Microsatellite instability in colorectal cancer: from molecular oncogenic mechanisms to clinical implications

Abstract

Background

Microsatellite instability (MSI) constitutes an important oncogenic molecular pathway in colorectal cancer (CRC), representing approximately 15% of all colorectal malignant tumours. In roughly one third of the cases, the underlying DNA mismatch repair (MMR) defect is inherited through the transmission of a mutation in one of the genes involved in MMR, predominantly MSH2 and MLH1, or less frequently, MSH6 or PMS2. In the overwhelming number of sporadic cases, MSI results from epigenetic MLH1 silencing through hypermethylation of its promoter. MMR deficiency promotes colorectal oncogenesis through the accumulation of numerous mutations in crucial target genes harbouring mononucleotide repeats, notably in those involved in the control of cell proliferation and differentiation, as well as DNA damage signalling and repair.

Design

In this review, we describe the molecular aspects of the MMR system and the biological consequences of its defect on the oncogenic process, and we discuss the various experimental systems used to evaluate the efficacy of cytotoxic drugs on MSI colorectal cells lines. There is increasing evidence showing that MSI CRCs differ from all CRCs in terms of prognosis and response to the treatment. We report the clinical studies that have evaluated the prognostic and predictive value of MSI status on clinical outcome in patients treated with various chemotherapy regimens used in the adjuvant setting or for advanced CRCs.

Conclusion

In view of this, the opportunity of a systematic MSI phenotyping in the clinical management of patients with CRC is further discussed.

Introduction

The importance of DNA mismatch repair (MMR) system in human oncogenesis clearly emerged in the early 90’s by the demonstration that its inactivation owing to germline mutations was the primary cause of hereditary non-polyposis colorectal cancer (HNPCC), also known as Lynch syndrome [17, 39, 89, 153, 189]. Colorectal cancers (CRCs) arising in the context of a Lynch syndrome, are mainly due to germline mutations in MSH2 or MLH1, more rarely MSH6 or PMS2, and account for 2–5% of all CRCs. These tumours are characterized by microsatellite instability (MSI) due to the accumulation of insertion/deletion mutations in tandem repeats of short DNA motifs (1–6 bp) called microsatellites. MSI also arises in approximately 12–15% sporadic cases of CRCs through somatic mutations, epigenetic inactivation of MMR, mainly by silencing of MLH1 through promoter hypermethylation. Identifying patients with Lynch syndrome, the most common inherited cause of CRC, is important not only for clinical management, notably because of the higher risk of developing a second primary cancer, but also for their family members some of whom have inherited the MMR gene defect. As MSI CRCs have a better prognosis than microsatellite stable (MSS) CRCs and behave differently in response to various therapeutic strategies including treatment with 5-fluorouracil (5FU), it may now be time to evaluate the relevance of systematic MSI testing rather than confining it to genetic screening [17, 39, 166, 171, 189]. In this review, we will summarize the molecular basis of MMR, the biological and clinical characteristics of MSI tumours, and the methods used to detect MSI. We will further describe the functions of MMR and the consequences of its defect on the response of cell lines to various drugs, notably those given to patients with CRC. Finally, we will report data regarding the impact of MSI on prognosis and clinical outcome of patients with CRC either treated or not by chemotherapy.

Post-replicative DNA mismatch repair

The DNA MMR system was originally identified in bacteria, in which its inactivation results in an increased rate of spontaneous mutations [134, 154]. The DNA MMR system has first been described for its role in post-replicative repair of the DNA polymerases errors that have escaped proofreading activity. Replicative polymerases may incorporate the wrong nucleotide creating Watson-Crick base-base mismatches or slip on short repeated sequences such as microsatellites, creating insertion-deletion loops (IDL). If left unrepaired, the daughter strand synthesized during the next round of replication carries a single nucleotide point mutation or a frameshift mutation that usually leads to a premature STOP codon resulting in the synthesis of a truncated protein (Fig. 1).

Fig. 1.

Eukaryotic DNA mismatch repair. Single nucleotide mismatches, as well as 1 nucleotide loops typically occurring at microsatellite mononucleotide repeated sequences, are recognized by the MutSα heterodimer formed by MSH2 and MSH6; this complex binds to DNA as a sliding clamp after MSH2 ADP has been exchanged for ATP (upper panel). The heterodimer MutSβ, formed by MLH1 and PMS2, then binds to MutSα. The MutSα-MutSβ complex moves along the DNA until it meets the DNA polymerase complex, PCNA and Exo I. The exonuclease then removes up to several hundreds bases from the newly synthesized DNA strand, to allow correct resynthesis of DNA by the replicative polymerase. IDL of two or more nucleotides are preferentially recognized by the MutSβ complex, formed by MSH2 and MSH3 (lower panel). After binding to DNA, MutSβ exchanges ADP for ATP and recruits the MutLα complex, a heterodimer of MLH1 and PMS2 heterodimer. Subsequently, excision and resynthesis are performed as described for the repair of single base mismatches. MLH1 may also form heterodimers with MLH3 and PMS1 to form respectively the MutLγ and MutLβ complexes, but their roles in human MMR function are not entirely clear

Bacterial MMR

In procaryotes, MMR system involves three specialized proteins: MutS, MutL, and MutH [107]. MutS is an ATPase acting as a homodimer that translocates along DNA, recognizes base-base mismatches, as well as small IDL, and promotes DNA loop formation [7, 134]. After loading onto the DNA, MutS undergoes a conformational change, allowing its interaction with MutL in an ATP-dependent manner [62]. MutL, another ATPase acting as a homodimer, couples mismatch recognition to further steps resulting in the removal of a large newly replicated DNA fragment containing the mismatch [12]. MutH is an endonuclease that can nick the newly synthesized nascent strands, which are tagged by the transient absence of methylation at GATC sequences [197].

Yeast and mammalian MMR

Yeast has six homologs of MutS (MSH1-MSH6), but mammals lack MSH1 homolog, which is involved in the repair of mitochondrial DNA [33]. With the exception of MSH4 and MSH5, which promote crossing-over and chromosome synapsis during meiosis [50, 82, 160], all other MutS homologs participate in mutation avoidance. Four homologs of MutL have been identified both in yeast (MLH1, MLH2, MLH3, PMS1) and mammals (MLH1, MLH3, PMS1, PMS2). These multiple homologs of MutS and MutL play specialized but partly redundant functions in yeast and mammals. MSH2 forms a heterodimer with either MSH6 or MSH3, giving rise to the formation of MutSα and MutSβ complexes, respectively, with different relative abilities to bind to base-base mismatches and to IDL [4, 63, 127, 128] (Fig. 1). Studies performed in yeast have shown that MutSα complexes are able to efficiently bind to most base-base mismatches (with the exception of CC mismatches), and to loops of one or two nucleotides, whereas MutSβ preferentially repairs heteroduplexes with larger IDL [6, 127]. In human cells, the mismatch recognition specificities of MutSα and MutSβ are very similar to those of their yeast counterparts. MutSα is present at around ten times higher levels than MutSβ and plays the major role, while MutSβ is dedicated to the repair of IDL larger than two repeated motifs [4, 63, 126, 144]. Cells deficient in both MSH3 and MSH6 have a high mutation rate, similar to that observed in MSH2-deficient cells, whereas disruption of either gene leads to a weak mutator phenotype, owing to the partially redundant functions of MSH3 and MSH6 [41, 51]. MSH2 is the most abundantly expressed MutS homolog, and its presence is necessary to prevent degradation of MSH6 and MSH3 [31].

