

(1) When the primary tumour overgrows and invades the visceral serosa, cancer cells can be shed into the peritoneal fluid and disseminate across the peritoneal cavity; (2) Tumour cells can enhance their survival by aggregating into multicellular clusters; (3) The increased survival allows the recruitment of immune cells (such as neutrophils) and fibroblast resulting in the formation of heterotypic clusters. (4) Those clusters also display an increased ability to attach to the peritoneum surface (mesothelial cells) and invade into the submesothelial space establishing a peritoneal metastatic tumour (5). Once tumour cell have invaded submesothelial stroma, they must modulate the peritoneal tissue in order to establish a supportive TME. As a consequence, resident fibroblast and other stromal cells are recruited, activated and converted into CAFs (6). The heterogeneity of cancer-associated fibroblasts is represented by the different colours attributed to CAFs.

Although tumour cell aggregates may attach to any intraperitoneal surface, the greater omentum appears to be the preferential site of initial attachment and tumour cells then seed into the rest of the peritoneal cavity [21]. The omentum represents a sheath overlying the abdominal organs which is composed of vascularises adipose tissue embedded with clusters of immune cells termed “milky spots”. Several authors have studied the metastatic pattern of cancer cells after intraperitoneal inoculation in animals and found that these cells preferentially accumulate in the milky spots of the greater omentum as well as the lymphatic lacunae of the diaphragm [21]. This binding may be mediated by the expression of vascular cell adhesion molecule-1 (VCAM-1) and collagen I fibres on the surface of these areas, and of the proangiogenic vascular endothelial growth factor receptor 3 (VEGFR3) by adjacent omental micro vessels [22]. Under healthy conditions, the peritoneum provides a slippery, non-adhesive surface through the microvilli of mesothelial cells, which produce large amounts of proteoglycans and hyaluronic acid, forming the glycocalyx. Ovarian cancer cells express the homing cell adhesion molecule CD44 on their surface, which interacts with hyaluronic acid and mediates intraperitoneal adhesion [23]. CD44 was also found on ovarian cancer-derived exosomes, which transfer this molecule to mesothelial cells, thereby promoting adhesion [24]. It has been shown in colorectal cancer that HGF expressed by associated fibroblasts stimulated tumour adhesion through up-regulation of CD44 via c-Met signalling pathway [25]. Another way in which tumour cell clusters can adhere to the mesothelium is via the tumour antigen CA-125, which is expressed by most tumours that metastasise to the peritoneum. CA-125 binds with high efficiency to mesothelin, a surface protein expressed on normal mesothelial cells lining the peritoneum [26]. This led to the development of mesothelin-targeting agents such as amatuximab and anetumab ravtansine which have been tested in different PSMs in phase I and II clinical trials [27, 28]. Furthermore, intraperitoneal adhesions can also occur as a result of healing processes in the peritoneum. Such processes can be induced by injuries in the course of a surgical procedure, but also as a reaction to the destruction of tissue by tumour growth. Both processes induce the secretion of pro-fibrotic TGF-β, which was found to correlate with the peritoneal adhesion index [29, 30]. Accordingly, high-grade PSMs are usually embedded in fibroblast-rich scar tissues. As both heat and chemotherapy only penetrate a few cell layers deep, such fibrotic PSMs are difficult to target during HIPEC [2]. In summary, it can be concluded that the path from tumour cells spreading into the peritoneal cavity to established surface metastasis is a multi-step process that involves the interplay with non-tumour cells. Fibroblasts appear to play an essential role in this progression.

