Abstract

The tumorigenesis of gastrointestinal and pancreatic tumors (GEPNETs) is poorly understood. We need a better understanding of the molecular alterations in GEPNETs to obtain an accurate classification, and it may also provide targets for therapeutic intervention. Purpose of Paper. The purpose of this paper was to critically examine recent advances in the molecular understanding of GEPNETs gained from genome-wide and transcriptome-wide profiling studies. Special emphasis was put on diagnostic, predictive, and therapeutic implications of profiling studies. Results. Pancreatic neuroendocrine tumours (PNETs) were characterised by a distinct pattern of chromosomal alterations and a higher degree of chromosomal instability (CIN) than ileal carcinoids. Subgroups of PNETs and ileal carcinoids were identified on the basis of specific chromosomal alterations. Exome sequencing identified mutations in MEN1, ATRX/DAXX, and mTOR pathway genes as being frequent events in sporadic PNETs. Expression profiles of PNETs and ileal carcinoids were found to be different, and allowed identification of subgroups of tumors, as well discrimination between benign and malignant tumors. The molecular data provided a number of candidate genes and pathways suitable for targeted therapy. For PNETs, candidate targets include BRAF, KRAS, TERT, EGFR, RET, MDM2, IGF, MET/HGF, ANG2, LCK, PDGFRB, AKT-mTOR, and SSTR2. Some of these targets have already been evaluated in clinical trials (mTOR and SSTR2). For ileal carcinoids, significantly fewer candidate targets were provided, including ERBB2 (HER2), RET, APLP1, and Notch. Conclusion. Profiling of GEPNETs is a powerful tool for discovery of novel targets for therapeutic intervention. Further studies, combining genome, epigenome, transcriptome, and proteome data are needed to enable us to identify clinically relevant targets in GEPNETs.

1. Introduction

Gastrointestinal and pancreatic neuroendocrine tumors (GEPNETs) are rare, and they account for 1-2% of all gastrointestinal malignancies. The incidence of pancreatic neuroendocrine tumors (PNETs) is 0.3 per 100,000 inhabitants per year and for ileal carcinoids the incidence is 0.7 per 100,000 inhabitants per year [1]. Mean age at diagnosis is 59 years for PNETs and 65 years for ileal carcinoids. The clinical presentation may be highly variable with some tumors following an indolent course and other tumors following an aggressive course with rapid tumor dissemination [2]. At diagnosis, there is distant disease in 64% of PNETs and in 30% of ileal carcinoids. The median survival for well-differentiated tumors is 42 months for PNETs and 88 months for ileal carcinoids [1]. Despite there being considerable differences in clinical presentation and biological behaviour, GEPNETs are classified according to a unified histopathological system instigated by the World Health Organisation [3]. All GEPNETs are regarded as being potentially malignant and they are graded on the basis of mitotic count and Ki67 index into two categories: neuroendocrine tumor (NET) of grades 1-2 and neuroendocrine carcinoma (NEC) of grade 3. However, there is a need for molecular classification of GEPNETs, which could improve the diagnostic accuracy and prognostic value in addition to providing information regarding therapeutic options. The advances in genome technology have greatly improved our understanding of the genetic alterations underlying the development of tumours [4]. A number of high-throughput techniques are available for the analysis of molecular alterations in cancers including array-based comparative genomic hybridisation (aCGH), single nucleotide polymorphism (SNP) arrays [5, 6], next-generation sequencing [4], expression arrays [5, 7], and protein profiling [8]. These technologies have only recently been applied to GEPNETs. A search on PubMed performed in September 2011 revealed 24 publications describing studies on PNETs or ileal carcinoids that used high-throughput technologies. Nine of these studies used aCGH/SNP arrays, 14 studies used expression arrays, and one study used next-generation sequencing. There were no studies on protein profiling. A summary of the studies on GEPNETs, togethe with a critical evaluation is given below.

2. Genomic Profiling of PNETs

PNETs may be part of hereditary cancer syndromes, for example, multiple endocrine neoplasia type 1 (MEN1), von Hippel-Lindau disease (VHL), neurofibromatosis type 1 (NF1), or tuberous sclerosis (TSC) [912]. The tumor suppressor genes responsible for these cancer syndromes have been cloned and characterised [1315]. However, a majority of PNETs (90%) occur as sporadic tumors without family history. Sporadic tumors may harbour mutations in the MEN1 gene (15–30%), while mutations in the VHL, NF1, or TSC genes are uncommon [1620]. Genetic characterisation of sporadic PNETs using LOH and conventional CGH has demonstrated recurrent chromosomal gains and losses in sporadic PNETs [2127]. However, introduction of array-based CGH and SNP arrays have greatly enhanced the accuracy by which chromosomal alterations can be determined in tumors. A summary of recent profiling studies on PNETs is given in Table 1.

