Are there two sides to colorectal cancer?

Authors

  • Barry Iacopetta

    Corresponding author
    1. Department of Surgery, University of Western Australia, Nedlands, Australia
    • Department of Surgery, University of Western Australia, 35 Stirling Highway, Nedlands 6009, Australia
    Search for more papers by this author
    • Fax: +61-8-9346-2416


Abstract

Colorectal carcinomas (CRC) that arise proximal (right) or distal (left) to the splenic flexure exhibit differences in incidence according to geographic region, age and gender. Together with observations that tumours in the hereditary cancer syndromes HNPCC and FAP occur predominantly in the right and left colon, respectively, the existence of 2 categories of CRC based on site of origin in the large bowel was proposed more than a decade ago. Differences between normal right and left colonic segments that could favour progression through different tumourigenic pathways are summarized in this review. Accumulating evidence suggests that the risk of CRC conferred by various environmental and genetic factors is different for proximal and distal tumours. Right- and left-sided tumours also exhibit different sensitivities to fluorouracil-based chemotherapy. Such differences are probably related to the molecular characteristics of the tumours, with the microsatellite instability and CpG island methylator phenotypes being associated with right-sided tumours and chromosomal instability with left-sided tumours. Future molecular-based classification systems for CRC that rely upon distinctive gene expression patterns may allow a clearer discrimination of subgroups than that provided by tumour site alone. Until then however, the existence of 2 broadly different groups of cancer defined by site of origin in the colon should be considered in the design of future epidemiologic studies as well as in the design of new clinical trials aimed at testing novel adjuvant therapies. © 2002 Wiley-Liss, Inc.

More than a decade ago Fearon and Vogelstein1 published a landmark review on genetic alterations and the development of colorectal cancer (CRC). During the same year, definitive evidence was also presented for a survival benefit from the use of fluorouracil-based adjuvant chemotherapy for CRC patients.2 These widely cited works have had enormous impacts on our understanding of the molecular basis of CRC and on its clinical management, respectively. With the benefit of hindsight it could be argued that a third article published in the same year was also deserving of a similar level of attention. Based on a review of the literature that included inherited forms of CRC, Bufill3 proposed the existence of 2 distinct categories of CRC according to the location of the tumour in the proximal or distal segments of the large bowel. Although Bufill cited several earlier authors who had put forward a similar idea,4, 5, 6 his article was the first to comprehensively review the evidence in favour of this concept. Since then several other groups, particularly in Europe, have continued to develop the theme of differences between right and left colon cancers.7, 8, 9, 10, 11, 12, 13 Despite this, many researchers fail to take into account the possibility that multiple categories of colon cancers with distinctive biologic and clinical properties could have implications for the evaluation of their data. In particular, clinical trials of adjuvant chemotherapies are conducted with apparent disregard for the likely existence of more than 1 CRC type.

While it may be convenient to categorize CRC into either proximal or distal location relative to the splenic flexure (Fig. 1), it is important to note that any definition based solely upon anatomical site of origin is a simplification. Underlying molecular features are responsible for determining tumour phenotype, and these are likely to show considerable overlap between right- and left-sided cancers. In the absence of clearly defined molecular markers that identify phenotypically distinct CRC subgroups, the site of tumour origin is a useful although somewhat crude way to distinguish these groups. One of the assumptions in this review is that proximal and distal tumours follow broadly different molecular pathways of carcinogenesis, thus explaining the epidemiologic, biologic and clinical differences described below. It should be borne in mind, however, that these routes are unlikely to be exclusive to 1 particular segment of the colon. The second assumption is that rectal carcinomas arise through similar mechanisms to left-sided colon tumours (descending and sigmoid colon) and have therefore been included with this category (Fig. 1).

Figure 1.

Right-sided tumours are classified as originating proximal to the splenic flexure (caecum, ascending colon, transverse colon), whereas left-sided tumours arise distally to this site (descending colon, sigmoid colon and rectum).

EMBRYOLOGIC, MORPHOLOGIC, PHYSIOLOGIC AND BIOCHEMICAL DIFFERENCES BETWEEN NORMAL RIGHTAND LEFT COLON

Various differences exist between the normal right and left colon that could predispose the tumours originating at these sites to develop along different pathways. For example, the proximal colon originates from the embryonic midgut and is perfused by the superior mesenteric artery, whereas the distal colon derives from the hindgut and is served by the inferior mesenteric artery. The capillary network surrounding the proximal colon is multilayered, whereas that of the distal colon is single-layered, possibly relating to the greater water absorption and electrolyte transport capacity of the former.14, 15 Blood group antigens A, B, H and Leb are expressed throughout the fetal colon but in the adult are found only in the proximal colonic mucosa.16, 17 Differences between adult proximal and distal colonic mucosa have also been reported for the expression of glycoconjugates as detected by lectin binding.18 Rectal mucosa expresses acidic mucin, whereas proximal colonic mucosa expresses neutral mucin.19 Average crypt length in the distal colon is greater than in the proximal colon.20 Some studies have suggested the existence of a proximal to distal proliferative gradient in the normal colon,21 although others have reported the opposite result22 or no segmental variation.23 The apoptotic index is lower in the right colon, possibly due to lower expression of the pro-apoptotic Bcl-2 homologue Bak compared to the left colonic mucosa.24

