Characteristics of CD44 alternative splice pattern in the course of human colorectal adenocarcinoma progression

In developed countries, colorectal cancer (CRC), or rather its progression to metastatic disease, accounts for 25% of tumour deaths [1]. Characterising the metastases associated processes is therefore of crucial importance for identifying ways of earlier and more sensitive diagnosis, more defined prognosis, and possibly in the selection of patients for targeted therapies.

A variety of genes have been described and extensively investigated in the literature as key candidates in the tumorigenesis and progression of colorectal carcinoma, including APC, p53, K-ras, BRAF, DCC, MSH, EGFR, SFK, TGFR2, SMAD4, etc. [2‐5].

Each individual step of the metastatic cascade is the result of complex molecular interactions, regulated by several already identified and/or unidentified genes [6]. One of the candidates of key importance has been CD44. In fact, it is one of the most investigated molecules in the metastatic process of several malignancies [7‐13], among them that of colorectal cancer [14‐16].

CD44 was first described to be a lymphocyte homing receptor [17]. The role of CD44 was later proven in fetal arteriogenesis [18], as well.

Furthermore, expression and function of various isoforms of CD44 were detected in chronic inflammatory diseases (rheumatoid arthritis, ulcerative colitis, resolution of lung inflammations, etc.) [19‐21].

A number of studies demonstrated the expression of CD44s (standard version) on most tissue types of epithelial origin (stomach mucous membrane, small intestinal mucosa, prostate, ductal epithelium of breast, skin, hair follicules, transitional cell membrane of utgenital tract) [22‐26]. Regarding colonic mucosa, immunohistochemical examinations demonstrated that CD44s (identified with antibody against the standard region of the molecule) is located at the level of basal crypts, similar to the “intestinal stem cells” of the small intestine sitting between Paneth-cells [27‐29]. This means that proliferative cells of colonic mucosa, as well as cells in the basal segment of differentiation zone are able to provide CD44s expression, while differentiated mucosal cells don’t show membrane positivity on immunostaining any more [30]. This fact gains special importance when considering the change of ‘CD44’ expression characteristics from the normal, the dysplastic and the malignant colorectal tissues [31].

Additionally, and probably not independently form the above mentioned findings, recent studies demonstrated ‘CD44’ as a potential tumour stem cell marker (in colorectal, gastric and pancreatic adenocarcinoma, as well as breast cancer and other cancers) [32‐39].

With the presence of CD44 protein in so many tissues, consideration has to be given to its physiological and pathological functions.

‘CD44’ has a variety of biological functions such as cell-cell, cell-ECM interactions, tumour cell migration [40, 41] or even chemoresistance [16, 41‐50]. This multitude of representations and functions is less surprising when considering that through its 9 variable exons (v2-v10), hundreds of different isoforms can be transcribed and translated from the single CD44 gene resulting in a mixture of glycoproteins with potentially different functions [51‐53]. Talking about ’CD44’ and studying its ’over expression’ therefore is rather debatable and not as straight-forward as it was previously thought. A basic understanding and consideration of alternative splicing should be warranted when dealing with this versatile gene. For the same reason, when using probes (i.e. primers or antibodies) against the standard region of CD44 (CD44s), one will detect a mixture of isoforms as they all share the same standard region [54, 55].

Determining the role of ’CD44’ is further complicated by the difficulties to identify all the isoforms potentially characterizing certain cell or tissue types. The trend is to focus on the expression of a single variable exon, which, of course, will only show the summation of isoforms containing the exon in question. However, these isoforms might only share a single variable exon, therefore they can be of different functions. In addition, several other CD44 variant isoforms can be present within the same cell, adding further functions to ’CD44’ [56‐59]. Furthermore, different cells of a tumour can express various, possibly different sets of CD44 isoforms. As most of our tests examine a group of cells from the tissue at the same time, the results are summated, failing to represent the cell-to-cell differences in details [6]. In colorectal carcinoma the most studied variant exons / protein domains are v3 and v6.

