Endocrine-Related Cancer (2011) 18 479–489

Different expression profiles in metastasizing midgut carcinoid tumors

Katarina Edfeldt, Peyman Bjo¨rklund, Go¨ran A˚ kerstro¨m, Gunnar Westin, Per Hellman and Peter Sta˚lberg

Department of Surgical Sciences, Uppsala University, SE-751 85 Uppsala, Sweden (Correspondence should be addressed to P Sta˚lberg; Email: [email protected])

Abstract The genetic events leading the progression of midgut carcinoid tumors are largely unknown. The disease course varies from patient to patient, and there is a lack of reliable prognostic markers. In order to identify involved in tumor progression, profiling was performed on tumor specimens. Samples comprised 18 primary tumors, 17 lymph node (LN) metastases, and seven liver metastases from a total of 19 patients. Patients were grouped according to clinical data and histopathology into indolent or progressive course. RNA was subjected to a spotted oligo microarray and B-statistics were performed. Differentially expressed genes were verified using quantitative real-time PCR. Self-organizing maps demonstrated three clusters: 11 primary tumors separated in one cluster, five LN metastases in another cluster, whereas all seven liver metastases, seven primary, and 12 LN metastases formed a third cluster. There was no correlation between indolent and progressive behavior. The primary tumors with Ki67 O5%, with low frequency of the carcinoid syndrome, and a tendency toward shorter survival grouped together. Primary tumors differed in expression profile from their associated LN metastases; thus, there is evidence for genetic changes from primary tumors to metastases. ACTG2, GREM2, REG3A, TUSC2, RUNX1, TPH1, TGFBR2, and CDH6 were differentially expressed between clusters and subgroups of tumors. The expression profile that assembles tumors as being genetically similar on the RNA expression level may not be concordant with the clinical disease course. This study reveals differences in gene expression profiles and novel genes that may be of importance in midgut carcinoid tumor progression. Endocrine-Related Cancer (2011) 18 479–489

Introduction a lethal outcome; overall five-year survival is w60% Intestinal carcinoid tumors originate from the neuro- (Modlin et al.2003). The annual incidence is about two endocrine enterochromaffin (EC) cells of the gut. The cases per 1 00 000 persons, but the rate is increasing, WHO classification from 2000 divides the tumors into probably due to increased sensitivity and awareness in three groups: well-differentiated tumors, well-differ- the diagnostic phase. However, there is still often a entiated endocrine carcinomas, and poorly differen- malignant spread to regional lymph nodes (LNs) and tiated carcinomas (Kloppel et al. 2004). The older and the liver at the time of diagnosis, because the available previously more commonly used classification divides biomarkers are of low specificity and the clinical the tumors according to their embryologic origin into course often diffuses (Schnirer et al. 2003). foregut, midgut, and hindgut tumors, which also The tumors are neoplasms of peptide- and amine- reflects its localization in the gastrointestinal tract producing cells. Some patients will develop the carcinoid (Arnold 2005). syndrome due to excessive secretion of serotonin and Midgut carcinoid tumors are the most common type other metabolites, which includes flushing, diarrhea, of carcinoid tumors in the gastrointestinal tract and abdominal pain, bronchospasm, and heart disease and arise in the lower jejunum, ileum, appendix, and other symptoms resulting from fibrosis (Soga et al. cecum. This malignant neoplasm grows slowly (Ki67 1999). Symptoms and progression rate can greatly differ proliferating index is often !2%) but nevertheless has between individual patients and is almost impossible

Endocrine-Related Cancer (2011) 18 479–489 Downloaded from Bioscientifica.comDOI: 10.1530/ERC-10-0256 at 09/27/2021 05:42:19AM 1351–0088/11/018–479 q 2011 Society for Endocrinology Printed in Great Britain Online version via http://www.endocrinology-journals.orgvia free access K Edfeldt et al.: Gene expression profiles in carcinoid tumors to predict due to lack of reliable prognostic indicators. EC cells and/or bowel mucosa (Kidd et al. 2006, Leja The biomarkers used correlate to total tumor burden, et al. 2009), and in one microarray, the gene expression such as chromogranin A in serum and 5-hydroxy- was compared with a cell line (GOT1; Arvidsson et al. indoleacetic acid (5-HIAA) in urine. Several analyses 2008). It is difficult to extract normal EC cells from the have stated that factors leading to a bad prognosis bowel mucosa, because these cells are scattered are advanced age, LN involvement, presence of more throughout the gastrointestinal mucosa, and isolating than five liver metastases, lack of symptoms at the time these cells will result in contamination from surround- of diagnosis, high levels of 5-HIAA, high levels of ing tissue. Owing to these difficulties, this study has a plasma chromogranin A or neuropeptide K, and the new approach. To clarify the gene expression pattern in carcinoid syndrome (Janson et al.1997, Soreide et al. midgut carcinoid tumors, with emphasis on genes 2000, Hellman et al.2002, Kolby et al.2004, Ahmed involved in progression and aggressiveness, an et al.2009). expression microarray assay was performed comparing Today, surgery is the only potential cure. To achieve primary midgut carcinoid tumors with metastases and prolonged survival and symptom relief, the patients are comparing groups of tumors with different clinical treated with cytotoxic agents, biological therapies, and features. tumor-targeted radionucleotides (Schnirer et al. 2003). Liver metastases can be treated with surgery, radio- frequency ablation, or liver embolization (Eriksson Materials and methods et al. 2008). Nevertheless many tumors are resistant to therapy and there is an urgent need to develop Tumor material and RNA preparation new therapies. Primary and metastatic midgut carcinoid tumors were Besides the lack of diagnostic and therapeutic tools, snap frozen in liquid nitrogen and kept at K70 8C. The a greater understanding of molecular arrangements in tumors came from a cohort of patients who were midgut carcinoid tumors is required. Loss of chromo- diagnosed with midgut carcinoid tumor and operated at some 18q has been found to be the most frequent Uppsala University Hospital between 1989 and 2007. genetic aberration (Zhao et al. 2000, Kytola et al. 2001, Patient records were evaluated, and totally 42 tumors Lollgen et al. 2001, Tonnies et al. 2001, Wang et al. from 19 patients were selected for further analysis. 2005, Kulke et al. 2008). No alterations in the tumor Clinical features are shown in Table 1; two patients had suppressor genes DPC4/SMAD4 and DCC, located on been treated with interferon a prior to surgery and ten 18, are found indicating that other, with somatostatin analogues. According to the WHO currently unknown genes are important for patho- classification, all of the tumors were well-differentiated genesis (van Eeden & Offerhaus, 2006). Moreover, neuroendocrine carcinomas. Endocrine tumors have previous research has treated neuroendocrine neo- often been treated as a homogenous group in the plasms as a homogenous group of tumors. This is most literature, but resent research has stated their hetero- likely insufficient, because recent studies have demon- geneity; thus, pancreatic neuroendocrine tumors are strated the heterogeneous nature and different gene foregut and not midgut carcinoid tumors. In 11 patients expression profiles among subgroups of tumors (Zhao (no. 1–5, 7–10, 17, and 19), RNA was extracted from et al. 2000, Kytola et al. 2001, Tonnies et al. 2001, both primary tumors and LN metastases, and from five Wang et al. 2005). patients (no. 11–15), RNA was extracted from the To our knowledge, only three expression microarray primary tumor, LN, and liver metastases. From one assays, which include midgut carcinoid tumors, have patient (no. 6), tumor material came from only the been performed until today. The gene expression primary tumor, and from another patient (no. 16), the profiles have only been analyzed in a limited number tumor material came from the primary tumor and liver of tumors; in total, three primary tumors, 16 liver metastasis, and from one patient (no. 18) RNA was metastases, and none LN metastases. There are some extracted from LN and liver metastasis. Total RNA suggestions of novel genes (PNMA2, SPOCK1, was extracted using TriZol Reagent (Invitrogen) SERPINA10, GRIA2, GRP112, OR51E1, CXCL14, according to the manufacturer’s instructions. Sections NKX23, NAP1L1, MAGE-D2, MTA-1, and APLP1) from all the tumors were stained with Mayer’s that may be of importance for tumorigenesis, but their hematoxylin and histopathologically evaluated, and role in the development of midgut carcinoid tumors is the tissues used contained at least 80%, and in most still unknown (Kidd et al. 2006, Arvidsson et al. 2008, cases more than 90%, of neoplastic cells. When Leja et al. 2009). In two of the microarrays, the gene needed, the tumor was macroscopically resected from expression in tumor tissues was compared with normal excessive stromal tissue, blood vessels, bowel mucosa,

