Cancer, Evolution, ALife: Attendees
INVITED TALK: Is cancer a form of cellular organised crime?
Dan V. Nicolau (Queensland University of Technology, Australia and University of Oxford, UK)
The traditional myth of cancer and carcinogenesis has been that of an evil adversary. This is betrayed by our language around the subject: “malignancy” , “invasion” , “aggressive” , and so on. But modern evidence suggests a far more complex view is needed. Carcinogenesis, all the way from a barely-abnormal microcolony to full-blown metastasis, appears to be driven by cellular stress – itself in turn caused by environments hostile to the tumour cells and by immune attack – leading to escape behaviour. In this talk, I discuss the outlines of an alternative view of carcinogenesis and progression, in which cancer represents increasingly organised criminal (i.e. anti-social) activity driven by want and necessity rather than evil intent. In many ways, this view suggests that radically new concepts and strategies of treatment are needed if we are to make progress on cancer as the leading cause of the death and suffering in the 21st century.
INVITED TALK: Universal Criticality beyond Edge of Chaos, derived from Proto-Anticipation
Yukio P. Gunji (Waseda University, Tokyo, Japan)
Although life is considered to be a universal and highly efficient computer, it has been regarded at the edge of chaos. While most of learning machines are fallen into either overfitting or failure of leaning, adequate generalisation performance is achieved by special device such as neural network with deep learning. Highly efficient and universal cellular automata are also found at the narrow region between chaos and order. Artificial swarms equipped with both sociality and freedom are also found only at the critical point of the phase transition. We, however, think that such views are derived from classical mechanism based on synchronous updating. We here show that highly efficient and universal computer are generally appeared from asynchronous updating and/or anticipation. If updating is asynchronous, probability dependent on empirical data is perpetually changed to overestimated or underestimated. It entails mixture of probable future and certain present, that is nothing but anticipation. In cellular automata anticipation is implemented by asynchronous time and tuning, and universal criticality beyond the edge of chaos is shown. In swarm models, anticipation is implemented by Bayesian inference equipped with changeable likelihood, and a swarm showing marching, splashing and/or massive tornado is shown. Since co-existence of these behaviour patterns reveal typical critical behaviour, it also shows robust criticality derived from anticipation.
INVITED TALK: Targeting cancer cells by altering pre-mRNA splicing
Michael R. Ladomery (Faculty of Health and Applied Sciences, UWE, Bristol, UK)
Pre-mRNA splicing, discovered in the 1970s, is the process through which introns are removed and exons joined together to create mature mRNAs. Through alternative splicing exons are be joined together differentially, for example by being skipped or by using alternative splice sites changing their length. The result of alternative splicing is that genes can express multiple protein isoforms that often exhibit different and even antagonistic properties. In humans the vast majority of protein-coding genes, over 94%, are alternatively spliced. Increasing evidence shows that aberrant alternative splicing contributes to the cancer phenotype, usually by favouring the expression of pro-oncogenic splice isoforms. With this in mind, there is a drive to develop novel cancer therapies based on the manipulation of alternative splicing. We describe two approaches. In the first approach we have targeted splice factor kinases. Splice factor kinases modify the activity of splice factors, proteins that regulate alternative splicing. We show that targeting the splice factor kinases SRPK1 and CLK1 convincingly impedes the growth of cancer cells in vitro and in vivo in mouse xenografts. In the second approach, we have designed and tested morpholino-based antisense oligonucleotides that target the ERG oncogene by modifying its splicing. The antisense reagents also inhibit the growth of cancer cells in vitro and in vivo. In summary, through these two approaches, we show that new anti-cancer therapies can be developed by altering the pre-mRNA splicing of cancer-associated genes.
Evolutionary computation in the service of the fight against cancer
Michail-Antisthenis Tsompanas, Larry Bull, Andrew Adamatzky and Igor Balaz (University of the West of England, UK and University of Novi Sad, Serbia)
Cancer is one of the most complicated health issues that the scientific community is facing. Due to this high complexity of the problem and the ability of cancer cells to adapt and evolve resistance in applied therapies, the scientific weaponry requires an appropriately resilient tactic. Consequently, evolutionary algorithms and artificial intelligence are the most reasonable candidate tools to be utilized in this fight. Recognizing this need, EVONANO initiative focuses on the implementation of evolving techniques in the simulated investigations of novel cancer treatments with the use of functionalized nano-particles (NPs). This approach has the ultimate goal of designing highly efficient and personalized treatments.