CD44v3 is the heparane-sulphate proteoglycane domain of CD44, however, the v3-containing isoforms of CD44 have hyaluronate binding potential and may contain a chondroitin-sulphate proteoglycan domain as well. Consequently, CD44v3 can bind different heparin-binding growth factors (GF), such as HBEGF, VEGF, HGF, βFGF, KGF and amphiregulin [60, 61].

CD44v6 functions as receptor of HGF/SF cooperating with c-met [61, 62], regulating EMT and taking part in the positive feedback loop of K-ras activation driving both Ras signal transduction pathway and CD44-splicing machinery [3, 63, 64]. Additionally, this is the domain most frequently found to be associated with metastatic phenotype in the literature [65].

Unsurprisingly, the studies of detecting ‘v3’ and ‘v6’ under different experimental and clinical settings do not lead to coherent results, and can even be contradictory. This also can be observed in the literature regarding colorectal cancer [14, 66‐71].

Recently, characteristic CD44 isoform patterns of different normal tissue types have been described in the literature [24‐26, 72, 73] and our earlier studies described CD44 isoform patterns characteristic of different tumour types, respectively (data publication in progress, Rásó-Barnett et al.). Most of the literature agrees that ‘CD44’ plays an important role during tumour progression based on ‘overexpression’ of specific variant regions containing isoforms at mRNA and/or protein level. However, no data is available concerning the pattern of CD44 isoforms and its modulation during tumour progression. Our previous study on CD44 alternative splice patterns showed no qualitative change of the melanoma specific ‘CD44-fingerprint’, while the expression level of all isoforms uniformly increased during the metastatic process (data publication in progress, Rásó-Barnett et al.).

The aim of our research was to find answers to the following two questions: (1) Is there a colorectal carcinoma specific CD44 isoform expression combination like to one identified in human melanoma? If there is such pattern, is it qualitatively / quantitatively stable during tumour progression? (2) What is the background of the incoherent quantitative results on the expression of the most investigated v3 and v6 variable exons / protein domains?

To answer the first question, we designed primer pairs covering the entire variable region (primers A-B, see Methods) for our quantitative and qualitative PCR reactions. The expressed, i.e. transcribed CD44 isoforms were identified using direct sequencing and next-generation sequencing.

Due to the large number of potential isoforms, we used a series a PCR reactions to create the CD44 fingerprint, a simple, easy to handle representation of the expressed CD44 isoforms. We used this fingerprint to track the changes of CD44 expression during tumour progression [41, 55, 74‐76].

To answer our second question, we designed a complex experimental animal model to examine the functional role CD44v3 and CD44v6 containing isoforms play in metastasis formation. The main question was whether or not there is a change in the expression of these variable exons that could be associated to the metastatic process of human colorectal carcinoma. We implanted the same colorectal cancer cell population orthotopically to form real metastases in liver, as well as heterotopically. We performed two types of heterotopic implantation: the first involved subcutaneous implantation into permissive and non-permissive hosts, the second was direct implantation into the spleen so that the implanted cells could directly colonize the liver. The latter method created a new type of secondary tumour, which albeit is in the same localization as the liver metastases of the true metastatis model, has not been through the metastatic cascade. This experimental setup allowed to functionally differenciate between phases of metastatic process and host related changes, regarding the expression activity of the examined variable CD44 exons.

