Download Cancer hallmarks through the network lens

Cancer Cell & Microenvironment 2015; 2: e943. doi: 10.14800/ccm.943; © 2015 by Enrico Capobianco
http://www.smartscitech.com/index.php/ccm
RESEARCH HIGHLIGHT
Cancer hallmarks through the network lens
Enrico Capobianco1, 2
1
Center for Computational Science, University of Miami, Miami, FL, USA
2
LISM-IFC, National Research Council of Italy, Pisa, IT
Correspondence: Enrico Capobianco
E-mail: [email protected]
Received: August 01, 2015
Published online: September 25, 2015
Given the role and relevance of the heterogeneity hallmark in cancer, a current need is to advance the
knowledge of the network dynamics building the complex architecture underlying the relationships between cell
sub-populations. This step would shed light over obscure mechanisms such as acquired resistance to drugs, for
instance. Network medicine has established itself as a systems integrative approach that leverages on the
interpretation offered by community-like behavior of bio-entities and clinical variables. Translated over cell
sub-population cells, there is a strong case for deciphering cooperative versus antagonist network dynamics in
order to determine better therapeutic strategies.
Keywords: cancer hallmarks; networks; communities; microenvironment
To cite this article: Enrico Capobianco. Cancer hallmarks through the network lens. Can Cell Microenviron 2015; 2: e943.
doi: 10.14800/ccm.943.
Cancer Hallmarks
Cancer hallmarks [1, 2] lead the cancer field, and stimulate
advances and discoveries, together with renewed interest and
debate over their sufficiency and adequacy. In particular,
immune response [3-6], drug resistance [7-9] and complex
interactions with micro-environment [10, 11] keep generating
influencing ideas, assuming critical roles in oncology
research.
Computational oncology scientists are heavily engaged in
modeling such effort, and particularly through the
experimental omics and discovery pipelines. These pipelines
are currently more focused on empirical data-driven solution
than model-centered ones. While the former follow various
methodological directions, especially integrative ones, the
latter partially lack insight and depth. Reasons for such
limitations are the role of evolving technologies, and the
complexity of the data generation processes. In our view,
these limitations suggest that there is also margin for
improvements, and it is in the networks direction that we
look in particular with this work.
Networks play a well-recognized integrative inference role
in elucidating complex cancer connectivity patterns and
cross-links [12]. In principle, and by operating systems-wise,
the network inference approaches can adapt to the hallmarks
of oncogenic signaling [13-15] and of cell-cell communication
[16]. Further specializations may occur by integrating clinical,
genetics, omics, and imaging data, in a so-called multiplexed
way [17, 18].
Despite all the complexities, an advantage of networks
compared to other techniques is that their structures can be
naturally simplified by partitioning strategies, from which
hierarchies of sub-modules or communities arise and appear
associated to characteristics shared by group members.
Importance
attractors
of
modularity
and
relationships with
Modular configurations in networks (Figure 1) provide
spatiotemporal examination of cellular components, thus
linking molecular pathways to phenotypes especially by
exploiting similarity metrics that underlie the network
topology. Typically, cancer hotspots consider two properties:
Page 1 of 5
Cancer Cell & Microenvironment 2015; 2: e943. doi: 10.14800/ccm.943; © 2015 by Enrico Capobianco
http://www.smartscitech.com/index.php/ccm
Figure 1. Cancer Networks. Adapted from [19], the central rhomboid in this representation involves two properties of cancer
hallmarks that might be translated in a network-centric context. Plasticity reflects the capacity of cancer to adapt under the pressure
of its inherent heterogeneity. Change in the system calls for resilience, leading to stability at some point in order to permit
establishment or re-modulation of phenotypic features. Cancer networks have configurations with several properties, such as: 1)
Redundancy, through hierarchized connectivity structure typical of community; 2) Robustness, to allow structural change in
response to anomalies or partial failures; 3) Heterogeneity, by embedding functional variety into the degree distribution; 4)
Adaptation, through flexibility of topological configurations in response to external stimuli and perturbations. There are
corresponding computational strategies that can be enabled to exploit these properties, from model selection and dimensionality
reduction to contextual enrichment. Cancer hallmarks represent systems forces whose presence and action requires the network to
dynamically evolve. In principle, one could also reverse this view, and consider the host network and its immune mediated
modulation.
hubbiness and centrality. Networks identify as hub nodes
those with relatively high connectivity degree, and as central
nodes or links those measuring high load or traffic, therefore
indicating relevance.
