DNA sequencing and proteomics validate tumour-derived cell lines
DNA sequencing was performed to confirm that primary tumour tissue-derived cell lines carried over key mutations present in the parent tumour and did not diverge too far from the parent tissue by accumulating new mutations, in order to assess their suitability as in vitro models of disease.
Inactivating mutations to the APC gene are the most common mutations in CRC (3, 4). They lead to constitutive Wnt signalling, a process considered to be the initiating factor in CRC development (3). Along with MMR genes such as MSH2 and MLH1, APC can be used as a predictive marker of CRC development (3). Of the 32 individual mutations to the APC gene detected across the 7 CA tissues, 26 were conserved in the tissue-derived cell lines. There were 4 instances of a mutation detected in only 1 out of 7 CA tissues being lost in the cell line derived from it, and one instance of a mutation arising in only 1 of the 7 cell lines but absent from all other samples, suggesting that these mutations are likely to be sequencing errors. There were two instances of a mutation being detected in both the tissue and cell line of only 1 patient, and these are assumed to be true low-frequency or passenger mutations.
TP53 usually suffers from biallelic deactivation in CRC – one copy mutated and the other lost in a chromosomal deletion (17q), and this loss is associated with malignant transformation (3). It has been suggested that CRC tumours with MMR defects usually retain a wild-type TP53 (3), however the 4 samples in this cohort with the c.215 C>A TP53 mutation also carried mutations in the MSH2, MSH6 and/or MLH1 genes, suggesting a hypermutated phenotype (4). MSH6 mutations in particular are prevalent in the hypermutated phenotype (4), and these mutations were very common in the LGCA samples (Additional File 3).
Interestingly, the LGCA tumours contained more mutations within these CRC-related genes than HGCA tumours; however, HGCA tumours had larger TMB scores (Additional File 4). This indicates that mutations in these CRC-related genes lead to CRC initiation, and that once the tumour is established and there are mutations present within the MMR genes, a range of other mutations arise that lead to progression of the tumour from LGCA to HGCA or get carried through as passengers.
The high degree of similarity between the tissues and cell lines, in terms of conservation of mutations in key CA-related genes and overlap of their proteomes, lends support to the idea that the cell lines are a reasonable representation of the parent tissue and a suitable in vitro model system. A recent study used a similar methodology to assess the conservation of mutations in primary cell lines derived from meningioma tissues (54).
The proteomes of the LGCA cell lines were less variable than those of the LGCA and HGCA tissues (Additional File 7), reflecting the heterogeneous mix of cell types in tissue, whereas the cell lines are comparatively homogenous. This aligned with the DNA sequencing data, which detected greater variability in terms of mutational burden between tissue samples than between cell lines. The tissue samples include muscle, blood vessels, immune cells and fat, and the process of tumourigenesis leads to large changes due to the loss of normal architecture and function as a dense bulk of tumour cells forms. Accordingly, the CA tissues exhibited significant decreases in proteins involved in muscle structure and contraction (CAV1, CAV2, ANK2, CNN1, TPM1, TPM2, MYL9, MYH11) that were not reflected in the cell lines. The downregulation of muscle system processes was also reported by Vasaikar et al. (27).
Colon adenosarcoma tissues and tissue-derived primary cell lines reveal unique biological functions of tumours
Analyses of the CA tissues revealed that the transition from NC to LGCA was most strongly characterised by increased preribosome activity, RNA exosome complex components and lipid synthesis and decreases in cytoskeletal organisation, caveolar signalling complex and normal colonic function, whereas the progression from LGCA to HGCA was characterised by increased immune signatures and decreased metabolism of nitrogen, tyrosine and drugs.
The caveolar macromolecular signalling complex is comprised of caveolins, which change membrane topology to form caveolae. These membrane structures facilitate trafficking and signal transduction (57). It has been previously reported that caveolin 1 (CAV1) loss is associated with malignant transformation in the colon, and that re-emergence can increase invasion and metastatic potential (58, 59). Accordingly, we saw a significant reduction of CAV1 in LGCA and HGCA tissues relative to matched NC tissues and significant upregulation in CAV1 isoform 2 in HGCA-derived cell lines relative to LGCA-derived cell lines. Furthermore, CAV2 was significantly down regulated in LGCA and HGCA tissues relative to NC, and non-significantly upregulated in HGCA-derived cell lines CAV2 usually forms a heterodimer with CAV1, so CAV2 loss may be a result of CAV1 loss leading to CAV2 degradation (60). CAV3 is a muscle-specific isoform and its loss leads to muscle degeneration (61). Interestingly, CAV3 was not detected in any tissues, but was significantly overexpressed in HGCA-derived cell lines relative to LGCA-derived cell lines. CAV3 has no previous links to CA, but its expression would be expected to decrease in colon tissues during tumour development as normal muscle structure and function is lost. Further investigation is required to determine the importance of CAV3 in CA.
