Purpose:

Little is known about the genetic alterations characteristic of small bowel adenocarcinoma (SBA). Our purpose was to identify targetable alterations and develop experimental models of this disease.

Experimental Design: Whole-exome sequencing (WES) was completed on 17 SBA patient samples and targeted-exome sequencing (TES) on 27 samples to confirm relevant driver mutations. Two SBA models with ERBB2 kinase activating mutations were tested for sensitivity to anti-ERBB2 agents in vivo and in vitro. Biochemical changes were measured by reverse-phase protein arrays.

Results:

WES identified somatic mutations in 4 canonical pathways (WNT, ERBB2, STAT3, and chromatin remodeling), which were validated in the TES cohort. Although APC mutations were present in only 23% of samples, additional WNT-related alterations were seen in 12%. ERBB2 mutations and amplifications were present in 23% of samples. Patients with alterations in the ERBB2 signaling cascade (64%) demonstrated worse clinical outcomes (median survival 70.3 months vs. 109 months; log-rank HR = 2.4, P = 0.03). Two ERBB2-mutated (V842I and Y803H) cell lines were generated from SBA patient samples. Both demonstrated high sensitivity to ERBB2 inhibitor dacomitinib (IC50 < 2.5 nmol/L). In xenografts derived from these samples, treatment with dacomitinib reduced tumor growth by 39% and 59%, respectively, whereas it had no effect in an SBA wild-type ERBB2 model.

Conclusions:

The in vitro and in vivo models of SBA developed here provide a valuable resource for understanding targetable mutations in this disease. Our findings support clinical efforts to target activating ERBB2 mutations in patients with SBA that harbor these alterations.

Translational Relevance

Small bowel adenocarcinoma (SBA) is a rare cancer, with limited data to guide therapeutic decisions. Both the absence of representative cell lines and patient-derived xenograft (PDX) models and the limited knowledge of the molecular alterations within this cancer have hindered design of tumor-specific therapeutic strategies. Currently, the approach to SBA is predicated on extrapolation from colorectal cancer. Here, through exome and targeted next-generation sequencing of over 40 SBA patient samples, and developing relevant cell lines and PDX models, we identified targetable genomic alterations in SBA. Alteration in ERBB2 was among the most frequent therapeutic targets identified. ERBB2 targeting resulted in significant growth inhibition in cell lines and tumor xenografts. Moreover, we found that activation of the ERBB2 signaling cascade predicts a poor outcome in SBA. These findings provide strong support for clinical efforts to identify and target ERBB2 genomic alterations in patients with SBA.

The incidence of small bowel adenocarcinoma (SBA) is 10- to 15-fold lower than that of colorectal adenocarcinoma despite their anatomic proximity. The reason for this discrepancy is not known currently, and our ability to explain it is limited by our lack of both molecular characterization and model systems available for studying this relatively rare cancer (1). Although both SBA and colorectal cancer are observed in the intestinal cancer syndromes familial adenomatous polyposis and hereditary nonpolyposis colon cancer, the majority of SBAs are sporadic, and little is known about the risk factors for these cases (2).

So, little is known about SBA that, at present, its clinical management is primarily extrapolated from that for colorectal cancer. The number of clinical trials exploring novel therapies for SBA is limited. In fact, a recent clinical trial of anti-EGFR therapy in RAS wild-type SBA demonstrated no responses, in sharp contrast with findings from colorectal cancer, reinforcing the critical need for improved understanding of the molecular mechanisms underlying SBA (3).

The one known major genetic difference between SBA and colorectal cancer is the frequency of APC gene alterations, which are far less prominent in SBA than in colorectal cancer. Although some reports of DNA sequencing for SBA have been published in the past 5 years (4–6), the patient cohorts in these studies were heterogeneous and small, and most of the studies targeted a limited number of genes. A 2017 study that compared 317 SBA cases with 6,353 colorectal cancer cases found that SBA tumors are characterized by a higher mutation rate, greater numbers of atypical BRAF mutations and ERBB2 point mutations, and lower rates of APC and SMAD4 mutations than colorectal cancer (7).

In the current study, our purpose was to identify targetable alterations and develop experimental models of SBA. We examined exomic mutations in samples from a cohort of patients with SBA at various clinical stages, both sporadic and familial forms; we also expanded our findings to a validation cohort in which we used targeted deep sequencing of archived paraffin SBA samples. Our results identified mutations and amplifications of growth factor ERBB2 in 23% of the tumors tested. We further demonstrate in vitro and in vivo that mutated ERBB2 is a relevant target in SBA.

Patient characteristics

Our cohorts comprised patients with SBA treated at The University of Texas MD Anderson Cancer Center (Houston, TX) between 1998 and 2016. The whole-exome sequencing (WES) cohort consisted of 16 matched samples of tumor and normal germline DNA (peripheral blood in 4 and adjacent histologically normal tissue in 12) plus one unpaired tumor sample selected because of availability of frozen tissue and ≥50% cellularity noted on gastrointestinal pathology review (H. Wang). For the targeted DNA sequencing (TES) cohort (n = 27), cases with paraffin-embedded tumor tissue with a cellularity threshold ≥30% were selected. Clinicopathologic characteristics and survival were collected from retrospective chart review.

This study was approved by the Institutional Review Board and Institutional Animal Care and Use Committee of MD Anderson and conducted in accordance with U.S. Common Rule. Prospective informed consent was obtained for SBA model generation.

Tumor xenograft and cell line models

To establish in vitro and in vivo patient-derived xenograft (PDX) models from freshly excised SBA specimens, we implanted one 4-mm3 piece of tumor subcutaneously into one NOD/SCID gamma mouse. After each transplanted tumor was established, pieces of that tumor were in turn transplanted into nude mice and allowed to grow for 3 weeks in preparation for anti-ERBB2 drug response testing. In parallel, pieces of each transplanted tumor were also cultured in vitro for development of cell lines; the cells were mechanically dissociated, subjected to centrifugation in a PBS solution containing antibiotics, and incubated at 37°C in Advanced MEM culture medium supplemented with collagenase IV (1 mg/mL) for 30 minutes. After the collagenase was removed, cells were fed with 1% Advanced MEM containing 1% FCS and plated into collagen-coated 6-well plates for cell growth.

