Current Colorectal Cancer Reports

, Volume 6, Issue 3, pp 158–167

Interpreting the Inconsistent Data Concerning the Role of 18qLOH as a Prognostic Marker for Colorectal Cancer

Authors

    • Dana Farber-Brigham and Women’s Cancer CenterBrigham and Women’s Hospital
Article

DOI: 10.1007/s11888-010-0060-3

Cite this article as:
Bertagnolli, M.M. Curr Colorectal Cancer Rep (2010) 6: 158. doi:10.1007/s11888-010-0060-3
  • 20 Views

Abstract

Loss of function of tumor suppressor genes is a condition necessary for the development of colorectal cancer (CRC). Deletions of the long arm of chromosome 18 are present in approximately 70% of CRCs, and this site contains several genes that regulate cellular functions implicated in tumorigenesis, including cell migration, proliferation, and differentiation. Recently, genome-wide association studies also showed a relationship between specific loci on 18q and risk of developing CRC. The presence of 18q loss in tumors has also been associated with poor outcome following CRC treatment. However, despite the more than 20 years since its identification in CRC, we still lack an adequate understanding of the clinical significance of 18q loss. This review details the existing data on the relationship between tumor 18q loss of heterozygosity (18qLOH) and treatment outcome for patients with CRC.

Keywords

Colorectal cancer18qLOHGenomicsPrognostic factors

Introduction

In a 1988 report of the genomic alterations seen in colorectal tumors throughout the adenoma—carcinoma sequence, Vogelstein et al. [1] identified deletions of the long arm of chromosome 18 in 73% of carcinomas and 47% of advanced adenomas, but only 11% to 13% of early adenomas. Subsequent work showed that colorectal cancer (CRC) exhibits at least three types of genomic instability, each producing tumor suppressor loss by a distinct mechanism. The pathway associated with 18q loss is termed chromosomal instability (CIN), a condition of faulty segregation of the sister chromatids during mitosis. As a result of CIN, CRCs exhibit losses at multiple loci, particularly 5q, 8p, 17p, and 18q. A different subset of CRCs, termed microsatellite instability-high (MSI-H) tumors, demonstrates an inability to repair single nucleotide mismatches, a defect identified by observing frequent alterations in di- and trinucleotide repeat sequences known as microsatellites. A third pattern of genomic alteration involves frequent promoter region hypermethylation, resulting in transcriptional silencing of downstream tumor suppressor genes. This DNA pattern, known as CpG island methylator phenotype (CIMP), is identified by polymerase chain reaction (PCR)-based methods that detect specific methylated DNA sites in tumor samples. MSI-H and CIMP are related. In sporadic disease, MSI-H is a subset of CIMP, because promoter methylation is a common method of inducing MSI-H by silencing hMLH1, a gene responsible for DNA mismatch repair.

The identification of distinct subsets of CRC based on their patterns of genomic instability allowed researchers to study whether these subsets demonstrate differences in tumor clinical behavior. For example, a substantial body of data indicated that CRC patients whose tumors demonstrate MSI-H have less aggressive disease. Studies of tumor specimens and clinical data gathered in randomized clinical trials have associated MSI-H status with better prognosis following treatment of stage II and III disease, and suggested that these patients do not benefit from treatment with 5-fluorouracil (5-FU)-based adjuvant chemotherapy [25].

Early studies showed that patients whose tumors lacked significant regions on 18q experienced poor overall survival compared with patients whose tumors maintained intact 18q alleles. The most common methods used to detect these DNA losses rely on assays that examine two distinct alleles at a particular chromosomal location, a condition known as “heterozygosity.” A normal tissue specimen is compared with a tumor sample from the same patient, and DNA loss is identified if one of the alleles present in the normal specimen is lost in the tumor, a condition termed “loss of heterozygosity” or LOH. This work led to multiple studies of the relationship between 18qLOH and treatment outcome for patients with CRC. Unfortunately, despite a considerable number of such reports, it is difficult to draw definitive conclusions from the current literature because of conflicting data and a lack of comparability between studies.

