GMMchi: gene expression clustering using Gaussian mixture modeling.

BACKGROUND: Cancer evolution consists of a stepwise acquisition of genetic and epigenetic changes, which alter the gene expression profiles of cells in a particular tissue and result in phenotypic alterations acted upon by natural selection. The recurrent appearance of specific genetic lesions across individual cancers and cancer types suggests the existence of certain "driver mutations," which likely make up the major contribution to tumors' selective advantages over surrounding normal tissue and as such are responsible for the most consequential aspects of the cancer cells' gene expression patterns and phenotypes. We hypothesize that such mutations are likely to cluster with specific dichotomous shifts in the expression of the genes they most closely control, and propose GMMchi, a Python package that leverages Gaussian Mixture Modeling to detect and characterize bimodal gene expression patterns across cancer samples, as a tool to analyze such correlations using 2 × 2 contingency table statistics. RESULTS: Using well-defined simulated data, we were able to confirm the robust performance of GMMchi, reaching 85% accuracy with a sample size of n = 90. We were also able to demonstrate a few examples of the application of GMMchi with respect to its capacity to characterize background florescent signals in microarray data, filter out uninformative background probe sets, as well as uncover novel genetic interrelationships and tumor characteristics. Our approach to analysing gene expression analysis in cancers provides an additional lens to supplement traditional continuous-valued statistical analysis by maximizing the information that can be gathered from bulk gene expression data. CONCLUSIONS: We confirm that GMMchi robustly and reliably extracts bimodal patterns from both colorectal cancer (CRC) cell line-derived microarray and tumor-derived RNA-Seq data and verify previously reported gene expression correlates of some well-characterized CRC phenotypes. AVAILABILITY: The Python package GMMchi and our cell line microarray data used in this paper is available for downloading on GitHub at https://github.com/jeffliu6068/GMMchi .

N-Glycomic and Transcriptomic Changes Associated with CDX1 mRNA Expression in Colorectal Cancer Cell Lines.

The caudal-related homeobox protein 1 (CDX1) is a transcription factor, which is important in the development, differentiation, and homeostasis of the gut. Although the involvement of CDX genes in the regulation of the expression levels of a few glycosyltransferases has been shown, associations between glycosylation phenotypes and CDX1 mRNA expression have hitherto not been well studied. Triggered by our previous study, we here characterized the N-glycomic phenotype of 16 colon cancer cell lines, selected for their differential CDX1 mRNA expression levels. We found that high CDX1 mRNA expression associated with a higher degree of multi-fucosylation on N-glycans, which is in line with our previous results and was supported by up-regulated gene expression of fucosyltransferases involved in antenna fucosylation. Interestingly, hepatocyte nuclear factors (HNF)4A and HNF1A were, among others, positively associated with high CDX1 mRNA expression and have been previously proven to regulate antenna fucosylation. Besides fucosylation, we found that high CDX1 mRNA expression in cancer cell lines also associated with low levels of sialylation and galactosylation and high levels of bisection on N-glycans. Altogether, our data highlight a possible role of CDX1 in altering the N-glycosylation of colorectal cancer cells, which is a hallmark of tumor development.

Dsh homolog DVL3 mediates resistance to IGFIR inhibition by regulating IGF-RAS signaling.

Drugs that inhibit insulin-like growth factor 1 (IGFI) receptor IGFIR were encouraging in early trials, but predictive biomarkers were lacking and the drugs provided insufficient benefit in unselected patients. In this study, we used genetic screening and downstream validation to identify the WNT pathway element DVL3 as a mediator of resistance to IGFIR inhibition. Sensitivity to IGFIR inhibition was enhanced specifically in vitro and in vivo by genetic or pharmacologic blockade of DVL3. In breast and prostate cancer cells, sensitization tracked with enhanced MEK-ERK activation and relied upon MEK activity and DVL3 expression. Mechanistic investigations showed that DVL3 is present in an adaptor complex that links IGFIR to RAS, which includes Shc, growth factor receptor-bound-2 (Grb2), son-of-sevenless (SOS), and the tumor suppressor DAB2. Dual DVL and DAB2 blockade synergized in activating ERKs and sensitizing cells to IGFIR inhibition, suggesting a nonredundant role for DVL3 in the Shc-Grb2-SOS complex. Clinically, tumors that responded to IGFIR inhibition contained relatively lower levels of DVL3 protein than resistant tumors, and DVL3 levels in tumors correlated inversely with progression-free survival in patients treated with IGFIR antibodies. Because IGFIR does not contain activating mutations analogous to EGFR variants associated with response to EGFR inhibitors, we suggest that IGF signaling achieves an equivalent integration at the postreceptor level through adaptor protein complexes, influencing cellular dependence on the IGF axis and identifying a patient population with potential to benefit from IGFIR inhibition.