For in vivo testing of sensitivity to ERBB2 inhibition, mice implanted with tumor as described in the previous paragraph were randomly assigned to receive chow containing dacomitinib (SelleckChem) or control (5 mice/group). The concentration of dacomitinib in the chow was calculated according to the following formula: DD = (SD × BW) FI, where DD is diet dose, SD is single daily dose, FI is daily food intake (3.5 g), and BW estimated mean body weight (24 g). Chow containing dacomitinib was prepared by Research Diets, Inc. to achieve an approximate daily drug dose of 10 mg/kg. The largest diameter (D) and the smallest diameter (d) of each mouse's tumor were measured every week, and the volume of each tumor was calculated according to the following formula: Vol = d2 × D/2.

For testing sensitivity of the newly established SBA cell lines to anti-ERBB2 agents, the cells were seeded (5.0 × 103 cells/well) in flat-bottomed 96-well microplates coated with rat collagen I (Corning). After 48 hours, the cells were treated with various concentrations of lapatinib (SelleckChem) or dacomitinib. Cell proliferation was estimated by using the MTS colorimetric assay (Promega Corp.). The significance of differences between treatment groups was defined using a standard t test on triplicate values.

Sequencing analyses

WES was performed by the sequencing core facility at MD Anderson under contract by Illumina to meet the typical depth of at least 50×, with 94% of the known genome being sequenced to at least 8× coverage while achieving a Phred base calling quality score of at least 30 over at least 80% of mapping bases. Quality control metrics were computed on a per lane basis using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Exome capture was performed using the manufacturer's protocol (Illumina Multiplexing_SamplePrep_Guide_1005361_D) as described on the MD Anderson–Illumina core website. Sequence reads were aligned to the hg19 human genome build. Aligned reads were then sorted into genome coordinate order and duplicate reads marked using http://broadinstitute.github.io/picard). Somatic single nucleotide variants (SNV) and INDELs were detected using Indelocator. To increase accuracy, additional filters were applied to high-confidence calls. Quality control modules within the Firehose pipeline (http://www.broadinstitute.org/cancer/cga/Firehose) were applied to all sequencing data for assessment of genotype concordance between tumor and normal paired samples. Contamination of samples was estimated using ContEst (8). All samples were required to have <0.5% contamination. MuTect 2.0 was applied to identify somatic SNVs for 16 of the samples. For the one tumor sample without a paired normal (I-755T), the GATK HaplotypeCaller (GATK version 3.7) was applied to identify the mutations. Identified variants were annotated by using Oncotator (version 1.2.7.0; ref. 9). To improve the specificity of mutation calls, only mutations with an allelic fraction >5% were considered. The remaining variants were removed if they were present at greater than 0.5% in the 1000 Genomes (phase I, version 3) or ExAC (version 0.3; ref. 10) databases and not present in COSMIC (version 74). Oncoprint and mutation mapping were done using c-bio portal tools (11, 12).

RNA sequencing (RNA-seq) was performed by the sequencing core facility at MD Anderson using the Applied Biosystems SOLiD system. Total RNA for whole-transcriptome sequencing was prepared according to vendor's protocols and individual prepared “barcode” libraries were quantified and pooled equally together for multiplexing. The sequencing runs were performed on SOLiD v 3.0. To identify SBA-related differentially expressed genes, the “reads per kilobase of exon per million mapped sequence reads” (RPKM) values of the human RefSeq genes were calculated using the RNA-seq flow in the Partek Genomics Suite (version 6.5 beta, Partek Inc.) and then log transformed. Single-factor ANOVA was used to detect differentially expressed genes among 18,890 protein-coding genes: P < 9.5 × 10−4 (FDR < 0.05) was used as the cutoff in the two-group comparison tumor versus adjacent normal tissue.

TES was performed by Genewiz, Inc. from formalin-fixed, paraffin embedded (FFPE) samples. Targeted whole-exome capture libraries were constructed from tumor DNA after sample shearing, end repair, phosphorylation, and ligation to barcoded sequencing adaptors. DNA was then subjected to hybrid capture using Agilent SureSelect Gene Enrichment workflow. The minimum coverage was set as 20×, although most of the samples had 300× to 400× coverage and a variant probability of 90%; required variant count was 2. The resulting SNPs/INDELs were further filtered to limit the potential of false positives (min. base Q = 30; min. forward/reverse balance = 0.05; min. frequency = 1.0%). The marginally filtered SNPs/INDELs generated above were annotated using the dbSNP138.txt (http://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/) and common_no_known_medical_impact_20140430.vcf from the NCBI. A list of genes included in this panel is shown in Supplementary Table S1.

Mutation rates and nucleotide transition signatures

The mutation rate for each sample was calculated by dividing the number of final mutation calls after filtering by the total genomic territory sequenced (in megabases). To identify mutation signatures that match known signature patterns found in other types of solid tumors from The Cancer Genome Atlas (TCGA), we applied the R package DeconstructSigs to estimate underlying source mutation signatures for our SBA tri-nucleotide mutation profiles (13). We obtained a measure of tumor heterogeneity for each tumor sample based on a technique that clusters mutation allele frequencies (14). We then estimated tumor cellularity and ploidy for each sample by applying Sequenza, which jointly analyzes mutation allele frequencies and copy number profiles (15).

IHC staining and ERBB2 copy number aberrations

IHC testing for ERBB2 (14/27 TES samples) and for microsatellite instability-high by MLH1, MSH2, MSH6, and PMS2 (42/44 TES and WES samples) was performed on FFPE tissue sections as reported previously (16). To detect HER2 overexpression, HER2 FISH was performed by Biotheranostics Inc. using the HER2IQ FISH PharmDx Kit (Agilent Technologies) according to the manufacturer's instructions. Appropriate positive and negative control tissues were assayed in parallel. The carcinoma component of the specimen was scored on the basis of evaluation of hematoxylin and eosin (H&E)-stained slides. HER2 and CEN-17 signals were counted for a total of 20 cells using the manufacturer's recommended interpretation of scoring. Specimens with a HER2/CEN-17 ratio >2.2 were considered HER2 gene-amplified (17). ERBB2 absolute copy-number alterations for each of the 17 WES samples was estimated using the BAM multiscale reference and NGS-quadratic correction (Nexus 9.0_biodiscovery, 2017). Segments were analyzed using SNP-FAST2 Segmentation, with high gain defined as higher than 0.6 and high loss as at least −1.2.

