| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Letters to the Editor |
The Burnham Institute La Jolla, California 92037
Boland et al. (1) , in a recent issue of Cancer Research, report on a meeting entitled "National Cancer Institute Workshop on Microsatellite Instability for Cancer Detection and Familial Predisposition: Development of International Criteria for the Determination of Microsatellite Instability in Colorectal Cancer," held in December 1997 in Washington, DC. The workshop was organized in an effort to clarify the concepts, definitions, and applications of what is commonly known as MSI1 in colorectal and other malignancies, as a diagnostic test for tumor genomic instability and cancer susceptibility. The meeting report (1) is divided into three parts, corresponding to the three sessions of the workshop: the first, "Definition and Technologies," was chaired by Steven N. Thibodeau; the second, "Clinical Associations and Preclinical Models," by Stanley R. Hamilton; and the third, "MSI in Non-HNPCC, Noncolonic Tumors," by David Sidransky. I have no major disagreements with the first and second, but I have major differences of opinion with the third.
Diagnostic Criteria for MSI
The first part of the report should be useful in clarifying the confusion in the MSI field due to the lack of unified criteria for its definition. [In a review of the literature, Hiroyuki Yamamoto and I counted over 200 papers reporting "MSI" in diverse tumors. I estimate that more than half of these papers were quite off the mark (Table 1)
.]
|
Use of Dinucleotide Microsatellite Repeats as a Source of Confusion
The confusion mainly originated from the widespread use of dinucleotide repeats in the screening of tumors for MSI (see a more detailed discussion on the reasons for the origins of the confusion in the "Conclusions" section at the end of this letter).
The use of dinucleotide repeats may lead to an erroneous diagnosis of MSI if there is mislabeling of the tissue samples. The highly polymorphic nature of these repeats (in contrast to mononucleotide repeats) results in a high probability of heterozygosity in the population (this is why these microsatellites are used for genetic mapping). Comparison of the PCR profiles from two different individuals may be easily mistaken as MSI because of the differences in mobility of the amplified sequences. A survey of the MSI literature reveals that this has been a relatively common occurrence, which may account for some (but not all) of the discrepancies in MSI incidence described for many types of tumors (Table 2 in Ref. 1
; see Table 1
in this letter). PCR artifacts may also generate spurious bands, especially when formalin-fixed tissue and/or small amounts of target DNA sequences are used. These extra bands, if they occur in tumor tissue DNA, also may be misinterpreted as MSI.
Low versus High MSI
Alterations in dinucleotide repeats also may be detected in some tumors because of their high spontaneous mutation rate, the long mitotic history of the tumors, and the clonal expansions occurring during tumorigenesis. These alterations are usually only of one (or two) repeated units [i.e., CA or CACA in the commonly used (CA)n dinucleotide repeat]. This is probably what the report describes as MSI-L. However, there is no experimental evidence to support that these alterations may be caused by a functional genomic instability (see below). Moreover, tumors with MSI-L are indistinguishable (with the exception of these alterations) in genotype and phenotype from tumors without MSI. This was reported at the workshop.
Therefore, if no distinction is made between MSI-H and MSI-L, the exclusive use of dinucleotide repeats necessarily leads to an overestimation of the prevalence of tumors with enhanced (i.e., true high) MSI.
Mononucleotide Repeats Are Sufficient for the Diagnosis of True MSI
As is specified in the report (1) , "Instability at [poly(A) microsatellite] loci appears in the majority of tumors with widespread MSI (i.e., MSI-H) but rarely in tumors defined as MSI-L." Therefore, the exclusive use of mononucleotide repeats is sufficient to detect true MSI and does not engender error due to sample mismatch (but see below).
It is not apparent whether the use of dinucleotide microsatellite repeats adds to the detection of true MSI achieved by the use of mononucleotide repeats. For instance, in a recent report (2) , 14 of the 15 tumors classified as MSI-H (those exhibiting alterations at multiple loci) exhibited mutations in mononucleotide repeats. The only exceptional case, exhibiting multiple mutations in di-, tri-, and tetranucleotide repeats but none in mononucleotide repeats, could be explained by the mislabeling of samples pointed out before. This tumor was also the only case among the 15 MSI+ tumors that did not exhibit alterations in hMSH2 or hMLH1 protein expression. The other example of a MSI-H tumor, which, however, was negative for alterations in the commonly used mononucleotide repeats (BAT25, BAT26, and BAT40), is an interesting case that deserves further evaluation of the monotonic (disrupted or undisrupted) nature of these repeats.
Therefore, the need to test three dinucleotide microsatellites (in addition to two mononucleotide microsatellites) to detect and classify MSI (Table 1
in Ref. 1
) is not obvious. We use only two mononucleotides and one dinucleotide in the screening of tumors for MSI (3, 4, 5)
. This is sufficient to diagnose the presence of true (high) instability. Thus, we consider a tumor to be MSI+ when deletions of more than two nucleotides occur in any of the two mononucleotide repeats. Contractions or expansions in the dinucleotide repeat usually serve to confirm the diagnosis of the tumors as MSI+. Our criteria has the advantage of eliminating false positives. The potential drawback of increasing the number of false negatives (see above) does not seem to be a serious problem, especially because their true unstable nature is not yet substantiated by characterization of the underlying MMR (or other) defects or by the detection of mutated cancer gene targets of the MMP.
