Multiomic sequencing of paired primary and metastatic small bowel carcinoids

Background: Small bowel carcinoids are insidious tumors that are often metastatic when diagnosed. Limited mutation landscape studies of carcinoids indicate that these tumors have a relatively low mutational burden. The development of targeted therapies will depend upon the identification of mutations that drive the pathogenesis and metastasis of carcinoid tumors. Methods: Whole exome and RNA sequencing of 5 matched sets of normal tissue, primary small intestine carcinoid tumors, and liver metastases were investigated. Germline and somatic variants included: single nucleotide variants (SNVs), insertions/deletions (indels), structural variants, and copy number alterations (CNAs). The functional impact of mutations was predicted using Ensembl Variant Effect Predictor. Results: Large-scale CNAs were observed including the loss of chromosome 18 in all 5 metastases and 3/5 primary tumors. Certain somatic SNVs were metastasis-specific; including mutations in ATRX, CDKN1B, MXRA5 (leading to the activation of a cryptic splice site and loss of mRNA), SMARCA2, and the loss of UBE4B. Additional mutations in ATRX, and splice site loss of PYGL, leading to intron retention observed in primary and metastatic tumors. Conclusions: We observed novel mutations in primary/metastatic carcinoid tumor pairs, and some have been observed in other types of neuroendocrine tumors. We confirmed a previously observed loss of chromosome 18 and CDKN1B. Transcriptome sequencing added relevant information that would not have been appreciated with DNA sequencing alone. The detection of several splicing mutations on the DNA level and their consequences at the RNA level suggests that RNA splicing aberrations may be an important mechanism underlying carcinoid tumors.


Introduction
Small intestine neuroendocrine tumors (SI-NETs) are the most common malignancy of the small bowel (Bilimoria et al. 2009).These tumors are characterized by their secretion of biogenic amines (such as serotonin and histamine), peptides (such as somatostatin and chromogranins), tachykinins, and/or prostaglandins (Pinchot et al. 2008;de Mestier et al. 2021).Secretion of these substances can lead to carcinoid syndrome, which is characterized by diarrhea, abdominal pain, bronchoconstriction, skin flushing, and valvular heart disease (Eltawil et al. 2022;Tran, Sherman, and Howe 2021;Moertel 1987).For this reason, SI-NETs have historically also been referred to as "carcinoid" tumors.
It has been postulated that SI-NETs are derived from enterochromaffin cells within intestinal crypts (Pinchot et al. 2008).Anatomically, small bowel carcinoids most frequently occur in the terminal ileum (Keck et al. 2018).Although these tumors tend to be indolent and insidious, they are often already metastatic at the time of diagnosis (Yao et al. 2008;Shah et al. 2019).The median age of diagnosis for small bowel carcinoid tumors is 61 years (Shah et al. 2019).Carcinoids are generally well-differentiated and slow-growing (Cunningham et al. 2011;'StatPearls' 2022).These tumors most often spread to regional lymph nodes, adjacent mesentery, and the liver (Pinchot et al. 2008).Improved imaging modalities, including the intravenous use of radiopharmaceutical gallium 68 Ga-or copper 64 Cu-dotatate paired with positron emission tomography (PET), may demonstrate metastases that are not appreciated with other, less sensitive imaging studies (Sanli et al. 2018).
Surgical removal of a localized, primary SI-NET (along with adjacent mesentery and lymph nodes) can be curative (Quinones et al. 2022).Treatment of metastatic carcinoid tumors, however, requires selection from a repertoire of therapeutic options, including surgical resection, administration of somatostatin analogues, peptide receptor radiotherapy with 177 Lu-dotatate, hepatic arterial embolization, hepatic radiofrequency ablation, external beam radiotherapy of selected isolated metastases, and administration of targeted therapies (Howe et al. 2017).
As the incidence of small bowel carcinoid tumors is increasing at an annual rate of 6.3%, the development and adoption of more effective targeted therapies is imperative (Ahmed 2020;Dasari et al. 2017;Maggard, O'Connell, and Ko 2004).Better elucidation of molecular drivers of carcinogenesis and metastasis underlying SI-NETs will help foster the development of new treatments.Knowledge of the full genomic landscape of small bowel carcinoid tumors is limited, as prior studies have deemed them to be relatively "mutationally silent" compared to other malignancies (Elias et al. 2021;Lim and Pommier 2021;Priestley et al. 2019).Loss of chromosome 18 as well as various loss of function mutations in cyclin-dependent kinase inhibitor 1B (CDKN1B, which encodes the cell cycle regulatory protein p27) have been reported in a minority of patients (i.e.9%) (Banck et al. 2013;Crona et al. 2015;Crona and Skogseid 2016;Cunningham et al. 2011;Francis et al. 2013;Kulke et al. 2008;Kytölä et al. 2001;Lim and Pommier 2021;Löllgen et al. 2001;Samsom et al. 2021;Zhang et al. 2020).
Reported aberrations include germline mutations in CDKN1B, which are known to cause Multiple Endocrine Neoplasia type IV (MEN4) (Seabrook et al. 2022).It has been suggested that loss of chromosome 18q may be an early event in the evolution of SI-NETs, whereas loss of CDKN1B and, therein, loss of tumor suppressor p27, occur later in malignant progression (Crona and Skogseid 2016;Di Domenico et al. 2017).
Recent studies have demonstrated that analysis of RNA may reveal the presence of causative molecular drivers of disease when DNA-level analysis has failed to do so (McCullough et al. 2020).Therefore, the integration of genomic data with transcriptomic data is the next logical step toward characterizing the molecular alterations underlying SI-NETs and, in doing so, identifying biomarkers for precision diagnosis, prognosis, and treatment.In this study, we investigated the genomic and transcriptomic landscapes of trio sets of germline, primary tumor, and metastatic tumor samples derived from 5 patients.We observed consistent loss of chromosome 18, as well as loss of function (LOF) mutations in CDKN1B.We identified metastasis-specific mutations in primary and metastatic carcinoid tumor pairs, several of which have been reported as driver mutations in other neuroendocrine tumor types.Transcriptome sequencing added relevant information that would not have been appreciated from DNA data alone.The detection of several splicing mutations on the DNA level, and of their consequences at the RNA level, suggest that RNA splicing aberrations may be an important mechanism underlying the development of SI-NETs.