Lipid biosynthesis has previously been linked to CSCs in colon cancer (62). Cholesterol biosynthesis is upregulated in patient-derived spheroid cultures and required for the maintenance of pluripotency (63). Cholesterol seems to promote CSC activity and persistence by negative regulation of TGF-β signalling, and inhibition of cholesterol synthesis has been suggested as a treatment option for CA (63). Similarly, steroids, specifically cholesterol-derived oxysterols, can influence Wnt and MAPK signalling, which are vital to CA development and progression and CSC function (64). Interestingly, CAV1 loss, as described above, causes fatty acid and cholesterol levels in the circulation to increase and the amount stored in the liver to fall due to the impairment to caveolae function at the basolateral membrane of enterocytes in the intestine (65). The upregulation of sterol and cholesterol biosynthesis in the LGCA tissues relative to their NC controls verifies previous findings that indicate their importance in CA, and particularly in CSC activity.
Mucins are highly-glycosylated proteins generally associated with epithelial cells (66). Aberrant glycosylation and expression of mucins is associated with CA (67, 68). MUC1 is a transmembrane mucin overexpressed in CA with links to NF-kB-mediated inflammation (69), CSC phenotypes (70) and metastasis (71), and it is under investigation as an immunotherapy target (68, 72). In contrast to MUC1, MUC2 suppresses CA and its loss leads to CA development (67). MUC2 and MUC5B are secreted mucins and are the predominant mucins comprising the mucus layer that forms on the lumenal surface of colon epithelial cells (73). MUC5AC upregulation has been linked to the serrated pathway of colon tumourigenesis (74). In line with these reports, MUC1 was significantly upregulated and MUC2 significantly downregulated in HGCA tissues relative to LGCA tissues, MUC5AC was only detected in CA tissues and not NC tissues, and MUC5B was significantly upregulated in HGCA tissues relative to LGCA tissues (Table 3). These findings collectively validate mucin expression trends in CA and strengthen the rationale for investigating their utility as biomarkers or potential treatment targets.
Analysis of the cell lines revealed the upregulation of ferroptosis and interferon-gamma (IFN-g)-mediated signalling pathways in HGCA-derived cells relative to LGCA-derived cells (Figure 1C, Additional File 10).
Ferroptosis is a recently recognised form of regulated cell death by iron-dependent lipid peroxidation that causes the cellular membrane damage and the accumulation of reactive lipid hydroperoxides (known as lipid-ROS) to lethal levels (75). Long-chain-fatty-acid-CoA ligase 4 (ACSL4) is an important modulator promoting ferroptosis via enhancing lipid peroxidation (76, 77). ACSL4 catalyses the esterification of free fatty acids, preferentially polyunsaturated fatty acids (PUFAs), and incorporates esterified PUFAs into phospholipids within the cell membrane, creating substrates for lipoxygenases for lipid peroxidation. Ferroptosis can be antagonised by neutralising lipid-ROS by coupling the oxidation reaction of glutathione (75). Glutamate-cysteine ligase (GCL), composed of catalytic (GCLC) and modifier (GCLM) subunits, is the rate-limiting enzyme in the glutathione biosynthesis pathway (78). Cellular iron is another factor regulating ferroptosis via the Fenton reaction to produce lipid-ROS, and acts as a cofactor for lipoxygenases that catalyse lipid peroxidation (79). When in circulation, iron forms a complex with transferrin, which binds the transferrin receptor protein-1 (TRFC) on the cell membrane to be taken up by the cell (79). Excess iron can be stored as ferritin or exported from the cell. Ferroptosis can be triggered when there is an excess of iron stored within the cell or when iron uptake is increased (79). In addition, there is evidence that p53 is involved in the regulation of ferroptosis. Xie et al. (80) reported that p53 loss increased the sensitivity of CRC cells to erastin-induced ferroptosis. This was due to the interaction between p53 and dipeptidyl-peptidase-4 (DPP4) – this complex translocates into the nucleus where DPP4 can act as a transcription cofactor. However, in TP53 mutants where p53 is reduced or absent, DPP4 forms a complex with NOX1 to promote lipid peroxidation. TP53 is one the most commonly mutated genes in CRC, and at least one mutation was detected in both the tissues and cell lines from 5 of the 7 samples (Additional File 3). Therefore, in many CRCs there will be a reduction or loss of p53 and a heightened capacity for ferroptosis, supporting the proposal that ferroptosis induction could target CRC cells. All the above results together suggest that ferroptosis may be an effective target for cancer therapies for CRC.
