Circulating tumour cells in gastrointestinal cancers: food for thought?

Abstract

Gastrointestinal (GI) cancers account for 35% of cancer-related deaths, predominantly due to their ability to spread and generate drug-tolerant metastases. Arising from different locations in the GI system, the majority of metastatic GI malignancies colonise the liver and the lungs. In this context, circulating tumour cells (CTCs) are playing a critical role in the formation of new metastases, and their presence in the blood of patients has been correlated with a poor outcome. In addition to their prognostic utility, prospective targeting of CTCs may represent a novel, yet ambitious strategy in the fight against metastasis. A better understanding of CTC biology, mechanistic underpinnings and weaknesses may facilitate the development of previously underappreciated anti-metastasis approaches. Here, along with related clinical studies, we outline a selection of the literature describing biological features of CTCs with an impact on their metastasis forming ability in different GI cancers.

Introduction

Gastrointestinal (GI) cancers are diagnosed in one out of four cancer patients worldwide. They comprise a group of malignant tumours originating from various locations along the GI tract, from the oesophagus to the anus, including supportive glands like liver and pancreas [1]. The most common GI cancers arise in the epithelial tissue of the oesophagus, stomach, pancreas, liver, and colorectum. With 4.9 million new cases and 3.9 million deaths in 2020, GI cancers are among the leading cause of cancer-related fatality in both genders [2].

The majority of GI cancer mortality is associated with metastasis—a complex, multistep process involving systemic spread of tumour cells from the primary site throughout the body, followed by colonisation of secondary organs. The standard of care for GI cancer patients includes approaches such as surgery, chemotherapy, radiotherapy, targeted therapy and immunotherapy. However, once the tumour has metastasised to distant organs, with lungs and liver being the most frequent sites, therapeutic options become increasingly narrower [3–6]. Typically, metastatic lesions become clinically visible in GI cancer patients in the years that follow surgical removal of the primary tumour [7–10]. Hence, identifying key drivers of the metastatic process is an essential step to facilitate the development of novel therapeutic strategies that may prevent and/or eradicate metastasis.

In case of most solid tumours, including GI malignancies, growing cancerous lesions may shed cells in the bloodstream which can travel to distant organs and form metastases. These pioneers of the metastatic process are referred to as circulating tumour cells (CTCs). Though relatively little is known about the mechanisms that influence dissemination and seeding of CTCs, it is well proven that increased numbers of CTCs in blood correlate with poor prognosis in patients [11]. Owing to the rarity of these cells in the peripheral circulation and a limited set of biomarkers for their identification, isolation of CTCs has been notoriously challenging [12]. Several technologies, each with their own set of advantages and disadvantages, are employed for CTC isolation in GI cancers. They can be broadly divided into two categories: antigen-dependent (based on the unique markers expressed by tumour cells but absent in other circulating blood cells) and antigen-independent (agnostic to markers but based on specific biophysical characteristics of CTCs that differ from haematological cells). So far, the antigen-dependent platform CellSearch [13] and the antigen-independent microfluidics technology Parsortix [14] have been approved by the Food and Drug Administration (FDA) for use in at least one cancer type. Of note, in GI cancers, at present only the CellSearch technology has been approved for monitoring CTCs in colorectal cancer (CRC) [15].

Based on the emergence of these technologies, over the past decades, substantial clinical and research efforts have been directed towards CTCs. In this review, we summarise recent insights into their biology and clinical relevance in the most prevalent GI cancers.

Clinical relevance of CTCs in GI cancers

Considering their pivotal role in metastatic spread, CTCs are extremely important for disease progression. In clinical settings, CTCs are utilised as prognostic biomarkers in GI cancer. CTC detection levels (and respective cut-off values) in the peripheral blood differ among cancer types and depend on the isolation method used [16]. The presence of three or more CTCs per 7.5 ml of blood using CellSearch is the FDA-approved gold standard for prognosis in CRC [17], yet more recent studies with the identical isolation method demonstrate that detection of even one or two CTCs per 7.5 ml of blood is associated with unfavourable prognosis in CRC patients [18, 19]. In addition to the prognostic value of CTC, the investigation of CTC suitability for early cancer diagnosis, evaluation of therapeutic response and monitoring of recurrence after surgery is enabled by recent technological advances [20–24]. Here and in Table 1, we provided the most relevant studies investigating the clinical relevance of CTCs in the five most prevalent GI cancers in the recent years.

Table 1.

Clinical relevance of CTCs in GI cancers (studies published between 2017 and 2022).