For PDXs, IHC analysis slides were stained for Ki67 (Cell Signaling Technology, #9027, 1:50, 30 min) and for phosphorylated-ERK1/2 (Cell Signaling Technology, pERK clone D13.14.4E; 1:150) as described previously (18).

Immunofluorescence, confocal analysis, and in vitro kinase assay

For immunofluorescence analysis, cells were seeded into 8-chamber collagen-coated chamber slides (Thermo Fisher Scientific). The cells were fixed with 3.7% formaldehyde (Tsoumis), extracted with chilled ethanol/methanol (1:1 volume), and blocked with 10% normal goat serum. Primary antibodies against E-cadherin and autophosphorylated EGFR (Cell Signaling Technology) were added, and the cells were incubated overnight and Z-sectioning was used for 3D virtual reconstruction, as described elsewhere (19).

Endogenous ERBB2 was pulled down using an anti-ERBB2 antibody (catalog 2165S, Cell Signaling Technology) from colorectal cancer cell line KM12L4 (wild type; Korean Cell Line Bank), and SBA cell lines SBA-6 (V841I) and SBA-16 (Y803H) and incubated with its substrate (Glu, Tyr; catalog CS0730; Sigma-Aldrich) at a ratio of 4:1 as per the manufacturer's instructions. Substrate phosphorylation was detected by dot blot assay using anti-phosphotyrosine PY69 (Cell Signaling Technology). Recombinant EGFR pretreated with EGF was used as positive control. IgG pull-down was used as negative control.

RPPA analysis of ERBB2-targeted therapy

To identify changes in ERBB2-mutant and wild-type tumors in response to ERBB2 inhibition, we used reverse-protein phase array (RPPA) analysis. Cellular proteins derived from the tumors were arrayed on nitrocellulose coated slides (Grace BioLabs) by an Aushon 2470 Arrayer (Aushon BioSystems). Each slide was probed with a validated primary antibody with a Pearson correlation coefficient between RPPA and Western blotting of greater than 0.7. Each dilution curve was fitted with a logistic model (“Supercurve Fitting” developed by the Department of Bioinformatics and Computational Biology in MD Anderson Cancer Center, http://bioinformatics.mdanderson.org/OOMPA).

Statistical analyses

Gehan–Breslow–Wilcoxon test and Fisher exact test were used to test associations between mutations and continuous and nominal variables. Pearson correlation was used to assess the linear relationship between continuous variables. Log-rank test and Cox proportional hazards models were applied for the time-to-event outcome analysis. P ≤ 0.05 was considered statistically significant. Factors in multivariate modeling included age, microsatellite instability status, small bowel location, TNM stage, histologic grade, perioperative chemotherapy use, and ERBB2 signaling cascade (ESC) mutation status.

Mutational landscape by WES

In the 16 paired WES samples, the overall mutation rate was 44.1 mutations/Mb in microsatellite instability–high (MSI-high) samples and 12.3 mutations/Mb in microsatellite stable (MSS) samples. From these mutations, we identified 3 underlying mutational processes (Supplementary Fig. S1; ref. 13). The most frequent mutation signature was Signature 1, a pattern associated with spontaneous deamination of 5-methylcytosine (5-mC) that accounts for a significant percentage of the critical somatic driver mutations observed in most cancers, including colorectal cancerss. For example, 60% of the APC nonsense mutations in SBA derived from 5-mC>T deamination. The second most common signature observed was a defective DNA mismatch repair signature (Signature 6), which is most frequently found in colorectal and uterine cancers and is highly prevalent in MSI-high tumors. As anticipated, the contribution of this signature to the mutation profiles was associated with MSI status, where percent contributions for MSI and MSS SBA tumors were 33.8% and 7.7%, respectively (P = 0.001; Wilcoxon rank-sum test). The third most frequent signature was Signature 5, whose etiology is currently unknown but is represented in all cancer types. Two cases (I.135 and I.1755) demonstrated high mutational rates but were MSS by intact expression of mismatch repair proteins and did not demonstrate evidence of the defective DNA mismatch repair mutational signature 6. The etiology of the high mutational rate in these two cases is not known, as neither case demonstrated POLE/POLD mutations, which has been associated with high mutational rates in colorectal cancer.

A frequency-based analysis of the somatic mutations in the WES cohort identified several tumor drivers with relatively higher recurrent mutation rates, such as APC (36%), NF1 (36%), ERBB2 (36%), KRAS (32%), or KMT2D (23%) (Fig. 1). Pathway enrichment analysis using the Ingenuity platform (IPA; Qiagen Bioinformatics) of the top 200 exomic mutations revealed several signaling pathways that were frequently mutated in the WES cohort, including WNT, ERBB2, STAT3, and chromatin remodeling (Supplementary Table S2).

Figure 1.

Whole exome–based somatic mutational landscape in SBA samples. Color-coded somatic aberrations are shown for by type (missense, nonsense, indels) or location (coding or noncoding) for each WES sample (n = 17), together with relevant clinicopathologic parameters. Only genes with somatic mutation in at least 3 samples are shown.

Figure 1.

Whole exome–based somatic mutational landscape in SBA samples. Color-coded somatic aberrations are shown for by type (missense, nonsense, indels) or location (coding or noncoding) for each WES sample (n = 17), together with relevant clinicopathologic parameters. Only genes with somatic mutation in at least 3 samples are shown.

Close modal

Mutation validation by TES

To further validate the frequently mutated genes found in the WES cohort, we used TES on an additional 27 samples. The targeted gene panel included over 250 genes that have been described to be associated with colorectal cancer or other solid tumors as reported in the COSMIC database (Supplementary Table S1). The most frequently mutated genes in the TES cohort included genes that code chromatin modifiers/epigenetic signaling, STAT3 and WNT signaling, and the ERBB2 signaling pathway (Fig. 2A).

Figure 2.

Gene set mutations in SBA. A, Oncoprint representation of frequently mutated genes in SBA across WES and TES cohorts. B, Nonsynonymous mutation mapping in ERBB2, RNF43, and ZNRF3 found in WES and TES cohorts. C, Kaplan–Meier plot representing overall survival (months) in patients with and without mutations in 8 members of the ESC (HR = 2.4, P = 0.03).

Figure 2.

Gene set mutations in SBA. A, Oncoprint representation of frequently mutated genes in SBA across WES and TES cohorts. B, Nonsynonymous mutation mapping in ERBB2, RNF43, and ZNRF3 found in WES and TES cohorts. C, Kaplan–Meier plot representing overall survival (months) in patients with and without mutations in 8 members of the ESC (HR = 2.4, P = 0.03).

