DOCK10 Regulates Insulin Hypersecretion in Insulinoma and Serves as a Diagnostic and Therapeutic Target

Abstract

Background & Aims

Insulinomas are rare pancreatic neuroendocrine neoplasms (pan-NENs) characterized by inappropriate insulin secretion. Despite advances in imaging techniques, the reliable identification of insulin-secreting lesions remains challenging. In addition, medical treatment options are limited and have seen little development in recent years, highlighting the unmet need for improved diagnostic tools and therapeutic strategies. This study aimed to identify the molecular mechanisms underlying insulin hypersecretion in insulinomas.

Methods

We established a biobank of human insulinoma surgical specimens and matched organoids. Comprehensive transcriptomic analyses—including bulk RNA sequencing, single-cell RNA sequencing, quantitative polymerase chain reaction, and immunohistochemistry—were conducted to identify genes enriched in insulin-secreting components. Functional validation was performed using MIN6 cells, a xenograft mouse model, and long-term cultured human insulinoma organoids.

Results

We identified dedicator of cytokinesis 10 (DOCK10) as a gene selectively overexpressed in insulin-secreting components of insulinomas. DOCK10 knockdown impaired glucose-stimulated insulin secretion in both mouse insulinoma cells and patient-derived organoids. Inhibition of the downstream effector Cdc42 with ML141 reduced insulin hypersecretion and improved survival in a MIN6 xenograft mouse model. These findings uncover a previously unrecognized role of the DOCK10–Cdc42 axis in regulating insulin secretion in insulinoma.

Conclusions

This study suggests that DOCK10 may serve as a diagnostic marker for insulin-secreting lesions and a potential therapeutic target in insulinoma. It provides mechanistic insights that may inform future strategies for precision diagnostics and treatment of functional pancreatic neuroendocrine tumors.

Graphical abstract

Summary.

Using insulinoma samples and organoids, we identified dedicator of cytokinesis 10 as a regulator of insulin secretion, specifically expressed in insulin-secreting lesions, suggesting its potential as both a diagnostic marker and therapeutic target.

What You Need to Know.

Background

Insulinomas may cause hypoglycemia even when small, and coexistence of secreting and non-secreting lesions makes localization difficult. Surgery remains mainstay as molecular mechanisms and therapies are limited.

Impact

We identified dedicator of cytokinesis 10 as a novel regulator of dysregulated insulin secretion in insulinoma, suggesting diagnostic and therapeutic potential, and established a long-term culture model for mechanistic and drug studies.

Future Directions

This study uncovers mechanisms of dysregulated insulin secretion and provides a translational platform through long-term human insulinoma cultures, advancing diagnostic and therapeutic strategies and insights into neuroendocrine and metabolic diseases.

Pancreatic neuroendocrine neoplasms (pan-NENs) are rare tumors, accounting for 2% to 3% of all pancreatic neoplasms,¹ yet their incidence has been rising due to advances in diagnostic imaging,2, 3, 4 including endoscopic ultrasonography (EUS), magnetic resonance imaging (MRI), and computed tomography (CT). The World Health Organization (WHO) classification distinguishes well-differentiated neuroendocrine tumors (NETs) from poorly differentiated neuroendocrine carcinomas (NECs), with NETs further graded by proliferative activity (Ki-67 index) into G1, G2, and G3 subtypes. Pan-NENs are also classified as functional or nonfunctional, based on clinical symptoms caused by hormone secretion.

Insulinomas are the most common functional pan-NENs.⁵ Although the majority are solitary, well-differentiated, and typically classified as low or intermediate grade (G1 or G2),5, 6, 7 with favorable prognosis when appropriately managed. However, approximately 7% present as malignant, with liver or lymph node metastases and a poor clinical outcome.⁸ Insulinomas occur either sporadically or as part of hereditary syndromes, such as multiple endocrine neoplasia type 1 (MEN1) or von Hippel-Lindau disease (VHL). Among these, MEN1-associated insulinomas are known for their early onset and frequent multiplicity.⁹ Despite their small size (0.5–2 cm), insulinomas frequently cause symptomatic hypoglycemia, and there is no clear correlation between tumor size and hormone secretion.^10,11 In metastatic cases, insulin-secreting and non-secreting lesions may coexist, complicating the clinical identification of the responsible tumor.¹²

Recent imaging techniques, such as DOTATATE positron emission tomography (PET)/CT and glucagon-like peptide-1 (GLP-1) receptor scintigraphy, are effective in detecting solitary lesions but lack specificity for insulin-secreting tumors due to variable receptor expression.¹³ Selective arterial calcium stimulation testing (SASI test) allows for the localization of insulin-secreting lesions,¹⁴ but it is invasive and only provides spatial resolution at the level of pancreatic vascular territories. Accurate localization of insulin-secreting lesions among multiple tumors therefore remains a clinical challenge.

The only curative treatment is surgical resection. Medical therapies such as diazoxide, somatostatin analogs, and everolimus are available but often suffer from limited efficacy or tolerability issues.6, 7, 8^,15 Peptide receptor radionuclide therapy (PRRT) has recently shown efficacy in functional pan-NETs, but its efficacy in insulinomas is limited due to their weak somatostatin receptor expression.^16,17 As many insulinomas follow a benign course, accurate identification of the insulin-secreting lesion can help avoid unnecessary surgery and reduce treatment invasiveness. Thus, new diagnostic and therapeutic strategies are urgently needed. Progress has been hampered by the rarity of the disease and the lack of robust in vitro models. Recent advances in high-throughput transcriptomics and the development of organoid cultures, particularly those based on intestinal organoid technology,¹⁸ have now enabled detailed genetic and functional studies across various solid tumors,¹⁹ including pancreatic neuroendocrine tumors (pan-NETs).²⁰ In this study, we performed integrative transcriptomic analyses using resected human insulinomas and patient-derived organoid models, identifying dedicator of cytokinesis 10 (DOCK10) as a previously unrecognized regulator of insulin secretion that is preferentially expressed in insulinomas. We further evaluated its role as a diagnostic marker and its potential as a therapeutic target for insulin-secreting lesions.

Results

Transcriptomic Characterization of Pan-NENs Using a Novel Organoid-based Biobank

Organoid cultures derived from surgical specimens allow for the specific expansion of tumor epithelial cells, excluding non-tumor components such as immune cells, stromal fibroblasts, and blood cells. Based on previous protocols,²⁰ we successfully established organoids from both primary pancreatic lesions and liver metastases of insulinoma derived from surgical samples during the study period (Figures 1 and 2A). These insulinoma-derived organoids were positive for insulin (Figure 2B) and expressed neuroendocrine markers such as synaptophysin and CD56, consistent with pathological diagnosis (Figure 2C). In contrast, organoids established from non-tumorous tissue were insulin-negative and positive for ductal epithelial markers including keratin 19 (KRT19), mucin 1 (MUC1), pancreatic and duodenal homeobox 1 (PDX1), and SRY-box transcription factor 9 (SOX9), indicating their ductal epithelial origin (Figure 2D). Using this method, we established a single-institution biobank consisting of surgical specimens and organoid cultures from patients with insulinoma and non-functional neuroendocrine tumors (NF-NETs) collected (Figure 1). Clinical characteristics of surgically resected primary pancreatic insulinoma and NF-NET cases over the past 5 years confirmed previous findings, showing that insulinoma lesions tend to be small, solitary, and histologically low-grade (Figure 1; Table 1). In total, we submitted 97 surgical samples and 22 organoid lines for bulk RNA sequencing (RNA-seq). The breakdown of surgical specimens included 32 normal pancreas samples, 42 NF-NETs, and 23 insulinomas. Among the organoids, 11 were derived from normal tissue, 7 from NF-NETs, and 4 from insulinomas (Table 2). Among the samples submitted for RNA sequencing, MEN1-associated cases were relatively frequent (Table 3). To investigate the mechanisms underlying insulin hypersecretion, we included all tumors that were histologically and clinically confirmed as insulinomas after surgical resection, regardless of tumor background, including MEN1-associated cases, liver metastases, and primary tumors. Principal component analysis (PCA) of the RNA-seq data revealed distinct transcriptional profiles between tumor and normal tissues in surgical samples (Figure 3A). When the analysis was limited to tumor samples, no clear separation was observed between primary pancreatic and liver metastatic sites in either insulinoma or NF-NET, and thus, these were combined for subsequent analyses (Figure 3B). Although insulinomas and NF-NETs could not be clearly separated by PCA (Figure 3B), heatmap clustering further demonstrated that both tumor types were transcriptionally distinct from normal pancreatic tissue, although they shared some overlapping expression profiles (Figure 3C). Differential expression analysis between insulinoma and normal tissue identified 2141 upregulated and 1192 downregulated genes in insulinomas (Figure 3D). Insulin is synthesized in the cytoplasm, stored in secretory granules, and secreted in response to various physiological stimuli. Accordingly, the insulin pathway can be broadly divided into 3 steps: (1) insulin biosynthesis; (2) insulin secretion; and (3) paracrine signaling mediated by hormones such as somatostatin and GLP-1.²¹ Functional analysis revealed a significant enrichment of genes involved in the insulin secretion pathway among those upregulated in insulinoma (Figure 3E), including ABCC8, CACNA1A, CACNA1C, CACNA1D, GCK, KCNJ11, RAB3A, SNAP25, STX1A, STXBP1, SYT7, and UNC13A (Figure 4). In contrast, the majority of genes associated with synthesis and paracrine pathways showed no significant changes in expression (Figure 4), suggesting that dysregulation of the insulin secretion machinery—rather than insulin biosynthesis itself—may underlie the clinical phenotype of insulinoma.

