当前位置: 首页 > 期刊 > 《临床肿瘤学》 > 2005年第10期 > 正文
编号:11332201
Loss of Heterozygosity for Chromosomes 1p and 16q Is an Adverse Prognostic Factor in Favorable-Histology Wilms Tumor: A Report From the Nati
http://www.100md.com 《临床肿瘤学》
     the Department of Pediatrics, Roswell Park Cancer Institute

    School of Medicine and Biomedical Sciences, University at Buffalo, State University of New York, Buffalo, New York

    Departments of Pediatrics and Oncology, Cross Cancer Institute and the University of Alberta, Edmonton

    Department of Pediatrics, University of Calgary, Calgary, Alberta, Canada

    Department of Biostatistics, University of Washington

    Fred Hutchinson Cancer Research Center, Seattle, Washington

    Department of Pathology, Children's Memorial Hospital, Chicago, Illinois

    Department of Pathology, Loma Linda University, Loma Linda

    Department of Pediatrics, Los Angeles Children's Hospital

    Department of Pediatrics, School of Medicine, University of Southern California, Los Angeles, California

    Department of Pediatric Surgery, The University of Texas at Houston Health Science Center, Houston

    Department of Pediatrics, The University of Texas Southwestern Medical Center, Dallas

    Department of Pediatrics, The University of Texas M.D. Anderson Cancer Center, Houston, Texas

    Department of Surgery, Children's Hospital, Boston, Massachusetts and the Department of Surgery, Harvard Medical School, Boston, Massachusetts

    Department of Pediatric Surgery, Denver Children's Hospital, Denver, Colorado

    Department of Radiation Oncology, Hospital of the University of Pennsylvania, Philadelphia, Pennsylvania

    Department of Pediatrics, Cooper Hospital, Camden, New Jersey

    Ochsner Clinic Foundation

    Tulane University School of Medicine, New Orleans, Louisiana

    Department of Radiation Medicine, St Joseph's Hospital, Tampa, Florida

    the Department of Radiation Oncology, Cleveland Clinic Foundation, Cleveland, Ohio

    ABSTRACT

    PURPOSE: To determine if tumor-specific loss of heterozygosity (LOH) for chromosomes 1p or 16q is associated with a poorer prognosis for children with favorable-histology (FH) Wilms tumor entered on the fifth National Wilms Tumor Study (NWTS-5).

    PATIENTS AND METHODS: Between August 1995 and June 2002, 2,021 previously untreated children with FH or anaplastic Wilms tumor, clear-cell sarcoma of the kidney (CCSK) or malignant rhabdoid tumor of the kidney (RTK), were treated with stage- and histology-specific therapy. Their tumors were assayed for LOH for polymorphic DNA markers on chromosomes 1p and 16q.

    RESULTS: LOH for 1p or 16q was rarely observed in CCSK (n = 90) or RTK (n = 22). The relative risk (RR) of relapse for patients with FH stage I to IV tumors with LOH, stratified by stage, was 1.56 for LOH 1p (P = .01) and 1.49 for LOH 16q (P = .01), whereas the RR of death was 1.84 (P = .03) and 1.44 (P = .15), respectively. When the effects of LOH for both regions were considered jointly among patients with stage I to II FH disease, the risks of relapse and death were increased for LOH 1p only (RR = 2.2, P = .02 for relapse; RR = 4.0, P = .02 for death), for LOH 16q only (RR = 1.9, P = .01 and RR = 1.4, P = .60) and for LOH for both regions (RR = 2.9, P = .001 and RR = 4.3, P = .01) in comparison with patients with LOH at neither locus. The risks of relapse and death for patients with stage III to IV FH tumors were increased only with LOH for both regions (RR = 2.4, P = .01 and RR = 2.7, P = .04).

    CONCLUSION: Tumor-specific LOH for both chromosomes 1p and 16q identifies a subset of FH Wilms tumor patients who have a significantly increased risk of relapse and death. LOH for these chromosomal regions can now be used as an independent prognostic factor together with disease stage to target intensity of treatment to risk of treatment failure.

    INTRODUCTION

    Wilms tumor is the most frequent malignant renal tumor in children. Approximately 460 new cases are diagnosed annually in the United States.1 The National Wilms Tumor Study (NWTS) Group (NWTSG) has completed four clinical trials2-6 which have resulted in the survival rate of children with Wilms tumor increasing from 20% before the regular administration of postnephrectomy chemotherapy to 90%.6 These results have been achieved while using shorter duration and total amount of chemotherapy and lower doses of radiation.5,6

    Therapy for patients with favorable-histology (FH) Wilms tumor is based on the risk of relapse using such variables as age at diagnosis, lymph node involvement, local or intravascular tumor extension, and presence of metastatic disease. Further refinement of the therapy and improvement in the outcome will depend on more accurate stratification of patients using novel prognostic factors. Refined risk stratification will allow targeting intensified therapy to only those at higher risk of recurrence, whereas patients at lower risk may be cured with less therapy than they currently receive. On the basis of our extensive study of clinical parameters, these novel prognostic factors will most likely be biologic in nature.

