Heparan sulfate proteoglycans undergo differential expression alterations in left sided colorectal cancer, depending on their metastatic character

Colorectal cancer (CRC) is a heterogeneous disease and, despite their similar histological aspect, there are substantial differences between left- and right sided CRCs (LSCRCs and RSCRCs), including their etiology, response to screening tests, the stage at which they are diagnosed, and their effect on mortality [1]. At the molecular level, differences in the expression of different biomarkers have been described between LS- and RSCRCs.

Interestingly, several of these markers are related to heparan sulfate proteoglycans (HSPGs) including: p53, related to the regulation of genes as SULF2 and heparanase [2, 3]; MUC1, involved in cell-cell dissociation and invasiveness, in cooperation with HSPGs [4]; the Wnt/β-catenin pathway, regulated by glypican-3 and -4 [5, 6]; cytokines like VEGF, EGF and TGF-beta and other markers, which bind to heparan sulfate (HS) chains, which regulate their activity [7‐9].

HSPGs comprise a specific small group of glycoconjugates composed of various core proteins post-translationally modified with HS glycosaminoglycan (GAG) chains. HS is a complex linear anionic polysaccharide whose synthesis occurs mainly in the Golgi apparatus. It is initiated by the formation of a tetrasaccharide linkage on the protein core after which the HS chain is elongated by the addition of alternating D-glucuronic acid (GlcA) and N-acetyl-D-glucosamine (GlcNAc) residues. Subsequently, a series of different enzymatic reactions act in an orderly manner on specific regions of the molecule, including N-sulfation, epimerization and various O-sulfations. This gives rise to highly sulfated regions (NS domains), which alternate with nonsulfated (NA domains) and mixed (NA/NS) domains [10, 11].

Since function ultimately relies on the fine structure of the chains, cells exercise accurate control over HSPG composition and sequence, which results in these molecules varying depending on factors like cell type and development stage, as well as due to pathological processes. The binding sites for a great variety of ligands, such as cytokines, chemokines, growth factors, enzymes and extracellular matrix (ECM) proteins [11, 12] are defined by specific sets of variably modified disaccharides, usually within the NS domains. These networks of complex interactions at the molecular level mean that HSPGs participate actively in the control of many normal physiological functions [10‐13]. Because of these interactions, HS is also involved in many pathologies, including inflammation, amyloid diseases, infectious diseases and cancer [14]. Given that function is dictated by the fine structure of the chains, the detailed analysis the full set of changes in the expression of HSPG core proteins and HS biosynthetic enzymes in cancer pathologies is of great interest, as is a detailed consideration of the effect of these particular signatures on invasion and metastasis.

Up- or downregulation of genes involved in the biosynthesis of HSPGs have been reported in many cancerous cells [15]. In the case of CRCs, various alterations have been described, for example, those relating to specific syndecans [16], the relative amounts and structure of glycosaminoglycans [17, 18] and the expression levels of certain enzymes involved in HS saccharidic chain structure [19, 20]. However, many of the previous studies reported in the literature analyze this pathology in a general way even though, as indicated above, CRC is a heterogeneous disease with respect to the anatomical location of the tumor.

We recently published the results of a study focused on RSCRCs [21] where, since the presence or absence of metastases in lymph nodes is a key predictor of progression, we subdivided the tumors into two groups according to this important feature. We found that the number of genes affected was higher in non-metastatic tumors, with around 40% of all genes analyzed being involved, and that most of the genes whose expression was altered in metastatic malignances also showed altered expression in non-metastatic tumors. Additionally, the PGs located at the cell surface showed significant differences in expression depending on the presence or absence of metastases, while alterations of those located in the ECM or within the cell were very similar in both tumor types. HS chains seemed to experience far more limited changes in metastatic CRCs than in non-metastatic tumors, while chondroitin sulfate (CS) chains, which are also carried by some HSPGs, were strongly affected, albeit differently in the two tumor groups [21]. In this current paper, we have investigated the expression patterns of all the genes involved in HSPG biosynthesis in LSCRCs, compared with healthy tissues from the same patients. As in the previous study focusing on RSCRCs, the tumors were subdivided into two groups according to presence or absence of metastases in lymph nodes, and the study included genes coding for HSPG core protein and GAG chain synthesis and modification. The aim of the work is to increase our knowledge of structural alterations of HSPGs in LSCRCs, comparing the data to that previously obtained for RSCRCs, in an attempt to define biomarkers which are different in metastatic and non-metastatic tumors which could be useful in the future to develop new chemical biology approaches to retard tumor progression by modulating deregulated biosynthetic pathways.

