Prolactin Levels May Predict Which Acromegaly Patients Stay Sick After Surgery

For patients with acromegaly, the news that surgery has removed their growth hormone-secreting pituitary tumor is often only the beginning of the story. Even when a skilled neurosurgeon removes what looks like the entire tumor, a substantial fraction of patients fail to reach biochemical remission, leaving them exposed to years of excess growth hormone and the joint disease, cardiovascular strain, and metabolic complications that follow. Why some tumors surrender after resection while others stubbornly keep secreting has long been a black box. A new single-cell study from researchers at Yonsei University College of Medicine in Seoul now offers a striking clue: the answer may lie, in part, in a hormone that is not supposed to be the star of the show at all — prolactin.

The study, published in the Journal of Translational Medicine, applied single-cell RNA sequencing to 55,648 individual cells drawn from 12 somatotroph pituitary neuroendocrine tumors, the benign but hormonally aggressive growths responsible for acromegaly. Rather than treating each tumor as a uniform mass of identical hormone-producing cells, the team dissected them cell by cell, cataloguing the transcriptional states that coexist within a single patient’s tumor. What emerged was a picture of structured heterogeneity: the tumor cells occupied a continuum of states the researchers labeled classic, intermediate, and neuronal-associated, all while retaining the molecular identity of their pituitary lineage. The tumors, in other words, are not blobs of sameness but layered ecosystems of related cell states arranged along a transcriptional gradient.

That finding alone would have been a useful contribution to the growing atlas of pituitary tumor biology. But the Yonsei team, led by neurosurgeons and endocrinologists from the Severance Hospital Pituitary Tumor Center, pushed the analysis one level higher. Using non-negative matrix factorization, a computational technique that decomposes complex expression matrices into additive components, they aggregated the single-cell data into patient-level profiles and asked whether the twelve tumors would sort themselves into meaningful groups. They did — and the dividing line was unexpected. The tumors separated cleanly into two clusters defined not by their cell-state composition, tumor size, or canonical oncogenic mutations, but by how strongly they expressed the gene encoding prolactin, a hormone classically associated with a different pituitary tumor type altogether.

The statistical robustness of this split is notable. Across 100 independent runs of the clustering algorithm, the two groups co-clustered with a mean within-group consensus of 0.925 and a between-group consensus of just 0.058, with an overall consensus silhouette width of 0.915 — figures that indicate the division is not an artifact of algorithmic noise. The team also ruled out obvious confounders: the correlation between prolactin expression and the proportion of classic tumor states was weak, and tumor size did not differ systematically between groups. One tumor showed an intermediate assignment, a reminder that biology rarely respects clean boundaries, but the core separation held firm.

Crucially, the prolactin signal was not confined to sequencing data. The researchers validated it at the protein level using immunohistochemistry on tumor sections, showing that tumors rich in PRL transcript were also rich in prolactin immunoreactivity. Even more striking, tumor prolactin expression correlated with patients’ preoperative serum prolactin levels — a routine blood test already measured in nearly every acromegaly workup. That correlation transforms an esoteric molecular observation into something clinically actionable: a biomarker that could, in principle, be read off a standard preoperative lab panel without any need for fresh tumor tissue or specialized sequencing.

To test whether the prolactin signature actually mattered for patients, the team turned to an independent clinical cohort of 188 individuals with acromegaly. Using logistic regression and stratified analyses, they asked whether prolactin-high tumor status predicted postoperative biochemical remission — the formal endpoint, determined by growth hormone and insulin-like growth factor 1 levels under oral glucose tolerance testing, that defines surgical success. It did. Patients whose tumors fell into the prolactin-high group were significantly more likely to fail to achieve remission after surgery. The association held even as the researchers accounted for other clinical variables, positioning prolactin as a candidate adjunctive risk marker alongside the anatomical and surgical predictors — tumor size, invasiveness, and extent of resection — that surgeons already weigh.

Molecularly, the prolactin-high tumors turned out to be an enigma. Gene set enrichment analysis showed that they lacked enrichment of the canonical oncogenic pathways one might expect to find in a more aggressive tumor. Instead, the differences between prolactin-high and prolactin-low tumors played out at the level of individual genes, with a notable cluster of receptor-associated candidates — including EPHA7, a receptor tyrosine kinase gene — distinguishing the two groups. The authors suggest that these receptor-level differences may shape how tumor cells respond to their hormonal and microenvironmental milieu, though the functional biology remains to be worked out. The finding is intriguing precisely because it does not fit the standard playbook: the high-risk tumors are not more oncogenic in any obvious way, just transcriptionally different in ways that matter for hormone control.

The team took several steps to guard against wishful thinking. In an external cohort of 60 somatotroph tumors from a public dataset, a discovery-derived prolactin-associated signature score — deliberately constructed with prolactin itself and pituitary lineage genes excluded to avoid circularity — tracked strongly with tumor prolactin abundance, with a Spearman correlation of 0.64. As a negative control, the same signature applied to 45 nonfunctioning pituitary tumors from the same dataset showed no such association, confirming that the signal is specific to somatotroph biology rather than a generic pituitary artifact. A sensitivity analysis excluding the one ambiguously assigned tumor left the genome-wide fold-change estimates essentially unchanged, with a concordance of 0.98.

The translational stakes are real. Acromegaly is a rare disease, but its consequences are expensive and disabling, and the current decision tree after surgery is largely reactive: patients are monitored, and second-line treatments — repeat surgery, radiation, or medical therapy with somatostatin analogs, dopamine agonists, or the growth hormone receptor antagonist pegvisomant — are deployed only after remission fails to materialize. A preoperative biomarker that flags high-risk patients could shift that calculus, prompting earlier counseling, closer biochemical surveillance, or more aggressive multimodal treatment planning from the outset. The Yonsei group is already moving in this direction: according to the paper’s competing interests statement, the university is preparing a patent application covering the use of prolactin-associated features and selected receptor-associated markers, including EPHA7, for molecular and clinical characterization of these tumors.

Caveats remain, and the authors are careful to name them. The single-cell discovery cohort comprised only twelve tumors, and while the clinical validation cohort of 188 patients is substantial, the association between prolactin-high status and remission failure was established retrospectively. The researchers themselves state that preoperative prolactin may serve as a potential adjunctive biomarker for risk stratification but that prospective external validation is required before it enters routine practice. Whether the prolactin-high state reflects a distinct developmental origin of the tumor cells, an epigenetic program, or a response to microenvironmental cues is unknown, and the functional role of EPHA7 and the other receptor-associated genes is an open question. Still, the study exemplifies a productive pattern in modern tumor biology: single-cell atlases that do not merely describe heterogeneity but convert it into clinically legible categories. For a disease in which the difference between remission and chronic illness can hinge on molecular features invisible to the surgeon’s eye, a routine blood test that whispers a warning before the first incision is a prospect worth taking seriously.

Subject of Research: Single-cell transcriptomic stratification of somatotroph pituitary tumors by prolactin expression and postoperative remission risk

Article Title: Single-cell transcriptomics reveals prolactin-associated molecular stratification and remission risk in somatotroph PitNETs

Article References: Single-cell transcriptomics reveals prolactin-associated molecular stratification and remission risk in somatotroph PitNETs. (n.d.). https://doi.org/10.1186/s12967-026-09065-2

Image Credits: AI Generated

DOI: 10.1186/s12967-026-09065-2

Keywords: acromegaly, pituitary neuroendocrine tumor, single-cell RNA sequencing, prolactin, somatotroph tumor, transcriptomics, tumor heterogeneity, EPHA7, biomarker, remission, growth hormone, non-negative matrix factorization