Colorectal cancer cell lines are representative models of the main molecular subtypes of primary cancer.

Human colorectal cancer cell lines are used widely to investigate tumor biology, experimental therapy, and biomarkers. However, to what extent these established cell lines represent and maintain the genetic diversity of primary cancers is uncertain. In this study, we profiled 70 colorectal cancer cell lines for mutations and DNA copy number by whole-exome sequencing and SNP microarray analyses, respectively. Gene expression was defined using RNA-Seq. Cell line data were compared with those published for primary colorectal cancers in The Cancer Genome Atlas. Notably, we found that exome mutation and DNA copy-number spectra in colorectal cancer cell lines closely resembled those seen in primary colorectal tumors. Similarities included the presence of two hypermutation phenotypes, as defined by signatures for defective DNA mismatch repair and DNA polymerase ε proofreading deficiency, along with concordant mutation profiles in the broadly altered WNT, MAPK, PI3K, TGFß, and p53 pathways. Furthermore, we documented mutations enriched in genes involved in chromatin remodeling (ARID1A, CHD6, and SRCAP) and histone methylation or acetylation (ASH1L, EP300, EP400, MLL2, MLL3, PRDM2, and TRRAP). Chromosomal instability was prevalent in nonhypermutated cases, with similar patterns of chromosomal gains and losses. Although paired cell lines derived from the same tumor exhibited considerable mutation and DNA copy-number differences, in silico simulations suggest that these differences mainly reflected a preexisting heterogeneity in the tumor cells. In conclusion, our results establish that human colorectal cancer lines are representative of the main subtypes of primary tumors at the genomic level, further validating their utility as tools to investigate colorectal cancer biology and drug responses.

Cancer cell lines for drug discovery and development.

Despite the millions of dollars spent on target validation and drug optimization in preclinical models, most therapies still fail in phase III clinical trials. Our current model systems, or the way we interpret data from them, clearly do not have sufficient clinical predictive power. Current opinion suggests that this is because the cell lines and xenografts that are commonly used are inadequate models that do not effectively mimic and predict human responses. This has become such a widespread belief that it approaches dogma in the field of drug discovery and optimization and has spurred a surge in studies devoted to the development of more sophisticated animal models such as orthotopic patient-derived xenografts in an attempt to obtain more accurate estimates of whether particular cancers will respond to given treatments. Here, we explore the evidence that has led to the move away from the use of in vitro cell lines and toward various forms of xenograft models for drug screening and development. We review some of the pros and cons of each model and give an overview of ways in which the use of cell lines could be modified to improve the predictive capacity of this well-defined model.

Direct and immune mediated antibody targeting of ERBB receptors in a colorectal cancer cell-line panel.

A significant proportion of colorectal cancer (CRC) patients are resistant to anti-ERBB1 [avian erythroblastic leukemia viral (v-erb-b) oncogene homolog, receptor for EGF] monoclonal antibodies (Mabs). We evaluated both immune and nonimmune effects of cetuximab (anti-ERBB1 Mab), trastuzumab (anti-ERBB2 Mab), pertuzumab (anti-ERBB2 Mab), and lapatinib (dual ERBB1 and ERBB2 tyrosine kinase inhibitor) in a large well-characterized panel of 64 CRC cell lines to find response predictive tumor characteristics. There was a significant correlation between the direct effects of cetuximab and lapatinib. Both agents were associated (P = 0.0004) with "triple' wild-type status in KRAS, BRAF, and PIK3CA exon 20. Most cell lines were resistant to the direct effects of anti-ERBB2 Mabs, suggesting that the effects of lapatinib might mainly be through ERBB1. Microarray mRNA expression profiles of sensitive and resistant cell lines showed that although ERBB1 receptor or ligand levels did not associate with cetuximab sensitivity, high levels of ERBB2 (P = 0.036) and amphiregulin (P = 0.026) predicted sensitivity to lapatinib. However, higher ERBB1 expression predicted susceptibility to cetuximab-induced antibody-dependent cellular cytotoxicity and occurred independently of KRAS/BRAF/PIK3CA mutations (P = 0.69). Lapatinib may be an effective alternative therapy to cetuximab in triple wild-type tumors. Microarray analysis provides suggestive biomarkers for resistance. ERBB1 levels, independent of mutation status, predict immune killing. Therefore, anti-ERBB1 antibodies may be considered in CRC tumors with higher ERBB1 expression and favorable FcγR polymorphisms.