    In 1994, Pediatric Oncology Group (POG) investigators showed that, among 232 children with Wilms tumor registered on NWTS-3 and -4, loss of heterozygosity (LOH) for polymorphic DNA markers on chromosome 16q, present in tumor tissue from 17.2% of those with favorable or anaplastic histology tumors, was associated with statistically significantly poorer 2-year relapse-free and overall survival percentages even when adjusted for stage or histology.7

    LOH for chromosome 1p, present in tumor tissue from 11% of children with Wilms tumor, was also associated with poorer relapse-free and overall survival, although these results were not statistically significant. By contrast, LOH for 11p, a region thought to contain at least two Wilms tumor–related genes, was found in 33% of cases, but was not associated with any difference in outcome.7

    NWTS-5 was designed to prospectively test the hypothesis that LOH for chromosome 16q or chromosome 1p in Wilms tumor tissue was associated with a poorer prognosis for children with FH Wilms tumor, all of whom were treated with stage-specific treatment regimens. The study was designed to detect clinically significant associations within stages of disease; namely stages I, II and III/IV. We now report the results of the assays for LOH on chromosomes 16q and 1p and the relationships with outcome.

    PATIENTS AND METHODS

    NWTS-5 was a multi-institutional clinical trial that nonrandomly assigned treatment regimens for several renal tumors and assessed the tumors for LOH. The study accrued patients between August 1995 and June 2002. This report utilizes follow-up data as of August 17, 2004. The protocol was approved by the institutional review boards of all institutions that entered patients onto NWTS-5. All parents or guardians signed informed consent before study enrolment, for therapy and biology studies as well as for subsequent banking of remaining samples for future research. The staging system utilized was essentially the same as previously described.5 Chest x-ray films, abdominal ultrasound, and computed tomography (CT) studies were required for staging. Histologic diagnosis was confirmed by central review by NWTSG pathologists in all cases. This study included all patients, whether they underwent initial nephrectomy followed by adjuvant chemotherapy, or prenephrectomy chemotherapy, as long as a pretherapy sample of tissue was available for analysis. Patients with FH stage I disease who were at least 24 months of age or whose kidney and tumor had a combined weight of more than 550 grams and those with stage II tumors were treated with vincristine and dactinomycin following regimen EE-4A as previously described.5 Patients with FH stage III and IV tumors were treated in addition with doxorubicin and postoperative radiation therapy to the tumor bed and other infradiaphragmatic and metastatic sites as necessary following regimen DD-4A.5

    Local investigators were requested to submit a sample of frozen tumor tissue and associated normal kidney, peripheral blood, serum, and urine from the patient and peripheral blood from both parents, although only the tumor tissue and a source of normal DNA (blood or kidney) were required. These were shipped on dry ice by overnight courier to the Cooperative Human Tissue Network, Pediatric Division (Columbus, OH), where they were stored at –80°C and then shipped in batches by overnight courier on dry ice to the NWTSG Biology Reference Laboratory in Edmonton, Alberta, Canada. Samples were again stored in a –80°C freezer until DNA was extracted.

    DNA was extracted from the peripheral blood or normal kidney and tumor using standard procedures.8 For Southern blots, 5-μg samples of the child's constitutional and tumor DNA were digested with the appropriate restriction enzyme, electrophoresed through 0.8% to 1.2% agarose gels and transferred to nylon membranes.9 DNA probes were prepared by labeling plasmid inserts, isolated in low-melt agarose, with [32P]dCTP by the random priming method.10 Prehybridization and hybridization of the membrane were performed under standard conditions8 and, along with washing, were performed in a hybridization incubator utilizing roller bottles. Following washing at high stringency (55° with 0.1 x 0.15 m NaCl/0.015 m sodium citrate, 0.1% sodium dodecyl sulfate [SDS]), the membranes were exposed to Kodak XAR5 film at –70° using intensifier screens. Polymerase chain reaction (PCR) analysis was performed with 100 ng genomic DNA, utilizing Taq polymerase (New England Biolabs, Ipswich, MA) in a PerkinElmer 9600 Thermal Cycler (PerkinElmer Life and Analytical Sciences, Shelton, CT). DNA was amplified using standard conditions, 94° for 5 minutes followed by 35 cycles at 55° for 1 minute, 72° for 1 minute and 1 minute at 94° with a final extension at 72° for 10 minutes. PCR products were separated on 6% to 8% nondenaturing polyacrylamide gels and detected by ethidium bromide staining.

