The Focal Form of Persistent Hyperinsulinemic Hypoglycemia of Infancy: Morphological and Molecular Studies Show Structural and Functional Differences With Insulinoma

A total of 13 cases of FoPHHI in infants (1.5–7 months of age) suffering from severe diazoxide-resistant hypoglycemia of neonatal onset cured by partial pancreatectomy were compared with 8 insulinomas occurring in 3 children (6–15 years of age) and 5 adults (31–81 years of age) with recent onset hypoglycemia, 4 diazoxide-sensitive and 4 not tested (Table 1). All insulinomas were considered sporadic because of the complete absence of familial history of MEN1-associated tumors.

Samples containing both the lesion and its adjacent healthy islets were fixed in Bouin’s solution for conventional microscopy and investigation of cellular endocrine composition, in 4% formalin to study ABCC8 and CDKN1C peptide expression as well as IGF2 peptide and mRNA expressions, and frozen in liquid nitrogen to search for 11p15 LOH. Analyses were performed on all cases containing sufficient material both inside and outside the lesion.

Immunohistodetections were performed to analyze endocrine cell composition of both lesions and adjacent islets as well as the presence of CDKN1C, ABCC8, and IGF2 peptides. Technical details are given in Table 2. Dr. M. Mikhailov and Prof. S.J. Ashcroft (Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, Headington, Oxford, U.K.) provided the ABCC8 antibody. As the immunizing peptide was not available, its specificity was confirmed by documenting the absence of cross-reaction with SUR2 (cardiac muscle), CFTR, or MDR (exocrine pancreas and renal tumor, respectively).

To produce IGF2 probe, total RNA was extracted from snap-frozen pheochromocytomas (24) with Trizol reagent and reverse transcribed (Life Biotechnologies, Gibco-BRL, Merelbeke, Belgium). The specificity of primers used for human IGF2 amplification (positions 372-390 and 726-707 of the cDNA, accession no. XM_006402, NCBI nucleotide sequence) was checked by BLAST search on Genbank database. Sensitive IGF2 probe was produced by incorporation of digoxigenin-11–2′deoxy-uridine-5′-triphosphate (DIG) (Roche Diagnostics, Mannheim, Germany) during the PCR. Amplified DNA samples were quantified by a Gel Doc 2000 scanning device (Bio Rad, Nazareth, Belgium). IGF2 PCR product identity was confirmed by DNA sequence analysis using a dye terminator sequencing system on the Genetic analyzer 3600 (Applied Biosystems, Foster City, CA).

Proinsulin mRNA was detected with a fluorescein isothiocyanate–labeled probe (final dilution 1:2; NCL-Proins, Novocastra, U.K.) (25). The DIG-labeled IGF2 probe (2.4 μg/ml, 18 h at 45°C) was detected with monoclonal anti-DIG antibody (1/50, 1 h, 30 min; Roche), and streptavidin-biotin-peroxidase (SB-P), followed by biotinylated tyramine amplification (1/800 during 9 min) (26), incubation with SB-P (1/800, 30 min), and diaminobenzidine (DAB) revelation system. In situ hybridization with unrelated probe against prostatic specific membrane antigen, also produced by DIG incorporation in PCR product, was included as negative control.

β- and δ-cell volume densities (Vv) were measured by point-counting (27) on sections double-immunostained for insulin and somatostatin and expressed as the ratio of β-cell Vv–to–(β + δ) cell Vv.

An image analyzer measured β-cell crowding (number of β-cell nuclei/1,000 μm² of β-cell cytoplasm) (4). This parameter reflects the nucleo-cytoplasmic ratio: a low value indicating abundant cytoplasm, as observed in hyperfunctioning cells, and a high value characterizing packed resting cells. For these parameters, pancreases of 15 organ donors, deceased from causes unknown to affect endocrine pancreas, were used as age-matched controls (aged 2–9 months for infants and 9–72 years for insulinomas).

Intracellular insulin and proinsulin mRNA contents were quantified by densitometry (28) with an image analyzer IBAS 2000 (Zeiss-Kontron, Munich, Germany) and expressed as specific optical density (SOD).