Fibroblasts in peritoneal surface malignancies

Already more than 35 years ago, Harold Dorak termed solid tumours as ‘wounds that do not heal’. This is because solid tumours induce a chronic activation of wound healing mechanisms [31]. The destruction of tissue architecture because of persistent growth and invasiveness of tumour cells results in the constant secretion of damage-associated molecular patterns (DAMPs). Furthermore, most cancers show a considerable degree of constant spontaneous cell death, leading to further release of DAMPs [32]. Fibroblasts express pattern recognition receptors (PRRs), such as Toll-like receptors 2 and 4, which sense DAMPs in their surrounding [33]. They migrate to the site of tumour-induced tissue damage, where they are irreversibly activated, and expand to restore the local tissue structures and function [34]. It is well established that these cancer-associated fibroblasts (CAFs) contribute to tumour development, progression and therapeutic resistance through diverse mechanisms. They secrete soluble molecules such as growth factors and chemokines that promote tumour cell growth and inhibit inflammation. Additionally, CAFs provide metabolic substrates for tumour cells, secrete ECM proteins and remodelling enzymes that promote tumour invasion and drive further tissue fibrosis. The different factors and molecular mechanisms involved have already been described in excellent reviews and are, therefore, not discussed here [33, 35,36,37]. However, it must be pointed out that these review articles describe the significance of CAFs in tumour biology in general. Unfortunately, the special conditions in PSMs are not described in detail, even though CAFs make such an important contribution to PSMs. Most PSMs are characterised by high infiltration of CAFs expressing fibroblast activation protein (FAP), and this is frequently associated with massive fibrosis and scarring [2, 34]. This characteristic can even be used to detect peritoneal metastases by PET/CT scans with [68Ga]-labelled FAPI radiotracers [38]. A meta-analysis revealed that 68Ga-FAPI PET/CT has a much better sensitivity for the detection of peritoneal metastases than the conventional FDG PET/CT [38].

There are no clearly-defined specific molecular markers to identify CAFs. FAP, α-smooth muscle actin (α-SMA), and platelet-derived growth factor receptor alpha (PDGFRα) are the most commonly used, although their expression is not mandatory [39]. The application of single cell RNA sequencing (scRNA-seq) technology on tumour samples revealed a surprising heterogeneity of CAF functional annotation, even within the same tumour. Dor Lavie and co-authors summarised the results of numerous scRNA-seq studies in a recent review [40]. They grouped the different types of CAFs into three major subsets: myofibroblastic CAFs (myCAFs), immune-regulatory CAFs (iCAFs), and antigen presenting CAFs (apCAFs). myCAFs, also termed ECM-remodelling or wound-healing CAFs, were identified by elevated expression of α-SMA (gene name ACTA2) and show contractile features. iCAFs facilitate tumour immune escape. They secrete cytokines and chemokines, recruit suppressive myeloid and regulatory T cells, exclude cytotoxic lymphocytes and dendritic cells and promote M2 macrophages and type 2 helper T cells. apCAFs express MHC class II-related genes such as CD74, HLA-DRA and HLA-DRB1. It has to be noted that the chosen subclustering into these functional types represents a summary and that individual scRNA-seq studies have made further or different classification. For example, a pan-cancer study by Ma et al. [41], which included primary tumours from liver, colorectal, prostate, breast, ovarian and endometrial origin, appointed five different subtypes: iCAFs, mCAFs (matrix CAFs), meCAFs (metabolic CAFs), pCAFs (proliferating CAFs) and fibroblasts with pericyte/muscle cell properties. In this study, iCAFs and meCAFs are in accordance with Lavie et al. However, the myCAFs defined in Lavie et al. are divided into a) mCAFs (showing ECM remodelling properties via expression of MMP11, CTHRC1 and specific collagens) and b) “real” myCAFs that have contractile properties, expressing mainly ACTA2, MYH11 and RGS5, also resembling pericytes, as mentioned elsewhere [42, 43]. This comparison illustrates the inconsistency of the current literature. There is no consensus on a nomenclature for the different functional subgroups of CAFs. The markers used in the individual studies must be carefully examined when comparing the results. Unfortunately, the number of available PSM-related scRNA-seq data sets is rather limited. At present, only scRNAseq data on ovarian cancer and gastric cancer-derived PSMs are available. Table 1 lists all accessible data sets and indicates the different fibroblast functional subtypes described per study. For further reference, differentially expressed genes between the different reported fibroblast subgroups are shown in Supplementary Table 1. These lists of differential markers are dependent on the context of each study.