An important premise of many studies is that tumor
processes may reveal quite different topological measures
from processes in normal conditions. Nevertheless, the
possibility of accurately measuring these different features
depends on many factors, which in turn suggests that a
systems approach is needed. Given the complexity of the
model, its reliability is also an element needing careful
evaluation. Complexity and reliability both increase with the
number of features under consideration and with the type of
measurement.
Network modules make a lot of sense biologically and
clinically as they do not act as separated entities, but
cross-linked aggregates. Consequently, differential network
configurations should be identified to reveal the diversified
propagation laws of the effects that ultimately determine the
phenotypes of interest, with reference to cancer hallmarks.
This transition involves changes in ‘attractor’ or stable
equilibrium states [20-23]. The states in a network are
associated to points of an N-dimensional state space,
depending on the number of nodes. Also, the modules of a
network are associated to ‘state communities’ which are
linked, as they assemble fractions of nodes and links in either
Page 2 of 5
Cancer Cell & Microenvironment 2015; 2: e943. doi: 10.14800/ccm.943; © 2015 by Enrico Capobianco
http://www.smartscitech.com/index.php/ccm
transient or persistent communication with each other.
Attractors, as stable points in the expression state space, tell
where the system would return after small perturbations
within a smooth landscape.
The impacts of external cues and also internal perturbation
of microenvironment origin should be accounted for.
Consequently, the treatment at systems scale of phenotypic
response patterns would be recommended. Such
non-stationary patterns can disrupt cancer network attractors,
when they are of systemic nature. Accordingly, these
dynamics drive the network state transitions through adaptive
mechanisms controlling and modulating cascades of events,
including compensatory effects at the pathway level. For
instance, critical aspects of sensitivity and resistance
mechanisms are emerging in clinical work. Inference on
signaling states across tumor phenotypes may follow
perturbation experiments having key modules as targets,
and/or considering component nodes/edges (i.e. edgetic
perturbations [24]). From the observed network configuration
changes, the dynamics of complex signaling pathways can be
inferred, for instance with regard to the effects of their
inhibition with regard to key mechanisms such as drug
resistance.
Endogenous and exogenous network dynamics
The complexity of signaling networks (cell cycle phases,
apoptosis, etc.) is due to non-linearly convoluted dynamics.
These latter depend on both the level and the activation state
of proteins which exert time-course effects on multiple key
pathways. Organizing the signaling data in a mathematical
framework is a complex task. Computational data-driven and
exploratory tools, as said before, remain the main proposal
currently available and preferred in general to theoretical
models. Intuitively, each hallmark refers to potential barriers
that a neoplasia must overcome in order to survive and
proliferate. Of a particular relevance from a clinical
perspective are angiogenesis, and tissue invasion and/or
metastasis formation.
Assessment of angiogenesis requires the consideration of a
spectrum of localized scales co-participating in measuring
the influence exerted by the cell states. Typically, while
models would hold for the entire network, data may refer to
just sampled sub-networks. Whole network parameter
estimates obtained at sub-network scale work under the
assumption of model consistency under sampling. A general
way to reduce the typical complexity of high-dimensional
networks is to turn to ensemble models averaging out
possible erratic aspects. Invasion and metastasis determine
patient survival, and require the acquisition of specific
phenotypes for the cells to be able to detach from the tumor
mass, migrate and survive in a different tissue.
It is now recognized [25] that the development of tumor
networks depends on the interaction of cancer cells with the
surrounding microenvironment. Further studies [26] of tumor
microenvironment have revealed the role of some specific
mechanism, for instance those involved in tissue infiltration
and metastasis formation. These examples recall both
epithelial-to-mesenchymal transition (EMT) and degradation
of the extracellular matrix (ECM), responsible for the cell
morphology change that enables the detachment from the
primary tumor mass and the migration through the
extracellular space, with the aim of invading the surrounding
tissue. Such detached cells can reach lymphatic or blood
vessels, thus becoming circulating tumor cells and possibly
forming tumor colonies (metastases) in other body parts [27].
Networks offer a unique opportunity to characterize tumor
phenotypes [28]. In particular, modularity suggests the
relevance of novel strategies addressing marker panels. This
choice in turn is almost necessary in view of combinatorial
treatments. Due to the complexity of heterogeneous
proliferative mechanisms, multi-marker signatures are
expected to improve prediction power over the singlemarker signatures, thus delivering network fingerprints.