Samples
Five patients (here referred to as 0006, 0007, 0008, 0009, 0018) undergoing surgical resection of their carcinoid tumors at Hoag Hospital (Newport Beach CA) or the University of Iowa Neuroendocrine Tumor Clinic consented to this study (written consent), which received IRB approval from the participating institutions (Hoag IRB number: 20180303;Iowa IRB number: 199911057).These patients had samples available from normal/constitutional tissue (i.e., gallbladder for patients 0006 and 0007, healthy bowel for patients 0008 and 0009, and peripheral blood for patient 0018), primary SI-NETs, and liver metastases from time of surgical resection.Deidentified tissue samples were sent for DNA/RNA extraction and sequencing as detailed below at the Translational Genomics Research Institute (TGen) (Phoenix AZ).

Sequencing
Tumor or constitutional DNA was extracted using the Qiagen AllPrep Kit or GeneRead FFPE DNA Kit (Cat. No. 80234, 180134), and tumor RNA was extracted using the Qiagen AllPrep Kit or RNeasy Mini Kit (Cat. No. 74104).Using 200 ng of input DNA, whole exomes were constructed for each sample using the Kapa Hyper Prep Kit (Cat.No. 07962363001) using Agilent SureSelect Human All Exon V7 baits.RNA libraries were constructed using 500ng of input RNA per sample and using the Illumina TruSeq Stranded Total RNA Kit with Ribo-zero (Cat. No. RS-122-2201).Paired-end sequencing of libraries was performed on the Illumina NovaSeq 6000 using S1 and SP flowcells for 100bp reads.Approximate sequencing targets were 200Â for tumor exomes, 100Â for constitutional exomes, and 150 million total reads for each RNA library.

