Histologic and Molecular Profile of Pediatric Insulinomas: Evidence of a Paternal Parent-of-Origin Effect

Abstract

Context:

Acquired insulinomas are rare causes of hyperinsulinemic hypoglycemia in children and are much less common than focal lesions of congenital hyperinsulinism. The latter are known to be associated with isodisomy for paternally transmitted ATP-sensitive potassium channel mutations on 11p15; however, the molecular basis for pediatric insulinomas is not well characterized.

Objective:

The purpose of this study was to characterize the histopathological and molecular defects in a large group of 12 pediatric insulinomas seen at The Children's Hospital of Philadelphia.

Results:

Twelve children with insulinomas were seen between 1971 and 2013, compared to 201 cases with focal congenital hyperinsulinism seen between 1997 and 2014. The age of insulinoma patients ranged from 4–16 years at the time of surgery. Features of MEN1 syndrome were present in five of the 12, including four cases with heterozygous mutations of MEN1 on 11q. Immunohistochemical analysis revealed nuclear loss of p57 staining consistent with loss of the maternal 11p15 allele in 11 of the 12 insulinomas, including all five MEN1-associated tumors. Imbalance of the paternal 11p allele was confirmed by single nucleotide polymorphism genotyping and methylation assays of the 11p imprinting control loci in four of five MEN1-associated tumors and six of seven sporadic insulinomas. In addition, single nucleotide polymorphism genotyping revealed extensive tumor aneuploidy beyond chromosome 11.

Conclusions:

These data indicate that MEN1 mutations are more common in insulinomas in children than in adults. Aneuploidy of chromosome 11 and other chromosomes is common in both MEN1 and non-MEN1 insulinomas. The novel observation of a paternal parent-of-origin effect in all MEN1 and most non-MEN1 tumors suggests a critical role for imprinted growth-regulatory genes in the 11p region in the genesis of β-cell endocrine tumors in children.

Insulinomas are well-differentiated pancreatic endocrine tumors causing symptomatic hypoglycemia due to unregulated secretion of insulin. With an incidence of two to four cases per million per year (1), these tumors are rare in adults and even less common in children. Insulinomas previously reported in neonates are now likely to be classified as focal lesions of congenital hyperinsulinism (FoCHI), which develop secondary to a paternally transmitted mutation in one of the two subunits of the ATP-sensitive potassium channel coupled with somatic maternal loss of heterozygosity (LOH) of 11p15. The 11p15 region contains imprinted genes that control islet cell growth, such as the paternally expressed growth promoter IGF2 and the maternally expressed growth suppressor CDKN1C that encodes p57^KIP2. Histologically, FoCHI lesions are characterized by a localized proliferation of endocrine tissue, including α- and β-cells, acinar and ductal elements. Maternal LOH of 11p15 can be demonstrated by immunohistochemistry (IHC) that shows loss of nuclear reactivity for p57^KIP2 (2).

In contrast to FoCHI, insulinomas result from a neoplastic proliferation of β-cells forming a tumor containing few intermixed normal structures and typically lacking IHC reactivity to glucagon or other islet cell hormones. A limited number of genetic alterations have been described in association with insulinoma including mutations in MEN1 (3), YY1 (4, 5), loss of maternal 11p (6–9), and LOH of chromosome 22q (10), and mutation analysis of specific congenital hyperinsulinism genes in insulinomas has only rarely been performed (11, 12). However, in one report, loss of p57^KIP2 by IHC was identified consistent with maternal LOH of 11p15 (11). The purpose of this study was to characterize the influence of 11p15 LOH on the clinical and histological features of pediatric insulinomas and to define the molecular changes within these tumors. In our series of 12 pediatric insulinoma cases, we identified unique genetic and genomic changes including copy number gains and losses as well as a novel paternal parent-of-origin effect in both MEN1-associated and non-MEN1 tumors.