Basically, first network profiles are built from modular
configurations, and on the basis of specific hallmarks or
phenotypes. Then, the fingerprint makes a synthesis of such
profiles, building an ensemble of them. Such fingerprint is
strictly correlated with the tumor stage, and thus with the
hallmarks.
Perspectives
Substantial intra-tumor heterogeneity can be interpreted
by metrics underlying the network topology. As a matter of
fact, there is the possibility that multiple mechanisms of drug
resistance or sensitivity co-act within the same tumor,
explaining the response to therapy or the deviations in tumor
evolution. Therefore, clonal sub-populations of cancer cells
may be able to join forces or modularize genotypically and
phenotypically within the same cancer. When such
hierarchies form to characterize intra-tumor heterogeneity,
consequences are expected in terms of metastasis. Modular
networks naturally match sub-structures induced by special
signatures.
The process of growth in cancer is highly complex and
dynamic. One aspect that has been established is that a
variety of contextual cells (fibroblasts, leukocytes,
endothelial, extra-cellular matrix etc.) are activated, with the
context usually defined as tumor. Recently, stromal
therapeutic agents have been pursued due to the increasingly
Page 3 of 5
Cancer Cell & Microenvironment 2015; 2: e943. doi: 10.14800/ccm.943; © 2015 by Enrico Capobianco
http://www.smartscitech.com/index.php/ccm
important role played by stroma (macrophages, chemokines,
cytokines, etc.) for anticancer drug activity and
chemoresistance. A few questions arise: how therapies affect
the tumor microenvironment? How changes in it can be
exploited for improving treatment? What’s the role of the
various cell sub-populations?
Conceiving the tumor microenvironment as a dynamic
multi-type cellular network linking clonal sub-populations
with stromal cells and immune cells, can suggest novel
computational models [28]. The resulting interactome is
expected to be heterogeneous and multiscale, thus requiring
integrative approaches to describe intra- and inter-cellular
interactions. So-called community effects should be taken
into account, as cell sub-populations may underlie
tumorigenic or progressive growth differently, in part
facilitated by specialized tasks and roles and in part engaged
in different dynamics with regard to phenotypic
characterization.
Changes in cell state can be triggered exogeneously, i.e.
due to microenvironmental signals, thus in addition to both
the specific intrinsic variability in gene expression and the
general patterns observed in view of cancer plasticity [29].
The latter concept is linked to heterogeneity, implying the
fact that cell sub-populations affect in a distinct way cancer
phenotypes (growth, differentiation, and apoptosis).
It can be hypothesized that during cancer development
individual cells acquire various states in dependence on the
microenvironment, thus representing a primary source of
variation. Therefore, an interesting question is deciphering
the regulatory mechanisms through which cells can
compromise between stability (enabling functional
differentiation) and plasticity (allowing state transitions),
influencing metastasis. Cell sub-populations vary in growth
rate, immunogenicity, metastatic power and drug response.
From such diversity, both functional and phenotypic
heterogeneity can be explained, but only rationally and not
yet dynamically through models. Basically, interference,
competition and synergisms are all possible dynamics
between clonal populations, suggesting a community-like
behavior [30].
References
1.
Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000;
100: 57-70.
2.
Hanahan D, Weinberg RA. Hallmarks of cancer: the next
generation. Cell 2011; 144: 646-674.
3.
Gajewski TF, Schreiber H, Fu YX. Innate and adaptive immune
cells in the tumor microenvironment. Nat Immunol 2013;
14:1014-1022.
4.
Kaufmann T and Simon HU. Targeting disease
immunomodulation. Cell Death Different 2015: 22: 185-186.
5.
Butt AQ, and Mills KH. Immunosuppressive networks and
checkpoints controlling antitumor immunity and their blockade in
the development of cancer immunotherapeutics and vaccines.
Oncogene 2014; 33: 4623-31.
6.
Melero I, Berman DM, Aznar MA, Korman AJ, Gracia JLP,
Haanen J. Evolving synergistic combinations of targeted
immunotherapies to combat cancer. Nat Rev Cancer 2015; 15:
45772.
7.
Holohan C, Van Schaeybroeck S, Longley DB, Johnston PG.
Cancer drug resistance: an evolving paradigm. Nat Rev Cancer
2013; 13: 714-726.
8.