Golgi proinsulin/β-cell area was evaluated on slides double-stained for proinsulin (immunoperoxidase) and insulin (immunofluorescence) digitized through a JVC KY-F58 color digital camera (Japan Victor, Brussels, Belgium). An image analyzer (KS-400 system; Zeiss-Vision, Munich, Germany) measured the surface ratio within both lesion and adjacent islets. ABCC8 and IGF2 peptide SODs were evaluated with the same devices with reference to the endocrine area labeled by immunofluorescence to synaptophysin.

Results are given as median and range. The statistical significance of results between lesions and controls or with their adjacent islets are assessed by the Wilcoxon’s rank-sum test and Wilcoxon’s matched-pairs test, respectively.

LOH of the 11p15 region was searched in all but four (Table 1, n°1, 2, 3, 13) previously published cases of FoPHHI (13,14) and in all insulinomas. To evaluate the deletion size, 12 different microsatellite markers covering 20 Mb from 11p15.5 region (containing imprinted genes H19, IGF2, and CDKN1C) to 11p15.1 region (containing ABCC8 and KCNJ11) were studied by PCR on either frozen material or microdissected slides of formalin-fixed specimens. Selected markers for the 11p15 region were from telomere to centromere (NCBI database): D11S922 (0.1 Mb), D11S4046 (0.8 Mb), tho (1.6 Mb), D11S1318 (2.1 Mb), D11S2351 (3.9 Mb), D11S1760 (4.2 Mb), D11S2071 (7.6 Mb), D11S569 (11.9 Mb), D11S1397 (15.0 Mb), D11S921 (16.1 Mb), D11S902 (16.3 Mb; intragenic in ABCC8), and D11S4114 (19.5 Mb). DNA was extracted from lesions and adjacent pancreas according to standard techniques. Fluorescent markers, custom synthesized by Genset (GensetOligos; Genset, Paris, France), were amplified using Amplitaq polymerase (Applied Biosystems). Gel data from PCR products were extracted and analyzed by an ABI 373 automated DNA sequencer and ABI Genescan software (Applied Biosystems).

In FoPHHI, the 13 lesions were usually invisible to the naked eye, measuring <10 mm. Microscopically ill delimited, the lesion seemed to result from the confluence of apparently normally structured islets separated by few exocrine acini maintaining a normal lobular pancreatic architecture (Fig. 1A). Endocrine cells, sometimes located in close contact with pancreatic ducts, were arranged in clusters of variable sizes separated by thin fibro-vascular strands or small cords of acinar tissue. Nuclei appeared polymorphous with size variations and hyperchromatism. Outside the lesion, islets were small with packed endocrine cells having scanty cytoplasm, thus showing a resting appearance (Fig. 1C). Immunohistochemistry revealed the presence of β-, α-, and δ-cells within the lesion, with non–β-cells predominantly localized at the periphery of the endocrine clusters, as in normal islets. Immunolabeling for insulin was homogeneous (Fig. 1E). Outside the lesion, labeling was more polarized and sometimes stronger. δ-Cells were particularly numerous in small islets (Fig. 1G).

Macroscopically, all insulinomas were well delimited, often palpable, recognizable and identifiable by their red-pink color. Their size varied from 8 to 24 mm. Microscopically the tumor was less well demarcated from adjacent pancreatic tissue, sometimes extending between exocrine acini or including normal islets. Endocrine cells were organized in nests (five cases) or ribbons (n°19, 20, 21: oldest patients), separated by fibrous and highly vascularized stroma (Fig. 1B). Three cases exhibited amyloid deposits in variable amounts (n°14, 15, 19). β-Cells were relatively uniform, cuboidal or cylindrical with moderately abundant acidophilic cytoplasm and a regular round or oval nucleus centrally located. The chromatin was granular, and there was a small nucleolus. Cellular and nuclear polymorphism was weak; mitoses and evidence of vascular invasion were absent. Outside insulinomas, islets showed regular nuclei and normally abundant cytoplasm and were thus very different from those located outside FoPHHI lesions (compare Figs. 1C and D). In the lesion, immunohistochemistry revealed almost exclusively β-cells, with either a diffuse and weak labeling for insulin or cellular variation in staining intensity (Fig. 1F). In three cases only (n°14, 16, 18), rare α- or δ-cells were scattered among tumoral β-cells. Islets outside insulinomas had much more intense insulin labeling and seemed relatively poor in δ-cells (Fig. 1H).

