Prognostic Significance of Major Histocompatibility Complex Class II Expression in Pediatric Adrenocortical Tumors: A St. Jude and Children's Oncology Group Study

Adrenocortical tumors are rare in children accounting for approximately 0.2% of all pediatric malignancies (1). Currently, prognosis is based on tumor size, surgical resectability, and metastasis at diagnosis (2–5). Adrenocortical adenomas carry a good prognosis; however, the outcome is variable for the 80% of cases classified as carcinomas. Many small, completely excised adrenocortical carcinomas later relapse, whereas others are cured (2–5).To determine the clinical behavior of pediatric resectability according to histologic criteria is often challenging.

Previous studies have evaluated MHC class II expression as a prognostic marker for pediatric adrenocortical tumor (6–8). We previously demonstrated that the expression of specific MHC class II genes was higher in pediatric adrenocortical adenomas compared with carcinomas (6). Consistent with these findings, Leite and colleagues reported that lower expression of HLA-DRA, HLA-DPA1, and HLA-DPB1 genes may contribute to more aggressive disease (7). However, Magro and colleagues concluded that MHC class II expression did not discriminate between benign and malignant pediatric adrenocortical tumors (8). Therefore, the relevance of MHC class II as an indicator of outcome for pediatric adrenocortical tumor remains to be established.

MHC class II markers are expressed mainly by hematopoietic cells of the immune system but have been observed in normal adult adrenal cortex cells with weak immunostaining in the inner fasciculate zone and strong staining in the reticularis zone (9). Conversely, fetal adrenocortical cells of both zones were invariably negative (9). Whether MHC class II gene expression in pediatric adrenocortical tumor arises from the adrenocortical tumor or infiltrating immune cells is uncertain.

Here, we evaluated: (i) the expression of MHC class II genes; (ii) the cell of origin expressing MHC class II; and (iii) the impact of MHC class II expression on clinical outcome in a large group of uniformly treated pediatric adrenocortical tumors. High MHC class II expression, in particular HLA-DPA1, was found associated with tumor-infiltrating immune cells and a more favorable progression-free survival (PFS).

The study included 52 participants in the St. Jude Children's Research Hospital (Memphis, TN) International Pediatric Adrenocortical Tumor Registry (IPACTR; ClinicalTrials.gov, ID number NCT00700414; ref. 2). The tumors were of any histology (adenomas, carcinomas, and undefined histology) and were not uniformly treated. The IPACTR patients typically underwent surgery and chemotherapy for advanced stage disease (2–4). A second group, comprised of 34 subjects with histologically defined adrenocortical carcinomas, was uniformly treated on the Children's Oncology Group (COG) ARAR0332 protocol (ClinicalTrials.gov, ID number NCT00304070; ref. 4), which is a nonrandomized, single-arm, prospective study. Classification of adenoma, carcinoma, or tumors of undetermined histology in the first cohort (IPACTR) was based on the modified criteria of Weiss and colleagues (10–12) by two independent pathologists. In the COG cohort, only cases with diagnosis of carcinoma, centrally reviewed by one of the authors (J.M. Hicks), were enrolled on this study. The criteria for carcinoma were based on a malignancy scoring system of 9 gross and histopathologic features, including tumor weight (>400 g), tumor size (>10.5 cm), extension into periadrenal soft tissue and/or adjacent organs, vena cava invasion, venous invasion, capsular invasion, mitotic index (>15 per 20 high power fields), and atypical mitotic figures, as previously defined by Wieneke and colleagues (13). Each item present is given a score of 1. The identification of 4 or more of the above features is considered to be malignant. If there is metastatic disease to lymph nodes or distant sites, the tumor is considered malignant as well. Written informed consent was obtained from parents or legal guardians in all cases. This study was approved by the St. Jude Children's Research Hospital Institutional Review Board and by the Cancer Therapy Evaluation Program of the NCI (Rockville, MD). Selected demographic and clinical features of the 86 patients are summarized in Table 1; detailed information is provided in Supplementary Table S1.

Gene expression data for pediatric adrenocortical tumors have been deposited in the National Center for Biotechnology Information, GEO profiles under accession code (COG cases, GSE76019; IPACTR cases, GSE76021).

Total RNA from tumor samples was used for RNA probe preparation and hybridization to Affymetrix Human Genome U133 Plus 2.0 or HT HG-U133+PM gene chip arrays according to the manufacturer's guidelines (Affymetrix; Supplementary Data, Section 1). Expressions of MHC class II and selected cell surface marker genes (Supplementary Table S2) were evaluated in the two cohorts of pediatric adrenocortical tumors (Supplementary Table S1).

