Abstract
By the strictest of definitions, a genetic driver of tumorigenesis should fulfill two criteria: it should be altered in a high percentage of patient tumors, and it should also be able to cause the same type of tumor to form in mice. No gene that fits either of these criteria has ever been found for ileal neuroendocrine tumors (I-NETs), which in humans are known for an unusual lack of recurrently mutated genes, and which have never been detected in mice. In the following report, we show that I-NETs can be generated by transgenic RT2 mice, which is a classic model for a genetically unrelated disease, pancreatic neuroendocrine tumors (PNETs). The ability of RT2 mice to generate I-NETs depended upon genetic background. I-NETs appeared in a B6AF1 genetic background, but not in a B6 background nor even in an AB6F1 background. AB6F1 and B6AF1 have identical nuclear DNA but can potentially express different allelic forms of imprinted genes. This led us to test human I-NETs for loss of imprinting, and we discovered that the IGF2 gene showed loss of imprinting and increased expression in the I-NETs of 57% of patients. By increasing IGF2 activity genetically, I-NETs could be produced by RT2 mice in a B6 genetic background, which otherwise never developed I-NETs. The facts that IGF2 is altered in a high percentage of patients with I-NETs and that I-NETs can form in mice that have elevated IGF2 activity, define IGF2 as the first genetic driver of ileal neuroendocrine tumorigenesis.
Introduction
Neuroendocrine tumors (NETs) can occur throughout the body, within the many organs that contain hormone-producing cells. Tumors found in these various locations can share common genetic drivers, as indicated by the fact that patients inheriting certain alleles of the MEN1 or MEN2/RET genes can develop neuroendocrine tumors in multiple locations, including pancreas, pituitary, thyroid, parathyroid, duodenum, thymus, adrenals and/or lungs (Wells et al. 2013, Kamilaris & Stratakis 2019). But familial syndromes account for only about 10% of the clinical cases of NETs (Scarpa 2019), while the other 90% of patients develop sporadic NETs that occur in single organs. Deep sequencing has identified genes that are recurrently mutated in many of the organ-specific, sporadic NETs. For instance, in sporadic pancreatic NETs (PNETs), recurrent mutations are observed in MEN1, DAXX or ATRX (Jiao et al. 2011). MEN1 loss also promotes PNET formation in mice (Crabtree et al. 2001). The fact that MEN1 is lost in a high percentage of patient samples and that MEN1 loss also causes PNETs to develop in mice defines this gene as a genetic driver of PNETs, and MEN1 is likely a driver of many other types of NETs as well.
But MEN1 is not a genetic driver of all types of NETs, including ileal NETs (I-NETs). I-NETs are the most common tumors of the small intestine. Approximately 2500 new cases of I-NETs occur in the United States every year (Dasari et al. 2017), and not only is the overall incidence of I-NETs increasing, but more importantly the percentage of patients with metastatic I-NETs at initial diagnosis is also increasing (Halperin et al. 2017). Metastatic I-NETs can be very aggressive, replacing as much as 80% of a patients’ normal liver. The genetic driver of this disease has long remained a mystery. Ileal neuroendocrine tumors are not found in patients with any of the multiple endocrine tumor syndromes, do not have mutations in the MEN1 and MEN2/RET genes that determine familial endocrine tumor syndromes, and do not have mutations in genes commonly mutated in other sporadic NETs such as ATRX or DAXX. Nor do I-NETs have mutations in oncogenes or tumor suppressor genes commonly mutated in human cancers, such as TP53, RB1, BRCA1, or KRAS. Probably the most compelling evidence for a genetic cause of I-NETs is the fact that chromosome 18 is lost in more than half of the tumors (Zhao et al. 2000, Wang et al. 2005), suggesting the presence of a key tumor suppressor gene or genes on chromosome 18. However, none of the chromosome 18 genes show recurrent mutations in I-NETs. In fact, the only gene that is recurrently mutated in I-NETs is CDKN1B, which is found on chromosome 12 and is mutated in 5–8% of I-NETs (Banck et al. 2013, Francis et al. 2013, Maxwell et al. 2015). This low percentage suggests that CDKN1B is not the genetic driver of I-NETs, and indeed I-NETs have not been reported in Cdkn1b mutant mice. The overall frequency of mutations in ileal neuroendocrine tumors is quite low (Banck et al. 2013, Francis et al. 2013) and is more similar to the low mutation frequency observed in pediatric tumors, which does not fit with the fact that I-NETs are usually detected in older patients.
Very few research tools are available to study I-NET biology. I-NETs do not easily yield patient-derived xenografts, and while I-NET cell lines have been reported (Pfragner et al. 1996, 2009, Kolby et al. 2001), these have not been made publicly available. There are also no autochthonous animal models of I-NETs. Conversely, the study of PNET biology is aided by the existence of several autochthonous mouse models (Hanahan 1985, Crabtree et al. 2001, Du et al. 2007, Shen et al. 2009, Contractor et al. 2016, Wong et al. 2020), by publicly-available cell lines, and by knowledge of several recurrently mutated genes. For these reasons, many more research studies are published about PNETs than I-NETs, in spite of the fact that the clinical frequencies of the two diseases are roughly the same (Dasari et al. 2017). Knowledge about a genetic driver of I-NETs, and availability of a mouse model for this disease, could stimulate research interest in ileal neuroendocrine tumors.
Among the many mouse models of pancreatic NETs is the transgenic model RT2 (Hanahan 1985), which has proven to be unusually versatile for the modeling of neuroendocrine tumors. Originally developed as a model of insulinomas, RT2 mice can also model a different NET subtype, nonfunctioning pancreatic neuroendocrine tumors, if the genetic background is changed from B6 to AB6F1 (Kobayashi et al. 2019). In a third genetic background, B6D2F1, RT2 mice can also generate a third type of NET, duodenal NETs (Grant et al. 1991). In this report, we demonstrate that in yet another genetic background, RT2 mice can develop ileal neuroendocrine tumors. Analysis of this first I-NET animal model also allowed identification of the first I-NET driver gene.
Materials and methods
Mouse experiments
Mouse experiments were approved by the Institutional Animal Care and Use Committee of Rutgers University. Mouse husbandry protocols have been previously described (Contractor et al. 2016). RT2 B6 mice were obtained from the National Cancer Institute (Frederick, MD), and A/J and B6 Igfbp1(+/−) mice were purchased from Jackson Laboratories. RT2 AB6(F1) animals were generated by mating RT2B6 males to A/J females, and RT2 B6A(F1) animals were generated by mating RT2B6 females to A/J males. Female B6 Igfbp1(−/−) mice were mated to male RT2B6 to generate RT2B6 Igfbp1(+/−) animals. Animals were killed using CO2 asphyxiation and cervical dislocation at 17 weeks of age. Pancreatic tumor volume was measured using a caliper, and livers were examined for metastasis. Intestinal tracts were analyzed for lesions within the lumen of the organs. Tumors were flash frozen and/or preserved in formalin.
