Disrupting cancer cells’ biocircuits with interactome-based drugs: is ‘clinical’ innovation realistic?

Keywords::

• drug
• interactome
• molecular network
• protein–protein interaction
• signalling pathway

Despite expectations, abundant scientific and clinical evidence now reveals that current reductionist science and single pathway-based cancer-targeted drugs provide isolated clinical success. Overall, efficacy is limited with temporary antitumor activity [1]. Can innovation in technology, systems science and genome annotation surpass daunting hurdles for understanding dynamic networks of interacting biological systems to overcome drug resistance? Such a network biology approach paves the way for disrupting signaling biocircuits in cancer cells through next-generation drugs and biomarkers discovery.

Network medicine is the most promising emerging approach to cure incurable diseases such as cancer and to overcome unresolved problems such as longevity [2,3]. To reach this future goal, not only is the understanding of complex dynamic signaling transduction networks that drive crucial cellular processes, including cell survival, growth and apoptosis at molecular level required [4,5], but also the exploration and possible prediction of the dynamic spatiotemporal personal genome–lifestyle interaction networks at the whole organism level [6].

Breakthroughs in the high-throughput technologies based on sequencing and arrays methods shape now the high hope of the clinical genome era. Shifting from reductionist biology and genetics to systems medicine by adopting omics data to the clinic is increasingly recognized as a clinical necessity. This effort requires collaboration of both innovators with a clinical orientation and physicians with a translational background and research perspectives [2–6].

Despite major science and technology advances, the crucial debate in the real clinical world is whether the myriad of challenges in understanding and predicting dynamic interacting biological systems, a cell’s decisions to survive or die, and tissue and organ homeostasis can be overcome. This understanding appears to be an essential step for the discovery of signaling network-based drugs and biomarkers to prevent or cure complex multifactorial disorders [6,7].

Although there is scepticism over whether the analysis of the interactome – the whole set of gene, protein and molecule interaction networks driving cell behavior – and discovery of drugs targeting interactome global dysfunction in cancer are realistic goals, current advances in science and technology provide a fascinating strategy for future success.

The goal of ‘clinical innovation’, defined here as a revolution in clinical practice for improving healthcare and healthy longevity [3,7], and curing cancer [1,8], in the second postgenome decade has not yet been achieved despite initial enthusiasm. Genome-wide association using arrays with tens of thousands of SNPs have identified several thousand rare or common genomic variants associated with phenotypes or various diseases. However, the clinical application of these significant but weak associations for predicting personalized risks for disease development is limited, a finding that is known as missing heritability. The power of next-generation sequencing (NGS) technologies is to identify the overall 3 million variants across the human genome that are involved in phenotypes and disease at low cost, accurately and within a short time frame to explain the explosion and popularity of sequencing technologies. Indeed, whole-genome sequencing (WGS) and whole-exome sequencing (WES) bring the hope for more accurate disease risk prediction and the better clinical utility of genome screening for overcoming genome-wide association-based assessment of missing heritability and variations with small disease effects. However, a more recent statistical model and mathematical framework study has revealed limitations of WGS tests for clinical prediction of common multifactorial diseases such as cancer [9]. As a personal genome is unique to each individual, patients also differ from each other with regard to their genome structure and function despite the fact that they have the same clinicopathologic features according to traditional medicine [10,11].

As major complex diseases, including cancer, diabetes neurodegenerative and other disorders, arise from the abundance not only of genetic mutations but also epigenetic alterations and deregulation of signaling pathways networks, gene expression patterns and cellular processes [6], latest efforts include more sophisticated approaches. For example, for cancer diagnostics and therapeutics, and for understanding diabetes molecular mechanisms, integrative analysis of personal genome profile, including WGS/WES, transcriptomics and proteomics data, can be used as we enter into the clinical genome and personalized medicine era [12,13]. However, with little appreciation of homogenous, accurate and detailed data on personal medical and family history, the strategy only focussed on omics data will continue to have limited ‘clinical innovation’ success. For example, clinical evidence shows good prognosis and high cure rates for early-stage cancer, but disease diagnosed at an advanced stage remains incurable, killing millions of patients [8]. These clinical results suggest a limited metastatic capacity of primary tumor cancer cells at stage I or II with mutations and simple protein–protein interactions (PPIs) and signaling pathways sensitive to surgery alone or surgery plus systemic therapy. By contrast, recent studies have revealed acquired mutations in cell subpopulations and complex signaling networks resistant to available therapies for more advances disease.

Overcoming resistance

Both patient-to-patient tumor diversity [14] and even an individual patient’s intratumor heterogeneity [15] and complexity in cellular signaling transduction through physical and functional interactions [5] explain the limitations of current therapeutic drugs and the need to move away from the one-size-fits-all principle in the pharmaceutical industry. Analysis of signaling pathways [16] and computational strategy-based systems medicine [2,3,6] to predict biological networks and drug targets [17] shape innovative approaches to overcome high rates of resistance [18], treatment failure and death [1].

