Context: Pituitary corticotroph adenomas are rare tumors that can be associated with excess adrenocorticotropic hormone (ACTH) and adrenal cortisol production, resulting in the clinically debilitating endocrine condition Cushing disease. A subset of corticotroph tumors behave aggressively, and genomic drivers behind the development of these tumors are largely unknown.

Objective: To investigate genomic drivers of corticotroph tumors at risk for aggressive behavior.

Design: Whole-exome sequencing of patient-matched corticotroph tumor and normal DNA from a patient cohort enriched for tumors at risk for aggressive behavior.

Setting: Tertiary care center.

Patients: 27 corticotroph tumors from 22 patients analyzed. 12 tumors were macroadenomas, of which 6 were silent ACTH tumors, 2 were Crooke’s cell tumors, and 1 was a corticotroph carcinoma.

Intervention: Whole-exome sequencing.

Main outcome measure: Somatic mutation genomic biomarkers.

Results: We found recurrent somatic mutations in USP8 and TP53 genes, both with higher allelic fractions than other somatic mutations. These mutations were mutually exclusive, with TP53 mutations occurring only in USP8-wildtype (WT) tumors, indicating they may be independent driver genes. USP8-WT tumors were characterized by extensive somatic copy number variation compared to USP8-mutated tumors. Independent of molecular driver status, we found an association between invasiveness, macroadenomas, and aneuploidy.

Conclusions: Our data suggest that corticotroph tumors may be categorized into a USP8-mutated, genome-stable subtype versus a USP8-WT, genome-disrupted subtype, the latter of which has a TP53-mutated subtype with high level of chromosome instability. These findings could help identify high risk corticotroph tumors, namely those with widespread CNV, that may need closer monitoring and more aggressive treatment.

See "Methods" section of the main manuscript and these captions for the submitted files:

Extended data 1: Clinical and pathological characteristics of 22 patients in cohort, Microsoft Excel (XLSX) file.

Individual patient demographic, clinical, and pathology characteristics for all patients in cohort.

Extended data 2: Detailed information on 27 specimens in cohort, Microsoft Excel (XLSX) file.

Per-specimen information for all specimens in cohort. WES libraries were multiplexed in the ratios given in “Multiplexing strategy” and “Samples per lane” columns and run on Illumina flowcells in High Output (HO) mode on the given number of lanes.  NGS QC statistics are provided.  Manual tumor purity estimates and supporting MSI QC data are also provided.

Extended Data 3: List of somatic mutations, Microsoft Excel (XLSX) file.

Somatic mutation calls and their QC data, including allelic fractions, for all specimens in the cohort. These have been deposited in COSMIC as paper identifier COSP48424. The tab “all” shows all QC-passing somatic mutation calls, whereas the tab “protein_altering_only” shows only the subset of those that are predicted to alter the protein-coding sequence of an mRNA, i.e. missense and nonsense SNVs, splice site variants, and indels overlapping coding exons. For each allele, a single gene and a single transcript isoform is selected in which to report the “c.” and “p.” nomenclature from HGVS (https://varnomen.hgvs.org/). The tab “legend” defines all the columns in the data tables. Patient 0301 is not included in this list because her somatic mutations have been previously published using a more comprehensive, multi-assay methodology (https://www.ncbi.nlm.nih.gov/pubmed/28487882) and already deposited in COSMIC (COSMIC sample ID COSS2629282, paper identifier COSP42647, https://cancer.sanger.ac.uk/cosmic/study/overview?paper_id=42647).  Coordinates are given in the human_g1k_v37 reference genome.

Extended Data 4: Genome-wide somatic CNV segment and QC diagnostic plots from SAAS-CNV for 27 specimens in cohort, Microsoft PowerPoint (PPTX) file.