Whole exome sequencing
Mouse DNA was prepared from tail, PNET or I-NET tissue using a Promega Wizard Kit. SureSelectXT Mouse All Exon Kit (Agilent) was used for capture of exonic DNA from the mouse samples. DNA was sequenced on an Illumina machine for 150 bp paired-end sequencing according to standard protocols. Reads were aligned to the mouse reference genome (mm9) with BWA 0.78-r455 (Li & Durbin 2009). Germline SNPs were called using GATK v3.8 (DePristo et al. 2011) and somatic SNPs were called using Mutect 11.4 (Cibulskis et al. 2013). Samples were manually compared for common mutations.
F2 analysis
A SNP assay for rs31828088 was designed and synthesized by ThermoFisher. The sequences of the outside primers were 5′AGCCCACCACAGCTGAAC and 5′CACAAGTCCAGGATATGTCTGAGAA. Reporter 1 sequence was 5′VIC-ACACTCGATAACATCTTGT-NFQ and reporter 2 sequence was 5′FAM-CACTCTGATAACGTCTTGT-NFQ. This assay was used to genotype DNAs from 279 RT2 AB6(F2) mice, which were previously reported (Contractor et al. 2016). Rs31828088 resides upstream of the INS2 gene, within a cluster of imprinted genes on mouse chromosome 7qF5. The B6 lineage and the A/J lineage encode different versions of this SNP.
Sequencing analysis of Armcx3
By whole exome sequencing, Armcx3 was initially judged to possess a common mutation in tumors from mouse 31462. However, this sequence was part of a 750 nt region that is duplicated on another chromosome within the mouse genome. To focus on the sequence of Armcx3 and not on the duplicated region, DNA isolated from tail, PNET and INET of mouse 31462 was subjected to PCR to amplify a portion of the genome that was specific to the Armcx3 gene and not part of the duplicated region. PCR primers were 5′-CCAGGAGCTTGTGATGAACG-3′ and 5′-AGCCCTTTTCTGTACAGCTCT-3′. After 40 cycles, the DNA was purified using a column from Qiagen. Sanger sequencing was primed using the oligonucleotide 5′-GACTGCAGGTCTGGTGATTG-3′ and was performed by Genewiz Inc.
Copy number analysis of Cdhr2
By whole exome sequencing, Cdhr2 was initially judged to have a copy number of 3 in tumors from mouse 31462. To analyze this further, DNA isolated from tail, PNET and INET of mouse 31462 was subjected to copy number analysis using real-time RTPCR. Assay Mm00392193_cn (Thermofisher) binds within the putative amplified region and was used to analyze the copy number of Cdhr2. DNA copy number was normalized using an assay against Tfrc (ThermoFisher).
Analysis of RNA
Flash-frozen mouse tumors were minced in TRIzol (ThermoFisher) and homogenized using a Polytron 1200E. Chloroform was then added and the upper aqueous phase was removed after microcentrifugation. Ethanol was added, and RNA was isolated using an RNAeasy column (Qiagen). RNA was converted into cDNA using RT reagents (ThermoFisher). qPCR was performed, using a Prism 7500 (Applied Biosystems). TaqMan assays for Igf2 and beta-actin were purchased from ThermoFisher. The following assays were designed with Primer3 software (http://bioinfo.ut.ee/primer3/) and used with SYBR green reagents (Thermofisher): H19 5′ATCTGCTCCAAGGTGAAGCT and5′GAAGTCATCCCGGGGTAGAG; H19os (A)5′CTATATGGGGATGGTGTCCAG and 5′GCCTAGTCTGAGCCCTGTTG; H19os (B) 5′TTTGCCAGCACACAATGTCA and 5′CACCCCATCTTTCAGACCCT; Ascl25′GGAGCTGCTTGACTTTTCCA and 5′TTTGGTCAGGCTGCACTAGA; Tnfrsf26 5′CGTACAGCTGATCGTGTGTG and 5′CTGTCCTGGCTCTGAGTCAA; Tnfrsf235′TCAACTGTCCCGATGGTGAA and 5′GTTCACAATCATGCAGGCCA; Tnfrsf22 5′CTCCTTCAAATGTCCCGCTG and 5′TGTTCCTGGGTGACACTTCT; Th 5′AAACCCTCCTCACTGTCTCG and 5′CGCACAAAGTACTCCAGGTG; Cd81 5′CAGTTCTATGACCAGGCCCT and 5′TGAGTATGTTGCCGCCTGA; Tssc4 5′CCAAGTGTTGTCCCCAGAGA and 5′GGTAGAACTCATGCCTCGGA; Kcnq1 5′CCTCATCGTGGTTGTAGCCT and 5′AAGCGGATACCCCTGATAGC; Scl22A18 5′CCATCCTGGCTTTTGTGGTC and 5′ATACACTGGCCTTGGTTCCA; Phlda2 5′CGCTCTGGGTCCGTGAAA and 5′GGGGCAAGGCTCAGCAAG; Cdkn1c 5′GCCAATGCGAACGACTTCTT and 5′ACGTTTGGAGAGGGACACC; and Napil4 5′GTCTTTCAGATGGAGGCCCT and 5′AAGACGTTCCTGCAAAGCTG.
Immunohistochemistry
Antibodies were purchased from the following suppliers: secretin (LSBio); chromogranin A (Abcam); serotonin (Immunostar); insulin (Agilent); and CDX2 (Cell Signaling). For antigen retrieval, slides were boiled for 16 min in 0.93% (v/v) Antigen Unmasking Solution H3301 (Vector Labs), then slowly cooled to room temperature for 30 min. Secondary antibodies were purchased from Vector Labs and diluted 500-fold in 10% goat serum and 1% BSA. Slides were treated with Vectastain Elite ABC peroxidase kit (Vector Labs) for 30 min, then treated with ImmPact DAB peroxidase substrate (Vector Labs SK-4105). The slides were counterstained with hematoxylin (Vector Labs). Immunohistochemistry results were evaluated by a pathologist (LT) with expertise in neuroendocrine tumors.
Statistical analysis
GraphPad Prism 7.04 software was used for analysis. Two-tailed t-test was used to evaluate RNA expression from sets of mouse and human tumors. Fisher’s exact test was used to compare metastasis frequencies. Nonparametric Mann–Whitney analysis was used to evaluate tumor volumes. Outliers were identified by Rout analysis (q = 1%).
Analyses of human tumor samples
Human ileal neuroendocrine tumors and matched normal tissues were provided by the Cooperative Human Tissue Network (CHTN), which obtained informed consents from patients. Analyses of human tumor samples were approved by the Institutional Review Board of CHTN.