Table 1 Fibroblast subtypes in peritoneal surface malignancies.

Ovarian cancer-derived PSMs

In a study on solid ovarian cancer samples, peritoneal metastasis-associated fibroblasts and fibroblasts from primary tumour expressed increased levels of collagen and ECM-remodelling proteins (MMP2/11), when compared to fibroblasts from normal adjacent tissue [44]. However, when fibroblasts from the primary cancer were compared to those from peritoneal metastases, the latter expressed a significant number of markers associated with iCAFs. This is in accordance with another study on high-grade serous ovarian cancer (HGSOC), which also found a prevalence of iCAFs in ascites, expressing IL-6, C1Q and CXCL1/2/10/12, among other markers [45]. A second CAF type in this study was defined as non-iCAFs. Carvalho et al. used the same scRNAseq cohort and confirmed the previous findings, additionally annotating the non-iCAF functional type as myCAFs [46]. In a study from Loret et al. on ascites and solid peritoneal metastases from HGSOC, four different CAF subtypes were identified (myCAF, iCAF, mCAF, STAR+ CAF). iCAFs were the prevalent type in ascites, whereas they were second prevalent in solid metastases, preceded by mCAFs. Interestingly, iCAFs were absent from primary tumours in treatment-naïve patients. After carboplatin-paclitaxel treatment, mCAFs and myCAFs showed increased expression of CXCL12 and IL-6, respectively, thus acquiring more iCAF-related properties [47]. Expression of these ligands could alter their communication with other cell types in the TME in an unfavourable manner [48, 49]. Zhang and co-workers divided HGSOC PSM samples based on the expression of a stress-response signature in stress-high and stress-low cancers [50]. Of note, stress-high cancers were related to poor prognosis, and only iCAFs were significantly enriched in stress-high samples. iCAF-like cells were also detected in another HGSOC study [51] which used the subtyping system of Qian et al. to assign CAF subtypes [43]. Notably, a subtype that was detected only in normal adjacent and metastatic omentum (FB_CALB2) was described to have mesothelial origin and expression of iCAF-related genes, including IL-6, IL-18, COL8A1, CXCL16, CCL2, CXCL1 and IL6ST. Other CAF types related to poor overall survival were: (a) mCAF-like cells (defined as FB_COMP) which were present equally in ovarian primary tumours, omental metastases and peritoneal metastases; and b) myCAF-like cells (defined as FB_MYH11), which were mostly present in normal ovarian tissue. Deng et al. identified four different types of fibroblasts in solid HGSOC samples: mCAF, STAR+, apCAF and TNF-related. However, STAR+, and TNF CAFs showed expression of immune-related factors and, combined with the de facto immunomodulatory apCAFs, formed a “supergroup” of immunoregulatory, iCAF-like cells. This group was the most prevalent in peritoneal metastasis [52]. Only few studies demonstrated subtypes other than iCAFs as the main subtype in ovarian cancer-related PSMs. PSM solid samples HGSOC and Low-grade SOC (LGSOC) [53] showed mainly a CAF type related to angiogenesis and evasion of apoptosis, related to vascular CAFs (vCAFs) and pericytes (defined here [54]), whereas primary tumours from the same study mostly contained matrix CAFs (mCAFs) (defined here [40]). mCAFs were also the main subtype identified in ascites of five HGSOC patients [55], showing proEMT properties.

Gastric cancer-derived PSMs

In a study on multiple metastasis locations from gastric adenocarcinoma, iCAFs were the main CAF type in PSMs, and the CXCL12-CXCR4 module was the main route of communication of iCAFs with several target cell types, including T cells, B cells, endothelial, dendritic cells, neutrophils and macrophages [56]. On the other hand, mCAF fibroblasts were the most prevalent in gastric cancer ascites samples [57]. Presence of the mCAF signature showed increased risk of metastasis and/or recurrence.