Bioinformatics
Paired DNA and RNA data (for germline, primary, and metastatic tissue samples) were input into two pipelines: TGen's Phoenix pipeline, and the Keck School of Medicine of USC's Genomics Platform (KGP) "Echo" pipeline.
RNAseq FASTQs (tumors only) were aligned to human genome build GRCh38 using STAR (v2.6.1d).Duplicates were marked using GATK's MarkDuplicates.Gene fusion predictions were made using STAR-Fusion (v2.6.1d).Quality control metrics for the resulting sorted, indexed BAMs were retrieved using Samtools Stats, Picard RNA Metrics, and Picard MultiMetrics.Variants and mutations were called using a dbSNP reference VCF (v146 hg38) and GATK's RNA HaplotypeCaller.Salmon, featureCounts, HTSeq Counts, and Cufflinks were used for gene quantification.The final variant candidate set was annotated with VEP.We utilized MultiQC to compile analysis logs and create a comprehensive quality control report.Differential expression and pathway analyses were run on primary/metastatic tumor pairs (not on germline, as constitutional transcriptomic data was not available) using iDEP (which employs the parametric and semiparametric packages limma and DESeq2) with GAGE analysis, KEGG gene sets, and Pathview for visualization (Ge, Son, and Yao 2018;Luo and Brouwer 2013;Kanehisa et al. 2021).

Limitations
The authors acknowledge the small sample size of this study, which is not a meta-analysis, and suggest that readers interpret results (particularly those of differential expression and pathway analyses) within the context of this limitation.Sample size appears to be a common limiting factor for genomic studies on SI-NETs (Wu et al. 2022).Many of these previous studies did not utilize RNA data, however, so one of the advantages of this present study is that it is among the first to leverage the transcriptome for identification of molecular drivers.

Bioinformatic summary statistics
Tumor samples had, on average, a 45-fold gene enrichment, a coverage ≥30Â for 95% of target bases, 87% aligned reads, and 96% RNAseq correct strand mapping.Constitutional samples had, on average, a 43-fold gene enrichment, a coverage of ≥30Â for 91% of target bases, and 99.8% aligned reads.

Patient-specific molecular findings
Patient 0006 is a 66-year-old male with grade 1 disease, treated with surgical resection then Octreotide (Table 1).He had large-scale copy number loss of chromosome 18 (Table 2).He also had a PYGL mutation (rs74464749) in both his primary and metastatic tumors (Table 3).He had a metastasis-specific mutation in MXRA5 (chrX:3,317,443:G:C).
Patient 0007 was a 62-year-old female with grade 2 disease (for both primary and metastatic tumors), treated with surgical resection then 177 Lu-dotatate (Table 1).She succumbed to her disease.She had large-scale copy number changes in both her primary and metastatic tumors; these included: gain of chromosomes 5, 7, 10, 14q, 15q, and 20, and loss of chromosomes 18 and 11q (Table 2).Loss of chromosome 11q has previously been noted as a potentially metastasisrelated event in primary/metastatic SI-NET pairs (Alexander and Ziv 2023; Elias et al. 2021).Amplification of chromosomes 7, 14, and 20 has also been reported in ileal neuroendocrine tumors (Mäkinen et al. 2022).This patient did not have any apparent pathogenic point mutations.
Patient 0008 was a 67-year-old male with grade 2 disease, treated with surgical resection, everolimus, then temozolomide (Table 1).He succumbed to his disease.He had one copy number alteration observed in both his primary and metastatic tumors, gain of chromosome 5 (Table 2).He also had metastasis-specific copy number changes including gain of chromosomes 7 and 10, and loss of chromosomes 9 and 18.As mentioned for patient 0007, amplification of chromosome 7 has been observed in ileal carcinoids (Mäkinen et al. 2022).Loss of the p-arm of chromosome 9 has been previously   reported in midgut neuroendocrine tumors (Alexander and Ziv 2023).He had metastasis-specific CDKN1B (rs797044482) and a UBE4B (chr1:10,105,515:G:A) mutations (Table 3).
Patient 0009 is a 52-year-old female with grade 1 disease (for both primary and metastatic tumors, grade 2 lymph nodes), treated with surgical resection (Table 1).The only copy number alteration observed in her tumors was loss of chromosome 18 in the primary/metastatic pair (Table 2).An ATRX point mutation (chrX:77,683,393:G:T) was identified in both her primary and metastatic tumors resulting in loss of heterozygosity (LOH) and allele-specific expression (ASE) in the mRNA (monoallelic expression of the wild-type allele in mRNA) (Table 3).She had a metastasis-specific mutation in SMARCA2 (rs752254994).
Patient 0018 is a 71-year-old female with grade 2 disease, treated with surgical resection then monthly lanreotide (Table 1).She had copy number loss of chromosome 18 in both her primary and metastatic tumors (Table 2).She had metastasis-specific, large-scale copy number loss of chromosome 16q, which has previously been reported in the literature as a metastasis-specific event (Alexander and Ziv 2023).She also had a metastasis-specific ATRX mutation (chrX:77,684,450:T:A) with monoallelic expression of the wild-type allele in mRNA (Table 3).

