Rapid intraoperative boundary diagnosis of somatotroph pituitary neuroendocrine tumors using real-time PCR

Abstract

Pituitary neuroendocrine tumors (PitNETs) are generally benign; however, functional subtypes cause hormone hypersecretion, leading to systemic complications. In somatotroph PitNETs, growth hormone (GH) excess increases mortality and necessitates complete tumor resection to achieve endocrinological remission. Although extracapsular resection along the pseudocapsule is preferred, the plane may be poorly defined with tumor cells infiltrating adjacent normal tissue, necessitating further resection of the outer layer. Conversely, excessive resection increases the risk of postoperative hypopituitarism; therefore, precise intraoperative delineation of tumor boundaries is essential. Approximately half of the somatotroph PitNETs harbor GNAS mutations. We used a microfluidic real-time PCR platform (GeneSoC^®) that detects mutations within approximately 15 min for intraoperative molecular boundary diagnosis. First, archival frozen samples were analyzed using Sanger sequencing, which identified GNAS R201C mutations in 11 of 24 samples (45.8%), with no other point mutations detected. After optimizing the real-time PCR using Sanger-validated samples, plasmids with defined variant allele frequencies (VAFs) were engineered to confirm detection accuracy. Real-time PCR demonstrated sensitivity and specificity of 1.000 in this cohort with a fluorescence intensity cutoff of 40, reliably detecting mutations ≥ VAFs 2%. Intraoperative analyses of multiple sites (23 samples) from five GNAS R201C-positive cases demonstrated concordance between real-time PCR results and those of droplet digital PCR. Notably, mutations were detected even at sites where tumor cells could not be clearly identified histopathologically. All patients achieved complete resection with preservation of pituitary function. These findings indicate that real-time PCR provides a rapid and objective method for intraoperative molecular boundary diagnosis of somatotroph PitNETs, facilitates safe and radical resection, and may also be applicable for other tumors harboring detectable genetic mutations.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-026-02255-5.

Introduction

Pituitary neuroendocrine tumors (PitNETs) are generally benign; however, their functional subtypes secrete excessive hormones, resulting in diverse systemic complications. Somatotroph PitNETs, in particular, secrete excessive growth hormone (GH), resulting in acromegaly, which is associated with increased risks of cardiovascular diseases, glucose metabolism disorders, respiratory dysfunction, osteoarticular diseases, and malignancies [5, 16, 19, 20]. These complications significantly increase mortality if GH secretion is left uncontrolled [2, 5, 19, 38]. Therefore, complete surgical resection is necessary for achieving endocrinological remission.

Endoscopic transsphenoidal surgery (eTSS) is the standard treatment approach for PitNETs. Complete resection is ideally achieved by extracapsular resection outside the pseudocapsule [9, 11, 27, 39]. The pseudocapsule is a fibrous, multilayered capsule of pituitary origin, which is formed by compression of the normal pituitary gland by the tumor [27]. However, this pseudocapsule may be poorly defined, with clusters of tumor cells infiltrating adjacent normal tissues [9, 14, 26–28], necessitating resection by further peeling off an additional outer layer [23]. Because the pseudocapsule originates from normal pituitary tissue, excessive resection carries the risk of damaging the adjacent normal pituitary tissue, potentially leading to postoperative hypopituitarism. Therefore, balancing complete tumor resection with preservation of pituitary function is crucial and necessitates precise intraoperative delineation of tumor boundaries.

Approximately half of somatotroph PitNETs harbor GNAS gene mutations (NM_000516.7, HGNC:4392), which are clonally preserved and represent a major driver event [3, 7, 8, 13, 18, 30, 31, 35, 38]. Conventionally, detection of these mutations has relied on methods such as droplet digital polymerase chain reaction (ddPCR) and Sanger sequencing. Despite being highly accurate, these techniques are impractical for intraoperative use because of time constraints. In contrast, real-time PCR using GeneSoC provides a rapid and sensitive alternative, capable of detecting genetic mutations within approximately 15 min. This technology has already been adopted for the intraoperative diagnosis of gliomas and malignant lymphomas, demonstrating diagnostic performance comparable to that of histopathology and ddPCR [15, 36]. Given that genetic mutations directly reflect the presence of tumor cells, we hypothesized that rapid detection of these mutations could be applied to intraoperative boundary diagnosis. Moreover, compared with highly invasive malignant tumors, PitNETs exhibit relatively well-defined borders owing to the pseudocapsule, suggesting that this approach may be particularly suitable.

