During cancer progression, metastatic spread occurs when cancer cells disseminate from the primary tumor and travel to a distant site to form a metastasis (Micalizzi et al., 2017). This can emerge through three major routes: the blood circulation, the lymphatic system, or via serosal or mucosal surfaces (Fidler, 1978). However, cancer cell dissemination through the blood is thought to be the main route of metastasis (Lambert et al., 2017), and the subset of cancer cells that have entered the blood circulation is named circulating tumor cells (CTCs). Intravasation of CTCs into the blood stream is believed to be one of the rate-limiting steps for metastasis formation and can occur through either an active invasion or passive shedding of cells from the tumor (Cavallaro and Christofori, 2001; Bockhorn et al., 2007; van Zijl et al., 2011). Only a minority of cancer cells reaching the blood circulation manages to survive shear stress, escape immune surveillance, avoid detachment-induced cell death, extravasate at the distant site, and finally establish a metastasis (Massague and Obenauf, 2016). Thus, the process of metastasis is both a complex and inefficient process (van Zijl et al., 2011; Reymond et al., 2013).

In addition to representing the primary tumor, CTCs have also been shown to exit from metastatic lesions (Kim et al., 2009). Such cells have the potential to travel back to the primary tumor site (called tumor self-seeding) or create another metastasis (Mentis et al., 2020). Therefore, CTCs could represent both the primary tumor and potential metastatic lesions (Kim et al., 2009), making CTC analyses highly relevant even years after surgical removal of the primary tumor. Hence, CTC analyses could provide important information about disease progression and relapse. Furthermore, molecular analyses of CTCs including mutational profiling, could provide the basis for personalized therapies in the future (Greene et al., 2012). Thus, CTCs are currently evaluated for clinical diagnostics. However, CTC analysis remains technically challenging, not only due to the rarity of CTCs among billions of normal blood cells, but also due to their inherent high degree of cellular plasticity complicating the choice of detection markers (Alix-Panabieres et al., 2017). Accordingly, CTC enrichment and detection strategies must be based on highly specific biomarkers to achieve the necessary assay specificity and sensitivity. Moreover, targeting a broader CTC population would be beneficial to ensure that the liquid biopsy better reflects the heterogenic cancer cell population.

As a part of discovering novel CTC targets, many strategies focus on proteins known to play an active role in metastatic seeding. Although solid tumors differ in their metastatic patterns, they share certain mechanistic similarities for metastasis formation, which are summarized as the metastatic cascade (Figure 1) (de Groot et al., 2017; Lambert et al., 2017; Riggi et al., 2018). Notably, the majority of steps in this process concerns the interaction between the cancer cells and the surrounding extracellular matrix (ECM). Therefore, novel clinically relevant CTC targets may be discovered within the pericellular layer called the glycocalyx.

The glycocalyx is a thick network of carbohydrates bound to glycoproteins, glycolipids, and proteoglycans (Figure 2A; reused from Okada et al., 2017). It is present on cell surfaces throughout the entire human body and constitutes a physical barrier between the cell and the surrounding microenvironment (Tarbell and Cancel, 2016). The glycocalyx plays a crucial role for receptor–ligand interactions of cancer cells and their surroundings, enabling migration as well as intra- and extravasation. Furthermore, the composition of the glycocalyx is thought to influence the transportation and survival of CTCs in the bloodstream (Mitchell and King, 2014). However, very few studies have investigated the glycocalyx of CTCs. At present, it is best studied in endothelial cells, where it serves as a physical and electrostatic barrier as well as a mechanotransducer toward other cells, the extracellular matrix (ECM), or shear forces of the blood (Reitsma et al., 2007; Butler and Bhatnagar, 2019). To convey signaling, growth factors, chemokines, and other interaction partners have to navigate through this dense structure. The glycocalyx, which extends beyond the length of most surface receptors, has a dual role in signaling by creating a physical hindrance for ligand receptor interactions or by promoting binding once interaction partners are in close proximity to each other (Kuo et al., 2018). Moreover, certain glycocalyx components are involved in chemokine storage and oligomerization, which strongly modulates their signaling strength (Salanga and Handel, 2011). Therefore, glycocalyx changes can have various effects on cellular behavior and, not surprisingly, cancer cells show specific alterations in their glycocalyx.