Differential expression & pathway analyses
The results of our analysis performed with iDEP's GAGE tool and KEGG gene sets suggested that the top 20 pathways with differentially expressed genes between metastatic vs. primary SI-NETs included: "pathways in cancer," "chemical carcinogenesis," and "viral carcinogenesis" (Figure 5; Table 4).The pathways within the "pathways in cancer" framework that appear to have the most significant changes in expression between metastases vs. primary tumors include: cytokine-cytokine receptor interaction, p53 signaling, extracellular membrane receptor and focal adhesion interactions, Wnt signaling, PI3K-Akt signaling, MAPK signaling, calcium signaling, TGF-b signaling, HIF-1 signaling, Notch and Hedgehog signaling, estrogen and androgen signaling, cell cycle, and block of differentiation (Figure 5).A recent study also demonstrated that Wnt signaling and focal adhesion pathways were associated with SI-NETs (Chen et al. 2023); however, this same study highlighted cytokinesis, iron ion homeostasis, platelet degranulation, and several metabolic processes to be differentially upregulated in SI-NET metastases compared to primary tumors (which we did not observe).

Mutational landscape of SI-NETs
Previous genomic analyses have failed to identify a consistent, putative driver mutation in small bowel neuroendocrine tumors.We endeavored to dive deeper into the molecular landscape of SI-NETs by supplementing exome analysis with transcriptomic analysis in paired primary small bowel tumors and liver metastases from the same patients.We were particularly interested in discovering drivers of metastatic potential, as metastatic SI-NETs pose the greatest morbidity and mortality risks.We intended to probe RNA for potential culprits that were, perhaps, not immediately evident from DNA analysis alone.RNAseq data enabled us to confirm the consequences of alterations found on the DNA level while also observing phenomena (like intron retention and ASE) that would not otherwise be apparent.
Tumor cells gain oncogenic and lose tumor-suppressive functions via various mechanisms, including point mutations, CNAs, and structural variants.In this study, we observed various CNAs, the most frequent being loss of chromosome 18.Loss of 18q has been previously reported as a potential mechanism of early oncogenesis in SI-NETs (Crona and Skogseid 2016) Loss of CDKN1B function is another previously reported characteristic of small bowel carcinoid tumors (though it has only ever been reported in a minority, i.e. 9% of tumors in larger studies) (Crona and Skogseid 2016;Crona et al. 2015;Maxwell et al. 2015).An insertion in CDKN1B (rs797044482, reported as "likely pathogenic" for neuroendocrine neoplasms in ClinVar) was identified in patient 0008's metastasis (Figure 1).Indel presence was confirmed in mRNA.Interestingly, this patient had a "second hit" to CDKN1B: uniparental disomy (and resulting LOH).B-allele frequency (BAF) is a measure of allelic balance, whereby a heterozygous SNP would have a BAF of 0.5 and a homozygous SNPor LOH eventwould result in a BAF that deviates from 0.5 (to 0 or 1).Patient 0008's uniparental disomy event was identified via CNA, as B-allele frequency deviated from 0.5 despite the fact that there was no copy number change at that location.This example exemplifies the ability of cancer cells to "mix and match" mechanisms to gain and lose gene products.