LOH as an Indicator of Tumor Suppressor Loss

The region of 18q lost during CRC development contains several important tumor suppressor genes. The longest gene on chromosome 18q is DCC (deleted in colorectal cancer), a gene first identified as a possible CRC tumor suppressor in a 1990 report by Fearon et al. [6]. This gene encodes a transmembrane receptor that mediates the effects of an adhesion protein, netrin-1, on axon outgrowth. Early studies using antibodies to identify DCC protein expression in tumor specimens by immunohistochemistry (IHC) showed that patients whose tumors were deficient in DCC experienced poor survival compared with those with intact protein [3, 7]. This clinical association was questioned when additional studies of human tumors failed to demonstrate somatic mutations in DCC, and mice heterozygous for Dcc loss (Dcc+/−) did not develop tumors at an increased rate, even when these animals were crossed with a strain bearing a germline mutation in Apc [8]. More recent work, however, found that DCC also was lost by promoter hypermethylation in CRC and that this characteristic combined with other adhesion-mediated signaling events to promote CRC progression and metastasis [9, 10].

In addition to DCC, the region of 18q commonly lost during CRC development contains two genes encoding regulators of signaling via transforming growth factor-β (TGF-β). These genes, known as SMAD2 and SMAD4, belong to a family of proteins similar to the gene products of the Caenorhabditis elegans gene Sma and the Drosophila gene “mothers against decapentaplegic” (Mad) [11]. TGF-β signaling has been linked to regulation of epithelial-mesenchymal interactions and CRC development. In keeping with this, germline mutations in SMAD4 are responsible for the human syndrome juvenile polyposis [12], an autosomal dominant condition characterized by multiple intestinal hamartomas. These tumors represent hyperproliferation of both epithelial and stromal tissue components, and patients with familial juvenile polyposis are at increased risk for developing CRC. The role of TGF-β signaling in CRC progression is complex and may even change from tumor suppressing to tumor promoting, depending on the tumor stage and the state of the stromal microenvironment. In response to its growth factor ligand, TGF-β receptor phosphorylates SMAD2, causing its dissociation from the receptor complex and permitting binding to SMAD4. The SMAD2-SMAD4 association is important for the translocation of SMAD4 protein into the nucleus, where it binds to target promoters and forms part of a transcriptional repressor complex. IHC can detect overexpression of SMAD4 within the nucleus of tumor cells, a finding that indicates activation of signaling via TGF-β. Small retrospective studies using archival specimens from patients with CRC showed that loss of SMAD4 was associated with poor treatment outcome [13]. These results require confirmation in larger, prospectively collected tissue sets.

Other potential tumor suppressors located at 18q are less well characterized in CRC. These include an additional mediator of TGF-β signaling, MADR2 [14]. This regulator is an important component of the signal transduction pathway downstream of TGF-βRI and TGF-βRII serine/threonine kinase receptors. Another 18q gene, ITF-2B, encodes a basic helix-loop-helix transcription factor that induces cell cycle arrest, likely by modulating expression of the cell cycle inhibitor, p21Cip1 [15]. ITF-2B signaling plays an important role in epithelial differentiation, and expression of this factor also is regulated by the CRC-associated oncoprotein β-catenin. Finally, Cables, a nuclear protein encoded by a gene on 18q11-12, regulates the activity of multiple cyclin-dependent kinases. This protein commonly is lost in CRC by promoter hypermethylation [16]. Additional tumor suppressors likely will be identified from data provided by the full sequencing of chromosome 18 [17], genome-wide association studies of CRC risk [18, 19], and new analytic methods that integrate copy number alterations, gene expression, and specific mutations in colon cancer-associated genes.

