Molecular Markers and Long-Term Recurrences in a Large Cohort of Patients with Sporadic Adrenocortical Tumors

INTRODUCTION

ACC 3 is a tumor with an extremely poor prognosis, the survival rate being 20% at 5 years. Although the most common clinical presentation of patients with ACC is that of Cushing’s syndrome, with or without associated virilism, a high proportion of patients (30–40%) already have metastatic disease at the time of diagnosis (1 , 2) . The probability of long-term control is highest if surgery is performed at an early stage of the disease. However, it is very difficult to distinguish between benign and malignant tumors at this stage. A tumor is unequivocally malignant only if there is local and/or regional invasion, and/or distant metastases (McFarlane stages III and IV). In cases of purely localized disease (stages I and II), the pathological examination is often unable to provide an unambiguous answer and, in ∼30% of localized tumors, gives no indication at all. Estimation of the likelihood of malignancy is still based on ill-defined parameters, and there is a crucial need for reliable prognostic markers to assist clinicians in management decisions, which may include the administration of adjuvant chemotherapy after surgery.

Recent progress has shed some light on the biology of adrenocortical tumors. Clonal analysis and comparative genomic hybridization have provided evidence for a multistep process of tumorigenesis, with sequential progression from normal to adenomatous and eventually to malignant cells (3, 4, 5, 6) . It has been shown that discrete genetic markers are strongly associated with the malignant phenotype. Abnormalities in the imprinted 11p15 region (involving maternal 11p15 LOH with duplication of the active IGF-II paternal allele and resulting in strong overexpression of the IGF-II gene; Refs. 7, 8, 9 ), 17p13 LOH (10 , 11) , 2p16 LOH (6) , and 11q13 LOH (6 , 12, 13, 14) are all highly specific for malignant tumors.

This study evaluated the prognostic value of tumor molecular markers at the 11p15 and 17p13 loci in a single series of 96 patients prospectively followed-up after resection of a localized adrenocortical tumor. We show that, in comparison with classical pathological assessment, molecular markers have independent prognostic value with LOH at 17p13 giving the highest relative risk of recurrence. We propose that these molecular markers should be routinely evaluated for the management of adrenocortical tumors.

PATIENTS AND METHODS

Patients

Between 1987 and 1998, sporadic adrenocortical tumor (other than Conn’s adenoma) was diagnosed in 127 adult patients referred to our departments. We excluded 10 patients with malignant adrenocortical tumors referred for recurrence. Three patients (with stage I tumor) were lost to follow-up. The remaining 114 patients were followed until the date of their death, their last examination, or the end of the follow-up period (June 1999). For patients without recurrence, the minimal follow-up period was 12 months.

Clinical data and hormonal status were evaluated as described previously (15) . The stage of the tumor was assessed according to the McFarlane classification (16) . Pathological data were assessed according to Weiss criteria (17 , 18) ; for each patient, a Weiss score (0 to 9) was determined according to the presence or absence of the following nine histological features: high mitotic rate, atypical mitoses, high nuclear grade, low percentage of clear cells, necrosis, diffuse architecture of tumor, capsular invasion, sinusoidal invasion, and venous invasion.

Informed consent for analysis of leukocyte and tumor DNA and access to the collected information was obtained from all of the patients in accordance with national ethical rules.

This study was prospective, and the patients were managed in a uniform and centralized way within the framework of the Comete network.

For postoperative follow-up, patients were examined twice a year after operation for 2 years and annually thereafter. Hormonal evaluation, chest X-rays, and computerized tomography scans of the abdomen were carried out at each evaluation.

Methods

Tumor fragments obtained during surgery were immediately frozen in liquid nitrogen and stored at −80°C until DNA and RNA extraction. Control leukocyte DNA for reference was available for all of the patients.

DNA and RNA Preparation.

Nucleic acids were prepared from leukocytes and surgically removed adrenocortical tumors as described previously (8) .

17p13 Allelic Status.