McCarthy N, Drug resistance: Making a point. Nat Rev Cancer
2014; 14: 6.
9.
Martz CA, Ottina KA,Singleton KR, Jasper JS, Wardell SE,
Peraza-Penton A, et al. Systematic identification of signaling
pathways with potential to confer anticancer drug resistance. Sci
Signal 2014; 7: ra121.
by
10. Whipple CA. Tumor talk: understanding the conversation between
the tumor and its microenvironment. Cancer Cell Microenviron
2015; 2: e773.
11. Brock A, Krause S, Ingber DE. Control of cancer formation by
intrinsic genetic noise and microenvironmental cues. Nat Rev
Cancer 2015; 15; 499-509.
12. Pe'er D, Hacohen N. Principles and strategies for developing
network models in cancer. Cell 2011; 144: 864-73.
13. Barabási AL, Gulbahce N, Loscalzo J. Network medicine: a
network-based approach to human disease. Nat Rev Genet. 2011;
12: 56-68
14. Huang SS, Fraenkel E. Integrating proteomic, transcriptional, and
interactome datareveals hidden components of signaling and
regulatory networks. Sci Signal 2009; 2: ra40.
Conflicting interests
15. Bandyopadhyay S, Chiang CY, Srivastava J, Gersten M, White S,
Bell R, et al. A human MAP kinase interactome. Nat Methods
2010; 7: 801-805.
The authors have declared that no competing interests
exist.
16. Morris MK, Saez-Rodriguez J, Sorger PK, Lauffenburger DA.
Logic-based models for the analysis of cell signaling networks.
Biochemistry 2010; 49: 3216-3224.
Acknowledgements
The author thanks his colleagues in cancer research for
many fruitful discussions.
17. Göke J, Chan YS, Yan J, Vingron M., NG HH. Genome-wide
Kinase-Chromatin Interactions Reveal the Regulatory Network of
ERK Signaling in Human Embryonic Stem Cells. Mol Cell 2013;
50: 844-855.
18. Welch CM, Elliott H, Danuser G, Hahn KM. Imaging the
coordination of multiple signalling activities in living cells. Nat
Page 4 of 5
Cancer Cell & Microenvironment 2015; 2: e943. doi: 10.14800/ccm.943; © 2015 by Enrico Capobianco
http://www.smartscitech.com/index.php/ccm
Rev Mol Cell Biol 2011; 12: 749-576.
Mol Syst Biol 2009; 5: 321.
19. Conway, JRW, Carragher NO, Timpson P. Developments in
preclinical cancer imaging: innovating the discovery of
therapeutics. Nat Rev Cancer 2014; 14: 314-328.
25. Johansson A, Hamzah J, Ganss R. License for destruction:
Tumor-specific cytokine targeting. Trends Mol Med 2014; 20:
16-24.
20. Dominietto M, Tsinoremas N, Capobianco E. Integrative analysis
of cancer imaging readouts by networks. Mol Oncol 2015; 9: 1-16.
26. Ungefroren H, Sebens S, Seidl D, Lehnert H, Hass R. Interaction
of tumor cells with the microenvironment. Cell Commun Signal
2011; 9: 18.
21. Kauffman S. Homeostasis and differentiation in random genetic
control networks. Nature 1969; 224: 177-178.
22. Huang S, Ernberg I, Kauffman S. Cancer attractors: a systems
view of tumors from a gene network dynamics and developmental
perspective. Semin Cell Dev Biol 2010; 20: 869-876.
27. Plaks V, Koopman CD, Werb Z. Circulating Tumor Cells. Science
2013; 341: 1186-1188.
28. Bolouri H. Network dynamics in the tumor microenvironment.
Semin Cancer Biol 2015; 30: 52-59.
23. Creixell P, Schoof EM, Erler JT, Linding R. Navigating cancer
network attractors for tumor-specific therapy. Nat Biotechnol
2012; 30: 842-848.
29. Brock A, Krause S, Ingber DE. Control of cancer formation by
intrinsic genetic noise and microenvironmental cues. Nat Rev
Cancer 2015; 15: 499-509.
24. Zhong Q, Simonis N, Li QR, Charloteaux B, Heuze F, Klitgord N
et al. Edgetic perturbation models of human inherited disorders.
30. Tabassum DP, Polyak K. Tumorigenesis: it takes a village. Nat
Rev Cancer 2015; 15: 473-483.
Page 5 of 5