In summary, despite all the differences between the individual studies, there is a clear prevalence of three functional CAF subtypes in PSM: mCAFs, myCAFs, and iCAFs (Fig. 2). Notably, iCAFs or fibroblasts presenting iCAF-related traits (iCAF-like) are the dominant fibroblast subtype in the majority of studies of ovarian cancer-derived PSMs. In addition, in a subset of studies, iCAFs were exclusively present in the PSM tumour microenvironment, and not in primary samples [44, 47, 50]. This finding could suggest an enrichment of the iCAF functional subtype in PSM, possibly related to immunoregulatory functions and adverse prognosis. In some studies, the prevalent peritoneal MAF subsets are not iCAFs, but rather myCAFs or mCAFs. This discrepancy could be eventually related to different clinical parameters, including disease subtype, tumour stages etc. The total number of patients in current studies is insufficient to correlate such variations with clinically relevant variables, highlighting the need for new, extensive cohorts with more patients in the scRNA-seq analysis of fibroblasts in ovarian cancer-derived PSMs. This limitation is even more evident in the other PSM types. Only two studies provide information about peritoneal MAFs in metastatic gastric cancer. Specifically, in gastric cancer, one study defined iCAFs as the main type, whereas the other defined mCAFs. An important difference between these cohorts is the cancer subtype (intestinal vs diffuse, respectively). Notably, scRNA-seq studies on fibroblasts from colorectal cancer-derived PSMs are absent.

Fig. 2: The most commonly found fibroblast subtypes in peritoneal surface malignancies.
figure 2

In case a specific definition by the authors of the studies was missing, reported markers from each study were evaluated according to markers from Ma et al., Lavie et al. and Cords et al., in order to define the functional subtypes of the reported fibroblast groups. The most frequently reported definition markers per functional subtype among the PSM scRNAseq studies are shown [40, 41, 54]. myCAFs and mCAFs, modulate the extracellular matrix and generate a stiff tumour microenvironment, favouring tumour metastasis and chemoresistance. However, only myCAF express in addition ACTA2 and show a contractile phenotype. iCAF-secreted factors can induce inflammatory pathways, but they have also been reported to attract immunosuppressive regulatory T cells and promote anti-inflammatory polarisation of myeloid cells.

Overall, iCAFs are the main subtype metastasis-associated fibroblasts (MAF) in PSMs, and certain studies claimed an enrichment of the IL-6/JAK/STAT pathway. This is in contrast to primary and metastatic cancers in other locations (pancreas, breast, lung, liver), where myCAFs and mCAFs are the dominant fibroblast types [40]. This difference could reflect the importance of the local microenvironment. Considering the plasticity of fibroblasts, it would be possible that the local microenvironment shapes the MAFs and vice versa. It should be noted that scRNAseq data only provides information about which cell populations are present in a tissue at a certain point in time. However, no conclusions can be drawn as to what led to this cell composition. In order to find out what could have led to the formation of MAFs, one must refer to other studies.

Origin of peritoneal metastasis-associated fibroblasts

The plasticity and heterogeneity of fibroblast phenotypes and functions are supposed to be primarily related to their diverse cellular origins [40]. Using mouse models, it was demonstrated that the majority of CAFs derived from local rather than from circulating precursors [58]. Thus, the source of stromal fibroblasts depends on the specific environment surrounding malignant cells. In PSMs this is the peritoneum or the greater omentum. Structurally, the peritoneum comprises the mesothelium, a basal lamina consisting of ECM, and the submesothelial stroma [38]. The latter include submesothelial fibroblasts, muscle cells, and finally endothelial cells and pericytes of the microvasculature. The omentum represents an adipose tissue with immune cell aggregates in the milky spots. Various cell types of the peritoneum or of the greater omentum have the capacity to serve as a source of peritoneal MAFs (Fig. 3 and Table 2).