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access 480 www.endocrinology-journals.org Endocrine-Related Cancer (2011) 18 479–489

lymphoid, and liver tissue. Quality and quantity of the extracted RNA was assessed by Agilent 2100 status Disease Bioanalyzer (Aglient Technologies, Santa Clara, CA, USA). Only one LN metastasis, of the 17 LN in total, contained lymphatic tissue (! 5%; tumor no. 18).

after To identify genes that are important for tumor months Follow-up surgery in progression and prognosis, the patients were classified and divided into two groups according to clinical data. A group with six patients featured a more aggressive after diagnosis Follow-up in months clinical pattern and a second group with ten patients had a less aggressive clinical behavior. The clinical criteria are viewed in Table 2. Of the clinical criteria, Radical surgery progression was defined according to RECIST 1.1 (Eisenhauer et al. 2009), and slow progression was defined as progression at more than 12-month intervals Age at surgery on radiological imaging and three patients that were included in the microarray did not classify into any of the groups (patient no. 12, 14, and 18). According to Age at

diagnosis the TNM staging for midgut endocrine tumors (Rindi

b et al. 2007), 18 patients were classified as stage IV and one patient (no. 2) was classified as IIIB (and displayed mol/day) 10 67 67 Yes 108 105 Stable 10 68 68 Yes 91 91 Progressive

U-5-HIAA no clinically evident liver metastases). m ( ! !

Gene expression array a Microarrays including a Oligo Set 1% 96 55 56 No 48 37 Dead 1% 1600 60 60 Yes 109 106 Dead 1% 71 74 75 Yes 93 94 Stable 1% 22 62 63 No 84 75 Stable 1% 131 77 77 No 110 110 Dead 1% 584 55 58 Yes 108 76 Stable 1% NA 51 51 Yes 46 45 Stable 1% 571 52 53 Yes 75 66 Stable 0.1% 116 46 46 Yes 50 50 Stable 1% 1% 229 61 61 No 37 33 Stable 1% 1% 80 66 66 Yes 39 35 Stable 1% NA 45 45 Yes 48 48 Stable 1% NA 66 67 No 36 25 Dead ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! containing 24 650 genes and 37 123 gene transcripts (version 3.0) were performed. The arrays were printed with a QArray2 (Genetix, Hampshire, UK) instrument Carcinoid syndrome Ki67 with 48 K2805 pins (Genetix) on Ultra GAPS slides (Corning, Lowell, MA, USA). Detailed protocols and the complete gene list are available at http://www. biotech.kth.se/molbio/microarray. Neoplasm samples were labeled with Cy5 and RNA reference with Cy3. PCA

cluster Gene expression analysis metastases Gender lymph node Image analysis was performed in GenePix 5.1 Soft- ware (Axon Instruments, Sunnyvale, CA, USA). After PCA

tumor ! cluster

primary removal of bad quality spots (if 70% of foreground pixels were below background intensity plus 2 S.D. in both channels or if the difference between ratio of medians and regression ratio exceeded 20% in one of liver Tumor no:

metastases the channels), analysis of the gene expression data was carried out in the freely available statistical computing language R (http://www.r-project.org) using a package available from the Bioconductor project (www. Tumor no: metastases lymph node bioconductor.org). The raw data was normalized using the LOWESS method (Yang et al. 2002). In order to search for differentially expressed genes tumor of primary Tumor no: between groups, an empirical Bayes moderated t test Patient characteristics was applied, using the Limma package (Smyth