    At the start of this study, polymorphic loci were selected to reasonably evenly span the minimal regions of LOH identified in our earlier report.7 To retain some commonality with markers assessed in the first study and to maximize rates of informativeness, one locus on each of chromosomes 1p and 16q was a Southern blot–based polymorphism (D1Z2 and D16S7 respectively). Midway in the study, newer genomic mapping information indicated that several loci that had been selected were in fact very near each other, leaving large regions unevaluated for LOH. Furthermore, new highly informative PCR-based polymorphisms were identified that mapped very close to the Southern blot–based loci and, in fact, the D1Z2 locus proved hard to evaluate. Thus, several new loci were therefore selected, and to maximize both informativeness and efficiency, the following algorithms were used to select loci for testing each chromosomal region. For chromosome 16q: D16S7 or D16S2621; D16S422 and D16S402 only if D16S422 was noninformative (NI); D16S518 and D16S3101 only if D16S518 was NI; D16S421; and D16S400 (Fig 1). For chromosome 1p: D1S80 and D1S243 only if D1S80 was NI, and D1S468 only if D1S243 was NI; D1S214 and D1S244 and D1S1612 only if both D1S214 and D1S244 were NI. D1Z2 results were not used and all cases were tested for D1S80 instead as above (Fig 2).

    LOH was considered to be present if one of the two alleles in the constitutional DNA was absent or definitely reduced in the tumor DNA as determined by visual inspection. In cases with reduction in intensity but not complete loss, so-called partial loss or allelic imbalance, the result was confirmed in duplicate. Quantitative measures of band intensity were not used. Each tumor was categorized as having LOH for a chromosomal region if loss was found at any single informative locus, as retaining heterozygosity (ROH) if heterozygosity was maintained at all loci, and NI if all loci were constitutionally homozygous.

    The assays on approximately 10% of the cases were repeated to ensure reproducibility as a measure of quality control. For a randomly chosen set of cases, DNA was extracted from a second aliquot of tumor and all LOH assays were independently repeated. In situations where the results for the two duplicates did not match, all assays from both samples were again repeated.

    The measure of therapeutic outcome was the percentage of patients who were alive and free of recurrent or progressive disease 4 years after diagnosis. Patients who relapsed or had progression of disease that necessitated a change in therapy were considered treatment failures. Development of new disease in the contralateral kidney and toxic deaths were not considered treatment failures in these analyses because the intent was to study the biologic effect of LOH on tumor progression. All deaths were counted in the survival analyses, however.

    Data were analyzed using standard statistical methods that included Fisher's exact, 2 and trend tests for contingency tables, Kaplan-Meier estimates of ordinary and relapse-free survival distributions,11 and log-rank comparisons.11,12 Relative risk (hazard ratio) estimates, together with 95% CIs, were based on the Cox proportional hazards model13 with and without stratification for other prognostic factors.

    RESULTS

    There were 2,449 patients who met the clinical criteria for this study. Of these children, 2,387 (97.5%) were entered onto the study from 214 institutions in the United States or Canada, representing an accrual of almost three quarters of the pediatric renal tumors expected to have occurred during this time period. The remaining 62 children were treated at eught institutions in Australia, New Zealand, Switzerland, or the Netherlands. Of the 2,449 cases, adequate pretherapy tissue for LOH analyses was received for 2,082 (85%). For 19 cases, the histology was unknown. Of the 2,063 patients remaining, 2,030 had informative LOH results for chromosome 1p, 2,025 for chromosome 16q, and 2,023 for both chromosome 1p and 16q. Follow-up data were missing for two patients, leaving 2,021 on whom analyses of outcome were based. Three stage I FH patients did not have enough information to be categorized into I/age < 24months/weight < 550 g or I/age 24 months or weight 550 g and so were excluded from the analyses involving these categories.

    The frequency of LOH by tumor histology (FH Wilms tumor, anaplastic Wilms tumor, clear-cell sarcoma of the kidney [CCSK] and rhabdoid tumor of the kidney [RTK]) is shown in Table 1. There was a significantly greater incidence of LOH at 16q in anaplastic Wilms tumors than in those with FH (32.4% v 17.4%; P = .001).

    The frequency of LOH by stage in FH Wilms tumors is shown in Table 2. There was no significant association between LOH on either chromosome 16q or 1p and stages II to IV. The frequencies of LOH for chromosomes 1p and 16q were both lower, however, in the subset of stage I patients who were < 2 years of age at diagnosis and had tumors weighing less than 550 g.

    The frequency of LOH in FH tumors varied by patient age at diagnosis. When the 1,659 patients with stage I to IV FH tumors were categorized as age 0 to 1 years, 2 to 3 years or older than 4 years, the frequency of LOH 1p was 9.9%, 8.5% and 15.4% respectively (P = .002), and for LOH 16q was 10.1%, 20.4%, and 20.7%, respectively (P < .0001).