Cancer type	No. of patients	Detection methods	Main findings/clinical trial no.	Ref.
Oesophageal cancer	115	ISET	Preoperative chemotherapy and CTC detection prior to treatment correlates with better short-term PFS for patients with stage II or III ESCC (NCT03005314)	[122]
	96	CellSearch	The presence of CTCs in patients during follow-up after tri-modality therapy was associated with significantly poorer DFS and OS regardless of the timing of chemoradiotherapy in non-metastatic oesophageal cancer patients (NCT00907543)	[26]
Gastric cancer	116	Centrifugal microfluidic system	CTC count (≥2 cells/7.5 ml blood threshold) after gastrectomy does not correlate with gastric cancer staging. CTC potentially used as an early diagnostic marker (85.3% sensitivity and 90.3% specificity)	[27]
	162	CellSearch	Dynamic changes in CTC count (≥2 cells/7.5 ml blood threshold) between baseline and at day 28 after treatment were significantly associated with PFS and OS (NCT01443065)	[28]
	93	CellSearch	Pre and postoperative CTC levels are prognostic markers for recurrence of advanced gastric cancer after resection	[123]
	228	Flow cytometry, immunofluorescent double staining	CK+CD44+ cells were significantly more common among patients with distant metastases and were associated with significantly shortened survival	[30]
	100	Ficoll	FGFR2+ CTCs (≥5 cells/10 ml blood) were associated with poorer RFS. CTCs might be helpful to identify an existing tumour with FGFR2 overexpression	[31]
	70	Immunofluorescence	CSV+PD-L1+ CTCs are significantly correlated with short survival duration and poor therapeutic response	[32]
	150	Ficoll	EpCAM−CEA+ cells count was significantly associated with 3-year RFS	[33]
Hepatocellular carcinoma	62	Multiplex fluorescence in situ hybridisation	Mesenchymal-like CTCs (≥1 cells/5 ml blood threshold) are correlated with the risk of early recurrence	[124]
	85	Flow cytometry	Preoperative GPC3+ CTCs (≥5 cells/8 ml blood threshold) are associated with poor prognosis and are a risk factor for microscopic portal vein invasion (UMIN000025989)	[125]
	105	Tapered slit filter	Postoperative CTCs were detected in 23.8% of HCC patients and associated with lower survival and higher recurrence in early HCC	[35]
	256	Canpatrol	CTC count and EMT status were not associated with predictive recurrence and clinical stages	[23]
	197	CellSearch	Postoperative CTCs (≥3 cells/7.5 ml blood) count is predictive of postoperative extrahepatic metastases	[22]
Pancreatic cancer	60	ISET, immunofluorescence	ALDH+ CD133+ CD44+ CTCs are predictors of tumour recurrence and associated with decreased disease-free and OS	[43]
	46	Specific TU-chip	Phenotypic-based CTC analysis was used to predict metastasis (hybrid: epithelial and mesenchymal) and OS (epithelial CTC)	[126]
	20	CellSearch	High number of CTCs (>10 cells/6 ml blood) was associated with OS and PFS. Expression of ALCAM, POU5F1B, and SMO mRNAs in CTCs is negatively correlated with shorter PFS and OS	[127]
Colorectal cancer	66	CanPatrol, in situ hybridisation	LGR5+ CTCs and mesenchymal-like CTCs correlated with the occurrence of metastasis	[128]
	153	CellSearch	CTC count (≥3 cells/7.5 ml blood) at baseline correlated with shorter OS (NCT01442935)	[129]
	121	Cyttel	Advanced CRC patients with CTC detection during chemotherapy had worse PFS and OS	[130]
	667	CellMax	CTC counts demonstrated high sensitivity in detecting CRC (stages I–IV)	[131]
	44	CellSearch	Presence of CTCs (≥2 cells/7.5 ml blood) was associated with disease progression and poor survival despite complete resection	[19]
	26	Cytology-based automated CTC detection	The number of CTCs in draining venous blood (mesenteric vein) was higher than in peripheral blood and increased significantly with stage progression	[132]
	1202	CellSearch	Elevated baseline CTCs (≥3 cells/7.5 ml blood) was detected in 41% of the patients. RAS mutation correlated with poor prognosis (NCT01640405, NCT01640444)	[45]
	349	CellSearch	CTC count (≥3 cells/7.5 ml blood) at baseline correlated with poor prognosis in metastatic CRC patients and might be useful as a biomarker for intensive first-line therapy decision (NCT01640405)	[133]
	168	EPISPOT, CellSearch	CTC detection at D28 and the D0–D28 CTC dynamics (after first line of treatment with FOLFIRI–bevacizumab) was associated with PFS and OS (NCT01596790)	[134]
	20	CellSearch	CTC levels are independent prognostic markers in KRAS mutated metastatic CRC patients	[135]

This table provides an overview of studies investigating the clinical relevance of CTCs in the five most prevalent GI cancers in recent years.

