Classic and Molecular Cytogenetic Analyses Reveal Chromosomal Gains and Losses Correlated with Survival in Head and Neck Cancer Patients Free

Carcinomas of the head and neck represent the sixth most frequent cancer worldwide and ∼90% to 95% are squamous cell carcinomas. Tobacco and alcohol consumption are the most important nongenetic risk factors associated with the development of head and neck squamous cell carcinomas (HNSCC; ref. 1). Estimated age-standardized rates per 100,000 for 1990 showed 12.8 men and 3.7 women of oral cavity and pharynx cancers and 6.5 men and 1.1 women for laryngeal cancer for Tropical South America, compared with 12.1 and 5.7 men, respectively, to oral cancer and pharynx and laryngeal cancers for all areas in the world (2). Tropical South America has one of the world's highest age-standardized rates of head and neck carcinomas (2, 3).

Abnormal karyotypes have been reported in >250 cultured HNSCC (4). Karyotypes are often complex, with many numerical and unbalanced structural aberrations. The most frequent losses involve chromosome arms 3p, 7q, 8p, 9p, 11q, 13p, 14p, 15p, 16p, and 18q and gains at 1q, 3q, 8q, and 15q and band 11q13. Among recurrent structural alterations, the most common are 8q isochromosomes, 3p deletions, and the presence of homogeneously staining regions at 11q13 (5–15). Previous cytogenetic studies have primarily been observational and have been drawn from predominantly European and North American cohorts. In this present study, we have examined cytogenetic aberrations in the context of accompanying clinical variables such as anatomic sites, histology, stage, grade, treatment, and outcome. Moreover, this study comprises a typical Brazilian patient cohort comprising multiracial ethnicity and the additional risk factors associated with South American lifestyle. This unique patient cohort has allowed us to associate genomic alterations with familial history of cancer and other prognostic factors in 67 consecutive head and neck tumors arising in a representative Brazilian patient population.

Patients. Sixty-seven head and neck tumor samples were obtained from the AC Camargo Hospital, São Paulo, Brazil from 1994 to 1996. The patients were accrued consecutively and the inclusion criteria were no previous new histologic diagnosis of head and neck cancer and a tumor larger than 1 cm in size. Three samples were excluded: an olfactory neuroblastoma (case 1), a maxillary antrum adenocarcinoma (case 66), and an oral cavity liposarcoma (case 67). A total of 64 squamous cell carcinomas were eligible to be included in this study, and the tumor sites were the following: oral cavity (34 cases), larynx (18 cases), and pharynx (12 cases). None of the patients had received radiotherapy or chemotherapy before surgery and sample collection for cytogenetic analysis. Selected patients with oral, oropharynx, hypopharynx, and supraglotic carcinomas were submitted to elective selective neck dissection. All patients with positive margins, tumors with perineural infiltration, or vascular embolization as well as those with metastatic lymph nodes confirmed by histologic examination of the surgical specimens were submitted to postoperative radiotherapy. None received adjuvant chemotherapy. All of the patients were advised of the procedures and provided written informed consent. This study was approved by the Brazilian Ethics Committee-CONEP 813/2000. The medical records of all patients were examined to obtain detailed clinical and histopathologic data, including information about consumption of alcohol beverages and tobacco, cancer family history, and other demographic data. Alcohol users were defined as the weekly consumption of 7 to 10 drinks of alcohol for >1 year anytime. Types of alcoholic beverages (wine, beer, or spirit) were converted into drink equivalents, and each was given a value of 13.6 g of absolute alcohol. Tobacco use was defined as patient responding positively to being a current or regular smoker for >1 year in their lifetime (unless smoking stopped >20 years ago). Information concerning the extent of usage, such as number of pack-years, was not obtained. For the family history of cancer, first-degree and second-degree relatives with cancer were considered as positive, and whenever possible, the evidence of cancer was based on assessment of medical records or ascertained from the death certificate. The histopathologic classification was based on the WHO International Classification of Diseases for Oncology (16). The clinical staging was determined using the tumor-node-metastasis staging system (17). For statistical analysis, the patients were grouped in two classes because of the small number of cases in each tumor stage [i.e., (T₁-T₂) versus (T₃-T₄) and (N₀) versus (N₁, N_2a, N_2b, N_3c)]. The patients were followed prospectively from surgery date to December 2003 and the follow-up time range was 1 to 103.5 months (median, 35.9 months). The median age of the patients was 58.0 years (range, 32-87 years), the male-to-female ratio was 9.7:1; 79.7% (51 of 64) were tobacco consumers and 76.6% (49 of 64) were alcohol users (Table 1).