Analysis of imprinting
Genomic DNA was prepared from frozen human tumors or matched normal tissue using a Wizard DNA prep kit (Promega). RNA was prepared from tumor or normal tissue using an RNAeasy kit following TRIzol extraction. RNA was converted into cDNA using reverse transcriptase reagents from ThermoFisher, or else as a negative control reverse transcriptase was omitted from the reaction. Primers 5′CTC TGT CCT CCC CTC CTT TG 3′ and 5′ AAC ACC CCA CAA AAG CTC AG 3′ were used to amplify the sequence flanking the rs680 SNP within the IGF2 coding sequence. Primers 5′CGG ACA CAA AAC CCT CTA GC 3′ and 5′GTC GTG GAG GCT TTG AAT CTC 3′ were used to amplify the sequence around the rs10840159 SNP within the H19 sequence. Amplifications from cDNA were performed for no more than 40 cycles in order to avoid genomic contamination. Following amplification, PCR products were enzymatically digested using ApaI (rs680 SNP) or Msc1 (rs10840159 SNP), then analyzed by agarose gel electrophoresis.
Pyrosequencing
Genomic DNA was treated with bisulfite, then subjected to pyrosequencing assays targeted against DMR0, DMR2 and H19 IGR. Pyrosequencing was performed by EpigenDx Inc (Hopkinton, MA, USA), using EpigenDx assays ADS1051, ADS006, and ADS596.
Results
Mouse modeling of I-NETs
In a previous study, RT2 AB6(F1) animals were generated by mating RT2 B6 males with A/J females; these animals produced highly metastatic, non-functioning pancreatic NETs (NF-PNETs) (Kobayashi et al. 2019). The reverse cross, in which A/J males were mated to RT2 B6 females, generates a slightly different line denoted RT2 B6A(F1); this mating was technically more challenging because RT2B6 mothers were extremely hypoglycemic and often died before or shortly after birth. But several litters of RT2 B6A(F1) mice were generated by fostering the pups to other mothers, and once these animals reached adulthood it became obvious that RT2 B6A(F1) mice had three phenotypes that were distinct from the nearly genetically identical RT2 AB6(F1) animals. First, pancreatic NETs produced by RT2 B6A(F1) were much smaller (Fig. 1A); second, liver metastasis was less common in RT2 B6A(F1) (Fig. 1B); and third, small, highly vascular tumors appeared within the ileum of the small intestines of RT2 B6A(F1) mice but not in RT2 AB6 (F1) mice (Fig. 1C). Thirty age-matched male RT2 B6AF(F1) animals were evaluated and 12 were found to have ileal tumors. Single ileal tumors were detected in nine of the animals, while each of the other animals had two ileal tumors. Pathological and immunohistochemical analysis revealed the ileal tumors to be neuroendocrine (Fig. 2A and B). Serotonin, a marker of human I-NETs, was detected in 22% of mouse ileal neuroendocrine tumors (Fig. 2C). The ileal tumors did not express secretin or insulin, which are common markers of duodenal and pancreatic NETs, respectively (Fig. 2D and E).
Phenotypic differences between RT2 AB6(F1) and RT2 B6A(F1) mice. (A) Comparison of pancreatic neuroendocrine tumor volume in age-matched RT2 mice from different genetic backgrounds (AB6F1 or B6AF1). 334 RT2 AB6(F1) and 46 RT2 B6A(F1) were evaluated. *** indicates a P value below 0.001 by Mann–Whitney analysis. (B) Comparison of liver metastasis frequency in age-matched RT2 mice with AB6F1 or B6AF1 genetic backgrounds. *** indicates a P value below 0.001 by Fishers’ exact test. (C) A section of mouse ileum, with tumor indicated by arrow. A full colour version of this figure is available at https://doi.org/10.1530/ERC-19-0505.
Citation: Endocrine-Related Cancer 27, 3; 10.1530/ERC-19-0505
Pathological analysis of ileal tumors from RT2B6AF1 mice. (A) Hematoylin and eosin staining of an ileal tumor from an RT2 B6A(F1) animal. (B, C, D and E) Immunohistochemical analysis of Chromogranin A (B), serotonin (C), secretin (D) and insulin (E) by an ileal tumor from an RT2 B6A(F1) mouse. (F and G) Immunohistochemical comparison of CDX2 expression by an I-NET (F) and a PNET (G) isolated from the same RT2 B6A(F1) mouse.
Citation: Endocrine-Related Cancer 27, 3; 10.1530/ERC-19-0505
We suspected that the ileal tumors in RT2 B6A(F1) mice might be primary ileal neuroendocrine tumors, which is a disease that has never previously been observed in mice. But it was also possible that the ileal tumors were metastatic lesions that originated from the primary PNETs generated by these animals; however, this seemed less likely because ileal tumors were not observed in the more metastatic RT2 AB6(F1) model (Fig. 1B). To test whether the ileal tumors were primary tumors, we employed two methods. First, immunohistochemistry was used to test for expression of the small intestinal marker CDX2; this test is often used on patient samples to distinguish whether a liver metastasis from an unknown primary NET originated from a primary PNET or lung NET, which do not express CDX2, or from a primary I-NET, which does express CDX2 (Strosberg et al. 2017). As shown in Fig. 2F and G, the mouse ileal tumors expressed the ileal NET-specific marker CDX2, while the mouse PNETs did not. Second, whole exome sequencing was performed on PNETs and ileal tumors isolated from the same two animals, because metastatic tumors should have mutations in common with precursor primary tumors, whereas independently arising primary tumors should not share common mutations. In tumors isolated from mouse 31508, there were no common mutations, indicating that the ileal tumor in this animal is a primary I-NET (Supplementary Fig. 1, see section on supplementary materials given at the end of this article and see also Supplementary Table 1 for nonsynonymous mutations and Supplementary Table 2 for copy number variations. Synonymous mutations are not shown). For a second mouse, identified as number 31462, initial analysis of whole exome sequencing data suggested the presence of two common mutations, a nonsynonymous SNP and a copy number variation (Supplementary Tables 3 and 4). However, further analysis revealed that both of the common mutations were actually false positives: the copy number amplification could not be found by real-time RTPCR analysis of the same region, and a C177A mutation could not be detected in the Armcx3 genes of either tumors using Sanger sequencing (Supplementary Figs 2 and 3). The Armcx3 sequence is part of a very long repeat that is also found in another part of the mouse genome, which was likely responsible for the false-positive result. Thus, the ileal tumor from mouse 31462 also arose independently from the PNET and is likely a primary ileal neuroendocrine tumor.
Loss of imprinting of IGF2 in human I-NETs
At the nuclear DNA level, the RT2 B6A(F1) line that develops I-NETs is identical to the RT AB6(F1) line that does not develop I-NETs. But since the lineage of the parents is reversed in these lines, I-NET formation might be caused by one or more imprinted genes, which are genes in which only the allele contributed by the father or by the mother is expressed. We also hypothesized that if an imprinted gene is required for I-NET formation in mice, then an imprinted gene might also be required for I-NET formation in patients. So we decided to analyze human I-NETs for loss of imprinting (LOI) of IGF2, as IGF2 LOI has been associated with several types of tumors (Leick et al. 2012) but has not previously been assessed for I-NETs.