Despite hundreds of published studies on CD44 within the last two decades, noconsensus opinion has been reached as of today apart from that it plays some role in tumour progression [69, 76]. One of the possible explanations is that from microarrays to immunohistochemistry, the overwhelming majority of studies fail to take into consideration that ‘CD44’ is a collection of isoforms generated via alternative splicing of its nine variable exons. Probes, primers and antibodies against the standard region will recognise a pool of different isoforms as they all share the same standard region, and not the isoform itself, which is functionally active and involved in the metastatic process. A good example of this is colorectal adenocarcinoma, where the cells of origin, i.e. the glandular epithelium of the large bowel, express standard form of CD44 (CD44s), but not variant ones (CD44v), while highly dysplastic colorectal adenomas, primary and metastatic colorectal cancers express a large variety of different-length CD44 isoforms, both on mRNA and protein levels [14, 31, 44, 77‐79]. Some of these new CD44 isoforms acquired during malignant transformation contain v3 and v6 domain which add a completely different range of functions to the pre-existing CD44s functions, as CD44v6 is a co-receptor of the met protein, and CD44v3 is one of the most important presenters of heparin-binding growth factors, such as HGF, the main ligand of Met [60, 62]. One must also consider that multiple variable regions might be present in a single isoform, adding further functional abilities. It is likely that the molecular interactions of CD44v3 are different from those of CD44v2-v10 variant despite both containing the v3 region, yet they will be ‘measured’ as a single product when using CD44v3 primers/probes/antibodies. Again, probes against any particular exons /protein domains od CD44 will demonstrate a wide range of CD44 isoforms, all of which share the same particular exon/domain. Furthermore, most of our tests are performed on mixtures of different tumour tissue elements: these assays work on a conglomerate of tumour cells, stromal cells, their proteins or gene products. Results of these reactions will lways represent some sort of average or summation. This might influence the qualitative expression pisture of distinct gene products of a tumour sample, but certainl will affect quantitative one. Considering this makes it easier to understand the contradictions of the literature.

Our aim was to identify a roster of CD44 isoforms expressed by colorectal cancer to see the extent of complexity we need to face when examining ‘CD44’. We also wanted to determine whether a potential “malignant” colorectal adenocarcinoma-specific isoform parttern exists. We then aimed to examine the stability of this pattern during tumour progression, i.e. whether it is qualitatively stable or new isoforms become dominant. This was tested in our complex iso- and xenograft experimental animal models. Reproducibility of v3 and v6 changes described in the literature was also tested in our experimental animal models. These models enabled us to separately study steps of tumour progression as well as the role of the tumour microenvironment.

First of all we attempted to identify the expressed isoforms from the qualitative CD44 ASP. We concluded that although not all the possible isoforms are expressed in reality, it is still rather technically impossible to identify the ones that do get expressed.All three attempts of ours on completing the isoform roster were only partially successful: (i) Our A – B primer overlapped the whole CD44 variable region, still just a few “dominant” isoforms could be identified as qualitative PCR products with direct sequencing. Some other isoforms could me identified according to the PCR product length, as well. Further CD44 variants were added by four other PCRs carried out with primer pairs designed onto particular exons of interest (v3 and v6). (ii) Next-generation sequencing theoretically should provide an allele-specific list of all transcribed isoforms. Still we reached limitations of the technique, but managed to identify more, sofar unidentified isoforms. (iii) I our previous publication on similar experimental set-up on melanoma (data under publication, Rásó-Barnett et al., paper accepted) identification attempt of “each” expressed CD44 isoform by cloning was reported to be fruitful, but again far from complete. We feel that the result of CD44 splicing is far richer than the sensitivity and effectivity of today’s tests. However, we do not yt know if this complexity of CD44 ASP carries any special functional importance. For now, we need to focus on functional characterization of distinct exons/domains of the complex molecule selectively.

Although the complete CD44 isoform list was not achieved, we identified simplified colorectal adenocarcinoma specific CD44 alternative splice pattern (ASP) which was qualitatively stable across samples from different colorectal cancer cell lines in vitro, and subcutaneous, intrasplenic and orthotopic implanted primary colorectal tumours and their liver metastases/colonies.

Hence, we focused on quantitative expression of two particular exons: v3 and v6. These two exons also appeared to be dominant on our reported qualitative ASP pictures. We separately examined their role in different steps of metastatic cascade in our complex animal experimental model system. Furthermore we managed to gain a glimpse of the role of tumour microenvironment (host) in the modulation of CD44 splice variant expression in conjunction with metastatic potential, as well.