Methods to Identify 18qLOH

Researchers studying the relationship between tumor DNA losses on 18q and CRC clinical outcome have used several different methods to assign 18q status. It is not possible to adequately interpret the available data concerning 18q loss and CRC prognosis without understanding the differences among detection methods. Assays to detect 18q deletions depend on the presence of DNA polymorphisms that create heterozygosity at regions located on 18q. This heterozygosity is identified by a size-dependent difference in electrophoretic migration of DNA products. DNA from normal tissue is compared with that from tumor, and LOH is revealed when a size difference between the two alleles is present in the normal sample and absent in the tumor. Early studies detected this loss using intragenic restriction fragment length polymorphisms (RFLPs) and comparing the DNA products using radioisotopic Southern blot analysis. This assay requires large amounts of high-quality DNA and generally is not suitable for study of the more readily available formalin-fixed, paraffin-embedded (FFPE) tissues. The most common method for identifying 18q allele loss utilizes the detection of small DNA sequences known as microsatellites. Microsatellites chosen for analysis commonly exhibit length polymorphisms and can be amplified readily by PCR using DNA extracted from FFPE tissues. Allele loss is detected using a fluorescence-based PCR assay to quantitate differences between tumor and normal samples. The size and amount of the fluorescently labeled specimen produced by PCR are charted on an electropherogram (Fig. 1), and the resulting peaks that correspond to the size and amount of DNA product are quantitated in terms of peak height and peak area. The sizes of the two alleles for heterozygous cases are assigned according to the two peaks of greatest height in the normal sample, and the values for peak area of the two alleles in the paired normal and tumor samples are used to assign a figure for allele loss (Fig. 1A and B). Because this assay is based on the alteration of allele ratio in the tumor compared with the corresponding normal sample, it cannot distinguish between allele loss and amplification. Therefore, this assay is more accurately termed “allelic imbalance.” Previous molecular and cytogenic studies of CRC showed that LOH is the most common cause of allele imbalance at the microsatellites examined; therefore, 18q allelic imbalance by microsatellite analysis often is referred to as “18qLOH.”
https://static-content.springer.com/image/art%3A10.1007%2Fs11888-010-0060-3/MediaObjects/11888_2010_60_Fig1_HTML.gif
Fig. 1

Electropherograms representing the results of polymerase chain reaction assay using the D18S69 microsatellite marker to compare DNA extracted from tumor with that from normal tissue. A, Allelic imbalance at D18S69, indicated by loss of size of dominant peak, allele 1 (arrow). B, Informative case due to the presence of two distinct alleles, but no allelic imbalance is observed. C, Noninformative case due to nonheterogeneous alleles. D, Noninformative case due to microsatellite instability at D18S69 with potential for misclassification as allelic imbalance

A microsatellite region is informative for analysis if it is heterozygous in normal tissue and if microsatellite instability is not present. Homozygosity at a microsatellite region renders the site noninformative for allelic imbalance determination (Fig. 1C). The presence of microsatellite instability at a microsatellite marker also prevents accurate interpretation for allelic imbalance, because loci may falsely appear to have allelic imbalance when one of the two alleles in the tumor shows a greater degree of instability than the other. In addition, tumors may be falsely classified as lacking allelic imbalance if an unstable allele in the tumor comigrates with the imbalanced allele (Fig. 1D). Because of these issues, 18qLOH analysis is improved greatly by examining two or more microsatellites, and this analysis is considered noninformative altogether for tumors that are MSI-H.

Among the studies in the current literature, there exists a high degree of variability in the methods used to detect allelic imbalance and to translate these results into a designation for a particular case. Some researchers scored 18qLOH using a single site [20], whereas others examined up to eight different loci [21]. When several loci are examined, the results often show some loci to be intact whereas others demonstrate allelic imbalance, and research teams using multi-microsatellite assays differed in their criteria for consistency among results in assigning 18q status. For example, some required that at least two well-amplified loci demonstrate 18q allelic imbalance in order to designate the case as having 18qLOH; others considered cases noninformative unless most loci were informative and in agreement. In addition, the degree of reduction in peak area required to designate allelic imbalance also varied among studies.

The most significant challenge to interpreting the current literature involves differences in the significance assigned to MSI-H cases in the study cohorts. Several investigators, particularly in more recent studies, considered tumors that were MSI-H to be noninformative for 18qLOH analysis and therefore excluded these cases from comparisons between 18qLOH and 18q-intact cases [2228, 29•, 30, 31•, 32]. However, many reports did not specifically state how MSI-H cases were handled [20, 21, 3339]. In a few significant reports, cases with MSI-H tumors were included in the analyses together with other cases that showed no 18qLOH [3, 40, 41]. This lack of clarity in study design is problematic for two reasons. First, unless a study clearly indicates how alleles demonstrating microsatellite instability are scored, it is not possible to assess the overall accuracy of the method used to identify 18q allelic imbalance. Second, patients with MSI-H tumors, particularly those who did not receive chemotherapy, demonstrate better clinical outcome than patients whose tumors do not have this feature. As a result, these cases may confound the outcome analyses.