IFN-g is a cytokine produced by immune cells in response to other cytokines or antigen stimulation to drive immune responses (81). It signals via the IFN-g receptor, which is expressed by most, if not all, cell types. Binding of IFN-g to its receptor initiates JAK-STAT signalling, primarily through STAT1 which binds conserved DNA elements called INF-g activation sites to induce the transcription of interferon-stimulated genes (81). The products of these genes regulate chemokine production, MHC molecules, antiviral and antibacterial factors, the function of regulators of metabolism, chromatin and transcription (81). IFN-g upregulates the production of proteasomal subunits and the formation of immunoproteasomes, which produce peptides that bind more efficiently to MHC class I molecules (82). In this way, IFN-g can increase the amount of antigen being presented to immune cells to build an immune response. The IFN-g-induced increase in immunoproteasomes and MHC class I components may allow for greater presentation of cancer antigens to help raise an anti-tumour immune response, and this might indicate a greater chance of immunotherapies being successful. IFN-g signalling was enriched in the HGCA-derived cell lines relative to LGCA-derived cell lines due to significant upregulation of guanylate-binding protein 1 (GBP1), vascular cell adhesion protein 1 (VCAM1), MCH class I components HLA-A and HLA-B, and CD44.
The LGCA tissues displayed decreased cell adhesion (FN1, VTN, VCAM, CEACAM5/6) relative to NC tissues (Figure 1A; Additional File 8B). Conversely, HGCA-derived cell lines displayed upregulation of proteins related to adhesion (CAV1/3, VCAM1, CD44) and cytoskeletal binding (FN1, CDH2) relative to the LGCA-derived cell lines, as previously reported (27) (Figure 1C; Additional File 10A). The observation of differences between the tissues and the cell lines in terms of adhesion and cytoskeletal binding are most likely due to the differences in the costs and benefits of adhesion between in vivo (tumour cell invasion and metastasis) and in vitro (growth as an adherent monolayer) conditions and may be influenced by a cell culture setting that lacks the 3D structural microenvironment present in vivo.
Interestingly, the data for FN1 and for MCH class I components from the tissues and cell lines were inconsistent. FN1 is one of many ECM proteins with aberrant expression in cancer, where it is associated with angiogenesis, invasion via matrix metalloproteinase activation, self-renewal, proliferation, and resistance to therapy, and high expression correlates with poor survival (83, 84). One study demonstrated that silencing FN1 leads to increases in apoptosis-related proteins and reduced NF-kB, suggesting that FN1 overexpression in CA may aid tumour cells to evade apoptosis and to resist therapy by increasing NF-kB anti-apoptotic signalling (84). Interestingly, in line with this finding, proteins involved in the positive regulation of NF-kB signalling such as ferroptosis-related proteins TRFC and HMOX1 were enriched in the HGCA-derived cell lines, despite FN1 being downregulated (Figure 2; Additional File 10B). MHC components HLA-A24, which preferentially presents tumour antigens with an aromatic residue at position 2 and a non-hydrophobic residue at the C-terminus, and HLA-B41, which displays self-peptides with Glu at position 2, were downregulated, whereas HLA-A69, associated with abnormal immune cell accumulation and suppression of the presentation of specific antigens, was upregulated; these factors all suggest a mechanism of immune avoidance by the cancer cells.