Figure 1.

Chronological overview of the case cohort used in this study. This diagram provides an overview of the case cohort and sample usage in this study. The dashed box on the upper right indicates the subset of patients with primary pan-NETs resected between January 2020 and January 2025, which were included in clinical feature analyses (n = 74). The pan-NEN biobank includes all surgically resected pan-NEN cases from May 2014 to January 2025. It comprises both tissue samples and organoids that could be established from the respective resected specimens. Bulk RNA-seq was performed on 97 surgical specimens and 22 organoid samples, all derived from 68 cases between May 2014 and September 2022, from which sufficient RNA could be obtained. A subset of cases was also subjected to qPCR (n = 11), immunostaining analysis (n = 12), and multiplex immunostaining in morphologically heterogeneous insulinomas (n = 2). Additionally, scRNA-seq (n = 1) and 2D organoid culture (n = 3) were performed using samples from the biobank.

Figure 2.

A biobank of pan-NEN organoids derived from tumor and adjacent normal tissues. (A) Schematic overview of organoid establishment using surgical specimens derived from tumor and adjacent normal tissue. (B) Immunofluorescent staining for insulin in organoid cultures from insulinoma: primary insulinoma (left), liver metastasis (right). (C) Immunofluorescent staining for synaptophysin (left) and CD56 (right) in organoids from insulinoma. Scale bar: 500 μm in (A); 100 μm in (B and C). (D) H&E staining (top row, first column). Immunofluorescence for insulin (top row, second column), KRT19 (top row, third column), MUC1 (top row, fourth column), PDX1 (bottom row, first column), and SOX9 (bottom row, second column), with DAPI. Scale bars: 500 μm in top row, first columns; 100 μm in all other columns.

Table 1.

Summary of Insulin Immunostaining in Surgical Samples of Pan-NETs Performed at our Hospital Over the Past 5 Years

	Total	Insulinoma	NF-NET	P value
No. cases	74	17	57
Age, y	47 (59–69)	54 (43–64)	61 (48–70)	.26
Sex, male/female	35/39	6/11	29/28	.26
Tumor size, mm	8 (11–17.8)	11 (10–13)	10 (8–19)	.29
MEN1	4 (5.4)	2 (11.8)	2 (3.5)	.19
Tumor main location				.03
Pancreatic head	19 (25.7)	6 (35.3)	13 (22.8)
Pancreatic body	18 (24.3)	0 (0.0)	18 (31.6)
Pancreatic tail	37 (50.0)	11 (64.7)	26 (45.6)
Multiple lesion	13 (17.6)	2 (11.8)	11 (19.3)	.47
Grading				.42
NETG1	57 (77.0)	15 (88.2)	42 (73.7)
NETG2	15 (20.3)	2 (11.8)	13 (22.8)
NETG3	2 (2.7)	0 (0.0)	2 (3.5)
Insulin immunostaining	36 (48.6)	16 (94.1)	20 (35.1)	<.01

NOTE. Data are presented as number (%) or median (interquartile range).

NOTE. Boldface P values indicate statistical significance.

pan-NET, pancreatic neuroendocrine tumor.

Table 2.

Summary of Surgical and Organoid Samples for RNA-seq

	Normal	Tumor
		Pancreas	Liver
Surgical samples (n = 97)
Insulinoma	8	15	8
NF-NET	24	27	15
Organoid samples (n = 22)
Insulinoma	1	2	2
NF-NET	10	5	2

NF-NET, non-functional neuroendocrine tumor; RNA-seq, RNA sequencing.

Table 3.

Summary of Cases Submitted to RNA-seq Analysis, Including Both Primary and Metastatic Tumor Samples, as well as Organoid Lines

	Total	Insulinoma	NF-NET	P value
No. cases	68	19	49
Age, y	59.5 (47–69.3)	59 (47–69.5)	59.5 (47–69.8)	.73
Sex, male/female	30/38	8/11	22/27	.84
MEN1	8 (11.8)	5 (26.3)	3 (6.1)	.02
Multiple lesions	21 (30.9)	7 (36.8)	14 (28.6)	.65
Grading				.97
NETG1	38 (55.9)	11 (57.9)	27 (55.1)
NETG2	22 (32.4)	6 (31.6)	16 (32.7)
NETG3	8 (11.8)	2 (10.5)	6 (12.2)

NOTE. Data are presented as number (%) or median (interquartile range).

NOTE. Boldface P values indicate statistical significance.

NF-NET, non-functional neuroendocrine tumor; RNA-seq, RNA sequencing.

Figure 3.

Transcriptomic profiling of surgical specimens. (A) Summary of surgical samples for RNA-seq. PCA plot of all surgical samples. Tumor includes both primary tumor and liver metastasis. (B) PCA plot of tumor samples only, including primary and liver metastases from both insulinoma and NF-NET. (C) Heatmap of DEGs across insulinoma, NF-NETs, and normal pancreas, showing row-wise z-scores of normalized counts. (D) Volcano plot of DEGs between insulinoma samples and normal pancreatic tissue obtained from surgical specimens. (E) Functional enrichment analysis of DEGs between insulinoma and normal pancreatic tissue from surgical specimens. Dot size represents gene ratio; color indicates q-value.

Figure 4.

RNA-seq–based gene expression of the insulin pathway in surgical samples of insulinoma variance-stabilized expression of key insulin signaling genes in insulinoma vs normal pancreas. Adjusted P-values from DGE analysis: ns, not significant; ∗P < .05; ∗∗P < .01; ∗∗∗P < .0001.

Candidate Genes Associated With Dysregulated Insulin Secretion in Insulinomas

To identify candidate genes involved in the dysregulation of insulin secretion in insulinomas, we analyzed the RNA-seq data to extract genes that were consistently and specifically dysregulated in insulinoma tissues and organoids. In organoid samples as well, PCA of the RNA-seq data revealed distinct transcriptional profiles between tumor and normal tissues (Figure 5A). Although 3333 genes were differentially expressed in surgically resected insulinoma samples (Figure 3D), insulinoma-derived organoids exhibited significant changes in the expression of 1596 genes compared with organoids derived from normal pancreatic tissue, including 1070 upregulated and 526 downregulated genes (Figure 5B). A total of 441 genes were found to be commonly differentially expressed in both insulinoma surgical specimens and insulinoma-derived organoids, compared with normal pancreatic tissues and corresponding normal organoids (Figure 6A). Fold change differences of these 441 genes showed a significant correlation between tissue and organoid datasets (Figure 6B; Supplementary Table 1). Among the downregulated genes, those commonly reduced in both tissue and organoids included CPA1, CPA2, PNLIP, CLPS, PNLIPRP1, CTRB1, and CTRC, which are typically secreted from normal exocrine pancreas.^22,23 These genes were excluded from further analysis, and we focused instead on genes upregulated in insulinoma tissue and organoid samples. Within this set, we identified genes whose expression levels were strongly associated with INS (insulin gene) expression (Figure 6C, D). We ultimately selected 4 candidate genes: DOCK10, calcium voltage-gated channel auxiliary subunit alpha2delta 1(CACNA2D1), neurotrimin (NTM), and ephrin A5(EFNA5). Some of these showed strong correlation with INS expression, whereas others were not strongly correlated but were markedly upregulated in insulinoma tissue and organoid samples. Ubiquitin C-terminal hydrolase L1 (UCHL1; also known as protein gene product 9.5 [PGP 9.5]) has been previously reported to be expressed in all islet cells and in 75% of insulinomas.²⁴ Consistently, it was also one of the genes significantly upregulated in our surgical specimens. Therefore, we decided to include UCHL1 as an additional candidate gene. We next validated the expression of these genes in clinical surgical samples using quantitative polymerase chain reaction (qPCR). Comparison among normal pancreas, NF-NETs, and insulinomas revealed that UCHL1, CACNA2D1, and DOCK10 were significantly and specifically upregulated in insulinomas (Figure 7A). Notably, DOCK10 and CACNA2D1 showed higher expression in insulinoma tissues and organoids compared with normal samples, and UCHL1 showed higher expression in insulinoma tissues compared with normal samples in the RNA-seq dataset (Figure 7B). To investigate protein-level expression of these candidate genes, we performed immunohistochemical staining on formalin-fixed surgical tissue sections from normal pancreas, NF-NETs, and insulinomas. CACNA2D1 was expressed in both insulinomas and NF-NETs (Figure 8). In contrast, DOCK10 and UCHL1 were negative in normal and NF-NET tissues, while showing strong and specific positivity in insulinoma tumor areas (Figures 9 and 10). These findings indicate that UCHL1 and DOCK10 are insulinoma-specific markers at both the transcript and protein levels.

