Introduction

In Europe in 2006, colorectal cancer was the second most common cancer, with an estimated incidence in both sexes of 412,900 cases, and the second most common cause of cancer mortality, with 207,400 deaths [1]. Colon cancer represents about two thirds of colorectal cancer cases, and TNM stage is the main prognostic factor. Historically, 20–25% of patients have had stage III disease and 15–20% have had stage II disease at diagnosis [2]. In many countries, mass screening programs using fecal occult blood tests likely will increase the proportion of patients with stage II disease in the coming years [3]. After surgical resection, the risk of recurrence is about 50–60% in patients with stage III disease and about 20–30% in those with stage II cancer [2]. To decrease this risk and increase overall survival, these patients are candidates for adjuvant chemotherapy.

After surgical resection of stage III colon cancer, adjuvant chemotherapy with 5-fluorouracil (5-FU) was associated with an estimated one-third relative reduction in the risk of colon cancer recurrence [4], and 5-FU-based chemotherapy was the standard adjuvant therapy for more than 10 years. However, since 2004 and the MOSAIC (Multicenter International Study of Oxaliplatin/FU-LV in the Adjuvant Treatment of Colon Cancer) study, the combination of oxaliplatin and fluoropyrimidine has been the new standard for stage III colon cancer patients, with a relative 20% increase in survival at 6 years [5•]. In patients with stage II colon cancer, overall survival rates at 5 years are 70–80%, and the absolute improvement with 5-FU chemotherapy is small, with a 3.6% reduction in the relative risk of death at 5 years [6, 7••]. Therefore, these results do not support the use of adjuvant chemotherapy in all patients with stage II colon cancer.

Colon cancers do not represent a homogeneous entity. Most develop sporadically (90%), but some are part of a hereditary cancer syndrome (<10%) or occur in the background of inflammatory bowel disease [8]. The adenoma–carcinoma sequence underlies the development of colon cancer in most cases. However, at least two major distinct pathways of carcinogenesis have been identified and have led to the conclusion that colon cancer is a genetically heterogeneous disease. These two pathways are the microsatellite instability (MSI) and the chromosomal instability (CIN) or microsatellite stability (MSS) pathways [9••]. More and more data suggest that the microsatellite status (MSI/MSS) is a prognostic and predictive marker of fluoropyrimidine efficacy, and these results potentially have clinical implications in adjuvant therapeutic strategy. The objective of this review is to highlight the importance of microsatellite status in decisions regarding adjuvant chemotherapy in patients with stage II colon cancer.

MSS and MSI Pathways

Studies of early models of colorectal carcinogenesis derived from families with familial adenomatous polyposis (FAP) or hereditary nonpolyposis colorectal cancer (HNPCC, Lynch syndrome) allowed us to describe and understand the two main known pathways [9••, 10]. The MSS and MSI status of colon cancer depends on FAP and HNPCC inherited disorders, respectively. The CIN or MSS pathway is the most common genetic pathway, occurring in approximately 85% of colorectal cancers, and is characterized by allelic losses, amplifications, and translocations. Genomic alterations associated with this pathway are well described and consist of aneuploidy and recurrent loss of heterozygosity at 17p, 18q, 5q, 1p, 8p, and 22q. Furthermore, MSS tumors are more likely than MSI tumors to bear APC, TP53, or KRAS mutations [9••, 10]. Despite the high frequency of MSS in colon cancer, the mechanisms for CIN are incompletely understood.