The Wnt signalling pathway is heavily implicated in CSC function, and is considered the first pathway to be altered in CA development (3). APC membrane recruitment protein 3 (AMER3), a positive Wnt signalling effector (85), was surprisingly downregulated in the cell lines and not detected in the tissues (Table 3, Additional Files 5 and 6). This may be because the Wnt signalling pathway is already constitutively activated in CA, making these enhancers redundant. Secreted frizzled-related protein 4 (SFRP4), which directly interacts with Wnt proteins and has been suggested as a marker of early-onset colon cancer (86), was overexpressed in CA tissues (Table 3). Furthermore, the cancer-associated scaffold protein syntenin-1 (SDCBP) is known to interact with Wnts (87) and is associated with colon CSC expansion, migration and chemoresistance (88); it was upregulated in HGCA tissues and non-significantly increased in the HGCA cell lines (Table 3). CD44 is of particular interest in CRC as it is considered a CSC marker and its transcription is partially mediated through Wnt signalling (89, 90). CD44 upregulation was observed in HGCA-derived cell lines relative to LGCA-derived cell lines (Figure 3; Table 3). CD44 may become a useful prognostic biomarker by aiding in tumour grading and estimating CSC presence. Structural maintenance of chromosomes protein 2 (SMC2) is involved in chromosome stability and DNA packaging as a component of the condensin complex (91-93). The SMC2 gene is a Wnt signalling target, and miRNA silencing of SMC2 reduces intestinal tumour cell proliferation (93). DNA supercoiling is vital to embryonic stem cell survival and SMC2 has been explored as a CSC-specific therapeutic target (92). SMC2 displayed a fold-change increase of approximate 2.5 from LGCA to HGCA, with no detection in the NC cell line or tissues (Table 3), suggesting that its expression may be related to CA initiation and its upregulation related to progression.
Increases in the LGCA and HGCA tissues, as well as the HGCA-derived cell lines, of proteins involved in immune system processes and drug metabolism (S100A8/S100A9, HLA-A, HLA-B, HLA-DRB1, HLA-DRB5) suggests that changes in the way the cancer cells interact with the immune system, in terms of suppression or avoidance of immune cells and tolerance of inflammation, is involved in tumour initiation and progression. The comparison of the proteomes of CA tissues with patient-matched tissue-derived primary cell lines provides unique insight in this respect. The cell lines, which are a purer population that retain the mutational signatures of the original tumour tissue, represents a unique and powerful method of analysing changes that are more relevant to the signalling within and between tumour cells without being overwhelmed by the large-scale changes occurring across the complexity of a tissue sample. For example, the HGCA cell lines revealed enrichment for IFN-g signalling and ferroptosis not seen in the comparison of NC, LGCA and HGCA tissues, which instead revealed less specific immune-related changes (e.g., “immune system process”, “leukocyte-mediated immunity”).
Overall, this suggests that within the tissue there are many unique physiological and structural changes occurring that are important aspects of the loss of normal function and the response of the body to the tumour. The primary cell lines supplemented these findings by revealing changes in proteins involved in immune interactions and ferroptosis that went undetected in the tissues.
Potential biomarker candidates and therapeutic drug targets for colorectal cancer
This study has identified various proteins of interest that warrant further research as potential CA biomarkers, as well as validating some that have previously been identified (Table 3). Proteins such as CAV1, CAV2 and CAV3, CA2, CEACAM5/6, CD44, CNN1, FN1, GBP1, HMOX1, HPGDS, MUC1 and MUC2, S100A8/S100A9, SDCBP, SMC2, and SFRP4 were significantly differentially expressed and have been previously reported in CA. However, a range of new protein biomarker candidates for CA including ACE2, ACSL4, AMER3, ANK2, CAV3, EXOSC1, EXOSC6, GCLM, LDB3, MUC5B, and TRFC were found in this study.
ACE2 is a component of the renin-angiotensin system (RAS) that catalyses the production of Mas receptor (MasR) ligands Ang1-9 and Ang1-7 (94). MasR signalling reduces inflammation and susceptibility to cardiovascular diseases (95). ACE2 loss has therefore been predicted to be a marker of poor prognosis in CA (96, 97). However, it was detected in only one NC tissue but found at high abundance in all LGCA and HGCA tissues (Additional Files 5 and 6). The role of ACE2 and the MasR in cancer is still unclear, with reports of decreased ACE2 in breast and pancreatic cancers (97), but contradictory reports of ACE2 and MasR overexpression in CRC, and MasR-mediated cancer cell migration in renal cell carcinoma (97, 98). The Human Protein Atlas supports the finding of elevated ACE2 levels in CRC and suggests that ACE2 is overexpressed in renal, pancreatic and liver cancers (https://www.proteinatlas.org/ENSG00000130234-ACE2/pathology). It is possible that ACE2 plays a role in CRC outside of the RAS, or that the outcomes of MasR signalling depend on their physiological context. The importance of ACE2 in CRC deserves further research.