50 normal reference value, measured in urine. 2004). To address the problem of multiple testing, Number in parenthesis refers! to hot spots. 123 1 2 3 26 22 34 NA NA NA 1 1 1 3 2 3 Female Female Male No No Yes 1–3% (17–18%) 3% (7%) 115 29 76 76 49 Yes 49 Yes 48 46 41 Dead 46 Stable 4 4 27 NA 1 3 Male No 567 5 6 7 30 NA 25 NA NA NA 1 1 1 3 – 3 Female Female Female No No No Varying (10%) 1–2% NA 104 63 70 63 Yes 70 Yes 96 99 92 95 Dead Dead 8 8 29 NA 1 3 Female Yes 9 9 28 NA 1 3 Male No 10 10 32 NA 1 3 Female No 11 11 18 36 3 2 Female Yes 12 12 19 37 3 2 Female No 13 13 23 38 3 3 Female No 14 14 24 39 3 3 Female Yes Table 1 Patient no. a b 15 15 20 40 3 2 Male Yes 16 16 NA 35 3 – Male Yes 17 17 31 NA 3 3 Male Yes 18 NA 21 41 – 2 Female No 19 42the 33P values NA were adjusted 1 using 3 the FemaleBenjamini Yes and

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access www.endocrinology-journals.org 481 K Edfeldt et al.: Gene expression profiles in carcinoid tumors

Table 2 Clinical parameters for expression array comparisons (patient no. 12, 14, and 18 did not fit this classification and were exempt from this comparison)

More aggressive R3 criteria Less aggressive R3 criteria Patient no. 1, 3, 8, 9, 15, and 42 Patient no. 2, 4, 5, 6, 7, 10, 11, 13, 16, and 17

O5 liver metastases !5 liver metastases O70 years old at diagnosis !70 years old at diagnosis Fast progression Slow progressiona Carcinoid syndrome No carcinoid syndrome Severe diarrheab Mild or no diarrhea Ki67 O5% Ki67 !5% No radical surgery Radical surgery aSlow progression defined as progression according to RECIST with more than 12-month intervals. bMore than 5 per day.

Hochberg method (Benjamini 1995). Hierarchical tumorigenesis, were selected for further analyses. clustering of genes for three-dimensional visualization Quantitative real-time PCR (RT-qPCR) was per- of expression pattern was performed by the principal formed to verify the microarray results and relative component analysis (PCA). PCA is used to reduce the mRNA expression was determined. DNA-free RNA dimensionality of a data set consisting of a large was prepared using the Nucleospin RNA II kit number of variables, while retaining as much as (Macherey-Nagel GmbH & Co. KG, Du¨ren, Germany). possible the variation present in the data set. The Successful DNase treatments were established by PCR data is transformed to a new set of variables, the analysis of all RNA preparations. Reverse transcription principal components (PCs), which are uncorrelated. of total DNA-free RNA was performed with random This simplifies the structure of the data and visualizes hexamer primers using the First-Strand cDNA hidden structures. PCA mapping was observed in Synthesis kit (GE Healthcare, Buckinghamshire, three-dimensional space in which the x-, y-, and z-axes UK) according to the manufacturer’s instructions. represent the first, second, and third PC respectively RT-qPCR reaction was performed on step 1 RT-PCR (Ma & Dai 2011). system (Applied Biosystems, Foster city, CA, USA). In addition to cDNA, all PCR reactions contained enrichment and gene network 200 mM of each primer and 1 ml probe and TaqMan analysis Gene Expression Master Mix (Applied Biosystems). All samples were amplified in triplicate, and non- The web-based gene ontology (GO) analysis tool template controls were included. Each sample’s DAVID (http://david.abcc.ncifcrf.gov/; Dennis et al. mean threshold value was corrected for the corres- 2003, Huang da et al. 2009) and Ingenuity Pathways ponding mean value for GAPDH rRNA, used as Analysis (Ingenuity Systems, www.ingenuity.com) internal control. were used for functional annotations and to determine The following labeled TaqMan probes were used statistical overrepresentation of GO terms among for step 1 RT-PCR System: GAPDH (4333764F), genes that differed between groups of tumors. A fold gamma 2 isoform (ACTG2; Hs00242273_m1), gremlin change cutoff of 2 was set to identify genes whose 2(GREM 2; Hs01934663_s1), regenerating islet- expression was differentially regulated and a P value of derived 3 alpha (REG3A; Hs00170171_m1), tumor !0.05 was considered significant. For all analysis, the suppressor candidate 2 (TUSC2; Hs00200725_m1), GO terms biological process, cellular compartment, runt-related transcription factor 1 (RUNX1; Hs01021 and top canonical pathways were considered. To map 971_m1), tryptophan hydroxylase 1 (TPH1; Hs0018 differentially expressed genes with genes already 8220_m1), transforming growth factor-b receptor II known to be involved in tumorigenesis, the Compu- (TGFBR2; Hs00559661_m1), and cadherin 6 (CDH6; tational biology center web site was used (http://cbio. Hs00191832_m1). mskcc.org/CancerGenes/SelectAndDisplay.action; U Higgins et al. 2007). Non-parametric Mann–Whitney test or Student’s t test was used to test significance between expression levels for RT-qPCR and Kaplan–Meier to test Quantitative real-time RT-PCR analysis differences in survival among clusters. Statistics were Genes with a fold change difference O2 and with calculated using SPSS (Chicago, IL, USA) and known gene function, involved in different types of differences were considered significant if P!0.05.