Mononucleotide Repeats Distinguish MSI-L from MSI-H
The use of the number of loci proposed in the report to distinguish MSI-L from MSI-H (Table 1
in Ref. 1
) is also unnecessary. MSI-L tumors are easily distinguishable from MSI-H with only two mononucleotide loci because MSI-L tumors do not have significant contractions in these repeats (see above). Because there is not a single genotypic (other than these sporadic microsatellite alterations) or phenotypic trait that distinguishes tumors with MSI-L from those without, the detection of MSI-L does not seem to be of any practical value, other than as a marker of clonality and therefore of neoplastic disease (Refs. 24 and 25 in Ref. 1
).
Need for Multiple Mononucleotide Repeats for the Diagnosis of MSI
The use of a single mononucleotide locus (BAT26) has already been proposed for the detection of MSI, even in the absence of matched normal tissue (6) . Although this is adequate for the identification of most positive cases, testing of more than one mononucleotide locus is necessary to avoid rare false positive results because of the existence of polymorphisms in the length of the repeat. If normal tissue is not available, the results of PCR amplification of BAT26 may be easily mistaken as somatic contractions in these repeated sequences. Although BAT26 is essentially monomorphic (7) , 2–4% of the population, especially blacks, carry significantly shorter alleles.2 PCR amplification of DNA from these individuals may be confused with the contractions characteristic of the MMP. The use of two mononucleotide repeats significantly diminishes the probability of simultaneous occurrence of these rare alleles.
In conclusion, the issues of the size of the microsatellite loci panel to be used for detection of MSI and related methodological aspects seem to be more complicated than necessary. Nevertheless, the use of three dinucleotide repeats, in addition to the two critical mononucleotide repeats recommended by the report (1) , will not hurt, as long as care is taken to avoid sample mismatch.
Clinical Associations and Preclinical Models
The second part of the report (1) is not discussed here because of its heterogeneous nature; it essentially lists individual reports at the workshop by different groups.
MSI in Non-HNPCC, Noncolonic Tumors
In contrast to the first part of the report (1) , the third part is largely confusing and misleading. The main problem is the incorrect extrapolation, from the detection of a few microsatellite alterations in some tumors to the existence of a genomic instability (MSI) playing a role in tumorigenesis. This axiom is flawed because, as the report states, "the classification of these tumors is difficult because the absolute background frequency of a given microsatellite alteration in somatic tissue is largely unknown."
MSI
The term MSI is a misnomer because these sequences are intrinsically unstable. What is relevant is not the existence of microsatellite alterations but their number. Tumors of the MMP, both hereditary (i.e., HNPCC) and "sporadic," accumulate hundreds of thousands of somatic clonal mutations in mononucleotide repeats (8) . The accumulation of this number of mutations represents true genomic instability, which is germane to tumorigenesis (i.e., due to underlying deficiencies in DNA MMR). As discussed before, detection of alterations (contractions) in a single mononucleotide repeat can be extrapolated with a high degree of certainty to the existence of genome-wide instability.
On the other hand, the detection of mutations in di-, tri-, or tetranucleotide repeats does not necessarily imply the existence of genomic instability. These alterations can be also due to spontaneous errors of replication of these highly unstable sequences. Whether there is a different kind of genomic instability generating mutations at these microsatellite sequences (but not at mononucleotide repeats) is an interesting possibility that, however, does not have experimental support (see above).
Sporadic (Non-HNPCC) versus Hereditary (HNPCC) Tumors with MSI
This part of the report (1) also magnifies another source of confusion by the artificial distinction of hereditary from sporadic tumors with MSI, as if they exhibited fundamental differences in genotype or phenotype. This is a widespread view in the field (9 , 10) . Thus, the report states that sporadic tumors with MSI "have not been commonly found to have mutations in the known MMR genes." But the inactivation of known MMR genes in sporadic tumors (9 , 10) is as common (or uncommon) as it is in HNPCC tumors (10 , 11) . Tumors with true MSI, with or without family history of cancer, are indistinguishable in genotype and phenotype, with the exception of the existence of a germ-line mutator mutation in the former but not the latter (5 , 9 , 10) . On the other hand, tumors with true MSI exhibit many peculiar features in genotype and phenotype that distinguish them from tumors without MSI (5 , 8) .
It seems unwarranted to distinguish one tumor from another in any significant manner, based upon whether the functional inactivation of MMR genes is achieved by nonsense, missense, or frameshift mutations, by mutations in regulatory regions, deletions, or epigenetic alterations (e.g., hypermethylation). What is relevant is whether the tumors have MSI or not, not whether the MSI is due to one or another form of inactivation of MMR genes. The evidence can be found in every one of the relevant papers cited in the report (1) . Whether inactivation of MMR genes by hypermethylation may be more common than inactivation by mutation in sporadic versus hereditary tumors is not yet clear. In any case, it would not affect the above conclusions. The existence of tumors with true MSI, in the absence of genetic (or epigenetic) inactivation of any of the characterized MMR genes, is not restricted to the sporadic category; on the contrary, it is also common in the hereditary form (HNPCC).