Replication error deficient and proficient colorectal cancer gene expression differences caused by 3'UTR polyT sequence deletions.

Replication error deficient (RER+) colorectal cancers are a distinct subset of colorectal cancers, characterized by inactivation of the DNA mismatch repair system. These cancers are typically pseudodiploid, accumulate mutations in repetitive sequences as a result of their mismatch repair deficiency, and have distinct pathologies. Regulatory sequences controlling all aspects of mRNA processing, especially including message stability, are found in the 3'UTR sequence of most genes. The relevant sequences are typically A/U-rich elements or U repeats. Microarray analysis of 14 RER+ (deficient) and 16 RER- (proficient) colorectal cancer cell lines confirms a striking difference in expression profiles. Analysis of the incidence of mononucleotide repeat sequences in the 3'UTRs, 5'UTRs, and coding sequences of those genes most differentially expressed in RER+ versus RER- cell lines has shown that much of this differential expression can be explained by the occurrence of a massive enrichment of genes with 3'UTR T repeats longer than 11 base pairs in the most differentially expressed genes. This enrichment was confirmed by analysis of two published consensus sets of RER differentially expressed probesets for a large number of primary colorectal cancers. Sequence analysis of the 3'UTRs of a selection of the most differentially expressed genes shows that they all contain deletions in these repeats in all RER+ cell lines studied. These data strongly imply that deregulation of mRNA stability through accumulation of mutations in repetitive regulatory 3'UTR sequences underlies the striking difference in expression profiles between RER+ and RER- colorectal cancers.

Cancer stem cells from colorectal cancer-derived cell lines.

Cancer stem cells (CSCs) are the subpopulation of cells within a tumor that can self-renew, differentiate into multiple lineages, and drive tumor growth. Here we describe a two-pronged approach for the identification and characterization of CSCs from colorectal cancer cell lines, using a Matrigel-based differentiation assay, and cell surface markers CD44 and CD24. About 20 to 30% of cells from the SW1222 cell line form megacolonies in Matrigel that have complex 3D structures resembling colonic crypts. The megacolonies' capacity to self-renew in vitro is direct evidence that they contain the CSCs. Furthermore, just 200 cells from SW1222 megacolonies initiate tumors in NOD/SCID mice. We also showed that CD44(+)CD24(+) cells enriched for colorectal CSCs in the HT29 and SW1222 cell lines, which can self-renew and reform all four CD44/CD24 subpopulations, are the most clonogenic in vitro and can initiate tumors in vivo. A single SW1222 CD44(+)CD24(+) CSC, when grown in Matrigel, can form large megacolonies that differentiate into enterocyte, enteroendocrine, and goblet cell lineages. The HCT116 line does not differentiate or express CDX1, nor does it contain subpopulations of cells with greater tumor-forming capacity, suggesting that HCT116 contains mainly CSCs. However, forced expression of CDX1 in HCT116 leads to reduced clonogenicity and production of differentiating crypt-containing colonies, which can explain the selection for reduced CDX1 expression in many colorectal cancers. In summary, colorectal cancer cell lines contain subpopulations of CSCs, characterized by their cell surface markers and colony morphology, which can self-renew and differentiate into multiple lineages.

Gastrointestinal differentiation marker Cytokeratin 20 is regulated by homeobox gene CDX1.

CDX1 is a transcription factor that plays a key role in intestinal development and differentiation. However, the downstream targets of CDX1 are less well defined than those of its close homologue, CDX2. We report here the identification of downstream targets of CDX1 using microarray gene-expression analysis and other approaches. Keratin 20 (KRT20), a member of the intermediate filament and a well-known marker of intestinal differentiation, was initially identified as one of the genes likely to be directly regulated by CDX1. CDX1 and KRT20 mRNA expression were significantly correlated in a panel of 38 colorectal cancer cell lines. Deletion and mutation analysis of the KRT20 promoter showed that the minimum regulatory region for the control of KRT20 expression by CDX1 is within 246 bp upstream of the KRT20 transcription start site. ChIP analysis confirmed that CDX1 binds to the predicted CDX elements in this region of the KRT20 promoter in vivo. In addition, immunohistochemistry showed expression of CDX1 parallels that of KRT20 in the normal crypt, which further supports their close relationship. In summary, our observations strongly imply that KRT20 is directly regulated by CDX1, and therefore suggest a role for CDX1 in maintaining differentiation in intestinal epithelial cells. Because a key feature of the development of a cancer is an unbalanced program of proliferation and differentiation, dysregulation of CDX1 may be an advantage for the development of a colorectal carcinoma. This could, therefore, explain the relatively frequent down regulation of CDX1 in colorectal carcinomas by hypermethylation.