Figure 5.

RNA-seq–based transcriptomic profiling of organoid samples. (A) Summary of organoid samples used for RNA-seq. PCA plot of organoid cultures derived from surgical specimens of insulinoma, NF-NET, and normal pancreatic tissue. (B) Volcano plot of DEGs between insulinoma-derived and normal pancreas in organoids.

Figure 6.

Identification of candidate genes regulating aberrant insulin secretion in insulinoma by integrative analysis of surgical samples and organoids. (A) Venn diagram comparing DEGs in insulinoma between surgical samples and organoid cultures, relative to normal pancreas by RNA-seq. (B) Expression patterns of 441 genes significantly differentially expressed in both surgical and organoid samples. (C) Top 10 genes most positively associated with INS expression among commonly upregulated genes. (D) Association between top 10 genes and INS expression (β, LMM effect size).

Figure 7.

Validation and comparison of candidate gene expression. (A) qPCR validation of DOCK10, CACNA2D1, UCHL1, NTM, and EFNA5 expression in surgical samples (insulinoma, n = 5; NF-NET, n = 6; normal, n = 5); mean ± SD; unpaired Student’s t-test; ns, not significant; ∗P < .05; ∗∗P < .01; ∗∗∗∗P < .001. (B) RNA-seq–based expression levels of DOCK10, CACNA2D1, and UCHL1 in surgical and organoid samples; adjusted P-values from DGE analysis; ns, not significant; ∗∗P < .01; ∗∗∗P < .0001.

Figure 8.

Immunofluorescence images of CACNA2D1 in insulinoma and NF-NET surgical samples. Representative images from 3 cases are shown. For each case, the upper left panel shows the H&E image, the upper right panel shows the low-magnification image, and the lower panel shows a high-magnification view of the region indicated by the yellow square. Green represents the protein signal, and blue (DAPI) marks the nuclei. Black dashed lines indicate the boundary between normal and tumor regions. Scale bars: 500 μm (low magnification) and 100 μm (high magnification).

Figure 9.

Immunofluorescence images of DOCK10 in insulinoma and NF-NET surgical samples. Representative images from 3 cases are shown. For each case, the upper left panel shows the H&E image, the upper right panel shows the low-magnification image, and the lower panel shows a high-magnification view of the region indicated by the yellow square. Green represents the protein signal, and blue (DAPI) marks the nuclei. Black dashed lines indicate the boundary between normal and tumor regions. Scale bars: 500 μm (low magnification) and 100 μm (high magnification).

Figure 10.

Immunofluorescence images of UCHL1 in insulinoma and NF-NET surgical samples. Representative images from 3 cases are shown. For each case, the upper left panel shows the H&E image, the upper right panel shows the low-magnification image, and the lower panel shows a high-magnification view of the region indicated by the yellow square. Green represents the protein signal, and blue (DAPI) marks the nuclei. Black dashed lines indicate the boundary between normal and tumor regions. Scale bars: 500 μm (low magnification) and 100 μm (high magnification).

Identification of DOCK10 as a Marker for Insulin Secretion-specific Alterations in Insulinomas

Insulinomas can sometimes present as metastatic lesions, with a mixture of insulin-secreting and non-secreting tumors within the same patient.¹² To identify insulin-secreting lesions among multiple tumors, the SASI test is employed. This diagnostic procedure involves intra-arterial injection of calcium gluconate into pancreatic arterial branches, followed by hepatic venous sampling to detect insulin “step-up” responses, which indicate the location of hyperfunctioning insulinomas. An extended version of this test, the Liver SASI test, involves injecting calcium gluconate via the proper hepatic artery and collecting blood samples from the right, middle, and left hepatic veins to identify functional liver metastases responsible for hormone secretion. We applied this approach in 2 patients with liver metastases to identify and surgically resect the insulin-secreting lesions.¹² In the first case, the pancreatic primary lesion was non-secreting, while insulin secretion originated from hepatic metastases drained by the right and middle hepatic veins. In the second case, the pancreatic lesion was insulin-secreting, while the hepatic metastases were non-secreting. In both cases, surgical resection of the insulin-secreting lesions led to significant symptom relief (Table 4). Given the distinct localization of functional and non-functional tumors in these cases, we investigated whether immunofluorescent staining for DOCK10 and UCHL1 could discriminate insulin-secreting from non-secreting tumors within the same patient. Fluorescent immunohistochemistry showed that the pattern of insulin staining did not consistently correspond to the clinical insulin-secreting phenotype. In contrast, DOCK10 was strongly and specifically positive only in insulin-secreting lesions, whereas UCHL1 was expressed in both secreting and non-secreting lesions (Figures 11 and 12). These findings suggest that DOCK10 can serve as a specific marker for insulin-secreting tumor populations in heterogeneous insulinomas.

Table 4.

Clinical Details of 2 Insulinoma Cases With Multiple Metastases, Where Insulin-secreting and Non-secreting Lesions Were Distinguished Using SASI and Liver SASI Tests

Case no.	Age, y	Sex	WHO classification	Hormonal tumor mapping					Resected lesion
				Pancreas		Liver			Pancreas	Liver
				Head	Body/ tail	RHV	MHV	LHV
1	39	M	NET G2	X	X	O	O	X	DP	Right hepatectomy
										Partial hepatectomy (Segment 1, 3, 4)
2	71	F	NET G2	O	X	X	X	X	SSPPD	Partial hepatectomy (segment 5)

DP, distal pancreatectomy; F, female; LHV, left hepatic vein; M, male; MHV, middle hepatic vein; RHV, right hepatic vein; SASI test, selective arterial secretagogue injection test; SSPPD, subtotal stomach-preserving pancreatoduodenectomy; WHO, World Health Organization.

Figure 11.

Expression of DOCK10 in multiple metastatic lesions containing a mixture of insulin-secreting and non-secreting tumors. Immunofluorescence staining of DOCK10 (green), insulin (red), and DAPI (blue) in multiple metastatic insulinoma lesions. Blue circle: insulin-secreting tumor; gray circle: non-secreting tumor. Scale bar: 200 μm.

Figure 12.

Expression of UCHL1 in multiple metastatic lesions containing a mixture of insulin-secreting and non-secreting tumors. Immunofluorescence staining of UCHL1 (green), insulin (red), and DAPI (blue) in multiple metastatic insulinoma lesions. Blue circle: insulin-secreting tumor; gray circle: non-secreting tumor. Scale bar: 200 μm.