Therefore, the classification of tumors as "non-HNPCC" is ambiguous, at the very least. For instance, it is well known that gastric and endometrial cancers are also characteristic of the Lynch (HNPCC) syndrome. However, the report addresses these tumors as "noncolonic, non-HNPCC" (Table 2 in Ref. 1 ). This classification is also contradictory with the recommendations given in the report to screen sporadic colorectal tumors for MSI to detect HNPCC cases: "The workshop members are of the opinion that MSI testing has one clinical purpose: to identify patients with HNPCC." But beyond its use in clarifying who, among the colon cancer patients, are actually HNPCC patients, there is another application of MSI analysis in the clinical arena: to identify patients with tumors of the MMP for studies designed to ultimately determine how to best treat these patients because surely they have a different disease than those with MMP tumors. Thus, accurate classification is paramount.
Mononucleotide Repeat-independent MSI
The tangle of this part of the report is further augmented in the following section: "A second group of noncolonic non-HNPCC tumors displays elevated frequencies of instability only at highly selected tri- and tetranucleotide repeats.... In certain tumors, larger repeats are more commonly altered than smaller repeats; this finding stands in stark contrast to what is found in HNPCC and most sporadic gastric and endometrial tumors ...." As explained above, these findings may be explained by assuming that: (a) HNPCC and other sporadic tumors (from colon or any other site) with true MSI (MSI-H) accumulate mutations in all these microsatellite sequences because they exhibit a profound genomic instability due to deficiencies in MMR (or other factors involved in maintaining the fidelity of replication of these repetitive sequences); and (b) that tumors from this "second group of noncolonic non-HNPCC tumors" do not have any genomic instability related to microsatellite replication fidelity. These microsatellite alterations are sporadically observed in these tumors due to clonal expansion of cells with these spontaneous mutations, which are more frequent in larger (i.e., repeats of larger repeated unit) and longer (i.e., repeats with more repeated units) repeats. This is in line with studies in yeast showing that, in MMR mutants, mononucleotide repeats are considerably less stable than dinucleotide repeats and that the spontaneous mutation rates at repeat loci goes up as the length of the repeating unit also goes up (12) .
Therefore, the argument that "larger repeats are more commonly altered than smaller repeats," used to support the concept that the tumors with this elevated microsatellite alterations at selected tetranucleotides phenotype involves a non-MMR pathway, is not convincing. Differences in mutation frequency in selected microsatellites loci may be due to their intrinsic differences in spontaneous mutation rates (see above). Topographical sequence constraints may also facilitate (or hamper) the slippage events responsible for their contractions or expansions (see also above). Whether the same microsatellite locus has a different mutation frequency in different tumors ("e.g., bladder cancer versuslung cancer") is not clear either. These differences may be due to differences in mitotic histories because the number of cell divisions undergone by the tumor cell (or its precursor) may be strongly dependent on the tissue of origin.
In conclusion, this part of the report fails to make a coherent argument in support of the contention that there are tumors with a microsatellite-related genomic instability, which differs in any significant manner from that possessed by MMP tumors.
MSI and Environmental Factors
The paragraph in the report (1) discussing the putative role of environmental factors in the generation of MSI (i.e., smoking and diet) is nothing more than a shaky speculation without any supporting experimental evidence. In this situation, the statement can only possibly enhance the already existing confusion.
Target Genes for the MMP
Finally, the paragraph following that discussed above and Table 3 in Ref. 1 are even more loaded with inaccuracies and errors. On the one hand, well-established facts are only considered as speculations, and on the other, findings without proven significance are reported as established truths. Thus, frameshift mutations are reported as "... presumably leading to inactivation of the affected allele." The functional consequences of a frameshift mutation are some of the few phenomena in biology that are clear-cut. This has been well known for over 30 years.
This part of the report (1) addresses the issue of "whether an affected gene is a true target of inactivation," by providing five different (apparently) sine qua non criteria: (a) frequency; (b) biallelic inactivation; (c) suppressor pathway; (d) inactivation in nonreplicative error; and (e) functional data.
High Frequency of Inactivation.
Table 3 in Ref. 1
lists transforming growth factor-ß type II receptor, IGF-IIR, BAX, MSH3/MSH6, PTEN, APC, and ß2-microglobulin as "potential target genes" in tumors with MSI from high or moderate to low frequency of mutations. Although transforming growth factor-ß type II receptor and BAX, as well as PTEN, are all in the "high" category, IGF-IIR is grouped with APC and MSH3/MSH6 as "moderate." Only the ß2-microglobulin gene is "demoted" to the "low" category. This classification is incorrect in numerous instances. IGF-IIR mutations are clearly less frequent in colon cancer with true MSI (3 of 35, 1 of 18, and 8 of 41, giving a total of 12 of 94 or 12%; Refs. 96, 98, and 107 in Ref. 1
, respectively), than mutations in BAX and hMSH3/hMSH6 (
50% for BAX and hMSH3 and 30% for hMSH6, both in HNPCC and sporadic). In gastric cancer of the MMP, the situation is also similar. IGF-IIR mutations are also less frequent than in the ß2-microglobulin gene (30–40%; Ref. 115 in Ref. 1
). Therefore, IGF-IIR mutations (12%) are listed as moderate, whereas ß2-microglobulin gene mutations (30–40%) are listed as low. Similarly, the PTEN gene is also listed as high, whereas no (or very few) mutations have been described in colon and stomach tumors of the MMP.