ALDH aldehyde dehydrogenase, CEA carcinoembryonic antigen, CK cytokeratin, CTC circulating tumour cell, CRC colorectal cancer, CSV cell surface vimentin, EMT epithelial–mesenchymal transition, EpCAM epithelial cell adhesion molecule, ESCC oesophageal squamous cell carcinoma, FGFR2 fibroblast growth factor receptor 2, GPC3 glypican 3, HCC hepatocellular carcinoma, LGR5 leucine-rich repeat-containing G-protein couple receptor 5, OS overall survival, PDAC pancreatic ductal adenocarcinoma, PD-L1 programmed cell death ligand-1, PFS progression-free survival, POU5F1B putative pou domain class transcription factor 1B, RFS recurrence-free survival, SMO smoothened.

In metastatic oesophageal cancer (OC), prognosis based on CTC analysis has been described as promising in numerous studies but with some limitations (e.g. lack of optimal cut-off value consensus) [25]. In non-metastatic OC patients undergoing tri-modality therapy (combination of radiotherapy, chemotherapy and surgery), CTCs were interrogated as potential prognostic indicators [26]. Using the CellSearch system, the presence of CTCs 6, 12 and 24 months after treatment was associated with significantly poorer disease-free and overall survival, regardless of the timing of chemoradiotherapy (neoadjuvant vs adjuvant chemoradiation). The results of an ongoing Canadian clinical trial (NCT02812680), using a large OC patient cohort (n = 200) may contribute to understanding the utility of CTCs and plasma microRNA in OC cancer management.

In gastric cancer (GC), a threshold of two or more CTCs per 7.5 ml of blood was used for differentiating patients with GC from healthy controls [27]. CTCs were identified in 85% of GC patients, but no associations between CTCs and clinicopathologic features such as histologic type, T stage, N stage, and mucin phenotype were found. Interestingly, more than 80% of the patients with early-stage disease (T1, N0) had detectable CTCs, implicating their use as an early diagnostic marker [27]. CTC levels, quantified at baseline and day 28 upon chemotherapy treatment initiation, were also reported to be a prognostic factor in GC, as they were associated with worse progression-free and overall survival [28]. Of note, CTCs often appear to be heterogeneous and in rare instances, some carcinoma-derived CTCs may even reduce the expression of their epithelial markers [29]. Therefore, antigen-based isolation using a single marker (e.g EpCAM) may underestimate the phenotypic diversity of CTCs. Efforts are made to increase the variety of CTC markers to further improve their characterisation [29]. For example, CD44, cytokeratin, fibroblast growth factor, cell surface vimentin protein, programmed cell death ligand-1, carcinoembryonic antigen, and HER2-positive CTCs were used as potential prognostic markers in GC [30–34].

In hepatocellular carcinoma (HCC), CTC counts before and after surgery were compared in early-stage HCC using different isolation methods, but contradictory outcomes were reported [22, 23, 35]. This illustrates again the importance of standardised CTC-detecting assays, able to capture heterogeneous and rare CTC in various contexts. In this direction, using a 113-patient cohort, CTC enumeration and phenotypic features were interrogated using CanPatrolTM technology. This technology is based on the enrichment of CTCs by combining red blood cell lysis, depletion of CD45-positive cells via magnetic bead separation method, and subsequent size-based isolation of CTCs [36]. Levels of CTCs and their respective phenotypes were not correlated with clinical stages or predictive of recurrence in HCC.

The discovery of biomarkers that could guide treatment decisions is also crucial in pancreatic ductal adenocarcinomas (PDACs). Neoadjuvant therapy in early-stage disease could improve patient survival [37–39]. The presence of three or more CTCs per 4 ml of blood is associated with shorter recurrence-free survival following surgery as well as worse overall survival [40]. CTCs expressing vimentin, a mesenchymal cell-associated surface marker, were detected in 76% of pancreatic cancer patients [41, 42]. The detection of vimentin-positive CTCs preoperatively correlated with change in the tumour burden (more advanced disease and metastasis) and short recurrence-free survival. Using isolation by size of epithelial tumour method, mesenchymal-like or stem-like CTC were also detected in pancreatic cancer [43].

