Genome-wide loss of heterozygosity predicts aggressive, treatment-refractory behavior in pituitary neuroendocrine tumors

Authors

Lin AL

Rudneva VA

Richards AL

Zhang Y

Woo HJ

Cohen M

Tisnado J

Majd N

Wardlaw SL

Page-Wilson G

Sengupta S

Chow F

Goichot B

Ozer BH

Dietrich J

Nachtigall L

Desai A

Alano T

Ogilive S

Solit DB

Bale TA

Rosenblum M

Donoghue MTA

Geer EB

Tabar V

Doi

10.1007/s00401-024-02736-8

PMID: 38758238 · DOI: 10.1007/s00401-024-02736-8 · Journal: Acta Neuropathologica (2024)

TL;DR

This study sequenced 92 pituitary neuroendocrine tumors (PitNETs) using MSK-IMPACT and whole-exome sequencing, comparing 23 aggressive, treatment-refractory tumors to 69 benign tumors. The authors found that genome-wide loss of heterozygosity (LOH) is significantly elevated in treatment-refractory PitNETs, with a recurrent chromosomal LOH pattern in corticotroph tumors involving 12 specific chromosomes. A random forest machine learning classifier using fraction of LOH achieved an accuracy of 0.88 (95% CI: 0.70–0.96) for predicting treatment-refractory behavior, outperforming TP53 mutational status alone.

Cohort & data

Retrospective cohort (n = 26): Patients with aggressive/treatment-refractory or higher-risk pituitary adenomas (PTAD); 22 of 26 (85%) were treatment-refractory (progression after surgery, medical therapy, and radiation); 10 (38%) were metastatic.
Prospective cohort (n = 66): Unselected patients presenting to MSKCC for pituitary surgery; only 1 patient met treatment-refractory criteria.
Analytic groups: 23 treatment-refractory and 69 benign PitNETs after stratification by response to radiotherapy.
Sequencing: All tumors sequenced via MSK-IMPACT; retrospective cohort also underwent whole-exome sequencing for USP8 mutation detection.
Copy-number analysis: FACETS v0.5.6 for allele-specific copy-number and LOH inference.
Validation: FISH on five patients to confirm chromosomal losses.
Dataset: ptad_msk_2024; data deposited in cBioPortal and dbGaP (phs001783).

Key findings

Treatment-refractory PitNETs had significantly higher tumor mutational burden (p = 1.3 x 10^-10) and fraction of LOH (p = 8.5 x 10^-9) compared to benign tumors (PMID:38758238).
TP53 mutations were the most common gene-level alteration in treatment-refractory tumors (12/23 treatment-refractory vs. 1/69 benign; Fisher’s exact test p = 4.2 x 10^-8); 11 of 12 were clonal.
A recurrent chromosomal LOH (rcLOH) pattern involving chromosomes 1, 2, 3, 4, 6, 10, 11, 15, 17, 18, 21, and 22 was found in 11/14 treatment-refractory corticotroph PitNETs vs. 1/14 benign corticotroph tumors (Fisher’s exact test p = 1.7 x 10^-4).
FISH confirmed that LOH events represent true chromosomal losses (monosomies/hypodiploidy) rather than copy-neutral LOH.
Tumors with TP53 mutations had higher fraction of LOH (p = 3.3 x 10^-8).
Random forest classifier identified fraction of LOH as the most important predictive feature for treatment-refractory behavior, outperforming TP53 status. A binary LOH threshold of 0.11 achieved AUC = 0.87, accuracy = 0.88 (95% CI: 0.70–0.96), sensitivity = 0.83, specificity = 0.90 on the test set.
Five treatment-refractory corticotroph tumors harbored mismatch repair gene mutations (MSH2, MSH6, MLH1), confirmed by loss of protein expression on IHC; 3 of 4 pre-treatment tumors were MSI-high.
Treatment-refractory corticotroph tumors uniquely harbored mutations in ATRX, DAXX, TERT, and TSC2, none of which were found in benign tumors.
USP8 gain-of-function mutations were rare in aggressive PitNETs (n = 3 of 22 retrospective patients); the one patient with a canonical USP8 hotspot mutation had a complete and durable response to radiation.
LOH was stable across sequential resections and was present in pre-radiation specimens, establishing it as an early genomic event rather than a treatment-induced artifact.