The differences in cell composition were quantified by measuring the volume density of β- and δ-cells. β-Cell Vv–to–(β + δ) cells Vv ratio was lower in FoPHHI than in insulinoma (0.93 [0.82–0.99] vs. 1.00 [0.99–1.00]; P < 0.001), implying that more δ-cells are present in FoPHHI. This ratio tended to be lower in islets located outside FoPHHI than in controls (0.55 [0.27–0.83] vs. 0.63 [0.44–0.77]). It was similar in islets adjacent to insulinoma and in their controls (0.87 [0.70–0.97] vs. 0.80 [0.70–0.92]).

In FoPHHI, insulin immunolabeling was similar or slightly weaker within the lesion than in adjacent islets (Figs. 1E and G). SOD measurements were not statistically different: 0.191 (0.123–0.171) in the lesion vs. 0.241 (0.121–0.268) in adjacent islets (Fig. 2A). In all insulinomas, insulin labeling was lower within the lesion than in adjacent islets (Figs. 1F and H). The difference estimated by measurement of insulin SOD was consistent (seven of seven cases) and large (0.114 [0.09–0.158] vs. 0.301 [0.185–0.345]; P ≤ 0.02) (Fig. 2A).

In FoPHHI, proinsulin labeling was strong in all lesions and formed a dense crescent outlining the enlarged Golgi area, sometimes with a faint cytoplasmic labeling (Fig. 3A). In insulinomas, the staining was lower, restricted to a slightly enlarged Golgi area (Fig. 3B) except in three cases with a trabecular pattern, two exhibiting some cells with diffuse cytoplasmic labeling, and the third presenting diffuse cytoplasmic labeling of all tumoral cells. Outside both lesions, proinsulin labeling was rather weak and limited to a small Golgi area. In FoPHHI, measurements confirmed a larger Golgi proinsulin/β-cell area within the lesion (n = 7) than in β-cells of adjacent islets (0.420 [0.329–0.500] vs. 0.178 [0.109–0.356]; P ≤ 0.02). No statistically significant difference between the lesion (0.254 [0.145–0.507]) and adjacent islets (0.132 [0.032–0.545]) was observed in insulinoma (Fig. 2B).

Proinsulin mRNA was detected in both types of lesions (Figs. 3C and D) and their adjacent islets. SOD was significantly higher in FoPHHI (n = 7; 0.348 [0.261–0.405]) than in insulinoma (n = 4; 0.135 [0.087–0.212]; P < 0.001). In contrast, there was no consistent difference between each lesion and their adjacent islets (FoPHHI: 0.336 [0.292–0.380] insulinoma: 0.216 [0.071–0.358]) (Fig. 2C).

The β-cell crowding reflects the cell activity, a low value indicating hyperfunctioning cells and a high value indicating resting cells. In FoPHHI (n = 8), crowding was significantly lower within the lesion (12.0 [10.3–13.9]) than in adjacent islets (15.2 [14.1–18.3]; P < 0.02), which had higher values than islets from age-matched controls (13.17 [11.23–15.58]; P ≤ 0.025) (Fig. 2D). In insulinoma (n = 7), no significant difference was observed between the tumor (12.7 [11.0–13.9]), adjacent islets (11.5 [10.7–15.5]), and islets from age-matched controls (11.4 [9.8–12.1]) (Fig. 2D).

ABCC8 peptide was expressed as homogeneous cytoplasmic staining of endocrine cells in both types of lesions (Figs. 3E and F) and also in the adjacent pancreatic islets (Figs. 3G and H). In FoPHHI (n = 7), ABCC8 peptide staining was weaker than in adjacent islets in all but one lesion (n°2, missense paternal mutation of ABCC8), and SOD was indeed significantly lower in the lesion (0.018 [0.003–0.03]} vs. 0.029 [0.021–0.039]; P < 0.05) (Fig. 2E). In insulinomas, ABCC8 peptide SOD was significantly higher than in adjacent islets (n = 7; 0.043 [0.024–0.080] vs. 0.032 [0.011–0.047]; P < 0.05) (Fig. 2E).