Data Linking Tumor 18qLOH to CRC Clinical Behavior

Many studies examined the relationship between 18qLOH and CRC clinical outcome. The available reports are almost equally divided between those finding that the presence of 18qLOH in a tumor predicted poor survival [3, 2426, 32, 35, 3741] and those showing no association between tumor 18q status and treatment outcome [2123, 27, 29•, 30, 31•, 33, 34, 36]. A single study from a randomized trial reported improved outcome for a subset of patients whose tumors demonstrated 18qLOH [28]. Table 1 provides a summary of results from the largest studies reported, together with a description of the methods used in each to detect 18q loss. Of these, 17 studies involved analysis of retrospective cohorts for which archival specimens were available. Ten of these analyses found worse outcome for patients whose tumors exhibited 18qLOH [2426, 32, 3741], although several of these positive associations were seen only within a subset of the cohort defined by stage. Four studies reported results of retrospective analyses of tumors from patients participating in either a randomized colon or rectal cancer treatment trial or a prospective cohort study. Of these, one found worse treatment outcome for patients with 18qLOH-positive tumors [3], one reported better outcome for a subset of patients with 18qLOH-positive tumors [28], and two found no association between 18q status and outcome [27, 29•]. Finally, two prospective clinical trials included analyses of the relationship between tumor 18q status and outcome as secondary end points in a colon cancer treatment study [30, 31•]. Neither of these studies found a significant association between tumor 18q status and outcome. A meta-analysis conducted in 2005 by Popat and Houlston [42] studied the prognostic value of 18q loss as measured by allelic imbalance and loss of expression of DCC protein by IHC. These investigators pooled data from 17 studies, resulting in a total of 2189 informative CRC cases. This analysis showed worse overall survival for patients whose tumors had 18q loss, and this relationship was maintained among the subset of patients who received adjuvant chemotherapy. These researchers did find evidence of heterogeneity and publication bias, limiting the validity of this meta-analysis.
Table 1

18q Loss of heterozygosity by genotyping or Southern blot analysisa

Report

Cohort description

Methods

Results

Retrospective studies using archival specimens

Laurent-Puig et al. 1992 [33]

109 surgical cases; CRC Dukes A–D

Southern blot analysis

73 informative cases

5-y OS 36.5% LOH+ vs 48% no LOH; P = NS

Jen et al. 1994 [40]

145 consecutive surgical cases; CRC stage II/III

Allelic imbalance; 5 loci; MSI-H cases included in the no-LOH category

135 informative cases

Overall stage II+III adjusted HR for LOH+ cases, 2.46 (95% CI, 1.06–5.71; P = 0.036)

Martinez-Lopez et al. 1998 [32]

144 cases CRC; stage I–III

Allelic imbalance; 2 loci; MSI-H tumors considered noninformative

121 informative cases

5-y OS 59% LOH+ vs 62% no LOH; P = 0.6

Within stage I/II only, 5-y OS 42% LOH+ vs 73% no LOH; P = 0.008

Ogunbiyi et al. 1998 [24]

151 surgical cases; CRC stage I–III

Allelic imbalance; 5 loci; MSI-H tumors considered noninformative

126 informative cases stages I–III

N = 50 stage II; 5-y disease-specific survival 66% LOH+ vs 94% no LOH; P = 0.016

N = 44 stage III; 5-y disease-specific survival 44% LOH+ vs 84% no LOH; P = 0.032

Carethers et al. 1998 [34]

70 CRC cases; stage II

Allelic imbalance; 5 loci; no exclusion of MSI-H cases

N = 70 stage II; 30 LOH+, 40 no LOH; OS multivariate HR, 1.22 (95% CI, 0.32–4.72; P = 0.78)

Lanza et al. 1998 [41]