Moreover, the inclusion of APC as a target for the MMP is dubious for several reasons. First, in Ref. 113 in Ref. 1 , The classification or replicative error-positive tumors may have been done by artificially homogenizing MSI-H and MSI-L tumors, which, as we have seen already, is incorrect and leads to confusion. Second, there are reports (13) in direct contradiction with the paper cited (Ref. 113 in Ref. 1 ) in Table 3 in Ref. 1 . Our own unpublished results indicate that APC deletion-insertion mutations are significantly less frequent in MMP colon tumors than in tumors without MSI2 .
The absence of p53 and ras in Table 3 in Ref. 1 is quite illuminating. Why aren’t these genes included, if, according to the prevailing view (10) , they are also mutated in MSI tumors as a result of their mutator phenotype? The answer is simple: these genes are not usually mutated in MMP tumors. This is one of the reasons why we concluded that these tumors develop through a different molecular pathway, the MMP pathway (8 , 14) .
Biallelic Inactivation by Simultaneous Alteration of the Other Allele’s Repeat Tract, Point Mutation, or Loss.
If a gene is inactivated by biallelic mutations, there is no doubt that no residual gene product remains. The concept is important and originated in Knudson’s "two-hit" model for retinoblastoma, which has been of great influence in the field for so many years. However, this criterion is not absolutely required for a bona fide nature of a suppressor gene. For instance, imprinting of one allele eliminates the need for biallelic mutation. Nevertheless, the criterion is useful (although not disqualifying). The problem is that Table 3 in Ref. 1 is mistaken in the two examples given for the genes not complying with the criterion, BAX and ß2-microglobulin. There is ample evidence in the very same papers listed in Table 3 in Ref. 1 (Refs. 70, 115, and 107, respectively, in Ref. 1 ; see also Ref. 15 in this paper) that biallelic mutations occur in these two genes in tumor cells of the MMP.
Involvement of the Candidate MSI Target Gene in a Bona Fide Growth Suppressor Pathway.
The criterion is doubly flawed. First, the MMP represents a new pathway for tumorigenesis, different from the suppressor pathway (14 , 16) . Therefore, the search for a specific target gene for the MMP (MSI) may reveal novel genes involved in tumorigenesis (this is why the mutator and the suppressor pathways are different because the target cancer genes are usually different). In other words, the search for target genes of the MMP may reveal genes not previously known to be tumor suppressors. Second, cancer genes under negative selective pressure (i.e., tumor suppressors) are not restricted to those regulating cell growth but also include genes playing roles in cell senescence, cell differentiation, cell survival (apoptosis), and the escape of immune surveillance. Table 3 in Ref. 1 is, therefore, bidirectionally incorrect. The reason why BAX is cited as "unclear" is, in itself, rather unclear. That the mechanism by which BAX contributes to tumorigenesis and to apoptosis is not yet well understood is not surprising because the gene was discovered recently. Besides, the precise oncogenic mechanism is not yet understood for many other oncogenes or tumor suppressor genes, including every one listed in Table 3 in Ref. 1 . In conclusion, although hMSH3, hMSH6, BAX, and ß2-microglobulin are not directly involved in cell growth, this should not disqualify them from being true targets of inactivation by the MMP.
Inactivation of the Same Growth Suppression Pathway in MSI-negative Tumors through Inactivation of the Same Gene or of Another Gene within the Same Pathway.
The criterion is invalidated by the previous argument. Further, it is intellectually inconsistent. If hMSH3 and hMSH6 are mutator genes, it is impossible that they could be involved in any other tumors than those of the mutator phenotype. Moreover, the other "no" answers in Table 3 in Ref. 1 (again, BAX and ß2-microglobulin, what a coincidence!) should be "yes," as reported in the very same papers (Refs. 70 and 115, respectively, in Ref. 1 ).
Functional Suppressor Studies in Vitro or in Vivo Models, such as Cell Lines or Animals.
Here again, BAX, hMSH3/hMSH6, and ß2-microglobulin are the "bad guys" because, according to Table 3 in Ref. 1 , there are no functional data to support their role as suppressor genes. However, functional evidence for the role of hMSH3 in restoring replication fidelity of microsatellite sequences has been documented in MMP cell lines (17) , and the functional consequences in cancer development of BAX inactivation (both in homozygous and heterozygous states) are well documented in knockout mice (18) . On the other hand, supportive evidence for a tumor suppressor function of PTEN or IGF-IIR in MMP tumors is nowhere to be found in any of the references provided. Moreover, the requirement for an involvement in a bona fide growth suppressor pathway is defeated by the distinctive feature of the MMP: although an inactivating mutation in a tumor suppressor gene may lead to altered growth, an inactivating mutation in a mutator gene does not (14 , 16) . Consequently, the reintroduction into a tumor cell of a wild-type mutator gene will not revert its transformed phenotype. It may complement the mutator defect but not the inactivated downstream target genes.
