NOTE: This assay is the introduction to my research vision I wrote five years ago but that did not make into the programme grant we wrote. I think this is still current and, as it is unlikely I will publish this text, I am releasing it in the public domain with very little editing. I should note, particularly, that some paragraph remains unreferenced.
The genome, biochemical networks and phenotypes | Somatic mutations and gene copy number variations (CNVs) accumulate over time, stochastically altering the abundance and the functions of gene products. At first glance, efforts to identify and characterize somatic mutations provided a comparatively simple model: a few hundreds genes (proto-oncogenes and tumour suppressor genes) are often mutated contributing mechanistically to tumourigenesis (driver mutations). Some driver mutations are very frequent, but driver mutations that are less frequent in a cancer type overall dominate in number within an individual tumour, presumably conferring a more subtle growth advantage than others taken individually [1,2]. Also CNVs are very common in cancer. Sometimes, a clear role of CNVs in tumourigenesis can be established; most of the times, however, the effects of CNVs are difficult to predict or characterize because of the very different possible dependencies between phenotype and concentration of a gene products (e.g., haplo-insufficiency, quasi-sufficiency, triplo-sufficiency, etc.) [3,4]. Concentration effects and “subtle driver mutations” complicate the interpretation of genomic studies and may be best described by a continuum model for tumorogenesis where the all-or-none effects of individual genomic alterations are the frequent exception rather than the rule [4,5]. Notwithstanding the invaluable insights that genomics studies have provided and will continue to provide in our understanding of cancer, diagnostics and therapy, the role of these genomic alterations in tumorogenesis will be better understood in the context of the alteration of molecular networks underlying the respective cancer-associated phenotypes [1,6].
Few phenotypes are selected by mutation, those that enable cancer evolution [7] by increasing clonal heterogeneity (by genetic mutation, aneuploidy or epigenetic instability) and that permit growing in a hostile environment (avoidance of immunosurvaillance, metabolic deregulation and stromal hijacking). Moreover, cell survival, cell fate determination and, later in cancer evolution, cell migration are the key phenotypes that make of cancer the devastating diseases it is. Genomic alterations select for these phenotypes by influencing a comparatively small number of biochemical networks. Indeed, cancer-associated somatic mutations cluster in pathways controlling cell-cycle or cell-death, RAS/PI3K/MAPK, TGFβ, APC, STAT, NOTCH, WNT, HH and mTor [1,4,7]. Unsurprisingly, somatic evolution of cancer reshapes a comparatively small number of biochemical pathways that control cellular and tissue homeostasis to offset, often in a subtle manner, the net proliferative rate of cells. The study of these pathways is no less daunting than the understanding of complex genomic alterations. However, biochemical networks have evolved to exhibit robustness in the presence of intrinsic noise present in biological systems (e.g. stochastic variations in transcription or cytokines concentrations). Robustness of biochemical pathways permit to stably encode for cellular functions and cellular states. It is therefore conceivable that the myriads of possible genomic alterations and individual gene-products simply concur to generate a discrete set of biochemical states corresponding to cancer-associated phenotypes.
Other “big data disciplines” (e.g., transcriptomics, proteomics and metabolomics) have provided the opportunity to study the working mechanisms of biological systems alongside genomics. Some groups have suggested that integrative biology [8,9], the effort to integrate data from these various disciplines, may permit avoiding the biases and inherent flows of individual –omics techniques and, at the same time, may deliver a new approach to the study of human disease. This approach is summarized by the term “network medicine” highlighting that molecular networks altered in disease can be both the target for future therapeutic strategies and the possible source of novel biomarkers. A biochemical network, common to many different cell types or even species, exhibit a different “network utilization” in different physiological and pathological contexts. Mutations can therefore offset the utilization of molecular networks and their dynamics. On the one hand, better understanding of how networks encode functional states and cellular decisions under physiological conditions and how these are altered in disease will offer more and better targeted therapeutic opportunities. On the other hand, defining cancer-associated network utilizations and engineering tools (probes and instrumentation) to reveal them will provide fundamental insights to optimize patient stratification for improved theranostics and prognostics.