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access 482 www.endocrinology-journals.org Endocrine-Related Cancer (2011) 18 479–489

Tumors with undetectable mRNA levels of the gene metastases, and all seven liver metastases. When of interest were arbitrarily set to a relative value comparing primary tumors with their respective of 0.0001. associated LN metastases, 13 of 16 matched pairs were situated in different clusters, thus expressing a Ethical permission different gene expression profile. Only three of 11 patients with primary tumors in cluster 1 suffered from This study was approved by the local ethical the carcinoid syndrome compared with the rest of committee, Uppsala, and informed consent was the primary tumors (in cluster 3) where five of seven obtained from all patients. patients had the carcinoid syndrome; three tumors displayed a high proliferation rate with Ki67 O5% Results in hot spots and all of these tumors were gathered in cluster 1 and seven patients eventually succumbed Comparison of gene expression profiles of from their carcinoid disease and interestingly all of primary tumors and metastases them resided in cluster 1. However, when analyzing In order to identify differentially expressed genes, the cumulative survival between patients from the two mRNA expression levels of primary tumors and clusters of primary tumors, there was no statistically metastases were globally analyzed. Of the 37 123 significant difference (PZ0.15; Fig. 2). A tendency transcripts, 7130 were significantly (P!0.05) altered toward shorter survival for patients in cluster 1 was between primary tumors and LN metastases. With a observed, but a larger patient cohort is needed to two-time fold change and known gene name, 289 genes investigate this properly. There was no correlation differed in expression level. The number of genes with between radical primary surgery and survival. a significant and at least a 1.5-time fold change difference among the tumor subgroups is summarized Quantitative real-time RT-PCR in Table 3, with detailed data in Supplementary RT-qPCR was performed to verify expression Table 1, see section on supplementary data given at differences of eight genes selected from the identified the end of this article. A comparison of expression clusters of the PCA and from the comparison of all levels between primary tumors in the group defined as primary tumors (nZ18) to all LN metastases (nZ17). less aggressive (number of tumorsZ6; Table 2) and the Geneselectionwasbasedonbothfold-change more aggressive group (number of tumorsZ10) differences and known gene function. ACTG2, revealed no significant differences (Table 3). Three GREM2, and REG3A were selected from the compari- patients (no. 12, 14, and 18) could not be classified son of all primary tumors to LN metastases. TUSC2 according to the set criteria and were therefore was chosen from the comparison of primary tumors excluded from this comparison. The results from the from clusters 1 and 3, and finally RUNX1, TGFBR2, microarray analysis have been deposit on NCBI’s GEO TPH1, and CDH6 were chosen from the comparison of (accession number GSE27162). LN metastases from clusters 2 and 3. All eight genes To further analyze gene expression patterns, we were verified as significantly (P!0.05) differentially performed unsupervised hierarchal cluster analysis expressed (Fig. 3). using the PCA method (Ma & Dai 2011). The expression data was reduced to three PCs that visualized three clusters (Fig. 1). Cluster 1 contained 11 primary GO enrichment tumors; cluster 2 contained five LN metastases, and To characterize the functional significance of genes cluster 3 contained seven primary tumors, 12 LN with a high expression in primary tumors and LN

Table 3 Number of significantly differentially expressed genes between subgroups of tumors from expression-array analysis

No. of genes*

Clinically more aggressive group (nZ6) Clinically less aggressive group (nZ10) 0 Primary tumors (nZ18) Lymph node metastases (nZ17) 289 Primary tumors (nZ16) Paired lymph node metastases (nZ16) 324 Primary tumors (nZ18) Liver metastases (nZ7) 266 Lymph node metastases (nZ17) Liver metastases (nZ7) 11** Interferon-treated tumors (nZ2) Not interferon-treated tumors (nZ2) 7

*P!0.05, fold change 2; **fold change 1.5.

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access www.endocrinology-journals.org 483 K Edfeldt et al.: Gene expression profiles in carcinoid tumors

Primary tumor When using the Computational Biology center web Lymph node metastases Liver metastases site and mapping the gene lists from the expression 50 2 microarray against known cancer genes, several genes 0 –50 1 were identified in the clusters. Gene lists are shown PC3 –100 –150 in Table 4. –80 3 –60

–40 Discussion

–20 P To enhance our understanding of pathophysiological C 2 0

20 –120 arrangements in midgut carcinoid tumors, this study –100 –80 –60 40 –40 aimed at identifying clinically relevant genes involved –20 0 20 PC1 60 40 60 in tumor progression and aggressiveness. Comparing pathological tissue with its normal counterpart is by its Figure 1 Principal component analysis. Principal component very nature a difficult task because the EC cells are analysis revealed three clusters: 11 primary tumors formed scattered throughout the gastrointestinal mucosa and cluster 1; five lymph node metastases formed cluster 2; and seven primary tumors, 12 lymph node metastases, and all every attempt at isolating these cells will result in of the seven liver metastases formed cluster 3. (PC1, 2, and contamination from surrounding tissue, thus making 3Zprincipal component 1, 2, and 3). interpretation difficult. Therefore, the goal in this study was to identify genes involved in tumor progression, metastases, and of significantly differently expressed which would mirror the problem faced by the clinician genes in clusters, GO term analysis was performed and in selecting the patients requiring aggressive clinical GO terms were identified. GO term analysis facilitates treatment from the large group of patients whose the interpretation of data providing biological and disease will follow an indolent course, despite having functional descriptions of gene products (Fig. 4). similar presentation and tumor burden at diagnosis. In Comparing the expression of genes in all of the an attempt to identify distinguishing genes of interest, primary tumors with all of the LN metastases, GO we characterized two groups of patients with aggres- terms that are unique and enriched in primary tumors sive (nZ6) and non-aggressive disease (nZ10). RNA were related to ‘the extracellular region’ and genes was isolated from primary tumors, LN metastases, and with a high expression in LN metastases were liver metastases from these groups (and from another associated with ‘the organelle and cell cycle’. LEF1, three patients) and an oligo microarray analysis was FOS, and MYC genes, known to be involved in performed. Surprisingly, there were no differentially colorectal cancer, were highly expressed in the LN expressed genes in the primary tumors from these metastases. GO enrichment analysis also revealed that two groups of patients when sorted according to genes highly expressed in the primary tumors in cluster 1 were significantly associated with ‘the plasma Survival functions membrane and cell adhesion’ in contrast to primary 1.0 Cluster: primary tumors in cluster 3, which were associated with tumors ‘chromosomal structure, cell cycle, and mitosis’. The 0.8 1 3 four genes involved in the p53 signaling pathway was 1-Censored 3-Censored highly expressed in cluster 3 primary tumors (GTSE1, 0.6 CCNB2, CCNE2, and RRM2), but not in cluster 1. Both the serotonin receptor signaling pathway and the 0.4 regulation of hormone levels were both associated Cum survival with LN metastases in cluster 3. 0.2 Cell adhesion molecules (CAMs) were differentially expressed in all three clusters. In cluster 2, 45 CAMs 0.0 were highly expressed compared with nine different 0 2040608010 12 CAMs in the LN metastases in cluster 3. The primary Follow-up after diagnosis (months) tumors in cluster 1 had a high expression of 21 CAMs compared with none in the primary tumors in cluster 3. Figure 2 Cumulative survival analysis for PCA clusters 1 and 3. Gene lists of differentially expressed CAMs are Kaplan–Meier survival analysis showing a tendency toward shorter survival for patients in cluster 1 compared with primary presented in Supplementary Table 2, see section on tumors in cluster 3; however, the differences were not supplementary data given at the end of this article. statistically significant (PZ0.15).