Our understanding of the second pathway is more consequent. The most common cause of MSI is a loss of the DNA mismatch repair (MMR) function. The MMR system contains several proteins that interact with one another and eliminate errors that may arise during DNA replication. In particular, it repairs single-base mismatches and insertion–deletion loops that may result from gains or losses of short repeat units within microsatellite sequences [9••, 10]. These microsatellites are small polymorphic repeated DNA sequences of one to five nucleotides localized throughout the genome. In cases of MMR dysfunction, small insertions or deletions alter microsatellite length and are visualized as band shifts on gel electrophoresis. MSI tumors are characterized by a high frequency of MSI (MSI-H) [11]. In patients with Lynch syndrome, a germline mutation in one of the MMR genes associated with an acquired second somatic event explains the MMR dysfunction in most cases; the four main genes implicated in this inherited disease are hMLH1, hMSH2, hMSH6, and hPMS2. Among Lynch syndrome patients, the rate of germline mutation of these genes is about 40% for hMLH1 and hMSH2, 5% to 10% for hMSH6, and less than 2% for hPMS2 [9••]. Nevertheless, Lynch colorectal cancers represent less than 3% of all colorectal cancers [12]; thus, most MSI tumors occur sporadically. In most sporadic cases, MMR dysfunction is the consequence of epigenetic mechanisms leading to hMLH1 silencing through its promoter hypermethylation [13]. As for MSS tumors, there are specific mutations associated with MSI tumors. Because of the existence of repeated coding sequences, some genes are mutated more frequently in MSI than MSS tumors [9••]. Furthermore, for unknown reasons, the rate of BRAF mutations is significantly higher in sporadic MSI tumors with a methylated hMLH1 promoter than in other MSI (Lynch syndrome) or MSS tumors [9••, 14••]. Sporadic MSI tumors resulting from promoter hypermethylation may be a subgroup of tumors with a high CpG island methylator phenotype (CIMP), but CIMP phenotype seems to be a poor prognostic marker in MSS tumors only [15].

Between these two main pathways, a class of colorectal tumors with low MSI (MSI-L) has been described and might exist as a separate group. In particular, some recent publications suggest that MSI-L tumors may be associated with poor prognosis in stage III colon cancer [16]. However, the biologic defect causing the MSI-L phenotype is not well understood, and most tumors can be classified as MSI-L if enough markers are tested [17]. No clear clinicopathologic or biologic difference has been found yet between MSI-L and MSS tumors, and these two tumor types are analyzed together in most clinical studies [12]. Therefore, there is much confusion surrounding MSI-L tumors, and research is needed to clarify the definition of this group and then its possible relevant clinical and biologic characteristics. In this review, MSI-L tumors are considered MSS tumors.

Diagnosis of MSI Tumors

Determining microsatellite status depends principally on identifying MSI tumors; the other tumors are MSS by exclusion. Some clinical, pathologic, and biologic characteristics are associated with MSI tumors, and some recommendations have been proposed to provide a basis for diagnosis in international studies.

In 1990, the International Collaborative Group on Hereditary Non-Polyposis Colorectal Cancer proposed the establishment of a set of selection criteria for families with HNPCC. The aim of these criteria was to allow simple and rapid selection of suspected HNPCC families. Selected families were offered genetic counseling, DNA testing, and surveillance in cases carrying the mutation. These criteria were revised 10 years later and are now called Amsterdam criteria II (Table 1) [18]. Apart from these familial history hallmarks, distinct clinical and pathologic features of colorectal tumors arising from the MSI pathway have been identified. Clinically, MSI tumors are associated with female sex, right/proximal colon cancer, and a lower stage at diagnosis compared with MSS tumors [14••, 19–23]. More precisely, whereas in HNPCC families MSI tumors occur frequently before the age of 50 years and in either sex, sporadic tumors are associated with women and older age [21]. Some pathologic features also are associated with the MSI pathway, and compared with MSS tumors, MSI tumors more frequently exhibit poor differentiation, a mucinous or signet-ring cell type, and increased tumoral and peritumoral lymphocytic infiltration, also called “Crohn’s-like inflammation” [10, 20–23]. The most sensitive pathologic feature of MSI status is the presence of tumoral lymphocytic infiltration that can be assessed and quantified [12].