Fermentation reactions leading to short-chain fatty acid production are up to 8-fold higher in the proximal compared to distal colon.25 Several hydrolytic and reductive bacterial enzymes involved in the production of mutagenic metabolites also show regional differences in concentration within the large bowel.26 Levels of the pro-mutagenic lesion O6-methyldeoxyguanosine, a marker of exposure to N-nitroso compounds, were found to be higher in the normal distal compared to proximal colonic DNA of CRC patients.27 In spite of this, there is evidence the stem cell mutational load is similar for both sides of the colon.28 There are marked regional differences in the metabolism of bile acids between the right and left colon,29 the expression of various isoforms of cytochrome P-450,30, 31 the activity of ornithine decarboxylase32 and the expression of Na+/H+ exchanger isoforms.33 Methylation of the estrogen receptor (ER) gene was shown to increase proximally to distally in the normal colonic mucosa of 2 of 3 CRC patients examined.34 The associated transcriptional silencing of ER was suggested to be an early predisposing event for CRC. Recent technical advances in laser capture microdissection, cDNA microarrays and proteomic analysis35, 36 provide powerful new tools with which to explore genome-wide differences between normal proximal and distal colonic mucosa. In addition to revealing differences between right and left segments, this approach may also reveal differences in gene expression patterns related to age, gender and ethnicity. Such findings could prove relevant not only to the risk of developing CRC but also to the type of cancer that develops.

EPIDEMIOLOGIC DIFFERENCES IN THE INCIDENCE OF RIGHT AND LEFT COLON CANCERS

The worldwide incidence of CRC in the year 2000 was estimated at 945,000 cases, of which more than half will eventually prove fatal.37 Approximately 60% of CRC in high-incidence populations arise in the left colon, whereas in low-incidence regions there is a predominance of right-sided cases.38 The influence of gender and age on the subsite distribution of CRC was reported as early as 30 years ago39 and has been confirmed by many others since. In one of the best recent studies based upon examination of almost 10,000 CRC cases in the state of Florida,40 a progressive decrease in age was observed for caecal (median age 73.7 years) through to rectal (69.4 years) carcinoma patients. A parallel decrease in the proportion of female patients was also reported (55% to 44%, respectively). Although some studies have suggested proximal cancers present at a more advanced stage,41 others40 have not confirmed this. The incidence of proximal tumours in Western countries has steadily increased while that of distal tumours has shown a corresponding decrease.4, 42–47 When populations migrate from low-risk to high-risk areas, the CRC incidence quickly approaches that of the host country, with the largest increase occurring in the distal colon.48 Taken together, the above geographic, age and gender influences on CRC subsite distribution, in concert with the observations on proximal shift and migrant studies, suggests the existence of 2 broadly different mechanisms of CRC. One favours the development of proximal tumours and occurs predominantly in females, in older individuals and in areas with low incidence. The other favours the development of distal tumours and occurs predominantly in younger males from high-incidence areas.

ARE DIFFERENT ENVIRONMENTAL FACTORS RESPONSIBLE FOR RIGHT AND LEFT COLON CANCERS?

Environmental factors that could potentially favour the development of right- or left-sided colon tumours include diet, physical activity, smoking, cholecystectomy, chemopreventive agents, reproductive and hormonal status. Unfortunately, the majority of epidemiologic studies on environmental risk factors do not report separate data for right and left colon cancer subgroups, presumably because information on tumour location was not always available. This review summarizes findings only from the relatively small number of papers where subsite data are given.

Risks associated with diet tend to be strongest in the distal colon for men and in the proximal colon for women.49 High dietary fat intake has been linked to an increased risk of proximal cancers49, 50 and high protein intake with an increased incidence of distal cancers.50 In a study of Chinese residing in North America, high carbohydrate intake was associated with increased right colon cancer in women but increased rectal cancer amongst men.51 High calcium intake has been linked to a reduced risk of distal cancer,52, 53 whereas long-term use of folate-containing multivitamins is associated with reduced risk of proximal cancer.54 High luminal exposure to iron appears to increase proximal cancer risk, whereas high iron levels in the serum increase the distal cancer risk.55