To further explore intratumoral heterogeneity at the single-cell level, we performed single-cell RNA sequencing (scRNA-seq) on a surgical specimen from a patient with insulinoma. The tumor measured 10 mm in diameter, was classified as G1, and was identified as a solitary insulin-secreting tumor based on the SASI test. Following previously reported marker gene expression profiles,^25,26 with chromogranin A (CHGA) specifically marking tumor cells, we clustered cells into tumor, stromal, and immune populations (Figure 13A and B). Graph-based clustering revealed that tumor cells could be divided into 2 distinct clusters (Cluster 1 and Cluster 2) based on their gene expression profiles (Figure 13C). A total of 808 genes were differentially expressed between these clusters (Figure 13D; Supplementary Table 2). DOCK10 was predominantly expressed in Cluster 1 and was nearly absent in Cluster 2, but UCHL1 was expressed in both clusters (Figure 13E). ZNF385D, a recently reported novel marker of β cells,²⁷ was also preferentially expressed in Cluster 1 (Figure 13E). Moreover, genes involved in insulin synthesis and secretion pathways were highly expressed in Cluster 1, indicating that this cluster represents the insulin-secreting population (Figure 13F and G). These results demonstrate that insulinomas exhibit transcriptional heterogeneity and can be subdivided into at least 2 distinct tumor cell populations, one of which is characterized by high insulin secretion and DOCK10 expression. Immunofluorescence staining of normal pancreatic tissue also showed positive DOCK10 expression in the islets of Langerhans, where it co-localized with insulin, strongly suggesting a potential link between DOCK10 and insulin secretion (Figure 14).

Figure 13.

DOCK10 marks the high insulin-secreting tumor population at the single-cell level. (A) scRNA-seq of 2984 cells from primary insulinoma, visualized by t-SNE and k-means clustering into 8 clusters. (B) Expression of selected marker genes overlaid on t-SNE map. (C) t-SNE of 370 tumor cells classified into Clusters 1 and 2 by graph-based clustering. (D) Heatmap of 808 genes differentially expressed between Clusters 1 and 2. (E) Violin plot of DOCK10, UCHL1, and ZNF385D expression in Clusters 1 and 2. (F) Violin plots of insulin synthesis pathway genes between Clusters 1 and 2. (G) Violin plots of insulin secretion pathway genes between Clusters 1 and 2. Statistical significance was corrected by Benjamini–Hochberg; ns, not significant; ∗∗P < .01; ∗∗∗∗P < .001.

Figure 14.

Expression of DOCK10 in pancreatic islets. H&E staining of pancreatic islets (left), DOCK10 immunofluorescence (second from left), insulin immunofluorescence (third from left), and the merged image of DOCK10, insulin, and DAPI (right). Green: DOCK10; red: insulin; blue (DAPI): nuclei. Scale bar: 50 μm. Analysis was performed on normal regions from the same specimens as in Figure 9.

DOCK10 Regulates Insulin Secretion via Cdc42 in MIN6 Cells

DOCK10 is a guanine nucleotide exchange factor (GEF) that activates small Rho GTPases, particularly Cdc42 and Rac1.²⁸ Although it has been implicated in cytoskeletal remodeling, cell migration, and B-cell differentiation,^29,30 its role in pancreatic function remains poorly understood. Previous studies have shown a positive correlation between DOCK10 expression and insulin secretion in human pancreatic islet transcriptomes,³¹ and data from the Human Protein Atlas suggest its expression in pancreatic β-cells. However, its molecular function in insulin secretion has not been elucidated. To explore the functional role of DOCK10 in insulin secretion, we used the murine insulinoma cell line MIN6, which retains the ability to secrete insulin in response to physiological glucose concentrations.³² Using small interfering RNA (siRNA)-mediated knockdown of Dock10 (Figure 15A), we performed static insulin secretion assays. Knockdown of Dock10 significantly suppressed glucose-stimulated insulin secretion (Figure 15B), whereas intracellular insulin content remained unchanged (Figure 15C), suggesting that DOCK10 regulates insulin release rather than synthesis. To further investigate the underlying mechanism, we examined the Cdc42-Pak1 pathway, which has been reported to mediate actin cytoskeleton remodeling and facilitate insulin granule exocytosis in MIN6 cells.³³ DOCK10 is known to activate Cdc42²⁸; therefore, we examined whether Dock10 knockdown affected Pak1 phosphorylation in response to glucose stimulation (Figure 15D). As expected, glucose-induced phosphorylation of Pak1 was diminished in Dock10-silenced cells (Figure 15E). Next, we evaluated whether pharmacologic inhibition of Cdc42 had similar effects. Treatment with the selective Cdc42 inhibitor ML141 resulted in suppression of Pak1 phosphorylation comparable to that observed with Dock10 knockdown (Figure 15F). Furthermore, ML141 significantly suppressed insulin secretion (Figure 15G), confirming the role of the DOCK10-Cdc42-Pak1 axis in insulin exocytosis. To evaluate the in vivo relevance of this pathway, we established a subcutaneous transplantation model using MIN6 cells in C57BL/6J mice, from which the cell line originates. In accordance with previous studies, only female mice were used throughout the experiments.³⁴ Based on previous reports indicating that tumor engraftment typically requires approximately 2 weeks,³⁴ we performed random blood glucose measurements, 4-hour fasting glucose assessments, and oral glucose tolerance tests (OGTTs) via gastric tube at 2 weeks post-transplantation (Figure 16A). Clinically, the diagnosis of human insulinoma relies on Whipple’s triad, which includes a fasting blood glucose level below 50 mg/dL. In many cases, even postprandial glucose levels remain within a relatively narrow range of approximately 60 to 100 mg/dL. In our model, MIN6-transplanted mice with 4-hour fasting glucose levels above 100 mg/dL (n = 9) exhibited no significant differences in basal glycemia or glycemic responses during OGTT compared with control mice (n = 8) (Figure 16B and C). Therefore, animals with fasting glucose levels below 100 mg/dL were considered to be successfully engrafted with insulin-secreting MIN6 cells and were used for subsequent analyses (n = 14). These mice showed significantly lower random and fasting glucose levels than control animals (n = 8) (Figure 16D) and displayed impaired glycemic responses during OGTTs, consistent with excessive insulin action (Figure 16E). Intraperitoneal administration of ML141 partially attenuated the hypoglycemic phenotype during OGTTs, although this effect did not reach statistical significance (Figure 16F). As previously reported, subcutaneous transplantation of MIN6 cells can result in fatal hypoglycemia.³⁴ Notably, long-term ML141 treatment (n = 7) significantly prolonged survival in MIN6-transplanted mice (Figure 16G). Subsequently, we evaluated serum insulin levels in mice after a 4-hour fast. Measurement of serum insulin in MIN6-transplanted (n = 9) and control mice (n = 8) revealed that MIN6-transplanted mice exhibited elevated insulin levels even in the fasted state (Figure 16H). Furthermore, comparison of the relative changes in serum insulin levels after a 4-hour fast before and after treatment in MIN6-transplanted mice administered ML141 (n = 4) or vehicle (n = 3) showed that insulin levels were markedly suppressed in the MIN6-transplanted mice (Figure 16I). This suggests that targeting the DOCK10-Cdc42 axis may modulate insulin secretion in vivo. Although improvement in blood glucose was not observed at the time of measurement, suppression of insulin secretion likely contributed to the extended survival of ML141-treated mice.

Figure 15.

DOCK10 regulates insulin secretion via CDC42 in MIN6 cells. (A) Dock10 mRNA levels by qPCR in control and DOCK10-knockdown (KD) MIN6 cells. (B) ELISA measurement of insulin secretion after stimulation with 3 mM glucose, 25 mM glucose, or 3 mM glucose + 30 mM KCl. (C) ELISA measurement of intracellular insulin levels in control and knockdown MIN6 cells (n = 3/group). (D) Schematic of Dock10–Cdc42 pathway regulating GSIS; ML141, a Cdc42 inhibitor. (E) Protein levels of phosphorylated PAK1 (p-PAK1), total PAK1, and β-actin after glucose stimulation (2-hour fasting) in control and knockdown cells by Simple Western; table shows p-PAK1/PAK1 ratio normalized to β-actin. (F) Protein levels after 1-hour fasting and ML141 or vehicle (DMSO) pretreatment, assessed by Simple Western; table shows normalized p-PAK1/PAK1 ratios. (G) ELISA measurement of insulin secretion after ML141 or vehicle pretreatment under the same stimulations (n = 3/group). Data were pooled from 3 independent experiments for panels in (A–C) and (G). Values are presented as mean ± SD. Statistical analysis was performed using unpaired Student’s t-test. ns, not significant; ∗∗P < .01; ∗∗∗P < .005.

Figure 16.