Efforts made to assess the clinical validity of CTCs in non-metastatic cancer are of interest too, particularly for the identification of early treatment opportunities. A meta-analysis including 20 studies (n = 3687 patients) demonstrated that CTC detection in blood (presence of one or more CTCs per 7.5 ml of blood) of patients with non-metastatic CRC was correlated with aggressive disease progression and reduced disease-free survival [44]. Postoperative CTCs correlated with poor recurrence-free survival [24]. A randomised phase III clinical trial analysed the correlation between baseline CTC, molecular profiling and clinical characteristics [45]. Elevated baseline CTCs and RAS mutations were associated with poor clinicopathologic prognostic factors, such as stage IV at diagnosis and involvement of at least three metastatic sites. Similarly, analysis of subgroups of CRCs patients (left vs right hemi-colon and colon vs rectal cancer) revealed a correlation between CTC positivity (presence of three or more CTCs per 7.5 ml of blood) and anatomical location of the tumour [21]. Furthermore, quantitative and phenotypic heterogeneity of CTCs was observed in distal compared to proximal CRCs and distinct features of CTCs in left-sided colon cancer may be accountable for poor prognosis observed within this subgroup of patients [46].

Overall, while seemingly heterogeneous, GI CTCs appear to be promising biomarkers for monitoring cancer progression. Ultimately, the clinical utility of CTCs will strongly depend on their ability to be implemented in standard clinical practice and to provide useful information to aid clinical decisions.

Biology of CTCs in GI cancers

Molecular features of CTCs

One of the initial steps in the metastatic cascade involves detachment of tumour cells from the primary tumour by breaching through the basement membrane and invading the adjacent tissue, which is counterintuitive given the poorly-motile, adult epithelial cell origin of carcinomas [47]. To understand how tumour cells equip themselves for this crucial step, various models have been proposed. In vitro and in vivo studies found that metastatic cells can travel individually via an epithelial-to-mesenchymal transition (EMT)—an embryonic development process of trans-differentiation of epithelial cells to cells with a mesenchymal-like phenotype, contributing to their ability to invade, withstand stress and disseminate [48, 49]. Recent studies have revealed that, rather than epithelial- and mesenchymal-like extremes, this process involves a spectrum of transitional phases where these two states are somewhat plastic [50, 51]. In contrast, besides travelling individually as single cells, CTCs can also travel as clusters of two or more cells. Such CTC clusters have higher metastatic seeding capability compared to single cells [52]. CTC clusters can be homotypic (consisting of only tumour cells) or heterotypic (where tumour cells are accompanied by non-tumour cells) [53, 54]. For instance, CTC–neutrophil clusters, observed in both breast cancer patients and mouse models, greatly contribute to metastasis given their high proliferative abilities [53]. In addition to heterotypic cluster formation with immune cells, heterogeneous clustering of tumour cells with cancer-associated fibroblasts (CAFs) has also been described in a lung cancer murine model, though their metastatic potential remains unclear [54]. Further, clusters of CTCs with polymorphonuclear myeloid-derived suppressor cells were found in metastatic melanoma and breast cancer patients [55]. This heterotypic interaction promotes the metastatic potency of CTCs via ROS/Notch/Nodal signalling.

In the context of GI cancer, CTCs have been reported as both single cells and multicellular aggregates. CTC clusters along with single CTCs were detected using size-based isolation in CRC patients, where their abundance and vimentin expression correlated with inferior prognosis [56]. Similarly, the presence of clusters has also been associated with worse survival in PDAC patients [57]. In GC, while the vast majority of detected CTC clusters consisted of two CTCs, CTC clusters with three-to-four cells were associated with therapeutic resistance and poor prognosis [58]. Of note, primary tumours can also shed non-cancer circulating entities such as cancer-associated macrophage-like cells in pancreatic cancer [59] and circulating non-malignant endothelial cell clusters in CRC [60]; however, their impact in disease progression remains unclear.

Though a prognostic value for CTC enumeration has been clinically proven, their in-depth molecular characterisation may lead to improved disease management and precision medicine. Isolation of CTCs without altering their transcriptome during processing has been considered a major challenge [12]. So far, most studies have performed bulk RNA sequencing or targeted sequencing analysis, while only a handful of published studies subjected CTCs to single-cell RNA sequencing (scRNA-seq) analysis, a powerful tool allowing for a higher-resolution dissection of tumour complexity. In one of the first efforts, RNA sequencing of orthotopic pancreatic cancer murine model-derived and patient-derived CTCs was performed using antigen-independent microfluidic isolation [61]. The data revealed at least three distinct CTC populations, highlighting their heterogeneity, where the majority possessed low proliferative signatures and enriched stem cell-associated genes like aldehyde dehydrogenase 1 family member a2 (Aldh1a2). CTCs displayed expression of both epithelial and mesenchymal markers along with high levels of insulin-like growth factor binding protein 5 (Igfbp5) transcript. When studied at the level of primary tumour using RNA in situ hybridisation, this extracellular growth factor binding was found to be focally expressed at the epithelial–stromal interface [61]. In HCC, patient-derived CTCs were characterised through a newly developed technology that combines image flow cytometry and single-cell mRNA sequencing [62]. In this proof-of-concept study, the authors describe differential gene expression between CTCs isolated from the same patient and across patients, as well as an enrichment of gene sets commonly associated with xenobiotic metabolism, coagulation, and peroxisomes, as expected considering their hepatic origin. Of note, this analysis was conducted using a limited number of cells obtained from two patients [62].