Yeast and mammals have four MutL homologs; yeast PMS1, MLH1, MLH2 and MLH3 correspond to human PMS2, MLH1, PMS1 and MLH3, respectively [137]. In humans, MLH1, the major MutL homolog, interacts with its three partners, PMS2, PMS1 and MLH3 through a unique region, to form MutLα, MutLβ and MutLγ, respectively [104]. Thanks to its latent endonuclease activity that is activated in the presence of a mismatch, MutSα, RFC (replication factor C), PCNA (proliferating cell nuclear antigen) and ATP, MutLα introduces nicks into DNA to help degradation of the mismatch-containing strand by EXO1 [96]. MutLα is the major component in MMR, operating in conjunction with either MutSα or MutSβ to repair both base-base mismatches and IDL. In vitro studies failed to identify a role for human MutLβ formed by MLH1 and PMS1, in postreplicative MMR [155]. Yeast MutLγ, the heterodimer of MLH3 and MLH1, acts in concert with MutSβ to repair a proportion of the IDL, decreasing the rate of frameshift mutations [60]. In mammals, MutLγ seems to recognize MutSα substrates, such as G/T mismatch and small IDL, rather than larger loops recognized by MutSβ [27]. MLH1 is required to stabilize PMS2 expression in human cells [31], but is dispensable for MLH3 stability, which is generally much less abundant than the two other partners of MLH1 [27]. Thus, MutLγ is unlikely to play a major role in MMR in vivo, but may act as a backup to MutLα in the absence of PMS2 [27].

Other functions of MMR

In addition to correcting misincorporated nucleotides, MMR proteins have further been implicated in double-strand break repair and recombination [76]. MMR also recognizes mismatches in heteroduplex recombination intermediates preventing completion of recombination between diverged sequences, thus contributing to genetic stability [156, 176]. MMR components are involved in DNA damage response, cell cycle regulation and in the p53-dependent apoptotic DNA damage response to a variety of compounds [19, 93, 141, 175].

Colorectal oncogenesis associated with MMR deficiency

MMR defects involved in hereditary and sporadic cases

The predisposition of some families to gastrointestinal and endometrial cancers was first reported in 1913 by Dr Whartin [194]. In 1966, Dr Lynch studied two large families who developed multiple primary cancers, notably adenocarcinomas of the colon, endometrium and stomach, and demonstrated an autosomal dominant pattern of inheritance [122]. In the early 90’s, linkage analysis of large kindreds with CRC, performed to identify regions of loss of heterozygosity in the tumour led to the mapping of two susceptibility loci to chromosome 2p21-22 [1, 139, 148] and 3p21 [139, 140]. Based on the similarities between microsatellite genetic instability and the mutator phenotype displayed by MMR-deficient bacteria and yeast, MSH2 and MLH1, the human homologs of E. coli MutS and MutL genes, located on chromosome 2p [59, 110] and 3p [21], respectively, were identified as candidate genes for Lynch syndrome. Mutations of two additional MutL homologs, PMS1 and PMS2, were further reported in patients with HNPCC [138]. The role of PMS1 was later challenged since the mutations occurring in this family was associated with a mutation in MSH2 which co-segregated with MSI CRC, while the PMS1 mutation did not [117]. Germline mutations of MMR genes are predominantly found in MSH2 and MLH1, accounting for approximately 40% and 50% of the families, in whom 200 and 250 pathogenic mutations have been identified, respectively [21, 59, 110, 138, 147].

Lynch syndrome is characterized by an autosomal dominantly inherited predisposition to early onset multiple tumours, most frequently in the colon, rectum, endometrium, stomach, accounting for approximately 2–5% of the total CRCs [74]. Cancer of the stomach, ureter, renal pelvis, small bowel, bile ducts and tumours of the brain are also associated with Lynch syndrome and are preferentially associated with MSH2 inactivation [195]; the lifetime risk of developing one of these extracolonic epithelial-derived tumours remains relatively low (<10%) [185]. The risk of developing endometrial cancer is highest in families with MSH6 mutations (64–71%) compared to patients carrying MSH2 or MLH1 mutation carriers (40–50%) [25, 80, 167]. The mean age at diagnosis for CRC is 43–46 years for MSH2 or MLH1 mutations carriers [15, 80, 147], and 51–57 years for MSH6 mutation carriers [15, 80]. Regarding MLH1 and MSH2 mutation carriers, the lifetime risk for CRC is estimated to be 60–80% [2, 69, 75, 174].

Approximately 10 to 15% of all CRCs display MSI in absence of germline mutations in one of the genes involved in MMR. In the overwhelming majority of cases, MMR defect results from inactivation of MLH1, due to biallelic methyation of its promoter [38, 81, 98]. Similarly to Lynch syndrome, sporadic MSI tumours are most frequently diploid and patients have a better prognosis than those with MSS tumours. However, whereas patients with Lynch syndrome are younger than those with sporadic CRC, sporadic MSI cases are older. Some studies have suggested that MSI phenotype could be associated with mutations in EGFR signalling pathway, such as PI3K/PTEN or BRAF. Mutations in PIK3CA, which encodes the catalytic subunit of PI3K, have been reported to be more frequent in MSI than in MSS CRCs [3], but these results have not been observed in other series of patients [142, 188]. Nevertheless, PTEN, which is the only tumour suppressor in this pathway, is not only mutated but also epigenetically silenced with higher frequency in MSI than MSS tumours [66]. Regarding the alternative EGFR signalling pathway, BRAF mutations, occurring usually at the V600E hot spot, are frequent in sporadic MLH1-deficient MSI tumours in contrast with hereditary MLH1-deficient tumours [42, 129, 192]. Finding the common V600E BRAF mutation, together with MLH1 promoter methylation, can serve as excluding a diagnosis of Lynch syndrome since BRAF mutation are virtually absent in hereditary cases [47, 121]. BRAF mutations are commonly associated with a group of sporadic carcinomas characterized by the CpG island methylator phenotype (CIMP), including sporadic MSI CRCs due to methylation of the of MLH1 promoter [84, 97, 196]. Although the concept of CIMP is not yet clearly established, CIMP is strongly associated with sporadic MSI CRCs where MLH1 and p16 promoter methylation occurs, and is less frequent with MSS CRCs, in which promoter methylation targets other tumour suppressor genes [43, 67]. The very high frequency of BRAF mutations in colorectal polyps with serrated morphology showing extensive DNA methylation further indicates that sporadic MSI CRCs may originate from sessile serrated adenomas, rather than from conventional adenomas [97].