The cancer cell glycocalyx is a highly dynamic structure from which most of the components have been linked to the acquisition of oncogenic phenotypes (Daniotti et al., 2015; Buffone and Weaver, 2020). For instance, aberrant glycosylations including hypersialylation support immune evasion mechanisms (Pearce and Läubli, 2016). Moreover, increased expression of bulky glycoproteins like mucin-1 has been linked to aggressive cancers (Paszek et al., 2014; Barnes et al., 2018) and associated with poor survival outcome in patients (Kufe, 2009). This might be explained by the bulkiness of the mucin-1 ectodomain shaped by numerous glycosylations, which facilitates integrin clustering, cell signaling, and cell proliferation (Paszek et al., 2014; Woods et al., 2017; Kuo et al., 2018). Naturally, mucin-1 is of interest as a target for therapy (Pillai et al., 2015), due to its high involvement in cancer. Likewise, CTCs were found to express high mucin-1 levels (Paszek et al., 2014), and mucin-1 has also been explored for CTC capture and detection (Muller et al., 2012; Strati et al., 2013; Schehr et al., 2016).

Another important component of the glycocalyx are proteoglycans (Figure 2B) with multiple implications in metastatic dissemination of cancer cells and tumor cell growth (Iozzo and Sanderson, 2011; Vitale et al., 2019). Proteoglycans can be secreted into the ECM or located intracellularly as well as on the cell surface either directly embedded in the plasma membrane or anchored by a glycosylphosphatidylinositol (GPI)-linker (Iozzo and Schaefer, 2015). Proteoglycans consist of two functional units: protein core and glycosaminoglycan (GAG) chains (Walimbe and Panitch, 2019). Most commonly, the assembly of GAG chains occur from a tetrasaccharide linker region covalently attached to serine residues within the protein core (Esko and Zhang, 1996). The GAG family is classified by their chemical composition and includes chondroitin sulfate (CS), dermatan sulfate (DS), keratan sulfate (KS), heparan sulfate (HS), and hyaluronic acid (HA) (Figures 2C,D) (Toole, 2004; Bulow and Hobert, 2006). In general, GAGs consist of long linear repeats of disaccharide units consisting of hexuronic acids and hexosamines. The hexuronic acid epimers comprise D-glucuronic acid (GlcA) for CS/HA and L-iduronic acid (IdoA) for DS, whereas the hexosamine units consist of an N-acetyl-D-glucosamine (GlcNAc) for HS/KS/HA, and an N-acetyl-D-galactosamine (GalNAc) for CS/DS (Schaefer and Schaefer, 2010; Ghiselli and Maccarana, 2016; Pomin and Mulloy, 2018). The structures of GAGs are extremely diverse, as their synthesis in the Golgi apparatus is not based on a precise template, but on a redundant network of enzymes that seems to be regulated based on tissue and cell types (Dick et al., 2012; Mikami and Kitagawa, 2013; Chen Y. H. et al., 2018).

Besides the variation in the monosaccharide composition, the molecular diversity of GAGs also results from varying polymer lengths and extensive post-translational modifications such as sulfations and epimerizations along the chain (Bulow and Hobert, 2006). GAG sulfation patterns often determine their biological function and serve as specific recognition motifs for a wide variety of growth factors, cytokines, chemokines, and pathogens (Xu and Esko, 2014; Mizumoto et al., 2015; Pinho and Reis, 2015). Therefore, alterations in the GAG composition of proteoglycans in cancers have received a lot of interest (Sweet et al., 1976; Chandrasekaran and Davidson, 1979; De Klerk et al., 1984). A well-studied example is the change in sulfation patterns of GAGs, which likely depends on the specific cancer type. Some studies have reported high expression of CS 4-O-sulfotransferases in both ovarian and breast cancers, while a study on cancerous lung tissues found elevated 6-O-sulfated CS, compared to nonmalignant tissue (Cooney et al., 2011; Oliveira-Ferrer et al., 2015; Li et al., 2017). Similarly, various HS sulfotransferases have been found upregulated in different cancers. These include 6-O-sulfotransferases in ovarian and colorectal cancer; 3-O-sulfotransferases in breast and pancreatic cancer; along with N-deacetylase and N-sulfotransferases in hepatocellular carcinomas (Tatrai et al., 2010; Song et al., 2011; Hatabe et al., 2013; Cole et al., 2014; Vijaya Kumar et al., 2014). Moreover, several studies have reported increases in CS quantity or in expression of CS polymerization genes in malignant tissues, suggesting that CS polymers are pro-tumorigenic (Momose et al., 2016; Li et al., 2017; Hou et al., 2019).

Interestingly, some proteoglycans may be modified with different GAG types simultaneously, as seen for syndecans, which can carry both CS/DS and HS, dependent on the structure of the ectodomain (Kokenyesi and Bernfield, 1994; Iozzo and Schaefer, 2015). Similarly, the HA-binding proteoglycan, versican, undergoes alternative splicing of exons encoding the GAG-attachment region resulting in altered GAG display. Notably, expression of distinct versican isotypes was shown to facilitate cancer progression in multiple cancer types (Dours-Zimmermann and Zimmermann, 1994; Theocharis et al., 2015; Zhangyuan et al., 2020). The protein cores of proteoglycans are, however, not just scaffolds for GAG extension, since they also directly bind ligands and mediate intracellular signaling in GAG-independent manners.