118 surgical cases; CRC stage II/III

Allelic imbalance; 2 loci; included MSI-H cases (N = 20) as informative in the no-LOH category

112 informative cases; 60 LOH+ cases, 60 no-LOH cases (includes 20 with MSI-H)

5-y OS 60% LOH+ vs 94% no LOH; P < 0.0001; multivariate HR, 7.13 (95% CI, 2.1–23.9; P < 0.001)

Kern et al. [35] 1989

54 surgical cases; CRC Dukes A, B, and C

RFLP probes, detected by PCR amplification; MSI-H cases not excluded from analysis

48 cases informative for 18q; 35 LOH+, 13 no LOH

For Dukes B and C only, 31 LOH+, 11 no LOH

OS worse for cases with LOH+ tumors, logrank P = 0.045

Jernvall et al. 1998 [25]

125 surgical cases; CRC Dukes A–D

Allelic imbalance; 7 loci; MSI-H cases considered noninformative

N = 195 informative cases; 5-y OS 67% no LOH vs 46% LOH+; HR 3.9, 95% CI, 0.9–9.1, P = 0.04

Dukes A and B only, 5-y OS 75% no LOH vs 50% LOH+ (P = NR)

Lindforss et al. 2000 [36]

64 consecutive surgical cases; CRC stage I–IV

Allelic imbalance; 2 loci; MSI-H cases not excluded from analysis

60 informative cases, no difference in survival between 18qLOH+ and no-LOH cases for stage II and III combined

Font et al. 2001 [37]

77 cases; CRC stage I–III

Allelic imbalance; 3 loci; no exclusion of MSI-H cases

77 informative cases; OS 48% LOH+ vs 75% no LOH; P = 0.1

N = 25 stage I/II; OS 23% LOH+ vs 73% no LOH; P = 0.006

N = 28 stage III; OS not given; P for comparison LOH+ to no LOH = 0.7

Bisgaard et al. 2001 [21]

64 consecutive surgical cases; CRC Dukes A–C

Allelic imbalance; 8 loci; MSI-H cases included, but category including MSI-H cases (LOH+ vs no LOH) not specified

64 informative cases; OS worse for cases with 18qLOH in univariate (P = 0.0194) but not multivariate analysis (P = 0.0899)

Choi et al. 2002 [26]

168 cases; CRC stage I–III

Allelic imbalance; 5 loci; MSI-H cases considered noninformative

139 informative cases; 107 = LOH+, 32 = no LOH; LOH at any marker vs LOH at none of the markers showed worse survival for LOH+

LOH at a single 18q microsatellite was associated with worse survival in stage II patients (39 patients LOH+, 14 patients no LOH)

Univariate P = 0.0196

Zhang et al. 2003 [20]

59 cases; CRC Dukes A–D

Southern blot analysis; DCC intron 5 polymorphic MspI restriction site; MSI-H cases not excluded from analysis

37 informative cases, 14 with LOH at DCC, 23 with no LOH; LOH at this site was not associated with worse prognosis (P = 0.71)

Diep et al. 2003 [38]

220 consecutive surgical cases; CRC Dukes stage A–D

Allelic imbalance; 2 loci; no exclusion of MSI-H cases

184 informative cases, 129 LOH+, 55 no LOH; LOH+ cases had worse survival overall, with 5-y OS ∼62% LOH+ vs ∼78% no LOH, P = 0.008

79 informative Dukes B cases, 59 LOH+, 20 no LOH; 5-y OS ∼68% LOH+ vs ∼95% no LOH, P = 0.010

Zauber et al. 2004 [23]

150 cases; CRC Dukes B

Allelic imbalance; 2 loci; MSI-H cases considered noninformative

97 informative cases, 67 LOH+, 30 no LOH; no difference in OS for LOH+ vs no LOH cases; HR 0.91; P = 0.86

Alazzouzi et al. 2005 [39]

86 cases; CRC Dukes C

Allelic imbalance; D18S1110; D18S1156; no exclusion of MSI-H cases

61 informative cases, 26 LOH+, 35 no LOH; no difference in disease-free survival (P = 0.79) or OS (P = 0.37) for cases with and without LOH