ATRX mutations as potential drivers
The tumor suppressor gene ATRX encodes a chromatin-remodeler that is a member of the SWI-SNF family of proteins; it is involved in transcriptional regulation, DNA recombination, nucleosome remodeling, and DNA repair (Bradley et al. 2019;Valenzuela et al. 2021).Somatic mutations in this gene have frequently been reported in gliomas, gastro/pancreatic neuroendocrine tumors, pheochromocytomas, and paragangliomas (Crona and Skogseid 2016;Jiao et al. 2011).ATRX has also been found to be mutated in gliomas and acute lymphoblastic leukemia (ALL) (Bradley et al. 2019).ATRX plays an essential role in brain development and is ubiquitously expressed at high levels in brain tissue (Valenzuela et al. 2021).
Though conflicting evidence exists, it has been reported that ATRX escapes X-inactivation, perhaps in a developmental stage and/or tissue-specific manner, with ATRX showing biallelic expression in XX-females and monoallelic expression in XY-males (Valenzuela et al. 2021).ATRX deficiency results in impaired nonhomologous end-joining and genomic instability (Bradley et al. 2019).Notably, ATRX mutations are frequently observed in female gastric cancer patients with high microsatellite instability (MSI), tumor mutational burden (TMB), and programmed death-ligand 1 (PD-L1) expression; these characteristics are purported to be predictive biomarkers for immunotherapy response (Ge et al. 2021).
Two of our patients (0009, 0018) had ATRX mutations.Patient 0009, an XX-female, had LOF mutations in ATRX in both her primary and metastatic tumors.Patient 0018, another XX-female, harbored a metastasis-specific ATRX mutation.Curiously, both mutations were predicted to be "tolerated" by VEP and showed ASE/monoallelic expression of the wild-type allele in RNA.

Metastasis-specific findings
As mentioned above, one metastasis-specific finding was patient 0008's CDKN1B mutation (rs797044482) (Figure 1).A metastasis-specific missense mutation in the gene Matrix Remodeling Associated 5 (MXRA5, chrX:3,317,443:G:C) was seen in patient 0006 (Figure 2).MXRA5 is a purported tumor suppressor gene (Tegally et al. 2020;Xiong et al. 2012;Cheng et al. 2017).It also plays a role in normal matrix remodeling and anti-inflammatory responses, the disruption  This is appreciable as a lack of mRNA reads corresponding to the downstream portion of the depicted exon (i.e., the thick blue line in the gene track) in the metastatic tumor (brown trace).This phenomenon is not observed in the primary tumor.Sashimi plots were generated in IGV.
of which is essential to metastatic progression (Poveda et al. 2017).This deleterious MXRA5 mutation resulted in the activation of a cryptic splice site and loss of mRNA transcripts.As patient 0006 is male, there was no dosage compensation for MXRA5.Patient 0006's primary and metastatic tumors had a splicing mutation in liver glycogen phosphorylase (PYGL, rs74464749), resulting in intron retention and expression of alternative splice variants in mRNA.
A metastasis-specific missense mutation in SWI/SNF Related, Matrix Associated, Actin Dependent Regulator of Chromatin, Subfamily A, Member 2 (SMARCA2, rs752254994) was called for patient 0009 (confirmed in mRNA, Figure 3).SMARCA2 is part of a chromatin-remodeling complex, and mutations in this gene have been reported in neuroendocrine tumors of the lung and thymus (Fernandez-Cuesta et al. 2014).A metastasis-specific mutation in Ubiquitination Factor E4B (UBE4B, chr1:10105515:G:A) was called for patient 0008, which resulted in intron retention in the mRNA (Figure 4); this was paired with a "second hit" of large-scale copy loss of chromosome 1.Mutations in UBE4B have previously been associated with neuroblastoma, another type of neuroendocrine tumor (Byron et al. 2016).

