Integrated clinical genomics: new horizon for diagnostic and biomarker discoveries in cancer

Keywords::

• breast cancer
• colorectal cancer
• copy number changes
• genome sequencing
• heterogeneity
• integrated analysis
• mutations
• transcriptomics

In the second post-genomic decade, the need to move from the ‘one-size-fits-all’ notion into personalized medicine is well recognised. However, the discovery of robust molecular diagnostics and predictive biomarkers to achieve this clinical goal is proving to be a highly complex task [1]. Over the past decade multiple studies using high-throughput technologies (HT) such as microarrays and PCR arrays have led to the development of gene-expression signatures. A growing number of such profiling signature tests, including Oncotype DX, MammaPrint and ColoPrint, have been commercially available as tests for more individualized treatment decisions in breast cancer and colorectal cancer (CRC). Recently, efforts are underway for the development of microRNA (miRNA) profiling signatures and their validation as tests for clinical use. However, abundant skepticism for the translational power of these simplified profiling signatures in the clinic has led to the current effort for more complex genomic discoveries based on genome sequencing and biological systems regulating gene-expression patterns.

Now, for the first time, genomic studies report integrated analysis data of biological samples from patients with breast cancer [2–7] and CRC [8,9] applying a variety of next-generation sequencing technologies [10,11] coupled with advanced microarray platforms. The goal of these integrated studies synthesizing genome sequencing and transcriptomics data is to provide a more comprehensive picture of gene-expression regulation enabling more robust pathway analysis. Signaling transduction pathways input and outputs govern cell behavior and outcome, and thus their analysis is crucial for biotechnology, pharmaceutical industry and clinical medicine. After analyzing clinical genome landscape-based data [2–9] and considering the complexity of gene-expression regulatory networks suggested by the ENCODE project [12–14], we discuss the perspectives and challenges of large-scale genome architecture-based information for the discovery of the next-generation diagnostics, predictive biomarkers and drug targets.

Cancer advances & limitations

Breast cancer and CRC are two of the most common cancer types with an estimated 2.4 million new patients collectively each year. Mortality rates, particularly in advanced stages, remain very high. Despite considerable basic and translational research progress, a small proportion ultimately reaches clinical implementation, revealing the extreme complexity of disease and the tremendous gap between labor and patient.

Personalized cancer medicine has been widely accepted as a research priority and essential strategy to improve oncological outcomes. However, progress in the clinic has been slow. It has been clear that we should move from the traditional and unrealistic ‘one-size-fits-all’ notion to personal genomics and biomarkers for predicting cancer risk and population risk stratification for personalized primary prevention. In patients, the primary goal is to tailor the best drug – or more often combination of drugs – with the higher probability for response and improved survival. However, this goal has proven to be extremely hard despite advances with the human genome sequence and completion of many genome-wide association studies.

High-throughput technologies

Inherited variants and somatic alterations deregulate gene expression and signaling pathways controlling cell growth, survival and apoptosis. The wide heterogeneity in both mutational landscape and gene-expression patterns among individual patients with the same cancer type and stage has guided biomedical research into driver mutation identification and gene-expression characterization in cancer. Over the past decade, the first generation of biomarkers was based on microarrays and PCR arrays. Gene-expression signature tests were based on differentially expressed genes measurements. Correlating this differential gene expression with clinical outcome, the multigene assays Oncotype DX (21 genes) and MammaPrint (70 genes) signatures for breast cancer and ColoPrint for CRC have been developed and become commercially available. However, there has been increasing skepticism for their clinical utility as prognostic and predictive tools. Now, beyond simple array-based gene expression and miRNA profiling data, integrated analysis is used to obtain and synthesize information from genome-wide mutations, copy number aberrations (CNAs) and transcriptome data. It is thought that such a comprehensive view of biological processes driving gene expression regulation is essential to assess critical molecular mechanisms underlying cancer. This is a more rational, deeper approach into molecular mechanisms driving genomes and cell function, and thus it may achieve the grand challenge of robust genome annotation-based development of molecular diagnostics, biomarkers and therapeutic targets.

Deep sequencing & integrated analysis

Innovation is needed for the discovery of novel molecular diagnostics and robust biomarkers for predicting prognosis, tumor responsiveness to therapy and recurrence risk stratification before therapy initiation. Despite a plethora of published reports on genetic discoveries, only a few molecular tests have been established in the clinic for routine treatment decision. These include the estrogen-receptor (ER) and HER2 status for breast cancer and microsatellite instability for CRC [15,16,101]. With accurate data, rapidly dropping costs and fast performance of genome sequencing and integrated analysis, the translation of more complex genomic discoveries in the clinic is now evaluated.

Breast cancer landscape

Recently, six genomic studies on breast cancer using HT have been reported. In three papers, whole-exome sequencing (WES) or both whole-genome sequencing (WGS) and WES respectively on biological samples from 100 patients were performed [2]. In these studies WGS was performed in 46 and 22 breast cancer cases and WES in 31 and 104 cases, respectively [3,4]. Genomic heterogeneity was the dominant finding in these studies. Stephens et al. reported a 73% diversity rate of mutational landscapes [2]. Of 40 cancer genes identified as harboring somatic driver point mutations and/or copy number changes, each individual cancer had a maximum number of six mutated cancer genes involved in tumorigenesis. Ellis et al. analyzed WGS and WES data from 77 pretreatment tumor biopsies from ER-positive breast cancer patients who underwent neoadjuvant therapy with anastrozole inhibitors (AIs) [3]. In this study, 18 significantly mutated genes were identified, including CBFB and RUNX1, which are also involved in triple-negative breast cancer (TNBC) [5]. Because of extreme heterogeneity in ER-positive subtype, the study was underpowered to identify predictors of endocrine therapy response. Banerji et al. have focused on the evaluation of genomic rearrangements in breast cancer patients of diverse subtypes [4]. They have discovered and report for the first time recurrent mutations in the CBFB transcription factor gene and deletions of its partner RUNX1. In addition, the recurrent translocation MAGI3–AKT3 found in TNBC may have clinical implications: the genomic MAGI3–AKT3 fusion shows AKT3 as a potential therapeutic target.