Fig. 3: The multi-cellular origins of cancer-associated fibroblasts (CAFs) in peritoneal surface malignancies.
figure 3

Signals from the tumour microenvironment can activate precursor cells from different tissue sources: peritoneum, adipose tissue/omentum, and bone marrow. The potential sources include mesothelial cells, submesothelial fibroblasts, pericytes, endothelial cells, mature adipocytes, adipose-derived mesenchymal stem cell, macrophages, and bone marrow-derived mesenchymal stem cells. The process which results from the conversion of those cellular precursors is indicated by the arrows. The diverse sources of CAFs lead to the emergence of diverse CAF subpopulations and heterogeneity within the tumour microenvironment.

Table 2 Possible cellular origin sources contributing to accumulation of activated fibroblasts in peritoneal fibrosis and cancer-associated fibroblasts within the tumour microenvironment.

Mesothelial cells

Mesothelial cells, which are specialised squamous epithelial cells, can transdifferentiate from mesothelial to mesenchymal states. The first evidence came from a study in peritoneal dialysis patients [59]. Indeed, later studies confirmed that similar transdifferentiation takes also place in PSMs of ovarian, endometrial and colorectal cancer patients [60,61,62]. The mesothelial-to-mesenchymal transition (MMT) process is characterised by a decrease of epithelial markers and increase in mesenchymal markers [59, 62]. Over time, there is a general downregulation of cytokeratins that are replaced by vimentin, resulting in loss of the cuboid epithelial morphology and acquisition of a spindle-shaped phenotype. Those fibroblast-like mesothelial cells lose contact inhibition, proliferate in a non-structured manner and acquire a migratory phenotype. The TGF-β1/Smad axis seems to be critical for full induction of this phenotype [63]. Sandoval et al. observed that peritoneal MAFs from biopsies of ovarian cancer co-expressed CAF-specific marker α-SMA alongside with the mesothelial markers cytokeratin and calretinin [62]. Similarly, in scirrhous gastric carcinoma, invading lead cells exhibited partial overlap in α-SMA and calretinin [64]. Miao et al. showed in gastric cancer that mesothelial expression of FAP was associated with poor prognosis, higher TNM stage and incidence of peritoneal dissemination by promoting cancer cells chemotaxis and adhesion [65]. During the colonisation of the peritoneum by colonic tumour cells, it was shown that CAFs found in the advanced primary tumours arise from MMT of mesothelial cells covering the visceral serosa [66]. Migrating mesothelial cells with spindle-like morphology were observed in the submesothelial region. These cells were negative for mesenchymal markers observed in invasive neoplasms. This resembles a transitional state of mesothelial cells, suggesting that MMT process is a critical event happening at the early stages of peritoneal dissemination. Interestingly, the in vitro treatment of omental mesothelial cells either with conditioned media of ovarian or colorectal cancer cells resulted in acquisition of MMT phenotypic features [62]. In a similar way, ex vivo cultures of mesothelial cells isolated from ascites of ovarian cancer patients undergo MMT and promote tumour growth in a xenograft mouse model [67]. Likewise, in cancer xenograft models the simultaneous injection of human mesothelial cells and tumour cells increased tumour fibrosis and peritoneal dissemination of gastric and ovarian cancer [68, 69]. Demuytere et al. suggested that MMT appears to be an important source of CAFs in the TME of CRC peritoneal metastases [70]. The immunohistostaining of biopsies showed overlap of mesothelial and CAFs markers. Recently, it was demonstrated in a pancreatic cancer model, that mesothelial cells are the cellular origin of antigen-presenting CAFs. By employing linage tracing methods Huang and colleagues identified the mesothelial-derived apCAFs as the main inducers of Treg formation in PDAC [71]. Interestingly, the treatment with mesothelin antibody inhibited the transdifferentiation. All together, these findings strengthen the concept that the activation of MCs through MMT process is a presumable source of peritoneal MAFs in PSMs.