Close modal

Pathways of interest across TES and WES cohorts

Twenty-three percent of the samples harbored APC mutations, 70% of which were truncating mutations (Fig. 2A). An additional 12% of samples displayed genetic alterations in other WNT-signaling inhibitors, including RNF43 and ZNRF3 (Fig. 2B), although the latter was not tested in the TES cohort. In total, 27 (61%) of the 44 samples had an alteration in a WNT signaling–related gene.

We also assessed WNT activation by assessing the differential expression of known WNT-related genes between normal and tumor samples using RNA-seq analysis of 15 WES samples (11 tumor and 4 normal). Several known secreted WNT inhibitors, including WIF1 and SFRP1, were significantly downregulated in >90% of the tumor samples (Supplementary Fig. S2A). Furthermore, a consistent upregulation of WNT-regulated stem cell markers, such as LGR5, HOPX, and BMI1, was evident across all tumor samples. Finally, CTNNB1 direct targets, such as AXIN2 and WISP1, were consistently upregulated in tumor samples, including those devoid of WNT pathway gene mutations (Supplementary Fig. S2A). Altogether, these results suggest that WNT activation is widely represented across SBA.

Another striking difference from colorectal cancer in the SBA samples was the relatively large number of samples with truncation mutations upstream of the SET domains of two histone–lysine N-methyltransferases, KMT2C and KMT2D (Supplementary Table S3). Twenty-nine percent of samples displayed nonsense mutations in either KMT2D or KMT2C, and 2 samples displayed nonsense mutations for both methyltransferases (Fig. 2A). Suppressor of zeste 12 homolog (SUZ12) is a zinc finger gene and a polycomb family member that also cooperates with KMT2C and KMT2D in the histone methylation process. SUZ12 was mutated in one third of the WES samples (Fig. 2A), underscoring the important interplay between various epigenetic regulatory mechanisms in SBA.

Of the greatest possible clinical relevance was our finding of ERBB2 alterations in 23% (10/44) of samples, with somatic point mutations in the ERBB2 gene in 7 samples and ERBB2 amplification in 3 samples. Of the 7 ERBB2 mutations, 4 were in the kinase domain (V777L, V842I, D769Y, Y803H), one in the extracellular domain (S310Y), and 2 samples were non-activating mutations (R678Q; Fig. 2B).

As previously mentioned, IPA of the top 200 WES mutations demonstrated that the STAT3 canonical signaling pathway was significantly altered in our SBA cohorts (Supplementary Table S2). STAT3 can serve as a network hub, where various receptor tyrosine kinase activators, including ERBB2, EGFR, TGFBR2, or IGFR2, can converge and induce its activation. In addition, 14% of samples demonstrated alterations in the tumor suppressor tyrosine phosphatase receptor type T (PTPRT), which has been shown to directly regulate STAT3 (20, 21). Further confirmation of ERBB2 and STAT3 signaling activity was provided by a transcriptome analysis of the 11 tumor samples from the WES cohort. IPA confirmed that STAT3 and ERBB2 were significantly deregulated (Supplementary Table S4).

Clinical relevance of ERBB signaling cascade (ESC) mutations in SBA

For the entire cohort of 44 SBA cases, the median follow-up interval was 84 months, and 53% were alive at the time of last follow-up. We evaluated the clinical impact of activating mutations in 8 members of the ESC (Supplementary Fig. S2B), using a grouping methodology similar to that described previously (22), and generated an ESC-mutated group (n = 27) and an ESC–nonmutated group (n = 17; Supplementary Fig. S2C). In univariate analysis, the ESC-mutated group demonstrated a poorer survival duration than the ESC–nonmutated group (Fig. 2C), with a median survival of 70.4 months compared with 109 months (log-rank test P = 0.03 and log-rank HR = 2.4). In multivariate modeling, ESC mutation was the only significant variable (HR = 3.6, 95% confidence interval, 1.2–10.7; P = 0.02).

SBA cell lines and in vitro targeting of ERBB2 mutations

To further validate ERBB2 as a target, we developed two SBA patient-derived cell lines: SBA-6 from human sample I-797, which harbors the ERBB2 mutation V842I, and SBA-16 from human sample I-577, which harbors the mutation Y803H (Fig. 2B). These lines, when injected into nude mice, developed tumors with typical adenocarcinoma features. In addition, both cell lines formed spheres when grown in low-attachment plates (Fig. 3A and B, middle) or formed 3-dimensional (3D) structures when attached on collagen- or vitronectin-coated plates (Fig. 3A and B, right). A typical 3D-reconstructed structure is shown in Fig. 3C, where red represents EGFR-Y1068, green represents E-cadherin, and blue is nuclear stain Topro-3. Altogether, these results suggest that tumor-derived small intestine cell lines are highly differentiated and maintain an organized cellular architecture.

Figure 3.

Morphologic similarities of tumors in situ and in tumor-derived cell lines in vitro. A, Representative images of sample I-797, showing in situ tumor staining with H&E (left), tumor-derived cells cultured in low-attachment culture dishes (suspension; middle), and tumor-derived cells cultured in extracellular matrix (ECM)–coated tissue culture dishes (attached; right). B, Representative images of sample I-577 under the same conditions as in A. Note the pseudocrypt formation by both cell lines. C, A 3D-reconstructed pseudocrypt derived from SBA-6 shown from base to top and stained with anti–EGFR-Y1068 (red), anti–E-cadherin (green), and nuclear stain (blue). Note distribution of active EGFR within the pseudocrypt structure.

Figure 3.

Morphologic similarities of tumors in situ and in tumor-derived cell lines in vitro. A, Representative images of sample I-797, showing in situ tumor staining with H&E (left), tumor-derived cells cultured in low-attachment culture dishes (suspension; middle), and tumor-derived cells cultured in extracellular matrix (ECM)–coated tissue culture dishes (attached; right). B, Representative images of sample I-577 under the same conditions as in A. Note the pseudocrypt formation by both cell lines. C, A 3D-reconstructed pseudocrypt derived from SBA-6 shown from base to top and stained with anti–EGFR-Y1068 (red), anti–E-cadherin (green), and nuclear stain (blue). Note distribution of active EGFR within the pseudocrypt structure.