Jonkers et al., 2006 [28] analysed 30 sporadic insulinomas with array CGH and found that chromosomal instability (CIN), defined as ≥20 chromosomal alterations per tumor, and telomeric loss, defined as ≥6 losses per tumor, were correlated with tumor progression and metastasis. Loss of 22q11.21–13.31 and gain of 7p21.1–11.2 were the best markers of malignant behaviour. A number of potential targets were identified including BRAF and KRAS, for which therapeutic agents are available [29, 30]. In a subsequent study, Jonkers et al. [31] analysed 27 sporadic insulinomas by array CGH and identified gain of 9q32 and loss of 22q13.1 as early events in tumor formation, while loss of 11q24.1 and 22q13.31 was a late event during tumor progression. ShcC was identified as a candidate gene in regions of chromosomal gains (9q22.2–q33.2). ShcC is a substrate for RET suggesting that inhibition of RET signaling is a potential therapeutic principle [3234].

Nagano et al., 2007 [35] analysed 15 PNETs by SNP-array and reported frequent losses on chromosomes 1, 3, 11, and 22 and gains on chromosomes 5, 7, 12, 14, 17, and 20. A number of putative tumor suppressor genes were located in regions of loss, including RASSF1A and PTEN. Putative therapeutic targets in regions of gain included TERT, EGFR, and RET, for which inhibitors are available [3638]. Hu et al. 2010 [39] performed array CGH analysis of 55 PNETs, confirming losses on chromosomes 1, 2, 7, 10, 11, 16, 17, 19, 20, and 22 and gains on chromosomes 1–10, 12–14, 17, and 18. Amplified genes included MDM2, MDM4, and WIP1 suggesting attenuation of p53 function in tumors. MDM2 was suggested as a potential therapeutic target [40].

Jiao et al., 2011 [41] performed whole-exome sequencing of sporadic PNETs and identified mutations in chromatin remodeling genes, MEN1 and DAXX/ATRX, in 44% and 43% of tumors. Tumors with DAXX/ATRX mutations had altered telomeres, which are maintained by a telomerase-independent mechanism (ALT) [42]. PNETs with mutated MEN1 or DAXX/ATRX were associated with prolonged survival compared to tumors without mutations. A subset of PNETs harboured mutations in PTEN, TSC2, and PIK3CA, which are part of the AKT-mTOR signaling pathway. Inhibition of mTOR was suggested as a possible treatment, and clinical trials using mTOR inhibitors for advanced PNETs has shown promising results [43, 44]. The relationship between mutational status in PNETS and response to mTOR inhibition remains to be evaluated.

In general, genomic profiling studies have provided consistent data, which allows construction of genomic progression models for PNETs (Figure 1). However, a detailed analysis of genomic alterations in PNET subgroups still remains to be done. Promising therapeutic targets have been generated, which should be subjected to preclinical or clinical testing.

3. Expression Profiling of PNETs

There have been a number of studies describing expression profiles in PNETs. These studies are summarised in Table 2. Maitra et al. [45], Hansel et al. [46], Bloomston et al. [47], Dilley et al. [48], Capurso et al. [49], Lowe et al. [50] and de Sá et al. [51] have analysed small numbers of PNETs and compared them with normal islets or normal pancreas. Significantly regulated genes were reported, as well as candidate genes for targeted therapy. The IGF-signalling pathway was suggested as a target by Maitra et al. [45], Hansel et al. [46], and Couvelard et al. [52], and HGF/MET signaling was suggested for targeting by Hansel et al. [46] and Dilley et al. [48]. For both of these signalling pathways, inhibitors are available for preclinical or clinical testing [5357]. Targeting ANG2 was suggested by Bloomston et al. [47], and inhibitors are available for testing [58]. Capurso et al. [49] suggested LCK (src family member) as target, and src inhibitors are also available for testing [59].