Indeed, more than a decade after the clinical adoption of cancer-targeted drugs, resistance was revealed to be the grand challenge [1]. From approximately 12 monoclonal antibodies and tyrosine kinase inhibitors approved by the US FDA and 200 in preclinical development and clinical staging in the treatment of various cancer types, only trastuzumab, imatinib and vemurafenib have demonstrated clinical antitumor activity measured by true overall survival benefit. A common conclusion for these successful drugs is their efficacy only in single-gene defect-based selection of patients; for example, trastuzumab for HER2-positive breast cancer patients, BCR-ABL1 kinase domain mutation-associated resistance to imatinib in patients with chronic myeloid leukemia and vemurafenib for BRAF V600-mutant metastatic melanoma [18]. However, even among these single-gene testing-based selected patients, resistance occurs in nearly all patients is a metastatic setting after a few months duration of clinical response, resulting in disease progression and, ultimately, death. Moreover, for many other targeted drugs, for example cetuximab, an anti-EGFR antibody that was investigated in Phase III randomized trials that selected and enrolled patients based on their KRAS status, the results showed no survival benefit in KRAS wild-type metastatic colorectal cancer as previously suggested. The results of large-scale clinical trials suggest that drugs inhibiting a single pathway by targeting single gene products have either a temporary antitumor effect or no therapeutic clinical response measured by overall survival benefit [18]. This disappointing finding should not be a surprise, if we consider the current simplified approach. This strategy includes identification of single mutations in a candidate cancer gene and then an effort to link this lesion with the expression profile of this gene and targeting its related single activated signaling pathway. However, cancer cells are driven by the abundance of genetic and epigenetic alterations and deregulation of several signaling pathways, as well as dysfunction of complex dynamic signaling transduction interaction networks [17–19].

Biomarker-based combination of available drugs versus new drug discovery

It takes up to 13.5 years and US$ 1.8 billion to bring a new drug from its discovery to animal models and Phase I, II, III and IV clinical trials for testing its safety and efficacy, and ultimately, to the market, [19]. Therefore, considering the cancer complexity-based uncertainty about the clinical success of a new drug, in the current economic crisis a more pragmatic approach for the pharmaceutical industry is to try to identify robust biomarkers for the clinic. Such molecular tools will allow, from the available pool of targeted drugs, the selection of one or a combination of drugs targeting a set of defected genes and activated signaling pathways in a certain patient with cancer. Indeed, this is a realistic approach if we consider the tens of billions dollars already invested in cancer drugs discovery.

Heterogeneity-based drugs

Personalized cancer diagnostics can emerge by sequencing the genomes of cancer cells from a patient’s tumor biopsies with WGS. WES and RNA sequencing. A combination of these and different sequencing platforms can provide high accuracy and validation in the identification of driver mutations, rather than passenger ones, and separation between inherited and somatic mutations. Such validation studies are crucial for effective targeted cancer therapy development. Indeed, a more recent validation study showed that acquired resistance to AC220 (quizartinib), an active investigational inhibitor of FLT3, in acute myeloid leukemia patients, was attributable to FLT3-ITD (internal tandem duplication) mutations [20]. In contrast to earlier studies that reported these mutations as passenger lesions, this new study showed that point mutations at three residues within the kinase domain of FLT3–ITD are driver mutations and thus these causative lesions represent optimal therapeutic targets for patients with acute myeloid leukemia [20].

Intratumor heterogeneity, more recently confirmed by a multiregion sequencing study [15], requires our special appreciation not only because it can explain resistance to current targeted therapy but also represents a basis for specific signaling pathway analysis and drug target selection. Therefore, besides the well-known patient-to-patient heterogeneity among patients with the same organ-specific clinicopathologic tumor characteristics, use of imaging techniques, such as endoscopic ultrasound, computed tomography, MRI and PET, and tumor-node-metastasis (TNM) staging (intratumor heterogeneity-based selection of drugs) can represent an important step towards personalized diagnostics and therapeutics.

Based on the simple concept of exome sequencing of multiple spatially separated primary and metastatic renal cancer samples, Gerlinger et al. found branched mutational evolution with 63–69% of all somatic mutation differing across tumor regions [15]. This wide heterogeneity even within an individual patient’s tumor shapes new personalized approaches by using proteomics, transcriptomics and epigenomics to link this mutational landscape with activated signaling pathways and combinations of drugs to inhibit this defected signal transduction and gene expression [12]. Although this WGS/WES and RNA sequencing mutations landscape-based approach can link mutant genes with signaling pathway inhibitors, there is scepticism as to whether and to what degree personalized medicine will reach its scope for curing cancer. Without considering systems medicine approaches for understanding dynamics of regulatory networks at both the molecular and whole-organism level [6], expectations for discovering highly effective cancer drugs for dramatic long-term durable survival benefit will probably remain elusive.