We identified 30 I-NET patients that were heterozygous for SNP rs680, which occurs within the transcribed portion of the IGF2 gene. Heterozygotes encoded one allele with a recognition site for the ApaI restriction endonuclease, and a second allele that lacks this recognition site. cDNA from the tumors of heterozygotes was amplified by PCR, and the PCR products were treated with ApaI. Agarose gel electrophoresis revealed that 17 out of 30 tumors expressed both alleles of IGF2, indicating loss of imprinting (Fig. 3A). IGF2 sequences could only be detected in samples treated with reverse transcriptase (Supplementary Fig. 4), eliminating the possibility of genomic DNA contamination.
Analysis of IGF2 imprinting in human ileal neuroendocrine tumor samples. (A) Tumor cDNA from patients heterozygous for the rs680 SNP in IGF2 were PCRd with primers that flanked the SNP. The DNA was then treated with restriction endonuclease ApaI, which digests in the middle of the PCR product produced by one of the alleles but which can not digest the other allele. Patient samples that show only one band, large or small, have retained imprinting, whereas patient samples that show both bands have lost imprinting. 5T and 136T are homozygous controls. (B) IGF2 transcription was compared in tumor samples from patients with loss of imprinting (LOI) or with retention of imprinting (ROI). Expression of IGF2 mRNA was quantified by Q-PCR. Statistical significance was determined by two-tailed T test. A total of 22 samples (11 with loss of imprinting and 11 with retention of imprinting) were analyzed. RPLP0 was used as a normalization control. * indicates a P value below 0.05 by two tailed t-test. (C) DNA methylation within three regions that have previously been linked to IGF2 imprinting were analyzed by pyrosequencing. Eighteen tumor DNAs were compared, ten from patients without loss of imprinting and eight from patients showing loss of imprinting. Methylation percentages within each assay were analyzed using two-tailed T test, and * indicates a P value below 0.05. (D) Expression of IGF2 mRNA by normal adjacent tissue was quantified by Q-PCR. The samples were classified according to whether or not the patients showed loss of imprinting. A total of 18 samples, 10 of which had LOI, were analyzed. RPLPO was used as a normalization control. Statistical significance was determined by two-tailed T test. * indicates a P value below 0.05. (E) cDNA was prepared from normal tissue available from patients showing loss of imprinting of IGF2 in tumor tissue. Imprinting was determined by PCR amplification around SNP rs680, followed by digestion with the restriction endonuclease ApaI. Presence of two bands indicates loss of imprinting. The location of the normal tissue is indicated. SB indicates small bowel tissue. RT indicates that the sample was treated with reverse transcriptase, and M indicates that the sample was mock treated with no enzyme. A full colour version of this figure is available at https://doi.org/10.1530/ERC-19-0505.
Citation: Endocrine-Related Cancer 27, 3; 10.1530/ERC-19-0505
As would be expected, loss of imprinting correlated with increased IGF2 transcription (Fig. 3B). LOI of IGF2 has previously been linked to DNA methylation of three regions, which are called DMR0, DMR2 and H19 IGR (Dejeux et al. 2009). DMR0 and DMR2 are within the IGF2 gene, and H19 IGR is near the adjacent gene, H19. Pyrosequencing analysis of bisulfite-treated tumor DNA showed no difference in DNA methylation of H19 IGR, but both DMR0 and DMR2 were more methylated in tumors from patients showing LOI (Fig. 3C). Increased methylation of DMR2 has also been linked to loss of IGF2 imprinting for another type of neuroendocrine tumor, insulinomas (Dejeux et al. 2009).
The H19 gene near IGF2 is also an imprinted gene. We identified 15 I-NET patients that were heterozygous for SNP rs1081459, which is found within the transcribed region of H19. One allele of this SNP produces a recognition site for the MscI enzyme, and the second allele does not. MscI digestion of PCR products from tumor cDNAs revealed no loss of imprinting of H19 in any of the ileal tumor samples (Supplementary Fig. 5). Therefore, loss of imprinting in patients with ileal neuroendocrine tumors was specific for the IGF2 gene, and did not reflect a general loss of control of genetic imprinting in I-NETs.
Matched normal tissue was available for 12 of the patients that showed IGF2 LOI. In all 12 cases, loss of imprinting of IGF2 was present not just in the tumor but also in the normal tissue (Fig. 3D). There was also increased transcription of IGF2 in the matched normal tissue from patients showing loss of imprinting (Fig. 3E). For 10 of the 12 cases in which IGF2 LOI was observed in matched normal tissue, both tumor and normal tissue originated from the small intestine, but the distance between tumor and the matched normal tissue was not recorded. However in the other two cases, the tumor was removed from the small intestine while the normal was removed from the large intestine. The pathology report for patient 50N was particularly detailed. The ileal tumor of this patient was removed 20 cm from the junction between the small and large intestine, while the matched normal was removed from an undescribed area within the large intestine. Thus for this patient, IGF2 LOI likely occurred across a 20 cm or larger region of the intestinal tract.
Effect of Igfbp1 on I-NET formation in mice
The high percentage of patients with LOI of IGF2 suggests that IGF2 is a genetic driver of ileal neuroendocrine tumorigenesis. But by the strictest of definitions, a genetic driver should not only be altered in a majority of patient tumors, but should also be able to drive formation of the same tumors in mice. To test whether Igf2 can promote I-NETs in vivo, we examined the effects of increased IGF2 activity in RT2 B6 mice. This mouse line ordinarily does not generate I-NETs. To increase IGF2 activity, male RT2 B6 mice were mated to female B6 Igfbp1(−/−) animals to create a generation of RT2 B6 Igfbp1(+/−) mice. Igfbp1 encodes a protein that prevents IGF2 from binding to its receptor (Kajimura et al. 2005, Baxter 2014); therefore, a decrease in copy number of Igfbp1 should lower expression of IGFBP1 and thereby increase the effective activity of IGF2 in RT2B6 Igfbp1(+/−) mice relative to RT2B6 animals. As shown in Supplementary Fig. 6, RT2B6 Igfbp1(+/−) mice indeed had decreased expression of Igfbp1, as measured in PNETs isolated from both lines.
A high proportion (38.5%) of RT2B6 Igfbp1(+/−) mice developed ileal tumors, which expressed chromogranin A and CDX2 (Supplementary Fig. 7). This result strongly supports Igf2 as a genetic driver of ileal neuroendocrine tumorigenesis.