Therefore, we aimed to determine whether intraoperative genetic analysis using real-time PCR targeting GNAS mutations could provide a precise boundary diagnosis of somatotroph PitNETs, thereby improving both the safety and radicality of surgical resection.

Materials and methods

Patient characteristics

Patients who underwent eTSS for somatotroph PitNETs between March 2024 and October 2024 were included. All patients included in this study had newly diagnosed somatotroph PitNETs. None of the patients received preoperative medical therapy. The patients’ characteristics are summarized in Table 1. Diagnosis was biochemically confirmed by elevated insulin-like growth factor I (IGF-I) levels above the age-adjusted reference range and failure of GH suppression during a 75-g oral glucose tolerance test (OGTT).

Table 1.

Patient characteristics and clinical outcomes

Case	Age	Sex	Tumor volume (cm3)	Knosp grade*	Pre-op medical therapy	Post-op endocrinological remission**	Post-op pituitary function	EOR	Bleeding	CSF leakage	Recurrence	Follow-up (months)
1	64	F	2.64	2	No	Yes	No change	GTR	No	No	No	18.1
2	55	M	1.51	2	No	Yes	No change	GTR	No	No	No	15.2
3	38	M	2.08	3 A	No	Yes	No change	GTR	No	No	No	18
4	70	M	0.99	1	No	Yes	No change	GTR	No	No	No	16.7
5	28	M	0.78	1	No	Yes	No change	GTR	No	No	No	15.3

F, female; M, male; pre-op, preoperative; post-op, postoperative; EOR, extent of resection; GTR, gross total resection; CSF, cerebrospinal fluid

* Knosp grading based on preoperative contrast-enhanced MRI to evaluate cavernous sinus invasion

** Endocrinological remission at 1 year postoperatively

Additionally, Sanger sequencing was performed on frozen tumor samples obtained between 2019 and 2023 from patients with confirmed somatotroph PitNETs.

Written informed consent was obtained from all participants. The study protocol was approved by the Institutional Review Board of Nagoya University (approval number: 2023–0461).

Genomic DNA extraction

Genomic DNA (gDNA) was extracted from 24 archived frozen tumor specimens stored at − 80 °C using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. DNA was extracted from formalin-fixed, paraffin-embedded (FFPE) tissues using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany). DNA concentrations were quantified using the Qubit dsDNA HS Assay Kit (Invitrogen, Paisley, Scotland).

Sanger sequencing

Mutations at codons 201 and 227 of the GNAS gene—accounting for nearly all reported pathogenic variants—were assessed [7, 31, 38]. For the codon 201 mutation located in exon 8, a 184-base pair (bp) region was amplified using the following primers: forward primer 5′-TTTCGGTTGGCTTTGGTGAG-3′ and reverse primer 5′-GGGACTGGGGTGAATGTCAA-3’. For the codon 227 mutation located in exon 9, a 165-bp region was amplified with the forward primer 5′-TTGACATTCACCCCAGTCCC-3′ and reverse primer 5′-ACAACACAGAAGCAAAGCGT-3’. PCR conditions were as follows: an initial denaturation at 94 °C for 2 min; 35 cycles of denaturation at 98 °C for 10 s, annealing at 62 °C for 30 s, and extension at 68 °C for 30 s; followed by a final extension at 68 °C for 5 min. Prior to PCR amplification, all samples were adjusted to 10 ng/µL. Details of the reagents used are provided in Supplementary Table 1a. Mutations were identified using Sanger sequencing, and sequence analysis was performed using A plasmid Editor, version 2.0.55 (M. Wayne Davis University, Utah, USA).