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access 484 www.endocrinology-journals.org Endocrine-Related Cancer (2011) 18 479–489

A B C 7 10 * 10 *3 10 *5 *8 8 8 8

7 mRNA ratio mRNA ratio mRNA ratio 6 6 6 *

4 4 4

2 2 2

25 27 18 ACTG2/GAPDH * REG3A/GAPDH 0 * GREM2/GAPDH 0 0 * Primary Lymph node Primary Lymph node Primary Lymph node tumors metastases tumors metastases tumors metastases

Figure 3 Relative gene expression. RT-qPCR data on selected genes with comparison between primary tumors (nZ18) and lymph node metastases (nZ17). Relative gene expression ratios are shown adjusted for GAPDH.A,ACTG2;B,GREM2;C,REG3A.*, outlier; filled circle, mean. well-recognized clinical criteria for more or less have a protective function, because the expression aggressive disease. Analyzing primary tumors against levels are high in primary tumors and greatly reduced LN metastases identified 289 genes that were differ- or absent in metastases. entially expressed. Some of these genes could be of ACTG2, GREM2, and REG3A are novel genes not relevance in tumor progression and to verify the earlier known to be involved in carcinogenesis in microarray data and identify genes that warrant further endocrine tumors. These genes have not, to our study, RT-qPCR was performed on ACTG2, GREM2, knowledge, been detected in other microarray studies. and REG3A. ACTG2 is located on chromosome 2p13 Two microarray studies have compared the gene and has been implicated in cisplatin resistance (Watson expression in midgut carcinoid tumors with normal et al. 2007) and cell cycle arrest (Li et al. 2008). mucosa and/or normal EC cells (Kidd et al. 2006, Leja GREM2 is located on chromosome 1q43 and belongs et al. 2009) and one with a cell line from a liver to the cystine knot superfamily. GREM2 has been metastasis (GOT1; Arvidsson et al. 2008). In the work shown to interact with bone morphogenetic by Leja et al. (2009), NKX23 had a lower expression (BMP) in the cross talk with the WNT/b- level in liver metastases compared with primary pathway (Kosinski et al.2007) and to have an tumors, and this is in accordance with the findings in increased expression in colorectal cancer (Segditsas this study; however, this has not been verified by et al. 2008). REG3A, located on chromosome 2p12, is RT-qPCR. None of the expression array has compared involved in liver and pancreatic regeneration and the gene expression between primary tumors and LN proliferation, and is a target of b-catenin signaling in metastases with emphasis on tumor progression and hepatoma cells (Cavard et al. 2006). These genes may level of aggressiveness of the disease. Thus, it is of no

A B C 10 10 10 *8

8 8 8 *13 mRNA ratio mRNA ratio 6 3 6 6

* mRNA ration 4 4 4

2 2 2 2

TPH1/GAPDH TUSC2/GAPDH RUNX1/GAPDH 0 0 * 0 *3 Cluster 1 Cluster 3 Cluster 2 Cluster 3 Cluster 2 Cluster 3 Primary tumors Primary tumors LN metastases LN metastases LN metastases LN metastases D E 10 10 *8

8 8 mRNA ratio

6 mRNA ratio 6

4 4

2 2

CDH6/GAPDH TGFBR2/GAPDH 0 0 Cluster 2 Cluster 3 Cluster 2 Cluster 3 LN metastases LN metastases LN metastases LN metastases

Figure 4 Relative gene expression. RT-qPCR data on selected genes with comparison between A, clusters 1 (nZ11) and 3 (nZ7) of primary tumors, TUSC2. B–E, clusters 2 (nZ5) and 3 (12) of lymph node metastases; B, RUNXI;C,TPH1;D,TGFBR2, and E, CDH6. *, outlier; filled circle, mean.

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access www.endocrinology-journals.org 485 K Edfeldt et al.: Gene expression profiles in carcinoid tumors

Table 4 Known cancer genes relatively highly expressed in the three clusters and subgroups of tumors with lower expression in the latter group

Cluster 2 Lymph node Primary tumors Lymph node Cluster 1 Primary tumors in compared to metastases in compared to metastases compared to cluster 3 lymph node cluster 3 lymph node compared to primary tumors in compared to metastases in compared to metastasesa primary tumorsa cluster 3b cluster 1b cluster 3c cluster 2b