Materials and Methods

Clinical review

A retrospective review of specimens diagnosed as “insulinoma,” “pancreatic adenoma,” or “pancreatic neuroendocrine tumor” between 1971 and 2013 was performed using the surgical pathology archives at The Children's Hospital of Philadelphia (CHOP). Hematoxylin and eosin-stained slides as well as any available IHC and cytochemical slides were examined to confirm the diagnosis of insulinoma. Cases presenting in early infancy or demonstrating morphological features that would now be recognized as diagnostic of FoCHI were excluded. Clinical information was abstracted from the medical records. Review of case material was performed in accordance with and under the approval of the Institutional Review Board.

Immunohistochemistry

IHC stains were performed on each case to include: chromogranin, synaptophysin, somatostatin, Ki67 (Dako, Inc), insulin, p57^KIP2 (ThermoFisher Scientific), glucagon (BioGenex Laboratories Inc), and pancreatic polypeptide (GeneTex). Slides labeled with antibodies to Ki67 and p57^KIP2 were scanned using the Aperio ScanScope CS slide scanner (Aperio Technologies), and the digitized slides were analyzed for determination of percentage nuclear positivity using the Aperio ImageScope software macro (version 10.0.1346.1807; Aperio Technologies) optimized for IHC.

Mutation analysis

Fresh-frozen normal and tumor tissue was preserved in RNAlater solution (Ambion; Life Technologies) to isolate genomic DNA (gDNA) and genomic RNA using the QIAGEN DNA/RNA extraction kit (QIAGEN Inc) according to the manufacturer's protocol. QIAGEN formalin-fixed, paraffin-embedded extraction kit was used to isolate gDNA from normal and tumor samples in formalin-fixed, paraffin-embedded scrolls. The coding regions and intron/exon boundaries of hyperinsulinism-related genes ABCC8, KCNJ11, GCK, GLUD1, and UCP2 along with CDKN1C, YY1, and MEN1 were amplified and directly sequenced on an ABI 3730 capillary DNA analyzer (Applied Biosystems) and analyzed using the Sequencher 4.9 program (Gene Codes).

Genome-wide single nucleotide polymorphism (SNP) array analysis

gDNA from eight of the 12 insulinomas with adequate quality were SNP genotyped by the CHOP Center for Applied Genomics and Cytogenomics Laboratory. SNP arrays were performed using either the Illumina 550v3, Quad610, or 850K Human array platforms, depending on which technology was available at the time of analysis. Copy number estimates based on alterations in probe intensities (as measured by log-R ratios) observed in conjunction with changes in genotype frequencies (as measured by B-allele frequencies) were used to assess percentage of mosaicism using methods described by Conlin et al (13).

Methylation studies

Methylation analysis of the two imprinting control regions (ICRs) on 11p15 was performed on gDNA using allele-specific methylated multiplex real-time quantitative PCR at the Genetic Diagnostics Laboratory at the University of Pennsylvania. The ICR1 region includes the paternally expressed IGF2, and the ICR2 region contains the maternally expressed CDKN1C. Probes and primers used in the assay were as described by Azzi et al (14).

Results

Patients

Table 1 shows the clinical and histological features of the 12 cases of pediatric insulinoma seen in order of increasing age at surgery (range, 4 to 16 years; median, 10.5 years). There were slightly more males than females (seven males, five females) with insulinoma. The median duration of symptoms, defined as the time between onset of symptoms and surgical resection, was 4 months (range, <1 month to 9 years).

Table 1.

Histological and Clinical Features of 12 Pediatric Insulinomas

Case No.	Clinical Features							Histological Features
	Gender	Age at Surgery, y	Duration of Symptoms, mo	Additional Endocrine Tumors	Responsive to Diazoxide	Tumor Size, cm	Tumor Location	Capsule	Architecture	Nucleomegaly	Mitoses, per 10 hpf	Ki67	Amyloid
1	F	4	2		No	0.7	Head	Incomplete	Mi, T, N	No	1	0.003	No
2	F	5	>6		No	0.8	Tail	Incomplete	T	Rare	0	2.6	No
3	F	5	N/A		N/A	0.9	N/A	Incomplete	T	Rare	0	0.7	No
4a	M	8	7	NOS	N/A	1.5	Tail	Focal	C	No	4	4.5	Yes
5	M	10	2		No	0.8	N/A	Yes	N, T	No	0	3	No
6	M	10	12		Yes	1.1	Head	Yes	T, N	No	0	1	No
7a	M	11	<1	INS, GLU	N/A	1.5	Tail	Incomplete	Mi, T	No	0	0.6	Yes
8a	F	11	9	INS	Yes	0.7	Head	Incomplete	T, N	No	0	7.6	No
9	F	13	N/A	INS, PRL, GLU	N/A	2	Mid-body	Incomplete	N, T	No	0	1.5	No
10a	M	13	2	GLU	No	1.1	Nodule at splenic hilum	Yes	Mi, N, T	No	0	3.7	Yes
11	M	14	2		No	0.7	Neck	Incomplete	N	No	0	1.6	No
12	M	16	108		No	1.5	Head	No	N	No	0	1.1	Yes
Median		10.5	4			1					0	1.55