The RT2 B6 and RT2B6 Igfbp1(+/−) mice were also assessed for pancreatic neuroendocrine tumor volume and for liver metastasis, as these were two other phenotypic differences that appeared to be affected by an imprinted gene in RT2 B6A(F1) mice (Fig. 1A and B). Indeed, the RT2 B6 Igfbp1(+/−) animals not only developed I-NETs but also showed clear decreases in PNET volume and in liver metastasis compared to RT2 B6 mice (Fig. 4A and B). These data suggest that an increase in IGF2 activity can drive I-NET formation while also decreasing growth and metastasis of PNETs.
Effect of loss of copy of Igfbp1 in RT2 B6 mice. (A) Comparison of pancreatic neuroendocrine tumor volume in 17-week-old mice. 38 RT2 B6 male mice and 13 RT2 B6 Igfbp1(+/−) male mice were evaluated. ** indicates a P value below 0.01 by Mann–Whitney analysis. (B) Comparison of liver metastasis in 17 week-old RT2 B6 and RT2B6 Igfbp1(+/−) male mice. ** indicates a P value below 0.01 by Fishers’ exact test. A full colour version of this figure is available at https://doi.org/10.1530/ERC-19-0505.
Citation: Endocrine-Related Cancer 27, 3; 10.1530/ERC-19-0505
Genetic association of IGF2/H19os to PNET phenotypes
Finally we also examined whether Igf2 or a related gene associated genetically with the PNET volume and liver metastasis phenotypes observed in RT2 mice that were F1 hybrids of the inbred B6 and A/J lines (Fig. 1A and B). In mice, Igf2 is part of a cluster of imprinted genes on chromosome 7qF5. A small nucleotide difference that distinguished chromosome 7qF5 of the A/J and B6 lineages was genotyped in DNAs from a previously described dataset of 279 RT2 AB6(F2) hybrid animals (Contractor et al. 2016). RT2 AB6(F2) mice had to be used instead of F1 mice for this linkage analysis because all of the mice in the F1 generation have identical genotypes, whereas animals in the F2 generation have genetic variability that may possibly associate with phenotypes. For this study only liver metastasis and PNET volume could be compared because I-NET incidence had not been examined in this older data set.
As shown in Fig. 5A and B, the genotype of 7qF5 linked to liver metastasis and PNET volume, respectively. Animals that inherited two copies of this chromosomal segment from the A/J genetic background were more likely to have larger PNETs and liver metastasis. Heterozygotes were not considered for this analysis because in this dataset there was no way to tell which imprinted allele was inherited from which parent. Figure 5C summarizes metastasis data from Figs 1B and 5A and shows that F1 and F2 animals inheriting 7qF5 from A/J or B6 mothers correlated with increased or decreased metastasis, respectively, while inheritance of 7qF5 from an A/J or B6 father had no clear association with metastasis. Figure 5C thus suggested that metastasis was influenced by a maternally imprinted gene on 7qF5. Interestingly, this would rule out Igf2, which is a paternally imprinted gene.
Linkage analysis of RT2 AB6(F2) mice. For figures (A) and (B), DNAs from age-matched RT2 B6A(F2) mice were genotyped for a small nucleotide polymorphism near the IGF2 locus. Only homozygotes were evaluated because the expressed allele of heterozygotes could not be determined. AA indicates animals that were homozygous for the allele encoded by the A/J background, which encodes adenine at the SNP position, whereas GG indicates animals that were homozygous for the allele from the B6 background, which encodes guanine. (A) Liver metatasis was compared. * indicates a P value below 0.05 by Fishers’ exact test. (B) Tumor volumes of pancreatic NETs were compared, and ** indicates a P value below 0.01 by Mann–Whitney analysis. (C) This table summarizes metastatic trends of F1 animals (obtained from Fig. 1B) or F2 animals (obtained from Fig. 3A). While 7qF5 inherited from fathers did not appear to influence metastasis, F1 and F2 animals that inherited 7qF5 from A/J mothers were more likely to be metastatic than F1 and F2 animals that inherited 7qF5 from B6 mothers, respectively. (D) Comparison of RNA expression of maternally imprinted genes in the chromosome 7qF5 region of the mouse genome. Pancreatic neuroendocrine tumor extracts were used for analysis. 'A' indicates tumor extracts from RT2 AB6(F1) mice, which are less metastatic, while 'B' indicates tumor extracts from RT2 B6A(F1) mice, which are more metastatic. 18 RT2 AB6(F1) tumors and 23 RT2 B6AF(F1) tumors were evaluated. Expression was normalized to β-actin; see also Supplementary Fig. 8 for normalization to Gapdh. H19os was analyzed by two separate assays, directed against different parts of the RNA. ** indicates statistical significance (P < 0.01) by two-tailed t test and * indicates a P value less than 0.05. Error bars correspond to s.e.m. Phlda2, Slc22A16, Tnfrsf23 and Tnfrsf26 were also assayed but were poorly expressed and are not included. (E) Comparison of Igf2 RNA expression within PNETs isolated from RT2 AB6(F1) and RT2 B6A(F1) mice. * indicates a P value below 0.05 by two tailed t-test. Expression was normalized with β-actin. A full colour version of this figure is available at https://doi.org/10.1530/ERC-19-0505.
Citation: Endocrine-Related Cancer 27, 3; 10.1530/ERC-19-0505
Expression of all maternally imprinted genes within chromosome 7qF5 was then compared for PNETs isolated from RT2 AB6(F1) and RT2 B6A(F1) animals. PNETs were compared instead of I-NETs because both lines make PNETs, whereas only RT2 B6A(F1) make I-NETs. Only one of the maternally imprinted genes, H19os (Berteaux et al. 2008), showed differential expression (Fig. 5D). Notably, the only known function of H19os is to increase expression of Igf2 (Berteaux et al. 2008, Tran et al. 2012). Igf2 expression was indeed elevated in the RT2 B6A(F1) PNETs that had elevated H19os (Fig. 5E). The most simple explanation for these data would be that RT2 B6A(F1) mice have different neuroendocrine tumor phenotypes than RT2 AB6(F1) mice due to higher expression of Igf2, which is caused by the presence of a more active allelic form of maternally imprinted H19os.
Discussion
Lack of knowledge about the genetic drivers of ileal neuroendocrine tumors has precluded attempts to model this disease by targeting specific mouse genes. Here, we were able to produce the first mouse model of I-NETs by altering the genetic background of transgenic RT2 mice to B6AF1. Although genetic background changes have previously allowed RT2 mice to produce different types of NETs (Kobayashi et al. 2019), the appearance of I-NETs in RT2 B6A(F1) mice nevertheless came as a surprise; it was previously thought that these slow-growing, difficult-to-detect, genetically enigmatic tumors might be unmodelable in mice.
Interestingly, although I-NETs appeared in RT2 B6A(F1) mice, they could not be detected in nearly genetically identical RT2 AB6(F1) mice, which suggested that allelic differences in an imprinted gene could increase I-NET incidence. We investigated human I-NET samples and discovered that loss of imprinting of IGF2 occurred in 57% of patient samples (Fig. 3A). LOI of IGF2 is also common in other tumors of the gastrointestinal tract, for instance occurring in 47% of gastric tumors, in 32% of esophageal tumors, and in 47% of colorectal cancer (Cui 2007, Leick et al. 2012).