ACTA2 APOC1 BANP ABCC4 BIRC3 ELF3 ADCY4 BCL2A1 CDK5R2 EME1 CCR7 ERBB3 ADCY5 BTG2 EIF4G1 KIAA0101 CD3E F5 BANP CA9 ELMO1 MAL CXCR4 FOXA1 CDH17 CCNB2 FN1 MTSS1 IL8 MNX1 CLSPN CTLA4 GUCY2C MYB LCP1 PKIB CRYAB DUSP1 HOPX MYBL2 LYZ PTPRN DIRAS3 DUSP2 MUC20 NQO1 MMP9 RAB28 DLG2 E2F2 PAK3 PAH PTPRC RAB3C DMD EGR1 PRKCG SERPINA3 SERPINA1 EPHA7 FOS SYNPO2 SGK1 TACSTD1 FHL1 FOSB SSX1 TPH1 FLNA FOSL1 TUSC2 TRPM7 GUCY2C GADD45B ULK4 TTR HOPX GAS1 ITIH5 IER3 LPP KIAA0101 MAOB KRT17 MUC20 LEF1 MYH11 LOX PTPN13 MAL PTPRN2 MYBL2 TAGLN MYC TNC NGFR TNXB NPM1 TPSAB1 NR4A1 TPSD1 PAH TTBK1 PKMYT1 TYRP1 PTPRC WFDC1 SERPINA3 ZBTB16 SGK1 SOCS3 SORBS1 SYT13 TFAP2A ULK4 aFold change O2. bFold change O3. cFold change O8. relevance to compare the results from this microarray had received radical primary surgery (Table 1); hence, with the other, above-mentioned, microarray studies. the survival analysis is unlikely to depend on the To find differentially expressed genes, the data was surgical results. In the comparison among clusters of analyzed and compared between groups of tumors and primary tumors, TUSC2 located on chromosome 3p21.3 hierarchal cluster analysis was performed using PCA, was identified. TUSC2 is frequently deleted in multiple which surprisingly revealed three clusters. Primary cancers and has an anti-tumor effect (Ito et al. 2004). tumors in PCA of cluster 1 displayed a tendency toward RT-qPCR verified low expression in primary tumors in shorter survival, had fewer patients with the carcinoid cluster 1, which reflects the tendency toward high syndrome, and showed high proliferation rate mortality and a high proliferation rate in these tumors. (Ki67 O5%) implying different pathophysiological All the LN metastases in cluster 2 are from patients mechanisms or different stages of differentiation. Of the who are still alive and thus the gene expression could seven patients, four that succumbed from the disease eventually be a marker for prognosis. No other clinical

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access 486 www.endocrinology-journals.org Endocrine-Related Cancer (2011) 18 479–489 features were correlated to this cluster, which possibly resided in a different cluster. The tumor content is suggests that gene expression levels are not concordant similar in all samples and no LN tissue (except !5% in with clinical features in metastases. In the comparison one sample) was present in the metastases. This pattern between clusters of LN metastases, RUNX1, TGFBR2, demonstrates that the expression profile changes TPH1, and CDH6 were verified using RT-qPCR. greatly through tumor progression and metastases RUNX1, located on chromosome 21q22, is a well- development. known oncogene in leukemia and other cancers (Cohen Interestingly, only 11 genes were expressed with a 2009, Hoi et al. 2010, Kumano & Kurokawa 2010) and significant difference between LN metastases and liver can directly promote proliferation (Hoi et al. 2010). metastases, indicating that a greater genetic change GO analysis identified ‘proliferation’ as one of the occurs in the development from primary tumors to more significant cellular events in the LN metastases in metastases than between local and distant metastases. cluster 2, which also had a high expression of RUNX1. There are obvious drawbacks when examining a The TGF-b signaling pathway has been considered limited number of tumors; however, this is to date one both as a tumor suppressor pathway and as a promoter of the most comprehensive array studies on midgut of tumor progression and invasion. Somatic mutations carcinoid tumors. To provide solid evidence on the in TGFBR2, located on chromosome 3p22, have been described gene involvement in carcinoid tumors, these shown in several different tumor types. Notably, LN findings will have to be validated in a larger metastases in cluster 3 displayed low TGFBR2 independent tumor set. expression. Midgut carcinoid tumors produce serotonin This study reveals that midgut carcinoid tumors fall and the gene TPH1, located on chromosome 11p15, into three clusters based on their gene expression codes for the rate limiting enzyme in the biosynthesis profiles; thus, there might be different molecular of serotonin. Serotonin deficiency is associated with subtypes of midgut carcinoid tumors. Some primary slower tumor growth and increased expression with tumors as well as some LN metastases form clusters, breast tumor progression (Pai et al. 2009). Interest- with interesting clinical associations. There is evidence ingly, TPH1 had a higher expression in the LN for genetic changes from primary tumors to metastases, metastases in cluster 3 compared with those in cluster because most primary tumors displayed a different 2. The GO analysis revealed that in LN metastases in expression profile than their associated LN metastases. cluster 3, highly expressed genes were associated with Novel genes, ACTG2, GREM2, REG3A, TUSC2, the ‘serotonin receptor signaling pathway and RUNX1, TGFBR2, TPH1, and CDH6, not previously regulation of hormone levels’, which may be related known to be involved in midgut carcinoid tumorigen- to a high expression of TPH1. It is interesting to esis, displayed different expression levels in tumor correlate the expression levels of TPH1 with 5-HIAA clusters and may be of importance for tumor levels. Patient levels of 5-HIAA were unfortunately progression. missing in four patients and therefore no correlation between 5-HIAA levels and TPH1 expression was detected (Table 1). If this lack of correlation reflects, Supplementary data the missing data remains an open question. CAMs play This is linked to the online version of the paper at http://dx. essential roles in development and progression of doi.org/10.1530/ERC-10-0256. cancer and show clinical significance as prognostic biomarkers or as potential therapeutic targets in a variety of malignancies (Okegawa et al. 2004). In Declaration of interest lymph node metastases in cluster 3, 31 CAMs were The authors declare that there is no conflict of interest that highly expressed and some of these molecules may be could be perceived as prejudicing the impartiality of the potential biomarkers and predictors of prognosis, for research reported. example, CDH6, which is located on chromosome 5 and is involved in cell–cell signaling. Over-expression of CDH6 has been detected in small cell lung cancers Funding (Yokoi et al. 2002) and is implicated as a prognostic This study was supported by Lions Cancer Foundation, marker in ovarian carcinomas (Kobel et al. 2008). Swedish Cancer Society and Medical Research Council. We When comparing primary tumors with their respect- are grateful to Hanna Go¨ransson Kultima for statistical and ive associated LN metastases, different expression bioinformatic assistance. P B is a fellow of Swedish profiles were detected. Of 16 primary tumors, 13 Association for Medical Research and supported by A˚ ke separated from their respective LN metastases and Wibergs Foundation.