RNA splicing results
As mentioned in the section above, patient 0008 had several metastasis-specific splicing mutations illuminated by RNA analysis.This included a mutation in UBE4B resulting in intron retention (Figure 4), and a mutation in MXRA5 leading to activation of a cryptic splice site and loss of mRNA.
Splicing mutations were also observed in the tumors of patient 0006.He had a splice-acceptor SNV in PYGL, which encodes liver glycogen phosphorylase, in both the primary and metastatic tumor in the liver.The splicing mutation resulted in intron retention.PYGL was listed as a mutated gene in SI-NETs in a previous study (Francis et al. 2013).Limitations, conclusions, and future directions Limitations of this study include its relatively small sample size, albeit for a rare tumor type, and the fact that we utilized whole exome data (thereby ignoring introns).It is possible that the molecular drivers of small bowel carcinoid tumorigenesis and progression/metastasis are in noncoding regions.Noncoding RNAsincluding microRNAs (miR-NAs), long noncoding RNAs (lncRNAs), small interfering RNAs (siRNAs), small nuclear RNAs (snRNAs), small nucleolar RNAs (snoRNAs), and PIWI-interacting RNAs (piRNAs)have been shown to play a role in the pathogenesis of various tumor types (Byron et al. 2016;Zeuschner, Linxweiler, and Junker 2020).A recent study also identified differentially expressed lncRNAs in primary/metastatic SI-NET pairs (Chen et al. 2023).Additionally, given the fact that two of the mutations identified in this study were in genes involved in chromatin remodeling (i.e., ATRX and SMARCA2), it will be important to study the epigenomic landscape of SI-NETs in the future (Kidd et al. 2021;Walter et al. 2018).
Proteomic analysis also appears to be informative in understanding the underlying mechanisms of carcinoid metastasis (Blažević et al. 2023).Lastlyas disparities exist in terms of carcinoid incidence, treatment, and survivalevaluation of the associations between race, ethnicity, and genetic ancestry/similarity with carcinoid genetic aberrations may be informative in a larger sample dataset (Takayanagi et al. 2022;Kessel et al. 2021).
In conclusion, we identified several candidate mutations potentially involved in the pathogenesis and metastatic cascade of SI-NET ("carcinoid") tumors.Mutations that were identified as metastasis-specific may provide insight into intermediate steps between initial tumorigenesis and metastasis (i.e., drivers of metastatic potential).We confirmed the presence of previously reported molecular aberrations (i.e., loss of chromosome 18 and LOF mutations in CDKN1B).It remains unclear whether there is a key tumor suppressor or set of tumor suppressors on chromosome 18, the loss of which is important to the etiology of small bowel carcinoids.Gene knock-outs could be performed to determine which regions on chromosome 18 are necessary and/or sufficient to drive the formation of SI-NETs.Additionally, it is worth noting that one potential explanation that could account for the fact that previous studies identified mutations in CDKN1B in a minority of SI-NETs is the possibility that distinct molecular subtypes of small bowel carcinoids exist.The significant utility of considering transcriptomic data in addition to genomic data was exemplified by our detection of phenomena such as intron retention, splicing variants, and ASE.The addition of RNAseq data also enabled us to investigate the consequences of mutations called on the DNA level in mRNA, and perform differential expression and pathway analyses to identify several pathways potentially involved in conferring metastatic potential to SI-NETs.
would recommend removing it to increase readability and focus on relevant information.

Discussion
It would be beneficial to discuss the similarities and discrepancies between the finding in this study and published literature in greater details.For example, the statement: Additionally, it is worth noting that one potential explanation that could account for the fact that previous studies identified mutations in CDKN1B in a minority of carcinoid tumors is the possibility that distinct molecular subtypes of small bowel carcinoids exist.What is already know about distinct molecular subtypes?Is there another explanation such as a difference in patient selection or pretreatment?These discrepancy deserve more discussion in order to embed this study properly in the known literature.
○ It would also be insightful to discuss the relation between epigenetic modification and the found RNA splicing results, as small intestinal neuroendocrine tumors are known to have epigenetic dysregulation.

○
Finally, Postel et.Al. should consider using the term small intestinal neuroendocrine tumor as this is more in line with current guidelines.
Overall, the study is interesting as it combines whole exome/whole transcriptome NGS and reveals interesting novel data.However, I have significant reservations, as outlines above.