Chang et al. 2005 [22]

207 cases; CRC, stage I–IV

Allelic imbalance; DCC locus; MSI-H cases considered noninformative

98 informative cases, 63 LOH at DCC, 35 no LOH

For stages I, II, and III, 45 were LOH+, 27 no LOH; comparison showed better 3-y disease-free survival in stage I, II, and III cases without LOH (79% vs 55%, P = 0.042)

Retrospective studies using specimens and outcome data from a randomized clinical trial or prospective cohort study

Halling et al. 1999 [27]

462 cases from 7 clinical trials of high-risk stage II and III colon cancer

Allelic imbalance; 5 loci; MSI-H tumors were considered noninformative

386 informative cases, 304 LOH, 82 no LOH; 76 additional cases were MSI-H

5-y OS 66% LOH+ vs 62% no LOH, P = 0.26

When MSI-H cases were included in the no-LOH category, 5-y OS 62% LOH+ vs 71% no LOH, P = 0.02

Watanabe et al. 2001 [3]

319 cases from 2 randomized trials of high-risk stage II and III CRC

Allelic imbalance; 5 loci; included MSI-H cases as informative in the no-LOH category

91 informative stage II cases; not prognostic; 221 informative stage III cases

5-y OS 50% LOH+ vs 69% no LOH (P = 0.005)

Excluding MSI-H cases, HR 2.75 (95% CI, 1.35–5.65; P = 0.006)

Barratt et al. 2002 [28]

393 cases from a randomized trial of surgery ± portal vein 5-FU in Dukes B and C colon cancer

Single markers examined individually; 2 loci; MSI at these loci excluded them from analysis

230 cases informative at D18S61; 144 LOH+, 86 no LOH

The subset of those treated with surgery alone had better 5-y OS if LOH+ (HR 0.50; 95% CI, 0.29–0.87; P = 0.014)

An interaction with treatment was observed, with patients with 18qLOH+ tumors deriving significantly less benefit from 5-FU

Ogino et al. 2009 [29•]

555 cases CRC stage I–IV from the Nurses’ Health Study and Health Professionals Follow-up Study

Allelic imbalance; 4 loci; MSI-H tumors were considered noninformative

532 informative cases with survival data available; OS 70% LOH+ vs 68% no LOH (P = 0.54); multivariate HR 1.18 (95% CI, 0.81–1.71)

160 stage III cases; OS univariate HR 0.91 (95% CI, 0.56–1.47); multivariate HR 0.93 (95% CI, 0.56–1.52)

Prospective biomarker studies as secondary end points in a randomized clinical trial

Popat et al. 2007 [30]

280 cases CRC stage I–III; randomized trial of postop. portal vein 5-FU; prospective secondary analysis of study subset

Allelic imbalance; 5 loci; MSI-H tumors considered noninformative

255 informative cases, stage I (N = 13), stage II (N = 149), stage III (N = 93); 124 were LOH+, 131 showed no LOH; no difference in 5-y OS between tumors with or without 18qLOH; univariate HR 1.08 (95% CI, 0.73–1.60; P = 0.7); corrected for covariates, HR 1.17 (95% CI, 0.79–1.74; P = 0.4)

Roth et al. 2009 [31]•

∼1200 stage II/III CRC cases from the PETACC 3-EORTC 40993-SAKK 60-00 trial; prospective secondary analysis of study subset

Pyrosequencing of 7 SNPs within 18q; MSI-H tumors considered noninformative

∼330 stage II cases: LOH+ had worse relapse-free survival than cases with no LOH by univariate (P = 0.03) but not multivariate analysis

∼860 stage III cases: no difference in relapse-free survival by either univariate or multivariate analysis

aStudies reporting ≥20 cases of colon cancer or CRC with 18qLOH.

5-FU 5-fluorouracil; CRC colorectal cancer; DCC deleted in colorectal cancer; HR hazard ratio; LOH loss of heterozygosity; MSI-H microsatellite instability-high; NR not reported; NS not significant; OS overall survival; PCR polymerase chain reaction; RFLP restriction fragment length polymorphism; SNPs single nucleotide polymorphisms.