Real-time PCR optimization

To enable rapid intraoperative genotyping, we used the GeneSoC^® (Kyorin Pharmaceutical Co., Ltd., Tokyo, Japan), a microfluidic real-time PCR device that detects mutations within approximately 15 min. This system uses mutation-specific fluorescent probes, in which an increase in the fluorescence intensity in the mutant channel (6-carboxyfluorescein [FAM]-labeled probe) indicates the presence of a mutation. Probes for the GNAS R201C mutation (FAM-labeled; UniqueAssay ID: dHsaCP2500562) and the corresponding wild-type sequence (hexachloro-fluorescein [HEX]-labeled probe; ID: dHsaCP2500563) were obtained from Bio-Rad (Pleasanton, CA, USA). PCR was performed using GeneSoC^® PCR Master Mix reaction (Kyorin Pharmaceutical Co., Ltd., Tokyo, Japan). Fluorescent probes were added at final volumes of 0.5 µL for the FAM-labeled probe and 1.0 µL for the HEX-labeled probe, as detailed in Supplementary Table 1b. Initial assay optimization employed two representative samples: one harboring the R201C mutation and one confirmed wild-type, as determined by Sanger sequencing. Real-time PCR conditions were as follows: heat shock/denaturation (HS/DN) at 96 °C and annealing/extension (AE) at 60 °C, with each cycle consisting of 30 s of HS, 5 s of DN, and 9 s of AE. GeneSoC employs a disposable sealed microfluidic cartridge, enabling PCR amplification and fluorescence readouts within a closed chamber. The cartridge was sealed immediately after the DNA loading, aerosol limitation, and post-PCR contamination. The PCR setup and DNA extraction were performed in areas physically isolated from the GeneSoC analysis area.

Construction of plasmids containing mutant and wild-type GNAS sequences

To further validate the assay’s sensitivity and specificity, a pcDNA3.1 plasmid construct was generated containing defined proportions of GNAS R201C-mutant and wild-type sequences. Initially, DNA fragments encompassing the mutation site (c.601 C > T) and the corresponding wild-type sequence were PCR-amplified from gDNA extracted from a patient harboring the GNAS R201C mutation. PCR amplification was performed using the following primers: forward primer 5′-CGAAGCTTCCGTTGAGCCTGACCTTGTA-3′ and reverse primer 5′-ATGGATCCCTTTACGAACAGCCAAGCCC-3′. Details of the other reagents are listed in Supplementary Table 1a. The resulting products were digested with HindIII and BamHI and ligated into the multiple cloning sites of the pcDNA3.1 vector using Ligation Mix (Takara Bio, Shiga, Japan). The purified recombinant plasmid DNA was further fragmented into 1,000 bp fragments using a Covaris Focused Ultrasonicator M220 (Covaris Inc., Woburn, USA). Defined mixtures of the mutant and wild-type plasmids were prepared with six variant allele frequencies (VAFs) (0%, 2%, 5%, 10%, 25%, and 50%) based on a molar basis.

Analysis of mutation detection accuracy

Real-time PCR was performed on each plasmid mixture, along with 24 samples of known mutation status that had been previously confirmed by Sanger sequencing. An optimal FAM fluorescence intensity cut-off value was defined to balance sensitivity and specificity. Applying this cutoff, the minimum detectable VAF was determined based on the fluorescence intensity of each plasmid mixture.

Additionally, the extracted DNA from tumor specimens was analyzed using the same protocol to determine whether VAF detectability with plasmid-derived standards could be generalized to clinical specimens. The VAFs were adjusted by combining ddPCR-quantified mutant-positive tumor DNA and ddPCR-confirmed wild-type DNA.

Assessment of optimal DNA input

Excessive template input in real-time PCR can artifactually drive the FAM signal to exceed the detection threshold within 50 cycles, resulting in false positives even in wild-type samples. In contrast, insufficient input can result in failure to reach the detection threshold and false negatives in mutation-positive samples. Therefore, we empirically defined an operational DNA input range that enables intraoperative real-time PCR without prior DNA quantification. Specifically, we evaluated the optimal range of cycle threshold (Ct) values at which the fluorescence intensity of the wild-type (WT; HEX channel) exceeded the threshold. Ct (WT) values were calculated using the crossing point method. Serial dilutions were prepared in distilled water at the indicated concentrations. In addition, frozen tumor samples (0.5, 2, 5, 15, and 25 mg) that were mutation-positive by Sanger sequencing were prepared for real-time PCR. The appropriate weight required to obtain a Ct (WT) value within the expected range was assessed.

Surgical procedure

In all cases, eTSS was performed. The tumor center was initially excised through a dural incision. If a mutation was detected, additional intraoperative samples were collected from anatomically distinct sites. In intracapsular resection, samples were collected from the tumor margins, whereas in extracapsular resection, samples were collected from the pseudocapsule. When the pseudocapsule was poorly defined and complete extracapsular resection could not be achieved, the surface of the adjacent pituitary gland tissue was additionally peeled off. When the tumor was in direct contact with the medial wall of the cavernous sinus, medial wall removal was performed [22]. Thereafter, specimens were collected from regions suspected of tumor invasion.