A Norwegian study reported that a high level of physical activity in men reduced the incidence of proximal but not distal colon cancer.56 The opposite was found in a North American cohort of male smokers,57 possibly relating to observations that heavy smoking is a greater risk factor for rectal compared to colon cancer.58, 59 A meta-analysis of 33 studies found that cholecystectomy was associated with increased risk of proximal but not distal colon cancer,60 a finding subsequently confirmed by a large Swedish study.61 At least 10 epidemiologic studies have reported a reduced incidence of CRC with prolonged use of nonsteroidal anti-inflammatory drugs (NSAIDs), but only 1 appears to have examined the influence according to subsite.62 This study found evidence for a reduction in both colon and rectal tumours with regular use of NSAIDs. In 17 familial adenomatous polyposis (FAP) patients treated with the NSAID sulindac, the incidence of adenomas decreased 7-fold in the proximal colon but only 2-fold in the distal colon.63 Depending on the chemopreventive agent used, animal models have also shown segmental differences in the reduction of experimentally induced colon cancers64, 65, 66 The reduction in distal tumours observed with the NSAID nabumetone66 may be linked to the proximal to distal gradient in expression of the target gene COX-2 in normal colonic mucosa.67

Women with at least 4 births show a reduced risk of colon but not rectal cancer compared to nulliparous women.68, 69 The risks associated with age at menarche and age at first pregnancy are reported to be significantly different between proximal and distal colon cancers.70 There have been numerous studies on hormone replacement therapy (HRT) and the risk of CRC in women, with several showing a reduced incidence of colon but not rectal cancer.71, 72, 73 Although 1 meta-analysis confirmed these individual findings,74 another found the reductions in risk were similar for both colon and rectal cancer.75

ARE DIFFERENT GENETIC FACTORS RESPONSIBLE FOR RIGHT AND LEFT COLON CANCERS?

The 2 known familial forms of CRC, FAP and hereditary nonpolyposis coli (HNPCC) exhibit markedly different clinical features that are suggestive of different mechanisms of disease.3 FAP is characterized by the presence of hundreds of polyps in the large bowel. These arise first in the rectum and distal colon before extending to more proximal segments. Close to 100% of FAP individuals will develop CRC in the left colon. In contrast, approximately 70% of large bowel tumours in HNPCC individuals arise in the right colon.76 In addition to the highly penetrant but rare APC or DNA mismatch repair gene mutations responsible for FAP and HNPCC, respectively, common low-penetrance genetic variants that predispose to right or left colon cancers also exist. At least 50 studies have examined associations between various polymorphisms and CRC risk,77 but relatively few have analysed according to subsite. Similar to the environmental risk factor studies described above, this review will only summarize findings where subsite data are given.

A polymorphism in apolipoprotein E is associated with reduced risk of proximal but not distal tumours.78 The TT genotype of the methylene tetrahydrofolate reductase (MTHFR) gene has also been associated with a reduced risk of proximal cancers, particularly in older individuals with high levels of folate, vitamin B6 and vitamin B12 intake.79 Possibly related to these observations, TT genotype patients with proximal tumours have a later age of diagnosis compared to those with the CC genotype.80 No such association was observed for patients with distal tumours. A polymorphism in the cystathionine beta-synthase (CBS) gene was also associated with reduced risk of proximal tumours but not distal tumours.80 Both the MTHFR and CBS genes are involved in folate and methyl group metabolism and it is tempting to speculate that functional polymorphisms in these genes might alter the risk of developing tumours with an aberrant DNA methylation phenotype (see below). Such tumours are most commonly found in the proximal colon.81, 82 A possible association between the E-cadherin CDH1 genotype and proximal cancers has been reported for familial CRC cases.83 Polymorphisms in genes that metabolize carcinogens may also be risk factors for CRC. Amongst this group, variants of the NAD(P)H:quinone oxidoreductase84 and CYP2C985 genes have been associated with increased risk of proximal cancer, while a combination of GSTM1 and GSTM3 alleles has been associated with reduced risk of distal cancers, particularly in males.86 Polymorphisms in NAT1, GSTM1 and GSTT1 have also been shown to modify the risks of proximal or distal tumour location in HNPCC individuals.87

MORPHOLOGIC AND MOLECULAR DIFFERENCES BETWEEN RIGHT AND LEFT COLON CANCERS

The density of aberrant crypt foci, generally thought to represent preneoplastic lesions, decreases from the proximal to the distal colon of patients with CRC.88 Mucinous carcinomas decrease in frequency by approximately 2.5-fold from the proximal to distal colon.89, 90 This is in line with the more frequent expression of mucin-associated M1 antigen in proximal (55%) compared to distal (13%) tumours.89