ML141, an inhibitor of CDC42, improves outcomes in an insulinoma mouse model. (A) Experimental timeline of OGTT and ML141 treatment following MIN6 or vehicle (Matrigel) transplantation. (B) Random blood glucose in unsuccessfully engrafted MIN6 mice (n = 9) and controls (n = 8). (C) Blood glucose time course during OGTT in unsuccessfully engrafted MIN6 mice (n = 9) and controls (n = 8) at 2 weeks post-transplantation. (D) Random blood glucose levels in MIN6-transplanted (n = 14) vs control mice (n = 8). (E) Blood glucose time course during OGTT at 2 weeks post-transplantation. (F) OGTT time course after ML141 (1 mg/body) (n = 7) or vehicle (n = 6) administration. (G) Survival curves of transplanted mice treated with ML141 or vehicle (n = 7/group); significance by log-rank test. (H) ELISA measurement of serum insulin levels after 4-hour fasting in MIN6-transplanted (n = 9) vs control mice (n = 8). (I) Relative post/pre changes in 4-hour fasting serum insulin levels (ELISA) in MIN6-transplanted mice administered ML141 (n = 4) or vehicle (n = 3). Data in (B–F, H–I) are mean ± SD; significance by unpaired Student’s t-test. ns, not significant; ∗P < .05; ∗∗P < .01; ∗∗∗P < .005; ∗∗∗∗P < .001.

ML141 Suppresses Insulin Secretion in Human Insulinomas

To evaluate whether ML141, a selective Cdc42 inhibitor, could also suppress insulin secretion in human insulinomas, we applied the compound to patient-derived insulinoma organoids. Organoids were cultured in 3D Matrigel domes and confirmed to secrete insulin via enzyme-linked immunosorbent assay (ELISA)-based assays (Figure 17A and B). However, with serial passaging, a progressive decline in insulin secretion capacity was observed, limiting long-term functional assays. To overcome this limitation, we optimized a long-term culture protocol. We observed that conventional 3D organoid cultures not only lost functional β-cell properties over time but also degraded the Matrigel matrix, likely due to exocrine enzyme secretion. We therefore adapted the protocol to a 2D monolayer culture system without passaging, which preserved insulin secretory function for over 1 month (Figure 17C and D). Using this optimized 2D culture system, we investigated the functional effects of ML141 on insulin secretion. Organoids were established from surgical specimens of 3 patients with pathologically confirmed insulinoma (Figure 18; Table 5). Treatment with ML141 resulted in a significant reduction in insulin secretion in all 3 organoid lines (Figure 19A and B), indicating that Cdc42 inhibition suppresses insulin release in human insulinomas as well. Taken together, these findings suggest that the DOCK10–Cdc42 signaling axis is an important regulator of insulin hypersecretion in insulinoma and represents a promising therapeutic target. Pharmacological inhibition of this pathway may offer a novel strategy to manage hypoglycemia in patients with inoperable or metastatic insulinoma.

Figure 17.

Comparison between 3D and 2D cultures of insulinoma organoids. (A) Images of human insulinoma organoids in 3D culture: Day 1 (passage 0) and Day 27 (passage 2). (B) ELISA measurement of C-peptide levels in 3D organoids from insulinoma and normal tissue at day 13, 20, and 27. (C) Images of human insulinoma organoids in 2D culture: Day 1 and Day 27. (D) ELISA measurement of C-peptide levels in 2D organoids from insulinoma and normal tissue at day 13, 20, and 27.

Figure 18.

Pathological diagnosis of insulinoma cases used for 2D culture. Staining for H&E, CHGA, synaptophysin, and insulin, top to bottom. Scale bar: 100 μm.

Table 5.

Clinical Details of 3 Insulinoma Cases Used for 2D Monolayer Organoids

Case no.	Age, y	Sex	Size, mm	WHO Classification	Hormonal tumor mapping		Resected lesion
					Pancreas		Pancreas
					Head	Body/tail
1	39	M	10	NET G1	N/A	N/A	DP
2	71	F	13	NET G1	O	X	SSPPD
3	38	F	30	NET G1	X	O	DP

DP, distal pancreatectomy; F, female; M, male; N/A, not available; SSPPD, subtotal stomach-preserving pancreatoduodenectomy; WHO, World Health Organization.

Figure 19.

ML141 suppresses insulin secretion in human insulinomas. (A) Schematic timeline of 2D organoid establishment and medium collection before and after ML141 treatment. (B) Relative C-peptide changes in 2D insulinoma organoids treated with ML141 or vehicle, measured by ELISA. Post-treatment levels were compared with pre-treatment. Data are mean ± SD (n = 3/group); significance by unpaired Student’s t-test. ∗P < .05; ∗∗∗∗P < .001.

Discussion

In this study, we investigated the molecular mechanisms underlying dysregulated insulin secretion in insulinoma by integrating transcriptomic profiling and functional assays using clinical samples and patient-derived organoids. Bulk RNA-seq of tumor tissues and organoids identified DOCK10 as a candidate gene associated with insulin hypersecretion. Its expression was validated through qPCR, immunohistochemistry, and SASI-based localization analyses. scRNA-seq, performed for the first time in insulinoma, revealed intratumoral heterogeneity, with DOCK10 expression restricted to insulin-secreting tumor cell populations. Functional assays using the murine insulinoma cell line MIN6 further supported a regulatory role for DOCK10 in insulin secretion.

To enable long-term culture of insulinoma cells, we adapted our organoid protocol to a 2D monolayer format, which maintained insulin-secreting capacity for over 1 month. Using this model, we showed that pharmacological inhibition of the DOCK10 downstream effector Cdc42 suppressed insulin secretion, highlighting the therapeutic potential of targeting this axis in insulinoma.

Although previous studies have reported gene expression analyses in insulinoma, they generally lacked functional validation and did not explore potential marker genes in depth.^25,35,36 Our study is the first to combine transcriptomic and functional analyses in both tissues and organoids from human insulinomas. Nonetheless, our study has limitations. The sample size was small, and scRNA-seq analysis was performed on only one specimen. These findings require validation in larger cohorts.

We acknowledge that the comparisons in this study were based on surgical specimens and organoids derived from tumor-bearing pancreata, and that the so-called “normal” tissues were obtained from non-tumorous regions of resected pan-NEN specimens, which may not fully represent a healthy pancreas. Moreover, organoids established from these non-tumorous regions predominantly expressed ductal markers, indicating a ductal epithelial origin. Ideally, pancreatic β cells would serve as a more appropriate comparator; however, the limited availability of normal tissue in surgical samples and technical challenges in isolating and culturing primary β cells precluded this approach.

Despite these limitations, our model appears to be relevant for studying insulin hypersecretion in insulinoma. We observed consistent gene expression profiles between insulinoma tissues and corresponding organoids, and insulin secretion was confirmed by ELISA. These findings suggest that our culture system successfully retains key functional properties of the original tumors. The sustained culture of insulin-secreting human cells using this 2D organoid format may also offer opportunities for disease modeling and therapeutic testing. As culture systems continue to improve, insulinoma-derived cells could even be explored as a potential source for transplantation therapies.

Although β-cell transplantation and iPSC-derived β-cell generation have been explored for diabetes therapy, these approaches remain costly and technically challenging, particularly in terms of long-term maintenance.³⁷ In contrast, our optimized 2D organoid system offers a more practical and stable platform for functional studies. However, 2D and 3D organoids differ in polarity, cell–cell and cell–matrix interactions, nutrient and oxygen gradients, and drug sensitivity, which could affect cytoskeletal dynamics and vesicle trafficking.³⁸ Future experiments using 3D or in vivo models will be important to validate and refine the physiological relevance of our findings. In addition, organoids may modestly reduce intratumoral heterogeneity and do not fully reflect the complex microenvironment of primary tumors or liver metastases, which could have slightly influenced the effects of ML141. Nonetheless, given the cellular heterogeneity observed in scRNA-seq, further studies using organoid-based single-cell transcriptomic analyses will be essential to more precisely define the cellular composition of these models and to refine their physiological relevance.

The diagnosis of insulinoma currently relies on clinical criteria such as Whipple’s triad and hormone testing to determine whether the tumor, as a whole entity, is secreting insulin. However, insulin-positive cells can also be found in NF-NETs. In our pathological analyses of primary pan-NETs, 35.1% of NF-NET specimens were immunopositive for insulin (Table 1). This finding suggests that even in primary NF-NETs, there may exist cells with the potential to become clinically insulin-secreting lesions. Consistent with previous reports, immunohistochemical insulin expression tended to be preserved in localized insulinomas but was often reduced in metastatic or large, high-grade lesions.^39,40 Accordingly, the clinical insulin-secreting phenotype in multiple metastatic lesions did not completely correspond to the pattern of insulin staining (Figures 11 and 12). In such cases, DOCK10 can classify insulinomas into 2 subgroups: one with high insulin-secreting capacity and the other with low or absent secretion. Our scRNA-seq analysis suggested that DOCK10 can classify insulinomas into a group with high insulin-secreting capacity and the other group. Importantly, clinical observations show that non-secreting primary tumors can sometimes give rise to insulin-secreting metastatic lesions. These findings suggest that the presence of DOCK10-positive cells capable of insulin secretion may underlie this behavior: once these cells metastasize, they can expand to become the dominant population, leading to symptomatic lesions. Conversely, even if the primary tumor is insulin-secreting, metastatic lesions may remain non-secreting if the proportion of DOCK10-positive cells is low. To validate this hypothesis, additional scRNA-seq data and in vivo metastasis models will be required. Nonetheless, our biobank and experimental framework lay the groundwork for studying NET pathogenesis, drug screening, and personalized medicine.