Inactivation of target genes through MSI-driven mutations

Microsatellites are ubiquitous in all genomes explored so far and are especially abundant in eukaryote genomes. Simple sequence repeats are not confined to intergenic non-coding repeated sequences but are also present in exonic sequences, most often mononucleotide repeats, in which an insertion or a deletion of a repeated motif leads to frameshift mutations. Accumulation of such mutations in genes harbouring a coding repeat is considered as the major molecular mechanism driving MMR-deficient cells oncogenic transformation. Inactivation of the transforming growth factor β receptor 2 (TGFBR2) and the activin type II receptor (ACVR2) genes have been established as playing a key role in MSI-driven colorectal oncogenesis (Fig. 2) [79, 94, 145]. These frameshift mutations cause loss of protein expression and tumour suppressor growth mediated through SMAD signalling [95, 125]. In absence of functional studies, discriminating between the genes that are true targets for colorectal MSI-driven mutagenesis from those whose mutations occur randomly and are left unrepaired due to the absence of MMR is not self-evident. High mutation prevalence in a given gene is generally regarded as strongly indicating its oncogenic role, although mutation rates of microsatellites are extremely variable, being influenced by flanking sequences, sequence composition, motif unit size, and recombination rate [26, 36, 37]. A bi-allelic inactivation and a role in a growth suppressor pathway are among other widely accepted criteria, but no consensus has been achieved, especially no mutation incidence cut-off value could be defined [18, 48]. To overcome this problem, several statistical analyses based on gene mutation frequency independently of functional considerations have been proposed [49, 199]. A database of human mononucleotide microsatellite mutations and their potential impact to tumorigenesis called SelTarBase for “Selective Targets in Human MSI-H Tumorigenesis Database” was recently built [200]. Today, 74 coding mononucleotide repeats located within the nearly 600 genes analyzed have been reported as predicted targets genes [200]. Classes of genes mutated in MSI tumours include DNA damage signalling and repair, apoptosis, cell growth and differentiation, as well as transcription factors [169, 189, 200]. The 1,291 potentially hypermutable human genes that host a coding mononucleotide repeat longer than seven base pair are implicated in a restricted number of functions, such as cell cycle and maintenance of DNA integrity [119]. In keeping, DNA damage signalling and repair genes are frequently mutated in MSI CRCs [64, 133], endometrial cancers [187], as well as urothelial cell carcinomas of upper urinary tract [135].

Fig. 2.

Molecular pathways involved in the development of MSI CRC. MSI CRCs arise either in a family context known as Lynch syndrome, which accounts for 2–5% of all CRCs or in sporadic cases, representing approximately 12–15% of all CRCs. Hereditary tumours develop at a young age due to the presence of a germline mutation in one of the MMR genes, followed by inactivation of the wild-type allele through gene conversion, loss of chromosome, mutation or methylation (especially for MLH1). Most sporadic cases of CRC displaying MSI are due to methylation-induced silencing of MLH1. Sporadic MSI CRCs are generally diagnosed at an age older than for all CRCs, and frequently carry the common BRAF V600E mutation. Defects in MMR results in the accumulation of somatic mutations in genes containing microsatellite repeated sequences, referred to as target genes. The identification of the relevant genes among the multitude that carry mutations in a proportion of MSI CRCs remains a challenge for understanding this oncogenic process. Several target genes involved in key cellular functions such as DNA damage signalling and repair, apoptosis, signal transduction and transcription regulation are suspected to play a critical role in tumour initiation and/or progression. Only genes carrying mutations in more than 10% MSI CRCs are shown; mutation frequency data are from the 201006.25, June 2010 SelTarbase release (http://www.seltarbase.org/) [200]

Experimental models used to evaluate the role of MMR on the cytotoxicity of chemotherapeutic agents

Considering that most cancer chemotherapy agents target DNA and create various types of adducts, the question of whether MMR-deficient tumours behaved differently or not has been largely investigated. Studies have been conducted either with matched, isogenic pair of cell lines, where the cell line differences are supposed to be restricted to the mutated MMR gene or by comparing a panel of MMR-deficient cell lines with a panel of MMR-proficient cell lines [85, 112, 158]. Both approaches have their pitfalls due to intrinsic characteristics of MMR-deficient cells : (1) cells accumulate a great number of mutations that may be selected during oncogenesis or during cell culture, (2) since these mutations occur preferentially in repeated sequences through insertion or deletion of the repeated motif, they are highly dynamic and may revert in a proportion of the tumour cells [83, 132]. Although MMR-deficient cell lines derived from MSI cancers may share a common defect in an MMR gene, they are highly variable with respect to the target gene mutation spectra; MSI cell lines may be seen as “mutation factories”, a characteristic that contributes to some of the discrepancies observed between laboratories. In order to provide useful information regarding the response of MSI tumours to treatments, this approach is reliable only if a reasonable number of cell lines are investigated. This approach is interesting as it recapitulates the MSI pathogenesis rather than investigating the impact of MMR defect per se, a situation that is seen only at the very beginning of the disease. This is particularly relevant since many genes involved in DNA damage signalling and/or repair carry microsatellite repeats that are frequently mutated in an MSI background [11, 64, 65, 101, 111, 112, 119, 133]. Furthermore, cell lines selected after introduction of a chromosome or transfection of a cDNA-expressing vector may not be truly isogenic as they undoubtedly differ in the mutation status of several target genes, a situation that may complicate the interpretation of the data (unpublished personal observations). In order to circumvent the problem caused by the heterogeneity in target gene mutations, an isogenic system, in which MLH1 expression is regulated by doxycycline at the level of the whole cell population has been established, giving the opportunity to properly investigate the role of the MMR defect itself [30]. Finally, experimental models to test the in vivo sensitivity have been achieved by xenografting human cancer cells either subcutaneously or orthotopically [20, 151, 158].

5-fluorouracil

5FU has been the gold standard adjuvant chemotherapy for CRC for many years, and is still used in association with other compounds. Most studies indicate that MSI CRC cell lines grown in vitro are more resistant to 5FU and its analogues than MSS cell lines [28, 130]. Interestingly, a direct role for MLH1 could be established in a cell line in which MLH1 deficiency was due to its promoter methylation, by reexpressing MLH1 after addition of the demethylating agent 5-azacytidine [10]. In keeping, the resistance to 5FU of orthotopically xenografted colon cancers was higher for tumours displaying MSI [151].

Platinum compounds

Cisplatin is a chemotherapeutic agent with widespread-spectrum of action on human malignancies that predominantly forms intrastrand crosslinks between adjacent purine. As these DNA adducts are repaired by the nucleotide excision repair (NER) system, cells deficient in this repair pathway are highly sensitive to cisplatin. In contrast to NER, a deficiency in MMR appears to contribute to increased resistance to cisplatin [5, 57, 58]. The 1,2 diguanyl crosslinks are poorly recognized by MutSα unless a mispaired thymine is incorporated opposite the 3′ guanine; thus, crosslinks that have undergone error-prone replication are efficiently recognized by MMR [202]. Cell death results from inefficient cycles of translesion synthesis through cisplatin DNA adducts followed by MMR-mediated removal of the newly synthesized mispaired DNA strand, explaining how loss of MMR confers resistance to cipslatin. Oxaliplatin is a third generation analogue of platinum compound that has shown clinical activity against colon cancer. Oxaliplatin and cisplatin form similar DNA adducts, and modelling studies suggest that both platinum compounds induce similar distortion of secondary DNA structure. However, oxaliplatin-induced adducts differ from those created by cisplatin in that they are poorly recognized by MutS complexes [204]. As expected, loss of MMR activity does not significantly affect oxaliplatin cytotoxicity [181].

Topisomerase inhibitors

Topoisomerases are nuclear enzymes that modify the topological state of DNA and participate in replication, transcription, repair, recombination, and chromosome segregation. Topoisomerases catalyze the relaxation of supercoiled DNA through the transient cleavage of DNA strands, creating single or double strand breaks upon action of topoisomerase I or topoisomerase II, respectively. Camptothecin and etoposide are among the many anticancer drugs targeting topoisomerases I and II. Irinotecan, a water-soluble camptothecin derivative has proven its clinical efficacy in CRC. Although not all studies agree [56], MSI cell lines are generally more sensitive to topoisomerase I inhibitors compared with MMR-proficient cell lines [85, 123]. Xenografted human MSI CRCs are also more sensitive to irinotecan but the difference was more modest than observed in vitro [20]. MSI-driven mutations occurring in genes involved double strand break signalling and repair, such as ATR, MRE11, RAD50 and DNA-PKcs alter efficiency and specificity of the repair process and are most likely responsible for the increased sensitivity of MSI CRCs to drugs generating DNA double strand break [86, 133, 158, 177, 190].