Formalin-fixed, paraffin-embedded tumor sections (4 μm) were stained for HLA-DPA1, HLA-DRA, and cell-surface markers (CD3, CD4, CD8, CD68, and CD163) using BenchMark XT (Ventana Medical Systems) or BOND-MAX (Leica Microsystems) automated stainers and reagents as listed in Supplementary Table S3 according to the manufacturer's instructions.

For each microarray, we used the Bioconductor affy software (www.bioconductor.org) to compute the mean values of robust multiarray average–normalized log2 scale signals across all probe sets annotated to MHC class II genes (14, 15) for comparison of total MHC class II gene expression with histology and prognosis. The positive quantile transformation (16) was used to further normalize microarray data for comparison of individual probe set values with histology and prognosis. We used a discrimination gap statistic, as described in Supplementary Data (Section 2), to select five individual probe sets in the IPACTR cohort, whose expression pattern most clearly distinguished adenomas versus carcinomas, for further evaluation in the COG cohort. For each of the five selected probe sets, a single-predictor Cox regression analysis was used to evaluate the univariate association of expression level with the estimate of PFS in the COG samples. These single-predictor Cox regression models were used to perform a one-sided test of the association of expression with PFS estimate at the P = 0.05/5 = 0.01 level (Bonferroni adjustment for multiple comparisons). Comparisons that showed a significant association (with the Bonferroni correction) were used in a two-predictor Cox regression model to compare expression with the PFS estimate after adjustment for tumor stage or tumor weight. PFS was defined as the time elapsed between protocol enrollment and the date of disease progression, death, or most recent follow-up. The Kaplan–Meier method was used to estimate PFS probability as a function of time. All statistical analyses were performed using R software (www.r-project.org), and all statistical tests are two sided except where otherwise stated.

The age at diagnosis of the 63 females and 23 males was 3.1 to 207.6 months (median, 38 months). Most patients presented with virilization (n = 47), Cushing syndrome (n = 7), aldosteronism (n = 2), or mixed hormonal secretion (n = 16). Fourteen patients had nonfunctional tumors. Disease was staged according to the COG ARAR0332 protocol (4): stage I (n = 28), II (n = 18), III (n = 28), and IV (n = 12). Primary tumor samples were obtained in all cases, with the exception of 3 biopsy samples (ACT080, ACT084, and ACT086) from COG patients who had stage IV disease, as the tumors were considered inoperable. The median follow-up time among progression-free patients was 52.3 months (range, 0.1–197.8 months). Clinical data for all patients are summarized in Table 1 and Supplementary Table S1.

The median expression of 28 MHC class II probe sets was significantly higher in adenomas than in carcinomas (P = 4.8 × 10⁻⁶, Wilcoxon rank-sum test) in the IPACTR cohort (Fig. 1). In addition, greater MHC class II expression (Fig. 2A) was associated with better PFS in the uniformly treated pediatric adrenocortical carcinomas in the COG cohort (Fig. 2B; P = 0.003, Cox regression). The 3-year PFS of subjects with less than median MHC class II expression was 47.1% [95% confidence interval (CI), 28.4%–77.9%)], whereas that of subjects with greater than median MHC class II expression was 88.2% (95% CI, 74.2%–100%; Fig. 2B).

To identify the individual MHC class II probe sets associated with histology and PFS, we computed a discriminating gap size (DGS) criterion (16) for differential gene expression in adenomas and carcinomas in the IPACTR cohort. The DGS criterion can also identify a gap in the expression distribution that may distinguish between two distinct biological processes (see Supplementary Data, Section 2, for details). The probe sets were ranked according to the value of this DGS criterion, and the top 5 were selected (HLA-DPA1, HLA-DQA1/2, HLA-DRA, HLA-DRB1, and HLA-DQB1; Supplementary Fig. S1).

Expression (< vs. ≥ median) of each of the five probe sets with greatest DGS was then compared with the PFS in the COG cohort (Fig. 3). We performed one-sided statistical tests to determine whether greater expression of each probe set was associated with improved PFS, with the Bonferroni adjustment for multiple tests. Greater expression of four of the five probe sets was significantly associated with greater PFS (one-sided raw P < 0.01 or 0.05/5; Supplementary Table S4).

In particular, expression of HLA-DPA1 (probe set 211991_s_at) was significantly associated with tumor weight (P < 0.0001) and disease stage (P = 0.032) but not with age (P = 0.397), gender (P = 0.179), endocrine syndrome (P = 0.071), surgery (P = 0.864), or chemotherapy (P = 0.106; Supplementary Table S5).