In summary, the confusion of this final part of the report is so egregious that it would surely trigger an amused response in the informed reader, were it not for the negative consequences that it would have if it were left unchallenged. The reader needs to consider the unnecessary experiments that have been performed due to the confusion in the field, which necessitated the organization of two workshops by the NCI to clarify it (see Table 1
). Although the first part of the report helps to clarify the confusion in the colon cancer MSI field, this part of the report will undoubtedly increase the chaos in the MSI field of "noncolonic, non-HNPCC" tumors and on the issue of target genes for the MSI.
Conclusions
In retrospect, it may be useful to reflect on the unpredictable nature of the events that sometimes influence the unfolding of scientific discoveries and their impact on scientific progress, for it is quite possible that the confusion created in the MSI field (see above) may have been avoided if the order of publication of the relevant papers would have just simply followed the order of their dates of submission. However, the publication dates were inversely related to their order of submission. Thus, although our paper (8) was submitted to Nature in November 1992 and resubmitted in February 1993 (19) , Thibodeau et al. (20) submitted their paper to Science in March 1993 and Aaltonen et al. (21) submitted theirs in April 1993. These papers were both published in May 1993 [Aaltonen et al. (21) , followed by Thibodeau et al. (20) ], whereas ours finally printed in June 1993.
It is tempting to speculate that, if our paper had been published first, perhaps the confusion in the field would not have been generated because we only mentioned in passing that tumors harboring ubiquitous contractions in mononucleotide repeats also contained mutations in di- and trinucleotide repeats. Thus, the following papers devoted to detecting MSI in tumors would most likely have used the same poly(A) repeats described in our paper and not the dinucleotide repeats reported in the Science papers. In this case, only tumors with true MSI would have been detected (see above).
In addition, if our paper had been published first, perhaps the generalized unawareness of the existence of definite genotypic differences between tumors with or without the MMP (see above) also would not have occurred. The significantly lower incidence of ras and p53 mutations (8) in tumors with true MSI versus tumors without MSI (including the MSI-L, dinucleotide-dependent MSI) would, thus, not have been obscured by the artificial homogenization of tumors with MSI-H and MSI-L (21) . But the speculation does not need to stop here. Other misconceptions in the field may have also been circumvented, such as the overestimation of the frequency of HNPCC tumors with MSI and the widespread belief in the existence of imaginary differences in genotype and phenotype between hereditary and sporadic tumors of the MMP (see above).
This somewhat depressing retrospective reflection is not lacking in irony, if we consider that the misleading descriptions in the first (21) and follow-up publications (9 , 10) were not unavoidable, but perhaps the result of an attempt to artificially highlight these nonexistent differences on the one hand and to unnaturally amalgamate the real differences on the other. As we have already discussed, it is difficult to comprehend why sporadic and hereditary tumors of the MMP could have functionally significant differences in their genotype if they have the same phenotype (hundreds of thousands of clonal somatic mutations in microsatellite repeats). On the other hand, it is easy to understand why differences in genotype (the presence or absence of the MMP) in a tumor cell would be accompanied by differences in its phenotype. However, these publications (9 , 10 , 21) led to the generalized doubly erroneous belief that: (a) tumors with or without the MMP are not significantly different in their pathways for cancer, and only the order and the speed of accumulation of mutations in cancer genes (APC, ras, and p53) differ; and (b) there are significant differences in genotype and phenotype between tumors of the MMP depending upon whether they are hereditary (HNPCC) or sporadic.
The reasons for these conclusions, which are unreasonable at first sight, may not be so difficult to understand, if we consider the details of our reiterated unsuccessful efforts to get our discovery of MSI published, for the submission of our paper to Nature in November 1992 was already the third attempt at publication (19) . The second attempt was in September 1992 to a highly prestigious journal. The outcome of our submission may have been already guessed by the reader.
This was a deplorable outcome, not only because it delayed scientific progress (see the evidence, for instance, in Ref. 22 ), but also because it set the stage for the future confusion in the MSI field. It was based on the editor’s misguided opinion that the findings lacked sufficient significance to merit publication. According to him, the existence in some tumors of these ubiquitous (more than 105) clonal somatic mutations in mononucleotide repeats and the existence of a severe genomic instability that was germane to tumorigenesis (8) were not connected by a "causal relationship." In defense of this editor’s self-detrimental decision, it can be said that he was definitively influenced by the unmistakable negative tone of the comments of one of the referees. The anonymous referee’s comments exhibited a similarity to the criticisms that our paper had already encountered in its first publication attempt.
These criticisms were made by a scientist to whom we had submitted the manuscript for publication in Proceedings of the National Academy of Sciences (USA) in the summer of 1992. The parable of this narrative will soon be unveiled to the patient reader, when he is told that this scientist was a coauthor of the Aaltonen et al. 1993 Science paper (21) . The reader may realize at this point that the Aaltonen et al. Science paper (21) was written with the knowledge of our paper’s content.