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access www.endocrinology-journals.org 487 K Edfeldt et al.: Gene expression profiles in carcinoid tumors

References Higgins ME, Claremont M, Major JE, Sander C & Lash AE 2007 Cancer genes: a gene selection resource for cancer Ahmed A, Turner G, King B, Jones L, Culliford D, genome projects. Nucleic Acids Research 35 D721–D726. McCance D, Ardill J, Johntson B, Poston G, Rees M (doi:10.1093/nar/gkl811) et al. 2009 Midgut neuroendocrine tumours with liver Hoi CS, Lee SE, Lu SY, McDermitt DJ, Osorio KM, metastases: results of the UKI NETS study. Endocrine- Piskun CM, Peters RM, Paus R & Tumbar T 2010 Related Cancer 16 885–894. (doi:10.1677/ERC-09-0042) Runx1 directly promotes proliferation of hair follicle Arnold R 2005 Endocrine tumours of the gastrointestinal stem cells and epithelial tumor formation in mouse skin. tract. Introduction: definition, historical aspects, classi- Molecular and Cellular Biology 30 2518–2536. (doi:10. fication, staging, prognosis and therapeutic options. Best 1128/MCB.01308-09) Practice & Research. Clinical Gastroenterology 19 Huang da W, Sherman BT & Lempicki RA 2009 Systematic 491–505. (doi:10.1016/j.bpg.2005.03.006) and integrative analysis of large gene lists using DAVID Arvidsson Y, Andersson E, Bergstrom A, Andersson MK, bioinformatics resources. Nature Protocols 4 44–57. Altiparmak G, Illerskog AC, Ahlman H, Lamazhapova D (doi:10.1038/nprot.2008.211) & Nilsson O 2008 Amyloid precursor-like protein 1 is Ito I, Ji L, Tanaka F, Saito Y, Gopalan B, Branch CD, Xu K, differentially upregulated in neuroendocrine tumours of Atkinson EN, Bekele BN, Stephens LC et al. 2004 the gastrointestinal tract. Endocrine-Related Cancer 15 Liposomal vector mediated delivery of the 3p FUS1 gene 569–581. (doi:10.1677/ERC-07-0145) demonstrates potent antitumor activity against human Benjamini Y 1995 Controlling the false discovery rate: a in vivo. Cancer Gene Therapy 11 733–739. practical and powerful approach to multiple testing. (doi:10.1038/sj.cgt.7700756) Journal of the Royal Statistical Society. Series B 57 Janson ET, Holmberg L, Stridsberg M, Eriksson B, 289–300. Theodorsson E, Wilander E & Oberg K 1997 Carcinoid Cavard C, Terris B, Grimber G, Christa L, Audard V, tumors: analysis of prognostic factors and survival in 301 Radenen-Bussiere B, Simon MT, Renard CA, Buendia MA & Perret C 2006 Overexpression of regenerating patients from a referral center. Annals of Oncology 8 islet-derived 1 alpha and 3 alpha genes in human primary 685–690. (doi:10.1023/A:1008215730767) liver tumors with beta-catenin mutations. Oncogene 25 Kidd M, Modlin IM, Mane SM, Camp RL, Eick G & Latich I 599–608. (doi:10.1038/sj.onc.1208860) 2006 The role of genetic markers – NAP1L1, MAGE-D2, Cohen MM Jr 2009 Perspectives on RUNX genes: an update. and MTA1 – in defining small-intestinal carcinoid American Journal of Medical Genetics. Part A 149A neoplasia. Annals of Surgical Oncology 13 253–262. 2629–2646. (doi:10.1002/ajmg.a.33021) (doi:10.1245/ASO.2006.12.011) Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, Kloppel G, Perren A & Heitz PU 2004 The gastroenter- Lane HC & Lempicki RA 2003 DAVID: database for opancreatic neuroendocrine cell system and its tumors: annotation, visualization, and integrated discovery. the WHO classification. Annals of the New York Genome Biology 4 P3. (doi:10.1186/gb-2003-4-5-p3) Academy of Sciences 1014 13–27. (doi:10.1196/annals. van Eeden S & Offerhaus GJ 2006 Historical, current and 1294.002) future perspectives on gastrointestinal and pancreatic Kobel M, Kalloger SE, Boyd N, McKinney S, Mehl E, endocrine tumors. Virchows Archiv: an International Palmer C, Leung S, Bowen NJ, Ionescu DN, Rajput A Journal of Pathology 448 1–6. (doi:10.1007/s00428-005- et al. 2008 Ovarian carcinoma subtypes are different 0082-4) diseases: implications for biomarker studies. PLoS Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Medicine 5 e232. (doi:10.1371/journal.pmed.0050232) Sargent D, Ford R, Dancey J, Arbuck S, Gwyther S, Kolby L, Bernhardt P, Sward C, Johanson V, Ahlman H, Mooney M et al. 2009 New response evaluation criteria in Forssell-Aronsson E, Stridsberg M, Wangberg B & solid tumours: revised RECIST guideline (version 1.1). Nilsson O 2004 Chromogranin A as a determinant of European Journal of Cancer 45 228–247. (doi:10.1016/j. midgut carcinoid tumour volume. Regulatory Peptides ejca.2008.10.026) 120 269–273. (doi:10.1016/j.regpep.2004.03.017) Eriksson J, Stalberg P, Nilsson A, Krause J, Lundberg C, Kosinski C, Li VS, Chan AS, Zhang J, Ho C, Tsui WY, Skogseid B, Granberg D, Eriksson B, Akerstrom G & Chan TL, Mifflin RC, Powell DW, Yuen ST et al. 2007 Hellman P 2008 Surgery and radiofrequency ablation for Gene expression patterns of human colon tops and basal treatment of liver metastases from midgut and foregut crypts and BMP antagonists as intestinal stem cell niche carcinoids and endocrine pancreatic tumors. World factors. PNAS 104 15418–15423. (doi:10.1073/pnas. Journal of Surgery 32 930–938. (doi:10.1007/s00268- 0707210104) 008-9510-3) Kulke MH, Freed E, Chiang DY, Philips J, Zahrieh D, Hellman P, Lundstrom T, Ohrvall U, Eriksson B, Skogseid B, Glickman JN & Shivdasani RA 2008 High-resolution Oberg K, Tiensuu Janson E & Akerstrom G 2002 Effect of analysis of genetic alterations in small bowel carcinoid surgery on the outcome of midgut carcinoid disease with tumors reveals areas of recurrent amplification and loss. lymph node and liver metastases. World Journal of Genes, & Cancer 47 591–603. (doi:10. Surgery 26 991–997. (doi:10.1007/s00268-002-6630-z) 1002/gcc.20561)