    There were no significant associations between LOH and outcome in patients with anaplastic Wilms tumors, CCSK or RTK (data not shown). Further analyses involve only those with stages I to IV FH Wilms tumor who had submitted adequate tissue for pathological evaluation and biological studies. Three patients in this group with stage I disease but unknown specimen weight were omitted, leaving 1,724 for analysis of treatment results.

    As shown in Tables 3 and 4, patients with tumor-specific LOH 1p or 16q had a significantly increased relative risk (RR) of relapse of 1.56 and 1.49 (stratified by stage), respectively. The pattern of the differences observed was similar for LOH 1p and 16q but the differences were slightly larger for LOH 1p. Although the stage-specific RRs for relapse appeared higher for stage I and II disease, and were individually statistically significant only for stage II disease, a likelihood ratio test for statistical interaction between stage and LOH provided no evidence to suggest that the RR in fact differed among stages (P = .39 and .22 for 1p and 16q, respectively). The summary RRs of death after stratification by stage were similarly elevated for cases with LOH 1p or 16q. However, the RR for LOH 16q did not differ significantly from one and the CIs for both RRs were substantially wider due to the many fewer events observed. Although several of the individual RRs of death for subgroups of patients who had stage I tumors were apparently statistically significant, these results involved only one to three events and must be interpreted with great caution.

    Although the study was not designed to examine the joint effects of LOH 1p and 16q, a subsequent exploratory analysis suggested that much of the adverse effect associated with LOH at either locus was in fact confined to the relatively small group of patients who had LOH at both loci. In these analyses, patients with stage I and II tumors were combined (Table 5) because all were treated with the same two-drug chemotherapy regimen (EE4A). Similarly, patients with stage III and IV tumors were all treated with three-drug therapy including doxorubicin (regimen DD4A) and were grouped together for these analyses (Table 6).

    Among 970 patients with low-stage disease, there was a difference in 4-year relapse-free survival (RFS) between patients with LOH for either 1p alone (80.4%) or 16q alone (82.5%), or for both chromosomes jointly (74.9%), relative to those with LOH for neither chromosome (91.2%; Table 5 and Fig 2). RRs of relapse were significantly different from 1 for each of the three LOH groups. The risk of death for patients in these groups was also increased, though not significantly so for LOH for 16q alone (Table 5).

    Table 6 and Figure 3 show the corresponding results for 686 patients with advanced stage disease, among whom only those with tumors with LOH for both chromosomes 1p and 16q had an increased RR of relapse (RR = 2.41; P = .01) and death (RR = 2.66; P = .04). A likelihood ratio test for differences in RR associated with LOH between the subgroups of patients with low- versus advanced-stage disease yielded an approximate 2 statistic of 7.82 on 3 df (P = .05) for relapse and 5.42 (P = .14) for death. The differences were primarily in the coefficients for LOH for chromosome 1p only and LOH for chromosome 16q only. The RR of relapse (stratified by stage) in the combined group of 1,656 patients with both low and advanced stage disease were RR = 1.25 (95% CI, 0.76 to 2.07; P = .39) for LOH for 1p only, RR = 1.28 (95% CI, 0.78 to 1.86; P = .20) for LOH for 16q only, and RR = 2.59 (95% CI, 1.62 to 4.15; P = .0001) for LOH for both loci. The corresponding relative risks of death were RR = 1.25 (95% CI, 0.54 to 2.93; P = .60), RR = 1.00 (95% CI, 0.51 to 1.98; P = 1.00) and RR = 3.11 (95% CI, 1.52 to 6.37, P = .002), respectively. These differences remained when analyses were stratified for age at diagnosis in addition to stage.

    The site of relapse did not correlate with the presence or absence of LOH for 1p, LOH for 16q or LOH for both 1p and 16q (Table 7).

    Since the categorization by LOH involved informative data at different loci in particular tumors, we also examined the outcome for patients with tumors categorized for LOH 1p using only the most telomeric three loci, for which 95.1% of patients were informative versus the three most centromeric for which 99.2% were informative. The RR of relapse was similar to that shown in Table 3 for patients classified using all the data and was not different whether the telomeric (RR = 1.76) or centromeric (RR = 1.67) loci were considered.

    For chromosome 16q, we categorized tumors by LOH using four pairs of adjacent loci and found that the rates of informativeness at each pair ranged from 75.7% to 85.5%. Again, the RR of relapse was similar for each method of LOH categorization (data not shown).

    In approximately 10% of cases, a second set of results was generated from DNA extracted from a second aliquot of tumor tissue from a different part of the original tumor. For chromosome 1p, results for 226 (99.1%) of the 228 informative cases were identical. In one of 30 cases showing loss in the repeat sample, LOH was not observed in the first sample, and conversely, in one of 30 cases showing loss in the first sample, LOH was not observed in the repeat sample. These discordant results were themselves reproducible, however, and so were not the result of error or misinterpretation. Similarly, 224 (98.3%) of the 228 cases had concordant results for chromosome 16q LOH, but two of 47 cases with LOH on the original sample did not have LOH on the repeat and two of 180 cases that retained heterozygosity on the first sample had LOH on the second sample. Again, these results were reproducible.