The prognostic value of HLA-DPA1 (probe set 211991_s_at) expression after adjustment for disease stage in the uniformly treated COG cohort was evaluated. As expected, advanced stage was associated with poor PFS. Five of 7 stage IV patients had progressive disease within one year and a sixth patient had disease progression within two years (Fig. 4A). To date, none of the 15 stage I–III patients with greater than median expression of HLA-DPA1 have died or experienced progressive disease, whereas the 12 stage I–III patients with less than median expression of HLA-DPA1 had a 3-year PFS of 58.3% (95% CI, 36.2%–94.1%; Fig. 4B).

Archival tumor sections of 26 IPACTR cases (15 carcinomas, 6 adenomas, 5 of undetermined histology) were stained for hematoxylin and eosin (H&E) to confirm the pathologic diagnosis. Eight adrenocortical tumors (ACT005, ACT008, ACT012, ACT030, ACT031, ACT032, ACT033, and ACT034) were stained with antibodies specific to HLA-DPA1 and HLA-DRA and markers of mononuclear hematopoietic cells (CD3, CD4, CD8, CD68, and CD163; Supplementary Table S3). H&E and immunostaining results for a representative adenoma (ACT030, row a), early-stage carcinoma (ACT012, row b), and late-stage carcinoma (ACT008, row c) are shown in Fig. 5A. HLA staining was predominantly cytoplasmic, with HLA-DRA brighter than HLA-DPA1 (Fig. 5A, columns 2 and 3). MHC class II staining was observed in two morphologic cell types, one that stained intensely for HLA-DRA and HLA-DPA1 and had relatively short cytoplasmic processes and one that stained less intensely and had longer cytoplasmic processes (Fig. 5A inset, columns 2 and 3). The cells with short cytoplasmic processes were also strongly positive for CD68 (Fig. 5A, column 4) and CD163 (not shown) but less positive for CD4 (Supplementary Fig. S2), suggesting tumor-infiltrating macrophages. Other cells showing long cytoplasmic processes were negative for CD68 (Fig. 5A, column 4) and CD163 but dimly positive for CD4, consistent with dendritic cells (Supplementary Fig. S2). Staining of sequential sections of the same tissue block for CD3, CD4, and CD8 revealed both types of T lymphocytes. Approximately 5% to 10% of the infiltrating cells were T lymphocytes (CD3⁺). Moreover, many of the tumor-infiltrating cells dimly positive for CD4 in the adenoma had morphologic characteristic of histiocytes and dendritic cells (Supplementary Fig. S2).

Because of the strong association between HLA-DPA1 gene expression and outcome, we analyzed the remaining 18 IPACTR adrenocortical tumors (10 adrenocortical carcinomas, 5 adrenocortical adenomas, and 3 of undetermined histology) with anti-HLA-DPA1 antibody. HLA-DPA1 protein was absent or undetectable in the tumor cells in 15 of these cases (83%). However, three cases (ACT040, ACT047, and ACT048; 17%) showed focal HLA-DPA1 immunostaining within the adrenocortical tumor cells (Fig. 5B; Supplementary Table S6).

Greater expression of MHC class II genes, HLA-DPA1 in particular, is associated with better prognosis for pediatric adrenocortical tumors independent of histopathologic diagnosis. In most cases, MHC class II genes were expressed by tumor-infiltrating hematopoietic cells rather than adrenocortical tumor cells. These results suggest that immune surveillance, evident by elevated expression of MHC class II and the presence of tumor-associated macrophages and dendritic cells, may limit adrenocortical progression, leading to less aggressive disease.

The histologic features of ACTs often do not allow unequivocal classification as adenomas or carcinomas, and such cases are currently classified as tumors of indeterminate histology, undetermined malignant potential, or simply “adrenocortical neoplasms” (13, 17, 18). In our IPACTR series, 7 of 52 tumors (13%) had undetermined histologic findings. Three of these cases expressed high levels of MHC class II based on microarray (ACT005 and ACT006) and IHC (ACT047), which is characteristic of pediatric adrenocortical adenomas. All three patients experienced tumor spillage during surgery but did not relapse and remained disease free at the time of the last follow-up (range, 20–56 months). Three of the remaining four adrenocortical tumors of undetermined histology (ACT034, ACT039, and ACT046) expressed high levels of HLA-DPA1 and are free of disease for more than 2 years. These clinical findings are consistent with the observed high MHC class II expression and less aggressive disease.

Prognostic information is also lacking for pediatric adrenocortical tumors classified as carcinomas, as their outcome varies substantially after complete resection (3–5, 10–13). Tumor size (weight or volume) is associated with prognosis, but small adrenocortical carcinomas too often behave aggressively (early local and distant relapse), whereas large carcinomas may have an indolent course. We suggest that the expression of MHC class II genes can be used to refine the risk classification of adrenocortical tumors. Overexpression of MHC class II genes could more decisively identify patients with undetermined or carcinoma histology who are unlikely to require adjuvant chemotherapy after complete surgical resection. Like childhood adrenocortical tumors with a Weiss score of 0 or 1 (10–13, 17), those with relatively large, completely excised carcinomas (stage II–III) with high expression of MHC class II genes could be spared intensive chemotherapy. Completely resected carcinomas with low MHC class II gene expression could be treated with adjuvant chemotherapy irrespective of their size.