Table 1 International guidelines for hereditary nonpolyposis colorectal cancer and for microsatellite instability
Full size table

These clinical and pathologic features guide the diagnosis of MSI tumors but do not affirm it. Biologic diagnosis of MSI is straightforward with polymerase chain reaction (PCR) assay; it consists of testing the DNA from the tumor as well as from normal tissue and looking for mutations altering microsatellite length in the tumor sample, which are visualized as band shifts on electrophoresis. Instability can be detected in both fresh and paraffin-embedded samples, and tissue processing should include microdissection to enrich tumor cell population [11]. In December 1997, the National Cancer Institute endorsed recommendations for the diagnosis of MSI in colorectal cancer. A panel of five microsatellites was validated and recommended as a reference panel [11]. Tumor classification was established according to the number of markers showing instability: MSI-H or MSI tumors were defined as having an instability in at least two of the five recommended markers, or in more than 30–40% of the markers if more were tested (Table 1). However, the availability of matching normal DNA is not an absolute requirement when the germline allele size is considered constant in the population. The use of quasi-monomorphic mononucleotide repeats is sensitive and specific enough to detect MSI colon tumors and may obviate the need for normal tissue for comparison [24]. These considerations and others led to a revision in the Bethesda guidelines in 2002 (Table 1) [12].

Which Testing Method to Use

Today, testing of microsatellite length is considered the “gold standard” for determining MSI phenotype. Nevertheless, because immunohistochemistry (IHC) allows identification of the specific gene implicated in the MMR deficiency, this method has been the focus of research to assess it. For HNPCC screening, IHC using four antibodies against hMLH1, hMSH2, hMSH6, and hPMS2 is almost as sensitive and specific as PCR [25]. However, germline mutations of hMSH6 and hPMS2 are involved in fewer than 10% of HNPCC patients, who represent approximately 2% of all patients with MSI tumors [9••]. Therefore, most studies used only antibodies against hMLH1 and hMHS2 for MSI determination in all colorectal cancers. With these two antibodies, the specificity of IHC is excellent—higher than 95%—but the sensitivity is moderate, varying from 62% to 95% among studies [19, 20, 23, 26]. The reasons for such variation include variability in fixation, storage, and staining quality within laboratories; heterogeneity within tumors; and difficulty in IHC interpretation [26].

Consequently, for Lynch syndrome diagnosis, IHC cannot yet replace PCR, although it constitutes an alternative and/or complementary method. Ideally, microsatellite status determination should be based first on PCR testing and second on IHC testing in MSI tumors to allow identification of the specific gene implicated in the MMR deficiency. Nevertheless, in routine practice, such an algorithm is not easy to conduct everywhere. The final decision to conduct one and/or the other test may be patient- and/or center-specific. It depends on local expertise, cost-efficiency, probability of an abnormal test, and whether the test is being used for screening or diagnosis [12]. Importantly, PCR and IHC testing cannot indicate the mechanism—genetic or epigenetic—of the MMR deficiency, and in HNPCC families, only gene sequencing allows identification of the specific germline mutation.

For MSI diagnosis, independent of Lynch syndrome screening, PCR and IHC tests gave approximately the same results in recent studies [23, 27–31], and either method may be used, according to local possibility and expertise.

MSI Tumors and Prognosis

Since the first studies analyzing the relationship between clinicopathologic variables and MSI tumors, researchers have suspected that patients with MSI tumors have a better prognosis after curative resection than patients with MSS tumors [22]. Interestingly, although MSI tumors are diagnosed at a significantly greater tumor invasion (T) stage and tend to have greater tumor volume than MSS tumors at the same T stage, MSI tumors have a significantly lower overall TNM stage (stage I–IV) than MSS tumors [10, 19, 20, 27]. However, MSI tumors are significantly more poorly differentiated, which is a classic poor prognostic factor [14••, 23, 32]. Therefore, many studies have assessed the prognosis of patients (stage I–IV) with MSI tumors, but the results of some were not statistically significant [20, 28, 33, 34]. Several reasons may explain the heterogeneity of published results: the definition of the MSI phenotype was not initially standardized, leading to inconsistent and confusing reports; many studies were retrospective, with inherent bias; and some studies were not powered enough [11, 20, 28].