Genes & alterations

TP53: Somatic mutations enriched in treatment-refractory PitNETs (52% vs. 1.4%); clonal in 11/12 cases; associated with higher fraction of LOH.
ATRX / DAXX: Mutations unique to treatment-refractory corticotroph tumors; implicated in alternative lengthening of telomeres (ALT).
TERT: Mutations found exclusively in treatment-refractory corticotroph PitNETs.
TSC2: Mutations found exclusively in treatment-refractory PitNETs.
MSH2 / MSH6 / MLH1: Mismatch repair gene mutations in 5 treatment-refractory corticotroph tumors, with loss of protein expression confirmed by IHC.
SF3B1: R625C hotspot mutation in 1 of 7 treatment-refractory lactotroph PitNETs; previously reported in aggressive prolactinomas.
GNAS: R201 hotspot in 1 treatment-refractory somatotroph PitNET; none in benign somatotrophs in this cohort.
USP8: Canonical gain-of-function mutations rare (1/22); not associated with treatment-refractory behavior.
CCND3: Amplification acquired in a recurrent tumor (patient TR-9).
CDKN2A / CDKN2B: Loss acquired in a recurrent, metastatic tumor (patient TR-9).

Clinical implications

Fraction of LOH greater than 0.11 is a candidate biomarker for predicting future aggressive, treatment-refractory behavior in PitNETs at the time of initial resection, with accuracy of 0.88 and negative predictive value of 0.95.
Mismatch repair deficiency, present in a subset of treatment-refractory corticotroph PitNETs, may render these tumors responsive to pembrolizumab and other checkpoint inhibitors; dramatic responses have been reported in this context.
Temozolomide was used in 77% of retrospective patients; one tumor developed acquired mismatch repair deficiency and hypermutation (93 mut/Mb) under temozolomide pressure with 76% of mutations attributable to the alkylator signature.
The rcLOH phenotype in corticotroph tumors resembles patterns seen in pancreatic neuroendocrine tumors and adrenocortical carcinoma, suggesting shared endocrine tumor biology that may inform cross-tumor therapeutic strategies.

Limitations & open questions

The cohort is limited in size (23 treatment-refractory, 69 benign), reflecting the rarity of treatment-refractory PitNETs. Larger prospective cohorts with longer follow-up are needed for validation.
The retrospective cohort is enriched for aggressive tumors (referral bias to MSKCC), so prevalence estimates of specific alterations may not generalize.
The prospective cohort may not capture the full spectrum of PitNET behavior given the relatively short follow-up period.
USP8 is not on MSK-IMPACT panels; whole-exome recapture was only performed for the retrospective cohort, so USP8 status is incomplete for the prospective group.
The functional mechanism by which rcLOH drives aggressive behavior (e.g., haploinsufficiency of specific tumor suppressors) remains undefined.
Two benign tumors were misclassified as treatment-refractory by the LOH classifier; whether these represent false positives or tumors that will eventually progress is unknown.

Citations from this paper used in the wiki

“A higher mutational burden and fraction of loss of heterozygosity (LOH) was found in the aggressive, treatment-refractory PitNETs compared to the benign tumors (p = 1.3 x 10^-10 and p = 8.5 x 10^-9, respectively).” (Abstract)
“TP53 [mutations] were overwhelmingly found in the treatment-refractory tumors (12 of 23 treatment-refractory tumors vs 1 of 69 benign tumors, two-sided Fisher’s exact test: p value = 4.2 x 10^-8).” (Results, p. 6)
“This specific pattern of recurrent chromosomal LOH (rcLOH), was observed in 78% (11/14) of the aggressive, treatment-refractory corticotroph PitNETs compared to only a single (1/14) benign corticotroph tumor (one-sided Fisher’s exact test, p = 1.7 x 10^-4).” (Results, p. 7)
“A machine learning approach identified loss of heterozygosity as the most predictive variable for aggressive, treatment-refractory behavior, outperforming the most common gene-level alteration, TP53, with an accuracy of 0.88 (95% CI: 0.70-0.96).” (Abstract)
“Five aggressive, treatment-refractory corticotroph tumors contained an oncogenic or likely oncogenic alteration in mismatch repair genes (MSH2, MSH6, and MLH1).” (Results, p. 7)

This page was processed by crosslinker on 2026-05-04.