Some of the largest studies deserve individual review, as they highlight the challenges in data interpretation. An early retrospective study used Southern blot analysis to determine 18q status in 73 informative Dukes A through D colorectal cancers [33]. These researchers found no significant difference in 5-year overall survival, with rates of 36% for patients with LOH-positive tumors and 48% for those whose tumors had no LOH. In contrast, Kern et al. [35], using RFLP probes in a PCR-based assay, examined 48 informative Dukes A through C tumors and showed inferior overall survival in patients whose tumors demonstrated 18qLOH (P = 0.045). Many of the available studies were conducted before the better outcome of patients with MSI-H tumors was fully appreciated, and these cases were handled in different ways among the studies. For example, a report by Jen et al. [40] examined 145 consecutive surgical cases from a single institution, limited to stage II or III colon or rectal cancer. Allelic imbalance at five microsatellites was used to identify LOH, and cases that were MSI-H were included in the analysis as having no LOH at 18q. This study found worse overall survival for patients with 18qLOH tumors, with an overall adjusted hazard ratio (HR) of 2.45 and a 95% confidence interval of 1.06 to 5.71 (P = 0.036). Individual analyses within the stage II and stage III subsets showed similar results. A second report examined 207 consecutive Dukes A through D CRCs for which survival data were available [38]. Allelic imbalance at two microsatellites was used to assign 18q status, and MSI-H cases were considered noninformative for LOH, resulting in a total of 184 cases for which tumor LOH was determined. The analysis also found that the presence of 18qLOH was associated with inferior treatment outcome, with a 5-year overall survival of approximately 62% for those with 18qLOH, compared with approximately 78% for those with no LOH (P = 0.0008). Two studies looked only at patients with stage II disease. Carethers et al. [34] studied 70 stage II colon cancer cases using allelic imbalance at five microsatellites, without excluding MSI-H cases from analysis. For overall survival, the multivariate HR was 1.22 (95% CI, 0.32–4.72; P = 0.78). Lack of correlation was also seen in 97 informative Dukes B CRCs examined in a similar manner, for which overall survival analysis showed an HR of 0.91 (P = 0.86) [23].

Each of the studies detailed above involved retrospective analyses of archival specimens. To date, six studies have reported analyses using specimens and outcome data obtained during a CRC clinical trial [3, 27, 28, 30, 31•] or a prospective cohort study [29•]. Unfortunately, these better-quality studies did not completely resolve the issue of whether the poor outcome observed with 18qLOH-positive tumors could be attributed to inclusion of the better-outcome MSI-H cases in the no-LOH comparison category. Halling et al. [27] examined 386 informative cases from seven clinical trials of high-risk stage II and III colon cancer using allelic imbalance at five microsatellite markers to detect 18qLOH. These investigators analyzed their outcome data in two ways. When cases with MSI-H tumors were included in the no-LOH category, they found that patients with 18qLOH-positive tumors had an inferior outcome, with a 5-year overall survival of 62% compared with 71% for those with no LOH (P = 0.02). They conducted a separate analysis that excluded the MSI-H tumors and found that now the outcomes were equal, with a 5-year overall survival of 66% for the 18qLOH-positive cases and 62% for those with no LOH (P = 0.26). The authors interpreted these results as indicating that inclusion of the MSI-H cases in the no-LOH category confounded the 18q result. In contrast, however, Watanabe et al. [3] examined 319 cases from two randomized trials of high-risk stage II and stage III CRC, using a similar 18qLOH determination method. For cases with stage III disease, these investigators obtained the same result for analyses with and without inclusion of MSI-H cases as informative in the no-LOH category. Of these, 221 patients had stage III disease. The stage III patients whose tumors showed no LOH had a 5-year overall survival of 69%, compared with 50% for those with 18qLOH-positive tumors (P = 0.005). When the cases with MSI-H tumors were excluded from the analysis, the association remained significant, with a 5-year overall survival of 74% in no-LOH versus 50% in 18qLOH-positive cases (P = 0.009). If MSI-H tumors respond poorly to chemotherapy, as some reports suggest, then it is possible the results of these two studies are actually in agreement because chemotherapy might have nullified the better outcome of the MSI-H cases.