Intraoperative boundary diagnosis using real-time PCR

Real-time PCR was employed for intraoperative boundary diagnosis using multiple samples collected as described above. Each specimen from the same site was divided for histopathological diagnosis and detection of GNAS mutations. The sample for genetic analysis was placed in a tube containing 100 µL of ultrapure water and 10 µL of proteinase K, then mechanically minced with a needle. The tube was incubated at 98 °C for 5 min to release gDNA. After a brief centrifugation, the supernatant was combined with reagents for real-time PCR. The results were compared with the corresponding histopathological findings to determine the presence of tumor cells. The definitive mutation status was confirmed using ddPCR.

Histopathological analyses

Surgically resected specimens were fixed in 10% neutral-buffered formalin and embedded in paraffin. Standard hematoxylin and eosin (H&E) staining was performed to determine the presence of tumor cells in each specimen. Specimens were incubated with an antibody against GH (Dako, Carpinteria, CA, USA; A0570) for immunohistochemical analysis. All histopathological assessments were independently performed by a pathologist who was blinded to the results of real-time PCR and ddPCR.

Droplet digital PCR

Extracted gDNA was combined with 2× ddPCR Supermix for Probes (Bio-Rad Laboratories, Pleasanton, CA, USA). Each primer and probe targeted both the GNAS R201C mutation and wild-type sequences (details in Supplementary Table 1c). The forward primer was 5′-TTTCGGTTGGCTTTGGTGAG-3′, and the reverse primer was 5′-GGGACTGGGGTGAATGTCAA-3′. The probes were identical to those employed in real-time PCR (UniqueAssay ID: dHsaCP2500562 for the FAM-labeled probe and dHsaCP2500563 for the HEX-labeled probe). Subsequently, droplets were generated using the QX200 Droplet Generator (Bio-Rad Laboratories, Richmond, CA) by adding 65 µL of droplet generation oil (Bio-Rad Laboratories). PCR amplification was performed using a C1000 Thermal Cycler (Bio-Rad Laboratories) under the following conditions: initial enzyme activation at 95 °C for 10 min; 40 cycles of denaturation at 94 °C for 30 s and AE at 53 °C for 1 min (ramp rate: 2 °C/second); and final enzyme deactivation at 98 °C for 10 min (ramp rate: 2 °C/second). The fluorescence intensity of individual droplets was measured using a QX200 Droplet Reader (Bio-Rad Laboratories) and analyzed with QuantaSoft droplet reader software (Bio-Rad Laboratories). The thresholds for ddPCR were set to FAM 3500 and HEX 2000.

Endocrinological assessment

Baseline hormone levels were assessed preoperatively, on postoperative day 1, as well as at 1 week, 3 months, 6 months, and 1 year after surgery. A 75-g OGTT was performed preoperatively and postoperatively. Additionally, postoperative hormonal responses to the administration of thyrotropin-releasing hormone, luteinizing hormone-releasing hormone, and corticotropin-releasing hormone were assessed. Endocrinological remission was defined as the normalization of sex- and age-adjusted IGF-I levels and a nadir serum GH level < 0.4 ng/mL during OGTT, both assessed ≥ 12 weeks after surgery [20]. All pituitary stimulation tests were conducted following the guidelines of the Japanese Society for Hypothalamic and Pituitary Tumors [34].

Statistical analysis

Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used for diagnostic performance assessment. Analyses were performed using the pROC package (version 1.18.5). The decision threshold shown in the figures was determined using Youden’s J statistic. Plots were generated using ggplot2 software (version 3.5.2). Scatter plots show individual data points with quasirandom jitter, implemented via the ggbeeswarm package, with error bars indicating the mean ± SD. In VAF-stratified plots, the mean fluorescence intensity for each group is shown with error bars indicating the mean ± SD. ggplot2 was used to generate amplification curves of the wild-type and mutant templates and plots of per-sample VAF and maximum fluorescence intensity, with exact values labeled above each bar. All analyses were performed using the R software (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria).

Results

DNA extraction and Sanger sequencing

gDNA extraction and quantification were performed on 24 archived frozen somatotroph PitNET samples. The DNA concentrations obtained are summarized in Supplementary Table 2. Sanger sequencing detected the GNAS R201C (c.601 C > T) mutation in 11 of 24 samples (45.8%); no other mutations at codons 201 or 227 were found.

Analytical performance of real-time PCR

Figure 1 demonstrates clear discrimination between mutant and wild-type GNAS R201C using real-time PCR, with mutation status confirmed by Sanger sequencing.

Fig. 1.