MSI phenotyping: who, how and what for?

Lynch syndrome is the most common hereditary colon cancer syndrome. However, whether large-scale screening should be performed or not remains controversial. A number of strategies have been defined to select high-risk subgroups that should benefit MSI screening. Clinical criteria to select families with high-risk CRC were first established in Amsterdam in 1991 by the International Collaborative Group on HNPCC, referred as Amsterdam I criteria, but they were too restrictive to allow effective diagnosis of HNPCC in small families, whereas large families with random clustering of CRCs could be falsely diagnosed as HNPCC [183]. Several guidelines for testing of colorectal tumours for MSI were further developed to increase sensitivity. The Amsterdam II Criteria proposed in 1998 were derived from the Amsterdam I criteria except that not only CRCs, but also extracolonic tumours are considered [186]. However, among families meeting the most stringent Amsterdam criteria I, 30–50% do not have pathogenic germline mutations in MMR genes and lack tumour MSI [116, 118, 149]. The Bethesda Guidelines developed in 1996 take into account the occurrence of extracolonic tumours as well as some characteristic pathology features such as tumour right-sided location and poor differentiation [180].

The methods and criteria used to define MSI in CRCs are still evolving and no consensus has been reached yet. MSI testing was originally performed on paraffin-embedded tumour tissues using a polymerase chain reaction-based assay to examine genetic alterations at four selected microsatellite loci [179]. A panel of five microsatellites, two mononucleotide and three dinucleotide repeats, was later validated and recommended as a reference panel at the National Cancer Institute workshop held in 1997 [18]. Tumours were classified as high-frequency MSI (MSI-H), if at least two of the five microsatellite markers showed instability, low-frequency MSI (MSI-L) if only one of the five markers showed instability, and microsatellite stable (MSS) if none of the markers showed instability [18]. Since dinucleotide markers are highly polymorphic, the interpretation of the profiles required the comparison with normal DNA from each patient. In order to avoid the need for matching normal DNA, an alternative molecular method based exclusively on quasi-monomorphic mononucleotide markers was developed and proved to be more specific and sensitive than the original National Cancer Institute panel [24]. Detecting MMR protein expression by immunohistochemistry (IHC) represents an alternative method that is widely used. IHC has the advantage of identifying the affected gene by detecting loss of its protein product and to be less expensive and available in numerous pathology departments. MSI testing and IHC are complementary, and loss of MMR protein expression by IHC has been shown to be highly concordant with DNA based MSI testing, especially when MSH6 and PMS2 expression are investigated together with MLH1 and MSH2 [115, 178, 193]. MSI-L and MSS CRCs are generally grouped together because they have similar pathological features and clinical outcome [146].

Because IHC is almost equally sensitive (83%) as the latest molecular MSI technique for tumours with mutations in MLH1 (91%), and MSH2 (87%) or MSH6 (77%), is as specific (90% versus 89%), has the advantage of directing gene mutation screening, and is easy to perform routinely, IHC tends to replace molecular MSI as a screening method for MMR-deficient tumours [39, 184, 185]. MSI tumours tend to share various clinicopathologic characteristics including preferential location in the proximal colon, poor differentiation with a mucinous aspect, a Crohn-like inflammatory reaction and an excess of infiltrating lymphocytes within the tumour that have been used to refine the screening of MSI tumours among familial [74, 87] and/or sporadic cases [68, 159]. Guidelines to detect Lynch syndrome or MSI tumours keep changing as our knowledge improves and should not be seen as definitely established, but evidence accumulates showing that it is time to diagnose MSI tumours in the general population [185]. Indeed, despite their overall better prognosis, MSI tumours do not benefit from 5FU-based adjuvant chemotherapy (see chapter 7.1). Systematic MSI phenotyping would better guide the choice of chemotherapy, increasing its efficiency while reducing its toxicity. Given that the cost of molecular MSI analysis is 10 times higher than by running IHC and is more time-consuming, IHC staining represents an attractive alternative method provided that it is performed in trained laboratories [143]. IHC screening using a 2-antibody panel, MSH6 and PMS2, has recently been shown to be as specific and efficient as a 4-antibody panel also comprising MSH2 and MLH1 [72, 170]. Loss of MSH2 or MLH1 results in concomitant loss of their respective partner, MSH6 or PMS2, while the reverse is not true. On the basis of these biochemical properties and recent observations, it may be attractive to investigate tumour expression of MSH6 and PMS2 as a first-line screening method, and to restrict MSH2 and MLH1 immunostaining to those displaying a defect in MSH6 or PMS2, in order to define the primary gene defect [72, 170].

Impact of microsatellite instability on the prognosis of patients with colorectal cancer

Most studies report that CRCs exhibiting MSI-H phenotype have improved survival compared to MSI-L or MSS tumour, but the prognostic value of MSI status varies significantly between studies. This may be because relatively few studies were population-based or controlled for tumour stage as well as other known confounding factors such as the age. Indeed, when patients with Lynch syndrome, who tend to be younger than average, are excluded from a series of patients, MSI tumours are mainly those arising in sporadic cases; the fact that they are older than the average may contribute to their shorter overall survival (OS). Moreover, patients receiving chemotherapy were most often analysed together with those who were treated by surgery alone. The favourable clinical outcome of patients with MSI-H tumours was already observed when this form of genetic instability was first discovered in CRCs [120, 179]. Subsequently, several retrospective studies either restricted to patients with stages II or III CRCs [34, 73, 109, 201] or including CRCs at all stages [14, 61, 165] found that MSI-H CRCs had a significantly better outcome compared with MSI-L or MSS tumours. In a large study that focused on patients developing CRC before 50 years of age at diagnosis, MSI was found in 17% of the 607 young patients examined and was associated, in a multivariate analysis, with a significant advantage in survival independently of all standard prognostic factors, including tumour stage [70]. In additional multivariate analyses, the survival advantage conferred by the MSI-H phenotype was generally shown to be independent of tumour stage and other clinical and pathologic variables [14, 61, 165]. Most of this risk reduction occurred in patients with stage III tumours and to a lesser extent to those with stage II tumours [14, 165], while MSI, which is less frequent in stage IV CRCs, appeared to confer little [165] or no [14] prognostic advantage. The better prognosis of patients with stages II and III tumours was observed in series of patients not receiving chemotherapy [14, 34, 109, 201]. Improved prognosis of patients with MSI tumours was further confirmed in a study based on five randomized trials of 5FU-based adjuvant chemotherapy trials in which 570 patients with stages II and III colon cancers were enrolled [157]. Indeed, among patients not receiving chemotherapy, those with MSI-H tumours had a significantly higher rate of five-year disease-free survival (DFS) and longer OS (P = 0.004) than those with MSI-L or MSS tumours [157]. These observations have recently been confirmed in another study based on five completed randomized clinical trials including patients with pathologically confirmed stage II (n = 218) or III (n = 239) colon cancers [166]. In univariate models in patients treated with surgery alone, MMR deficient status was associated with improved DFS (P = 0.03) and a trend towards improved OS (P = 0.06) [166]. These trends remained in multivariate models for both DFS and OS but were not significant (P = 0.17 and P = 0.24; respectively) [166].