FOOTNOTES
1 The abbreviations used are: MSI, microsatellite instability; HNPCC, hereditary nonpolyposis colorectal cancer; MSI-L, low-frequency MSI; MSI-H, high-frequency MSI; MMR, mismatch repair; MMP, microsatellite mutator phenotype; IGF-IIR, insulin-like growth factor type II receptor. 
2 S. Schwartz, H. Yamamoto, and M. Perucho, unpublished observations.
3 To whom requests for reprints should be addressed, at The Burnham Institute, 10901 North Torrey Pines Road, La Jolla, CA 92037. Phone: (619) 646-3112; Fax: (619) 646-3190; E-mail: mperucho{at}ljcrf.edu
1 The abbreviations used are: MSI, microsatellite instability; MSI-H, high-frequency MSI; IGF-IIR, insulin-like growth factor type II receptor.
2 To whom requests for reprints should be addressed, at E-mail: Crboland{at}ucsd.edu
Received 6/16/98. Accepted 12/10/98.
REFERENCES
University of California La Jolla, California 92093-0688
Mayo Clinic Rochester, Minnesota 55905
The University of Texas M. D. Anderson Cancer Center Houston, Texas 77030-4 095
Johns Hopkins University Baltimore, Maryland 21205-2196
University of Utah Health Sciences Center Salt Lake City, Utah 84132
University of Maryland Hospital Baltimore, Maryland 21201
Roswell Park Cancer Institute Buffalo, New York 14263
Leiden University Leiden AL 2333, the Netherlands
University of Pavia 27100 Pavia, Italy
National Cancer Institute Bethesda, Maryland 20892-7346
In this issue of Cancer Research, Dr. Perucho (1) has taken an opportunity to comment on our recently published workshop report (2) . Some of the comments represent differences of opinion with what we wrote in the meeting report; others deal with issues not raised in the report, which we shall not address here.
In December 1997, over 100 investigators from throughout the world assembled to reach a consensus on what should be called MSI1 . Because this has become an intensively studied area, it was felt that uniformity of terminology and techniques should be developed to facilitate future conversations. Dr. Perucho was among those invited, although he chose not to participate. Those who attended the workshop and contributed to the writing of the manuscript sacrificed their individual opinions to the consensus of opinions developed through group process. Many hours were spent discussing the issues mentioned in the text of the report. The first draft was written at the workshop and read back to the participants for additional modification to develop the broadest possible consensus. Everyone left with a better understanding of the state of the art but with the full knowledge that this dynamic field will continue to evolve in the future. Many participants left without having each of their personal opinions and speculations specifically expressed. Such is the nature of a consensus conference.
One of the principal aims of the workshop was to develop uniform nomenclature for MSI. This was successfully reached, and the clarity of the field is not benefited by use of the terms microsatellite mutator phenotype for MSI or "true MSI" for MSI-H. Terminology is an issue of personal preference, but when a large group inclusive of most experts and interested parties in the field reaches consensus on nomenclature, it would be reasonable to accept this.
A considerable amount of time and effort was expended on a debate that resulted in the selection of the five microsatellite markers in the reference panel and the criteria for interpretation. Notwithstanding Dr. Perucho’s contention that three markers are sufficient (1) , it was perfectly clear to the workshop attendees that the use of fewer than five markers would carry a risk of error. As indicated in the two multicenter studies (3 , 4) , some proportion of tumors will, by chance, have a mutation detected in a single microsatellite locus but might still have the MSI-H phenotype. That was, after all, the point of the first third of the meeting report. The use of fewer markers was an opinion that was considered and rejected by the group. It is the case that a single mononucleotide repeat marker will find most but not all MSI-H tumors (5) . By using just one marker, one will miss some MSI-H tumors and falsely identify some tumors with the low-frequency MSI phenotype as MSI-H. The reader is referred to the meeting report (2) for the supporting rationale and the importance of this distinction.
Dr. Perucho and colleagues had taken care to examine over 200 papers that deal with MSI (Table 1 in Ref. 1 ). As he indicates, there is wide variation in the estimates of MSI. This was, after all, why the workshop was convened. It was clearly the opinion of the workshop participants that the principal problems were both an inappropriate failure to distinguish MSI-L from MSI-H tumors and technical problems in the performance and interpretation of the assays for MSI. The workshop participants determined that a panel of five microsatellite markers, verified by several laboratories, was necessary to address this problem. It is difficult to accept the suggestion that some or most of the confusion in the field has been due to the mislabeling of samples, as suggested in Dr. Perucho’s letter (1) . The selection of these five markers was not too complicated; rather, it was only as complicated as it needed to be to deal with the problem. Moreover, the meeting report (2) makes no attempt to stifle creativity. It was stated that this panel was selected for those investigators who wanted a consensus panel for their research, but it was also pointed out that other panels could be used with equal authority if they were validated as were those at the workshop.