    The vast majority of tumors with LOH for either chromosome 16q or 1p had undergone LOP for all informative loci tested. Markers for 16p or 1q were not included, and so the frequency of whole chromosome loss as the basis of LOH is not known. Considering only the loci described herein for chromosome 16q, the minimal region of LOH shared among all tumors was between D16S421 and D16S402 (Fig 1) and for chromosome 1p, between D16S468 and D1S244 (Fig 2). Detailed analyses of the extent of the deletions, utilizing additional markers in subsets of tumors will be published separately.

    DISCUSSION

    This study was undertaken to confirm and expand upon the findings from our preliminary study, which involved a convenience sample of patients registered on the third or fourth National Wilms Tumor Study (NWTS-3 and -4) and on the Pediatric Oncology Group Wilms Biology Study (POG 9046). The preliminary study included 232 patients. The results of this pilot study suggested that LOH for either chromosome 16q or 1p was associated with an increased risk of relapse, although the association with LOH 1p was not statistically significant.7 These results led to the next study of the NWTSG, which was designed to have sufficient power to detect associations between LOH for these two regions and outcome within clinical stages of FH Wilms tumor.

    The study was successfully carried out with about 70% of the expected number of pediatric renal tumors in North America registered on the trial and the requisite biologic samples submitted for 85% of eligible patients. The results of our study should therefore be applicable to the Wilms tumor population as a whole. This study included all patients whether they underwent initial nephrectomy followed by adjuvant chemotherapy, or prenephrectomy chemotherapy for tumors judged initially unresectable (9.3% of patients), as long as a pretherapy sample of tissue was available for analysis. The results are therefore potentially applicable to patients treated on International Society of Pediatric Oncology (SIOP) protocols which utilize prenephrectomy chemotherapy.

    The greater number of cases in NWTS-5 relative to the pilot study allowed a determination of the incidence of LOH 16q and 1p in the spectrum of childhood renal tumors. LOH of 16q or 1p was not observed in RTK and rarely in CCSK (Table 1), consistent with their distinct histologic appearance and clinical behavior relative to Wilms tumor. The incidence of LOH 16q (17.4%) and 1p (11.3%) in FH Wilms tumor was essentially identical to that observed in our previous study. LOH for 16q was significantly more common in anaplastic Wilms tumors than in favorable histology disease. A similar trend was observed for LOH 1p, but the difference was not significant. Whether this reflects more frequent involvement of the putative underlying tumor suppressor genes in the genesis of anaplasia or simply greater genomic instability in anaplastic tumors is not known but a similar finding has been noted by others.14

    LOH was significantly less frequent in younger patients. Interestingly, the pattern was somewhat different for the two chromosomes with a lower incidence of LOH 16q in the third of patients younger than 2 years versus those age 3 to 4 years or older than 4 years, whereas LOH 1p was less common in the two thirds of patients younger than 4 years compared with those who were older. Despite this association between LOH and age, the differences in outcome remained similar after stratification of analyses for age. This was most likely due the fact that all analyses were already stratified by stage, itself known to be associated with age.

    A subset of children younger than 2 years with stage I FH Wilms tumors weighing less than 550 g has been identified previously as having a superior outcome relative to those who are either older or who have a larger tumor.15 Interestingly, this subset had the lowest incidence of LOH for both chromosome regions compared with other Wilms tumors. Patients with bilateral tumors, who also tend to be younger, also had a lower incidence of LOH 1p. The relative infrequency of LOH is therefore an additional factor identifying these young children with small tumors as a biologically distinct group, adding to the rationale for treating them differently from other children with stage I tumors.16

    However, there were no further increases in the incidence of LOH by stage for stages II, III, and IV (Table 2). This suggests that LOH for these two regions, and by implication loss of function of the critical genes on 1p and 16q, is not involved in the degree of tumor invasiveness or in the development of metastases.

    When all stages of FH Wilms tumor were considered together, there was a significantly increased risk of relapse associated with LOH for 1p (RR = 1.56) and 16q (RR = 1.49; Tables 3 and 4). As with the incidence of LOH 1p and 16q, these RRs are consistent with those observed in our preliminary study. The earlier, smaller study suggested that the greater prognostic effect was associated with LOH 16q, although the CIs were large. Even in the current, much larger study, the 95% CIs overlap. As a result, it is not certain whether LOH for 16q or 1p is the stronger prognostic factor. Although LOH was examined at additional loci in this study as compared with our earlier pilot study, this is unlikely to have had a significant effect on the results since 99% of tumors with LOH had undergone LOH at all informative loci assayed. This shows that, for the great majority of tumors, the region affected by LOH is large and would have been detected by either methodology.