In a more recent study focusing on metastatic GC, gel-based cell manipulation was employed for antigen-agnostic size-based isolation of single CTCs from patients’ blood. The subsequent transcriptomic analysis of single CTCs revealed a characteristic gene expression profile, implying that a majority of these cells may have undergone an EMT [63]. Moreover, the results suggested that the EMT was induced by adhesion of CTCs to platelets within the blood vessels. Additionally, these EMT-induced cells exhibited cell cycle arrest and acquired chemoresistance. Of note, a small fraction of CTCs was epithelial and did not express any platelet-adherence associated genes. These epithelial cells were metabolically more active than other CTCs and correlated with a poor prognosis of the patients [63].

Although single-cell analysis is informative, naturally occurring low numbers of CTCs are a limitation for robust interpretation. To combat this, attempts have been made to culture CTCs and expand them ex vivo, aiming to augment the material for subsequent experimental interrogation [16]. Autologous cell lines were generated from colon CTCs, and high numbers of CTCs (~300 cells) were needed for ex vivo culture [64, 65]. Such ex vivo models can be instrumental in assessing functional properties of CTC and accelerating drug development efforts [66].

Altogether, with regard to molecular features, the scientific literature indicates various degrees of heterogeneity in GI CTCs, yet, further studies are required to gain more knowledge on different types of GI cancers and how their biology influences CTC generation dynamics.

Metastatic patterns

The metastatic spread to distant organs does not appear to be a random process—as different cancer types preferentially metastasise to specific sites—a process referred to as organotropism [67]. Organ-specific metastasis is dictated by several factors including the anatomical location, circulation pattern, organ-specific niches, tumour intrinsic factors, and the interaction between the tumour cells and host microenvironment. Differences in the genomic makeup of metastases according to their organ location were also identified [68]. Furthermore, the vascular architecture within the GI tract contributes to the specific metastatic pattern of GI cancers. The hepatic portal system transports the blood, through the portal vein, from most of the GI tract to the liver. The liver is a densely vascularised organ, receiving dual blood supply (both from the hepatic arteries and the portal vein) and harbouring a rather fenestrated endothelial layer, favouring a permissive environment for CTC extravasation. The described vascular organisation is in line with the clinical observation depicting the liver as the most common site of metastases in GI cancers followed by peritoneal metastases [69, 70]. Occasionally however (e.g. distal rectum), the venous blood drainage bypasses the liver and directly flows into the lung via the inferior vena cava, resulting in a higher proportion of lung metastases in rectal cancer compared to colon cancer (20% vs 8%) [69]. In HCC, CTCs exit the liver through the hepatic veins, pass via the heart (via the inferior vena cava) and then reach the lungs, where most metastases are found [71].

Currently, the spatial representation of CTCs within anatomically distinct regions of the human circulatory system has been limited mostly due to detection (e.g. low abundance of CTCs) and accessibility issues (access to different blood vessel locations). Characterisation of CTCs within distinct compartments could be informative and provide an insight into their distinct phenotypes (i.e. abundance, morphology, heterogeneity), specific molecular features, as well as their intrinsic metastatic ability. Interestingly, comparing mesenteric venous blood versus central venous blood compartments, higher detection rate and abundance of CTCs was shown in 200 post-surgery CRC patients using the CellSearch system [72]. Further, transcriptional heterogeneity in CTCs was explored upon drawing the blood from four key vascular compartments (i.e. portal, hepatic, peripheral veins, and peripheral artery) in 10 HCC patients [73]. A single-cell level gene expression analysis (133 cells in total) revealed that CTCs isolated from different vascular compartments clustered by the patient of origin, indicating that interpatient heterogeneity is higher than intrapatient, intervascular compartment heterogeneity. Yet, the gene expression profiles of CTCs across the different compartments were further analysed within individual patients and highlighted interesting differences (e.g. cell cycle and immune response genes) between compartments but also within the same blood vessel. Notably, the heterogeneity of CTCs was significantly decreased in the peripheral artery compared to the other vascular compartments, suggesting a selection of CTCs that passed through the lung capillaries [73]. A higher number of sequenced CTCs will be required for further validation of these interesting observations. In pancreatic and bile duct cancer, a study collected portal vein blood from 41 patients and cultivated FACS-isolated CTCs and immune cells using a patient-derived ex vivo platform [74]. By adding portal blood mononuclear cells to the platform, they reconstituted the CTC and immune cell interactions that are characteristic of the portal venous system. After seven days in culture, CTCs and immune cells formed clusters which promoted CTC survival and growth in vitro [74].