17p13 allelic status was determined by Southern blotting using two different probes: the pYNZ22 probe (D17S5) mapping to 17p13.3 and recognizing a TaqI multiallelic polymorphism (19) and the pBHP53 genomic p53 probe (17p13.1) recognizing a BamHI biallelic polymorphism (20) . It was also assessed by PCR analysis with primers flanking an AccII polymorphism in the fourth exon of the human p53 gene (amino acid 72; Ref. 21 ). Genomic leukocyte and tumor DNA was amplified as described previously, digested with AccII, and subjected to electrophoresis in a 2% agarose gel.

11p15 Allelic Status.

11p15 allelic status was investigated by Southern blotting as described previously (9) .

To determine 17p13 and 11p15 allelic status, the restriction pattern of tumor DNA was compared in each case with that for leukocyte DNA.

IGF-II mRNA Tumor Content.

Tumor IGF-II mRNA content was evaluated by dot-blot analysis as described previously (9) and compared with normal control adrenal gland mRNA content. Normal control adrenal gland RNA was obtained from glands surgically removed during nephrectomy for kidney cancer.

Statistical Analysis

The primary study outcome was disease-free survival in patients with local disease at initial operation (McFarlane stages I and II). Disease-free survival was the time in months between initial operation and documented recurrence or to the end of the study (June 1999). Survival was censored if the patient did not exhibit recurrence or died from other causes.

Survival curves were constructed using the method of Kaplan-Meier and were compared and tested for statistical significance by log rank tests (univariate analysis).

Cox’s proportional hazards regression models were used to identify the factors independently predicting prognosis (multivariate analysis).

The χ² and Mann-Whitney tests were used to compare groups for noncensored qualitative and quantitative variables, respectively.

The threshold for significance was P ≤ 0.05.

The calculations were performed using the SAS package (SAS Institute, Inc., Cary, NC).

RESULTS

Features at Presentation.

Patient characteristics are summarized in Table 1 ⇓ . The patients were 18–81 years old. The ratio of women:men was 3:2. The adrenal tumor was in the right adrenal gland in 66 patients and in the left in 48. Sixty-eight percent of the patients (n = 77) were referred for endocrine symptoms, and 86% (n = 98) exhibited an abnormal hormonal profile.

Ninety-six patients had tumors restricted to the adrenal gland: 60 patients had tumors ≤5 cm in size (stage I), and 36 patients had tumors >5 cm in size (stage II). Eighteen patients had malignant tumors at diagnosis: 1 patient had a tumor confined to the adrenal gland with involvement of local lymph nodes (stage III), and 17 patients had stage IV tumors. Six patients had tumors invading adjacent organs (kidney, 1; veins, 5), and 11 patients had distant metastases. One hundred three patients (90%) underwent curative surgery.

All 18 of the patients with obviously malignant tumors had a Weiss score of 4 or more, and 82 (85%) patients with localized tumors had a Weiss score of 3 or less.

Molecular Abnormalities.

Most patients (98%) were informative for at least one 11p15 marker, and 78% patients were informative for at least one of the three 17p13 markers.

The frequencies of 17p13 LOH, 11p15 LOH, and overexpression of the IGF-II gene are shown in Table 1 ⇓ . Twenty-three of 34 tumors with 17p13 LOH displayed complete LOH and 11 displayed mosaicism. Thirty-five of 47 tumors with 11p15 LOH showed complete loss of one allele and duplication of the remaining allele and 12 showed mosaicism. Overexpression of the IGF-II gene resulted in very high IGF-II mRNA levels (10–1000 × those in normal adrenal tissue; Fig. 1 ⇓ ).

Fig. 2 ⇓ shows the frequency of these three molecular defects (17p13 LOH, 11p15 LOH, and overexpression of the IGF-II gene) according to tumor stage (Fig. 2A) ⇓ and Weiss score (Fig. 2B) ⇓ . Molecular abnormalities were frequent in stage III and IV tumors, which were obviously malignant, and were also frequent in stage II tumors, occurring in 63–69% of tumors (Fig. 2A) ⇓ . According to pathology data (Fig. 2B) ⇓ , a high frequency of molecular abnormalities was already observed in tumors with a Weiss score of 2.