Close modal

The Y803H mutation found in sample SBA-16 has not been previously reported in the COSMIC database, whereas mutation V842I has previously been reported to increase the kinase activity of ERBB2 and improve response to lapatinib (23). An in vitro kinase assay using immunoprecipitated endogenous ERBB2 confirmed that both mutations are kinase activating (Fig. 4A). Exposure to an ERBB2 inhibitor caused significant decreases in cell proliferation, with IC50 at 25 nmol/L for lapatinib (SBA-6 and SBA-16) and at 1 nmol/L (SBA-6) or 2.5 nmol/L (SBA-16) for dacomitinib (Fig. 4B). A reduction in downstream ERBB2 signaling was seen after exposure to lapatinib (100 nmol/L) for 72 hours (Fig. 4C). In particular, phospho-AKT, phospho-ERK1, and phospho-MEK1 demonstrated reductions, as did total STAT3.

Figure 4.

In vitro sensitivity of ERBB2 kinase domain mutant cell lines SBA-6 and SBA-16 to ERBB2 inhibition. A,In vitro kinase assay using endogenous pulled-down ERBB2 from kinase mutant SBA cell lines [SBA-6 (V842I) and SBA-16 (Y803H)] or kinase wild-type (WT) KM12L4 colorectal cancer cells. Representative dot-blot conditions are shown from original single blot containing multiple replicates and controls. B, Log-dose versus response in SBA tumor-derived cell lines SBA-6 (left) and SBA-16 (right) treated with ERBB2 inhibitor lapatinib (Lapa) or dacomitinib (Daco). C, Following treatment with Lapa at 100 nmol/L for 72 hours cells lines were analyzed and fold change of normalized medians for each protein derived from RPPA for downstream ERBB2 signaling proteins are shown: phospho-AKT_pT308, phospho-ERK1-pT202_Y204, phospho-MEK1-pS217_S221, and total STAT3.

Figure 4.

In vitro sensitivity of ERBB2 kinase domain mutant cell lines SBA-6 and SBA-16 to ERBB2 inhibition. A,In vitro kinase assay using endogenous pulled-down ERBB2 from kinase mutant SBA cell lines [SBA-6 (V842I) and SBA-16 (Y803H)] or kinase wild-type (WT) KM12L4 colorectal cancer cells. Representative dot-blot conditions are shown from original single blot containing multiple replicates and controls. B, Log-dose versus response in SBA tumor-derived cell lines SBA-6 (left) and SBA-16 (right) treated with ERBB2 inhibitor lapatinib (Lapa) or dacomitinib (Daco). C, Following treatment with Lapa at 100 nmol/L for 72 hours cells lines were analyzed and fold change of normalized medians for each protein derived from RPPA for downstream ERBB2 signaling proteins are shown: phospho-AKT_pT308, phospho-ERK1-pT202_Y204, phospho-MEK1-pS217_S221, and total STAT3.

Close modal

ERBB2 targeting in SBA PDX models

To determine whether the 2 ERBB2 kinase activating mutations are responsive to ERBB2 inhibitors in vivo, we implanted nude mice with one of 3 different SBA tumors, one ERBB2 wild-type and 2 ERBB2 mutant. Because both SBA-6 and SBA-16 cell lines displayed slightly higher in vitro sensitivity to dacomitinib than to lapatinib, we decided to treat the mice with dacomitinib.

All 3 of the PDX models, the wild-type ERBB2 mouse model PDX20 (derived from human sample T20) and the 2 ERBB2-mutant models PDX6 (derived from SBA-6 cells) and PDX16 (derived from SBA-16), displayed a KRAS or NRAS mutation in either codon 12 or 13. Dacomitinib significantly reduced tumor growth in both ERBB2-mutant models, whereas it had no effect in the wild-type model (Fig. 5A). Microscopic analysis of H&E-stained sections representative of control and treated tumors in each model showed similar adenocarcinoma features with a somewhat richer mucinous component in the wild-type PDX20 model than in the mutant PDX6 and PDX16 models (Fig. 5B, left; Supplementary Fig. S3). The uniform glandular morphology was drastically disrupted in PDX6 and PDX16 by dacomitinib, with evident areas of necrosis, whereas no significant changes were evident in the PDX20 tumors (Fig. 5B; Supplementary Fig. S3). Both Ki-67 and phospho-ERK stains were reduced in ERBB2-mutant models (P = 0.033 and P = 0.048, respectively) with dacomitinib treatment in contrast to control, whereas no difference was seen in the ERBB2 wild-type model (PDX20; Fig. 5C and D). Unsupervised clustering of significantly differentially expressed proteins (P < 0.05) from RPPA analysis of tumors after 3 weeks of treatment revealed 2 types of differences among the mouse models used. Differences in protein expression between ERBB2-mutant and wild-type models may reflect differences in prosurvival mechanisms and autophagy-related responses, as suggested by differences in HIAP/BIRC3 and ATG7 expression (Fig. 5E, yellow vertical bar). Differences related to dacomitinib response in ERBB2-mutant and wild-type models included reductions in the mutant models of expression of proteins directly related to tumor-specific glycolysis: FASN, MCT4, PKM2, LDHA, and SDHA (Fig. 5E, green vertical bar).

Figure 5.

In vivo sensitivity of ERBB2 kinase domain–mutant tumors and wild-type control tumors to ERRB2 inhibition. A, Sensitivity to dacomitinib (Daco) or control (Con) of ERBB2-mutant SBA PDX models PDX6 (derived from SBA-6 cells) and PDX16 (derived from SBA-16 cells) and ERBB2 wild-type (WT) SBA PDX model PDX20. B, H&E-stained representative examples of Daco-treated and control-treated ERBB2-mutant tumors. C, Ki-67 and phospho-ERK from Daco-treated and control-treated PDX models at 3 weeks. D, Representative examples of phospho-ERK and Ki-67 stained Daco-treated and control-treated ERBB2-mutant PDX16 tumors. E, Heatmap representation of normalized median RPPA values for tumors from each of these models at 3 weeks showing cluster classification of significantly (P < 0.05) differentially expressed proteins between WT and ERBB2 mutants, yellow bar, or control versus dacomitinib treatment, green bar.

Figure 5.