However, some studies failed to demonstrate significant differences in recurrence-free survival (RFS), DFS or OS between patients with MSI-H and MSS/MSI-L, whether analysis was restricted to stages II and III CRCs [99] or to tumours at all stages [13, 108, 162]. One was based on approximately one third of patients participating in a large cooperative group trial conducted from 4 randomized colon cancer adjuvant treatment studies (n = 542), among whom 173 had not received chemotherapy [99]. The hazard ratio (HR) for RFS when comparing MSI-H to MSS/MSI-L patients was 0.77 (95% CI, 0.40 to 1.48) in the untreated group, indicating that there was a trend towards improved RFS in patients with MSI-H CRC although the difference was not statistically significant [99]. Another study was performed on 416 consecutive patients with stages I to IV CRCs, among whom 202 did not receive chemotherapy. DFS of patients with stage III as well as pooled data of patients with stages II and III was better in patients exhibiting an MSI-H phenotype (n = 35) than in patients with MSS tumours (n = 218), but the differences did not reach statistical significance (P = 0.28) [108]. There was no difference in OS between the subgroups, neither in the study cohort as a whole nor in early stage or locally advanced cancers (stages I to III) [108]. These results further confirmed a previous study based on a series of 181 unselected stages I to IV CRCs, in which survival analysis revealed no statistically significant difference in prognosis between MSI and MSS cases, although a trend towards better survival for MSI cases was observed [162]. The marginal influence of the MSI status on prognosis observed in these latter studies may be due to unidentified confounding factors in retrospective studies or insufficient sample size to detect moderate effect of MSI [108, 162]. Yet, the prognostic impact of MSI-H on survival has clearly been established in a meta-analysis of 32 studies comprising a total of 7,642 patients with stages I to IV CRCs, including 1,277 MSI cases [152]. The analysis included untreated patients, as well as patients receiving 5FU-based adjuvant chemotherapy. The combined HR estimate for OS associated with MSI-H was 0.65 (95% CI, 0.59 to 0.71); this benefit persisted when restricting analyses to patients with stage II or III disease participating in clinical trial [152]. The association between MSI and favourable prognosis was recently confirmed in another meta-analysis based on pooled data from 31 published studies [71].

The possible molecular mechanism underlying the better outcome of patient with MSI CRC remains largely unknown. A number of MSI tumour characteristics may play a role in the evolution of the disease. MSI tumours generally present with a greater depth of invasion but with a lower stage, as defined by the American Joint Committee on Cancer. In patients having developed MSI CRC, distant metastases were found less frequently in studies focusing on young patients, [70], sporadic CRCs [88] or in a large series of consecutive and unselected patients undergoing surgery for CRC [124]. In addition, aneuploidy, a hallmark of tumours with chromosomal instability, is an independent marker of poor prognosis in patients with CRC [191]. Given that alterations in tumour cell DNA content are inversely associated with MSI-H phenotype, whether both MSI and DNA ploidy contributed to improved clinical outcome was investigated using data obtained from 6 randomized 5FU-based phase III adjuvant studies [173]. MSI-H and DNA ploidy status were found to be independent prognostic variables, with ploidy being the strongest marker; diploidy was associated in better survival in both MSI and MSS patients groups [173]. Finally, tumour-infiltrating lymphocytes (TIL) may also contribute to the survival advantage of patients with MSI CRC by eliciting a protective anti-tumour immune response against tumour-specific carboxy-terminal frameshift neopeptides accumulating as a result of frameshift mutation hotspots in MMR-deficient cells [23, 168, 172]. MSI tumours are characterized by strong tumour lymphocyte infiltration [131], a criterion that has recently been proposed as a marker to more accurately predict MSI in CRCs [68]. TIL in MSI CRCs are activated and cytotoxic containing higher ratios of CD8/CD3 mRNA copy numbers than TIL infiltrating MSS tumours [150]. Interestingly, a recent study reported that CD8⁺ and CD45RO⁺ T cell densities were higher in tumours showing a lymphocytic response or MSI, while FOXP3+ was not associated with either of these features [163]. These observations are in keeping with a recent study reporting that CD8⁺FOXP3⁺ cells display strong immunosuppressive properties in vitro [32].

Predictive impact of microsatellite instability for adjuvant chemotherapy

Patients with MSI-H tumours are generally treated according to the same strategy as those with MSI-L or MSS phenotype, with the stage of the tumour being the major determinant for choosing patients who will receive adjuvant chemotherapy. However, evidence accumulates showing that MSI status may also be important in predicting clinical outcome and response to chemotherapy.

5-fluorouracil-based adjuvant chemotherapy

The predictive role for MMR status has been evaluated in numerous studies investigating the outcome of patients receiving adjuvant 5FU-based therapy or undergoing surgery alone (Table 1). These studies led to contradictory results, with some reporting a benefit from 5FU for patients with MSI tumours, while others concluded that 5FU had no benefit or even a deleterious effect. Various reasons may account for these discrepancies, (1) some studies focused on stage III patients only, while in others, patients with disease at various stages were mixed, (2) in some, patients having rectal carcinomas were excluded, restricting investigations to patients with colon cancer exclusively, (3) multivariate analysis were not always performed. Above all, the patients were generally not allocated to a 5FU-based treatment at random, allowing for selection bias or any of the many other traps associated with nonrandomized comparisons.

Table 1.

Impact of microsatellite instability on clinical outcome of adjuvant CRC patients

Reference	Type of study	Number of patients	Tumour stage	Number of MSI tumours (%)	Number of patients receiving adjuvant CT	Comparison	Survival analysis criteria	Results
5-fluorouracil-based adjuvant chemotherapy
Elsaleh et al., 2000 [52]	R	656	III	56 (8.5)	272	Adj CT: MSS vs MSI	5 year-OS	Longer survival for MSI
Hemminki et al., 2000 [78]	P/NR	95	III	11 (12)	95	Adj CT: MSS vs MSI	3 year-RFS	Longer survival for MSI
Ribic et al., 2003 [157]	RCT	570	II + III	95 (16.7)	283	Adj CT: MSS vs MSI	5 year-DFS	No benefit or detrimentala
						MSI: surgery alone vs adj CT	5 year-OS
Carethers et al., 2004 [29]	R	204	II + III	36 (17.6)	66	MSI: surgery alone vs adj CT	OS	No benefit
de Vos Tot Nederveen Cappel et al., 2004 [40]	R	92	III	92 (100)b	28	MSI: surgery alone vs adj CT	5 year-OS	No benefit
Benatti et al., 2005 [14]	R	1263	All stages	256 (20.3)	304	MSI: surgery alone vs adj CT	5 year-OS	No benefit
Westra et al., 2005 [198]	RCT	273	III	44 (16)	273	Adj CT: MSS vs MSI	5 year-DFS	Longer survival for MSIc
Jover et al., 2006 [92]	P/NR	754	All stages	66 (8.8)	260	MSI: surgery alone vs adj CT	OS	No benefit
Lanza et al., 2006 [109]	R	718	II + III	114 (15.9)	193	Adj CT: MSS vs MSI	6 year-OS	No benefit
						MSI: surgery alone vs adj CT
Kim et al., 2007 [99]	RCT	542	II + III	98 (18)	369	Adj CT: MSS vs MSI	5 year-RFS/OS	No significant differenced
Lamberti et al., 2007 [108]	P/NR	416	All stages	52 (13)	89e	Adj CT: MSS vs MSI	OS	No significant difference
Sargent et al., 2010 [166]	RCT	457	II + III	70 (15)	229	Adj CT: MSS vs MSI	5 year-DFS	No benefit
	RCT	1027f	II + III	165 (16)	512	MSI: surgery alone vs adj CT	5 year-OS	No benefit or detrimental
Oxaliplatin-containing adjuvant chemotherapy
Kim et al., 2010 [103]	R	135	All stages	12 (8.8)	121	Adj CT: MSS vs MSI	3 year-DFS	No significant difference
							3 year-OS
Zaanan et al., 2010 [203]	R	233	III	32 (14)	5FU, n = 124	MSI: 5FU vs FOLFOX	3 year-DFS	Longer survival with FOLFOX
					FOLFOX, n = 109
Irinotecan-containing adjuvant chemotherapy
Bertagnolli et al., 2009 [16]	RCT	702	III	96 (13)	5FU, n = 348	IFL: MSI vs MSS	5 year-DFS	Longer survival for MSI treated with IFL
					IFL, n = 354	MSI: 5FU vs IFL