In contrast to the rest of the GAGs, HA stands out by not being covalently attached to a protein core (Figures 2B,D). Instead, HA is synthesized as an unmodified polysaccharide at the plasma membrane, where it is extruded from the cell surface and cleaved off into the ECM (Weigel and DeAngelis, 2007; Itano, 2008). In most cells, HA is an abundant structural component of the glycocalyx, where it interacts with receptors and surface proteoglycans via their hyaluronan-binding motifs and regulates the viscosity of the glycocalyx by its ability to retain water (Toole, 2001, 2004). Upon binding, HA triggers activation of a range of signaling pathways involved in cell proliferation, differentiation, motility, and adhesion, thereby influencing processes such as development, tissue homeostasis, and carcinogenesis (Turley et al., 2002; Toole, 2004; Liu et al., 2019). Although HA is extensively involved in cancer, it will not be discussed in detail throughout this review, as it is not considered a proteoglycan due to its lack of a protein core.

Altogether, proteoglycans compose a highly heterogeneous group of proteins that diverge by structural alterations of the protein core as well as by differences in their GAGs with regard to chain number, type, length, and post-translational modifications. Notably, proteoglycans are important integrators for cell signaling events with direct implications for carcinogenesis and cancer progression (Iozzo and Sanderson, 2011; Pinho and Reis, 2015; Nikitovic et al., 2018). In spite of this, the functions of proteoglycans and their GAGs in relation to CTCs are currently understudied. Therefore, this review will highlight examples of proteoglycans involved in the metastatic cascade with potential links to CTC biology. More specifically, we will discuss how proteoglycans play active roles in cancer cell proliferation, migration, survival, plasticity, and invasion with a dedicated focus on the function of both the protein core and the GAG chains. Finally, we provide an overview of proteoglycans that are currently evaluated for CTC technologies and briefly highlight some of the technical aspects to consider when studying proteoglycans.

Circulating tumor cell detection assays have spurred increasing clinical interest since the prognostic value in progression-free and overall survival was established in patients with metastatic colorectal (Cohen et al., 2009), breast (Cristofanilli et al., 2004), prostate (de Bono et al., 2008), and lung (Krebs et al., 2011) cancer. CTC enumeration from patient blood samples has also demonstrated clinical relevance for several other cancer types such as pancreatic cancer (Kurihara et al., 2008; Bidard et al., 2013; Effenberger et al., 2018) or hepatocellular carcinoma (Schulze et al., 2013; Qi et al., 2018). The presence of detectable levels of CTCs in the peripheral blood is associated with the metastatic capacity of the disease (Allard et al., 2004; Cristofanilli et al., 2004). However, low levels of CTCs have been reported in non-metastatic disease for several cancer indications before and after surgery (Thorsteinsson et al., 2011; Franken et al., 2012; Gazzaniga et al., 2013). Additional studies suggest that CTCs are even shed from premalignant lesions and this opens the possibility for using CTC detection for early diagnosis of cancer (Husemann et al., 2008; Stott et al., 2010; Rhim et al., 2012; Zhang et al., 2014; Tsai et al., 2016; Murlidhar et al., 2017). CTC assays might also have potential as a tool for predicting treatment efficacy and monitoring disease (Schochter et al., 2019; Yang et al., 2019), thereby providing real-time, non-invasive information about the disease by liquid biopsies. Furthermore, many CTC assays do not only enable enumeration but also downstream analyses such as genomic, transcriptomic, proteomic, or phenotypic characterization of cancer cells. Therefore, studying CTCs can also bring novel insight into aspects of metastasis formation, which are still not fully understood (Chaffer and Weinberg, 2011). Despite the interest and potential in analyzing CTCs, the methods are rarely implemented in the clinical setting, as CTC identification requires highly specific markers and an extreme assay sensitivity. Many CTC methods struggle to reach the needed sensitivity, as it is a technical challenge to detect few cancer cells in billions of normal blood cells.