Moreover, peritoneum dissemination is also observed in GI cancers and peritoneal carcinomatosis is considered to be the end stage of the disease. Especially in CRC, tumour cells form clusters to evade anoikis and shed into the lymphatic system, leading to peritoneal dissemination as opposed to the invasion-metastasis cascade that occurs during hematogenous dissemination [75]. CTCs may travel through lymphatic vessels but how cancers choose their route of dissemination is debated [76].

Altogether, CTC properties in GI cancers are affected by multiple factors including anatomical and biochemical features. Mechanical cues such as shear stress, size restriction and mechanical trapping may also impact CTC composition. A better understanding of these phenomena in GI cancer is likely to reveal unexpected, yet potentially druggable metastasis-relevant patterns.

Tumour microenvironment

Until a few decades ago, cancer was thought to be exclusively a disease of abnormal tumour cells, generated through the accumulation of genetic aberrations. It is now widely appreciated that, additionally to the neoplastic cell components, microenvironmental elements such as stromal and immune cells play a pivotal role in cancer progression. In epithelial malignancies like colon and oesophageal carcinomas, tumours with more than 50% stromal composition correlate with unfavourable outcome [77].

In GI cancers, CAFs constitute a significant component in the stroma and their implication in metastasis, through various mechanisms, has been reported [78–80]. CAFs activate CXCL12/CXCR4 axis by integrin β1 clustering on the cell surface, enabling GC cell invasion [81]. Co-injection of patient-derived CAFs along with pancreatic cancer cells in mice leads to higher metastatic burden with an increasing proportion of co-injected CAFs [82]. Recent scRNA-seq studies have highlighted heterogeneity of CAFs in pancreatic cancer [83]. Along with plasticity amongst different phenotypic signatures, CAFs also change their role of pro-resistance to pro-invasion as pancreatic cancer progress through clinical stages [84]. Though studies have shown that CAFs support metastasis and interact with CTCs as described earlier, the impact of these interactions remains incompletely understood. CAFs can also engage and alter non-cellular components like the extracellular matrix (ECM) [85], and alterations in biomechanical properties such as ECM stiffness may trigger migration of cancer cells [86]. In CRC, matrix metalloproteinase-independent migration of cancer cells can be achieved by remodelling the basement membrane through CAF-induced biophysical forces [87]. Further, fragments of ECM components like collagen, laminins, elastins, and proteoglycans have been implicated as circulatory biomarkers and liaisons to metastasis [85, 88]. Molecular analysis of pancreatic CTCs revealed high expression of core matrisome ECM glycoproteins, such as SPARC, MGP and SPON2 [61]. Short-hairpin RNA-mediated SPARC knockdown in pancreatic cancer cells resulted in reduced migration in vitro as well as decreased metastatic burden in vivo following an orthotopic primary tumour-derived or tail vein injection into NSG mice [61]. Thus, ECM-related proteins impact metastasis but how they influence CTC dissemination remains unclear.

In addition to CAFs, immune cells are an impactful component in the GI tumour microenvironment (TME). Pro-tumour immune populations include M2 macrophages, N2 neutrophils, regulatory T cells and myeloid-derived suppressor cells; each contributing to the tumour aggressiveness via key effector molecules like colony stimulating factor-1, interleukin (IL)-6, metalloproteases, vascular endothelial growth factor, prostaglandin E2, transforming growth factor-β and IL-10 [89]. High levels of C-X-C chemokine motif ligand 5 (CXCL5) in the TME facilitate metastasis in GC by promoting invasion and migration via induction of EMT through activation of ERK signalling pathway in cancer cells [90]. Additionally, CXCL5 prompts activation of pro-tumour neutrophils via ERK and p38 signalling, resulting in elevated inflammatory cytokines like IL-6 and IL-23 that support metastatic potential of GC cells [90]. Tumour-associated macrophages are another immune cell type extensively studied for their contribution at each step of the metastatic cascade [91]. M2 polarisation of macrophages has been associated with inflammation in the TME, which in turn fosters metastatic progression of GI cancers [92]. A strong correlation between M2 and tumour neo-vessels has been described, suggesting their role in promoting angiogenesis and evolution of the tumour vasculature. Disorganised and collapsed neo-vessels accompanied by swift overgrowth of tumour cells leads to the development of hypoxic regions [93]. It has been reported that hypoxia triggers the intravasation of clustered breast CTCs highlighting the importance of the tumour biochemical landscape in the metastatic dissemination [94].