The other three studies performed integrated deep-sequencing analysis [5,6] or dynamics of transcription factor binding events [7]. Shah et al. performed integrated analysis of samples from 104 primary TNBC tumors [5]. The investigators profiled the genomes of these patients with Affymetrix^® SNP 6.0 (Affymetrix, CA USA) arrays (all cases), RNA-seq (80 cases; Illumina GAII, CA USA), and WES/WGS (65 cases; tumor and normal DNA). This study gives emphasis to the clonal frequencies at the time of diagnosis that is, according to the authors, essential for understanding the biology and therapeutic responses of patients with TNBC.

CNAs & transcriptome

Analysis of transcriptomics and copy number changes can provide a comprehensive view of gene-expression regulation, enabling the discovery of robust biomarkers and drug targets. Such an integrated analysis was performed by Curtis et al. [6]. Using Affymetrix SNP 6.0 and Illumina HT-12 v3 platforms, the authors looked at genomic and transcriptional profiling, respectively, in samples from 2000 breast cancer patients. The investigators found that the total number of inherited copy number variants, SNPs and acquired somatic CNAs was associated with expression in approximately 40% of genes. Interestingly, the researchers found that the landscape was dominated by cis- and trans-acting CNAs. An analysis of paired DNA–RNA profiles revealed a high-risk, ER-positive 11q13/14 cis-acting subgroup and a favorable prognosis subgroup devoid of CNAs. These data enhance further extensive evaluation of the relationship between CNAs and transcription factor-binding events because it may lead to a novel transcriptome-based molecular classification of breast cancer. Ross-Innes et al. performed a transcriptome analysis using chromatin immunoprecipitation (ChIP) sequencing (ChIP-seq) on a genome-wide scale [7]. This genome-wide mapping of ER-binding events in primary and metastatic frozen tissues samples from eight ER/PR-positive patients was associated with distinct clinical outcomes. Therefore, ChIP-seq transcriptome analysis in clinical samples may provide exciting perspectives for identifying robust biomarkers for predicting response or resistance to endocrine therapy in ER-positive disease.

Colorectal cancer genomic landscape

Similarly to breast cancer, limitations also exist for personalized treatment decisions in patients with CRC. Genome sequencing and integrative omics analysis in more than 276 and 70 samples, respectively, from CRC patients have recently been reported [8,9]. In the Cancer Genome Atlas project a total of 24 mutated genes were identified and in addition to the well-known APC, TP53, SMAD4, PIK3CA and KRAS mutations, mutations were also found in ARID1A, SOX9 and FAM123B genes. The study revealed that CNAs and amplification of HER2 justify the evaluation of the anti-HER2 antibody trastuzumab in HER2-positive CRC patients. The study confirmed the crucial role of the Wnt-signaling pathway that was found to be deregulated in 93% of CRC and of the MYC gene in CRC. The second study [9] was performed WES, RNA-seq, CNAs and transcriptomics and found, in addition to expected mutated genes, several other new recurrent mutations involved in Wnt3 signaling pathway genes and multiple gene fusions. Seshagiri et al. emphasize the R-spondin gene fusions as potential therapeutic targets and biomarkers for identifying approximately 10% of colon cancer patients with these gene fusions [9].

Future perspectives & conclusion

As genome sciences and technology are rapidly being evolved, new efforts for using more complex genomics are emerging to advance genomic medicine and improve human health. Moving from array-based methods to analyze gene expression to deep sequence and now to integrate genomic analysis we dramatically increase the chance to discover robust and reliable genomic tools to guide personalized treatment decisions and improve oncological outcomes. The first published reports using a variety of HT, such as next-generation sequencing and modern microarrays in clinical samples, provide new exciting perspectives [5–9]. With deep sequencing, transcriptome and gene copy number analysis of patients’ samples, a more comprehensive view of the dysfunction of genes and genomes of individual tumors is provided. This sophisticated deeper insight into structural and functional genome and transcriptome architecture raises more rational opportunities to translate these complicated genomic discoveries into clinic use.

However, the road to reach medical practice from genome sciences is too long and multiple challenges are now revealed by the ENCODE project. With the ambitious goal of mapping the functional elements of the human genome, important advances in understanding the function of genes and the genome, beyond simple sequence, have been made. At the same time these results from 30 recent papers reveal an extreme complexity of interacting biological systems regulating gene expression and genome function. Approximately 80% of the genome is functional, in contrast to the notion of junk noncoding DNA. Transcriptional regulation is a highly complex and dynamic networking process. However, the project has completed the evaluation of only 100 out of a total of approximately 1500 transcription factors and approximately 20 other DNA-binding proteins, such as histones. The principles of dynamic regulatory networks orchestrating gene expression through DNA-binding events, epigenomic modifications and noncoding RNAs are so far poorly understood [12–14]. Although it is now unknown whether such a deep knowledge of structural and functional genome architecture is a precondition to achieve massive clinical success, small steps towards the translation of current and merging integrated genomic analysis and network biology-based approaches to reach genomic network medicine can be made [15–20].

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.