Adj adjuvant; CT chemotherapy; FOLFOX 5FU/leucovorin/oxaliplatin; IFL irinotecan/5FU/leucovorin; NR non-randomized; P prospective; R retrospective; RCT randomized controlled trial

^aTrend towards deleterious effect on MSI tumours: 5-yr DFS 70.7 vs 88.0%, P = 0.07. ^bAll patients are from HNPCC families and display MSI tumours. ^cNot significant in multivariate analysis (P = 0.06). ^dTrend towards benefit for MSI in terms of RFS but results are not significant; identical OS. ^e89 stage III patients received CT, no information for stage II patients. ^fPooled data set from patients included in references [157] and [166]. Deleterious effect of 5FU on OS for stage II patients with MSI tumours (P = 0.04)

Numerous reports concluded that survival of patients with MSI CRC tumours were not improved by 5FU-based chemotherapy [14, 29, 40, 92, 109] (Table 1). Fluorouracil-based adjuvant chemotherapy has even been suspected to be detrimental for patients with MSI colon cancer [157, 166] (Table 1). In a retrospective study of 204 consecutive stages II and III CRC patients, multivariate analysis showed no difference in survival between MSI-H and MSS groups when the analysis did not take into account whether patients were treated or not by 5FU (P = 0.88) [29]. There was a significant OS benefit among stages II and III patients who were treated with 5FU-based chemotherapy compared to those who were not (P = 0.04) [29]. When categorized by the MSI status, there was no difference in survival among patients with MSI tumour, irrespective of whether they received 5FU-based chemotherapy (P = 0.52) [29]. By contrast, there was a significant survival benefit among patients who had MSS or MSI-L tumours and who received 5FU (P ≤ 0.05) [29]. Another study including 718 consecutive patients with adjuvant CRC (393 stage II and 325 stage III) showed a significantly better clinical outcome for patients whose tumour lacked expression of either MLH1 or MSH2 [109]. In stage III CRCs, the survival advantage conferred by absence of detectable expression of MMR protein in tumours was more evident among patients undergoing surgery alone than among patients who further received adjuvant 5FU chemotherapy [109]. Similar conclusions were drawn in a prospective study investigating 505 patients with stage II (n = 296) or stage III (n = 209) CRCs [90]. Consistent with previous reports, patients with stage II or III CRCs treated with 5FU chemotherapy had better DFS and OS, but the survival benefit of 5FU treatment was again restricted to the group of patients with MMR proficient tumours [91, 92]. In HNPCC families, the 5-year survival of subjects with stage III colon cancer that were treated with (n = 28) and without (n = 64) adjuvant 5FU did not differ [40]. Benatti et al. reported that the use of 5FU-based treatment was not a significant factor for survival in a cohort of patients with stage II (n = 491) or III (n = 461) CRCs [14]. There was no difference in survival among patients with MSI tumour (22%), irrespective of whether they received 5FU-based chemotherapy, even when adjusting for stage [14]. This also held true for patients with MSS stage II tumours, while those with MSS stage III CRC had a significant survival advantage when receiving 5FU therapy [14].

The impact of MSI status to predict the benefit of 5FU-based adjuvant chemotherapy was further evaluated in a study based on five randomized phase III trials in which 312 patients with stage II and 258 with stage III CRC were enrolled [157]. Among patients who received chemotherapy, those with MSI-H tumours were associated with a slightly lower 5-year survival rate compared to the MSI-L or MSS tumours (70.7% vs 75.5%; P = 0.66). Chemotherapy with 5FU in patients with MSI-H tumours was associated with a worse outcome (HR for death, 2.14; P = 0.11). Altogether, these results suggest that 5FU-based chemotherapy does not improve survival of patients with MSI-H CRC and may even be detrimental with a nonsignificant, 2-fold excess in mortality among stage III MSI CRC patients and a 3-fold increase in mortality among stage II MSI CRC patients [157].

In order to provide an independent validation of the findings of Ribic et al., an international collaboration was established to investigate whether patients with MSI tumour benefit from 5FU-based chemotherapy or not. This study was based on the data obtained from 457 patients with stage II or III colon cancer previously enrolled onto five completed, randomized clinical trials of 5FU-based treatment versus no-treatment control [166]. Since the findings obtained with these patients were highly consistent with those previously obtained by Ribic [157], the two data sets were pooled. In patients not treated with FU-based therapy, deficient MMR status was associated with improved DFS (P = 0.009) and OS (P = 0.004). Only patients with stage III disease and MMR-proficient tumours benefited from chemotherapy [166]. There was a statistically significant decreased OS in patients with stage II disease and MMR-deficient tumours who were treated compared with patients who underwent surgery alone (P = 0.04); in keeping, there was a trend for worse DFS in this group, but the difference was not statistically significant [166]. Neither DFS nor OS was improved by chemotherapy in patients with stage III MMR-deficient tumour [166].

On the other hand, several studies failed to demonstrate a survival difference among patients treated by 5FU between MSI and MSS tumours [99, 108], whereas others reported a greater benefit of adjuvant 5FU chemotherapy for MMR-deficient tumours [52, 53, 78, 198]. One study used data from 391 stage III colon cancer patients who participated in a prospective randomized trial evaluating adjuvant 5FU-based chemotherapy, in which MSI phenotype was determined in 273 patients [198]. In univariate analysis, MSI-H tumours were associated with a longer DFS (P = 0.04), but in a multivariate model adjusting for nodal involvement, histology, invasion, and grade of tumour, the association of MSI-H status with DFS no longer reached statistical significance (P = 0.06) [198]. The retrospective series of patients with stage III CRC published by Elsaleh et al. reported that among patients who received 5FU-based chemotherapy, those whose tumour displayed MSI had longer survival [52, 54]. In the earliest report, patients receiving 5FU were on average 13 years younger than those who did not, which may represent an additional bias [52]. Likewise, another study conducted in a group of 95 stage III CRC patients receiving adjuvant 5FU-based chemotherapy after surgery showed that the 3-year RFS was 90% and 43% for MSI and non-MSI tumours (P = 0.02), respectively [78]. A better clinical outcome of patients with MSI tumour was also reported in a large cooperative group trial conducted from 4 randomized colon cancer treatment studies where 369 patients were treated by 5FU chemotherapy, among whom the 61 patients whose tumour exhibited an MSI-H phenotype had a trend for longer RFS compared to MSS/MSI-L patients, but the differences were not statistically significant [99]. In keeping, Lamberti et al. found no significant difference in OS among patients with stage III CRC treated with 5FU adjuvant chemotherapy when comparing MSI (n = 5) and MSS (n = 81) tumours (P = 0.55) [108].