There is no debate regarding whether the phenotypes of MSI tumors are sporadic or related to hereditary nonpolyposis colorectal cancer. The difference between these is that hereditary nonpolyposis colorectal cancer tumors are found in association with germ-line mutations in DNA mismatch repair genes, and sporadic MSI tumors are not. Hypermethylation of the hMLH1 promoter accounts for most, but perhaps not all, of the sporadic MSI tumors (6 7 8 ). MSI tumors are otherwise similar. We are not certain where the alleged confusion arose.
With respect to specific genes potentially inactivated by MSI, the meeting report recommended guidelines to help sort out the likelihood that these genes might be true targets of inactivation rather than innocent bystanders (2) . In one of his original papers on this subject, Perucho et al. (9) caution that clonal mutations in tumors with MSI cannot be easily interpreted. Because he and others now postulate that some of these genes may be real targets, we chose five important characteristics to help strengthen the case for the involvement of a specific gene.
We feel that such criteria are not irrelevant because empirical data indicate instances of inactivation in tumors with and without MSI. The table was a first attempt to identify the loci that are true targets in this pathway. It is not a simple task to prove which mutations occur because genes are functionally involved in tumorigenesis and which happen as bystander or passenger events. The distinction is of considerable importance, and the evidence for functional involvement of some genes is stronger than for others. For example, at least two independent studies found that PTEN is mutated significantly more frequently (85% versus 35%) in MSI-positive than in MSI-negative endometrial tumors (10 , 11) , and there is abundant evidence to support the function of PTEN as a tumor suppressor gene in the germ line of cancer families (12 13 14) , knockout animal models (15 , 16) , and studies of gene function per se (17 , 18) . Similarly for IGF-IIR, abundant studies show mutation and loss of heterozygosity in a variety of human tumors (19 20 21) .
Thus, there is no less supportive evidence for a tumor suppressor function of PTEN or IGF-IIR in MSI tumors than for hMSH3, ß2-microglobulin, or BAX. The point of this criterion was that frequent mutations (in MSI-negative or MSI-positive tumors) of genes possessing known tumor suppressor functions are less likely to occur as bystander events than are mutations of genes without such known functions. In this context, it may be somewhat artificial to classify PTEN and IGF-IIR as more targeted than hMSH3, hMSH6, ß2-microglobulin, or BAX. It should be stated that, at the workshop, we benefited from the contribution of many investigators in the field who shared published and unpublished data (including large amounts of negative data) to help generate the table on mutated genes in tumors with MSI.
Finally, our response to Dr. Perucho’s concerns about priority is simply to quote from the original article with which he takes issue (22) : "Perucho and colleagues have previously observed consistent alterations in simple repeated sequences in a subset of sporadic colorectal cancer. Our results suggest that the tendency to form such alterations can be inherited and may be directly related to a defective gene on chromosome 2." There can thus be no question that Dr. Perucho’s priority and contributions have been appropriately and explicitly acknowledged by his colleagues in this and several other publications (23) . Dr. Perucho has made seminal discoveries in this field and deserves full credit for them. As noted in our workshop report, other investigators have also made substantial and independent observations that have together led to the rapid advances in this important area of research.
Received 12/ 2/98. Accepted 12/10/98.
REFERENCES
Nature London W8 4NX, United Kingdom
I can confirm Dr. Perucho’s account (1) of the treatment of his paper in Nature (2) . When we received it, we knew that it had already been refused publication by one other journal (the anonymous "prestigious" publication he now mentions), in part because of the recommendation of a named referee. Publication in Nature was delayed by our proper request for the removal of a general introduction to the manuscript before it was sent to reviewers. The appearance of the competing manuscript during this interval suggested that the named referee cannot but have been conscious of a sharp conflict of interest, to put the best light on the matter. Some months later, I drew Nature’s readers’ attention to the case by means of an article entitled "Competition and the Death of Science" (3) . It is beyond belief that putative reviewers do not disqualify themselves in circumstances like these—and that editors do not remind reviewers of the seemliness of doing so.
Received 12/ 7/98. Accepted 12/10/98.