Is the work clearly and accurately presented and does it cite the current literature? Partly
Is the study design appropriate and is the work technically sound?Yes

Are sufficient details of methods and analysis provided to allow replication by others? Partly
If applicable, is the statistical analysis and its interpretation appropriate?I cannot comment.A qualified statistician is required.
Are all the source data underlying the results available to ensure full reproducibility?Yes

Are the conclusions drawn adequately supported by the results? Yes
Competing Interests: No competing interests were disclosed.
I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

Mackenzie Postel
REVIEWER: "What was the timing between the surgical resection and other treatments as these could affect mutational load.In the result section, patient-specific molecular finding paragraph it is not clearly enough stated." AUTHOR RESPONSE: For each patient, surgical resection was the first intervention performed (i.e., sequence was: surgery, medication).Molecular findings, therefore, represent tissues before use of chemotherapeutics/other medications.In the Results section, for each patient, we have used the phrasing: "treated with surgical resection then ___" to give a sense of the chronology of treatment.
-REVIEWER: "Grading.Was the tumor grade determined in the included tissue in the study or on initial biopsy?Was there a discrepancy between the primary tumor and liver metastases?" AUTHOR RESPONSE: We are happy to clarify.WHO grading by mitotic count and/or Ki-67 labeling index was performed at the time of surgical resection of the tissues used in this study.Primary/met grading was consistent for each pair (no discrepancy between tumors in any given patient).All this to say, in our manuscript, "grade" can be taken to refer to the grade of a given patient's primary tumor as well as their met.We have endeavoured to make this more clear in the text.
-REVIEWER: "What was the origin of the normal tissue?Intestine or healthy liver tissue?" AUTHOR RESPONSE: Thank you for your question.The "normal" or constitutional samples were used mainly to differentiate germline from somatic alterations in DNA.We did not have RNA data available for constitutional samples, therefore we did not perform differential expression analysis, eQTLs, or any other transcriptomic comparison between normal and tumor (which would, as I believe you are suggesting, require matched healthy liver or intestine).We simply wanted to establish the patients' germline variants to distinguish them from somatic mutations.Constitutional samples were derived from: gallbladder for patients 0006 and 0007, healthy bowel for patients 0008 and 0009, and peripheral blood for patient 0018.We have added a statement to the Methods section clarifying the sources of constitutional samples.
-REVIEWER: "Was the collection of primary tumor and liver metastases synchronous or metachronous?" AUTHOR RESPONSE: Thank you for your comment.If you mean metachronous as in two sites with independent primary malignancies, then no -the liver tumors represented metastases from primary small bowel tumors.If you are asking whether primary/metastatic tumors were collected at the same time vs. in sequential surgeries, they were collected at the same time (patients were metastatic at time of presentation/there was no lag time in terms of diagnosis of primary and met as they were diagnosed simultaneously).In this study, one primary tumor and one metastatic tumor were analyzed for each patient (no multiple, synchronous primary tumors for a given patient).I hope this answers your question.
-REVIEWER: "No mentioning of analysis on normal tissue.If this was deemed not relevant, it should be clarified why the was no differential analysis between normal and primary tumor tissue as this is theoretically also interesting.If there was no significant differential expression, it is also an important observation." AUTHOR RESPONSE: Thank you for your comment.We did not have RNAseq data available for normal tissue.Should we have the opportunity to gather constitutional transcriptomic data for these patients in the future, it would certainly allow us to investigate the interesting comparison you suggest.We have clarified in the Methods section that differential expression analysis focused on primary/met tumor pairs (and excluded normal tissue) because we only had RNA for tumors.
-REVIEWER: "Table 5 does not add relevant information and the findings are not further discussed, so I would recommend removing it to increase readability and focus on relevant information." AUTHOR RESPONSE: Table 5 has been removed per the reviewer's helpful suggestion.
-REVIEWER: "It would be beneficial to discuss the similarities and discrepancies between the finding in this study and published literature in greater details." AUTHOR RESPONSE: Thank you for this consideration.This manuscript was submitted several months ago but had not been assigned reviewers until relatively recently.In this time gap, more multi-omicliterature has been published re: SI-NETs.We have incorporated several of these new studies, including a recent paper by the reviewer.
-REVIEWER: "It would also be insightful to discuss the relation between epigenetic modification and the found RNA splicing results, as small intestinal neuroendocrine tumors are known to have epigenetic dysregulation."