Several CTC enrichment technologies ranging in complexity have been developed (Kowalik et al., 2017; Dianat-Moghadam et al., 2020). From whole blood, CTCs can be enriched along with leukocytes by density fractionation or a simple lysis of the erythrocytes. The crude cell enrichment can be analyzed by direct antibody staining and examined by, e.g., microscopy (Hillig et al., 2015; Werner et al., 2015) or flow cytometry (Hristozova et al., 2012; Watanabe et al., 2014; Lopresti et al., 2019). Because of the rarity of the CTCs, an additional cancer cell-specific enrichment step is, however, often preferred. CTCs can be enriched from the leukocytes based on distinct biophysical properties such as size, density, deformability, or charge (Liu et al., 2015; Mitchell et al., 2015; Shaw Bagnall et al., 2015). Following CTC enrichment, the detection of CTCs will still rely on a staining step, distinguishing for example cytokeratin (CK)-positive CTCs from the remaining CD45-positive leukocytes. Other systems for CTC isolation use cancer- or tissue-specific antibodies to enrich for CTCs (Coumans and Terstappen, 2015) or even leukocyte cell surface proteins like CD45 to deplete for leukocytes (Ozkumur et al., 2013; Karabacak et al., 2014). The positive selection of CTCs is evidently very dependent on highly specific cancer or tissue markers. In order to demonstrate high potential for clinical application, extensive validation of CTC capture platforms must reveal robust clinical sensitivity and specificity (Parkinson et al., 2012). Most pilot studies do not present large-scale clinical data and should hence be interpreted with caution. Inclusion of healthy controls becomes crucial to demonstrate the specificity of the capture and/or detection strategy. Alternatively, some studies apply downstream molecular analyses to verify the tumor origin of detected CTCs, e.g., by mutation detection, which supports the reliability of the CTC assay (Gasch et al., 2013; Muller et al., 2014). From a more technical perspective, pre-analytical conditions such as blood tubes, storage time, and temperature as well as choice of antibody clones can have a huge effect on assay performance (Qin et al., 2014; Ilie et al., 2018; Wu et al., 2020), making comparisons across studies difficult. Furthermore, in the light of exploring proteoglycans as potential CTC targets, the consideration of technical assay parameters become crucial for, e.g., sustaining the GAG stability.

The current clinical standard for CTC enumeration is the CellSearch^® platform, which is approved by the American Food and Drug Administration (FDA) for monitoring patients with metastatic breast, colorectal, and prostate cancer. CellSearch^® relies on cell enrichment using anti-EpCAM antibody-coated ferrofluid and CTC detection via fluorescent anti-CK antibody labeling (Liberti et al., 2001; Allard et al., 2004; Coumans and Terstappen, 2015). EpCAM-based capture approaches are, however, rarely efficient for epithelial cancers with downregulated EpCAM expression, likely due to EMT, or cancers of mesenchymal origin. Therefore, several studies have been focusing on finding novel markers, which can distinguish EpCAM-low or -negative CTCs from normal blood cells with high specificity and sensitivity to broaden the spectrum of detectable CTC subpopulations (Lampignano et al., 2017; Nicolazzo et al., 2019). As a part of this, multiple strategies using proteoglycans for CTC enrichment or identification are currently under investigation. See Table 1 for an overview on the most used proteoglycans and their applications. Some of them are highlighted in the following.

A well-known example is the CellSearch^® Circulating Endothelial Cell Kit, which can be used for the enrichment of circulating melanoma cells that are EpCAM-negative by nature (Rao et al., 2011; Khoja et al., 2013). After capture using ferrofluid coupled with antibodies against melanoma cell adhesion molecule (MCAM), circulating melanoma cells are identified by staining with antibodies against high molecular weight melanoma-associated antigen (HMW-MAA), also known as CSPG4.

As described earlier, CSPG4 has been linked to many aspects of the metastatic cascade, including proliferation, migration, as well as ECM-remodeling and is expressed across many cancer types (Ilieva et al., 2018). Moreover, CSPG4 is expressed in a majority of melanoma lesions (Real et al., 1985) and is a well-characterized surface marker for melanoma (Ilieva et al., 2018). Multiple retrospective studies using the MCAM/CSPG4 CellSearch^® Kit have found that CTC levels detected at baseline correlates with overall survival in late-stage melanoma (Rao et al., 2011; Khoja et al., 2013; Bidard et al., 2014) (Table 1). Recently, two large prospective studies also evaluated the prognostic significance of MCAM/CSPG4-positive CTCs in cutaneous melanoma. In a study of 93 stage IV patients, Hall et al. (2018) found that presence of CTCs at baseline was associated with shorter progression-free survival after 6 months compared to CTC-negative patients. Later, the same research group showed that CTC-positivity at baseline for stage III patients (n = 243) was an independent predictor of relapse-free survival within 6 and 54 months (Lucci et al., 2020). The CTC levels in these studies were not associated with primary tumor characteristics, such as ulcerations, tumor thickness, and mutational status. Therefore, MCAM/CSPG4-positive CTC numbers may add additional information on top of clinicopathological characteristics for clinicians to foresee the disease course in the future.