Finally, a recently published meta-analysis was performed for treated and non-treated MSI population on seven studies representing 3,690 CRC patients (810 stage II and 2,444 stage III), including 454 MSI cases [45]. A total of 1,444 patients received 5FU-based chemotherapy, whereas 1,518 patients did not. There was no survival difference among MSI patients whether or not they received chemotherapy, whereas MSS patients had a better response to chemotherapy, suggesting that MSI could be considered as a predictive marker of chemoresistance to 5FU [45]. Another recent meta-analysis, based on seven studies stratified by 5FU-based adjuvant chemotherapy of 2,863 stages II and III CRC patients of whom 396 were MSI tumours, reported similar results [71]. The MSS 5FU-treated patients showed a significant better OS as compared to untreated MSS patients (OR = 0.52; 95% CI, 0.4 to 0.6; P < 0.0001).The pooling data of the three randomised studies confirmed the benefit of 5FU treatment for MSS patients (OR = 0.62; 95% CI, 0.47 to 0.82; P = 0.003), while the beneficial effect of 5FU was not obvious for the 214 MSI patients (OR = 1.03; 95% CI 0.25 to 4.3; with evidence heterogeneity: P = 0.03) [71]. Such clinical outcomes are consistent with findings of preclinical data, which showed that 5FU mediated cytotoxicity may be dependent on intact DNA MMR gene function [28].

The lack of benefit from 5FU-based chemotherapy in patients with MMR-deficient tumours justifies to routinely assess MMR status at least for patients being considered for 5FU therapy.

Oxaliplatin-containing adjuvant chemotherapy

The addition of oxaliplatin to 5FU improves the adjuvant treatment of stage II and III colon cancer by reducing the risk of recurrence and increasing OS [8, 9, 106]. However, adding oxaliplatin did not improve OS in unselected patients with stage II disease [9]. Based on these trials, infusional fluorouracil, leucovorin, and oxaliplatin, known as FOLFOX regimen, is now the current standard adjuvant therapy for stage III colon cancer. Up to now, the predictive impact of MSI status upon the chemosensitivity to FOLFOX remains poorly explored. A recent retrospective study was performed on 135 patients treated by FOLFOX chemotherapy after curative resection for colon adenocarcinoma (13 stage II and 108 stage III: 12 MSI) or with metastases to liver only (14 stage IV: no MSI) [103]. There was no significant difference in terms of DFS and OS according to the MMR status [103] (Table 1). In order to determine the value of MSI phenotype to predict benefit from adjuvant chemotherapy, we recently performed a retrospective study including 233 unselected patients with stage III colon cancer treated by 5FU and leucovorin alone (n = 124) or with oxaliplatin (n = 109). In this series, among the patients with stage III colon cancer displaying MSI, those treated by FOLFOX had a significantly longer 3-year DFS than those receiving 5FU chemotherapy (P = 0.01) [203]. However, because no disease recurrence was observed among patients with MSI tumours treated by FOLFOX, multivariate analysis to test whether MSI was an independent factor could not be performed [203]. Interestingly, in the subset of patients with MSS tumours, adding oxaliplatin to 5FU remained beneficial, but the improvement in DFS was slight and was not statistically significant (P = 0.15). Thus, MSI phenotype seems to have a favourable impact on the efficacy of adjuvant treatment by FOLFOX in patients with stage III colon cancer although our results require confirmation in samples from randomized trials. Since future trials comparing FOLFOX to 5FU in adjuvant treatment will probably never be done, prospective validation will not be feasible. In view of this, the effort currently made to analyze materials of previously completed trials such as the MOSAIC study represents a unique chance to translate these observations into clinical application [8].

Irinotecan-containing adjuvant chemotherapy

Unlike oxaliplatin, irinotecan added to 5FU as adjuvant therapy in patients with stage III colon cancer did not confer a statistically significant improvement in DFS or OS compared with 5FU alone [164, 182]. Bertagnolli et al. evaluated the impact of MSI phenotype in 723 of the 1,264 stage III CRC patients randomly assigned to receive either 5FU-based chemotherapy alone or with irinotecan [16]. Patients with MMR-deficient tumours (n = 96) treated by 5FU and irinotecan showed an improved 5-year DFS as compared with those whose tumour normally expressed MMR proteins (P = 0.03) [16]. This relationship was not observed among patients treated with 5FU-based chemotherapy alone. A trend towards longer DFS was observed in patients with MMR-deficient tumours treated by 5FU and irinotecan as compared with those receiving 5FU alone (P = 0.07) [16]. Loss of tumour MMR function may predict improved outcome in patients treated with the 5FU and irinotecan regimen as compared with those who do not receive irinotecan.

In summary, adding oxaliplatin to 5FU improves clinical outcome for stage III colon cancer; results for stage II are controversial, though high risk stage II colon cancer most probably benefit from the addition of oxaliplatin to 5FU [8, 9, 161]. MSI is an important molecular marker for prognosis and lack of benefit from 5FU adjuvant chemotherapy. Preliminary data indicate that adding either oxaliplatin or irinotecan to 5FU may overcome the poor response observed in patients with stage III MSI colon cancer. Based on these data, patients with stage III colon cancer should receive FOLFOX, whatever the MMR status. For patients having high risk stage II colon cancer, those with MSS tumours could be treated with 5FU or FOLFOX. Besides, in the present state of our knowledge, patients having high risk stage II MSI colon cancer tumours should be given FOLFOX rather than 5FU, if treatment is needed.

Predictive impact of microsatellite instability for chemotherapy in metastatic colorectal cancer

Significant developments have been made in chemotherapy regimens for the treatment of metastatic CRC (mCRC) over the past decade with introduction of new cytotoxic drugs including fluoropyrimidine, oxaliplatin, irinotecan and biological agents. Several studies have investigated novel predictive factors that could allow to choose the best individualised therapy but none of these markers has been introduced into routine clinical practice except for KRAS mutations that are considered as a predictive marker associated with resistance to anti-EGFR treatment [114, 182]. The predictive value of MSI status on the efficacy of chemotherapy in mCRCs has been investigated in only few studies and one meta-analysis, representing a difficult issue because of the reduced likelihood of metastases at diagnosis (Table 2). Moreover, the chemotherapeutic protocols given to patients with advanced CRC are highly variable, and the parameters to evaluate the impact of MSI differed among the studies being the response rate (RR), progression-free survival (PFS) or OS.

Table 2.

Impact of microsatellite instability in chemotherapy metastatic colorectal cancer

Reference	Type of study	Number of patients	Number of MSI tumours (%)	Protocols of chemotherapy	Outcome analysis criteria	Results
5-fluorouracil-based chemotherapy
Liang et al., 2002 [113]	NR	244	52 (21)	5FU + leucovorin	RR/OS	Better for MSI
Brueckl et al., 2003 [22]	RCT	43	7 (16)	5FU + folinic acid	RR/OS	Better for MSI
Oxaliplatin-containing chemotherapy
des Guetz et al., 2007 [44]	R	40	9 (22.5)	FOLFOX	RR/PFS/OS	No significant difference
Müller et al., 2008 [136]	RCT	108	4 (4)	FUFOX or CAPOX	RR/PFS/OS	No significant differencea
Chua et al., 2009 [35]	P	118	2 (2)	FOLFOX	RR/PFS/OS	No significant difference
Kim et al., 2009 [102]	R	171	10 (6)	FOLFOX or CAPOX	RR/PFS/OS	No significant difference
Irinotecan-containing chemotherapy
Fallik et al., 2003 [55]	R	72	7 (9.7)	5FU + irinotecan	RR	Better for MSI
Koopman et al., 2009 [105]	RCT	515	18 (3.5)	Capecitabine or CAPIRIb	RR/PFS/OS	No significant difference
Kim et al., 2010 [100]	R	200	23 (11.5)	Irinotecan-based regimen	RR/PFS	No significant difference

CAPIRI capecitabine + irinotecan; CAPOX capecitabine + oxaliplatin; FOLFOX 5FU + leucovorin + oxaliplatin; FUFOX 5FU + oxaliplatin; NR non-randomized; P prospective; R retrospective; RCT randomized controlled trial