Cytogenetic Study. Fresh tumor samples were obtained under sterile conditions and immediately processed. Chromosomal preparations and cytogenetic analysis were carried out by standard techniques following short-term harvesting of primary cultures (9-12 days) as previously described (15). Metaphase chromosomes were subjected to GTG banding (18). The mean number of metaphases analyzed in each sample was 20 (range, 7-35 cells). The karyotype description and the requirements for clonality followed the International System for Human Cytogenetic Nomenclature criteria (19). A further requirement for clonality was that cytogenetic changes had to be found in at least two culture flasks.

Fluorescence In situ Hybridization. Slides stored at −20°C were used for fluorescence in situ hybridization (FISH) and spectral karyotyping analyses. Probe labeling, hybridization, suppression hybridization, detection, fluorescence microscopy, and microphotography were done as previously described (20, 21). The PAC134J15 (110-kb insert) was isolated from a screen with probe D22S543 (22). The DNA probe was labeled with biotin-14-dUTP (Life Technologies, Carlsbad, CA) by nick translation (21). FISH was done in 17 cases. An average of 100 nonoverlapping interphase cells with intact morphology and 10 metaphase cells based on 4,6-diamidino-2-phenylindole counterstaining were scored to determine the number of hybridization signals for the chromosome 22 probe. Normal controls were phytohemagglutinin-stimulated normal male lymphocytes. A case was considered to have a numerical chromosome abnormality when the percentage of cells showing an abnormal number of hybridization signals was higher than the mean value plus 2 SDs obtained for the same chromosome in the normal control. The results were verified independently by a second observer.

Spectral Karyotyping. A commercially available spectral karyotyping kit from Applied Spectral Imaging (Carlsbad, CA) was used to analyze 12 metaphases from two cases. Slide treatment, post-hybridization detection, and washes were done according to published protocols and the manufacturer's instructions (23). Spectral images were acquired and analyzed with a SD 200 spectral bio-imaging system (Applied Spectral Imaging Ltd., MigdalHaemek, Israel) attached to a Zeiss microscope (Axioplan 2) by means of a C-mount consisting of an optical head with a special Fourier-transformed spectrophotometer (SAGNAC common path interferometer) to measure the spectrum, and a cooled CCD camera for imaging. The images were stored in a computer for further analysis using the SKYVIEW (ASI, Carlsbad, CA) software. Based on the measurement of the spectrum for each chromosome, a spectral classification algorithm was applied. 4,6-Diamidino-2-phenylindole images were acquired from all metaphases analyzed using a 4,6-diamidino-2-phenylindole–specific optical filter.

Statistical Analysis. Associations with the monosomy and trisomy groups and risk factors were verified with 5% of significance level using the Fisher's exact test. The tobacco and alcohol consumption were determined using criteria described above. The Kaplan-Meier method was applied to obtain survival curves and to calculate the actuarial estimators of survival (24). The log-rank test was used to compare survival curves of the different categories of the same variable. The patients were followed up and the interval in months between surgery and death or until the last date of follow-up in December 2003 was calculated.

Cytogenetic Analysis. The karyotypes of the 67 tumors after GTG banding are presented in Table 1. Chromosome banding analysis showed complex karyotypes with multiple numerical and structural alterations in almost all cases. Six cases showed normal karyotypes. Although in low frequency, normal cells were detected in all cases, except in five samples.