IGF2 peptide was immunodetected in endocrine cells from FoPHHI and insulinomas with variable staining (Figs. 4A and B). IGF2 mRNA was detected in β-cells in both types of lesions (Figs. 4C and D). IGF2 peptide was also observed in islets outside both lesions where, importantly, IGF2 mRNA was undetectable. IGF2 peptide SOD was similar inside and outside lesions (FoPHHI, n = 4; 0.114 [0.083–0.126] vs. 0.121 [0.107–0.177]; insulinoma, n = 4; 0.110 [0.094–0.299] vs. 0.095 [0.079–0.104]).

CDKN1C peptide was completely absent from all tested cases of FoPHHI (n = 9) (Fig. 4E) but also from insulinomas with 11p15 LOH (cfr. microsatellites analysis) (n°15, 17, 18), whereas it was detected in the five other cases in either a few cells (four cases) or numerous cells (n°16) (Figs. 4F). It was always present in islets located outside both types of lesions (Figs. 4G and H).

A genetic map of microsatellites tested is given in Fig. 5. For informative cases, allelic loss is documented if one allele is significantly decreased (<50%) in cells from the lesions compared with the same allele coming from cells located outside the lesions. All tested cases of FoPHHI showed extensive 11p15 LOH involving imprinted genes in 11p15.5 and ABCC8 and KCNJ11 in 11p15.1. Interestingly, three of the eight insulinomas (n°15, 17, 18) showed LOH of a similar extent. However, importantly, these three insulinomas were not different from the five others for any other parameter studied (Fig. 2). DNA extraction efficacy is probably lower in formalin-fixed material rather than in frozen material, explaining the differences in microsatellites detection between cases.

Severe hypoglycemic hyperinsulinemia characterizes FoPHHI and insulinoma and both are curable by selective surgical resection. The present study demonstrates that FoPHHI is a specific homogeneous entity contrasting with insulinoma but also that some insulinomas have a similar molecular background to FoPHHI.

The constant morphological aspect of the FoPHHI lesion is characterized by large endocrine clusters intermingled with exocrine tissue, whereas insulinomas of this series have the previously reported trabecular or solid pattern (29). Their cellular composition also differs: insulinomas are here almost exclusively composed of β-cells, whereas in FoPHHI, as in normal islets, non–β-cells always delineate β-cells cores. This difference is apparently not age-related. Indeed, our quantification demonstrates that β- and δ-cells proportions within both lesions differ from those outside lesions. Interestingly, islets outside insulinomas have a cellular composition identical to that of normoglycemic controls, whereas islets outside FoPHHI tend to have more δ-cells than normoglycemic controls. This may result from hyperinsulinemic hypoglycemia or reflect an immature development of these islets, which resemble those of neonates (27).

We also used various morphological approaches to evaluate several functional characteristics of both types of lesions. In this evaluation, FoPHHI appears very homogeneous. Indeed, compared with adjacent islets, all lesions have a high Golgi proinsulin/β-cell area, indicative of stimulated proinsulin production. In agreement with previous studies (30), there is no obvious abnormal prohormone processing. Interestingly, despite this high proinsulin production, proinsulin mRNA expression is not significantly different inside and outside the lesion. This is not due to limitations of our technique as its accuracy, reproducibility, and sensitivity previously have been demonstrated (25). This apparent discrepancy clearly indicates that in FoPHHI, proinsulin production regulation occurs at a translational rather than a transcriptional level in lesional β-cells. No significant differences in insulin storage were observed, although it appeared slightly lower within lesions. As lesional proinsulin production is over twice that outside, similar insulin content values imply its abnormal release. This is indeed what happens to β-cells without functional K_ATP channels.

In insulinomas, insulin content, although variable, is consistently lower within the lesion than in surrounding islets. As previously reported (31), this could either result from low insulin production or from its abnormal release. Golgi proinsulin/β-cell area, although slightly larger inside the lesion, was not significantly different outside, and great individual variations were observed, in agreement with previous studies (32). Furthermore, diffuse cytoplasmic proinsulin labeling indicative of impaired proinsulin-insulin conversion, was found in various tumoral cells of three cases, as previously reported in insulinomas (33). Proinsulin mRNA content was similar in insulinomas and adjacent islets, as in FoPHHI. However, proinsulin mRNA concentration was lower than in FoPHHI, suggesting a lower functional state.