    While it appears in the current study that LOH at either 16q only or 1p only was associated with higher risks of recurrence for low stage (I/II) patients in comparison with advanced stage (III/IV), we were unable to rule out the possibility that the true relative risks were the same for the two groups except at the borderline of statistical significance (P = .05). We propose, nonetheless, that a likely basis for a greater association of LOH at a single chromosome with outcome in stage I/II disease is that the treatment for patients with low-stage disease differs significantly from that for patients with high stage disease. Two-drug chemotherapy with vincristine and actinomycin is used for stages I and II, whereas doxorubicin and radiation therapy are added for stages III and IV. Thus, the more intensive treatment for advanced-stage disease may overcome the effect of loss of the putative tumor suppressor gene(s) located within these chromosomal regions. The practical implication is that for low-stage tumors with LOH, the use of the more intensive regimen may overcome this adverse prognostic feature. We therefore examined the pattern of recurrence to determine whether distant or local relapses were predominant in those with LOH. The pattern of recurrence (Table 7) was similar to Wilms tumor patients in general and did not provide guidance as to whether the addition of radiation or doxorubicin might be most effective for these patients with stage I/II tumors with LOH.

    The adverse effect on outcome was greatest for those with LOH for both loci; and, for patients with advanced stage disease, was restricted to this LOH subgroup (Table 6). However, the fact that the upper bounds on the 95% CIs of RR for LOH at a single locus include the value RR = 1.5 indicate that one cannot rule out separate as well as joint effects, even among patients with advanced disease. If one considers primarily the results for the combined group of patients with stage I-IV disease, there are two possible interpretations. LOH at 16q and LOH at 1p may both be associated with a poor outcome, with effects on the RR that are additive or even superadditive. Consequently, the patients with the worst outcomes are those in the relatively small group with LOH at both loci. Alternatively, the only subgroup of patients that have a worsened outcome may be the subgroup with LOH at both 1p and 16q. Under either model, the strongest effect is associated with LOH of both chromosomal regions. When patients with low-stage disease are considered separately, LOH at either locus alone may convey the same amount of risk as LOH at both together (Table 5). However, the statistical test that could help justify such separate consideration, namely the comparison of joint LOH RRs between the low- and advanced-stage subgroups, achieved only borderline significance. Analogous tests for statistical interaction between effects of LOH at each locus separately and individual stages were negative (Tables 3 and 4).

    The RRs of death for patients with tumors with LOH were increased comparably to the RRs of relapse for every pattern of LOH and stage of disease. These differences were statistically significant for LOH for both chromosomal regions in low and advanced stage tumors and for LOH 1p alone in low stage disease. Thus, these patients with recurrent disease do not all successfully undergo salvage treatment.

    As part of the study design, we repeated the entire DNA analysis for approximately 10% of the cases. Selection of cases was based on a random selection of 10 numbers between one and 100. For each case whose study number ended in one of these digits, DNA was extracted from a second piece of tumor and all PCR reactions repeated. The purpose of this exercise was to demonstrate reproducibility of the results. It was therefore surprising to find cases in which the results were discordant but reproducible. Thus, cells from one part of the tumor harbor LOH but not in other parts, demonstrating tumor heterogeneity for this molecular genetic change. This would be consistent with this event's having occurred subsequent to the development of the tumor rather than initiating it because in the latter situation one would expect the LOH to be present in all cells. Reassuringly, however, by testing only one part of each tumor, LOH 1p was missed in only one of 198 (0.5%) samples and LOH 16q in 2 of 178 (1.1%) cases.

    LOH that occurs in a recurrent pattern suggests the location of an underlying recessive gene that is involved in the pathogenesis of the disease being studied. LOH is thought to represent the loss of the second allele, the first having been inactivated by a more localized mutation. An exception to this involves imprinted genes in which there is only one active allele. In the case of LOH of chromosomes 1p and 16q however, there is no indication of bias to loss of specific parental allele.7 In any event, LOH at chromosome 1p and 16q would be expected to represent loss of function of genes at these two locations. On the one hand, it is not unreasonable in this genetically heterogeneous tumor to find different genes involved in different tumors. Each gene might contribute to the malignant phenotype (tumor initiation), tumor invasiveness (tumor progression) and/or responsiveness of the tumor to therapy. On the other, it is more difficult to explain the simultaneous involvement of two genes at two different chromosomal locations.

    Two explanations for this observation were considered. First, it is possible that the putative genes at 1p and 16q interact in a synergistic fashion such that this coincident loss is selected for by a survival advantage, particularly in the face of therapy. Alternatively, it is possible that involvement of these two regions results from a single chromosomal mechanism. This would seem more likely than the four independent events required to inactivate two different genes through independent events. An unbalanced translocation between chromosome 16q and 1q, which has been frequently noted in cytogenetic studies of Wilms17,18 followed by duplication of the normal chromosome 1 and subsequent non-disjunction could result in monosomy 16q (manifested as 16q LOH), isodisomy of 1p (manifested as LOH 1p) and trisomy of 1q. This possibility is under active investigation since, if true, the gene of interest might well be on the long arm of chromosome 1 (1q) rather than the short arm, with LOH 1p serving as a surrogate marker of the event.