AUTHOR RESPONSE:
This would be a fascinating future direction.However, currently, we do not have epigenetic data available for these patients.
-REVIEWER: "Finally, Postel et.Al. should consider using the term small intestinal neuroendocrine tumor as this is more in line with current guidelines." While mentioned in the discussion (small sample size, non-coding RNA's are assigned slightly more importance than may be warranted) this reviewer suggests the methods section should note that this study is not a meta-analysis, the differential expression is highly limited, non-parametric analyses have limited utility, and the pathway analysis and other interpretations should be presented in the context of the study limitations, rather than a sentence in the discussion.Minor additions to the data addressing these limitations would provide readers a way to understand that extending the results to a biological cause or even to a pathway is pure speculation.I would much prefer the authors use phrases like "evidence suggests genes X, Y, Z that were found through our analysis are implicated in the functions shown in our genetic pathway figures."The way it's written, there is more confidence in the interpretation of the biological role of the detected gene expression than may be warranted.It would be beneficial to highlight similar findings in published journals concerning similar tumor types, similar benign vs. metastatic cases, and perhaps making a case for a set of commonly activated genes or types of genes that appear in published/public studies of metastasis/benign tumors.
Overall the study brings up good points, the NGS analysis is fine, the WTS and WES are an important combination, and releasing these data into the public domain will spur further studies from these important data.The biological analysis is extensive but may be a little overstated.The conclusions are warranted.The text and writing style are acceptable, the references are wellchosen but could be bolstered by finding any studies between met/benign tumors in similar tissues that might have found similar over/underexpressed genes.
Recommend publish with minor revisions.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound?Yes

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate?Yes Are all the source data underlying the results available to ensure full reproducibility?Yes

Are the conclusions drawn adequately supported by the results? Yes
Competing Interests: No competing interests were disclosed.

I confirm that I have read this submission and believe that I have an appropriate level of
Point mutations in primary and metastatic tumor pairs -highlighting those that are metastasis-specific.

Figure 1 .
Figure1.CDKN1B sashimi plots for patient 0008's primary/metastatic tumor pair.A loss-of-function frameshift mutation resulting in intron inclusion is observed in the metastatic tumor alone.This is perceptible as the orange signal that reads through intronic regions (i.e., the thin blue lines in the gene track).Sashimi plots were generated using the Broad Institute's Integrative Genomics Viewer (IGV).

Figure 2 .
Figure2.Activation of a metastasis-specific MXRA5 cryptic splice site in patient 0006 leads to exon truncation.This is appreciable as a lack of mRNA reads corresponding to the downstream portion of the depicted exon (i.e., the thick blue line in the gene track) in the metastatic tumor (brown trace).This phenomenon is not observed in the primary tumor.Sashimi plots were generated in IGV.

Figure 3 .
Figure 3.A metastasis-specific SMARCA2 missense mutation (rs752254994) is observed in the mRNA of patient 0008.Sashimi plots were generated in IGV.

Figure 4 .
Figure4.A metastasis-specific UBE4B splicing mutation results in intron retention in the mRNA of patient 0008.Intron retention is apparent as the presence of mRNA (i.e., the orange signal/bars) corresponding to intronic material (i.e., the thin blue lines in the gene track).Sashimi plots were generated in IGV.

Figure 5 .
Figure 5. "Pathways in cancer" is one of the top 20 differentially expressed pathways between metastatic and primary small intestine neuroendocrine tumors (SI-NETs) as determined by GAGE Analysis with KEGG gene sets.Genes colored bright red are most upregulated, and those colored bright green are most downregulated.The most significant sub-pathways within this framework appear to be: cytokine-cytokine receptor interaction, p53 signaling, extracellular membrane receptor and focal adhesion interactions, Wnt signaling, PI3K-Akt signaling, MAPK signaling, calcium signaling, cell cycle, TGF-b signaling, HIF-1 signaling, Notch and Hedgehog signaling, estrogen and androgen signaling, and block of differentiation.These pathways may provide clues regarding which molecular events confer SI-NET metastatic potential.

Table 1 .
Patient demographics and sample information.

Table 2 .
Copy number alterations in primary and metastatic tumor pairs.