In vivo sensitivity of ERBB2 kinase domain–mutant tumors and wild-type control tumors to ERRB2 inhibition. A, Sensitivity to dacomitinib (Daco) or control (Con) of ERBB2-mutant SBA PDX models PDX6 (derived from SBA-6 cells) and PDX16 (derived from SBA-16 cells) and ERBB2 wild-type (WT) SBA PDX model PDX20. B, H&E-stained representative examples of Daco-treated and control-treated ERBB2-mutant tumors. C, Ki-67 and phospho-ERK from Daco-treated and control-treated PDX models at 3 weeks. D, Representative examples of phospho-ERK and Ki-67 stained Daco-treated and control-treated ERBB2-mutant PDX16 tumors. E, Heatmap representation of normalized median RPPA values for tumors from each of these models at 3 weeks showing cluster classification of significantly (P < 0.05) differentially expressed proteins between WT and ERBB2 mutants, yellow bar, or control versus dacomitinib treatment, green bar.

Close modal

This large-scale analysis demonstrates alterations in previously reported genes of interest in SBA, such as APC, KRAS, and ERBB2, but also genes less recognized as implicated in SBA biology, such as the tumor suppressor PTPRT, the epigenetic regulators of H3K27me3 (SUZ12, KMT2C, and KMT2D), and the WNT activators RNF43 and ZNRF3. This work demonstrates that in novel SBA models, which harbor activating mutations in ERBB2, small-molecule inhibitors of ERBB2 have anticancer activity both in vitro and in vivo. Given the rarity of SBA, in which only limited clinical trials have been conducted, laboratory efforts are critical to identifying and informing novel clinical approaches for patients with this cancer.

An important finding from this study is that 23% of samples harbor either ERBB2 missense activating mutations or gene amplifications. Our results accord with previous reports identifying ERBB2 as a targetable molecule in SBA (5–7). We found not only that small-molecule ERBB2 inhibitors were effective against tumor growth both in vitro and in vivo, but also that downstream changes resulting from ERBB2 inhibition included inhibition of glycolysis and lactic acid transport, suggesting a metabolic phenotype in ERBB2-mutant tumors (24). In addition, we found that SBA with genomic alterations in the ERBB2 signaling cascade demonstrated worse survival, although given the limited sample size, further replication of this finding is needed.

As expected, APC mutations were found in less than 30% of the SBA samples tested, confirming the known difference between SBA and sporadic large intestine tumors, in which APC mutations represent the dominant somatic mutation (>70%). Furthermore, we identified loss-of-function alterations in the upstream WNT repressors RNF43 and ZNRF3 in samples without APC truncations. In total, 27 (61%) of the 44 samples had an alteration in a WNT signaling–related gene. Transcriptome analysis revealed that a WNT-driven phenotype is not the exception in SBA but is probably the norm because all tumors displayed downregulation of secreted WNT suppressors such as WIF1 and SFRP1 and upregulation of LGR5 and LGR6, which mark actively dividing stem cell populations (25), or of HOPX and BIM1, which mark WNT-driven dormant stem cell populations. We found that WNT ligand activation by ZNRF3 and RNF43 loss of function could be responsible for WNT activation in at least 12% (5/42) of SBAs. This pathway of WNT activation is vulnerable to Porcupine, an enzyme catalyzing the acylation of WNT proteins (25). Trials investigating the anticancer effects of Porcupine inhibition are currently ongoing, although primarily focused upon colorectal cancer (NCT01351103). Prior work in SBA has noted abnormal nuclear accumulation of β-catenin as a marker of WNT activation in 20% to 41% of cases, and in one study correlated with worse overall survival (26, 27). In colorectal cancer, despite near universal alterations in the APC gene, abnormal β-catenin expression is observed in a limited number of cases, often heterogenous within a tumor, and of unclear prognostic significance (28, 29). Thus, as in colorectal cancer, additional factors besides single alterations in the WNT pathway genes are likely involved with downstream WNT-related protein expression.

One of the most intriguing findings of our study is that 70% of the SBA samples (WES and TES combined) have somatic mutation in KMT2C and/or KMT2D methyltransferases, whereas only approximately 10% of colorectal cancer samples have such a mutation (30–32). Methylation of histone H3 lysine 4 is one method that cells use to mark promoters, enhancers, and superenhancers for further recruitment of transcription coactivators, such as p300 (33). Several studies have underscored the roles of KMT2C and KMT2D as important suppressors of various cancers (34, 35). Nonsense mutations constitute 37% and 60%, respectively, of the total KMT2D and KMT2C mutations in many malignancies, including those of the esophagus and prostate (36–39). Importantly, recent studies comparing intensive mapping of enhancer modulation by KMT2C and KMT2D between normal crypts and tumor areas in colorectal cancer revealed differential enrichment that translates into modified expression of genes of various signaling pathways, including the WNT pathway (40). Moreover, KMT2D deficiency alone can induce tumor formation by transcriptional stress, which triggers abnormalities in early replicating fragile sites within the chromosome (41, 42). Recently the N-terminal domain of KMT2C was demonstrated to interact with the histone H2A deubiquitinase and tumor suppressor BAP1 and may reflect the mechanism seen in SBA as the majority of KMT2C mutations, 72%, occurred in the N-terminal domain (43).

A final finding of potential clinical relevance was the consistent finding of STAT3 pathway activation in SBA. A known suppressor of STAT3 activation, PTPRT, was mutated in 14% of SBA samples. These mutations were all identified within the catalytic domains of PTPRT and occurred in samples without upstream alterations in receptor tyrosine kinases such as ERBB, EGFR, or TGFBR2.

Although this study represents one of the largest known molecular analyses of SBA tumor samples, our analyses were limited by sample size, and validation in additional datasets is needed. In particular, as frozen tumor samples were collected from a referral hospital, there is likely a selection bias for more indolent cases, as the higher than expected MSI-high rate may reflect. However, the use of in vivo and in vitro models for the identified ERBB2 mutations lends strong support to the importance of ERBB2 alterations in this tumor type. Given the overall low frequency of ERBB2 alterations in this rare cancer, however, clinical trials investigating ERBB2 targeting are unlikely to occur. Thus, the development of novel model systems, as reported here, represents a critical step forward in our efforts to both understand this cancer and guide development of potential novel therapies for patients with metastatic SBA.