Despite these limitations, however, there is strong evidence that patients with MSI tumors have a more favorable prognosis than those with MSS tumors [14••, 27–30]. Population-based and retrospective studies of microsatellite status in untreated patients (no adjuvant chemotherapy) who participated in prospective phase 3 adjuvant trials found that in most cases, MSI phenotype was an independent prognostic factor for relapse-free and overall survival (Table 2) [14••, 27–30]. In 2005, a systematic review was performed to obtain a more precise estimate of the prognostic significance of microsatellite status; 32 studies were included, with a total of 7642 patients, including 1277 with an MSI tumor. Compared with MSS, MSI phenotype was associated with a 35% reduction in the relative risk of death (hazard ratio [HR], 0.65; 95% CI, 0.59–0.71), with no evidence of heterogeneity (P = 0.16; I 2 = 20%) or publication bias (P > 0.1). The survival advantage of MSI tumors was confirmed when analysis was restricted to subgroups of clinical trial patients (HR, 0.69; 95% CI, 0.56–0.85), patients with metastatic cancer (HR, 0.64; 95% CI, 0.46–0.89), or patients with stage II and III disease (HR, 0.67; 95% CI, 0.58–0.78) [28]. In a large cohort of patients with colorectal cancer (718 patients), MSI phenotype was an independent good prognostic factor for cancer-specific survival on multivariate analysis (HR, 0.26; 95% CI, 0.13–0.52; P = 0.0001). Furthermore, this survival advantage was statistically significant in both stage II (P = 0.0006) and stage III (P = 0.0052) disease, and the 6-year survival rate for patients with MSI stage II tumors was 97% [23]. These results are concordant with those from a retrospective study in which patients with MSI T3N0 sporadic colon cancer had a 5-year survival rate of 90% (no recurrence) [31].

Table 2 Prognostic and/or predictive value of microsatellite status in adjuvant clinical trials
Full size table

The better overall survival of patients with MSI tumors was acknowledged by the vast majority of retrospective studies (Table 2) [14••, 27–30]. Its reasons are not well understood, although some hypotheses exist. The MSI pathway, with its pronounced genetic instability, is associated with an increased rate of mutation in many genes and may be responsible for at least two tumoral growth disadvantages. First, microsatellite sequences are sometimes present in the exons of critical regulatory genes. Mutations in such genes implicated in cell growth and survival are most frequent in MSI tumors and possibly may increase the rate of “spontaneous” apoptosis of tumoral cells [9••]. Second, the tumoral lymphocytic infiltration observed in MSI tumors most likely is the result of aberrantly expressed proteins resulting from frameshift mutations [35••], and probably plays a large role in the better prognosis of patients with MSI tumors.

MSI and the Predictive Benefit of Adjuvant Chemotherapy with 5-FU (Stage II and III)

5-FU, used alone or in combination with oxaliplatin, remains the pivotal drug in colon cancer adjuvant treatment [4, 5•]. For this reason, soon after the discovery of the MSI pathway, some studies assessed the predictive value of MSI phenotype for 5-FU chemosensitivity.

In vitro data have shown that MSI colorectal cancer cell lines are more resistant than MSS cell lines to 5-FU, despite similar amounts of 5-FU incorporated in DNA. Furthermore, in MSI cell lines with hMLH1 promoter hypermethylation, chemoresistance to 5-FU can be overcome by promoter demethylation and restoration of normal MMR function [36]. Later, it was demonstrated that 5-FU chemosensitivity is a result of the capacity of MMR components to recognize 5-FU—modified DNA and induce apoptosis [37]. Despite these in vitro data, MSI phenotype initially was suggested to be a positive predictor of the benefit of adjuvant chemotherapy with 5-FU [38]. However, subsequent studies did not confirm these preliminary results. Some retrospective studies failed to demonstrate any predictive value of MSI phenotype [32]; in many other studies, patients with MSI tumors did not benefit from 5-FU adjuvant chemotherapy, whereas those with MSS tumors did [27, 30]. The reasons for such heterogeneous results are the same as those discussed for prognostic assessment in the previous section. Furthermore, the prognostic significance of MSI phenotype has served to mask its predictive value [38], which could be assessed correctly only in studies comparing two groups of patients, one treated with 5-FU and one untreated [23, 27, 30].