Abbreviations: F, female; M, male; N/A, data not available; INS, insulinoma; PRL, prolactinoma; GLU, glucagonomas; NOS, endocrine tumor, not otherwise specified; C, cribriform; Mi, microcystic; N, nested; T, trabecular; hpf, high-power field.

^aMEN1 mutation identified.

Five of the 12 cases were classified as having MEN1 syndrome based on clinical features and/or mutation analysis. Case 4 had a microadenoma identified in a prior partial pancreatic resection specimen. At surgery, case 7 had multiple insulin- and glucagon-positive microadenomas in addition to an insulinoma and later developed a prolactinoma. One case (no. 8) had a germline MEN1 mutation and a history of an insulinoma resected 4 years earlier. The sample from case 9 was too degraded to obtain adequate DNA for MEN1 mutation analysis, but multiple pancreatic endocrine tumors had been found at surgery, and she later developed a prolactinoma. Case 10 had an insulinoma resected 2 years before presentation for a second insulinoma at CHOP; he had not been previously tested for MEN1 mutations, but subsequently developed a prolactinoma. During resection of the second insulinoma, multiple glucagonomas were also found within the pancreas. None of the patients in the MEN1 group had evidence of parathyroid adenomas commonly associated with MEN1 syndrome.

Tumor characteristics

In contrast to normal pancreas and FoCHI, the insulinomas demonstrated the characteristic morphology of well-differentiated endocrine neoplasms, being predominantly composed of monotonous proliferations of small- to medium-sized cells with pale cytoplasm and minimal nuclear pleomorphism (Figure 1, 1A–1C and 2A–2C). Tumor sizes ranged from 0.7 to 2.0 cm in diameter (median, 1 cm) consistent with previous reports (12). All tumors, except one, were found within the pancreas and were primarily located in the head or tail; in case 10, the insulinoma was located within a pancreatic nodule at the splenic hilum. An incomplete (n = 7) or complete (n = 3) fibrous capsule was noted in most cases, as well as fibrous bands traversing the tumor (n = 9). Architectural patterns primarily included microcystic, trabecular, nested, or a combination of these. Interposed ductal or acinar cells were largely absent; when present, these elements were mainly found within the fibrous bands. Overall, tumor nuclei appeared bland; however, rare, prominent nucleoli were seen (cases 2 and 3). Mitotic figures were absent in most cases and were only rarely seen in two cases (nos. 1 and 4); no cases showed intratumoral necrosis. Amyloid was present in four cases (nos. 4, 7, 10, 12), and two cases also contained numerous infiltrating eosinophils (nos. 4 and 10). No malignant features, such as invasion into lymphovascular spaces or adjacent organs, were identified in any of the cases, and none of the patients had evidence of metastatic disease. No specific morphological features were identified in the five patients in the MEN1 group compared to the seven non-MEN1 patients. Only patients in the MEN1 group had multiple endocrine tumors.

Figure 1.

Hematoxylin and eosin and immunostained sections of normal pancreas, FoCHI, and insulinoma (case 8). Compared to normal islets (1A), FoCHI demonstrates expanded endocrine tissue; ducts and acinar tissue are retained (1B). Loss of lobular architecture and homogenous population of endocrine cells characterize insulinoma (1C). Both FoCHI and insulinomas contain proliferation of insulin-producing β-cells (2B, 2C, insulin IHC). Similar to normal islets (3A), glucagon-positive α-cells are seen at the periphery of enlarged islets in FoCHI (3B); no additional hormone expression is seen in insulinoma (3C). Both lesions demonstrate loss of nuclear p57 expression within β-cells (4B, 4C, p57 IHC). All images are at 20× original magnification.