Beyond science: interactome drug targets

Understanding biological principles driving interacting signaling pathways for biomarkers and drug target discovery is an overambitious goal of future systems pharmacology and medicine [2,17–19]. Although the number of genes and their products involved in each cancer type is much lower than the approximately 23,000 genes and 100,000 proteins of the human genome, there is a need to simplify cancer analysis by grouping the long lists of genes and proteins into pathology (cancer) pathways. Over the last decade, with the advent of throughput technologies, including arrays- and sequencing-based methods, a tremendous quantity of differentially expressed genes and proteins data is being assembled. Individual gene expression is regulated by signaling transduction networks and, ultimately, the expression profile of multiple interacting genes controls cellular process such as survival, growth, proliferation, apoptosis and metastasis, which represent the hallmarks of cancer [1]. Genetic and epigenetic alterations affect gene expression and biological networks in cancer and, therefore, pathway and network analysis is crucial for understanding the biology underlying initiation and metastasis of cancer pathways [5], and for developing novel diagnostics and therapeutics.

Moving from the simplified microarray-based differentially expressed gene profiling [17] to highly complex dynamic interacting biological systems that drive gene expression and signaling pathways [2,5,19], we can pave the way to reach the major goal of cellular behavior prediction leading to effective drug and biomarker development. Unsurprisingly, the clinical utility of microarray-based data is modest if we consider the substantial limitations of pathway analysis methods to provide validated biological results. Over the past decade, three generations of pathway analysis methods have evolved, including the approaches of over-representation analysis, the functional class scoring and the pathway topology-based approaches. Despite these advances using omics data (genomics, proteomics and metabolomics), along with functional annotations (pathway database) as input to virtually all pathway analysis methods, there are still annotation and methodological challenges [16]. These limitations include lack of cell- and disease-specific information, incomplete structural and functional annotation of the human genome in healthy and diseased states, and the inability to integrate the dynamic spatiotemporal changes of biological systems and their interactions in analyses to assess and predict signalling pathways and cell behaviour [16].

Next-generation methods & drugs

To overcome challenges in signaling pathway analysis and dynamic predictive models of gene expression regulatory networks and cell behavior, guided emerging technological and systems science innovation is increasingly being developed. For example, transcriptome analysis with RNA sequencing and chromatin immunoprecipitation followed by WGS is now moving from models to patients’ tumor sample-based transcriptome analysis for breast cancer biomarker development [21]. Epigenome analysis with whole-genome bisulfite sequencing for characterization of epigenetics involved in cancer, such as DNA methylation and DNA-binding histones modifications, can contribute to epigenetic anticancer drugs targeting the whole genome [22] and better understanding of chromatin remodeling and gene expression regulation programs [23]. Besides the transcriptome and epigenome, noncoding RNA, including miRNAs and RNA methylomes also have a fundamental role in the regulation of gene expression [23].

We just recently started to explore the human interactome, including DNA, RNA and molecules networks, in both physical, such as PPI and DNA-binding proteins, and functional (gene–gene interactions) interactions. Using high-performance mass spectrometry to generate liquid chromatography–tandem mass spectrometry phosphoproteome maps can provide quantitative analyses allowing computational strategies to study PPI networks [24], and paving a way for future PPI inhibitors development.

Expert commentary & five-year view

Large-scale, international and individual academia cancer genome projects using high-throughput NGS and arrays-based strategies are currently underway, and will provide deeper insights into cancer genome structural variation and how it affects gene expression and deregulates signaling pathways. In addition, innovation in technologies monitoring signaling transduction interaction networks in living cells using biosensors can provide robust visualized network data [25]. To achieve a personal profile of the activating signaling network, novel techniques analyzing an individual patient’s tumor biopsy-based pathway circuits are required. Such a progress is limited to cell line models, but exciting research has more recently been reported for transcriptome-based biomarkers using chromatin immunoprecipitation followed by WGS technologies in primary and metastatic breast cancer samples [21]. Current NGS-based studies are focussing on the completion of databases with driver genetic and epigenetic alterations for several major cancer types, the advances in understanding the dynamics of molecular elements of the genome and accurate analysis of signaling pathway networks underlying primary and metastatic tumors. The achievement of this goal will require a global assembly of omics, signal transduction networks and clinical data. Simultaneously, progress in bioinformatics, systems computational and mathematical strategies can lead to the development of accurate pathway networks analysis with clinical validity for discovering the next generation of drugs targeting cancer cells. Disrupting the deregulated biocircuits of an individual patient’s tumor, identified by tumor biopsy-based signaling network analysis, shapes a future perspective of pragmatic personalized cancer genomic medicine.

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.