Several optimization methodologies of the design for a type of NP were tested thus far, such as simple genetic algorithms, differential evolution, novelty search, haploid-diploid representations and surrogate-assisted models. Then, the application of more than one types of NPs applied at the same time as a treatment, was investigated. As there is no knowledge of the best amount of types of NPs, this is left to be found by the evolutionary algorithm. This algorithm uses what is termed as a metameric representation; a variable length of genome that can be divided into similar segments each one representing one different type of NPs. After appropriate changes in the source code of PhysiCell simulator to accommodate the simulation of more than one types of NPs, the evolutionary algorithm was tested. Preliminary results show that the multiple types of NPs used as a treatment are more promising than single NPs, as expected.
Adaptive Novelty and Automated Generation of Cancer Treatments in Simple Agent-based Model
Igor Balaz, Tara Petric, Marina Kovacevic, Namid Stillman, Antisthenis Tsompanas (University of Novi Sad, Serbia and University of the West of England, UK)
Malignant tumours are heterogeneous structures that can rapidly acquire drug resistance. To overcome therapy resistance, clinical practices are using so called combinatorial therapy – combining multiple drugs in order to synergistically eliminate various clones that emerge in a tumor. The use of nanoparticles as drug carriers can overcome the limitations associated with traditional drug therapy by improving site-specific targeting of drugs, increasing in vivo stability, extending the drug’s blood circulation time, and allowing for controlled drug release. In order to design efficient treatment tailored to specific tumor, we are facing an immense state of possible solutions generated by: (i) various drugs for targeting specific cancer cells, (ii) nanoparticles that can modify physico-chemical properties of those drugs, and (iii) heterogeneous and highly adaptable tumor that can become resistant to the primary therapy.
To automate creation of therapies for such complex and changeable scenarios, we need open-ended approach where the exploration of an initially small portion of possible solutions can be expanded by ongoing generation of adaptive novelties, whenever simulated tumour makes an adaptive leap. Therefore, we created a simple engine for open-ended virtual evolution of combinatorial oncological treatments. It is a system of non-interacting evolvable nano-agents in the 2D environment of cancer- and healthy cells. Since tumour can adapt and thus became resistant to nano-agents, the adaptive landscape of nano-agents is under constant modification so the context in which their evolution occurs is continuously changing. As a result, the population of nano-agents continuously reshapes itself and creates combinatorial, optimized strategies.
Vessel-on-chip in-vitro model for nanomedicine validation
Jean Cacheux, Sergio Davila and Isabel Rodriguez (IMDEA Nanoscience C/ Faraday, Spain)
Tumour-on-a-chip devices recapitulating in vitro the particular physiological scenario of the tumour microenvironment, are an ideal platform for the development and validation of anti-cancer nanomedicines. The principal physiological scenario to consider when it comes to nanomedicine delivery is the microvasculature given that the main driver for nanomedicine selective extravasation into a tumour is attributed to the so called enhanced permeability and retention (EPR) effect. However, drug delivery by nanocarriers has not generated the expected results and according to a recent publication, the delivery rate of nanomedicines to the target tumours is on average 0.7% of the total injected dose. Moreover, Sindwani and co-workers recently revealed that the EPR effect may not be the dominant mechanism for nanoparticle extravasation, instead authors argue that nanoparticles extravasate towards the tumour using active transcellular processes through the vascular endothelium.
Hence, there is a need to improve our understanding of the nanoparticle extravasation process to be able to improve the therapeutic efficacy and clinical translation of cancer nanomedicines. Towards this aim, we are developing an artificial tumour microvessel on-a-chip. This comprises of microchannels with a diameter in the range of the tumour capillaries and semi-circular geometry. These channels serve as scaffold of human endothelial cells to form a vessel-like endothelium. Connections are established to control the fluid flow and shear stress of a blood-analogue shear-thinning fluid, recapitulating closely the physiological conditions of the tumour vessel microenvironment. In the artificial microvessel, the dynamics of gold nanoparticles and interaction with the artificial endothelial-cell monolayer are studied. Time-lapse fluorescence microscopy images of labelled nanoparticles indicate that they adhere readily to the endothelial cell membrane and get internalized soon upon interaction. Overtime, a substantial accumulation and clustering of nanoparticles in the cytoplasm near the nucleus was seen. The mechanisms of in-out intracellular trafficking of nanoparticles in endothelial cells are under study.