The abnormal expression of HSPGs in cancer and stromal cells can serve as a biomarker for tumor progression and patient survival [25]. HS fine structure is determined by the cell-type specific expression of only certain isoforms of some of the biosynthetic enzymes, notwithstanding the existence in some specific cases of regulation at the level of translation or enzymatic catalysis [28‐30]. In a previous work, we have described the alterations that take place in RSCRCs which affect both the core proteins of HSPGs and the different enzymes responsible for the synthesis of the GAG chains, as well as the differences in these changes depending on the presence or not of metastasis in malignances [21]. In the present study, we provide a similar analysis focused on LSCRCs in order to determine whether there are any differences between left- and right sided pathologies. As in the previous study, here we have considered the comparative analysis of tumors at the T3 stage, where the muscularis propria is affected, and classified the tumors depending on the presence or absence of lymph node metastasis.

Two gene families, syndecans and glypicans, account for most cell surface HSPGs in humans, along with a few part-time proteins [22]. Transcripts for all syndecan species were detected in LSCRCs, but only syndecan-1 mRNA appeared overexpressed in most tumors, independent of the presence of metastasis. Although some previous studies using colon carcinoma cells have described alterations in the transcriptions of syndecan-2 and -4 [16], our previous work on RSCRCs was only able to detect overexpression of syndecan-1 in metastatic tumors [21]. Interestingly, in this work the analysis of the expression of syndecan-1 protein using immunohistochemistry provided the noteworthy finding that non-metastatic LSCRCs displayed lower immunoreactivity to that detected in normal tissues from the same patients, while the metastatic tumors showed a dramatic decrease in staining. Furthermore, non-metastatic tumors have a certain level of staining in the extracellular matrix, suggesting the shedding of cell membrane-bound proteoglycans. These results are very similar to those previously described in RSCRCs, where we suggested that the expression of syndecan-1 involves additional post-transcriptional mechanisms, such as protein translation, degradation, inhibition by feedback loops or miRNA regulation [21]. There is evidence for the post-transcriptional regulation of syndecan-1 expression in, for example, pancreatic cancer and peritoneal macrophages [30, 31]. Our data also correlate well with other previous immunohistochemical studies which have described a loss of expression of syndecan-1 in CRCs, some of which have been found to correlate with tumor stage and metastasis [32‐35]. Upregulation of syndecan-1 has been described in some types of tumors, and it has been postulated that this aberrant expression may play a key role in promoting growth factor signaling in cancer cells [36]. In contrast, other malignances have been found to show downregulation of this molecule, indicating that this HSPG could well serve as a prognostic marker in a cancer-type-specific manner [36].

The glypican family comprises six cell surface HSPGs that are involved in the regulation of several signaling pathways, where, depending on biological context, they either stimulate or inhibit activity [37]. As such, tumor progression is affected by their activity, with abnormal expression being linked to various human tumors [25]. In the samples analyzed in this study, their relative expression patterns were quite similar to those observed in RSCRCs [21]. However, very few transcriptional changes were detected: in isoforms 4 and 6 in metastatic tumors, where, moreover, these alterations were markedly different from those observed in ascending tumors, where there is a great underexpression of glypican 1 in all types of tumors; and in isoforms − 3 and − 6 in non-metastatic tumors [21]. Unlike the other isoforms, relatively little is known concerning the expression or functional roles of glypican-4 and -6 in tumors. However, the ability of glypican-4 to uncouple pluripotent stem cell differentiation from tumorigenic potential has been recorded [38], while the reduced expression or loss of function of glypican-6 has been described in retinoblastoma and autosomal-recessive omodysplasia [39, 40].