Demonstrates the real-time PCR results on two representative samples: one harboring the GNAS R201C mutation (c.601 C > T) and the other confirmed as wild type by Sanger sequencing. The optimal conditions were defined to identify the presence of mutations a GNAS R201C mutation-positive sample; b GNAS R201C mutation-negative sample

The maximum fluorescence intensity by real-time PCR in 24 archived samples are presented in Fig. 2a. ROC curve analysis established a fluorescence intensity threshold of 40, corresponding to a sensitivity and specificity of 1.000 for GNAS R201C mutation detection (Fig. 2b). Values ≥ 40 were considered positive.

Fig. 2.

a The maximum fluorescence intensity by real-time PCR in 24 archived samples with known mutation status confirmed by Sanger sequencing b Receiver operating characteristic curve analysis demonstrating sensitivity and specificity of 1.000 c Real-time PCR results of six plasmid mixtures with defined variant allele frequencies (VAFs) ranging from 0 to 50% (0, 2, 5, 10, 25, and 50%). The maximum FAM fluorescence intensity increased proportionally with VAF d The maximum FAM fluorescence intensity at different VAFs (n = 3). The dotted line indicates fluorescence thresholds of 40. Mutations were reliably detected in the samples with VAFs as low as 2% e Tumor-derived DNA analyzed under the same conditions demonstrated results consistent with those of the plasmid mixtures

In the plasmid mixtures containing six defined VAFs (0–50%), the maximum fluorescence intensity in the mutant FAM channel increased proportionally with VAF (Fig. 2c). Using the threshold of 40, mutations were reliably detected in samples with VAFs as low as 2% (Fig. 2d).

Additionally, analysis of tumor-derived DNA under the same conditions yielded results consistent with those obtained using the plasmid mixtures (Fig. 2e).

Using high-input wild-type plasmids (1 × 10², 1 × 10¹, and 1 pM), the FAM signal did not exceed the threshold within 50 cycles (Fig. 3a). The lowest Ct (WT) value was 28.4. Therefore, even with high-template inputs, false-positive results are unlikely when the Ct (WT) value is above this level. In addition, low-input plasmids (1 × 10^-2 pM, 5 × 10^-2 pM, and 1 × 10^-1 pM) containing 2% of the GNAS R201C mutation were also analyzed. At least 1 × 10^-1 pM DNA consistently produced FAM fluorescence > 40 within 50 cycles (Fig. 3b). In these analysis, the Ct (WT) value was 39.4, indicating that when the Ct (WT) value is below this level, mutant alleles at ≥ 2% abundance can be reliably detected. Thus, GeneSoC results were considered reliable when the Ct (WT) value ranged from 28.4 to 39.4. In the analysis of samples weighing 0.5, 2, 5, 15, and 25 mg with the mutation confirmed by Sanger sequencing, an approximate range of 2–15 mg was found to be suitable to obtain Ct (WT) values within the expected range (Fig. 3c).

Fig. 3.

a, b Real-time PCR performed with high (a) and low (b) amounts of artificial DNA. The dotted line indicates the fluorescence threshold of 40. The x-axis shows the cycle-threshold (Ct) value at which the wild-type (HEX) signal first exceeds 40, and the y-axis shows the maximum FAM fluorescence intensity. Real-time PCR results were considered reliable when the Ct value of the wild-type allele ranged from 28.4 to 39.4 a Plasmids containing only the wild-type sequence at high-input levels (1 × 10², 1 × 10¹, and 1 pM) b Plasmids containing 2% of the GNAS R201C mutation at low-input levels (1 × 10 ⁻², 5 × 10 ⁻², and 1 × 10 ⁻¹ pM) c Analysis of samples weighing 0.5, 2, 5, 15, and 25 mg with the mutation confirmed by Sanger sequencing. An approximate range of 2–15 mg was found to be suitable for obtaining Ct (WT) values within the expected range of 28.4 to 39.4

Intraoperative boundary diagnosis using real-time PCR

Among the nine somatotroph PitNETs cases treated between March 2024 and October 2024, five (23 samples) demonstrated the GNAS R201C mutation intraoperatively using real-time PCR. Among the five patients, four were male and one was female; ages at surgery ranged from 28 to 70 years (mean, 51 years). The method enabled complete analysis—from sample collection to the final results—in approximately 25 min.

In case 1, five samples were collected from distinct regions: (a) tumor center, (b) tumor margin, (c) pseudocapsule, (d and e) adjacent pituitary gland tissue. Real-time PCR demonstrated fluorescence intensity values exceeding the cutoff of 40 in samples (a–d), indicating the presence of the mutation (Fig. 4a–e, g).