Interestingly, CTC-negative melanoma patients have been found to have better progression-free or relapse-free survival compared to CTC-positive patients (Hall et al., 2018; Lucci et al., 2020). However, a significant proportion of late-stage melanoma patients still appear to have no CTCs detectable by CellSearch^® (Rao et al., 2011; Roland et al., 2015; Hall et al., 2018). This has also been reported for other CSPG4-based methods (Ulmer et al., 2004; Ruiz et al., 2015) as well as for CSPG4-independent isolation methods (Khoja et al., 2014; Aya-Bonilla et al., 2019). This may simply be due to the rare nature of CTCs. However, there could be CSPG4-negative CTC subpopulations, which are not captured by CSPG4-dependent strategies. In fact, one study found that of 31 melanoma patients with CTCs detectable by other markers, only 42% had CSPG4-positive CTCs (Gray et al., 2015), suggesting a need for multi-marker approaches.

During the past decade, multiple other studies investigated the potential of CSPG4 for CTC capture and/or identification in melanoma (Table 1). Up to 4 CTCs per mL blood was found using CSPG4 immunomagnetic capture (Sakaizawa et al., 2012), which is similar to the reported CTC numbers using the MCAM/CSPG4 CellSearch^® kit (Khoja et al., 2013). However, a study by Ruiz et al. (2015) using a CSPG4-based immunofluorescent microscopy approach without prior enrichment step identified a mean of 14.9 CTCs per 1 mL blood samples from melanoma patients (n = 40), without potential CTC hits in healthy control samples (n = 10). These variations in CSPG4-positive CTCs can be explained to some extent by the use of different CTC enrichment strategies, varying markers for CTC identification or other technical differences in the assays.

Commonly mutated genes in melanoma, such as BRAF and NRAS (Colombino et al., 2012), are upstream activators of ERK signaling (Savoia et al., 2019). As CSPG4 expression has been connected to ERK signaling (Ampofo et al., 2017), these mutations might be particularly important in CSPG4-positive CTCs. Indeed, a recent study found that CTCs enriched by CSPG4-based method presented more RAS/RAF mutated cells compared to CTCs isolated only by physical properties (Gorges et al., 2019). Since some therapeutic approaches target the serine/threonine-protein kinase B-raf (BRAF) (Holderfield et al., 2014), it is possible that the CSPG4 expression might also decrease in response to this form of treatment, which could affect the prospect of using CSPG4 alone to monitor CTC numbers. Indeed, initial longitudinal study of CTC heterogeneity in 10 stage IV melanoma patients suggested that expression of CSPG4 on CTCs may be downregulated in response to BRAF and mitogen-activated protein kinase kinase (MEK)-inhibiting therapy (Tsao et al., 2018). However, to our knowledge, none of the major studies on CSPG4-positive CTCs in cutaneous melanoma have yet found any correlation between CTC levels and BRAF-mutational status or adjuvant therapy.

Another recent study revealed that the transcriptomic profile of CSPG4-enriched CTC populations from six patients was dominated by up-regulation of tumor necrosis factor alpha (TNFα)/nuclear factor kappa B (NF-κB) as well as signal transducer and activator of transcription (STAT) pathways (Aya-Bonilla et al., 2019). Both signaling pathways have central roles for cell proliferation as well as cell survival (Wu and Zhou, 2010; Igelmann et al., 2019). Furthermore, in silico analysis found other genes upregulated in the CSPG4-enriched population to be connected to metastasis, tumor growth, and melanoma biology (Aya-Bonilla et al., 2019), which indicates an interesting biological role of CSPG4-positive CTCs for melanoma progression.

Overall, CSPG4 is a relevant surface marker for melanoma CTCs and is hence evaluated in many studies. Although little is still known about the biological role of CSPG4-positive CTCs, they might represent a specific subpopulation. This potential bias should be considered when using only CSPG4 for CTC capture or CTC identification.

CD44 has also been explored for CTC analysis. CD44 is widely expressed (Goodison et al., 1999), and as previously described it acts as a receptor for a variety of ligands. Particularly well-described is the interaction with HA, which constitutes a major part of the glycocalyx and ECM (Banerjee et al., 2016). Upregulation of CD44 confers tumorigenicity, metastatic capacity, and drug resistance to primary tumor cells as well as CTCs (Naor et al., 2002; Fitzgerald and McCubrey, 2014). The abnormal expression of CD44 splice variants is associated with treatment refractoriness, recurrence, and prognosis (Katoh et al., 2015), and overexpression of both CD44s and variants serves a long list of biological functions across many cancer types (Chen C. et al., 2018). Since isoform switching introduces new cancer-related antigens, development of both anti-CD44s and anti-CD44v antibodies has attracted much interest.