Betaglycan and CD44v3 are part time membrane HSPGs, meaning that they occur either with or without HS chains [22]. Although the expression of CD44v3 in CRCs has been described as being related to more advanced pathological stage and poorer prognosis [41], in this study no statistically significant differences in any type of LSCRCs were found, mirroring our previous findings for RSCRCs [21]. The other part time HSPG analyzed was betaglycan, whose expression in tumor cells appears to play an important role in the progression of the pathology [42]. However, in relation to CRCs, although this molecule appeared underexpressed in non-metastatic RSCRCs [21], in this study no significant differences between tumor and healthy tissues was detected in LSCRCs.

Another cell-associated HSPG is serglycin, which constitutes a separate category since it has the peculiarity of being located intracellularly [23]. Transcript levels of this gene in this work were significantly reduced, both in metastatic and non-metastatic tumors, following a similar pattern to that previously observed in RSCRCs [21]. Serglycin is mainly found in hematopoietic and endothelial cells, and the principal GAG chains found bound to this core protein are CS, except in mast cells where CS type E or heparin may be present, depending on the cell’s origin [23, 43]. Mast cells in LSCRCs were drastically reduced in tumors compared to non-tumor colon mucosa, which could be, at least in part, the reason for the decrease in protein expression. A number of previous studies have described alterations in serglycin in different tumors [44‐46], and it is also worth noting that results analogous to those described in this work, involving both downregulation of transcription and reduction in the population of mast cells, have also been obtained in RSCRCs [21], suggesting that this is a common feature of both CRC types.

Three HSPG species are located at the ECM: agrin, perlecan and collagen XVIII, and the latter two showed significant downregulation in tumoral samples, which is interesting considering that these two species also appeared modified in RSCRCs [21]. Perlecan expression, both at the transcription and the protein level, was diminished in tumors, independent of the presence or absence of lymph node metastasis. Perlecan, a critical regulator of growth factor-mediated signaling and angiogenesis, is an essential element in maintaining basement membrane homeostasis [47], likely indicating that it has a role to play in the progression of CRCs.

Collagen XVIII appeared downregulated to a statistically significant extent only in non-metastatic LSCRCs, while in RSCRCs its expression was significantly reduced in both metastatic and non-metastatic tumors [21]. However, decreases of about 50% were found in 70% of metastatic LSCRCs, a result that approached significance (p = 0.07), leading us to suggest that the results observed could be dependent on the individual sample analyzed, and that the real effect might occur similarly in all CRCs, regardless of their location. Several reports in other malignances describe different types of alterations in collagen XVIII depending on type of tumor, e.g., its expression increases in ovarian and pancreatic cancer, while it diminishes in liver and oral cancer [25].

In summary the patterns of alterations in the levels of expression of HSPG core proteins in CRCs is quite similar for ECM molecules, syndecans and serglycin, independent of tumor location, while glypicans display differences between RS- and LS- malignances.

The tissue-specific expression of individual HSPGs will determine when and where HS chains are expressed. For GAG chain generation, it is necessary to regulate the activity of many different enzymes, mainly GTs, located in the lumen of the Golgi apparatus [10]. The initial step in the biosynthesis of the chains is the creation of a tetrasaccharide linkage region, followed by polymerization through the consecutive addition of alternating GlcA and GlcNAc. A number of works have described variations in HS levels, both increases and decreases, in different tumor types [15, 48], including for CRCs, where decreases have been reported [17, 49]. However, in this work, it was not possible to determine the existence of significant differences in the transcription levels of any of these genes in LSCRCs. This finding contrasts with the results previously described for metastatic RSCRCs, in which B3GAT1 expression decreased, particularly in non-metastatic RSCRCs, where several genes responsible for the synthesis of the linker (XYLT1, XYLT2, B3GAT1) and the polymerization of the chain (EXT1) were downregulated [21].