Ankyrins are adapter proteins that organise integral membrane proteins including cell junction proteins and cell adhesion molecules, ion channels and transporters by anchoring them to the spectrin-based membrane skeleton within the cell (99-101). In particular, ANK2 is known to bind and immobilise L1CAM, attenuating its role in axon outgrowth (102). ANK2 was present in the network of proteins significantly downregulated in LGCA tissues relative to NC. In this network, ANK2 was linked with CACNA1D, a subunit of voltage-dependent calcium channels that confers the “long-lasting (L-type), high-voltage activated” phenotype to the channel. These channels facilitate the movement of calcium ions into cells to allow calcium-dependent processes to occur, including muscle contraction. The loss of ANK2 could reflect the physiological changes that occur in the gut during cancer development, including the loss of muscle in the tumour. Alternatively, low levels of ANK2 would lead to a reduced capacity to anchor other integral membrane proteins to the spectrin skeleton, preventing the collection of multiprotein complexes. This may cause aberrant signalling, either due to members of signalling pathways being unable to group efficiently within the membrane, or alternatively, by allowing proteins that require anchoring to isolate them from a signalling complex to move freely through the membrane and interact with their partners. Defects in such signalling pathways as a result of downregulated ANK2 may contribute to CA development or progression. ANK2 has not yet been associated with CA in the literature and may represent a new biomarker for CA development. Our finding that ANK2 is significantly downregulated in LGCA tissues relative to matched NC suggests that impaired ability to localise and stabilise other transmembrane proteins and interact with adhesion molecules may be implicated in CA progression.
ACSL4 was significantly upregulated in HGCA cell lines relative to LGCA cells lines (Table 3), suggesting that these cells have an elevated potential to undergo ferroptosis (Additional File 10A). Induction of ferroptosis to target therapy-resistant CRC tumour cells has been proposed (76), with ACSL4 potentially representing a novel biomarker for success of this treatment.
TFRC was found to be significantly upregulated in the HGCA cell lines relative to the LGCA cell lines (Table 3), indicating the increased capacity of these cells to take up iron, possible overaccumulation of iron and subsequence of ferroptosis. Furthermore, heme oxygenase 1 (HMOX1), which increases cellular iron levels by metabolising heme, was also found significantly upregulated in these cells. HMOX1 has previously been shown to be expressed in CRC tissues and cell lines (103) and to promote erastin-induced ferroptosis (104, 105). TFRC has not been previously linked with CRC but may represent a useful biomarker in this context
GCLM was significantly upregulated in HGCA cell lines relative to LGCA cell lines (Table 3), indicating that the HGCA cells have a higher capacity for glutathione production to defend against ferroptosis induced by ACSL4, TFRC and HMOX1. Though there is currently no established link between GCLM and CA, this suggests that GCLM could be a potential target for ferroptosis-inducing therapies for CA. GCL inhibition by buthionine sulfoximine has been demonstrated to induce ferroptosis in cultured pancreatic cell lines (106).
LIM domain-binding protein 3 (LDB3), also known as Z-band alternatively spliced PDZ-motif protein (ZASP) or protein cypher, is essential for myofibrillar development and muscle contraction by maintaining the Z-line (107, 108). It was detected in LGCA tissues but not expressed in HGCA tissues (Table 3). While LDB1 is associated with cell proliferation and drug resistance in CRC (109), LDB3 has no current links with CA and its loss of expression in HGCA may be useful as a marker for CA progression.
The CA tissues displayed increased levels of preribosomal components (CIRH1A, NIP7, BYSL) and RNA processing members (EXOSC1, EXOSC6). U3 small nucleolar RNA-associated protein 4 (CIRH1A/UTP4) strongly promotes CA cell proliferation and reduces apoptosis (110); it was not detected in NC tissues but was present in CA tissues, and may represent a potential biomarker for CA initiation and a therapeutic target to reduce tumour proliferation. Exosome complex component 6 (EXOSC6) is a non-catalytic component of the RNA exosome complex that performs RNA processing and degradation (111). It is important for cell growth and division and has been implicated in the progression of breast cancer (112). However, its role in CA is undefined. Based on its low expression levels in the NC and stepwise increases in abundance in LGCA and HGCA tissues, it may be a candidate for further research as a potential marker of CA progression.