As a CTC isolation tool, anti-CD44 antibodies have been used to capture CTCs from cancer patient blood (see Table 1). Yan-Bin et al. (2020) investigated the CD44-positive CTC abundance in NSCLC patients by immunomagnetic enrichment and evaluated the correlation to clinical characteristics. None or very few CD44-positive cells were detected in the 30 controls in contrast to frequent CTC observations in the 128 patient samples. Detected CTCs associated negatively with serum TNF-related apoptosis-inducing ligand (TRAIL) levels, suggesting that CD44-positive CTCs could be more vulnerable to TRAIL-induced apoptosis through death receptor 4 and 5 signaling (Yan-Bin et al., 2020). A small study on gastric cancer patients (n = 26) and healthy controls (n = 10) associated increased prevalence of EpCAM- and CD44-positive CTCs in patients with tumor depth, disease progression, and venous invasion (Watanabe et al., 2017). Consequently, CD44-based CTC detection was suggested to reflect the malignant potential of the tumor. The authors, however, disregard EpCAM-positive cells found in all healthy controls and the few double positive cells found in 2 healthy controls as either non-specific immunological reactions or contaminating skin cells. Again, this discrepancy highlights the demand for CTC validation, as for example via genomics. Another study analyzed CD44-positive CTCs isolated by immunomagnetic enrichment from 30 oral squamous cell carcinoma (OSCC) patients and 15 healthy controls (Patel et al., 2016). Self-renewal and proliferation capability of the CD44-positive cells were observed by increased sphere-forming capacity unlike the CD44-negative sorted population. Moreover, cisplatin resistance assays confirmed a drug-resistant phenotype associated with the CD44-positive population. This was specifically associated with high transcript levels of CD44v6, as opposed to CD44s, as well as elevated levels of the stemness marker NANOG. Furthermore, the different expression levels strongly correlated with the primary tumor profile and, importantly, clinicopathological parameters such as late-stage, loco-regional aggressiveness, and relapse. The findings suggest that detection of CD44v6-positive CTCs could be used to predict disease progression, therapy outcome, and recurrence.

In addition, CD44 is being evaluated for novel therapeutic approaches against CTCs. For instance, in vivo homophilic CD44-mediated CTC clustering of metastatic breast cancer cells in mice was largely inhibited by the administration of anti-CD44 neutralizing antibody, leading to decreased metastatic capacity (Liu et al., 2019). In summary, numerous studies of applications to target CD44-positive CTCs underline its potential in therapy and as a valuable marker for prognosis and treatment response.

Another interesting proteoglycan for clinical purposes is glypican-3 (GPC3), which is upregulated amongst several cancer entities with highest positive case rates in hepatocellular carcinoma (HCC). Importantly, GPC3 has been reported to discriminate between HCC and non-malignant lesions (Zhu et al., 2001; Wang et al., 2008) or other liver-associated cancers like cholangiocarcinoma (CCA) (Man et al., 2005). Nowadays GPC3 is included in a diagnostic HCC panel together with glutamine synthetase and heat shock protein 70, according to guidelines of the European Society for Medical Oncology (ESMO) (Vogel et al., 2019) and the American Association for the Study of Liver Diseases (AASLD) (Marrero et al., 2018). Furthermore, GPC3 might also be used as a serum biomarker (Capurro et al., 2003; Hippo et al., 2004) and is exploited for different targeted cancer therapy approaches (Sawada et al., 2012; Feng et al., 2013).

Several studies have been utilizing GPC3 for analysis of CTCs in HCC patients (Table 1). Anti-GPC3 antibodies have been used for positive immunomagnetic enrichment of CTCs (Court et al., 2018; Hamaoka et al., 2019). Hamaoka et al. (2019) found in a prospective, single-institution study that most of the 85 examined HCC patients had GPC3-positive CTCs with a median of 3 CTCs in 8 mL blood samples, whereas negative controls (in total n = 27) such as healthy individuals (n = 12) or individuals with inflammatory diseases (n = 4), only had a median of 1 GPC3-positive cell in the blood samples (Hamaoka et al., 2019). Moreover, patients with 5 or more CTCs showed shorter disease-free survival compared to patients with fewer GPC3-positive CTCs. Another study by Court et al. (2018) analyzed GPC3 in a capture cocktail together with antibodies against asialoglycoprotein receptor (ASGPR) and EpCAM. Importantly, the combined capture approach with all three targets, isolated higher CTC numbers in patients than each antibody alone. This approach detected CTCs in 96.7% of all HCC patients (n = 61) with a median of 6 CTCs per 4 mL blood. In contrast, in healthy controls (n = 8) maximum one potential CTC hit was found. Moreover, CTC numbers were increased in more advanced stages compared to early stages. This effect was even more pronounced for the subfraction of vimentin-positive CTCs, which presumably are generated by EMT processes. This highlights the importance of CTC capture strategies independent of potential EMT target proteins like EpCAM, which are often downregulated during EMT. Indeed, varying or low EpCAM expression has already been reported for CTCs originating from other cancer entities (Hyun et al., 2016; de Wit et al., 2018) and should be taken into consideration when designing or interpreting CTC capture assays.