During HS biosynthesis, various sulfation and epimerization reactions take place which are responsible for the fine structure of the saccharide chain. The first reaction involved in polymer modification is the removal of acetyl groups from GlcNAc residues, after which the amino group is sulfated, catalyzed by four different isoforms of N-deacetylase/N-sulfotransferases [10]. The tissue distribution of NDST1 and NDST2 broadly overlap [50], and transcripts for both were detected in all samples analyzed in this study, with NDST2 appearing downregulated in all LSCRCs, while NDST1 transcription was downregulated in all metastatic tumors and in 60% of non-metastatic. In contrast, previous work with RSCRCs has shown that both isoforms were underexpressed, but only in the non-metastatic patients [21]. NDST4 was undetectable in most samples in the current work, while NDST3 transcripts were detected in only a small percentage of tumors. Expression of NDST3 and NDST4 is principally restricted to the period of embryonic development [51]. That said, in certain tumor types, expression of these molecules has been described, for example, NDST4 in breast cancer [51], although in RSCRCs neither was detected [21].

The next steps in the synthesis of the HS chains include the epimerization of GlcA into IdoA, an action catalyzed by the enzyme C5-GlcA epimerase, along with O-sulfation at C2 of uronic acid [11, 12]. An overexpression in the transcription levels of the two genes involved, GLCE and HS2ST1, was detected in non-metastatic LSCRCs in this study, although not in metastatic tumors, in contrast to our previous study in RSCRCs which found no alterations in the expression of these genes [21].

The addition of an O-sulfate group at C6 is mediated by enzymes encoded by the genes HS6ST1–3, each of which is specific to a particular substrate and differs in its tissue expression [52]. Transcripts for the three isoforms were identified in the healthy tissue studied here, but HS6ST3 mRNAs were not detected in tumor samples, neither metastatic nor non-metastatic. Meanwhile, HS6ST1 appeared downregulated in metastatic LSCRCs. These results show a pattern different from that previously described in RSCRCs, where HS6ST2 was not detected in either tumor type or in healthy tissue, and HS6ST1 appeared deregulated in non-metastatic tumors [21].

The last, and largest, family of enzymes involved in the biosynthesis of HS is the 3-O-sulfotransferases, which comprises seven different members (HS3ST1–6). In LSCRCs, only isoform − 6 expression appeared altered, its expression being diminished in non-metastatic tumors, although it was not detectable in metastatic forms. Again the data differ from those obtained in RSCRCs, where none of the isoforms appeared modified in metastatic tumors, and HS3STB1 and HS3ST5 were underexpressed only in non-metastatic ones [21]. 3-O-sulfation is a relatively rare modification, and it has only been found to influence a small number of proteins thus far [53]. That said, several studies have reported 3-O-sulfation alterations in different tumors [15, 48], and it has been suggested that certain patterns of 3-O-sulfation may be responsible for the appearance of cancerous phenotypics [25].

Once the HS biosynthetic process has been completed, 6S groups present at the glucosamine residues can be post-synthetically edited from the chain, suggesting that it may have special regulatory importance. The reaction is carried out by two endosulfatases located on the cell surface, SULF1 and SULF2 [27], and alterations in the expression of these genes in various tumor types, either up- or downregulation, have been reported [15, 48]. In the case of LSCRCs, transcript levels of none of these genes were found to be altered, mirroring the results previously observed in RSCRCs, where there were also no alterations [21].

Analysis of the HS structure by immunohistochemistry showed differences between normal mucosa and tumors as regards the intensity and distribution of the molecules. It is possible that these differences are caused by structural changes brought about by alterations in the transcription levels of HS biosynthetic enzymes.

Some of the HSPGs analyzed in this study are hybrid molecules, with both HS and CS side chains [54], making it interesting to extend the study to the genes involved in the biosynthesis of this GAG. In addition, changes in CSPGs associated with CRCs, such as versican and decorin, have been found [55], as well as alterations of the CS chains in RSCRCs [21].