Calponin-1 (CNN1) regulates smooth muscle contraction by binding actin, calmodulin and tropomyosin (113, 114). Downregulation of CNN1 leads to a loss of membrane integrity in smooth muscle, the uterus and peritoneum, causing blood vessels to become leaky and allowing cancer cell intravasation (115, 116). CNN1 is considered cancer-suppressive, and indeed it is known to be downregulated in cancers (115, 117). It is thought to be a better marker than αSMA for smooth muscle cell differentiation (117). Tumour vasculature has an immature phenotype characterised by incomplete pericyte coverage, irregular shapes and growth patterns, and permeable membranes, which may in part be due to CNN1 loss (117). This allows tumour cells to enter vessels and metastasise (117). In the CA tissues and tissue-derived cell lines, CNN1 levels fell in a stepwise manner from NC to LGCA and LGCA to HGCA. Collectively, CNN1 loss may indicate the initiation of CA and it may represent a useful predictor of invasion and metastatic potential in CA.
Carbonic anhydrases catalyse the interconversion of CO2 and bicarbonate and are crucial for pH maintenance (118). Carbonic anhydrase 1 (CA1) and 2 (CA2) are found in the cytoplasm of NC mucosal cells (119). CA2 is used as a marker for enterocytes in the human colon (120). CA2 overexpression has been associated with poor prognosis and higher grade in CA patients and is correlated with local and lymphatic invasion in rectal cancer (118, 121). Alternatively, CA2 loss has been proposed as a biomarker for CRC due to its downregulation in tumours relative to the NC and its inhibitory effect on proliferation when its expression was induced in CRC cell lines (122). CA1 is not detected in most CA cases (119). Both CA1 and CA2 were found to be expressed at low levels in the NC tissues with a non-significant increase in LGCA tissues and subsequent significant reduction in HGCA tissues relative to LGCA and NC, which was also reflected in the HGCA cell lines. This hints at a potential role for CA1 and CA2 loss as markers for progression to HGCA, possibly through loss of enterocytes, meriting further investigation. Similarly, CA9 has been detected in CA and is associated with intestinal stem cells, cellular proliferation and colon carcinogenesis (118, 123). CA9 was not detected in the CA cell lines or in NC tissues but was present in LGCA and HGCA tissues, providing further evidence that its expression may be related to CA development.
S100A8 (MRP8) and S100A9 (MRP14) are commonly found as a heterodimer called calprotectin that binds Ca2+ and Zn2+ ions and plays an important role in inflammation caused by infection, autoimmunity or metabolic diseases (124). Together, S100A8 and S100A9 account for almost half of the total protein content of neutrophils and are released to induce chemotaxis of leukocytes, cytokine release and apoptosis (124-126). Expression of S100A8 and S100A9 is upregulated during inflammation, a key aspect of CRC (124). Accordingly, other studies have reported significant overexpression in CRC (21, 27). Similarly, the abundance of S100A9 here was found to increase in a stepwise manner from NC to LGCA and LGCA to HGCA (Figure 4B). Unsurprisingly, they were not detected in the cell lines because they are not produced by the cancer cells themselves, but by neutrophils recruited to the site of the tumour. The detection in the blood of S100A8/S100A9 released by neutrophils in the tumour microenvironment suggests they could be utilised as a serum biomarker for diagnosis or prognosis of CA.
GBP1 expression is induced by IFN-g, and offers broad protection against pathogens via oxidative killing and antimicrobial peptide delivery to autophagosomes (127). The effects of GBP1 on prognosis appear to be context specific. It is overexpressed in inflammatory bowel disease where it is thought to prevent apoptosis in order to maintain barrier function (127). Similarly, it restricts proliferation and apoptosis in cancer, which can establish a senescent, therapy-resistant state (128). Aberrant overexpression of GBP1, induced by INF-g in the inflamed bowel, may be a marker of cells capable of developing resistance to therapy via reduced proliferation and apoptosis.
Hematopoietic prostaglandin D synthase (HPGDS) was expressed at high levels in the NC tissues, but not detected in LGCA or HGCA tissues (Table 3). HPGDS is an enzyme that converts prostaglandin H2 into prostaglandin D2 that functions in smooth muscle contraction and inhibition of platelet aggregation (129). It uses glutathione (GSH) and either Ca2+ or Mg2+ as cofactors (130). In the colon of mice, HPGDS has been shown to reduce inflammation and colitis-associated cancer (131, 132), and therefore its loss may be a marker for increased CA risk induced by inflammation.
Functional assays, including miRNA silencing or CRISPR knock-outs, would clarify the roles of these potential CRC biomarkers.