Fig. 4.

Real-time PCR, histopathology and droplet digital PCR (ddPCR) results in case 1. a–e: Real-time PCR and histopathology findings. In sample d, real-time PCR detected the mutation, whereas H&E staining and GH immunostaining did not clearly confirm tumor cells. Sample locations: a, central; b, marginal; c, pseudocapsule; d and e, adjacent pituitary gland. Histopathology panels: left, H&E staining; right, GH immunostaining. f: ddPCR findings (upper, FAM channel; lower, HEX channel). The results were concordant with the real-time PCR results. The pink line indicates the threshold (FAM 3500, HEX 2000). g, h: The maximum FAM fluorescence intensity by real-time PCR and VAF values determined by ddPCR for each sample. Clinical samples (a–e) are shown in gray, PC in blue, and NC in red, with exact values labeled above the bars. Sample e showed a VAF of 0.8%, which was below the detection threshold of real-time PCR. PC, Positive control; NC, Negative control

In case 2, samples were collected from (a) tumor center, (b) tumor margin, (c) pseudocapsule, (d) adjacent pituitary gland tissue, and (e) the medial wall of the cavernous sinus. Real-time PCR detected mutations in samples (a–c) and (e) (Fig. 5a–e, g).

Fig. 5.

Real-time PCR, histopathology and ddPCR findings in case 2. a–e: Real-time PCR and histopathology findings. In sample e, real-time PCR detected the mutation, while H&E staining and GH immunostaining did not clearly identify tumor cells. Sample locations: a, central; b, marginal; c, pseudocapsule; d, adjacent pituitary gland; e, medial wall of cavernous sinus. Histopathology panels: left, H&E staining; right, GH immunostaining. f: ddPCR findings (upper, FAM channel; lower, HEX channel). The results were concordant with the real-time PCR results. The pink line indicates the threshold (FAM 3500, HEX 2000). g, h: The maximum FAM fluorescence intensity by real-time PCR and VAF values measured by ddPCR for each sample. Clinical samples (a–e) are shown in gray, PC in blue, and NC in red, with exact values labeled above the bars. Sample d had a VAF of 0.7%, which was below the detection threshold of real-time PCR. PC, Positive control; NC, Negative control

Contrast-enhanced magnetic resonance imaging (MRI) for all cases is provided in Supplementary Fig. 1. Imaging findings were reviewed to identify regions suspicious for tumor invasion. In these regions, real-time PCR results showed no clear discrepancies with imaging findings and were interpretable.

Concordance of histopathology and ddPCR with real-time PCR

In case 1, H&E staining and GH immunostaining confirmed tumor cells in samples (a–c), but not in sample (d and e) (Fig. 4a–e). ddPCR revealed the presence of the GNAS mutation in samples (a–d), in concordance with the results of real-time PCR. In contrast, sample (e) showed a VAF of 0.8%, which was below the detection limit of real-time PCR (Fig. 4f–h). Subsequently, real-time PCR on DNA extracted from FFPE tissues of sample (d) confirmed the presence of mutation with maximum FAM fluorescence intensity of 54.45 (Supplementary Fig. 2a).

In case 2, H&E staining and GH immunostaining confirmed tumor cells in samples (a–c), but not in sample (d and e) (Fig. 5a–e). ddPCR detected mutations in samples (a–c) and (e), consistent with real-time PCR results (Fig. 5f–h). Real-time PCR on FFPE tissues of sample (e) were positive for the mutation (Supplementary Fig. 2b).

In the remaining three cases (13 samples), intraoperative real-time PCR results were consistent with those of ddPCR in all samples. In two samples, the mutation was detected by real-time PCR and ddPCR, whereas tumor cells were not clearly identified by histopathology. Supplementary Table 3 summarizes the results of all five cases (23 samples) that were analyzed using real-time PCR, histopathology, and ddPCR.

Surgical and endocrinological outcomes

Complete resection was confirmed in all five patients intraoperatively by real-time PCR and on postoperative magnetic resonance imaging. No postoperative complications were observed, including those attributable to intraoperative sampling. Postoperative endocrinological remission at 1 year was confirmed in all patients. Pituitary function was preserved in all patients. The median follow-up duration was 16.7 months (range, 15.2–18.1 months), and no tumor recurrence has been observed to date (Table 1).