Couvelard et al. [52] analysed a large series of benign and malignant PNETs and identified a number of differentially expressed genes, of which CD34, E-selectin, MKK4, and MDR1 were verified. IGF1 was suggested as a promising target for therapy. Duerr et al. [60] studied a series of PNETs and could differentiate the expression profile of benign tumors from those of malignant tumors by hierarchical clustering. Malignant tumors showed high expression of FEV, ADCY2, NR4A2, and GADD45β. PDGFRβ was shown to be activated by phosphorylation in a high proportion of PNETs, suggesting PDGFR β as a target for therapy. Missiaglia et al. [61] have performed expression profiling on the largest series of PNETs and were able to separate insulinomas from nonfunctioning tumors by hierachical clustering. PTEN and TSC2 was shown to be downregulated in PNETs and found to be correlated with short survival. PTEN and TSC2 are part of the AKT-mTOR signaling pathway and inhibition of mTOR was suggested for therapy. These data are highly concordant with the findings in the whole exome sequencing study by Jiao et al. [41]. SSTR2 is also highly expressed in non-functioning PNETs and has confirmed the suitability of somatostatin-receptor-mediated therapy for PNETs [62]. Roldo et al. [63] have examined the expression of miRNAs in a large series of PNETs and reported miR-204 to be a marker for insulinomas, and miR-21 to be a marker of metastatic behaviour. No therapeutic targets were identified.

In general, expression profiling studies on PNETs have provided less consistent data than genomic studies. This could be due to the low number of tumors studied or to inherent differences in the profiling techniques. Larger groups of well-characterised tumours should be studied using optimally designed studies, or alternatively by pooling of different studies in a meta-analysis.

4. Genomic Profiling of Ileal Carcinoids

Ileal carcinoids present as sporadic tumors, and familial cases are very rare. Genetic analysis of sporadic ileal carcinoids using LOH and conventional CGH has shown recurrent chromosomal aberrations, with loss of chromsome 18 as a frequent and early event in tumorigenesis [6467]. High-resolution CGH and SNP arrays have been used to further analyse chromosomal aberrations in ileal carcinoids. A summary of these recent profiling studies is given in Table 3.

Kim et al. [68] analysed 15 ileal carcinoids by SNP arrays and found loss on chromosomes 9, 16, and 18 and gain on chromosomes 4, 14, 17, and 20. ERBB2 (HER2) was suggested as a candidate gene for targeted therapy. Kulke et al. 2008 [69] analysed 24 ileal carcinoids by SNP-array and confirmed loss on chromosomes 9, 16, and 18 and gain on chromosomes 4, 5, 7, and 14. Hierarchical clustering showed two separate groups of tumors, a smaller group characterised by 2 or more clustered gains on chromosomes 4, 5, 7, and 14 and a larger group without clustered gains. Candidate genes for targeted therapy were from amplified regions on chromosome 14q (DAD1), 11p (OR4A5), and 17q (PRKCA). Andersson et al. [70] analysed 52 ileal carcinoids by array CGH, which showed loss of chromosome 18 as a common and early event in tumor formation. Tumors could be differentiated into two groups, a larger group with loss of chromosome 18 and a smaller group with an intact chromosome 18 but with gain of chromosome 14. Gain of chromosome 14 was also shown to be a strong predictor of short patient survival. Cunningham et al. [71] analysed 61 tumors by array CGH and SNP arrays, and found similar chromosomal alterations in sporadic and familial ileal carcinoids. Again, loss of chromosome 18 was shown to be a frequent and early event in tumor formation with minimal regions of deletions at 18q21.1–q21.31, 18q22.1–q22.2, and 18q22.3–q23.

In general, genomic profiling studies have provided concordant data from large numbers of tumors, allowing construction of a genomic progression model for ileal carcinoids (Figure 2). However, the majority of alterations involved whole chromosomes or large parts of chromosomes, making identification of individual oncogenes or tumor suppressor genes difficult. This is particularly true for chromosome 18, where identification of a putative tumor suppressor gene is still lacking. Also, for this reason, few candidate genes have been generated for targeted therapy.