CS chain extension takes place through the sequential addition of alternative GlcA and GalNAc residues. Five genes, CSGALNACT1, CSGALNACT2, CHSY1, CHPF and CHSY3 encode the GTs involved in this process [56]. CSGALNACT2 transcription was downregulated in non-metastatic LSCRCs, which coincides with what has been described in RSCRCs, although in the latter the rest of the genes involved, except CHPF, also experienced a decrease in their transcription levels [21]. Interestingly, CHPF appeared overexpressed in all LSCRCs, as it was in metastatic RSCRCs [21]. In other previous studies of CRCs, different changes have been described, although it must be stressed that these studies involved colorectal cancers from different locations and not only RSCRCs [57].

CS repeating disaccharide building units can be modified by epimerization of GlcA residues and by various sulfations [57]. O-sulfation at C4 of GalNAc residues is carried out by enzymes encoded by four genes (CHST11–14), although in this work only CHST11 and − 12 appeared underexpressed in all LSCRCs. This concurs with our previous study in RSCRCs that showed a very similar expression pattern for these genes, although in that case CHST14 was also underexpressed in all tumors, irrespective of their metastatic features [21].

O-sulfation at C6 of GalNAc residues is performed by three different genes, and CHST3 translation was downregulated in non-metastatic tumors, although immunohistochemistry showed the opposite result, i.e. that it was upregulated. This apparent contradiction between the two sets of results is similar to the case described above for syndecan 1, and suggests the involvement of additional post-transcriptional mechanisms [21]. Discordances between mRNA and protein in complex biological samples have been widely analyzed and discussed [58], and subsets of proteins displaying negative correlation with mRNA expression values have been described in some tumors [59]. In addition, CHST7 appeared underexpressed in metastatic LSCRCs in this work. Although very similar patterns of expression for the three genes were previously found in RSCRCs, no alterations in transcription were observed [21].

Both genes encoding the enzymes involved in the modification reactions of uronic acid residues, DSE and UST, were significantly altered, although DSE to a lesser extent, with the difference not reaching significance in metastatic tumors. The alterations observed once again followed a pattern similar to that previously observed in the ascending tumors, where both enzymes appeared downregulated [21]. The alterations in the CS chains as a result of the differences in expression of biosynthetic genes were analyzed by immunohistochemistry using the specific antibody CS-56, with clear differences found in the amount and location of the staining, although it must be taken into account that CS-56 antibody reacts preferentially with CS-D (sulfated at C-2 and C-6), but is also able to recognize other types of structures, including CS-A, -B, -C, and -E [60].

In terms of survival, the small sample size and the retrospective character of this study should be recognized as a considerable limitation, but, despite this, statistically significant differences in the underexpression of two genes were detected, along with a trend in two additional genes, and this behavior seems to be maintained regardless of lymph node involvement or not. Some of the other genes found to be significantly dysregulated in this work might also show a relationship with survival in a bigger sample, therefore prospective studies with a larger study population are necessary.

In sum, analyzing the differential expression of the genes involved in HSPG biosynthesis in LSCRCs has highlighted that some of them showed significant changes in their transcript levels. When comparing the results with those previously described for ascending tumors, similarities can be seen for different gene families, as well as notable differences that evidence the heterogeneous character of the two pathologies. In contrast to RSCRCs, where non-metastatic tumors showed more alterations than metastatic tumors, in the case of LSCRCs the percentage was very similar, around 30%. With regard to HSPG core proteins, the patterns of alterations in expression of ECM molecules, syndecans and serglycin were quite similar, while glypicans displayed differences between RS- and LS- malignances.

The synthesis of the HS chains did not show, in contrast to RSCRCs, differences in the expression of the genes responsible for polymerization, but differences were evident in those controlling sulfation and epimerization. Differences between LS- and RSCRCs were preferentially found in the expression of genes involved in C6 and C3 sulfation of glucosamine, but not in NDSTs or SULFs. Finally, the synthesis of the CS chains showed differences with respect to the RSCRCs which preferentially affected the polymerization of the chains and the 6C sulfation of the GalNAc residue. The differences described in this article can help to understand some of the molecular differences between the proximal and distal colorectum previously described for various biomarkers.