The highest level of evidence available in the current literature is provided by two studies that included determination of the relationship between tumor 18q status and treatment outcome as a secondary end point in a CRC clinical trial. Unfortunately, these data have yet to resolve the controversy concerning 18qLOH as a prognostic marker. Popat et al. [30] studied patients who underwent potentially curative resection of CRC in China, followed by randomization to a trial of intraportal vein infusion of 5-FU. For a randomly assigned subset of these patients, FFPE blocks containing normal and tumor tissue were submitted and analyzed in a prospectively designed biomarker study. A total of 280 stage I through III colon and rectal cancer cases were genotyped; 25 of these were determined to be MSI-H and therefore were excluded from outcome analyses. Unfortunately, 82% of the participants also received nontrial adjuvant chemotherapy. The method to determine 18q allelic imbalance used five standard microsatellites, with homozygotes considered noninformative, and a tumor was scored as exhibiting 18qLOH if any one of the five markers demonstrated allelic imbalance. This study found no difference in outcome between patients with 18qLOH and those without it, with 5-year overall survival of 60% and 61%, respectively (P = 0.7). A more conclusive prospective study was reported in abstract form by Roth et al. [31•], representing work conducted as part of the PETACC 3-EORTC 40993-SAKK 60-00 trial. This study randomly assigned 2378 patients with stage II or III colon cancer to receive postoperative adjuvant therapy with either 5-FU plus leucovorin alone or in combination with irinotecan. The investigators analyzed 18q by pyrosequencing seven single nucleotide polymorphisms in approximately 1,400 cases, and tumors with MSI-H were excluded from the analysis. They found that 18qLOH had prognostic value within stage II (P = 0.03) but not stage III (P = 0.91) subsets. Review of the specific details of this study awaits its publication in manuscript form.

Tumor 18qLOH and Pitfalls in Biomarker Development

The story of tumor 18qLOH as a prognostic marker for CRC has much to teach us concerning the challenges of clinical biomarker development. Our inability to draw a conclusion from the very large number of available clinical studies results from several key factors. First, there has been no consistently applied method of determining loss of 18q alleles; therefore, it is almost impossible to adequately compare most of the studies in the current literature. An effort to correct this problem, sponsored by the National Cancer Institute’s Program for the Assessment of Clinical Center Tests (PACCT), recently completed a cross-laboratory validation study of methods used to detect 18q allelic imbalance in CRC specimens [43].

Efforts to correlate tumor genotype with clinical phenotype are currently hampered by our lack of understanding of the biological consequences of 18q loss. For example, we do not know whether clinical behavior is dictated by the loss of a single key tumor suppressor on 18q or by the condition of CIN itself. By far the greatest hindrance to this research has been a lack of adequate study cohorts available for CRC biomarker development and validation. Retrospective, observational study designs may produce false-positive biomarker associations because they run the risk of producing comparison populations that are not equal with regard to important factors dictating clinical response. In addition, heterogeneous responses may be masked by nonuniform study cohorts. For example, data from newer studies increasingly suggest that CRC biomarkers are stage specific and may be associated with chemotherapy, and the available studies of 18qLOH contain few stage-specific or treatment-specific results. As a result of inherent bias and disease heterogeneity, large numbers of stage-matched, uniformly treated patients are needed to adequately assess the clinical significance of a biomarker. Data from cohorts meeting these criteria are just now emerging. In particular, a trial currently underway from the Eastern Cooperative Oncology Group, E5202, addresses the utility of MSI-H and 18qLOH tumor status as a method for selecting high-risk patients with stage II colon cancer to receive adjuvant chemotherapy.

Conclusions

Unfortunately, we do not currently have the data required to define the role of tumor 18qLOH in the clinical management of CRC patients. Examining the reasons for the lack of clarity concerning this prognostic marker provides insight into the field of CRC biomarker development. Fortunately, new tools are emerging that will help define the biological significance of genes located on 18q, and new data obtained in the context of cancer clinical trials should resolve the methodologic issues plaguing this field.

Disclosure

No potential conflict of interest relevant to this article was reported.

Copyright information

© Springer Science+Business Media, LLC 2010