Therapeutic Manipulation of Macrophages for the Treatment of Cancer using Nanoparticles
Fernando Torres Andón
In most cancer patients, chronic inflammation and immune suppression are the dominant effects in the tumor microenvironment. The infiltration of tumor associated macrophages (TAM), with a pro-tumoral (M2-like) phenotype, in tumor tissues supports tumor growth, invasion and metastasis. Indeed, high density of TAM in tumors is correlated with resistance to therapies and poor prognosis . Different immunostimulant drugs, such as TLR-agonists, have been shown to repolarize M2-like macrophages towards an antitumor M1-like phenotype. However, their indiscriminate delivery poses a risk of unwanted systemic immune activation.
We have decided to use nanotechnology as a strategy to improve the delivery of these drugs towards TAM in the tumor microenvironment . For this, considering the molecular properties of TAM, we work on the design of Polymeric Nanostructures (PNs) to target and re-educate TAM. These PNs are composed of biodegradable polysaccharides (i.e. hyaluronic acid) which are functionalized to optimally reach TAM in vivo. In addition, these nanomedicines (PNs) are loaded with immunostimulatory drugs aimed to re-educate TAM into M1-antitumor macrophages. Classical immunostimulant drugs were loaded into the PNs to reach and activate their respective receptors, TLR-3 and TLR-7/8. These TLR-loaded-PNs were characterized by their shape, size, surface charge and drug encapsulation efficiency. We have optimized appropriate in vitro and in vivo models to evaluate the antitumor effect of these PNs and their mechanism of action through activation of macrophages. In vitro 2D cell cultures, using primary human monocyte derived macrophages alone or combined with cancer cells, have been used to study the toxicity of the TLR-loaded-PNs and their ability to program macrophages into an M1-anti-tumor phenotype. Phenotype and functional evaluations of these cells are performed by Alamar Blue, FACS and ELISA assays. These experiments demonstrated their favorable biocompatibility profile and their ability to enhance the cytotoxic activity of PNs-treated-macrophages towards cancer cells. The in vivo biodistribution/biocompatibility and antitumoral efficacy of TNs is tested using appropriate pre-clinical murine tumor models of fibrosarcoma and lung cancer (i.e. tumors rich in macrophages). At the end of the treatment, tumors are excised and analyzed by flow cytometry and by immunohistochemistry to assess the infiltration of immune cells (i.e. phenotype of macrophages in the tumor). TLR-loaded-PNs were able to reduce tumor growth. Future experiments will be carried out to test these PNs in orthotopic murine lung cancer models. With this strategy we expect to enable greater progress in the treatment of tumors and ultimately lead to improved outcomes for cancer patients.
Systems-based biomarkers for precision oncology
Federica Eduati (Eindhoven University of Technology, Netherlands)
Cancer is a very heterogeneous disease, with patients responding differently to drug treatments. Although a variety of treatments is already available, we often lack the ability to assign the best available treatment to each patient. Biomarkers can be used to predict patients’ response to a drug, therefore supporting the decision of the therapeutic intervention to adopt for each patient. With the advent of small molecules inhibitors targeting specific proteins, the search for biomarkers has focused in particular on individual mutations. However individual mutations have revealed in general to be poor biomarkers of drug response, due to the complexity of cancer and its ability to develop different resistance mechanisms to drug therapy. In this talk I will present the approaches we have developed to integrate different types of omics data with prior knowledge on the system (e.g. on the structure of the cellular networks), in order to be able to take into account the complexity of cancer. In this regard, we use different types of computational models, ranging from machine learning techniques to more mechanistic mathematical models that allow the description of signalling networks. The characterisation of the dynamics of patient-specific signalling networks is obtained using the data from a microfluidics platform for screening of drugs in patient biopsies that we recently developed.