In summary, GPC3 is currently evaluated as a therapeutic target, serum biomarker, and importantly for CTC analyses, where it has been used both for capture (Court et al., 2018; Hamaoka et al., 2019) and identification (Ogle et al., 2016) of CTCs in HCC. Since only few studies have been performed, further studies are needed to prove the feasibility of GPC3 for clinical CTC analyses. Although GPC3 is a well-established diagnostic marker for HCC, further characterization or validation of the potential GPC3-positive CTC hits, for example via molecular analyses, is to our knowledge missing so far.

As described, there is a great diversity among the proteoglycans associated with different cancer types. The different proteoglycans facilitate distinctive processes in the metastatic cascade and their universal expression suggests that proteoglycans are an essential feature for all cancers. The complexity is further expanded when considering the GAG composition. An increasing number of studies indicate that at least some of the functions of proteoglycans are exerted through specific GAG chains. However, only a few studies focused on targeting the GAG part or GAG composition of proteoglycans when isolating CTCs.

A wide variety of qualitative and quantitative methods has been developed for studying glycocalyx components. As the biosynthesis of glycans is non-template driven and complex, their analysis may often be challenging. Several approaches take advantage of the large repertoire of glycan-binding proteins and antibodies to distinguish between different glycan classes. For large screenings, glycan microarrays have been developed that may probe for different glycan classes or subclasses. This approach has been used to screen breast CTCs for glycan markers, which identified a specific O-glycan epitope as a potential target (Wang et al., 2015). Microarrays and cell-based libraries have also been developed to screen for GAG-binding proteins (Rogers and Hsieh-Wilson, 2012) and these may be useful for identifying GAG-based CTC targeting reagents. However, microarrays for detection of cell-surface GAGs, which could be useful for identifying GAG biomarkers on CTCs, have not been constructed. CS and HS-specific antibodies, such as CS56 and 10E4, and GAG-binding proteins, like fibroblast growth factor, are also commonly used in flow cytometry and microscopy-based assays (Figure 6A). These may assess the relative levels of GAGs, however, they do not convey specific structural information due to their low specificity or affinity toward their targets (Yamagata et al., 1987; Smetsers et al., 2004; ten Dam et al., 2007). For this, GAGs will have to be isolated and analyzed, often by chromatography, mass spectrometry, or nuclear magnetic resonance. The structural characterization of GAGs is challenging due to heterogeneity of the polymers. Hence, analysis is often limited to disaccharide analysis, which does not allow for sequencing of intact GAG chains. This is even more technical challenging for CTCs because of the limited input material due to their low abundance. Similarly, while different proteoglycan core proteins can be probed with antibodies, fine structural analysis of their GAG attachment sites is only achieved by glycoproteomic methods. While these analyses may be laborious, they are highly descriptive and may provide novel insight into structural alterations on cancer cells, both on the protein and GAG level. For example, one study found that several major ECM proteoglycans had elevated levels of N-glycosylation in pancreatic cancer tissues (Pan et al., 2014). In addition, another study identified novel CS linkage region modifications in CS glycopeptides from the inter-α-trypsin inhibitor complex, which is abundant in plasma from cancer patients (Gomez Toledo et al., 2015). To our knowledge, glycoproteomics has not been used for analysis of CTCs, and could potentially help identify novel targets.