Our isograft model (Balb/C – C26) proved that quantitative changes only happen during real metastasis formation, considering that v3 and v6 expression levels of both orthotopic and heterotopic implanted primary tumours were similar, but increased expression levels were only detected in real liver metastases and not in the colonies. This means that the higher expression level of v3 and v6 is not the effect of the liver microenvironment, but it is the actual pre-existing feature of distinct metastatic cells within the primary tumour. While at orthotopic implantation into large bowel wall tumour cells must undergo each classical steps of metastasis formation (EMT, intravasation, survival in circulation, extravasation, MET, proliferation in metastatic target organ), intrasplenic implantation gives the chance to a proportion of tumour cells to get straingt into the circulation by skipping the early phases of the cascade. Comparison of the two models can potentially differenciate between genes, gene products that act in different in different phases of metastatic process.

The same theory was tested in a human xenograft model as well. HT29 human colorectal adenocarcinoma xenograft showed similar results to the C26 isograft, while both HT25 and HCT116 behaved differently, which proposed further explanation to the contradictions of CD44 literature. We hypothesise that a higher v3/v6 expression level is a necessary, but probably not satisfactory prerequisite of cellular metastatic capability. This is proven by the fact that real liver metastases of all three CRC cell lines expressed v3 and v6 in the same range, despite the different expression levels of their colonic primaries Figure 4A,B. This also means, that the number of cells expressing high levels of CD44v3 and v6 is not necessarily high within the primary tumour, as theoretically a single ‘highly expressing’ clone is enough for metastasis formation. In a similar experimental model system circulating tumour cells of malignant melanoma were found to express CD44v3 and v6 containing varints on a high level in real metastatic system (Rásó-Barnett et al. under publication) Correlating this with our result would mean that HT29 had the lowest, while HT25 and HCT116 had the higher percentage of ‘highly expressing’, metastatically potent cells. This also carries the possibility of a primary tumour showing relatively higher expression levels than its metastasis, which would have already been re-selected by the new microenvironment. Translating this into clinical practice would mean, that the ‘lack’ of overexpression would not exclude metastatic potential or even a worse prognosis. As primary tumour site seemed to be interesting in terms of selection of metastatically potent, CD44v3/v6 high subclones, we tried to find support os this theory in another, quite different model system.

It is likely to assume that tumour microenvironment (tumour host organism) have a central role in metastatic phenotype presentation of primary colorectal cancer.

While normal tissues (such as colonic mucosa) do not express variant isoforms of CD44, tumours seem to perform a wide variation of CD44 isoforms. Presuming that different domains of different CD44 isoforms can perform a variety of new functions, there is no more sense in investigating CD44 “in general”.

Moreover, expression pattern of CD44 isoforms was found to be stable, meaning that colorectal type expression profile remained unchanged from tumour cell suspensions (in vitro) throughout primary and metastatic colon cancer xenografts. In contrary, quantitative changes do exist in the expression of distinct variant exons.

CD44 variant isoforms, especially the functionally well-characterized v3 and v6 containing isoforms, seem to massively take part in expressing the metastatic phenotype. Our results support that higher-level v3/v6 co-expression can represent this “quasy-metastatic-gene” function at the primary tumour site and at the early phase of the metastatic cascade. Nontheless, it must be emphasized that only a minor proportion of the primary tumour mass is sufficient to hold “metastatic-phenotype” for the complete primary tumour. Although each colorectal tumour types use metastatic subclones of high CD44v3/v6 expression rate, selective examination of metastatic tumour cell group is not yet resolved. This means that quantitative evaluation of not only CD44 “in general”, but even of v3/v6-containing isoforms are inappropriate for the prognosis of metastatic behavior of a single tumour case because of the summative way of the measurement techniques. Moreover this can be the background of the quite diversive results on the predictive value of “CD44” in the literature.

This study was supported by grants from the TÁMOP-4.2.1./B-09/1/KMR-2010-0001 and ETT 087-07/2009.