Discussion

Novelty and clinical implications

This study demonstrated that real-time PCR provides a rapid and objective method for intraoperative boundary diagnosis within the short timeframe of tissue collection. The microfluidic real-time PCR platform (GeneSoC^®) enables parallel analysis of multiple samples and is particularly advantageous in difficult cases with a poorly defined pseudocapsule or tumor extension into normal tissues. Our approach provides real-time objective confirmation of tumor cell absence in such cases, avoiding unnecessary resection and preserving pituitary function. In functional PitNETs, although histologically benign, uncontrolled hormone hypersecretion due to residual tumors can result in serious comorbidities and increased mortality. Our approach facilitates radical resection while minimizing the risk of postoperative endocrine complications. Intraoperatively, we performed a peel-off resection of the adjacent pituitary gland in areas where the pseudocapsule was not clearly identifiable. Although this technique is not associated with postoperative hypopituitarism when resection is limited, improper tissue manipulation may compromise this concept [23]. In this study, sampling was limited to visually indeterminate boundary regions and was not performed extensively. Regarding endocrinological remission rate, a recent systematic review by Nie et al. reported the rate of 56.2% following primary surgery for somatotroph PitNETs [25]. Although endocrinological remission was achieved in all patients in the present study, further investigation in larger cohorts is required.

In this study, we assumed that GNAS mutations in somatotroph PitNETs represent the initiating genetic alterations in tumorigenesis, and their detection indicates the presence of tumor cells. During intraoperative analysis, the presence of genetic mutations closely corresponded with the distribution of tumor cells, enabling a clear delineation of tumor margins. These findings indicate that GNAS mutations can serve as a reliable molecular marker for intraoperative boundary diagnosis of somatotroph PitNETs. Notably, in some peel-off samples obtained from regions without macroscopic suspicion of tumor cells, real-time PCR yielded mutation-positive results concordant with ddPCR. This suggests that intraoperative real-time PCR may offer objective information beyond conventional surgical observation and may contribute to improved remission outcomes. This approach may help standardize complete extracapsular resection independent of the surgeon’s experience or skill. Although this study focused on somatotroph PitNETs, this approach may also be applicable to other tumors with well-characterized molecular markers. As an extension of this work, we are currently investigating the USP8 mutations in corticotroph PitNETs. Through enabling rapid mutation-specific intraoperative diagnostics, this approach represents a step toward a new paradigm in molecular-guided neurosurgery.

Real-time PCR assay demonstrated excellent sensitivity, enabling the detection of the GNAS R201C mutation in samples at a VAF of ≥ 2%, corresponding to a tumor cell content of ≥ 4%, owing to the heterozygosity of the gene. We further observed a proportional relationship between the maximum FAM fluorescence intensity in real-time PCR and the VAF measured by ddPCR. This observation indicates the potential utility of semi-quantitative assessment of tumor burden. Such information could inform intraoperative decisions, particularly in regions where the tumor is closely adherent to critical structures, supporting a strategy to leave minimal residual tissue when the VAF is low. Nevertheless, even in these cases, there remains a risk of compromising postoperative endocrinological remission, highlighting the need to balance functional preservation with endocrinological remission. Currently, a clear quantitative relationship has not yet been established, and an acceptable residual VAF that would still enable postoperative endocrinological remission has not yet been defined. Larger cohorts are required to validate clinically appropriate fluorescence cutoff values based on these factors.

Comparison with histopathology

PitNETs present distinct challenges for intraoperative histopathological diagnosis. These tumors consist of uniform, densely packed monomorphic cells. Assessment of tumor boundaries can be particularly difficult, because it may require the detection of subtle variations in cellular density and uniformity. In such areas, H&E staining of frozen sections can compromise tissue morphology, limiting the evaluation of fine structural changes. Additionally, this approach may be constrained by technical and temporal constraints [4, 12, 21]. Although immunohistochemical assessment methods, including hormone and transcription factor staining, can help in diagnosis, non-tumorous pituitary cells may also exhibit immunoreactivity, necessitating careful correlation with morphological features [1, 17, 24, 33]. Moreover, intraoperative immunohistochemical evaluation can be challenging. Recently, novel approaches—such as stimulated Raman histology combined with deep learning—have been reported to enable rapid intraoperative diagnosis, although their clinical application is still limited [29].