IHC staining demonstrated reactivity with neuroendocrine markers chromogranin and synaptophysin in all tumors. Eleven of 12 cases were positive for insulin, with variable intensities ranging from faint to strong. In contrast to normal pancreatic islets and the expanded endocrine tissue within FoCHI, most insulinomas were negative for other hormonal markers (Figure 1, 3A–3C); one tumor showed rare cells positive for glucagon and pancreatic polypeptide, whereas three additional insulinomas (nos. 2, 10, and 12) contained scattered somatostatin-positive cells. Mitotic activity was low across all tumors (median, 0 mitoses per 10 high-power fields). Cellular proliferation determined by Ki67 IHC was variable but low overall, ranging from 0.003 to 7.6% of cells (median, 1.6%). In three cases (nos. 7, 8, and 10), the proliferation rate based on mitotic count or Ki67 index fell into the range of intermediate-grade tumors (15); however, no definitive evidence of malignant growth was seen histologically. No significant differences in age at diagnosis, tumor size, number of intratumoral mitoses or percentage of Ki67-positive cells were found between MEN1 and non-MEN1 patients (median age at diagnosis, 11 vs 9 years; P = .39; median tumor size, 1.4 vs 0.9 cm; P = .08; median number of mitoses per 10 high-power fields, 0 vs 0; P = .36; percentage of Ki67-positive cells, 4.1 vs 1.5%; P = .09).

In contrast to normal islets that contained nuclei positive for p57 by IHC, all but one of the 12 pediatric insulinomas (no. 11) demonstrated diffuse loss of nuclear p57^KIP2 expression (Figure 1, 4C) similar to FoCHI and consistent with maternal LOH of 11p15 in all five cases in the MEN1 group and in six of the seven non-MEN1 insulinomas.

Mutation analysis

Tumor DNA from four of the five MEN1 patients (Table 2) contained mutations previously reported in association with MEN1 syndrome. Case 4 (MEN1:c.783+1 g>c) and case 7 (MEN1:c.783+1 g>t) each carried a splice site mutation identified in gDNA from both tumor and normal tissues (16, 17). Case 8 carried an in-frame, three base-pair deletion leading to loss of a single amino acid (MEN1:c.358_360delAAG; p.Lys120del) (18), and case 10 carried a nonsense mutation (MEN1:c.378 G>A; p.W126X) (19). Parental samples were not available to determine parent-of-origin in case 4; however, nonmaternal inheritance was confirmed in case 7, and the mutations in cases 8 and 10 were paternally inherited.

Table 2.

IHC, Molecular, and Cytogenomic Studies of Chromosome 11 in Pediatric Insulinomas

Case No.	p57 IHC	MEN1 Mutation	Chromosome 11 Profile by SNP Array			% Methylation of ICR2 in Lesion DNA (CDKN1C)	% Methylation of ICR1 in Lesion DNA (IGF2)
			Copy No.	LOH	Mosaicism
MEN1 cases (n = 5)
4	Negative	Yes	2	Copy-neutral LOH	∼80%	28	66
10	Negative	Yes	2	Copy-neutral LOH	∼100%	19	90
7	Negative	Yes	1	Monosomy	∼85%	15	81
8	Negative	Yes	ND	ND	ND	2	78
9	Negative	Suspected	ND	ND	ND	ND	ND
Non-MEN1 cases (n = 5)
6	Negative	No	1 (11p), 2 (11q)	11p monosomy	∼50%	8	32
12	Negative	No	3	Trisomy, no LOH	∼70%	50	30
1	Negative	No	2	No LOH	No mosaicism	40	65
2	Negative	No	2	No LOH	No mosaicism	40	64
11	Positive	No	2	No LOH	No mosaicism	67	67
Insufficient data (n = 2)
3	Negative	ND	ND	ND	ND	ND	ND
5	Negative	ND	ND	ND	ND	ND	ND

Abbreviation: ND, no data.