Overall, the complex interactions between the TME and cancer cells have a great impact on metastasis. However, the available evidence in GI cancer is limited to only few cell types and more comprehensive studies are required to fully understand these processes. Further exploration is needed to determine how biochemical and metabolic conditions of GI tumours influence CTC properties during the metastatic cascade.

Microbiota’s contribution to metastasis

Accumulating evidence points to the microbiota as a new component of the TME. The GI tract concentrates complex and large microbial communities (i.e. bacteria, archaea, eukaryotes, viruses) that can regulate cancer onset, progression, metastasis, and response to therapy [95–97]. Numerous preclinical and clinical studies have revealed that particular bacterial species, among other microorganisms, may have tumour-promoting or tumour-preventing effects in various types of cancer. Microbiota can directly facilitate tumourigenesis through mutagenesis by inducing DNA damage directly, interfering with mechanisms that maintain genome integrity or by activating oncogenic signalling pathways [96]. On the other hand, the microbiota and their metabolites can prevent cancer through their interaction with the host immune system. Similarly, microbes play a dual role in the metastatic spread: some species stimulate antitumoural immunity, while others promote a pro-tumourigenic inflammation at the metastatic sites [96]. These effects are regulated via signalling through metabolites, an immunosuppressive microenvironment, EMT, and gut vascular barrier impairment [97]. An imbalance of gut’s microbial community, termed dysbiosis, causes the disruption of the mucosal barrier, which allows the microbes, mostly bacteria, to spread to other organs by entering the bloodstream or lymphatic system directly through the disrupted epithelial and vascular barriers. Bacteria can also travel from the primary tumour environment to distant organs by invading cancer or immune cells [98–100]. Escherichia coli, for example, migrate to the liver and contribute to the pre-metastatic niche maturation by promoting the recruitment of innate immune cells and the formation of an inflammatory pro-metastatic environment [101]. Interestingly, when comparing the composition of the microbiota of primary CRC tumours and hepatic metastasis, it was found that they are both colonised by similar bacteria, such as Fusobacterium nucleatum, Bacteroides fragilis and Prevotella species [102].

A link between the gut microbiome, its metabolites, and immune responses in the liver was described in one primary and three metastatic mouse models of liver cancer [103]. Mechanistically, this study demonstrated that the microbiota modifies the bile acid inducing an antitumour effect with the accumulation of CXCR6-positive natural killer T cells. Conversely, lipopolysaccharide, produced by Escherichia coli, promotes CRC metastasis in a syngeneic mouse model [104]. Lipopolysaccharide increases the secretion of cathepsin k, which in turn mediates M2-like macrophage polarisation and promotes metastasis. In the clinical settings, recent studies highlight the emerging role of the microbiota as a diagnostic and prognostic marker in GI cancers [105–107]. In GC, the presence of Helicobacter pylori, a bacterium colonising the stomach and classified as a carcinogen, is also correlated with a better prognosis, highlighting its controversial role [108, 109]. By modelling the gut microbiota, H. pylori is involved at early stage of gastric carcinogenesis but is also suggested to be absent in the later stages of tumourigenesis. In a recent pan-cancer study analysing 35 cancer types, a potential prognostic value of fungi was suggested [106]. The authors characterised the cancer mycobiome of 17,401 tissue and blood samples and found cancer-type specific mycobiomes. Noteworthy, gut microbiota do not only affect cancer progression but also impact responses to chemotherapy, radiation and immunotherapy, mainly due to a role in drug metabolism [110, 111]. Recently, faecal microbiota transplantation was proposed as a promising therapeutic strategy to compensate for microbiota dysbiosis and to improve antitumour immune response [112]. However, long-term consequences of modifying the composition of gut microbiota, e.g. via faecal microbiota transplantation, will need to be addressed in larger randomised studies. Microbiota-targeted treatments in cancer show different responses across tumours, possibly due to the heterogeneity of the microbiome composition between patients and cancer types [113]. In pancreatic cancer, for example, the absence of microbiota correlates with a better response to PD-1 immunotherapy in preclinical models [114]. In particular, bacterial ablation modulates the immune component of the TME, as reflected by the reduced numbers of myeloid-derived suppressor cells, an increase in M1 macrophage differentiation, as well as an increased fraction of intratumoural T cells.