Heterogeneity, causality and phenotypes | Phenotypic heterogeneity, including genetic and epigenetic polymorphism, and polyphenism, is at the basis of both unicellular and complex lifeforms. These three levels of phenotypic heterogeneity are recapitulated in cancer and constitute often insurmountable obstacles to effective therapeutic intervention. Intra-tumour heterogeneity, either within the primary tumour, within a metastasis or between different metastatic foci is indeed the primary cause for the emergence of drug resistance and tumour relapse. The genetic basis for phenotypic heterogeneity within a tumour is rather established. However, other non-genetic factors can be regarded as equally important.
For instance, upon treatment, a fraction of tumour often exhibit drug resistance. In part, this can be caused by pre-existing tumour cell clones carrying mutations that, by chance, will confer resistance to any given drug. Alternatively, this may be caused by tumour initiating cancer cells, stem-like cells that are usually quiescent, less vulnerable to treatment and that can regenerate the tumour upon termination of the therapy. Moreover, non-Darwinian mechanisms for the emergence of drug resistance have been proposed as well, whereby cells trigger a transient drug-resistant phenotype that, in time, can be then converted to a stable inheritable state by subsequent somatic evolution. Fractional killing may also be explained by non-genetic heterogeneity. For instance, Spencer et al. have shown that in a clonal population of cells, TRAIL elicits a heterogeneous phenotypic response with cells undergoing apoptosis at different times or surviving indefinitely. The authors elegantly demonstrate that this phenomenon is caused by stochastic variations in the abundances of the many proteins involved in the apoptotic molecular network.
Genomics, transcriptomics, proteomics and metabolomics allow the characterization of tens of thousands of biomolecules at the same time. Furthermore, the increasing sensitivity of these techniques provides – or may provide in the future – single cell “–omics” characterization. However, the invasiveness of these techniques will limit their applications to the study of individual time points. Thus, causality can be established only by inference. Techniques capable to provide low invasiveness and biochemical information on living cells are thus extremely useful to complement models derived by ‑omics techniques and to provide a tool for testing hypothesis derived from analysis of big data.
It is thus evident that time-lapse imaging of individual living cells with biochemical information is strategic for the understanding of the heterogeneous response of biological systems and to establish causality between biochemical events and cellular decisions. At the same time, genetic heterogeneity within a tumour and between tumours induces differences in network utilizations with significant consequences for prognosis and treatment. Also in this context, biochemical imaging techniques are necessary to understand the phenotypic heterogeneity of a tumour and, at the same time, may be useful to define network-based biomarkers.
The next generation of Systems Biology | Several groups have identified the need to integrate fluorescence microscopy in the systems level study of the cell and organisms [10-15]. The term “Systems Microscopy” has been suggested for the description of microscopy tools applied to this field [14]. In order to strategically complement other approaches, Systems Microscopy has to deliver single cell resolution, temporal characterization of living cells and high quality quantitative data and has to be applied to the most appropriate biological context (e.g., for epithelial cancers, adherent 2D, 3D, organotypic cultures or in vivo rather than in suspension or cellular homogenates) [10]. Whereas –omic techniques can sample the biological space over the fullness of biochemical moieties (genes, RNAs, proteins, metabolites) albeit with poor sampling of individual cellular behaviours and spatio-temporal organization, Systems Microscopy samples the fullness of the spatio-temporal organization of molecular networks but reports about a limited number of gene products or biochemical events [10]. Therefore, Systems Microscopy elegantly complements big data studies.