In the 64 cases of HNSCC, numerical changes most frequently involved the loss of chromosomes Y (26 cases), 19 (14 cases), 22 (12 cases), 9 (11 cases), 16 and 20 (9 cases each), 10 and 21 (8 cases each), and 13 (7 cases). Gains most frequently affected the chromosomes 22 (23 cases), 8 and 20 (11 cases each), 17 and 18 (6 cases each). In the oral cavity, laryngeal, and pharyngeal carcinomas, the most common numerical chromosome alterations were trisomy 22 (44.1% of cases), trisomy 20 (20.2% of cases), and trisomy 22 (41.7% of cases), respectively. Loss of chromosome Y in male patients was detected in 14 of 29 oral cavity tumors, 7 of 12 pharyngeal, and 5 of 17 laryngeal carcinomas. Numerical chromosomal alterations identified in all samples are showed in Fig. 1.

The most frequent structural changes detected in the 64 HNSCC involved chromosomes 22 (12 cases), 6 (6 cases), and 12 (5 cases). Overall, 42 out of 64 cases showed clonal chromosome aberrations involving chromosome 22. Chromosome 22 was rearranged in 6 of 34 oral cavity tumors and in 6 of 18 laryngeal carcinomas. Losses of chromosome 22 were observed in 5 of 12 pharyngeal, 3 of 34 oral cavity, and 2 of 18 laryngeal carcinomas.

In addition, FISH was applied in 14 cases to evaluate the frequency of aneuploidies of chromosome 22. Interphase FISH yielded one signal (i.e., monosomy) in 12 cases and all the 14 cases showed three or more signals, with frequencies higher than the control (5% and 6%, respectively). The frequency of monosomy 22 ranged from 10% to 40% and the trisomy 22 from 9% to 27%. A small submetacentric chromosome was detected and was interpreted as del(22)(q13.1) (Fig. 2A). Using the probe PAC134J15, we confirmed the marker as a chromosome 22 derivative in eight cases (Fig. 2B).

To determine whether a cryptic chromosomal rearrangement had led to misclassification using conventional cytogenetic methods, we did spectral karyotyping analysis. Spectral karyotyping analysis was done in two cytogenetic preparations (cases 48 and 59). The analysis of 12 cells from case 48 revealed clonal losses involving chromosomes Y, 13, 19 and a gain of chromosome 22. Monosomies 7 and 9 and trisomies 4 and 12 were nonclonal. Thirteen metaphase cells from case 59 revealed losses of chromosomes Y and 10 and gains of chromosome 22. Figure 2C shows a cell from case 48 analyzed using the chromosome-painting probe.

Statistical Analysis. The mean age of the subjects was 57 years (range, 32-87 years). Of the 64 patients, 85.9% were coded T3 or T4 and 40.6% had more than one clinically positive lymph node. Around 78% of the patients were alcohol and/or tobacco users and only 25% had a positive family history of cancer. The absolute and relative frequencies of some selected clinical and demographic variables are shown in Table 2.The association test was accomplished for all chromosome alterations (data not shown). No significant associations were observed in univariate and survival analyses (data not shown) despite the high frequencies of trisomies and monosomies observed.

In spite of a small sample size, a significant association (P = 0.012) was observed between a family history of cancer and monosomy 17. Monosomy 22 and family history of cancer was marginally significant (P = 0.058). Borderline significance (P = 0.054) was detected between alcohol consumption and trisomy 20. Univariate analysis revealed a possible association between the presence of all monosomies and trisomies and the occurrence of death, but only monosomy 22 and trisomy 20 showed significant associations (Table 2).

The impact of the numerical chromosomal alterations on the overall survival probability was assessed by the log-rank test. The survival probability was worse in the presence of alterations in chromosomes 10, 15, 20, and 22, and the difference between the survival curves was significant (P < 0.05; Fig. 3). Although monosomy 22 was associated with a significant reduction on the survival expectancy (P = 0.021), in the analysis by T-stage, a nonsignificant association of survival rates (P = 0.061) and T3 -T4 staged cases was verified. T3 and T4 tumors with trisomy 10 (P = 0.020), monosomy 15 (P = 0.016), and trisomy 20 (P = 0.008) were significantly associated with survival rates decreased (data not shown). Patients with at least one of these chromosomal alterations were younger than 50 years in 25% of cases (7 of 28 cases), 14.3% (4 of 28 cases) were nonusers of alcohol and tobacco, 21.4% (6 of 28 cases) were nontobacco users, and 42.8% (12 of 28 cases) died of disease before 24 months. Taken together, the data presented indicates that the presence of cytogenetic aberrations of chromosomes 10, 15, 20, and 22 were a significant determinant of poor outcome in head and neck carcinomas.