M.T. Tetzlaff is a consultant/advisory board member for Novartis LLC. M.J. Overman is a consultant/advisory board member for Bristol-Myers Squibb, Roche, and MedImmune. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Adam, D. Menter, S. Kopetz, M.J. Overman

Development of methodology: L. Adam, F.A. San Lucas, M.T. Tetzlaff, S. Kopetz, M.J. Overman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Adam, H. Wang, M.T. Tetzlaff, S. Kopetz, M.J. Overman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Adam, F.A. San Lucas, R. Fowler, Y. Yu, W. Wu, Y. Liu, H. Wang, J. Ensor Jr, G. Manyam, S.T. Arold, C. Huff, S. Kopetz, P. Scheet, M.J. Overman

Writing, review, and/or revision of the manuscript: L. Adam, F.A. San Lucas, R. Fowler, H. Wang, D. Menter, J. Ensor Jr, S.T. Arold, S. Kopetz, P. Scheet, M.J. Overman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Adam, F.A. San Lucas, R. Fowler, M.J. Overman

Study supervision: L. Adam, M.J. Overman

This work was supported by the Kavanagh Family Foundation and by the NCI through Cancer Center Support Grant P30CA16672. S.T. Arold was supported by funding from King Abdullah University of Science and Technology.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Raghav
K
,
Overman
MJ
. 
Small bowel adenocarcinomas–existing evidence and evolving paradigms
.
Nat Rev Clin Oncol
2013
;
10
:
534
44
.
2.
Overman
MJ
,
Hu
CY
,
Kopetz
S
,
Abbruzzese
JL
,
Wolff
RA
,
Chang
GJ
. 
A population-based comparison of adenocarcinoma of the large and small intestine: insights into a rare disease
.
Ann Surg Oncol
2012
;
19
:
1439
45
.
3.
Gulhati
P
,
Raghav
K
,
Shroff
R
,
Varadhachary
G
,
Javle
M
,
Qiao
W
, et al
Phase II study of panitumumab in RAS wild-type metastatic adenocarcinoma of small bowel or ampulla of vater
.
Oncologist
2018
;
23
:
277
e26
.
4.
Alvi
MA
,
McArt
DG
,
Kelly
P
,
Fuchs
MA
,
Alderdice
M
,
McCabe
CM
, et al
Comprehensive molecular pathology analysis of small bowel adenocarcinoma reveals novel targets with potential for clinical utility
.
Oncotarget
2015
;
6
:
20863
74
.
5.
Laforest
A
,
Aparicio
T
,
Zaanan
A
,
Silva
FP
,
Didelot
A
,
Desbeaux
A
, et al
ERBB2 gene as a potential therapeutic target in small bowel adenocarcinoma
.
Eur J Cancer
2014
;
50
:
1740
6
.
6.
Yuan
W
,
Zhang
Z
,
Dai
B
,
Wei
Q
,
Liu
J
,
Liu
Y
, et al
Whole-exome sequencing of duodenal adenocarcinoma identifies recurrent Wnt/beta-catenin signaling pathway mutations
.
Cancer
2016
;
122
:
1689
96
.
7.
Schrock
AB
,
Devoe
CE
,
McWilliams
R
,
Sun
J
,
Aparicio
T
,
Stephens
PJ
, et al
Genomic profiling of small-bowel adenocarcinoma
.
JAMA Oncol
2017
;
3
:
1546
53
.
8.
Cibulskis
K
,
McKenna
A
,
Fennell
T
,
Banks
E
,
DePristo
M
,
Getz
G
. 
ContEst: estimating cross-contamination of human samples in next-generation sequencing data
.
Bioinformatics
2011
;
27
:
2601
2
.
9.
Ramos
AH
,
Lichtenstein
L
,
Gupta
M
,
Lawrence
MS
,
Pugh
TJ
,
Saksena
G
, et al
Oncotator: cancer variant annotation tool
.
Hum Mutat
2015
;
36
:
E2423
9
.
10.
Lek
M
,
Karczewski
KJ
,
Minikel
EV
,
Samocha
KE
,
Banks
E
,
Fennell
T
, et al
Analysis of protein-coding genetic variation in 60,706 humans
.
Nature
2016
;
536
:
285
91
.
11.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
12.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
13.
Rosenthal
R
,
McGranahan
N
,
Herrero
J
,
Taylor
BS
,
Swanton
C
. 
DeconstructSigs: delineating mutational processes in single tumors distinguishes DNA repair deficiencies and patterns of carcinoma evolution
.
Genome Biol
2016
;
17
:
31
.
14.
Mroz
EA
,
Rocco
JW
. 
MATH, a novel measure of intratumor genetic heterogeneity, is high in poor-outcome classes of head and neck squamous cell carcinoma
.
Oral Oncol
2013
;
49
:
211
5
.
15.
Favero
F
,
Joshi
T
,
Marquard
AM
,
Birkbak
NJ
,
Krzystanek
M
,
Li
Q
, et al
Sequenza: allele-specific copy number and mutation profiles from tumor sequencing data
.
Ann Oncol
2015
;
26
:
64
70
.
16.
Overman
MJ
,
Pozadzides
J
,
Kopetz
S
,
Wen
S
,
Abbruzzese
JL
,
Wolff
RA
, et al
Immunophenotype and molecular characterisation of adenocarcinoma of the small intestine
.
Br J Cancer
2010
;
102
:
144
50
.
17.
Martincorena
I
,
Luscombe
NM
. 
Non-random mutation: the evolution of targeted hypermutation and hypomutation
.
Bioessays
2013
;
35
:
123
30
.
18.
Amaria
RN
,
Prieto
PA
,
Tetzlaff
MT
,
Reuben
A
,
Andrews
MC
,
Ross
MI
, et al
Neoadjuvant plus adjuvant dabrafenib and trametinib versus standard of care in patients with high-risk, surgically resectable melanoma: a single-centre, open-label, randomised, phase 2 trial
.
Lancet Oncol
2018
;
19
:
181
93
.
19.
Adam
L
,
Mazumdar
A
,
Sharma
T
,
Jones
TR
,
Kumar
R
. 
A three-dimensional and temporo-spatial model to study invasiveness of cancer cells by heregulin and prostaglandin E2