Disease-free Survival Analysis.

We carried out a prognostic study for the 96 patients diagnosed with localized adrenal disease. In this group, the median duration of follow-up was 44 months (range, 5 months-11.5 years). The Kaplan-Meier estimate of disease-free survival was 87% at 5 years (Fig. 3A) ⇓ . During the follow-up period, 4 patients died of unrelated cause: pancreatic cancer (1 case), amyloidosis (1 case), and stroke (2 cases).

Eleven (12%) patients displayed recurrence 2–35 months after initial surgery (median = 10 months; Table 2 ⇓ ). Three had abdominal recurrence, and the others exhibited metastatic recurrence. Four of these patients had a Weiss score of 3. Only 1 of the 11 patients had no molecular abnormalities.

In univariate analysis, disease-free survival was not linked to age or sex. The absence of secretion or the presence of a mixed secretion (most often glucocorticoid and androgen secretion) was associated with a poor prognosis [relative risk (95% CI), 3.4 (1–11.7) and 4.6 (1.35–15.9), respectively; Table 3 ⇓] . 17p13 LOH was found to be a strong predictor of shorter disease-free survival [relative risk (95% CI), 27.3 (3.5–213.6)] as were 11p15 LOH [relative risk (95% CI), 9.4 (2–43.5)] and overexpression of the IGF-II gene [relative risk (95% CI), 4.8 (1.4–16.5)]. Univariate analysis also showed both tumor size [≤5 cm versus >5; relative risk (95% CI), 17.9 (2.3–140)] and, particularly, histological grade [Weiss score ≤3 versus ≥4; relative risk (95% CI): 15.15 (4.4–52.1)] to be significant predictors of disease-free survival (Table 3) ⇓ .

The prognostic value of these factors is illustrated by Kaplan-Meier curves of disease-free survival (Fig. 3, B–F) ⇓ .

Multivariate analysis was carried out for 73 patients (patients informative for the three molecular markers). The predictive power of the three molecular markers (17p13 LOH, 11p15 LOH, and overexpression of the IGF-II gene) was compared with that of other clinical and biological factors. Proportional hazards regression analysis showed that 17p13 LOH was the only molecular marker that acted as an independent factor predictive of disease-free survival (P = 0.009; relative risk, 21.5). χ² analysis also showed a strong correlation between 17p13 LOH and 11p15 LOH or overexpression of the IGF-II gene (P < 0.0001). Of the clinical and biological factors, older age (P = 0.04; relative risk, 1.06/year) and higher histological grade (P = 0.04; relative risk, 4.2) were also independent predictive factors (Table 4) ⇓ .

DISCUSSION

One major problem in the management of adrenocortical tumors is distinguishing localized tumors with malignant behavior from those that are benign. No single parameter has been identified that detects malignancy with certainty, and the likelihood of malignancy is still estimated on ill-defined parameters (1 , 2) . The major limitation of prognostic studies is the rarity of ACCs. Only a few studies have been devoted to a comparison of benign and malignant adrenocortical tumors. Weiss et al. (17) assessed the value of nine histological criteria for analyzing the behavior of adrenocortical tumors and showed that tumors with four or more of these criteria almost always recurred or metastasized. However, this study was undertaken in a small series, consisting of only 43 patients. In a subsequent study, Weiss et al. (18) showed that a high mitotic rate was the only one of these pathological criteria associated with poor survival.

During the last decade, our understanding of the mechanisms of adrenal tumorigenesis has advanced, and a variety of genetic abnormalities have been described, most associated with the malignant phenotype. In particular, abnormalities of the 11p15 region and LOH at the 17p13 locus are highly specific to malignant adrenocortical tumors. We followed up, in a prospective study, a large series of patients managed in a uniform and centralized way. We determined 17p13 and 11p15 allelic status and IGF-II mRNA content in sporadic adrenocortical tumors from 114 patients. We found that these molecular abnormalities were frequent in tumors that were clearly malignant (stages III and IV), occurring in ∼85% of such tumors.