Although we identified DOCK10 as an important regulator of insulin hypersecretion, it is likely not the sole factor involved. DOCK10 has previously been implicated in B cell differentiation, and knockout mice exhibit only mild phenotypes. Therefore, therapeutic targeting of the DOCK10 pathway must consider potential effects on the immune system, possibly requiring local (intra-pancreatic) drug delivery strategies.

DOCK proteins are involved in various diseases including cancer, neurological disorders, and immune dysfunctions.⁴¹ Among the 11 mammalian DOCK proteins, DOCK10 belongs to subfamily D and has been shown to activate both Cdc42 and Rac1, which are associated with cancer metastasis.⁴² Despite these associations, no prior studies have reported a role for DOCK10 in pancreatic insulin secretion. However, our immunofluorescence analysis showed that DOCK10 was expressed in the islets of normal pancreatic tissue (Figure 14), strongly suggesting its involvement in insulin secretion. Cdc42 activation is known to occur during the first phase (∼2–3 minutes) of glucose-stimulated insulin secretion (GSIS) in β-cells⁴³ and is involved in cytoskeletal remodeling required for the second phase of insulin granule translocation.^44,45 Mice with β-cell–specific Cdc42 knockout (Rip-Cdc42cKO) show impaired glucose tolerance and reduced GSIS despite normal islet morphology.⁴⁶ To our knowledge, this is the first study to demonstrate that DOCK10 regulates insulin secretion and that inhibition of this pathway suppresses hypersecretion in both murine and human insulinoma models. In ML141-treated mice, fasting blood glucose levels remained unchanged, whereas insulin secretion was reduced. Although overall insulin levels increased, possibly reflecting the influence of the OGTTs performed between samplings, the relative decrease in secretion suggests that ML141 effectively suppressed insulin release in vivo. Although no selective DOCK10 inhibitors are currently available, pharmacologic inhibition of its downstream effectors is under investigation. For example, MBQ-167—a dual Rac/Cdc42 inhibitor—has shown antitumor efficacy in preclinical models and is currently undergoing evaluation in a phase I clinical trial for advanced breast cancer. Nevertheless, the clinical safety of systemic Cdc42 inhibition remains uncertain. Given the widespread physiological roles of Cdc42 in cell polarity, cytoskeletal dynamics, and immune function, future therapeutic strategies will likely require targeted drug delivery approaches and careful evaluation of immune-related side effects.

Our findings suggest that insulin hypersecretion—rather than increased insulin production—is the key pathological feature in insulinoma. This mechanism has been underexplored, and DOCK10 is unlikely to be the sole contributor. Larger-scale studies and novel analytical frameworks will be required to fully elucidate the regulatory network. Notably, although insulin secretion has been extensively studied in diabetes, insulinomas have rarely been used as a model for hypersecretion. Our work provides a new angle that could help inform therapeutic strategies for both insulinoma and diabetes.

In conclusion, we established a transcriptomic and functional framework for studying human insulinoma, identified DOCK10 as a regulator of insulin hypersecretion, and demonstrated that its downstream pathway may be targetable. These insights may enhance the molecular diagnosis and therapeutic targeting of insulinoma, while also providing a new paradigm for investigating insulin regulation in both tumor biology and metabolic disease.

Materials and Methods

Ethics Approval

Human tumor samples were obtained from patients who underwent surgery. The study was approved by the Ethics Committee of the Institute of Science Tokyo (approval numbers: M2000-1080 and M2020-306). Written informed consent was obtained from all patients. For previously collected samples and patient information, an opt-out system was used in accordance with institutional and ethical guidelines. All animal experiments were approved by the Animal Study Committee of the Institute of Science Tokyo (approval numbers: A2024-082C3) and conducted in compliance with institutional and national regulations.

Antibodies and Reagents

Details are provided in Table 6.

Table 6.

Antibodies Used for Immunofluorescence Staining and Western Blotting

	Company	Product code	Dilution
Primary antibodies for immunostaining
Insulin	Cell Signaling Technology	4590S	1:200
Synaptophysin	Dako	M7315	1:1
NCAM (CD56)	Gene Tex	GTX01917	1:100
KRT19	Abcam	AB220193	1:200
MUC1	Cell Signaling Technology	4538S	1:200
Pdx1 XP	Cell Signaling Technology	5679S	1:400
SOX9	Abcam	AB185230	1:2000
DOCK10	Atlas antibodies	HPA058106	1:2500 with amplification
UCHL1	Cell Signaling Technology	Anti-DOCK10	1:200
CACNA2D1	R&D Systems	7666-MB	1:100 with amplification
Secondary antibodies for immunostaining
AlexaFluor-488 goat anti-rabbit IgG	Thermo Fisher	A32731	1:1000
AlexaFluor-594 donkey anti-rabbit IgG	Thermo Fisher	A21207	1:1000
AlexaFluor-488 goat anti-mouse IgG	Thermo Fisher	A11001	1:1000
Primary antibodies for Western blotting
PAK1	Cell Signaling Technology	2602S	1:100
phospho PAK1	Cell Signaling Technology	2601S	1:10
βactin	Abcam	AB6276	1:100
Secondary antibodies for Western blotting
Anti-rabbit secondary HRP antibody	Bio-Techne	042-206	1:1
Anti-mouse secondary HRP antibody	Bio-Techne	042-205	1:1

Inhibitors used in this study
Inhibitor	Company	Product code
ML141	Selleck Chemicals LLC	S7686

Establishment of Organoid Cultures From Pan-NENs and Normal Tissues

Pan-NEN and normal tissue samples were obtained by surgical resection. Organoid cultures were established using a previously reported method.²⁰ Surgical specimens were washed with phosphate-buffered saline (PBS), minced into ∼1 mm fragments, and digested with Advanced Dulbecco’s Modified Eagle Medium (DMEM)/F12 (Thermo Fisher Scientific) containing 1 mg/mL collagenase type II (Gibco) and 125 μg/mL Dispase II (Thermo Fisher Scientific) at 37°C for 1 to 2 hours with agitation. The suspension was filtered through a 100-μm cell strainer, centrifuged, and resuspended in Advanced DMEM/F12. After a second centrifugation, cells were embedded in Matrigel (Corning) and seeded into 24-well plates. Organoids were cultured in Tumor or Normal medium (Table 7) under 5% CO₂ and 20% O₂, with medium changes every 2 to 3 days. For 2D culture, plates were precoated with 10% Matrigel in PBS, and cells were directly seeded in medium.

Table 7.

Factors and Reagents Used in the Organoid Culture Medium in This Study

Material	Company	Product code	Final concentration	Final concentration
			Normal pancreas	Tumor
Advanced-DMEM	Thermo Fisher	12634-010	1×	1×
GlutaMAX	Gibco	35050-061	2 mM	2 mM
Penicillin/streptomycin	nacalai tesque	26253-84	100 U/mL	100 U/mL
Hepes	nacalai tesque	17557-94	10 mM	10 mM
B27 supplement	Thermo Fisher	17504044	1×	1×
N-Acetylcystein	Sigma-Aldrich	A9165-5G	1.25 mM	1.25 mM
Gastrin1	Sigma-Aldrich	G9145	10 nM	10 nM
Recombinant murine EGF	Peprotech	31509	50 ng/mL	50 ng/mL
Recombinant mouse Noggin	R&D Systems	1967	25 ng/mL	25 ng/mL
A-83-01	Tocris	2939	5 μM	500 nM
Recombinant mouse R-spondin1	R&D Systems	3474-RS	1 μg/mL	1 μg/mL
N2 supplement	Thermo Fisher	17502048	1×
Recombinant Human FGF10	Peprotech	100-26	100 ng/mL
Nicotinamide	Sigma	N0636	10 mM
FSK	Tocris	1099	10 μM
PGE2	Tocris	2296	3 μM
SB202190	Sigma-Aldrich	S7067-5mg		10 μM
Recombinant Mouse Wnt-3a	R&D Systems	1324-WN		300 ng/mL
Y27632	Wako	030-24021	10 μM	10 μM
Matrigel	Corning	356231	–	–

Evaluation of Insulin Secretion by C-Peptide ELISA

The level of insulin secretion from organoid cultures was assessed by measuring C-peptide in the culture supernatant. Culture medium was collected at indicated time points and analyzed using a commercially available C-peptide ELISA kit (Mercodia; #10-1141-01), according to the manufacturer’s instructions.