Diagnostic plots and visualizations of the genome-wide copy-number profiles, as output by the SAAS-CNV tool (the plots are described in https://pubmed.ncbi.nlm.nih.gov/30039361/). The derivation of the tetraploidy claim from SAAS-CNV data is given on Slide 55 as a detailed example for patient 0310 tumor A. Other slides apply clusters defined on Slide 55 to data from other tumors to support tetraploidy or potential tetraploidy. The values “log2mBAF” and “log2ratio” are unadjusted values that come from columns “log2mBAF.Median” and “log2ratio.Median”, respectively, in Extended Data 7. The gray dashed baselines inferred for each sample in the SAAS-CNV QC diagnostic plots come from columns “log2ratio.base.Median” and “log2mBAF.base.Median” of Extended Data 7.

Extended Data 5: Somatic CNV classification worksheet, Microsoft Excel (XLSX) file.

Worksheet showing our manual classification of each chromosome arm into a gain (G), loss (L), or copy-neutral loss of heterozygosity (H) state for each of the 27 tumors in our cohort, on the basis of SAAS-CNV data from Extended Data 4. See the “legend” tab of the XLSX file for a more detailed legend. This worksheet also gives the final classification of each tumor as “highly aneuploid” (yes or no) and its level of “chromosome/arm-level sCNV” (none/low, medium, high), as shown in Figure 1. sCNV categories in last column correspond to categories highlighted in Figure 3.

Extended Data 6: Mutational signatures, Microsoft PowerPoint (PPTX) file.

(A) Single base substitution mutational signature decomposition result. Samples included here are samples with SNV count (exome region only) >= 15 at AF cutoff of >=10%, and the signature reference was determined using bootstrap approach (see Methods and Extended Data 6g).

(B) Transcriptional bias analysis of samples used in the mutational signature decomposition analysis (see Methods).

(C) Relative substitution frequencies in the trinucleotide context for radiation-naïve and radiation-treated patients, as well as for COSMIC signatures 12 and 16 from COSMIC catalog v2 vs v3.

(D) Relative substitution frequencies in the trinucleotide context for each patient.

(E) Bootstrap result of mutational signature fitting for each tumor using full COSMIC v2 signature reference (boostrap N = 1000). Each dot refers to the estimation of signature weight from a single bootstrap. The red dot refers to the signature weight estimation of the original mutational profile (without bootstrap). The two blue crosses for each signature refer to the 10% and 90% quantiles from the bootstrap signature weight distributions, thus representing p = 0.1 (one-sided). The numbers in the parentheses next to patient and specimen ID refer to the somatic SNV counts (exome region only) with AF >= 10%.

(F) Bootstrap result of mutational signature fitting for “naive” and “radiation” samples using full COSMIC v2 signature reference (boostrap N = 1000). The plotting conventions are same as in Extended Data 6e.

(G) A table describing, for the full set of COSMIC v2 N=30 mutational signatures, the number of samples with signature exposure below different cutoffs (at p = 0.1, one-sided) out of the 9 tumor samples analyzed for mutational signatures. The signatures highlighted in yellow were the ones with sample number <= 6 at exposure cutoff of below 0.2, which we chose to include in the final signature reference for mutational signature fitting (see Methods).

Extended Data 7: Raw, machine-readable somatic CNV segment calls from SAAS-CNV for 27 specimens in cohort, ZIP file of per-sample segment call TSVs.

Raw, machine-readable TSV data as output by SAAS-CNV and used for all sCNV analysis, including all plots in Figure 3 and Extended Data 4. All segments are provided without filtering. Segment coordinates are in the human_g1k_v37 genome reference (except prefix “chr” is added) and are given by the columns “chr”, “posStart”, “posEnd”. The column “CNV” shows the SAAS-CNV classification of the segment that is shown in Extended Data 4 plots and in Figure 3, except in the latter, “normal” and “undecided” segments aren’t shown. The “copy ratio” in Figure 3 is from column “log2ratio.Median.adj”, i.e. after adjustment with respect to sample-specific inferred baselines that are shown in Extended Data 4 diagnostic QC plots.