    If the first model involving two separate events is correct, then recessive Wilms-related tumor suppressor genes are located on both 1p and 16q. We are therefore analyzing the minimal overlapping regions of LOH to narrow the possible location and will perform mutation analyses on candidate genes with plausible biologic characteristics. Identification of the important gene(s) under the other possible model will be much more challenging, however, because the presence of trisomy implies increased gene dosage as the pathogenic mechanism. It is very possible therefore that more than one gene is affected but furthermore since the sequence of the relevant gene(s) is not altered it is much more difficult to prove a mechanistic role in tumorigenesis. This will require large scale studies of expression profiles and statistical analyses of associations between gene dosage and phenotypic features such as outcome.

    Since constitutional loss or deletions of chromosome 16q or 1p have not been described in Wilms tumors, and genetic linkage studies have excluded these regions in families segregating Wilms tumor, it is unlikely that these genes are involved in the development of Wilms tumors per se. Rather, because LOH is associated with prognosis, it is more likely that these genes at 1p and 16q are involved with other phenotypic features of the tumor. The fact that the effect on outcome is greater for less intensively treated tumors (stage I and II) suggests that the function of these genes may relate to the ability of the cells to tolerate certain chemotherapy, in particular vincristine and actinomycin, or irradiation.

    LOH for chromosome 1p has been shown to be adversely prognostic in several other tumors including neuroblastoma, meningioma, and breast and endometrial carcinoma.19-22 The region of LOH for each of these tumors includes 1p36.2-36.3. It remains possible that a single gene could be involved with the phenotype of a number of disparate malignancies rather than being a gene that specifically affects the biologic behavior of Wilms tumor.

    Historically the surgicopathologic stage and classification into favorable versus anaplastic histology have been used for the stratification of Wilms tumor therapy. Further advances in the treatment of this disease will depend on the better identification of the small subset of patients who remain destined to relapse. The alternative approach, generalized intensification of therapy for all patients, would result in overtreatment of the majority with benefit for only a few. Our findings suggest the next step in the process. Patients with low-stage disease but with LOH for chromosome 1p and 16q have only a 75% RFS, clearly worse than their counterparts without LOH who have a 91% RFS. Patients with combined LOH may benefit from the addition of doxorubicin to their therapy. The efficacy of this approach should be studied. Likewise, patients with advanced stage disease, whose tumors harbor LOH for 1p and 16q, also have an inferior outcome, even when treated with three-drug therapy. These patients may benefit from intensification of their therapy with the addition of other chemotherapeutic agents that are active against Wilms tumor, such as cyclophosphamide and etoposide23 or carboplatin and etoposide.24,25

    Clearly, additional and more sensitive and specific markers of outcome are still needed. Ongoing studies of telomerase levels,26 markers of apoptosis,27 markers of mitotic activity,28 and now gene expression arrays may individually or together provide additional prognostic information.

    Authors' Disclosures of Potential Conflicts of Interest

    The authors indicated no potential conflicts of interest.

    Acknowledgment

    We thank the investigators of the Children's Oncology Group and the many pathologists, surgeons, pediatricians, radiation oncologists, and other health professionals who managed the children entered on the National Wilms Tumor Studies; Audrey Evans, MD, for assisting with the review of the flow sheets; and Kevin Dietrich and Kay Ziebart for the technical aspects of the DNA analyses.

    NOTES

    Supported in part by US Public Health Service Grant No. CA-42326.

    Presented at the United Kingdom Children's Cancer Study Group Biology of Childhood Cancer Meeting, London, United Kingdom, December 1-3, 2002, and at the Annual Meeting of the International Society of Pediatric Oncology, Cairo, Egypt, September 10-13, 2003.

    Terms in blue are defined in the glossary, found at the end of this issue and online at www.jco.org.

    Authors' disclosures of potential conflicts of interest are found at the end of this article.