Possibly the first report of a site-related molecular difference was the description in 1982 of a progressive increase in the incidence of aneuploidy from proximal to distal tumours.91 This was followed by evidence for differential expression of the c-myc oncogene and the suggestion that it defined 2 genetically distinct forms of CRC.6 Further evidence was provided by Delattre et al.,7 who found that almost all distal but only 40% of proximal tumours showed allelic losses at chromosomal regions in 17p, 18 and 5q. Approximately 80% of proximal tumours were found to be diploid compared to only 40% of distal tumours. Left-sided tumours have been shown to express higher levels of growth factors including IGFII92 and EGF/TGF-alpha.93 The cyclooxygenase COX-2 is overexpressed in 90% of rectal tumours but in only 20% of colonic tumours.94

Interestingly the incidence of somatic mutations in the Ki-ras oncogene does not show regional differences,7, 95 suggesting that it may be relevant to the progression of CRC from both segments of the colon. Mutations in the TP53 gene are approximately 1.5–3-fold more frequent in distal tumours.95, 96 Large regional differences in the incidence of the microsatellite instability (MSI+) phenotype have been reported, with up to 10-fold higher frequencies seen in sporadic proximal compared to distal tumours.95, 97, 98 The majority of sporadic MSI+ tumours arise following hypermethylation of the hMLH1 gene and are closely related to a tumour group referred to as CIMP+ and characterized by concurrent methylation of multiple CpG islands.99 It is not surprising therefore to find that DNA hypermethylation is also more frequent in proximal compared to distal tumours by a factor ranging from 4–13-fold depending upon the number and type of CpG islands examined.81, 82, 100

CONCLUDING REMARKS

Since Bufill's review more than a decade ago,3 studies from diverse fields have provided further evidence for the existence of at least 2 mechanisms of CRC. These include additional data showing that the risks associated with various environmental and genetic factors differ for proximal and distal cancers. Perhaps the strongest evidence, however, comes from the molecular characterization, or profiling, of colorectal tumours. The existence of at least two major groups of CRC has been proposed based on the molecular features of CIMP+, MSI+ and chromosomal instability (CIN). One group (CIMP+/MSI+) occurs predominantly in the right colon and the other (CIN) in the left colon.12, 101, 102 A summary of the major clinical and molecular properties of proximal and distal tumour subgroups is shown in Table I.

Table I. Clinical and Molecular Features of Proximal and Distal Colon Tumours
FeatureProximalDistal
Age at diagnosisOlderYounger
GenderMore femalesMore males
Mucinous tumoursFrequentInfrequent
Familial cancer syndromeHNPCCFAP
5FU chemotherapy benefitGoodMarginal or none
PloidyMostly diploidMostly aneuploid
Loss of heterozygosityInfrequentFrequent
TP53 mutation20–30%50–60%
MSI+25%2–3%
CIMP+25–40%3–10%

As suggested earlier for the study of differences between normal proximal and distal colonic mucosa, recent technical advances open the way for a more detailed molecular profiling of colorectal cancers.35, 36 These advances include laser capture microdissection, cDNA microarray technology, proteomics and tissue arrays.103 Comparison of gene or protein expression patterns between proximal and distal, CIMP+ and CIMP-, or MSI+ and MSI- tumours should reveal fascinating insights into the different mechanisms of CRC. Furthermore, this work should lead to a novel, molecular-based classification of CRC similar to that described recently for B-cell lymphomas104 and breast tumours.105 One of the most important clinical applications of this work is for the improved selection of patients to receive chemotherapy. Molecular profiling will eventually allow chemoresponsive patients to be identified with much greater accuracy than is currently achieved using morphologic-based parameters.106 With the aim of identifying molecular markers of chemosensitivity, work from a retrospective cohort study found the MSI+ phenotype to be associated with excellent survival benefit from standard 5-fluorouracil (5FU)-based chemotherapy.107 Knowing that almost all MSI+ tumours are found in the proximal colon,95, 97, 98 this observation led to the discovery that patients with right-sided tumours gained more benefit from chemotherapy than those with left-sided tumours.107 Results from a prospective, randomized trial also confirm that patients with colon cancer, but not rectal cancer, derive significant benefit from 5FU chemotherapy.108 Clinical trials of adjuvant therapies for CRC have so far failed to take into account the possibility that proximal and distal tumours may differ in their response to treatment. Ironically, the final report from the landmark clinical trials of 5FU chemotherapy in CRC109 was published 5 years later in the same journal in which Bufill3 had so eloquently drawn attention to the many differences between proximal and distal tumours.

It is likely that reanalysis of existing epidemiologic data on environmental and genetic risk factors and of clinical trial data on adjuvant chemotherapy will reveal further differences between right and left colon cancers. Future epidemiologic studies and clinical trials of CRC would be advised to take into account the possibility of tumour site differences in their study design. Although some progress has been made towards a more accurate molecular-based system of CRC classification, the anatomic site of origin of this tumour continues to provide a convenient discriminator for 2 subgroups having important biologic and clinical differences.

Acknowledgements

The author is grateful to F Grieu for critical reading of this manuscript.

Ancillary