We envisage two (not mutually exclusive) approaches to Systems Microscopy: high throughput screening platforms and single cell biochemical multiplexing. High Content Screening (HCS, also known as imaging cytometry or high throughput imaging) is the current tool of choice for Systems Microscopy. Relying on robotics, automation and unsupervised or semi-supervised data analysis, HCS enables the screening of large numbers of cellular perturbations (e.g., siRNA or compound libraries) with commercial instrumentation making the correlation of these perturbations with morphological estimators and fluorescent markers possible. Several groups have also highlighted the importance of integrating quantitative biophysical imaging techniques such as Fluorescence Correlation Spectroscopy (FCS) and Foerster Resonance Energy Transfer (FRET) in Systems Microscopy in order to deliver data of high quality. Despite this, HCS has been integrated with these techniques only in a few academic-based efforts [16-19]. HCS expands the sampling of biological space of imaging technologies to deliver another set of “big data” but with single cell resolution.
We are pursuing a different approach to Systems Microscopy that maps in space and time an increasing number of fluorescent markers within the living cell. Fluorescence is not amenable to the simultaneous detection of many fluorescent molecules because of the broad excitation and emission spectra of common fluorophores. Therefore, we are determined to develop new techniques (bioprobes and instruments) that exploit all properties of light (photon arrival times, colour and polarization) efficiently to maximize the biochemical resolving power of microscopy. We aim to monitor nodes of molecular networks (e.g., quantifying the dynamic phosphorylation of several substrates) in living cells in response to stimuli, discerning between physiological and pathological (oncogene-driven) network behaviour (topology). The integration of Optogenetics tools (e.g., light-inducible oncogenic signalling) enables perturbational analysis of biochemical networks and facilitates the execution of complex biochemical imaging assays fully automated with no requirement for sample manipulation other than by light. Therefore, these techniques will be strategic for the study of biological networks at low throughput with high quality data; thanks to this all-optical approach, they may also be integrated with HCS increasing the quality and quantity of information and decreasing steps in chemical manipulations of the samples (e.g., addition of doxycycline to stimulate the expression of a gene)
[1] Vogelstein et al. (2013) “Cancer Genome Landscapes” Science
[2] Wood et al. (2007) “The genomic landscapes of human breast and colorectal cancers” Science
[3] Solimini et al. (2012) “Recurrent hemizygous deletions in cancers may optimize proliferative potential.” Science
[4] Davoli et al. (2013) “Cumulative Haploinsufficiency and Triplosensitivity Drive Aneuploidy Patterns to Shape the Cancer Genome” Cell
[5] Berger et al. (2011) “A continuum model for tumour suppression.” Nature
[6] Jorgensen&Linding (2010) “Simplistic pathways or complex networks?” Current Opinions in Genetics and Development
[7] Hanahan&Weinberg (2010) “Hallmarks of Cancer: The Next Generation” Cell
[8] Erler&Linding (2010) “Network-based drug and biomarkers” J Pathology
[9] Barbasi et al (2011) “Network Medicine: A Network-based Approach to Human Disease” Nat Rev Genet
[10] Megason et al. (2007) “Imaging in Systems Biology” Cell
[11] Verveer&Bastiaens (2008) “Quantitative microscopy and systems biology” Histochem Cell Biol
[12] Ankers et al. (2008) “Spatio-temporal protein dynamics in single living cells” Current Opinion in Biotechnology
[13] Pepperkok&Ellenberg (2006) “High-throughput fluorescence microscopy for systems biology” NAt Rev Mol Cell Biol
[14] Lock&Stromblad (2010) “Systems microscopy: An emerging strategy for the life sciences” Exp Cell Res
[15] Conrad&Gerlich (2009) “Automated microscopy for high-content
RNAi screening” J Cell Biol
[16] Esposito et al. (2007) “Unsupervised fluorescence lifetime imaging microscopy for high content and high throughput screening” Mol Cell Proteomics
[17] Mathews et al. (2008) “A high-content screening platform utilizing polarization anisotropy and FLIM microscopy” SPIE BIOS
[18] Talbot et al. (2008) “High speed unsupervised fluorescence lifetime imaging confocal multiwell plate reader for high content analysis” J Biophotonics
[19] Barber et al. (2013) “The Gray Institute ‘open’ high‐content, fluorescence lifetime microscopes” J Microsc