^aOnly disease control rate (complete/partial remission + stable disease vs progression disease) is significantly better for MSI (P = 0.02). ^bPatients were randomized between first-line capecitabine, second-line irinotecan, third-line CAPOX (sequentiel arm) versus first-line CAPIRI, second-line CAPOX (combination arm); RR and survival were analyzed for first-line treatments only

5-fluorouracil-based chemotherapy

The data regarding the predictive value of MSI-H in the treatment of mCRCs based in 5FU chemotherapy are very limited. The results from two studies suggested that patients with advanced MSI CRC benefit from 5FU-based chemotherapy [22, 113]. Liang et al. observed in 244 mCRC patients a significantly better outcome among MSI-H mCRCs treated with 5FU/leucovorin in terms of RR (65.7% vs 35.1%; P = 0.001), and OS (24 vs 13 months; P = 0.0001), compared to patients with non MSI-H tumours [113]. In their series in which an unusually high proportion of metastatic tumours displayed MSI (>20%), the authors concluded from a multivariate analysis that the better prognosis of stage IV sporadic CRCs with MSI might be associated with better chemosensitivity, rather than lower aggressiveness in biologic behaviour. Similarly, another study including 43 mCRC patients treated in first-line chemotherapy with 5FU and folinic acid showed slightly better RR (72% vs 41%, P = 0.07) and a significant longer median survival (33 vs 19 months; P = 0.02) for MSI (n = 7) patients compared to patients with MSS tumours [22].

Oxaliplatin-containing chemotherapy

A retrospective study reported the predictive value of MSI in patients treated with palliative first-line combination of 5FU and oxaliplatin (FOLFOX) in 40 patients with mCRC, among whom nine exhibited an MSI phenotype. In this study, in which a frequency of MSI was unexpectedly high (22.5%), there was no significant difference between MSI-H and MSS cases in terms of RR, PFS and OS [44]. Similar conclusions were drawn from a series of 118 mCRC patients who underwent FOLFOX treatment from three successive phase II trials [35]. Of note, the incidence of tumour with MSI-H status was especially low in this study, representing only 2% of patients [35]. Likewise, Kim et al. also reported similar outcomes of patients with MSI and MSS tumours in a study including 171 patients who received capecitabine and oxaliplatin (CAPOX) or FOLFOX in first-line combination chemotherapy for advanced CRC [102]. MMR defect displayed by 6% cases did not significantly influence the RR, PFS or OS in patients with mCRC [102]. Another study investigated the value of MSI-H in 108 patients with mCRC who were enrolled in a prospective, randomised trial comparing two 5FU/oxaliplatin-based first-line chemotherapy (FUFOX vs CAPOX) [136]. MSI-H was found in four cases and was correlated with a lower rate of disease control compared to non-MSI-H patients (P = 0.02) [136]. However, there was no correlation between MSI-H and RR, PFS or OS [136].

Irinotecan-containing chemotherapy

Fallik et al. have evaluated the impact of MSI-H phenotype on the tumour response in 72 mCRC patients treated with 5FU and irinotecan after disease had progressed under first-line 5FU-based therapy [55]. In this study, among the seven MSI-H tumours, four responded to chemotherapy, whereas only seven of the 65 MSI-L and MSS tumours did (P = 0.009) [55]. Thus, MSI appeared to predict the tumour sensitivity to 5FU + irinotecan, but the impact on survival was not investigated [55]. Later, Koopman et al. evaluated the impact of MSI status from 515 mCRC patients who were enrolled in a randomised phase III study (CAIRO study) between first-line capecitabine, second-line irinotecan and third-line capecitabine + oxaliplatin (sequential treatment arm) versus first-line capecitabine + irinotecan and second-line capecitabine + oxaliplatin (combination treatment arm), of whom only 3.5% displayed MSI [105]. For first-line treatments, this study showed no significant difference in terms of objective RR (25% vs 31%; P = 0.63), median PFS (4 vs 6.9 months; P = 0.28) and median OS (10.2 vs 17.9 months; P = 0.41) when comparing patients with MSI versus MSS tumours [105]. At last, Kim et al. recently reported the predictive value of MSI in a series of 200 mCRC patients receiving irinotecan-containing regimen as their first-line chemotherapy [100]. MMR deficiency was found in 23 patients (11.5%). The objective RR was 47% in MSS and 56.5% in MSI tumours (P = 0.11). The median PFS tended to be longer with 8.8 months in patients with MSI tumours compared to 6.8 months in patients with MSS tumours, but the difference was not statistically significant (P = 0.09) [100].

In summary, these studies suggest that MSI phenotype may have a predictive value of better clinical outcomes for patients with mCRC treated by 5FU alone [22, 113], but not for patients treated by combination chemotherapy as 5FU with oxaliplatin or irinotecan. These observations are surprising in light of the lack of benefit survival from adjuvant 5FU chemotherapy in patients with MSI tumours. Nevertheless, these results should be interpreted with caution because of the absence of randomized studies, small number of patients in these studies and the highly variable incidence of MSI in mCRCs. In 2009, des Guetz et al. have investigated the predictive value of MSI status on the effect of chemotherapy given to patients with a mCRC by a meta-analysis study. Statistical calculations were performed on six studies representing 964 patients, with 287 patients receiving 5FU-based chemotherapy, and 678 patients receiving combinations of 5FU or capecitabine with oxaliplatin and/or irinotecan [46]. In this meta-analysis, MSI status did not predict the effect of chemotherapy in terms of RR, which was similar in MSI and MSS mCRC tumours, but the impact on survival was not investigated [46]. To our knowledge, there is no study evaluating the predictive value of MSI phenotype in patients treated by chemotherapy combined with targeted therapies such as monoclonal antibodies against vascular endothelial growth factor (VEGF) or epidermal growth factor receptor (EGFR).

General conclusions and perspectives

This review summarizes the increasing evidence showing that MSI CRCs differ from all CRCs in terms of prognosis and response to various treatments. Based on these data, we believe that it is now time to consider performing large scale MSI phenotyping to improve the clinical management of patients with CRC. In this perspective, it becomes essential to define a consensus method for the screening of MSI tumours that is reliable, inexpensive and robust enough to be used in routine laboratories, a goal that has not been achieved yet. In keeping, MSI phenotyping should be systematically investigated in future clinical trials as it represents a powerful biomarker to identify the appropriate personalized treatment regimen based on tumour molecular characteristics. Identifying genes and pathways that are altered preferentially in MSI CRCs should be pursued as it represents a promising opportunity for designing targeted treatments.

Abbreviations

5FU

5-fluorouracil

ACVR2

Activin type II receptor

CAPOX

Capecitabine and oxaliplatin

CIMP

CpG island methylator phenotype

CRC

Colorectal cancer

DFS

Disease-free survival

EGFR

Epidermal growth factor receptor

FOLFOX

5-fluorouracil and oxaliplatin

HNPCC

Hereditary non-polyposis colorectal cancer

HR

Hazard ratio

IDL

Insertion-deletion loop

IHC

Immunohistochemistry

mCRC

Metastatic colorectal cancer

MSI

Microsatellite instability

MMR

Mismatch repair

MSS

Microsatellite stable

NER

Nucleotide excision repair

OR

Odds ratio

OS

Overall survival

PFS

Progression-free survival

RFS

Relapse-free survival

RR

Response rate

TGFBR2

Transforming growth factor β receptor 2

TIL

Tumour-infiltrating lymphocytes

VEGF

Vascular endothelial growth factor