No mutations were identified in tumor DNA involving any of the five hyperinsulinism-related genes, the tumor suppressor gene CDKN1C, or a ubiquitous transcription factor, YY1, which has been reported to harbor mutations in 30% of sporadic adult insulinomas (4, 5).

Chromosome 11 copy number abnormalities in tumor DNA

Genome-wide SNP array data from tumor DNA were available for eight patients to infer copy number and regions of LOH. Five of these eight patients demonstrated a chromosome 11 abnormality leading to an altered ratio of maternal and paternal alleles (Table 2). The three MEN1 cases (nos. 4, 7, and 10) all showed mosaic LOH of entire chromosome 11. Cases 4 and 10 had copy-neutral LOH (Figure 2B), and case 7 had monosomy. These three cases also showed loss of the maternally expressed p57 by IHC; therefore, for cases 4 and 10, the SNP array was consistent with paternal uniparental disomy (UPD) and for case 7 was consistent with paternal monosomy. SNP array data were not available for case 8; however, a paternally derived MEN1 heterozygous mutation was identified in blood and was homozygous in the tumor, consistent with loss of the maternal allele.

Figure 2.

Copy number and genotype frequencies for chromosome 11 obtained from genome-wide SNP array of tumor DNA. A, Example of normal copy number and genotype frequencies for entire chromosome 11. B, Case 4 showing mosaic copy-neutral loss of heterozygosity. C, Case 12 shows mosaic trisomy. D, Case 6 shows mosaic monosomy of 11p and normal copy number for 11q.

Within the five non-MEN1 cases, two (nos. 6 and 12) were mosaic for a copy number abnormality in chromosome 11, whereas the other three (nos. 1, 2, and 11) had a normal chromosome 11. Four of these five cases had loss of p57 IHC, suggesting maternal 11p LOH. Case 6 was 50% mosaic for 11p monosomy but had no alterations in 11q (Figure 2D). Case 12 had copy number findings consistent with mosaicism for three copies of chromosome 11 with approximately 70% of cells carrying this change (Figure 2C).

Methylation analysis

To confirm the parent-of-origin effect, we determined the methylation of imprinting control regions ICR1 and ICR2 on chromosome 11p15 using tumor DNA. At both loci, approximately 50% methylation indicates normally imprinted alleles. As shown in Table 2, all four MEN1 cases tested showed hypermethylation of ICR1 and hypomethylation of ICR2. This altered pattern of methylation was consistent with paternal UPD in cases 4 and 10, paternal monosomy in case 7, and either paternal UPD or paternal monosomy for case 8.

Within the five non-MEN1 cases tested, case 6 was hypermethylated at ICR1 and hypomethylated at ICR2, confirming the 11p monosomy seen on SNP array and the maternal LOH identified by loss of p57 IHC. Methylation of ICR1 and ICR2 was variably altered in case 12 likely due to mosaic trisomy. Case 12 was also negative for p57 IHC, suggesting that the allelic imbalance was likely due to an extra paternal copy. In case 11, both loci were equally hypermethylated without any concomitant copy number change on SNP array. The retained p57 staining in this case may reflect the presence of two populations of cells, some with loss of maternal 11p and some with loss of the paternal allele; the latter cells may explain the positive p57 staining. Methylation was skewed at both loci for cases 1 and 2, consistent with most of the other pediatric insulinomas and with loss of p57 IHC, although LOH for 11p was not detected by SNP array; as in case 11, this might also reflect heterogeneous cell populations.