.
Cancer Res
2001
;
61
:
81
7
.
20.
Zhang
X
,
Guo
A
,
Yu
J
,
Possemato
A
,
Chen
Y
,
Zheng
W
, et al
Identification of STAT3 as a substrate of receptor protein tyrosine phosphatase T
.
Proc Natl Acad Sci U S A
2007
;
104
:
4060
4
.
21.
Wang
Z
,
Shen
D
,
Parsons
DW
,
Bardelli
A
,
Sager
J
,
Szabo
S
, et al
Mutational analysis of the tyrosine phosphatome in colorectal cancers
.
Science
2004
;
304
:
1164
6
.
22.
Li
M
,
Zhang
Z
,
Li
X
,
Ye
J
,
Wu
X
,
Tan
Z
, et al
Whole-exome and targeted gene sequencing of gallbladder carcinoma identifies recurrent mutations in the ErbB pathway
.
Nature genetics
2014
;
46
:
872
6
.
23.
Ross
JS
,
Gay
LM
,
Wang
K
,
Ali
SM
,
Chumsri
S
,
Elvin
JA
, et al
Nonamplification ERBB2 genomic alterations in 5605 cases of recurrent and metastatic breast cancer: An emerging opportunity for anti-HER2 targeted therapies
.
Cancer
2016
;
122
:
2654
62
.
24.
Zhao
YH
,
Zhou
M
,
Liu
H
,
Ding
Y
,
Khong
HT
,
Yu
D
, et al
Upregulation of lactate dehydrogenase A by ErbB2 through heat shock factor 1 promotes breast cancer cell glycolysis and growth
.
Oncogene
2009
;
28
:
3689
701
.
25.
Nusse
R
,
Clevers
H
. 
Wnt/beta-catenin signaling, disease, and emerging therapeutic modalities
.
Cell
2017
;
169
:
985
99
.
26.
Lee
HJ
,
Lee
OJ
,
Jang
KT
,
Bae
YK
,
Chung
JY
,
Eom
DW
, et al
Combined loss of E-cadherin and aberrant beta-catenin protein expression correlates with a poor prognosis for small intestinal adenocarcinomas
.
Am J Clin Pathol
2013
;
139
:
167
76
.
27.
Aparicio
T
,
Svrcek
M
,
Zaanan
A
,
Beohou
E
,
Laforest
A
,
Afchain
P
, et al
Small bowel adenocarcinoma phenotyping, a clinicobiological prognostic study
.
Br J Cancer
2013
;
109
:
3057
66
.
28.
Brabletz
T
,
Jung
A
,
Reu
S
,
Porzner
M
,
Hlubek
F
,
Kunz-Schughart
LA
, et al
Variable beta-catenin expression in colorectal cancers indicates tumor progression driven by the tumor environment
.
Proc Natl Acad Sci U S A
2001
;
98
:
10356
61
.
29.
Zhang
S
,
Wang
Z
,
Shan
J
,
Yu
X
,
Li
L
,
Lei
R
, et al
Nuclear expression and/or reduced membranous expression of beta-catenin correlate with poor prognosis in colorectal carcinoma: a meta-analysis
.
Medicine
2016
;
95
:
e5546
.
30.
The Cancer Genome Atlas Network
. 
Comprehensive molecular characterization of human colon and rectal cancer
.
Nature
2012
;
487
:
330
7
.
31.
Giannakis
M
,
Mu
XJ
,
Shukla
SA
,
Qian
ZR
,
Cohen
O
,
Nishihara
R
, et al
Genomic correlates of immune-cell infiltrates in colorectal carcinoma
.
Cell Rep
2016
;
17
:
1206
.
32.
Brannon
AR
,
Vakiani
E
,
Sylvester
BE
,
Scott
SN
,
McDermott
G
,
Shah
RH
, et al
Comparative sequencing analysis reveals high genomic concordance between matched primary and metastatic colorectal cancer lesions
.
Genome Biol
2014
;
15
:
454
.
33.
Lai
B
,
Lee
JE
,
Jang
Y
,
Wang
L
,
Peng
W
,
Ge
K
. 
MLL3/MLL4 are required for CBP/p300 binding on enhancers and super-enhancer formation in brown adipogenesis
.
Nucleic Acids Res
2017
;
45
:
6388
403
.
34.
Calo
E
,
Wysocka
J
. 
Modification of enhancer chromatin: what, how, and why?
Mol Cell
2013
;
49
:
825
37
.
35.
Heintzman
ND
,
Stuart
RK
,
Hon
G
,
Fu
Y
,
Ching
CW
,
Hawkins
RD
, et al
Distinct and predictive chromatin signatures of transcriptional promoters and enhancers in the human genome
.
Nat Genet
2007
;
39
:
311
8
.
36.
Rao
RC
,
Dou
Y
. 
Hijacked in cancer: the KMT2 (MLL) family of methyltransferases
.
Nat Rev Cancer
2015
;
15
:
334
46
.
37.
Sausen
M
,
Phallen
J
,
Adleff
V
,
Jones
S
,
Leary
RJ
,
Barrett
MT
, et al
Clinical implications of genomic alterations in the tumour and circulation of pancreatic cancer patients
.
Nat Commun
2015
;
6
:
7686
.
38.
Gao
YB
,
Chen
ZL
,
Li
JG
,
Hu
XD
,
Shi
XJ
,
Sun
ZM
, et al
Genetic landscape of esophageal squamous cell carcinoma
.
Nat Genet
2014
;
46
:
1097
102
.
39.
Grasso
CS
,
Wu
YM
,
Robinson
DR
,
Cao
X
,
Dhanasekaran
SM
,
Khan
AP
, et al
The mutational landscape of lethal castration-resistant prostate cancer
.
Nature
2012
;
487
:
239
43
.
40.
Akhtar-Zaidi
B
,
Cowper-Sal-lari
R
,
Corradin
O
,
Saiakhova
A
,
Bartels
CF
,
Balasubramanian
D
, et al
Epigenomic enhancer profiling defines a signature of colon cancer
.
Science
2012
;
336
:
736
9
.
41.
Kantidakis
T
,
Saponaro
M
,
Mitter
R
,
Horswell
S
,
Kranz
A
,
Boeing
S
, et al
Mutation of cancer driver MLL2 results in transcription stress and genome instability
.
Genes Dev
2016
;
30
:
408
20
.
42.
Tubbs
A
,
Nussenzweig
A
. 
Endogenous DNA damage as a source of genomic instability in cancer
.
Cell
2017
;
168
:
644
56
.
43.
Wang
L
,
Zhao
Z
,
Ozark
PA
,
Fantini
D
,
Marshall
SA
,
Rendleman
EJ
, et al
Resetting the epigenetic balance of Polycomb and COMPASS function at enhancers for cancer therapy
.
Nat Med
2018
;
24
:
758
69
.