Currently, little is known about the interactions between the microbiota and CTCs across various cancer types. A recent study demonstrated that the distribution of the microbiota within a tumour is organised in micro-niches and promotes cancer progression in oral squamous cell carcinoma and CRC patients [100]. Using single-cell RNA-sequencing and in situ spatial-profiling technologies, the identity and in situ location of intratumoural microbial communities within the tumour were revealed. Co-culturing of tumour-isolated bacteria and CRC spheroids within a collagen gel containing myeloid cells, resulted in the recruitment of myeloid cells to the tumour spheroids, modification of the transcriptome of CRC cells and facilitation of cancer cell migration [100]. Interestingly, a conserved intracellular bacterial profile in murine and human breast cancer was previously reported [115]. These intracellular bacteria, while not being required for primary tumour growth, promoted CTC survival by enhancing resistance to mechanical stress through reorganisation of the actin cytoskeleton. Depletion of these intracellular bacteria reduced lung metastasis in experimental animals. Using ISH, an enrichment of bacteria in CTC clusters and in lung metastases was detected when compared to single CTCs and primary tumour site, indicating a favourable role of specific bacteria in metastasis [115]. This concept might be extended to other cancers. We anticipate that in the GI field, future research may uncover further dynamics that govern the complex interaction between CTCs and the microbiota, providing mechanistic insights on whether and how microbiota influences the cancer dissemination and outgrowth at distant sites.

Conclusions

As extraordinarily important precursors of metastasis in various cancer types, research on the biological features of CTCs has increased over the last decades. However, compared to other solid malignancies, studies exploring GI CTCs are relatively sparse. In this review, we have focused on CTCs and their interplay with various local and systemic factors that influence metastatic behaviour in GI cancers (Fig. 1).

Fig. 1. Circulating tumour cells (CTCs) in gastrointestinal (GI) cancer metastasis.

Primary tumours originating in various GI organs are complex entities consisting of heterogeneous tumour cell populations entangled with tumour microenvironment (TME) components like stromal cells, immune cells and microbiota. During the process of metastasis, cancer cells intravasate into circulation as various CTC entities, including single cells, homotypic clusters and heterotypic clusters. The liver and the lungs are the predominant metastatic sites, compared to peritoneum, bones and brain. Illustrations were created with BioRender.

Future work in this area of research may involve several aspects. For instance, location of the blood draw can impact on CTCs abundance and characteristics in GI cancers (e.g. peripheral vs central locations) [72, 116, 117]. The origin and plasticity of GI cancer cells during disease progression is another important feature in the spatiotemporal dynamics of the metastatic process [118, 119]. Along these lines, a recent study identified a novel rare tumour cell population, named high-relapse cells, responsible for metastatic relapse in CRC patient samples and mouse models [120]. Further research should also take into consideration the timing of sample collection and therapeutic intervention, given recently observed effects of the circadian rhythm on CTC intravasation and metastatic ability [121]. Practically, in the case of GI cancers, both timing and location of blood collection could help capturing more (and maybe more disease-relevant) CTCs, eventually including those with higher metastatic propensity. Devices that integrate CTC capture with molecular and functional testing at specific time points may facilitate a routine application in the clinical context.

Finally, although recent research advances outlined in this review highlight the relevance of CTCs in GI cancer, several open questions remain. Owing to primary tumour heterogeneity in the GI tract, are there diverse morphological manifestations of CTCs, such as the presence of heterotypic clusters? If so, what type of non-tumour cell types partners with CTCs in GI cancers? Is the anatomical location and timing of blood draw impacting on CTC abundance and characteristics? How is the TME affecting dissemination of GI cancers? Can CTCs be exploited to improve diagnosis and treatment of GI malignancies? Answering these questions will be key to a better understanding of the metastatic process in cancers of the GI tract, and may provide prospective avenues for the development of innovative treatments.

Author contributions

SA, MN, AG and NA wrote the article. SA, MN and AG performed literature search. SA designed the figure.

Funding

The Aceto laboratory is supported by the European Research Council (101001652), the strategic focus area of Personalised Health and Related Technologies at ETH Zurich (PHRT-541), the Future and Emerging Technologies programme of the European Commission (801159-B2B), the Swiss National Science Foundation (212183), the Swiss Cancer League (KLS-4834-08-2019), the Basel Cancer League (KLbB-4763-02-2019), the ETH Lymphoma Challenge (LC-02-22) and the ETH Zürich. Open access funding provided by Swiss Federal Institute of Technology Zurich.

Data availability

Not applicable.

Competing interests

NA is co-founder and member of the board of PAGE Therapeutics AG, Switzerland, listed as inventor in patent applications related to CTCs, paid consultant for companies with an interest in liquid biopsy, and a Novartis shareholder.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.