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access 488 www.endocrinology-journals.org Endocrine-Related Cancer (2011) 18 479–489

Kumano K & Kurokawa M 2010 The role of Runx1/AML1 Smyth GK 2004 Linear models and empirical bayes methods and Evi-1 in the regulation of hematopoietic stem cells. for assessing differential expression in microarray Journal of Cell Physiology 222 282–285. (doi:10.1002/ experiments. Statistical Applications in Genetics and jcp.21953) Molecular Biology 3 Article3. (doi:10.2202/1544-6115. Kytola S, Hoog A, Nord B, Cedermark B, Frisk T, Larsson C 1027) & Kjellman M 2001 Comparative genomic hybridization Soga J, Yakuwa Y & Osaka M 1999 Carcinoid syndrome: a identifies loss of 18q22-qter as an early and specific event statistical evaluation of 748 reported cases. Journal of in tumorigenesis of midgut carcinoids. American Experimental & Clinical Cancer Research 18 133–141. Journal of Pathology 158 1803–1808. (doi:10.1016/ Soreide JA, van Heerden JA, Thompson GB, Schleck C, S0002-9440(10)64136-3) Ilstrup DM & Churchward M 2000 Gastrointestinal Leja J, Essaghir A, Essand M, Wester K, Oberg K, carcinoid tumors: long-term prognosis for surgically Totterman TH, Lloyd R, Vasmatzis G, Demoulin JB & treated patients. World Journal of Surgery 24 1431–1436. Giandomenico V 2009 Novel markers for entero- (doi:10.1007/s002680010236) chromaffin cells and gastrointestinal neuroendocrine Tonnies H, Toliat MR, Ramel C, Pape UF, Neitzel H, carcinomas. Modern pathology 22 261–272. Berger W & Wiedenmann B 2001 Analysis of sporadic (doi:10.1038/modpathol.2008.174) neuroendocrine tumours of the enteropancreatic system Li GY, Kim M, Kim JH, Lee MO, Chung JH & Lee BH 2008 by comparative genomic hybridisation. Gut 48 536–541. Gene expression profiling in human lung fibroblast (doi:10.1136/gut.48.4.536) following cadmium exposure. Food and Chemical Wang GG, Yao JC, Worah S, White JA, Luna R, Wu TT, Toxicology 46 1131–1137. (doi:10.1016/j.fct.2007.11.007) Hamilton SR & Rashid A 2005 Comparison of genetic Lollgen RM, Hessman O, Szabo E, Westin G & Akerstrom G alterations in neuroendocrine tumors: frequent loss of 2001 Chromosome 18 deletions are common events in chromosome 18 in ileal carcinoid tumors. Modern classical midgut carcinoid tumors. International pathology 18 1079–1087. (doi:10.1038/modpathol. Journal of Cancer 92 812–815. (doi:10.1002/ijc.1276) 3800389) Ma S & Dai Y 2011 Principal component analysis based Watson MB, Lind MJ, Smith L, Drew PJ & Cawkwell L 2007 methods in bioinformatics studies. Briefings in Bioinformatics Expression microarray analysis reveals genes associated [in press]. (doi:10.1093/bib/bbq090) with in vitro resistance to cisplatin in a cell line model. Modlin IM, Lye KD & Kidd M 2003 A 5-decade analysis Acta Oncologica 46 651–658. (doi:10.1080/ of 13 715 carcinoid tumors. Cancer 97 934–959. (doi:10. 02841860601156157) 1002/cncr.11105) Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J & Okegawa T, Pong RC, Li Y & Hsieh JT 2004 The role of cell Speed TP 2002 Normalization for cDNA microarray adhesion molecule in cancer progression and its appli- data: a robust composite method addressing single and cation in cancer therapy. Acta Biochimica Polonica 51 multiple slide systematic variation. Nucleic Acids 445–457. Research 30 e15. (doi:10.1093/nar/30.4.e15) Pai VP, Marshall AM, Hernandez LL, Buckley AR & Yokoi S, Yasui K, Saito-Ohara F, Koshikawa K, Iizasa T, Horseman ND 2009 Altered serotonin physiology in Fujisawa T, Terasaki T, Horii A, Takahashi T, human breast cancers favors paradoxical growth and cell survival. Breast Cancer Research 11 R81. (doi:10.1186/ Hirohashi S et al. 2002 A novel target gene, SKP2, bcr2448) within the 5p13 amplicon that is frequently detected in Rindi G, Kloppel G, Couvelard A, Komminoth P, Korner M, small cell lung cancers. American Journal of Pathology Lopes JM, McNicol AM, Nilsson O, Perren A, Scarpa A 161 207–216. (doi:10.1016/S0002-9440(10)64172-7) et al. 2007 TNM staging of midgut and hindgut (neuro) Zhao J, de Krijger RR, Meier D, Speel EJ, Saremaslani P, endocrine tumors: a consensus proposal including a grading Muletta-Feurer S, Matter C, Roth J, Heitz PU & system. Virchows Archiv: an International Journal of Komminoth P 2000 Genomic alterations in well- Pathology 451 757–762. (doi:10.1007/s00428-007-0452-1) differentiated gastrointestinal and bronchial neuroendo- Schnirer II, Yao JC & Ajani JA 2003 Carcinoid – a crine tumors (carcinoids): marked differences indicating comprehensive review. Acta Oncologica 42 672–692. diversity in molecular pathogenesis. American Journal of (doi:10.1080/02841860310010547) Pathology 157 1431–1438. (doi:10.1016/S0002-9440(10) Segditsas S, Sieber O, Deheragoda M, East P, Rowan A, 64780-3) Jeffery R, Nye E, Clark S, Spencer-Dene B, Stamp G et al. 2008 Putative direct and indirect Wnt targets identified through consistent gene expression changes in APC- Received in final form 5 May 2011 mutant intestinal adenomas from humans and mice. Accepted 2 June 2011 Human Molecular Genetics 17 3864–3875. (doi:10.1093/ Made available online as an Accepted hmg/ddn286) Preprint 2 June 2011

Downloaded from Bioscientifica.com at 09/27/2021 05:42:19AM via free access www.endocrinology-journals.org 489