Genome-wide aneuploidy in tumor DNA

As shown in Supplemental Table 1, extensive genome-wide aneuploidy characterized by copy number gains, losses, and copy-neutral LOH were seen in all three cases with an MEN1 mutation. Cases 4 and 10 showed near-triploidy status for the genome, whereas case 7 showed a genotype consistent with near-haploidy. Among the five non-MEN1 cases tested by SNP array, three lacked extensive genome-wide aneuploidy: case 1 had a normal karyotype, case 2 had only trisomy of chromosome 12, and case 6 did not have any whole chromosome aneuploidies but had several arm changes. In contrast, the other two non-MEN1 cases (nos. 11 and 12) had extensive chromosomal aneuploidy, consistent with near-triploidy. Whole chromosome aneuploidy was seen in both MEN1 and non-MEN1 cases with chromosomes 7, 9, 19, and 20, showing extra copies in all cases with near-triploidy for the genome. Of note, interstitial copy number changes were not seen in any of the cases. For both MEN1 and non-MEN1 patients, more copy number gains than losses were seen, with the highest average gains seen in chromosomes 9 and 20 and highest average losses seen in chromosomes 1p and 11p. Chromosomal arm changes were seen only in non-MEN1 cases (nos. 6, 11, and 12), with chromosome 5 arm changes seen in two of five cases.

Discussion

The present report describes the largest series of pediatric insulinomas in the literature to date. A large proportion of these children (42%) had identified or presumed mutations of MEN1, in marked contrast with the 7.7% reported rate of MEN1 in adult insulinomas (20). Even excluding two cases with prior surgery for insulinoma, at least three of the 12 insulinoma cases had MEN1 (25%); this is similar to the 22% incidence of MEN1 reported in another pediatric series, consistent with a higher incidence of MEN1 in children compared to adults (10). In cases without MEN1 mutations, no mutations in YY1 were found, in contrast to recent identification of YY1 mutations in up to 32% of adult insulinomas (5). Both MEN1 and non-MEN1 pediatric insulinomas had varying abnormalities of chromosomal numbers that were sometimes quite extensive but appeared to be restricted to copy number changes of either whole chromosomes or whole chromosome arms. An important novel finding was that 11 of the 12 pediatric insulinomas showed evidence of a paternal parent-of-origin effect due to maternal LOH for the 11p15 region, including all five MEN1 cases.

Only a few small series of pediatric insulinomas have been reported (Table 3), including a recent series of nine cases from the combined 12-year experience in Manchester and London (12). Early reports included the series of “insulinoma” cases collected from the literature by Mann et al (21) in 1969, of whom one-third were probably FoCHI, based on age at diagnosis. In the 60-year experience at the Mayo Clinic reported in 1991, only 6% of insulinomas occurred in children under age 20 years (22). Based on these reports and the current series, the youngest age at diagnosis for pediatric insulinomas appears to be 3–4 years (12, 22–24). Thus, insulinomas are unlikely to be a cause of hypoglycemia before 2 years of age. This allows clear separation of acquired insulinomas from FoCHI, which usually manifests within the first few weeks of life (25). In the present series and in the report by Padidela et al (12), only 25–50% of the pediatric insulinoma cases were responsive to treatment with diazoxide. Importantly, malignant insulinomas have only rarely been reported in pediatric patients (26).

Table 3.

Reported Pediatric Insulinoma Series

	Boley et al (23)	Mann et al (21)	Service et al (22)	Padidela et al (12)	CHOP	Total Cases
Year reported	1960	1969	1991	2014	2015
Total cases in series, n	15	14	13	9	12	51
Age range, y
2–4	0	0	0	1	0	1
4–10	9	5	2	4	4	24
10–20	6	9	11	4	8	38
Youngest age, y	N/A	8	4	4	4
MEN1 mutation	N/A	N/A	N/A	2/9	5/12	7/21
Responsive to diazoxide	N/A	N/A	N/A	5/9	2/8	7/17

Abbreviation: N/A, not available.

As described by Sempoux et al (11) and confirmed by the present study, the histopathological appearances of insulinomas and FoCHI are distinctly different (9). Insulinomas consist of a monotonous proliferation of insulin-staining β-cells usually bounded by a well-defined capsule. In contrast, FoCHI lesions have no capsule and consist of an overgrowth of mixed endocrine cells also incorporating exocrine cells and ducts. Based on our series and previously reported cases, the histopathology of pediatric insulinomas is essentially identical to that described in adult insulinomas and appears to be the same in both MEN1 and non-MEN1 tumors.