5. Expression Profiling of Ileal Carcinoids

A limited number of studies describing expression profiles in ileal carcinoids have been published. These studies are summarised in Table 4. Three studies, Duerr et al. [60], Arvidsson et al. [72], and Leja et al. [73], have analysed small series of ileal carcinoids, generating a limited number of candidate targets for therapy, including RET signaling, Notch signalling, and APLP1. Edfeldt et al. [74] reported a comprehensive study of 42 ileal carcinoids that differentiated tumors as belonging to three different groups by principal component analysis. Genes discriminating the three groups included TUSC2, RUNX1, TPH1, TGBR2, and CDH6. Reubel et al. [75] have analysed miRNA expression in 16 ileal carcinoids and they were able to show that expression of miRNA-133a is downregulated during tumor progression.

In general, expression profiling studies on ileal carcinoids have provided less consistent data than genomic studies. This is most likely due to the small number of tumors studied in addition to differences in the profiling techniques.

6. Summary

Genomic profiling showed losses on chromosomes 1, 3, 11, and 2 and gains on chromosomes 5, 7, 12, 14, 17, and 20 in the early development of PNETs. Telomeric losses, attenuation of p53 function, and chromosomal instability (CIN) occurred during tumor progression. Losses on chromosomes 3, 6q, and 21q and gains on chromosomes 4 and 7 characterised metastatic tumors. Inactivating mutations of MEN1 and ATRX/DAXX were common, while inactivating mutations of genes in the mTOR pathway (PTEN, TSC2, and PIK3CA) occurred in a subset of tumors. PNETs of B-cell phenotype (insulinoma) showed different genetic alterations compared to the majority of PNETS, with loss on chromosome 22 and gain on chromosome 9q as early events in tumorigenesis. Telomeric losses, chromosomal instability, loss on chromosome 3p, 6q, 11q, and 22q, and gain on chromosome 7p occurred during malignant progression. Expression profiling of PNETs showed different subgroups by hierarchical clustering, differentiating benign PNETs from malignant PNETs, and non-functioning PNETs from functioning PNETs (insulinoma). Genes associated with malignant PNETs included MET, IGFBP3, CD34, E-selectin, MKK4, MDR1, FEV, ADCY, NR4A2, GADD45β, and FGF13. Expression of miRNA-21 was also associated with metastatic behaviour. In functioning PNETs (insulinoma), expression of SERPINA1 (alpha-1-antitrypsin) was correlated with malignancy. Profiling studies have suggested a number of candidates to be evaluated for targeted therapy of PNETs. These targets and signaling pathways include BRAF, KRAS, TERT, EGFR, RET, MDM2, IGF, MET/HGF, ANG2, LCK, PDGFRB, AKT-mTOR, and SSTR2.

Genomic profiling of ileal carcinoids showed alterations mainly involving large parts of chromosomes or whole chromosomes. The number of alterations per tumor was lower than that for PNETs. Ileal carcinoids could be separated into two groups by hierarchical clustering. The majority of ileal carcinoids showed loss on chromosome 18, which is an early and most likely pathogenetic event in this group of tumors. Minimal regions of deletions on chromosome 18 included 18q21.1–q21.31, 18q22.1–q22.2, and 18q22.3–q23. No mutations have so far been detected in genes located in these regions. During tumor progression, chromosomal alterations accumulate, most frequently loss on chromosomes 3p, 11q, and 13. A small proportions of ileal carcinoids show clustered gains on chromosomes 4, 5, 7, 14, and 20. These alterations may be present early in tumor development. Chromosome 18 was usually intact in this group of tumors. Gain of chromosome 7 was associated with solid growth pattern and metastatic behaviour, while gain of chromosome 14 was correlated with shorter patient survival. Expression profiles of ileal carcinoids were different from those of PNETs and allowed differentiation of carcinoid tumors into three groups by principal component analysis. During tumor progression, there was downregulation of expression of CXCL14, NKX2-3, ACTG2, GREM2, REG3A, and miRNA133a. Profiling studies have suggested a limited number of candidates for targeted therapy including ERBB2 (HER2), RET, APLP1, and Notch.

7. Conclusion

High-throughput technologies have greatly facilitated the molecular characterisation of GEPNETs. Recent studies applying genomic and expression profiling to GEPNETs have permitted molecular classification of tumors that goes beyond the traditional histopathological classification. However, data from different studies, notably expression studies, are not always concordant and provide limited information on candidates for targeted therapy. Thus, additional high-quality studies are needed, which should include large numbers of clinically well-characterised tumors, stringent study design and bioinformatics analysis. Future studies must also address stromal cells, cancer stem cells, and bulk cells separately. The recent identification of cancer stem cells by Gaur et al. [76] in intestinal carcinoids will stimulate such studies.