Real-time PCR provides an objective, morphology-independent method of tumor detection based on the mutation status and may be an alternative when on-site histopathological evaluation is limited. In our study, real-time PCR detected the mutations in areas where tumor cells could not be clearly identified by histopathology, highlighting the potential utility of real-time PCR as a complementary intraoperative diagnostic tool. In some samples, although the tumor cells were morphologically altered or crushed on H&E staining, the mutation was still detected by real-time PCR. The corresponding FFPE specimen from the same site was also mutation-positive, ruling out sampling error. These findings suggest that this method remains reliable as long as DNA integrity is preserved. Taken together, intraoperative real-time PCR should be regarded as a complementary adjunct to histopathological evaluation rather than a replacement.

Clinical and biological significance of GNAS mutations

Activating mutations of the GNAS gene, located in exons 8 or 9, at codons 201 or 227, are present in approximately 40–60% of somatotroph PitNETs. GNAS encodes the stimulatory G protein alpha subunit, a key mediator of the cyclic adenosine monophosphate (cAMP) signaling pathway. These mutations constitutively activate the cAMP pathway, leading to excessive GH secretion; they are also observed in McCune–Albright syndrome and various endocrine tumors [3, 8, 13, 18, 30, 35, 37, 38].

Clinically, GNAS-mutant PitNETs have been reported to be smaller [7, 10, 18, 30, 31, 37, 38] or comparable in size [8, 13, 32] to wild-type tumors, less invasive [18, 31, 37, 38], or not significantly different [7, 8, 10, 13]; however, these findings remain inconsistent across studies.

Moreover, these tumors tend to present with increased preoperative GH and IGF-I levels and show a better response to somatostatin analog therapy than wild-type tumors. They are frequently observed in slightly older patients and are often GH-monoproducing, densely granulated tumors [3, 6–8, 13, 18, 30, 31, 37, 38]. These increases in hormone levels may lead to early clinical presentation and diagnosis prior to significant tumor progression. Prior studies demonstrated either no difference in postoperative endocrinological remission rates between GNAS-positive and negative tumors [7, 38] or higher rates in GNAS-positive tumors than in GNAS-negative tumors [8].

Limitations

This study has some limitations. First, the number of cases analyzed intraoperatively was limited. Although concordance with ddPCR was observed in all samples, a larger sample size is required to validate the generalizability of these findings. Future prospective multicenter studies with larger cohorts should evaluate whether the use of intraoperative real-time PCR improves long-term outcomes, such as endocrinological remission or tumor recurrence, and support its clinical utility and standardization in surgical practice. Second, this proof-of-concept study is limited to tumors harboring GNAS R201C mutations, and its applicability to mutation-negative cases remains restricted. In such cases, surgical decisions still rely on histopathological and gross anatomical assessments. In addition, the current approach is limited by feasibility constraints. Third, despite careful handling and processing, the risk of sample contamination cannot be entirely excluded.

Conclusions

Intraoperative real-time PCR is an effective tool for boundary diagnosis of somatotroph PitNETs. This method improves the surgeon’s ability to distinguish the tumor tissue from surrounding structures, including the normal pituitary gland tissue, thereby potentially improving both the safety and radicality of tumor resection. Furthermore, this method may be applicable for other tumors harboring detectable mutations. Our study provides initial evidence supporting the feasibility of the intraoperative molecular diagnosis and boundary detection of somatotroph PitNETs, highlighting the need for further multicenter validation.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

PitNETs

Pituitary neuroendocrine tumors

GH

Growth hormone

VAFs

Variant allele frequencies

eTSS

Endoscopic transsphenoidal surgery

ddPCR

Droplet digital polymerase chain reaction

IGF-I

Insulin-like growth factor I

OGTT

Oral glucose tolerance test

FFPE

Formalin-fixed, paraffin-embedded

WT

Wild-type

H&E

Hematoxylin and eosin

MRI

Magnetic resonance imaging

Author contributions

Conception and design: K.T., F.O., and R.S.; development of methodology: Y.S., S.M., and F.O.; acquisition of data: Y.S., K.T., Y.N., K.I., E.O., S.D., T.H., and T.M.; analysis and interpretation of data: Y.S., K.T., and F.O.; writing of the manuscript: Y.S., K.T., and Y.N.

Funding

This work was supported by JSPS KAKENHI Grant Number JP23K08540, and by the program Creating training hubs for advanced medical personnel (Supporting the fostering of doctors with advanced clinical and research capabilities).

Data availability

All data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of the Nagoya University Hospital (approval number: 2023–0461) and complied with all provisions of the World Medical Association Declaration of Helsinki. Tumor samples were collected intraoperatively after obtaining informed consent from the patients.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.