Explainable AI for Nanoparticle Design in Cancer Treatment
Sepinoud Azimi, Marina Kovacevic, Sebastien Lafond, Igor Balaz (Åbo Akademi University, Finland and University of Novi Sad, Serbia)
Cancer treatment has come a long way, however, the existing traditional treatments, e.g., radiation and chemotherapy, although effective, have major side effects since they affect not only the tumor, but the other healthy cells as well. In recent years, nanomedicines (NM) have emerged as a possible alternate strategy. NM have been shown to improve site-specific targeting, the potential to modulate both the pharmacokinetic and pharmacodynamic profiles of drugs, thus improving efficacy and reducing side-effects. Although promising, this approach has its own major obstacles. Design of novel and efficacious NM requires thorough understanding of their physicochemical properties and their behavior in biological environment. However, it is often difficult, even impossible to obtain that information experimentally. Moreover, systematic variation of NM composition can be time consuming and expensive. This can be solved by using simulation-based approaches. Molecular Dynamics (MD) simulations are used to investigate structural variations and their influence on the behavior of NMs thus suggesting favorable chemical modifications. However, as the state space of simulation inputs could be dramatically large, it is not feasible to run simulations for all possible settings as this would require years of CPU time,
As a way to deal with such problem, an Artificial Intelligence (AI) model could be trained on a subset of MD simulations of the aforementioned large state space to predict the efficiency of the nanoparticle design. Although this could be a step forward, the nested non-linear structure of AI models is still a black box in the majority of the cases. In other words, although the AI models provide valuable information on whether a particular design is efficient or not, they fail to explain what exactly makes them arrive at their predictions. The large drawbacks that such lack of transparency entails, have resulted in the emergence of the field of explainable AI. In this talk we discuss potential approaches in developing more interpretable AI models trained to predict the efficiency of the nanoparticle designs. We discuss the existing methods that could be used in this context to explain predictions of deep learning models which could analyze the decision in terms of the input variables.
Determination of the physiological effect of functionalized nanoparticles (CSC dynamic models)
Petra Gener (Vall d’Hebron Institut de Recerca, Spain)
Resistance to therapy and metastatic disease seem to be sustained by the presence of cancer stem cells (CSCs) within the tumors. These cells retain the capacity of repopulating the tumor, while being insensitive to conventional anticancer therapies, antimitotic agents or radiation [1,2]. Accordingly, the percentage of CSCs within a tumor often increases after anti-cancer treatment like chemotherapy or radiation. This leads to cancer recurrence and metastatic growth since only few CSCs are necessary and sufficient for tumor regeneration and tumor spread . Clinicians often observe tumors initially shrinking after multimodal treatment, while remaining resistant clones of CSCs survive and eventually cause tumor re-growth and relapse, often rising very aggressive tumor types with unfortunately, very limited treatment alternatives [4–6]. Base on this acuity, several cancer treatment strategies have been designed and developed to target CSCs over the last decade. Nonetheless, their clinical relevance has been very limited [7,8]. One way to ameliorate the clinical outcome, is to use artificial intelligence to design an effective treatment (e.i. nanomedicine) and to predict the tumor behavior in respect. Yet, the cellular and animal models are crucial to feed the artificial intelligence with data in respect of tumor behavior, its dissemination and interaction with the treatment (nanoparticles). On other hand the cellular and animal models may serve to verify the consequence of the design nanomedicine in biological systems. We will discuss here the usefulness of in vitro and in vivo models CSC model as well as an innovative 3D in vitro models and CSCs in vivo syngeneic models that involve fibroblast and immune system, to mimic the tumor reality as much as possible.
Convenient and versatile platform based on functionalized gold nanoparticles for assessment of biological activity
Dariusz Witt (ProChimia Surfaces, Poland)
Spherical gold nanoparticles (AuNPs) are one of the most widely used gold nanostructures in drug-delivery applications. A large number of AuNPs can be easily synthesized with relatively high monodispersed size by the reduction of aqueous chloroauric acid. To obtain hybrid materials, dynamic ligand exchange is performed after synthesis of ligands and originally coated gold nanoparticles. The preparation of thiols terminated with quinoline-8-ol will be presented. The synthesis involves formation of alkane thiols with hexa-ethylene spacer to improve solubility under aqueous conditions and provide exposure of quinoline-8-ol (active pharmaceutical ingredient, API) to interaction with molecular targets. The appropriate construction of functionalized gold nanoparticles (AuNPs) allows to control solubility, zeta potential, pH stability, and size of nanoparticles. These surface modifications ensure nanoparticles stabilization and determine their properties. The obtained functionalized gold nanoparticles are characterized by UV-VIS spectroscopy. The size of AuNPs can be determined by DLS and TEM, however zeta potential that reflects charge density can be measured by DLS. The pH stability or fluorescence (for AuNPs with selected dye) for functionalized gold nanoparticles will also be presented.