Until now, no prospective randomized adjuvant chemotherapy study looking at microsatellite status had been published. Currently under way is a phase 3 study in stage II colon cancer (Eastern Cooperative Oncology Group [ECOG] 5202) that is classifying tumors according to MSI status and chromosome 18q allelic imbalance (AI). However, in this trial, patients in the low-risk group (MSI or no 18q AI) receive no adjuvant chemotherapy whereas those in the high-risk (MSS or 18q AI) group are treated with FOLFOX (leucovorin, 5-FU, and oxaliplatin) with or without bevacizumab. Despite opposite results and controversies, most studies reported that patients with MSI tumors did not benefit from 5-FU adjuvant chemotherapy (Table 2) [28, 30, 32]. In a recent meta-analysis without individual data, which included seven studies representing 3690 patients with stage II (25%) or III disease (75%), 5-FU adjuvant chemotherapy had no effect on relapse-free survival (HR, 0.96; 95% CI, 0.62–1.49; P = 0.86) or overall survival (HR, 0.70; 95% CI, 0.44–1.09; P = 0.12) in patients with MSI tumors [39]. In comparison, chemotherapy had a beneficial effect on relapse-free survival among patients with MSS tumors (HR, 0.77; 95% CI, 0.68–0.87; P < 0.001) [39]. MSI phenotype is predictive of the absence of 5-FU adjuvant chemotherapy benefit [23, 27, 30].

Whereas the value of MSI phenotype for predicting 5-FU chemoresistance is supported by many studies, only a few studies have assessed its predictive value in polychemotherapy regimens. In metastatic colorectal cancer, the FOLFOX regimen seems to give the same results in MSI tumors as in MSS tumors [40]. In the adjuvant setting (stage III), one retrospective study reported that the addition of oxaliplatin to 5-FU significantly improved relapse-free survival compared with 5-FU alone in patients with MSI tumors [41•]. Recently, it was suggested that a combination of irinotecan, 5-FU, and leucovorin may improve outcome compared with 5-FU/leucovorin in patients with MSI stage III colon cancer [42]. Nevertheless, another study (PETACC 3 [Pan-European Trials in Adjuvant Colon Cancer]) did not confirm this suggestion [14••]. Thus, these results are preliminary and must be interpreted with caution. Other studies are needed to determine the possible predictive value of MSI phenotype for 5-FU/oxaliplatin or 5-FU/irinotecan regimens.

Adjuvant Chemotherapy in Patients with Stage II Colon Cancer

Although statistically significant, the absolute improvement observed with 5-FU chemotherapy is small in patients with stage II colon cancer, with a 3.6% reduction in the relative risk of death at 5 years, and does not support the routine use of adjuvant chemotherapy for all patients with stage II colon cancer [43]. At present, the recommendations of oncology societies are to discuss the use of adjuvant chemotherapy with patients who have high-risk stage II disease [43, 44]. Many prognostic factors described in colon cancer may be used to define high-risk stage II tumors. Among the most validated “classic” factors are T4 stage (visceral peritoneum perforation, presence of adherence to or invasion of local organs); number of examined lymph nodes, with a cutoff between 8 and 12; tumor grade with a poorly differentiated histology; bowel obstruction or tumor perforation; and presence of peritumoral lymphovascular invasion [43, 44]. Among these factors, the most significant on multivariate analysis were T4 stage, number of examined lymph nodes, and tumor grade [45], [46•]. Importantly, the value of these factors is prognostic, not predictive; they are used only to “identify” high-risk stage II tumors and to select patients who may benefit the most from adjuvant chemotherapy.