Overexpression of IGF2 and BUB1B is associated with poor prognosis in adult adrenocortical tumor (19, 20). In pediatric adrenocortical tumor, IGF2 is overexpressed in both adenomas and carcinomas (21) and therefore does not correlate with HLA-DPA1 levels or outcome (P > 0.10; Supplementary Table S5). BUBR1, the protein encoded by BUB1B, has a critical role in regulating the spindle assembly checkpoint and implicated in chromosome instability and cancer progression (22). Similar to adult adrenocortical tumor and other adult cancers (23, 24), overexpression of BUB1B inversely correlated with PFS in pediatric adrenocortical tumor from both cohorts (P = 0.003; Supplementary Table S5).

MHC class II molecules are surface glycoproteins on antigen-presenting cells (e.g., dendritic cells, monocyte/macrophages, and B lymphocytes) that presumably present tumor peptides to CD4⁺ T cells (25). Previous studies of pediatric adrenocortical tumor have not definitively addressed the cell of origin expressing MHC class II, although it was suggested to originate from adrenocortical tumor cells and downregulated concurrently with tumor progression (6–8). However, MHC class II is not expressed in the fetal adrenal cortex (26, 27), the putative origin of pediatric adrenocortical tumor (28), but rather, predominantly by tissue macrophages and a small population of CD45R⁺, IgM⁺ lymphoid cells (28). Our findings demonstrate that the cells expressing MHC class II genes in pediatric adrenocortical tumors, primarily adenomas, are macrophages and dendritic cells. Dendritic cells are capable of initiating innate and adaptive tumor responses (29), suggesting a mechanism for suppressing pediatric adrenocortical tumor progression. In contrast, three adrenocortical tumors (ACT40, ACT047, and ACT048) displayed focal tumor–cell expression of HLA-DPA1 in addition to the infiltrating hematopoietic cells. Two of these patients (ACT047, ACT048) were more than 7 years of age at diagnosis and presented with virilization and large tumors. Interestingly, these cases experienced tumor spillage during surgery but did not relapse during more than 20 months of follow-up. Whether tumor cells expressing MHC class II are capable of participating in regulatory immune mechanisms and have prognostic implications remains to be determined.

We have demonstrated that MHC class II gene expression in pediatric adrenocortical tumors originates from tumor-infiltrating immune cells and is associated with more favorable prognosis. Our data support the inclusion of MHC class II expression in addition to standard criteria (e.g., tumor size, mitotic activity, atypical mitoses, and other histologic features) as a tool for determining pediatric adrenocortical tumor diagnosis. Future studies will define the interaction of infiltrating hematopoietic cells with the pediatric adrenocortical tumor microenvironment, which may refine risk classification and facilitate the development of immunotherapies for these tumors.

No potential conflicts of interest were disclosed

Conception and design: E.M. Pinto, C. Rodriguez-Galindo, J.M. Hicks, R.C. Ribeiro, G.P. Zambetti

Development of methodology: E.M. Pinto, J.K. Choi, S. Pounds, J.M. Hicks, G.P. Zambetti

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.M. Pinto, C. Rodriguez-Galindo, J.M. Hicks, B.C. Figueiredo, R.C. Ribeiro, G.P. Zambetti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.M. Pinto, J.K. Choi, S. Pounds, Z. Liu, G. Neale, D. Finkelstein, J.M. Hicks, A.S. Pappo, G.P. Zambetti

Writing, review, and/or revision of the manuscript: E.M. Pinto, C. Rodriguez-Galindo, J.K. Choi, S. Pounds, J.M. Hicks, B.C. Figueiredo, R.C. Ribeiro, G.P. Zambetti

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.M. Pinto, G.P. Zambetti

Study supervision: E.M. Pinto, R.C. Ribeiro, G.P. Zambetti

We thank the study participants and the clinicians and research staff at participating institutions. We thank Jay Morris and Melanie Loyd for assistance with the expression arrays; Charlene Henry, Raven Childers, and the St. Jude Children's Research Hospital anatomic pathology staff for assistance with immunostaining studies. We thank Sharon Naron for editing the manuscript.

This work was supported by a grant from Centocor, Inc., National Institutes of Health Cancer Center support grant CA21765, and by the American Lebanese Syrian Associated Charities (ALSAC).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.