Cytogenetic analysis of head and neck tumors has revealed extensive genetic heterogeneity and karyotype complexity. The pattern of chromosomal alterations observed in our G-banding study and the identification of the most frequently aberrant chromosomes were consistent with other reports on the cytogenetics for head and neck carcinomas (4, 10, 14, 25, 26).

Although head and neck carcinomas have been intensively studied using classic cytogenetic techniques (4) and comparative genomic hybridization (27, 28), the correlation between specific chromosomal abnormalities and clinical outcome, tumoral staging, anatomic sites, treatment, and family history of cancer is still poorly understood. In this study, we found different profiles of chromosome gains and losses depending on the tumor location. In 1995, Jin et al. (7) had already suggested that there might be karyotypic variations among HNSCC from different anatomic sites. Subsequently, the same group reported clonal chromosome aberrations in 56 laryngeal tumors; rearrangements involving 22p11-q11 were detected in 20% of the cases (29). We found similar chromosome alterations in 18 laryngeal tumors, including rearrangements involving 22q in six cases (33%). However, a small submetacentric marker del(22)(q13.1) was also observed in oral carcinomas (17% of cases) but not in pharyngeal tumors. A previous analysis by our group using microsatellite markers revealed a higher frequency of allelic losses on 22q12.1-13.3 in laryngeal tumors as compared with oral tumors and other HNSCC (30).

Changes involving chromosome 22 were the most frequent alteration detected by G-banding and confirmed by FISH and spectral karyotyping analyses: gains were detected in 25 cases, losses in 13 cases, and structural rearrangements in 12 cases. Abnormalities affecting 22q are common in squamous cell carcinomas of larynx (11, 29, 31, 32), oral cavity (6, 7, 13, 31, 33), and HNSCC from several anatomic sites (34–37). In agreement with our G-banding results, comparative genomic hybridization data have revealed copy number changes in >50% of the analyzed cases showing deletions and, more frequently, gains of chromosome 22. Recently, we studied 40 primary oral tumors by quantitative real-time PCR and identified a deletion of the DIA1 gene (mapped on 22q13) and D22S274 (22q13.31) sequences in 25% of the cases; these losses were significantly correlated with family history of cancer and a reduced probability of survival (38). As shown here, there was a significant association (P = 0.012) between a familial history of cancer and the loss of chromosome 17, whereas this association was only marginally significant (P = 0.058) for chromosome 22 loss. Loss of heterozygosity, deletions, and other rearrangements involving chromosome 17, particularly the TP53 gene, are the most common mutations described in HNSCC (reviewed in refs. 25, 39). TP53 mutations have been correlated with exposure to alcohol and tobacco and seem to be associated with a short recurrence interval (40). A novel candidate cancer susceptibility gene located on chromosome 17p is ELAC2. Carriers of mutations in ELAC2 have a higher risk of developing prostate cancer (41).

According to univariate analysis, overall poor survival probabilities were observed for the monosomies 15 and 22 and trisomies 10 and 20 (P < 0.05). Although monosomy 22 had a significant log-rank test P value, it was not significant (P = 0.061) in T3-to-T4 stages of the disease. The small sample size of our study precluded the use of multivariate analysis to determine whether monosomy 22 was a possible independent prognostic factor. Overall, these results indicated that the loss of 22 was correlated with a worse outcome. In agreement with our results, Ashman et al. (32), using comparative genomic hybridization analysis, showed that the deletion of 22q was associated with a reduced survival in head and neck cancer patients, suggesting that this abnormality is an independent prognostic marker.

We also assessed the impact of the chromosomal alterations on the overall survival rates. The probability of survival decreased with trisomy 10 (P = 0.008), trisomy 20 (P = 0.002), monosomy 15 (P = 0.005), and monosomy 22 (P = 0.021). Trisomy 10 has been most commonly observed as a nonrandom abnormality in hematologic diseases, such as acute myeloid leukemia (42, 43) and less frequently in some human solid tumors (4). In our series, all the five patients with trisomy 10 showed recurrence and/or metastasis and died of cancer. This finding suggests that trisomy 10 is associated with a more aggressive behavior in head neck tumors. The analysis of the distribution of this abnormality in the different anatomic sites showed that trisomy 10 was present in four cases of oral cavity tumors and in one laryngeal carcinoma.