IGF2 is the first example of a gene that is altered in a majority of patients with I-NETs. Previous deep sequencing analysis of I-NETs had revealed no genes that were mutated in more than 8% of samples. Notably, loss of imprinting can be difficult to detect by deep sequencing approaches.
Patients with ileal neuroendocrine tumors generally have multiple tumors (Gangi et al. 2018), which can arise independently (Katona et al. 2006). This has suggested that mechanisms that cause multifocality, such as viral infections (Miao et al. 2014) or tumor-predisposing alleles (Bergthorsson et al. 2001), may be important drivers of I-NET tumorigenesis. Indeed, within certain families there is a predisposition to I-NETs (Sei et al. 2015, Du et al. 2016, Neklason et al. 2016, Dumanski et al. 2017), and such familial predispositions can associate with multifocality (Sei et al. 2015). But familial NETs tend to be rare, and the cause of non-familial multifocal I-NETs has been less clear. In this report, we found two patients in which loss of imprinting was present in both an ileal tumor and in normal large intestine, indicating that LOI can occur within a sizeable area of the intestinal tract. A large area of loss of imprinting of a secreted genetic driver protein like IGF2 may create an extended environment within which multiple I-NETs can form. Fields of loss of IGF2 imprinting have also been reported around prostate cancers (Bhusari et al. 2011). In the future, it will be interesting to determine whether multifocal I-NETs are located within large areas of IGF2 LOI.
The high incidence of loss of imprinting in human samples strongly suggested that IGF2 is a driver of ileal neuroendocrine tumorigenesis. To test this idea further, we elevated the activity of IGF2 by removing a copy of its negative regulator, Igfbp1. Loss of copy of Igfbp1 caused RT2 B6 mice, which normally do not make I-NETs, to produce I-NETs. Alteration of a gene in a high percentage of human tumors, combined with the ability of the gene to cause the same tumors to form in mice, is the strictest definition for a genetic driver.
The nearly identical lines RT2 B6A(F1) and RT2 AB6(F1) mice had other phenotypic differences in addition to differential ability to form I-NETs. Both lines developed pancreatic NETs, but these tumors were smaller and less metastatic in the RT2 B6A(F1) line that also developed I-NETs. Interestingly, decreasing the copy number of Igfbp1 in an RT2 B6 mouse line was sufficient to phenocopy all three of these traits: RT2 B6 Igfbp1 (+/−) mice not only developed I-NETs, but also showed decreases in both PNET volume and liver metastasis compared to RT2 B6 mice. This suggests that IGF2, whose activity would increase due to decreased IGFBP1, not only causes I-NETs to form but can also decrease PNET metastasis. The idea that IGF2 could increase incidence of I-NETs is not surprising, given the high percentage of I-NETs with LOI of IGF2 in Fig. 3A; however, the idea that IGF2 might decrease metastasis is less intuitive. Interestingly, expression of IGF2 is known to decrease as PNETs progress from localized disease to metastasis both in patients (Henfling et al. 2018) and in mice (Contractor T & Harris CR, unpublished observations). Also there are examples of other genes that have opposing effects on tumor incidence and metastasis, including TGFβ and CDKN1B (Besson et al. 2004, Massague 2012).
In a hybrid RT2 AB6(F2) mouse line, the phenotypes of PNET volume and liver metastasis associated with a maternally imprinted gene at mouse chromosome 7qF5 (Fig. 5). H19os, a maternally imprinted, positive regulator of Igf2, resides at this location, and expression of both H19os and Igf2 was elevated in the smaller, less metastatic PNETs produced by RT2 B6A(F1) mice. Our current model is that the H19os allele from B6 mice is more expressed than the H19os allele from A/J mice, such that RT2 B6A(F1) animals that inherit this allele from their B6 mothers will have elevated expression of Igf2, which then leads to formation of I-NETs and to smaller, less metastatic PNETs. By this model RT2 AB6(F1) animals, which inherit the less expressed allelic form of H19os from their A/J mothers, would have lower expression of Igf2, and this would preclude formation of I-NETs but promote larger, metastatic PNETs.
Transgenic cancer mouse models, especially ones that depend upon SV40 T-antigen expression, are usually considered to be inferior to more contemporary knock-in models. The chief criticism about use of SV40 T-antigen is that this is a monkey virus protein that is not responsible for tumorigenesis in humans, while the chief criticism about use of transgenic models is that transgenes insert into incorrect genomic locations, allowing control elements from heterologous genes to drive ‘leaky’ expression of the transgene in heterologous cell types. But the current work demonstrates that leaky expression of SV40 T-antigen can actually be useful, particularly when little is known about the actual genetic drivers for a type of cancer. Leaky expression of T-antigen by RT2 mice has now allowed I-NETs, as well as metastatic nonfunctioning PNETs (Kobayashi et al. 2019), to be modeled in animals for the first time. These animal models could lead to improved clinical treatments for patients with these two highly metastatic diseases. It should be noted that there are currently no animal models for rectal carcinoids, a type of NET that arises in a disproportionate number of African-Americans (Modlin et al. 2003), nor for lung carcinoids, which are among the most common of all NETs (Dasari et al. 2017). Crossing RT2 B6 mice with additional inbred mouse lines may allow these other types of NETs to be modeled in animals.
Supplementary materials
This is linked to the online version of the paper at https://doi.org/10.1530/ERC-19-0505.
Declaration of interest
Chris Harris, Tanupriya Contractor and Richard Clausen were employees of the Raymond and Beverly Sackler Foundation, which is a 501c3 nonprofit. The other authors have nothing to disclose.
Funding
This work was funded by the Raymond and Beverly Sackler Foundation.
Author contribution statement
Acquisition, analysis and interpretation of data was performed by Tanupriya Contractor, Richard Clausen, Chris Harris, Laura Tang, Jeffrey Rosenfeld, and Grant Harris. Conception and design of experiments was performed by Chris Harris and Tanupriya Contractor. Writing, review and/or revision of manuscription was performed by Chris Harris, Darren Carpizo, and Tanupriya Contractor.
Acknowledgments
The authors thank Arnold Levine, Evan Vosburgh and Doug Hanahan for important discussions, Sandra Harris and Xin Yu for help with the manuscript, and the Raymond and Beverly Sackler Foundation for financial support.