Crowding measurement is another way of evaluating the functional state of endocrine cells. This parameter proved to be of major interest to differentiate focal and diffuse PHHI (4), even on per-operative sections (11). In FoPHHI, islets adjacent to the lesions have a resting appearance leading to high crowding, thereby reflecting the influence of permanent hyperinsulinemic hypoglycemia. Surprisingly, despite hyperinsulinemic hypoglycemias, islets located outside insulinomas have no resting appearance and no increased crowding. Several explanations are proposed. First, islets adjacent to FoPHHI endured hypoglycemia from birth, whereas those adjacent to insulinomas functioned normally for years. Second, counter-regulation mechanisms are more effective in adults than in infants and could reduce both severity and duration of hypoglycemias. Finally, our data indicate a very high proinsulin production in FoPHHI unobserved in insulinoma, which may thus result in more permanent, higher, and inadequate insulin release. Preoperative diazoxide treatment was only used in four of eight insulinomas and is thus unlikely involved in the differences observed between the islets.

This study is the first to report ABCC8 peptide expression in PHHI. The different expression in both lesions is interesting and surprising. Indeed, ABCC8 peptide amounts are significantly lower in FoPHHI and significantly higher in insulinomas than in their respective adjacent islets. Even the infants that could not be tested for ABCC8 mutation were resistant to diazoxide treatment, a hallmark of K_ATP channel dysfunction in hypersecreting β-cells. Assuming an ABCC8 mutation was indeed present in all FoPHHI, the decrease in ABCC8 protein is in full agreement with lesional LOH. The individual differences probably reflect different mutations. Indeed, not all mutations lead to a truncated protein, and it is likely that in some cases (case n°2), the protein is still present but abnormal and nonfunctional.

The increased ABCC8 protein expression in insulinomas was unexpected. It could be a characteristic of those tumoral cells. However, it could also reflect the normal expression level and imply that immunolabeling of adjacent islets is actually below normal. As ABCC8 protein is present in similar amounts in islets outside both insulinomas and FoPHHI, this low expression could reflect an adaptative response to hypoglycemia. Functional consequences of the higher ABCC8 protein expression in insulinomas are unclear, as it does not prevent abnormal insulin release. Rather than an increased β-cell mass, an abnormal regulation of insulin secretion certainly explains hyperinsulinism in insulinomas, but the reasons for this are unclear. To our knowledge, the eventual implication of ABCC8 mutations in insulinomas has never been studied. The increased ABCC8 peptide labeling findings in insulinomas in our series does not necessarily imply that the peptide functions normally. However, the fact that four of eight insulinomas of this series were sensitive to diazoxide does not favor the ABCC8 hypothesis.