The final aim of the study was to evaluate the prognostic value of these markers in localized tumors, particularly tumors assigned a low grade on the basis of surgical (McFarlane) and pathological (Weiss) criteria. We confirmed that pathological variables such as tumor size and histological grade were strong predictors of recurrence. Indeed, most patients who relapsed had tumors >5 cm in diameter and a Weiss score of ≥4. Although powerful, the multifactorial analysis proposed by Weiss (17 , 18) may be difficult to apply to individual cases and requires trained pathologists. The score is not totally reliable and may differ from one area to another in the same tumor. Therefore, other diagnostic markers of malignancy are required. The three molecular markers tested were found to be strong predictors of disease-free survival in univariate analysis. The best predictor, strongly associated with a risk of relapse and with shorter disease-free survival, was 17p13 LOH. This was also the only molecular marker of independent prognostic significance in a multivariate analysis taking into account biological and pathological data.

This study has several limitations. Like Weiss score, these prognostic parameters are only available after surgery. However, it is possible to consider the use of this type of molecular evaluation for biopsy samples and for identifying tumors at risk of malignancy. Although this study is the largest available prospective study for a relatively rare disease, the frequency of events (recurrences in 11 cases) was low, and our results require confirmation in a larger series of patients.

The strong negative prognostic impact of 17p13 LOH suggests that the 17p13 locus probably includes specific target genes, the inactivation of which contributes to tumor progression. In this study, 17p13 allelic status was assessed at two loci: 17p13.3 (D17S5 marker) and 17p13.1 (p53 markers). Unlike other tumor models such as breast cancer (22 , 23) , in adrenocortical tumors, if reduction to homozygosity was detected with the 17p13.3 marker and patients were informative with one of the two p53 gene analysis systems, the 17p13 LOH also involved the p53 gene. However, p53 mutations (24, 25, 26, 27) 4 occur in only ∼25% of ACCs. This difference between p53 mutations and 17p13 LOH frequencies suggests that another tumor suppressor gene in this region is important for adrenal tumorigenesis. The allelic loss of 17p13 in adrenal tumors should now be investigated with a set of well-mapped markers to define accurately the smallest common region, which can then be used to test a number of recently cloned putative tumor suppressor genes such as the HIC-1, OVCA1, OVCA2, CRK, and ABR genes (28, 29, 30, 31) .

The other two molecular markers studied here concern the 11p15 region. This region is subject to parental imprinting. We characterized previously the 11p15 LOH as a parental isodisomy with loss of the maternal allele and duplication of the paternal allele (8 , 9) . This structural abnormality often leads to overexpression of the IGF-II gene, which is normally only expressed from the paternal allele (7, 8, 9) . Overexpression of the IGF-II gene is a common feature of various cancers (32 , 33) , and we have shown that this growth factor is involved in the proliferation of malignant adrenal H295R cells (34) . The greater prognostic value of 11p15 LOH than of overexpression of the IGF-II gene suggests that the loss of expression of genes normally expressed from the maternal allele is important in adrenal tumorigenesis. Indeed, most genes in the 11p15 region are maternally expressed and encode products with antiproliferative function. One of these products, p57KIP2, inhibits G₁ cyclin-cyclin-dependent kinase complexes (35 , 36) . H19 could act as a tumor suppressor gene (37) , TSSC3 favors apoptosis (38) , and TSSC5, the function of which is still unknown, has been shown to be mutated in lung cancers (39) . Loss of H19 gene expression has been described in malignant adrenocortical tumors (9) . The p57KIP2 gene has also been evaluated in adrenal tumors and has been shown to be down-regulated in malignant tumors (40 , 41) , resulting in an increase in the kinase activity of G₁ cyclin-cyclin-dependent kinase complexes (41) .

In conclusion, this prospective study establishes the prognostic value of molecular markers (particularly 17p13 allelic status) in the management of localized adrenocortical tumors. After complete resection of a localized tumor with molecular abnormalities, the patient should be followed in more detail, whatever the pathological evaluation. Whether adjuvant therapies such as mitotane or chemotherapy are of benefit and should be used in such patients remains to be tested in a prospective trial.