References
Banck MS, Kanwar R, Kulkarni AA, Boora GK, Metge F, Kipp BR, Zhang L, Thorland EC, Minn KT, Tentu R, et al.2013 The genomic landscape of small intestine neuroendocrine tumors. Journal of Clinical Investigation 2502–2508. (https://doi.org/10.1172/JCI67963)
Baxter RC 2014 IGF binding proteins in cancer: mechanistic and clinical insights. Nature Reviews: Cancer 329–341. (https://doi.org/10.1038/nrc3720)
Bergthorsson JT, Ejlertsen B, Olsen JH, Borg A, Nielsen KV, Barkardottir RB, Klausen S, Mouridsen HT, Winther K, Fenger K, et al.2001 BRCA1 and BRCA2 mutation status and cancer family history of Danish women affected with multifocal or bilateral breast cancer at a young age. Journal of Medical Genetics 361–368. (https://doi.org/10.1136/jmg.38.6.361)
Berteaux N, Aptel N, Cathala G, Genton C, Coll J, Daccache A, Spruyt N, Hondermarck H, Dugimont T, Curgy JJ, et al.2008 A novel H19 antisense RNA overexpressed in breast cancer contributes to paternal IGF2 expression. Molecular and Cellular Biology 6731–6745. (https://doi.org/10.1128/MCB.02103-07)
Besson A, Gurian-West M, Schmidt A, Hall A & Roberts JM 2004 p27Kip1 modulates cell migration through the regulation of RhoA activation. Genes and Development 862–876. (https://doi.org/10.1101/gad.1185504)
Bhusari S, Yang B, Kueck J, Huang W & Jarrard DF 2011 Insulin-like growth factor-2 (IGF2) loss of imprinting marks a field defect within human prostates containing cancer. Prostate 1621–1630. (https://doi.org/10.1002/pros.21379)
Cibulskis K, Lawrence MS, Carter SL, Sivachenko A, Jaffe D, Sougnez C, Gabriel S, Meyerson M, Lander ES & Getz G 2013 Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nature Biotechnology 213–219. (https://doi.org/10.1038/nbt.2514)
Contractor T, Kobayashi S, da Silva E, Clausen R, Chan C, Vosburgh E, Tang LH, Levine AJ & Harris CR 2016 Sexual dimorphism of liver metastasis by murine pancreatic neuroendocrine tumors is affected by expression of complement C5. Oncotarget 30585–30596. (https://doi.org/10.18632/oncotarget.8874)
Crabtree JS, Scacheri PC, Ward JM, Garrett-Beal L, Emmert-Buck MR, Edgemon KA, Lorang D, Libutti SK, Chandrasekharappa SC, Marx SJ, et al.2001 A mouse model of multiple endocrine neoplasia, type 1, develops multiple endocrine tumors. PNAS 1118–1123. (https://doi.org/10.1073/pnas.98.3.1118)
Cui H 2007 Loss of imprinting of IGF2 as an epigenetic marker for the risk of human cancer. Disease Markers 105–112. (https://doi.org/10.1155/2007/363464)
Dasari A, Shen C, Halperin D, Zhao B, Zhou S, Xu Y, Shih T & Yao JC 2017 Trends in the incidence, prevalence, and survival outcomes in patients with neuroendocrine tumors in the United States. JAMA Oncology 1335–1342. (https://doi.org/10.1001/jamaoncol.2017.0589)
Dejeux E, Olaso R, Dousset B, Audebourg A, Gut IG, Terris B & Tost J 2009 Hypermethylation of the IGF2 differentially methylated region 2 is a specific event in insulinomas leading to loss-of-imprinting and overexpression. Endocrine-Related Cancer 939–952. (https://doi.org/10.1677/ERC-08-0331)
DePristo MA, Banks E, Poplin R, Garimella KV, Maguire JR, Hartl C, Philippakis AA, del Angel G, Rivas MA, Hanna M, et al.2011 A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nature Genetics 491–498. (https://doi.org/10.1038/ng.806)
Du YC, Lewis BC, Hanahan D & Varmus H 2007 Assessing tumor progression factors by somatic gene transfer into a mouse model: Bcl-xL promotes islet tumor cell invasion. PLoS Biology e276. (https://doi.org/10.1371/journal.pbio.0050276)
Du Y, Ter-Minassian M, Brais L, Brooks N, Waldron A, Chan JA, Lin X, Kraft P, Christiani DC & Kulke MH 2016 Genetic associations with neuroendocrine tumor risk: results from a genome-wide association study. Endocrine-Related Cancer 587–594. (https://doi.org/10.1530/ERC-16-0171)
Dumanski JP, Rasi C, Bjorklund P, Davies H, Ali AS, Gronberg M, Welin S, Sorbye H, Gronbaek H, Cunningham JL, et al.2017 A MUTYH germline mutation is associated with small intestinal neuroendocrine tumors. Endocrine-Related Cancer 427–443. (https://doi.org/10.1530/ERC-17-0196)
Francis JM, Kiezun A, Ramos AH, Serra S, Pedamallu CS, Qian ZR, Banck MS, Kanwar R, Kulkarni AA, Karpathakis A, et al.2013 Somatic mutation of CDKN1B in small intestine neuroendocrine tumors. Nature Genetics 1483–1486. (https://doi.org/10.1038/ng.2821)
Gangi A, Siegel E, Barmparas G, Lo S, Jamil LH, Hendifar A, Nissen NN, Wolin EM & Amersi F 2018 Multifocality in small bowel neuroendocrine tumors. Journal of Gastrointestinal Surgery 303–309. (https://doi.org/10.1007/s11605-017-3586-8)
Grant SG, Seidman I, Hanahan D & Bautch VL 1991 Early invasiveness characterizes metastatic carcinoid tumors in transgenic mice. Cancer Research 4917–4923.
