Endocrine Abstracts

Introduction: Genetic of pituitary tumors (PT) is incompletely understood. The aim of this study was to search for somatic copy number variations (CNV) in pituitary tumors and evaluate their prognostic impact on tumor recurrence.

Methods: CGH array analysis (Agilent SurePrint G3 Human CGH+SNP Microarray 4×180 K) was performed on 196 fresh-frozen PT (67 gonadotroph, 31 corticotroph, 38 prolactinomas, 60 somatotroph). PT were classified according to the five-tiered classification grading based on invasion on magnetic resonance imaging, immunocytochemical profile, Ki-67, mitotic index, and p53 positivity. ACGH data were analyzed after centralization, normalization and circular binary segmentation steps. Gains and losses were considered when Log2 ratio > 0.14 and <−0.15 respectively. Hierarchical clustering using Jaccard index was first performed considering altered vs unaltered probes. Effect of quantity of alterations on recurrence was then studied using logistic regression models.

Results: 124 patients (63%) presented recurrence during the 5 years of follow-up. Most of PT were macro-adenomas (182) and 84 were invasive PT. Sex ratio males/females was 1.3:1. 84 PT (43%) presented a highly-disrupted genome (>5% of altered probes). Clustering could classify PT according to the tumor type: the cluster of PT with rare alterations gathered gonadotroph tumors together, while the cluster of PT with numerous alterations was heterogeneous and did not separate the prolactinomas from corticotroph and somatotroph PT. Indeed, the quantity of altered probes was higher in prolactinomas (median=38% of probes, min=0%, max=96%) compared to other groups: corticotroph (median=12%, min=0%, max=76%), somatotroph (median=4%, min=0%, max=99%), and gonadotroph which presented rare alterations (median=0%, min=0%, max=22%). Alterations in prolactinomas were preferentially gains (median=35%, min=0%, max=96%) than deletions (median=0%, min=0%, max=24%). When the whole cohort was considered, we were not able to identify a common alteration shared by the different types. We found that the quantity of alterations could not predict recurrence (P value=0.52) whereas age at diagnosis and clinico-pathological grades could (P value=0.0006, 0.0003 and 0.00007 for age, grade 2a and 2b respectively).

Conclusions: aCGH analysis of PT showed many CNV which could concern the entire genome in some prolactinomas or somatotroph tumors. The CNV occurrence was highly dependent on tumor type but did not predict recurrence within 5 years of follow-up. PT, except gonadotroph, are characterized by high genomic instability.

Grant PITUIGENE ClinicalTrials.gov Identifier: NCT01903967