An increase in the chromosome 20 copy number has been reported as a frequent aberration in several solid tumors, including oral carcinomas (44, 45). Trisomy 20 was found in all anatomic sites and 10 out of 11 patients died of the disease after 7 to 36 months. Amplification of 20q13 was considered a predictor of poor survival in breast cancer (46). Two candidate genes, the ZNF217 putative oncogene and CYP24 (encoding vitamin D 24 hydroxylase) are mapped at 20q13.2. Gains on chromosomes 8q and 20 were reported to be predictors of a poor outcome in hepatoblastomas (47). Although gains in chromosome 20 are a common event reported in several tumor types, their biological significance remains unknown.

Karyotypic and allelic losses of chromosome 15 are associated with aggressiveness in several tumors (48–50). Recurrent allelic deletions on chromosome 15 have been reported in HNSCC (51). This abnormality was found in one oral cavity carcinoma, one pharyngeal and three laryngeal carcinomas. All of these patients died of their disease within 8 to 42 months after surgery.

Loss of chromosome Y and a gain of chromosome 8 were the most frequent chromosomal alterations in our samples; however, these abnormalities were not associated with clinical and histopathologic variables. The role of the Y chromosome in tumorigenesis has been somewhat controversial. Several groups have reported a loss of Y chromosome in leukemia, lymphoma, and solid tumors (4). In HNSCC, Jin et al. (6) reported loss of the Y chromosome in 20 out of 21 cases, 14 of which showed aneusomy Y as the only chromosomal alteration. Most studies of hematologic diseases have shown a loss of the Y chromosome in both normal and malignant marrows, with the frequency showing an age-dependent increase (52).

Although there is no evidence for the existence of a tumor suppressor gene on the Y chromosome, several genes mapped to this chromosome have been implicated in disease progression (53–55). Recent studies have shown that the Y chromosome contains more functioning genes than previously thought (56, 57). As expected, older patients (>50 years) were most frequently affected in our study (44 out of 58 male patients). Chromosome Y aneusomy was absent in 23 patients ages >50 years. We detected a loss of chromosome Y in 26 out of 58 (45%) head and neck carcinomas, 24 of which were T3-T4; 14 of 26 cases showed involvement of the lymph nodes and 16 of 26 died of the disease. Five patients ages <50 years showed loss of chromosome Y; four of them were T4 stage and two died of the disease. Although these data were not statistically significant, they showed a possible association between loss of the Y chromosome and a poor outcome.

Chromosome 8 overrepresentation, the second most common copy number gain, was detected in 11 cases, including pharyngeal, laryngeal, and oral carcinomas. A gain of chromosome 8 is a frequent alteration in neoplasia and has been detected in a wide variety of tumor types, particularly HNSCC (4). Huang et al. (58) reported that +8q2 was the second most important early chromosomal event in all subtypes of HNSCC. Recently, Silva Veiga et al. (15) detected gains and/or amplifications at 8q23.2 in 19/19 HNSCC cases by FISH and suggested that an increase in copy number at 8q23.2 was a potential early marker in HNSCC.

In summary, we found significant associations between trisomies of chromosomes 10 and 20 and monosomies of chromosomes 15 and 22 suggesting that these abnormalities may be genetic biomarkers for a poor prognosis and an increased risk of death in these patients. These genetic markers could be useful for distinguishing among patients with similar clinical and histopathologic characteristics, but distinct probabilities of survival. In addition, chromosome 8 gains could be involved in early stages of the disease. Our findings support the idea that karyotypic variations occur among HNSCC arising from different anatomic sites. Comprehensive molecular profiling methods of these sites should help to identify the multiple target genes involved in head and neck carcinogenesis.

We thank Dr. Heather McDermid (Department of Biological Sciences, University of Alberto, Edmonton, Alberta, Canada) for supplying the PAC134J15 used here and Paula Marrano and Zong M. Zhang for their excellent technical assistance.