Sempoux et al (11) have previously described the differences in molecular features of FoCHI and insulinoma. As expected, all 13 of their FoCHI cases had loss of p57 by IHC and maternal LOH for 11p15. However, they also found maternal LOH for 11p15 in three of their eight insulinoma cases, including two of three pediatric tumors. Our results suggest that maternal LOH for 11p15 in insulinomas may be even more common, at least in pediatric tumors. This was especially true for insulinomas associated with MEN1 syndrome (five of five). Moreover, six of our seven non-MEN1 pediatric insulinoma cases also had maternal LOH for 11p15 by p57 IHC; thus, loss of chromosome 11 appears to be an early, critical event in initiating tumor formation in sporadic insulinomas. The remaining non-MEN1 case (no. 11) had positive p57 staining, but SNP arrays and methylation studies indicate that this case also had imbalanced expression of the 11p imprinted region.

We reviewed previously reported pedigrees with MEN1 syndrome for evidence of a parent-of-origin effect on either endocrine tumors in general or insulinomas specifically (27–29). Data on the last point are sparse because insulinomas are rare in MEN1 syndrome and most reports do not clearly distinguish between insulinoma and other pancreatic endocrine tumors. However, in these reports, there was clear evidence for transmission of MEN1 from both males and females for tumors other than insulinomas (29). For insulinomas, there were several examples of paternally transmitted MEN1; however, we were unable to find any examples of insulinoma due to MEN1 syndrome inherited from the mother (29). This supports the likelihood that maternal LOH for the 11p15 region may play an important role specifically for the growth of insulinomas, similar to its role in the overgrowth of islet cells in FoCHI.

The development of endocrine tumors in MEN1 syndrome is presumed to follow the two-hit mechanism of cancer in which an inherited inactivating mutation of MEN1 becomes isodisomic due to a somatic event leading to LOH. Our results showing a parent-of-origin effect in pediatric MEN1 insulinomas suggest that this “second hit” is likely to be a deletion of the entire maternal chromosome 11 leading simultaneously to isodisomy for the MEN1 mutation and to maternal LOH for 11p15, and that this may be specifically required for overgrowth of islet cells. Presumably, subsequent events in tumorigenesis lead to the multiple other chromosomal gains and losses that we observed in the MEN1 cases.

In contrast to the strong evidence for 11p maternal LOH in pediatric insulinomas, Dejeux et al (30) demonstrated that overexpression of the IGF2 growth promoter in the 11p imprinted region was biallelic in adult insulinomas. This suggests that an 11p parent-of-origin effect may be specific to pediatric insulinomas. In the present series, case 11 occurred in an older patient (age, 14 years), and the tumor retained p57 by IHC, indicating expression from the maternal allele. Although Sempoux et al (11) found loss of p57 staining in several adult insulinomas, the possibility that changes in the 11p15 locus in insulinoma are different in older individuals/adults vs children deserves further investigation.

As shown in Supplemental Table 1, all but one of the nine insulinomas genotyped for copy number aberrations demonstrated mosaic, genome-wide aneuploidy, indicating that both the MEN1 and non-MEN1 tumors had unstable genomes. Chromosomal arm changes appeared to be restricted to the non-MEN1 group and, within this group, gain of 5p was particularly common. Studies of adult insulinomas have shown that extensive chromosomal alterations and genetic instability are common in these tumors (31). The changes involved not only whole arm and whole chromosome changes, but also multiple interstitial gains and losses throughout the genome that are usually associated with advanced tumor stage. In contrast, our pediatric insulinomas appeared to have a milder degree of aneuploidy with copy number changes involving only whole chromosomes or whole chromosomal arms, without interstitial chromosomal changes. This suggests that the pediatric tumors might have been resected at earlier stages of development and had not had time to acquire additional genetic abnormalities.

In summary, The Children's Hospital of Philadelphia series shows that there is a higher frequency of MEN1-associated insulinomas in children compared to adults, as well as a novel paternal parent-of-origin effect due to maternal LOH for 11p15 in both MEN1 and sporadic insulinomas. Demonstration of loss of p57 expression due to maternal LOH of 11p15 in both FoCHI and insulinomas further points to the critical role of this region in regulating islet cell growth and of its dysregulation being important to neoplastic proliferation of pancreatic β-cells.