    REFERENCES

    Crist WM, Kun LE: Common solid tumors of childhood. N Engl J Med 324:461-471, 1991

    D'Angio GJ, Evans AE, Breslow N, et al: The treatment of Wilms' tumor. Cancer 38:633-646, 1976

    D'Angio GJ, Evans AE, Breslow N, et al: The treatment of Wilms' tumor: Results of the second National Wilms' Tumor Study. Cancer 47:2302-2311, 1981

    D'Angio GJ, Breslow N, Beckwith B, et al: Treatment of Wilms' Tumor: Results of the Third National Wilms' Tumor Study. Cancer 64:349-360, 1989

    Green DM, Breslow NE, Beckwith JB, et al: Comparison between single-dose and divided-dose administration of dactinomycin and doxorubicin for patients with Wilms' tumor: A report from the National Wilms' tumor study group. J Clin Oncol 16:237-245, 1998

    Green DM, Breslow NE, Beckwith JB, et al: Effect of duration of treatment on treatment outcome and cost of treatment for Wilms' tumor: A report from the National Wilms' Tumor Study Group. J Clin Oncol 16:3744-3751, 1998

    Grundy PE, Telzerow PE, Breslow N, et al: Loss of heterozygosity for chromosomes 16q and 1p in Wilms' tumors predicts an adverse outcome. Cancer Res 54:2331-2333, 1994

    Sambrook J, Fritsch EF, Maniatis R: Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1989

    Rigaud G, Grange T, Pictet R: The use of NaOH as transfer solution of DNA onto nylon membrane decreases the hybridization efficiency. Nucleic Acids Res 15:857-857, 1987

    Feinberg AP, Vogelstein B: A technique for radiolabelling DNA restrictions endonuclease fragments to high specific activity. Anal Biochem 132:6-13, 1983

    Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. Am Statist Assoc J 53:457-481, 1958

    Peto R, Peto J: Asymptotically efficient rank invariant test procedures. J R Statist Soc A 135:185-192, 1972

    Cox DR: Regression models and life tables. J R Stat Soc B 34:187-220, 1972

    Grundy RG, Pritchard J, Scambler P, et al: Loss of heterozygosity on chromosome 16 in sporadic Wilms' tumour. Br J Cancer 78:1181-1187, 1998

    Green DM, Breslow NE, Beckwith JB, et al: Treatment outcomes in patients less than 2 years of age with small, stage I, favorable-histology Wilms' tumors: A report from the National Wilms' Tumor Study. J Clin Oncol 11:91-95, 1993

    Green DM, Breslow NE, Beckwith JB, et al: Treatment with nephrectomy only for small, stage I/favorable histology Wilms' tumor: a report from the National Wilms' Tumor Study Group. J Clin Oncol 19:3719-3724, 2001

    Kaneko Y, Knodo K, Rowley JD, et al: Further chromosome studies on Wilms' Tumor cells of patients without aniridia. Cancer Genet Cytogenet 191-197, 1983

    McDowell H, Howard P, Martin J, et al: Chromosome 1 studies in Wilms' tumor. Cancer Genet Cytogenet 43:203-209, 1989

    Maris JM, White PS, Beltinger CP, et al: Significance of chromosome 1p loss of heterozygosity in neuroblastoma. Cancer Res 55:4664-4669, 1995

    Bello MJ, de Campos JM, Kusak ME, et al: Allelic loss at 1p is associated with tumor progression of menigiomas. Genes Chromosomes Cancer 9:296-298, 1994

    Borg A, Zhang Q-X, Olsson H, et al: chromosome 1 alterations in breast cancer: allelic loss on 1p and 1q is related to lymphogenic metastases and poor prognosis. Genes Chromosomes Cancer 5:311-320, 1992

    Arlt MF, Heroz HT, Mutch DG, et al: Frequent deletion of chromosome 1p sequences in an aggressive histologic subtype of endomerial cancer. Hum Mol Genet 5:1017-1021, 1996

    Yankelevich,M, Dolgopolov,I, Ravshanova, R et al: Efficacy and toxicity of ifosfamide/cyclophosphamide, carboplatin, and etoposide (ICE/CCE) chemotherapy with or without GM-CSF in relapsed or refractory Wilms' tumor: A single institution study. Int J Ped Hematol Oncol 6:331-338, 1999

    de Camargo B, Melaragno R, Saba e Silva N, et al: Phase II study of carboplatin as a single drug for relapsed Wilms' tumor: Experience of the Brazilian Wilms' Tumor Study Group. Med Pediatr Oncol 22:258-260, 1994

    Pein F, Tournade MF, Zucker JM, et al: Etoposide and carboplatin: A highly effective combination in relapsed or refractory Wilms' tumor—A phase II study by the French Society of Pediatric Oncology. J Clin Oncol 12:931-936, 1994

    Dome JS, Chung S, Bergemann T, et al: High telomerase reverse transcriptase (hTERT) messenger RNA level correlates with tumor recurrence in patients with favorable histology Wilms' tumor. Cancer Res 59:4301-4307, 1999

    Takamizawa S, Scott D, Wen J, et al: The survivin:fas ratio in pediatric renal tumors. J Pediatr Surg 36:37-42, 2001

    Arcellana-Panlilio MY, Egeler RM, Ujack E, et al: Decreased expression of the INK4 family of cyclin-dependent kinase inhibitors in Wilms tumor. Genes Chromosomes Cancer 29:63-69, 2000(Paul E. Grundy, Norman E.)