In 2006, the American Society of Clinical Oncology’s recommendations for the use of tumor markers in gastrointestinal cancer did not include the use of microsatellite status to determine prognosis or to predict the effectiveness of 5-FU adjuvant chemotherapy [47]. Since then, further evidence has accumulated confirming the prognostic value of microsatellite status in stage II and III colon cancer patients treated or not treated with adjuvant 5-FU-based chemotherapy [14••, 30], [46•]. Presently, the results from recent large independent studies on the value of microsatellite status are statistically robust and concordant [14••, 30, 46•]; in France, they already have brought about modifications in recommendations for adjuvant chemotherapy for patients with stage II disease (http://www.tncd.org).

In accordance with the findings that patients with MSI tumors have a better prognosis than those with MSS tumors [14••, 30, 46•] and do not benefit from 5-FU-based adjuvant chemotherapy [27, 30], patients with MSI stage II colon cancer should not receive fluoropyrimidine adjuvant chemotherapy. Microsatellite status should be determined before making any decision regarding adjuvant chemotherapy with fluoropyrimidine alone in patients with stage II disease, and either PCR or IHC testing may be performed. In the subgroup of patients with MSI stage II colon cancer, the prognosis of those with T4 stage and/or fewer than 8 to 12 lymph nodes is unknown and should be assessed.

For patients with MSS stage II colon cancer, the use of adjuvant chemotherapy should be based on the presence of classic prognostic factors and on discussions with the patient about the benefit/risk of this treatment, and each treatment decision should be made on a case-by-case basis. Two prognostic models have been developed to help in decision making with regard to adjuvant therapy: the Mayo Clinic adjuvant systemic therapy tool (http://www.mayoclinic.com/calcs) and the Adjuvant! On-line calculator (http://www.adjuvantonline.com). These two models recently were validated in an independent population-based dataset of patients, with acceptable and similar reliability [48]. Moreover, besides these models that include only classic prognostic factors, other pathologic features, such as the extent of intratumoral immune reaction, or molecular features such as the 12-gene colon cancer recurrence score show promise for identifying high-risk stage II colon cancers [46•, 49••]. Although fluoropyrimidine-based chemotherapy demonstrated a benefit [7••, 50], the adjuvant chemotherapy regimen also may be discussed. A subgroup analysis of the MOSAIC study found no difference in benefit between FOLFOX4 and 5-FU/leucovorin regimens in patients with stage II disease [5•]. However, the probabilities of disease-free survival at 5 years in patients with high-risk stage II disease were 82.3% and 74.6% in the FOLFOX4 and 5-FU/leucovorin groups, respectively (HR, 0.72; 95% CI, 0.50–1.02). The FOLFOX regimen is more toxic, and direct evidence does not support its use in most patients with stage II disease. Therefore, fluoropyrimidine may be considered the standard adjuvant chemotherapy for patients with high-risk MSS stage II colon cancer, and FOLFOX may be an alternative for some patients with T4 stage and/or fewer than 10 to 12 examined lymph nodes.

Conclusions

Concordant and accumulated results from large independent studies allow us to acknowledge that microsatellite status is now a robust prognostic factor in colon cancer. In patients with stage II MSI colon cancer, better prognosis and 5-FU chemoresistance do not support the use of 5-FU adjuvant chemotherapy. Microsatellite status should be determined before making any decision regarding adjuvant chemotherapy in patients with high-risk stage II colon cancer. In stage III colon cancer, translational research is ongoing to better define the benefit of the FOLFOX4 regimen according to microsatellite status. Because individualization with regard to treatment and strategy is critical, future trials in colon cancer will have to consider microsatellite status an important molecular marker in the adjuvant setting.