MIN6 Cell Culture

The MIN6 cell line, derived from mouse pancreatic insulinoma, was purchased from AddexBio. MIN6 cells were cultured as described previously.³² Cells were cultured in DMEM (Thermo Fisher Scientific) with 25 mM glucose, 15% heat-inactivated fetal bovine serum (FBS), 1% penicillin-streptomycin (Pen/Strep), and 50 μM 2-mercaptoethanol, and maintained as previously described.³² Unless otherwise stated, assays were performed at 70% to 80% confluence.

Electroporation

Electroporation was conducted using the Nucleofector 4D 96-well strip or 4D-X single-cuvette device (Lonza; #V4SC-2096 and #V4XC-2012) according to the manufacturer’s instructions. The SF solution and CM138 program were used, based on a previous report,^32,47 to achieve siRNA-mediated knockdown. siRNA sequences are listed in Table 8.

Table 8.

Primer Sequences for qPCR and Information on siRNAs Used for Gene Knockdown

Target	Species	Forward/Reverse	Sequence (5'-3')
DOCK10	Human	Forward	GAAGAGCAGTGAGAAGACGGAC
		Reverse	TCAGGAGACTCCTGGTTTCTGC
UCHL1	Human	Forward	CAGTTCAGAGGACACCCTGCTG
		Reverse	CCACAGAGCATTAGGCTGCCTT
CACNA2D1	Human	Forward	GCTATTCACGGATGGAGGAGAAG
		Reverse	CCATCCACTGAATAGGTCCTCTG
EFNA5	Human	Forward	CTCCAAATGGACCGCTGAAGTTC
		Reverse	GCTTTAGACAGGACCTTCTTCCA
NTM	Human	Forward	CCACACCAATGCCAGCATCATG
		Reverse	TGAGAAGCAGGTGCAAGACCAG
β-actin	Human	Forward	GGATGCAGAAGGAGATCACTG
		Reverse	CGATCCACACGGAGTACTTG
Dock10	Mouse	Forward	TCTCCGAGAAGTTGGCTTTGCC
		Reverse	GACTCGCTATCTGTGCCTGCTT
β-actin	Mouse	Forward	CACTGTCGAGTCGCGTCC
		Reverse	TCATCCATGGCGAACTGGTG

siRNAs
Gene	Company	Product code
siDock10	Thermo Fisher	s231866
ON-TARGET plus Non-targeting Pool	Dharmacon	D-001810-10-05

qPCR, quantitative polymerase chain reaction; siRNA, small interfering RNA.

Protein Extraction and Quantification Following Glucose Stimulation

Knockdown MIN6 cells were washed twice and incubated for 2 hours in freshly prepared modified Krebs-Ringer bicarbonate buffer (MKRBB; 5 mM KCl, 120 mM NaCl, 15 mM HEPES, pH 7.4, 24 mM NaHCO₃, 1 mM MgCl₂, 2 mM CaCl₂, 1 mg/mL radioimmunoassay-grade bovine serum albumin [BSA]). Cells were then stimulated with MKRBB containing 25 mM glucose for 5 minutes. After stimulation, cells were lysed in buffer (50 mM Tris–HCl, pH 7.6, 150 mM NaCl, 1% CHAPS, 0.5% sodium deoxycholate, 1% sodium dodecyl sulfate [SDS]) supplemented with 1:100 Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific). Protein concentrations were measured using a BCA protein assay kit (Nacalai Tesque).

Western Blot Analysis Using an Automated Capillary Electrophoresis System (Simple Western)

Capillary electrophoresis automated system on Jess Simple Western system (Bio-Techne) was used. Sample preparation, antibody application, and running conditions were performed according to the manufacturer’s instructions. We applied 1 μg protein per lane. Samples were analyzed using the 12–230 kDa Wes Separation Module (Bio-Techne). Expression of β-actin were analyzed using RePlex assay (Bio-Techne) in every experiment. Using the Compass for Simple Western software (Bio-Techne), we analyzed the experimental data and quantified protein expressions. Specific bands were detected for p-PAK1 at approximately 90 kDa, PAK1 at 70 kDa, and β-actin at 50 kDa. Protein expression levels were normalized against β-actin. For evaluation of phosphorylation, the ratio of phosphorylated to total protein, such as p-PAK1/PAK1 ratio, was calculated.

Static Insulin Secretion Assay and Inhibitor Treatment

The method for static insulin secretion assay was described previously.⁴⁸ Briefly, MIN6 cells (5 × 10⁵ cells/well) were seeded in 24-well plates and analyzed 4 days after seeding for ML141 assays. Because electroporation of siRNA reduces cell viability, MIN6 cells (1.2 × 10⁶ cells/well) were seeded and analyzed the following day in the DOCK10 knockdown assay. Cells were pre-incubated for 30 minutes in Krebs-Ringer bicarbonate buffer (KRBB; 114.3 mM NaCl, 4.76 mM KCl, 1.90 mM CaCl₂, 0.95 mM MgCl₂, 23.8 mM NaHCO₃, 10 mM HEPES, pH 7.4) containing 0.02% BSA and 3 mM glucose, followed by 1-hour incubation in KRBB with either 3 mM glucose, 25 mM glucose, or 3 mM glucose plus 30 mM KCl. Culture supernatants were collected and analyzed for immunoreactive insulin using an ELISA kit (Mercodia, #10-1247-01). Total protein content was measured using a BCA protein assay kit (Nacalai Tesque) for normalization. Intracellular insulin content was measured by extracting cells with acid ethanol (0.18 M HCl in 96% ethanol, 1:3 [v/v]). For ML141 treatment, cells were cultured in FBS-free medium from the day before treatment, treated with 50 μM ML141 for 1 hour, and subjected to static insulin secretion assay as described above.

Animal Experiments

Female C57BL/6J mice (6–8 weeks old) were purchased from Japan Clea and housed at the Center for Experimental Animals, Institute of Science Tokyo. Mice were maintained under controlled conditions (18–23°C, 40%–60% humidity, 12-hour light/dark cycle) at a density of 100 cm² per mouse. Animal health was closely monitored, and no mice became ill before developing hypoglycemia.

Subcutaneous Transplantation of MIN6 Cells and Oral Glucose Tolerance Test

Subcutaneous transplantation of MIN6 cells was performed as previously described.³⁴ Cell pellets were suspended in Matrigel (80% v/v) containing 10% FBS and 10% PBS at a final concentration of 1 × 10⁷ cells/mL. A 200-μL suspension (2 × 10⁶ cells) was injected subcutaneously into the right flank or dorsal region using a 22G needle. As a control group, mice were injected with Matrigel (80% v/v) containing 10% FBS and 10% PBS at the same anatomical site. Transplanted mice were maintained for 2 weeks to allow engraftment. To assess insulin secretion capacity, an OGTT was performed after a 4-hour fast. Baseline blood samples were collected at 0 minutes for glucose and insulin, followed by oral administration of glucose (200 mg/kg body weight). Blood glucose levels were measured at 30, 60, and 120 minutes post-administration. Following the first OGTT, mice received an intraperitoneal (i.p.) injection of ML141 (1 mg/mouse in 20 μL DMSO, 80 μL PEG300, 10 μL Tween 80, and 80 μL ddH₂O) or vehicle. ML141 or vehicle was administered again at the start of fasting 4 hours before the second OGTT the next day. Survival was monitored daily for 2 weeks after the initial OGTT. Blood samples were collected during the initial 4-hour fast and again the following day after a second 4-hour fast. An OGTT was performed between the 2 samplings. Serum was separated, and serum insulin levels were measured using an ELISA kit (Mercodia, #10-1247-01).