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  K. Imai and H. Yamamoto Carcinogenesis and microsatellite instability: the interrelationship between genetics and epigenetics Carcinogenesis, April 1, 2008; 29(4): 673 - 680. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. F. Wong, T. H. Cheung, K. W. K. Lo, S. F. Yim, L. K. Y. Chan, O. Buhard, A. Duval, T. K. H. Chung, and R. Hamelin Detection of microsatellite instability in endometrial cancer: advantages of a panel of five mononucleotide repeats over the National Cancer Institute panel of markers Carcinogenesis, May 1, 2006; 27(5): 951 - 955. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Bacani, R. Zwingerman, N. Di Nicola, S. Spencer, T. Wegrynowski, K. Mitchell, K. Hay, M. Redston, E. Holowaty, D. Huntsman, et al. Tumor Microsatellite Instability in Early Onset Gastric Cancer J. Mol. Diagn., October 1, 2005; 7(4): 465 - 477. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Findeisen, M. Kloor, S. Merx, C. Sutter, S. M. Woerner, N. Dostmann, A. Benner, B. Dondog, M. Pawlita, W. Dippold, et al. T25 Repeat in the 3' Untranslated Region of the CASP2 Gene: A Sensitive and Specific Marker for Microsatellite Instability in Colorectal Cancer Cancer Res., September 15, 2005; 65(18): 8072 - 8078. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. B. Hatch, H. M. Lightfoot Jr., C. P. Garwacki, D. T. Moore, B. F. Calvo, J. T. Woosley, J. Sciarrotta, W. K. Funkhouser, and R. A. Farber Microsatellite Instability Testing in Colorectal Carcinoma: Choice of Markers Affects Sensitivity of Detection of Mismatch Repair-Deficient Tumors Clin. Cancer Res., March 15, 2005; 11(6): 2180 - 2187. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Duval, M. Raphael, C. Brennetot, H. Poirel, O. Buhard, A. Aubry, A. Martin, A. Krimi, V. Leblond, J. Gabarre, et al. The mutator pathway is a feature of immunodeficiency-related lymphomas PNAS, April 6, 2004; 101(14): 5002 - 5007. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Plaschke, S. Kruger, B. Jeske, F. Theissig, F. R. Kreuz, S. Pistorius, H. D. Saeger, I. Iaccarino, G. Marra, and H. K. Schackert Loss of MSH3 Protein Expression Is Frequent in MLH1-Deficient Colorectal Cancer and Is Associated with Disease Progression1 Cancer Res., February 1, 2004; 64(3): 864 - 870. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G.M. Nash, M. Gimbel, J. Shia, A.T. Culliford, D.R. Nathanson, M. Ndubuisi, Y. Yamaguchi, Z.S. Zeng, F. Barany, and P.B. Paty Automated, Multiplex Assay for High-Frequency Microsatellite Instability in Colorectal Cancer J. Clin. Oncol., August 15, 2003; 21(16): 3105 - 3112. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Iacopetta Aberrant DNA Methylation: Have We Entered the Era of More Than One Type of Colorectal Cancer? Am. J. Pathol., April 1, 2003; 162(4): 1043 - 1045. [Full Text] [PDF]
|
|
|
|
|
|
K. Suzuki, T. Dai, I. Suzuki, Y. Dai, K. Yamashita, and M. Perucho Low Mutation Incidence in Polymorphic Noncoding Short Mononucleotide Repeats in Gastrointestinal Cancer of the Microsatellite Mutator Phenotype Pathway Cancer Res., April 1, 2002; 62(7): 1961 - 1965. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Liu, J. Chen, S. Salahshor, S. Kuismanen, E. Holmberg, H. Gronberg, P. Peltomaki, and A. Lindblom Screening families with endometrial and colorectal cancers for germline mutations J. Med. Genet., September 1, 2001; 38(9): e29 - 29. [Full Text] [PDF]
|
|
|
|
 |
|
 T. Watanabe, T.-T. Wu, P. J. Catalano, T. Ueki, R. Satriano, D. G. Haller, A. B. Benson, and S. R. Hamilton Molecular Predictors of Survival after Adjuvant Chemotherapy for Colon Cancer N. Engl. J. Med., April 19, 2001; 344(16): 1196 - 1206. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Duval, S. Rolland, A. Compoint, E. Tubacher, B. Iacopetta, G. Thomas, and R. Hamelin Evolution of instability at coding and non-coding repeat sequences in human MSI-H colorectal cancers Hum. Mol. Genet., March 1, 2001; 10(5): 513 - 518. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. L. Bacon, S. M. Farrington, and M. G. Dunlop Sequence interruptions confer differential stability at microsatellite alleles in mismatch repair-deficient cells Hum. Mol. Genet., November 1, 2000; 9(18): 2707 - 2713. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. A. Eckert and G. Yan Mutational analyses of dinucleotide and tetranucleotide microsatellites in Escherichia coli: influence of sequence on expansion mutagenesis Nucleic Acids Res., July 15, 2000; 28(14): 2831 - 2838. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. E. Wilentz, M. Goggins, M. Redston, V. A. Marcus, N. V. Adsay, T. A. Sohn, S. S. Kadkol, C. J. Yeo, M. Choti, M. Zahurak, et al. Genetic, Immunohistochemical, and Clinical Features of Medullary Carcinoma of the Pancreas : A Newly Described and Characterized Entity Am. J. Pathol., May 1, 2000; 156(5): 1641 - 1651. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. E. Hile, G. Yan, and K. A. Eckert Somatic Mutation Rates and Specificities at TC/AG and GT/CA Microsatellite Sequences in Nontumorigenic Human Lymphoblastoid Cells Cancer Res., March 1, 2000; 60(6): 1698 - 1703. [Abstract] [Full Text]
|
|
|
|
|
|
R. Pyatt, R. B. Chadwick, C. K. Johnson, C. Adebamowo, A. de la Chapelle, and T. W. Prior Polymorphic Variation at the BAT-25 and BAT-26 Loci in Individuals of African Origin : Implications for Microsatellite Instability Testing Am. J. Pathol., August 1, 1999; 155(2): 349 - 353. [Abstract] [Full Text] [PDF]
|
|