Halperin DM, Shen C, Dasari A, Xu Y, Chu Y, Zhou S, Shih YT & Yao JC 2017 Frequency of carcinoid syndrome at neuroendocrine tumour diagnosis: a population-based study. Lancet: Oncology 525–534. (https://doi.org/10.1016/S1470-2045(17)30110-9)
Hanahan D 1985 Heritable formation of pancreatic beta-cell tumours in transgenic mice expressing recombinant insulin/simian virus 40 oncogenes. Nature 115–122. (https://doi.org/10.1038/315115a0)
Henfling M, Perren A, Schmitt AM, Saddig CM, Starke AA, Riedl RG, Versleijen-Jonkers YMH, Sprij-Mooij DM, Ramaekers FCS, Hofland L, et al.2018 The IGF pathway is activated in insulinomas but downregulated in metastatic disease. Endocrine-Related Cancer 1005–1018. (https://doi.org/10.1530/ERC-18-0222)
Jiao Y, Shi C, Edil BH, de Wilde RF, Klimstra DS, Maitra A, Schulick RD, Tang LH, Wolfgang CL, Choti MA, et al.2011 DAXX/ATRX, MEN1, and mTOR pathway genes are frequently altered in pancreatic neuroendocrine tumors. Science 1199–1203. (https://doi.org/10.1126/science.1200609)
Kajimura S, Aida K & Duan C 2005 Insulin-like growth factor-binding protein-1 (IGFBP-1) mediates hypoxia-induced embryonic growth and developmental retardation. PNAS 1240–1245. (https://doi.org/10.1073/pnas.0407443102)
Kamilaris CDC & Stratakis CA 2019 Multiple endocrine neoplasia Type 1 (MEN1): an update and the significance of early genetic and clinical diagnosis. Frontiers in Endocrinology 339. (https://doi.org/10.3389/fendo.2019.00339)
Katona TM, Jones TD, Wang M, Abdul-Karim FW, Cummings OW & Cheng L 2006 Molecular evidence for independent origin of multifocal neuroendocrine tumors of the enteropancreatic axis. Cancer Research 4936–4942. (https://doi.org/10.1158/0008-5472.CAN-05-4184)
Kobayashi S, Contractor T, Vosburgh E, Du YN, Tang LH, Clausen R & Harris CR 2019 Alleles of Insm1 determine whether RIP1-Tag2 mice produce insulinomas or nonfunctioning pancreatic neuroendocrine tumors. Oncogenesis 16. (https://doi.org/10.1038/s41389-019-0127-1)
Kolby L, Bernhardt P, Ahlman H, Wangberg B, Johanson V, Wigander A, Forssell-Aronsson E, Karlsson S, Ahren B, Stenman G, et al.2001 A transplantable human carcinoid as model for somatostatin receptor-mediated and amine transporter-mediated radionuclide uptake. American Journal of Pathology 745–755. (https://doi.org/10.1016/S0002-9440(10)64017-5)
Leick MB, Shoff CJ, Wang EC, Congress JL & Gallicano GI 2012 Loss of imprinting of IGF2 and the epigenetic progenitor model of cancer. American Journal of Stem Cells 59–74.
Li H & Durbin R 2009 Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 1754–1760. (https://doi.org/10.1093/bioinformatics/btp324)
Massague J 2012 TGFbeta signalling in context. Nature Reviews: Molecular Cell Biology 616–630. (https://doi.org/10.1038/nrm3434)
Maxwell JE, Sherman SK, Li G, Choi AB, Bellizzi AM, O’Dorisio TM & Howe JR 2015 Somatic alterations of CDKN1B are associated with small bowel neuroendocrine tumors. Cancer Genetics 208 564–570. (https://doi.org/10.1016/j.cancergen.2015.08.003)
Miao R, Luo H, Zhou H, Li G, Bu D, Yang X, Zhao X, Zhang H, Liu S, Zhong Y, et al.2014 Identification of prognostic biomarkers in hepatitis B virus-related hepatocellular carcinoma and stratification by integrative multi-omics analysis. Journal of Hepatology 840–849. (https://doi.org/10.1016/j.jhep.2014.05.025)
Modlin IM, Lye KD & Kidd M 2003 A 5-decade analysis of 13,715 carcinoid tumors. Cancer 934–959. (https://doi.org/10.1002/cncr.11105)
Neklason DW, VanDerslice J, Curtin K & Cannon-Albright LA 2016 Evidence for a heritable contribution to neuroendocrine tumors of the small intestine. Endocrine-Related Cancer 93–100. (https://doi.org/10.1530/ERC-15-0442)
Pfragner R, Wirnsberger G, Niederle B, Behmel A, Rinner I, Mandl A, Wawrina F, Luo J, Adamiker D, Hoger H, et al.1996 Establishment of a continuous cell line from a human carcinoid of the small intestine (KRJ-I). International Journal of Oncology 513–520. (https://doi.org/10.3892/ijo.8.3.513)
Pfragner R, Behmel A, Hoger H, Beham A, Ingolic E, Stelzer I, Svejda B, Moser VA, Obenauf AC, Siegl V, et al.2009 Establishment and characterization of three novel cell lines - P-STS, L-STS, H-STS - derived from a human metastatic midgut carcinoid. Anticancer Research 1951–1961.
Scarpa A 2019 The landscape of molecular alterations in pancreatic and small intestinal neuroendocrine tumours. Annales d’Endocrinologie 153–158. (https://doi.org/10.1016/j.ando.2019.04.010)
Sei Y, Zhao X, Forbes J, Szymczak S, Li Q, Trivedi A, Voellinger M, Joy G, Feng J, Whatley M, et al.2015 A hereditary form of small intestinal carcinoid associated With a germline mutation in inositol polyphosphate multikinase. Gastroenterology 67–78. (https://doi.org/10.1053/j.gastro.2015.04.008)
Shen HC, He M, Powell A, Adem A, Lorang D, Heller C, Grover AC, Ylaya K, Hewitt SM, Marx SJ, et al.2009 Recapitulation of pancreatic neuroendocrine tumors in human multiple endocrine neoplasia type I syndrome via Pdx1-directed inactivation of Men1. Cancer Research 1858–1866. (https://doi.org/10.1158/0008-5472.CAN-08-3662)
Strosberg JR, Halfdanarson TR, Bellizzi AM, Chan JA, Dillon JS, Heaney AP, Kunz PL, O’Dorisio TM, Salem R, Segelov E, et al.2017 The North American Neuroendocrine Tumor Society consensus guidelines for surveillance and medical management of midgut neuroendocrine tumors. Pancreas 707–714. (https://doi.org/10.1097/MPA.0000000000000850)
Tran VG, Court F, Duputie A, Antoine E, Aptel N, Milligan L, Carbonell F, Lelay-Taha MN, Piette J, Weber M, et al.2012 H19 antisense RNA can up-regulate Igf2 transcription by activation of a novel promoter in mouse myoblasts. PLoS ONE e37923. (https://doi.org/10.1371/journal.pone.0037923)
Wang GG, Yao JC, Worah S, White JA, Luna R, Wu TT, Hamilton SR & Rashid A 2005 Comparison of genetic alterations in neuroendocrine tumors: frequent loss of chromosome 18 in ileal carcinoid tumors. Modern Pathology 1079–1087. (https://doi.org/10.1038/modpathol.3800389)
Wells SA, Jr, Pacini F, Robinson BG & Santoro M 2013 Multiple endocrine neoplasia type 2 and familial medullary thyroid carcinoma: an update. Journal of Clinical Endocrinology and Metabolism 3149–3164. (https://doi.org/10.1210/jc.2013-1204)
Wong C, Tang LH, Davidson C, Vosburgh E, Chen W, Foran DJ, Notterman DA, Levine AJ & Xu EY 2020 Two well-differentiated pancreatic neuroendocrine tumor mouse models. Cell Death and Differentiation 269–283. (https://doi.org/10.1038/s41418-019-0355-0)
Zhao J, de Krijger RR, Meier D, Speel EJ, Saremaslani P, Muletta-Feurer S, Matter C, Roth J, Heitz PU & Komminoth P 2000 Genomic alterations in well-differentiated gastrointestinal and bronchial neuroendocrine tumors (carcinoids): marked differences indicating diversity in molecular pathogenesis. American Journal of Pathology 1431–1438. (https://doi.org/10.1016/S0002-9440(10)64780-3)