RNA Extraction and qPCR

The methods of qPCR were previously described.⁴⁹ Briefly, total RNA was extracted from frozen samples from surgical specimens and organoids using the RNeasy Mini Kit (Qiagen), following the manufacturer’s instructions. First-strand cDNA was synthesized using a PrimeScript IV first strand cDNA Synthesis Mix (Takara Bio) and used as a template for PCR amplification. qPCR analysis was performed using the QuantiTect SYBR Green PCR Kit (Thermo Fisher Scientific) or the Universal Probe Library system (Sigma-Aldrich) in a reaction volume of 25 μL. Relative expression was determined using ΔΔCT method and normalized to β-actin (human). Primers are listed in Table 8.

Histopathological Analysis

Surgical specimens were reviewed by at least 2 pathologists, and a definitive diagnosis of pan-NEN was made only when both cytomorphology and immunohistochemistry were consistent with the diagnosis. Tumor grading was assessed based on the Ki-67 index.

Immunofluorescence Staining

Immunofluorescence staining using human surgical samples was performed as previously described.⁵⁰ Primary antibodies are listed in Table 6. Microwave treatment (500 W, 10 minutes) in 10 mM citrate buffer was applied for staining insulin, synaptophysin, NCAM (CD56), KRT19, MUC1, Pdx1, SOX9, DOCK10, UCHL1, and CACNA2D1.

Primary antibodies against insulin, synaptophysin, NCAM (CD56), KRT19, MUC1, Pdx1, SOX9, and UCHL1 were visualized with Alexa-488-conjugated secondary antibodies (Molecular Probes). DOCK10 and CACNA2D1 staining was performed using the standard ABC method (Elite ABC kit, Vectastain) combined with tyramide signal amplification. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). Images were acquired using an epifluorescence microscope (BZ-X700, KEYENCE). Levels adjustment was performed in Photoshop (Adobe) using an adjustment layer, and the same settings were applied identically to all images within the same staining group.

RNA-seq and Data Processing

Tumor and adjacent non-tumor tissues were excised as small blocks (a few millimeters in size) after surgery, snap-frozen in liquid nitrogen, and stored at −80 °C. Total RNA was extracted from these fresh-frozen tissues. For organoids, culture medium was removed from each well, followed by 2 washes with PBS. Lysis buffer was then added directly to the wells for RNA extraction. RNA was extracted using the same method as for qPCR, and RNA-seq was subsequently performed. Total RNA concentration was calculated by Quant-IT RiboGreen (Invitrogen, #R11490). To assess the integrity of the total RNA, samples are run on the TapeStation RNA screentape (Agilent, #5067-5576). Only high-quality RNA preparations, with RIN greater than 7.0, were used for RNA library construction.

A library was independently prepared with 1 ug of total RNA for each sample by Illumina TruSeq Stranded mRNA Sample Prep Kit (Illumina, Inc; #RS-122-2101). The first step in the workflow involves purifying the poly-A containing mRNA molecules using poly-T-attached magnetic beads. Following purification, the mRNA is fragmented into small pieces using divalent cations under elevated temperature. The cleaved RNA fragments are copied into first strand cDNA using SuperScript II reverse transcriptase (Invitrogen, #18064014) and random primers. This is followed by second strand cDNA synthesis using DNA Polymerase I, RNase H, and dUTP. These cDNA fragments then go through an end repair process, the addition of a single ‘A’ base, and then ligation of the adapters. The products are then purified and enriched with PCR to create the final cDNA library.

The libraries were quantified using KAPA Library Quantification kits for Illumina Sequencing platforms according to the qPCR Quantification Protocol Guide (KAPA BIOSYSTEMS; #KK4854) and qualified using the TapeStation D1000 ScreenTape (Agilent Technologies; #5067-5582). Indexed libraries were then submitted to an Illumina NovaSeq (Illumina, Inc), and the paired-end (2 × 100 bp) sequencing was performed by the Macrogen Incorporated.

Data Processing and Analysis

Preprocessing of the sequencing data was performed using the nf-core RNA-Seq pipeline (https://github.com/nf-core/rnaseq, version 3.10.1). Briefly, reads with low quality and length shorter than 20 bp were trimmed with TrimGalore (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/, version 0.6.7) and aligned to the human genome GRCh38 (Ensembl version 86) using STAR (version 2.7.9a).⁵¹ Transcript counts were then quantified with salmon (version 1.9.0).⁵² Tximport (version 1.28.0)⁵³ was used to import transcript-level abundances and summarize these counts to gene-level. For each analysis, gene expression counts were filtered for genes expressed in at least 80% of samples from the respective dataset and then normalized across samples using the DESeq2 (version 1.40.2)⁵⁴ median of ratios normalization method.

Differential Gene Expression Analysis

Analysis of surgical samples

Differentially expressed genes (DEGs) were identified for patients with insulinoma and NF-NET separately by comparing primary tumor and liver metastasis samples combined against healthy tissue samples while accounting for patient ID using DESeq2 (version 1.40.2).⁵⁴ Genes with adjusted P-value < .05, mean expression > 10 and |log₂ fold change| > 1 were considered significant. DEGs were further characterized through functional enrichment analysis using Gene Ontology terms and RGD human pathways.

Identification of insulin secretion candidate genes

Differential gene expression analysis in organoids was performed as described above. However, due to a limited sample size, primary cell culture samples grown from healthy tissue were classified as normal independent of the diagnosed tumor type of the patient and compared with primary tumor and liver metastasis organoids combined. To identify genes that are potentially involved in the abnormal insulin secretion in insulinomas, we selected all genes significantly upregulated in insulinomas compared with healthy tissue in both surgical and organoid samples and assessed their association with INS expression through linear mixed-effects models on variance stabilized expression counts while accounting for patient ID.

Details on RNA-seq Analysis

Overrepresentation of Gene Ontology terms in up- and down-regulated DEGs from surgical samples was tested with topGO (version 2.46.0) using the elim algorithm and fisher test. A P-value below .05 was considered significant. Enrichment of RGD human pathways in these genes was investigated with clusterProfiler (version 4.8.2) with a significance threshold of q-value < .05.

Heatmaps of DEGs were constructed with ComplexHeatmap (version 2.16.0), and overlaps of DEGs between comparisons were visualized as Venn diagrams using VennDiagram (version 1.7.3).

scRNA-Seq

Human insulinoma surgical sample was dissociated by the solution containing Advanced DMEM/F12 (Thermo Fisher Scientific), 1 mg/mL collagenase type II (Gibco), and 125 μg/mL Dispase II (Thermo Fisher Scientific) to generate single-cell suspension. Approximately 5000 cells with 85% survival rate were applied to Gel Bead-in-emulsion (GEM) generation using Chromium Controller (10x Genomics) to capture 3000 cells each in individual emulations together with reverse transcription reagent and a gel bead having a unique cell barcode. The following reverse transcription reaction enables cDNA to be labeled by the unique cell barcodes in each emulsion. cDNA amplification followed by library construction for single-cell gene expression were performed using Chromium Next GEM Single Cell 3ʹ Reagent Kits v3.1 (10x Genomics). The resulting libraries were quantified by Qubit dsDNA Assay (Thermo Fisher Scientific) and TapeStation D1000 ScreenTape (Agilent Technologies). The libraries were sequenced by a NovaSeq 6000 platform (illumina), generating approximately 300M paired-end reads per sample. The sequence reads were trimmed into Read1: 28 bp and Read2: 90 bp and processed in a Cell Ranger pipeline ver 7.0.1(10x Genomics), in which mapping, gene expression count, cell calling, and clustering analyses were performed. The single-cell gene expression profiling described above was conducted by Azenta Japan Corp (formerly GENEIWZ Japan). The obtained data were analyzed using the Loupe Browser v8.1.

Statistical Information

Statistical analysis was performed using GraphPad Prism (version 10) for macOS (GraphPad Software). No data were excluded for analyses. Significant differences in comparisons of 2 groups were determined by Student’s t-test. For data that were not normally distribution, the Mann-Whitney test was used. For certain clinical data, such as insulin immunostaining of surgical samples between insulinoma and NF-NET, the χ² test was used. P values less than .05 were considered as statistically significant. Bulk RNA-seq analyses of surgical and organoid samples were performed using R Statistical Software (version 4.3.1), and correction for multiple testing for these analyses was performed where applicable using the Benjamini and Hochberg method.

Code Availability

All R-based custom codes created for the analysis of bulk RNA-seq data from surgical and organoid samples are deposited at https://github.com/Systems-Immunology-IKMB/Neuroendocrine_tumors.

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

During the preparation of this work the authors used ChatGPT and Copilot in order to improve the clarity and readability of the manuscript, including assistance with language editing and sentence refinement. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.