Submesothelial fibroblasts

Peritoneal MAFs can also arise from resident submesothelial fibroblasts. Under normal conditions, fibroblasts in the peritoneum only occur in the submesothelial connective tissue [38]. Their CD34 expression suggests that they originate from blood-borne fibrocytes [72]. The absence of α-SMA, desmin, or FAP indicate that they are in a quiescent state. The specific conditions in the microenvironment of the peritoneal metastasis, such as tissue destruction, hypoxia and starvation, lead to attraction, activation and epigenetic reprogramming of the resident submesothelial fibroblasts, resulting in a constantly activated phenotype [73, 74]. Chen et al. by applying lineage tracing methods, identified the SM fibroblasts as the major precursors of myofibroblasts in peritoneal fibrosis [75]. In a similar approach, it was shown that around 25% of ACTA2+ CAFs were generated via activation of resting fibroblasts in a gastrointestinal cancer model [76]. Tumour cell secretion of TGF-β family members facilitates the activation and recruitment of resident fibroblasts in a Src- and Smad-dependent manner, respectively [77, 78]. Resident fibroblasts in the metastatic niche might be pre-activated even before the tumour cells arrive at the distant site. In prostate cancer, the activation of local fibroblasts in the pre-metastatic niche, similarly to those in the primary tumour, accelerates and supports the growth of metastatic colonies [79]. Those activated fibroblasts increase the expression of ECM proteins and chemokines. Cai et al. reported that activated normal omentum fibroblasts behave upon TGF-β1 treatment similarly to CAFs isolated from metastatic ovarian cancer patients, and those cells could also be found in omentum of these patients without metastasis [80]. Activation of omental normal fibroblast into CAFs promote the peritoneal metastasis of ovarian cancer [81].


Pericytes are mural cells involved in the microcirculation. It has been observed in a xenograft mouse model that platelet-derived growth factor-BB (PDGF-BB), which is overexpressed in several tumour types, induces a pericyte-to-fibroblast transition [82]. Inhibition of the PDGF-BB-PDGFRβ signalling reduced this transition as well as tumour invasion and metastasis. A study on gastric cancer-derived exosomes showed that they are able to induce the transdifferentiation of pericytes into CAFs by an PI3K/AKT and MEK/ERK pathway-dependent mechanism [83]. Although both AKT, as well as ERK, are responsive to PDGF, the potential role of this factor was, unfortunately, not investigated. Transition from pericytes to fibroblasts was also observed in a peritoneal dialysis model [84]. Macrophages were found to induce the transformation via the gasdermin/IL-1β axis. Recently, a single-cell RNA sequencing study showed the contribution of pericytes in human kidney fibrosis [85]. This observation supports the previously described origin of myofibroblasts in fibrosis by a lineage tracing study [86]. Thus, the transdifferentiation of pericytes into fibroblasts seems to be a general response and not restricted to malignancies.

Endothelial cells

It was shown in a spontaneous pancreatic carcinoma model that TGF-β1 induces endothelial cells to undergo a phenotypic conversion into fibroblast-like cells [87]. This was associated with the emergence of mesenchymal marker fibroblast-specific protein-1 (FSP1) and down-regulation of CD31/PECAM. Invasive colon cancer cells were found to potentiate the formation of endothelial cell-derived fibroblasts via the TGF- β1/tubulin-β3 axis [88]. Lineage tracing studies demonstrated that endothelial cell-to-fibroblast transdifferentiation is also involved in kidney and peritoneal fibrosis of peritoneal dialysis patients [89, 90].