The characteristic organization of the FoPHHI lesion with its islet-like pattern is indicative of adenomatous hyperplasia. The previously reported increased β-cell proliferation (18,19) could result from localized growth stimulating factors or from loss of local inhibitory mechanisms (1). This is in total agreement with 11p15 LOH observed in all of our FoPHHI and in those previously published (12–14). Indeed, this LOH changes the expression of 11p15 imprinted genes, resulting in an imbalance of corresponding gene products that could lead to excessive β-cell proliferation and adenomatous hyperplasia. In this region, maternally expressed H19 and CDKN1C genes and paternally expressed IGF2 (34–36) are known to play an important role in cell proliferation. IGF2 has been implicated in β-cell hyperplasia as shown in transgenic mice overexpressing this gene (37,38). Our investigations demonstrated similar amounts of IGF2 peptide in FoPHHI and adjacent islets, whereas IGF2 mRNA was only present within the lesion. This discrepancy between IGF2 peptide and mRNA has previously been reported for Wilms’ tumors and pheochromocytomas (24,39). This could result from accelerated degradation or from nonregulated excessive secretion of the peptide. In contrast to our results, Kassem et al. (40) report increased IGF2 peptide amounts in FoPHHI. However, this increase is weak, and discrepancies between the two studies could be explained by the differences in measurement methods. Indeed, we mainly observed IGF2 in β-cells but sometimes also in non–β-cells, as reported for fetal pancreases (41). We then chose an antibody against synaptophysin, a general endocrine marker (42), and not insulin for measuring the reference area in order to avoid restricting reference area to β-cells alone, contrary to Kassem et al. (40). Furthermore, synaptophysin identifies endocrine cells, whatever their functional state, because its detection is independent of the presence of granules. Our detection of IGF2 mRNA restricted to the lesion seems to contradict Fournet et al. (13), who could not demonstrate by semiquantitative RT-PCR assay an increased IGF2 mRNA concentration within the lesion. This difference may again be explained by different mRNA detection methods. Indeed, small differences in mRNA content could escape large RT-PCR amplification but be detected by less sensitive but accurate in situ hybridization. Furthermore, this last technique offers the opportunity to analyze which cell type contains mRNA. Presence of IGF2 in insulinomas has previously been reported (43). Interestingly, the pattern of IGF2 peptide and mRNA expression is similar in insulinomas and FoPHHI. We can thus conclude from our results that IGF2 overexpression exists at mRNA level in both lesions.

The function of H19 gene is still unclear, one proposition being that it directly or indirectly modulates IGF2 mRNA cytoplasmic levels (44). As both FoPHHI and insulinomas (with or without LOH) equally expressed IGF2, H19 is likely to play a role by itself, its loss combined with IGF2 overexpression, explaining previously reported differences in β-cell proliferation (18). In agreement with this interpretation, Fournet et al. (13) found a decreased expression of the H19 gene in FoPHHI and proposed that IGF2/H19 imbalance rather than IGF2 upregulation induced pancreatic hyperplasia in FoPHHI.

Interestingly, three of our eight insulinomas had the same 11p15 LOH as all FoPHHI cases. This LOH is thus not so exceptional in insulinomas. It was indeed already reported on the basis of a long deletion involving the whole length of chromosome 11 (21). In our study, the long arm of chromosome 11 was not analyzed. A more recent paper (22), using another technique based on DNA copies quantification, also demonstrated 11p15 LOH in some insulinomas. Although molecular background and tumoral mechanisms certainly appear more heterogeneous in insulinomas, they may be similar to FoPHHI in cases with 11p15 LOH.

Finally, molecular homogeneity in FoPHHI is further supported by the constant loss of CDKN1C peptide expression in our series, as recently published by Kassem et al. (40). This gene encodes for a cyclin-dependent kinase inhibitor and is a negative cell proliferation regulator (45) implicated in the pathogenesis of several tumors (17). 11p15 LOH results in loss of the protein, this loss potentially also playing a role in the occurrence of the lesion. CDKN1C peptide is also absent in insulinomas with 11p15 LOH, whereas it is always present in insulinomas devoid of LOH, indicating, as proposed by Kassem et al. (40), its potential usefulness in tumoral 11p15 LOH detection.

In conclusion, this study, combining morphology with quantitative and molecular data demonstrates the extraordinarily homogeneous character of FoPHHI. This hyperplastic lesion, likely resulting from an imbalance between 11p15 imprinted gene products related to constant LOH of this region, produces great amounts of proinsulin in response to abnormal insulin release consecutive to decreased ABCC8 peptide expression. Adjacent islets are resting, and their composition is similar to that of normoglycemic neonates. These results evidence intra-pancreatic regulation of islet β-cell activity at a translational level. In accordance with previous studies, we confirm the heterogeneity of insulinomas. Therefore, FoPHHI is not an insulinoma of the infant, although some insulinomas share molecular background similarities.

This study was supported by FRSM Belgium (Grant no. 3.4594.99) and by the European Network for Research into Hyperinsulinism in Infancy (ENRHI).

The authors thank Dr. Michael Mikhailov and Prof. Steve Ashcroft for the gift of the antibody against ABCC8 peptide. They also thank Mr. Stéphane Lagasse for his precious help in preparing the photomicrographs and Dr. Claire de Burbure for reviewing the manuscript.