Mutations in the AXIN1 gene in advanced prostate cancer.

BACKGROUND: The Wnt signalling pathway directs aspects of embryogenesis and is thought to contribute to maintenance of certain stem cell populations. Disruption of the pathway has been observed in many different tumour types. In bowel, stomach, and endometrial cancer, this is usually due to mutation of genes encoding Wnt pathway components APC or beta-catenin. Such mutations are rare in hepatocellular carcinomas and medulloblastomas with Wnt pathway dysfunction, and there, mutation in genes for other Wnt molecules, such as Axin, is more frequently found. OBJECTIVE: Although evidence of abnormal activation of the Wnt pathway in prostate cancer has been demonstrated by several groups, APC and beta-catenin mutations are infrequent. We sought mutations in genes encoding Wnt pathway participants in a panel of prostate cancer clinical specimens and cell lines. DESIGN, SETTING, AND PARTICIPANTS: DNA was obtained from 49 advanced prostate cancer specimens using laser microdissection followed by whole genome amplification and 8 prostate cancer cell lines. MEASUREMENTS: The DNA samples were screened for mutations in the genes encoding APC, beta-catenin, and Axin. The subcellular distribution of beta-catenin expression was assessed in the clinical specimens using immunohistochemistry. RESULTS AND LIMITATIONS: Abnormal patterns of beta-catenin expression, suggesting Wnt pathway dysregulation, were observed in 71% of specimens. One APC mutation, two beta-catenin gene mutations, and 7 DNA sequence variations in the Axin gene were detected. Four different Axin polymorphisms were also found in the cell lines. The study does not provide definite evidence that the observed sequence changes alter protein function, promoting neoplasia, but the potential functional relevance of these variants is discussed. CONCLUSIONS: These data contribute to our understanding of the role of Wnt dysregulation in prostatic tumourigenesis and support the current interest in the pathway as a therapeutic target. Of particular interest, we identified three new potentially functionally relevant AXIN1 mutations.

Multiple rare variants in different genes account for multifactorial inherited susceptibility to colorectal adenomas.

Clear-cut inherited Mendelian traits, such as familial adenomatous polyposis or hereditary nonpolyposis colorectal cancer, account for <4% of colorectal cancers. Another 20% of all colorectal cancers are thought to occur in individuals with a significant inherited multifactorial susceptibility to colorectal cancer that is not obviously familial. Incompletely penetrant, comparatively rare missense variants in the adenomatous polyposis coli gene, which is responsible for familial adenomatous polyposis, have been described in patients with multiple colorectal adenomas. These variants represent a category of variation that has been suggested, quite generally, to account for a substantial fraction of such multifactorial inherited susceptibility. The aim of this study was to explore this rare variant hypothesis for multifactorial inheritance by using multiple colorectal adenomas as the model. Patients with multiple adenomas were screened for germ-line variants in a panel of candidate genes. Germ-line DNA was obtained from 124 patients with between 3 and 100 histologically proven synchronous or metachronous adenomatous polyps. All patients were tested for the adenomatous polyposis coli variants I1307K and E1317Q, and variants were also sought in AXIN1 (axin), CTNNB1 (beta-catenin), and the mismatch repair genes hMLH1 and hMSH2. The control group consisted of 483 random controls. Thirty of 124 (24.9%) patients carried potentially pathogenic germ-line variants as compared with 55 ( approximately 12%) of the controls. This overall difference is highly significant, suggesting that many rare variants collectively contribute to the inherited susceptibility to colorectal adenomas.

Genetics of colorectal cancer: hereditary aspects and overview of colorectal tumorigenesis.

Familial adenomatous polyposis and hereditary non-polyposis colorectal cancer are dominantly inherited conditions with 100% and 80% life-time risk of developing colorectal cancer, respectively. The genetic mutations responsible for these two conditions lie in the adenomatous polyposis coli (APC) and mismatch repair genes. These same genes also play a key role in the formation of sporadic colorectal cancers, which arise on a background of a similar spectrum of mutations to the hereditary cancers. This article examines the genetic mechanisms underlying the hereditary colorectal cancers, as well as genetic predisposition to colorectal cancer in the general population in the absence of a clear-cut genetic syndrome. Colorectal cancer arises as the cumulative effect of multiple mutations within the cell, allowing it to escape growth and regulatory control mechanisms. This step-wise progression of mutations facilitates the histological transition from normal mucosa to adenoma to carcinoma. The latter part of this paper focuses on the key genetic events underlying this process and provides an overview of the genetic mechanisms responsible for colorectal tumorigenesis.