If succeeding in finding specific binding moieties, the GAG chains would be an alternative novel approach for CTC enrichment or detection. We have previously shown the use of the recombinant VAR2CSA malaria protein (rVAR2) as a novel CTC-targeting reagent (Figures 6A, 7) (Agerbaek et al., 2018; Bang-Christensen et al., 2019; Sand et al., 2020). rVAR2 binds to a distinct type of CS, termed oncofetal CS, expressed by placental as well as cancer cells (Salanti et al., 2015). The native VAR2CSA binds to a specific CS oligosaccharide motif in the placenta during normal physiological conditions (Gowda, 2006; Ayres Pereira et al., 2016; Toledo et al., 2020). A study using a library of cells with knockouts of GAG biosynthesis genes, indicated that 4-O-sulfated CS is essential for rVAR2 binding (Chen Y. H. et al., 2018). The specific oncofetal CS-carrying proteoglycans have been examined by screening of rVAR2 binding to more than 3500 cell surface proteins (Salanti et al., 2015) as well as by rVAR2-affinity chromatography coupled to glycoproteomics, using tumor and placenta samples (Toledo et al., 2020). These studies showed that the distinct oncofetal CS is displayed on multiple proteoglycans such as CSPG4 or CD44 in cancer cells, indicating an important function of oncofetal CS in the disease development. Moreover, rVAR2 binds to cancer cells independently of tumor origin and oncofetal CS is expressed both in primary and metastatic lesions (Salanti et al., 2015). This has also been shown in a metastatic murine model, where rVAR2 binding furthermore inhibited integrin signaling and seeding of CTCs (Clausen et al., 2016). As studies have also indicated that rVAR2 binds to cancer cells independent of EMT processes (Agerbaek et al., 2018; Bang-Christensen et al., 2019), oncofetal CS could be an advantageous target for CTC enrichment. In line with this, rVAR2-coated paramagnetic beads have been used to capture CTCs from blood samples of different carcinoma patients (n = 44) and glioma patients (n = 10) in small proof-of concept studies (Agerbaek et al., 2018; Bang-Christensen et al., 2019). Therefore, the rVAR2-based approach offers an alternative capture approach, demonstrating how GAG-targeting can allow the capture of CTCs independently of single target proteins, like EpCAM.

Actually, the reversed approach can also be utilized for CTC capture (Figure 6B). For example, Gopinathan et al. (2020) coated a synthetic HS-based octasaccharide probe (SCH45) onto magnetic beads, that were used in combination with a microfluidic chip to isolate CTCs from 65 advanced or metastatic cholangiocarcinoma patients. Single CTCs or CTC clusters were detected in all samples with ≥ 1 CTCs per 1 mL blood, even in patients with no distant metastases. Previous comparable CellSearch^®-based studies found only 17% of CCA patients positive for ≥2 CTCs per 7.5 mL blood (Yang et al., 2016). However, only three healthy controls were included in the HS-based study. Moreover, the authors reported that studies evaluating whether this approach could be employed to capture EpCAM-negative CTCs are currently ongoing (Gopinathan et al., 2020). In addition it would be interesting to identify the binding target of the SCH45-coated beads in order to characterize the captured CTC population.

Another approach exploited the GAG-receptor function of CD44 in order to capture CTCs (Wang M. et al., 2018). Purified HA, the ligand of CD44, was coated to a microfluidic chip and showed 91% retrieval of CD44-overexpressing A549 cells spiked into blood. Also other cancer cell lines from different cancer entities were captured with comparable efficiencies. Although also lacking healthy controls, the study found between 1–18 putative CTCs per 1 mL blood from 9 of 10 NSCLC and 5 of 5 breast cancer patients as detected through CK- and DAPI-staining.

The utilization of GAGs for CTC technologies is a relatively new approach. Most studies have been limited to smaller pilot studies so far and further molecular characterization of the putative CTC hits is needed to prove their cancer-origin and thus the reliability of the CTC assay. Clearly, the establishment of specific GAGs as biomarkers for clinical CTC diagnostics needs extensive validation in large-scale studies in the future. However, GAGs have the potential to capture or identify broader and more heterogenous CTC populations as they are often independent of a single protein and thus might be less prone to gene expression changes associated with different or transient cancer cell phenotypes.

Circulating tumor cell analyses have the potential to allow prognostic and predictive insights by convenient liquid biopsies. However, novel biomarkers are needed to enable the necessary assay sensitivity and specificity to detect CTCs. Another unsolved problem is that most CTC assays introduce biases in regards to which CTC subpopulations can be captured as they are often based on single biomarkers. Therefore, a CTC capture approach based on a combination of several biomarkers could be beneficial. Another solution for this problem could be the targeting of cancer-specific changes in the GAGs (the “GAGome”), or other known glycocalyx components, which should in principle, allow the capture of more heterogenous CTC populations. Studies on clinical CTCs and their proteoglycans, GAGs, or general glycocalyx structure are still not strongly represented, probably due to associated technical challenges of glycocalyx characterizations. However, structural insights would be beneficial for improving or defining novel CTC capturing strategies based on proteoglycans or their GAGs and to explore whether these strategies then better reflect the heterogenic cancer cell population.

TA, SB-C, AJ, CL, CS, NS, and MA conceived and drafted the manuscript. TA designed the figures with scientific input from co-authors. All authors critically revised the manuscript.

TC, AS, and MA are shareholders in VAR2 Pharmaceuticals holding the intellectual property rights to use rVAR2 for binding oncofetal chondroitin sulfate and to capture circulating tumor cells. VarCT Diagnostics is a subsidiary to VAR2 Pharmaceuticals.

The remaining authors declare that the research was conducted in the absence of any other commercial or financial relationships that could be construed as a potential conflict of interest.