Omental cells

The omentum, which consists of well-vascularises and innervated layer of adipose tissue, is a common place of peritoneal metastasis. Omental adipocytes can promote peritoneal dissemination of PSMs via their transdifferentiation into fibroblasts [91]. The ability of cancer cells to induce such transdifferentiation was found for gastric cancer and ovarian cancer [92, 93]. Tumour cell-derived soluble factors, such as TGFβ1 inhibit adipogenesis of adipose-derived mesenchymal stem cells (AD-MSCs) while increasing the fibroblastic differentiation followed by ECM deposition [94]. The greater omentum is known to contain abundant AD-MSCs [95]. It was reported that AD-MSCs migrate towards tumour engraftments in animal models [96]. In a xenograft model, combination of lineage tracing with scRNA-seq analysis revealed that AD-MSCs can differentiate into diverse CAFs populations [97]. In an in vivo study, Kitayama et al. have demonstrated that MSC-like cells found in peritoneal fluid of PSM patients play an important pro-metastatic role [98]. Stimulation of AD-MSCs with TGF-β induced their differentiation to a fibroblast-like phenotype expressing type I collagen, vimentin, α-SMA and FAP. This was associated with an enhanced rate of metastasis formation. Human AD-MSCs isolated from omentum of ovarian cancer patients promote survival and chemoresistance of tumour cells in vitro [99]. Tang et al. and others have demonstrated the contribution of AD-MSCs to the peritoneal MAF generation and their support of peritoneal dissemination [100, 101]. It was found that ascites of ovarian cancer patients are enriched in lysophosphatidic acid. This bioactive phospholipid stimulates the differentiation of AD-MSCs into CAFs via TGF-b1/smad axis [102]. Furthermore, it was shown in an ovarian cancer mouse model that adipose-derived α-SMA+ stroma is the preferred engraftment location for peritoneal metastasis formation [103].


The omental milky spots increase in size and number in response to the initial colonisation of tumour cells, mainly by recruitment of macrophages which acquire an anti-inflammatory polarisation and suppress anti-tumour immune activities [104]. It was observed that macrophages from malignant ascites of gastrointestinal cancer patients transdifferentiate into fibroblasts though activation of TGF-β signalling and cell-matrix adhesion [105]. The authors found that in an animal model of PSM, these cells promote tumour growth. A scRNA-seq analysis to unveil omentum activation processes showed the existence of a novel population which co-expressed both fibroblast and macrophage markers, suggesting the presence of macrophage-derived fibroblast cells [106].

Not only cells of the peritoneal cavity but also cells from other sites of the body are potential sources of peritoneal MAFs. For example, a study in gastric cancer revealed that bone marrow-derived fibrocytes are recruited from the peripheral blood into the TME where they support tumour growth and fibrosis by differentiating into fibroblasts [107]. Cancer patients were found to have higher numbers of fibrocytes and bone marrow-derived MSCs (BM-MSCs) in their peripheral blood [108]. Evidence arising from tracing studies suggests that BM-MSCs can also be a source of fibroblasts [74, 109]. Following exposure to TGF-β, BM-MSCs show DNA hypomethylation and their gene expression shifts towards myCAF phenotype signature. Mishra et al. have shown that exposure of BM-MSCs to tumour conditioned media increased expression of α-SMA and FSP1 simultaneously with sustained expression of CXCL12, inducing their differentiation into a CAF-like population [110]. In a study on gastric cancer, Quante et al. utilised GFP-labelled BM-MSCs to illustrate that MSCs are recruited to tumour site in a TGF-β- and CXCL12-dependent manner, accounting for around 20% of all CAFs found within the tumour mass [111]. Supporting these observations, Worthley et al. reported that female patients with gastrointestinal neoplasias, who received BM transplants from male donors, presented Y chromosome-positive CAFs within the TME [112]. Finally, some studies even propose that malignant cells or tumour stem cells might also account for the CAF pool through the EMT process [113, 114]. However, earlier data suggest that genetic mutations present in both cancer cells and CAFs are mostly mutually exclusive [115].

Conclusion and outlook

The spontaneous death of disseminating tumour cells and the tumour-induced tissue destruction at the adhesion site trigger a chronic tissue repair reaction in which various types of fibroblasts are involved. They are recruited from various cell sources and attempt to heal the wound and restore the homoeostasis. However, in doing so, they support the survival, adhesion, growth and spread of tumour cells and make them more resistant to chemotherapy. Although this role of fibroblasts can also be observed in many other types of cancer, it is of particular importance in PSM. Firstly, they exhibit strong fibroblastic infiltrates. Perhaps more importantly, the site of spread and adhesion is more accessible than in other metastatic malignant tumours. A better understanding of the origin of peritoneal MAFs and their interplay with tumour and immune cells may help to identify specific drug targets. This could contribute to the development of new drugs that complement the current treatment options.