The research carried out in the research group is mainly focused on aspects of drugs being used in cancer treatments.
Cancer pharmacology - Drug repositioning
New strategies for drug discovery are needed. One such strategy is drug repositioning/repurposing, i.e. when a new indication for an existing drug is identified and explored. In this approach, known on-patent, off-patent, discontinued and withdrawn drugs with unrecognized cancer activity can be rapidly advanced into clinical trial for this new indication since much or all of the required documentation to support clinical trials can rely on previously published and readily available data. During the past years we have systematically screened several innovative model systems with focus on colorectal carcinoma and acute myelocytic leukemia using our library of annotated and clinically tested drugs. In this effort we have identified several potentially useful candidates for repositioning including the anti-parasitic drugs mebendazole, nitazoxanide and quinacrine. Current research activities are focused on identification of compounds with immunooncology properties. With new methodology being developed for screening for immune modulating activity we will use our in-house library of clinically used compounds for identification of new candidates for drug repositioning.
Cancer pharmacology - High-throughput screening - in 3D
Understanding how cells within tumors respond to drugs is a critical issue in anticancer drug development. Much of what we currently know about cancer, both in terms of treatment and underlying molecular mechanisms, has been learned through growing cancer cells in the laboratory. One major limitation of conventional cell culture is that cancer cells are grown in a single ‘two-dimensional’ layer. In such cultures all cells have unlimited access to nutrients and oxygen and they can grow very quickly. In real tumors, where blood vessels grow chaotically, not all of the cells have a steady inflow of nutrients and oxygen. As a consequence, within tumors there are regions where cancer cells are starved and divide slowly or do not divide at all, while managing to stay viable. They also become insensitive to standard chemotherapy, which mainly targets actively dividing cells. After the treatment has ceased, these surviving dormant cells can often resume growth. They are also believed to be one of the major causes of cancer relapse. Thus, there has recently been an increased interest in laboratory cell models that would incorporate not only dividing, but also quiescent cells and that would mimic tumors in patients more accurately. One way of achieving this is growing cancer cells in three dimensions (3D). 3D cultures are essentially small tumors, in which cells have limited supply of oxygen and nutrients. These models could be a useful tool in the hunt for new drugs that are toxic to dormant cancer cells. We have developed a platform for high-throughput screening in 3D cultures. The platform makes it possible to monitor both viability and gene-expression responses to drugs in thousands of 3D cultures in parallel. Using this approach, we have found that inhibitors of oxidative phosphorylation (OXPHOS) are toxic to dormant cancer cells, and that inhibition of cholesterol synthesis (statins) increases the effect.
Cancer pharmacology - Targeting glioma
Glioblastoma (GBM) is the most frequent and malignant primary brain tumor in adults, with an almost invariably fatal outcome. The current conventional treatment with surgery followed by radiotherapy and chemotherapy (temozolomide) is largely ineffective. Tumor heterogeneity is a characteristic feature of GBM and is likely to be the source therapy failure and relapse. Within the GBM tumors there is a small subset of cells that can self-renew, generate a differentiated progeny, and most importantly, initiate new tumors, which are defined as Glioma stem cells (GSC). To investigate this further, patient specimens are used in a pipeline for identification, isolation, categorization and expansion of the GSC’s. From a single GBM tumor numerous clones can be derived and clone libraries have been established. These clones are extensively characterized for morphology, growth rate, initiating capacity, genetic profiling, tumor forming ability in immunocompromised mice (A), expression of phenotypic markers (B), as well as radiation and drug resistance (D). All these characteristics aid in dividing them into clones with a resistant profile or a sensitive profile (C). For instance, a clone resistant to one treatment is often multiresistant and has similar characteristics and genetic profile as other resistant clones. Since all treatments are ineffective, we are exploring the possibility of combining two or more drugs to kill the GBM cells. The clones are used for screening of compound libraries with the intent to find new drugs that are more toxic to GBM cells than normal cells.
This project is founded by grants from the Knut and Alice Wallenberg Foundation, and performed in collaboration with Neurooncology at the Institution of Genetics and Pathology, UU, headed by Professor Bengt Westermark (firstname.lastname@example.org).
Cancer pharmacology - Personalized medicine
In close collaboration with Uppsala University Hospital we perform ex vivo chemosensitivity assays on primary patient cells. The assay results in a resistance classification for each assayed compound which is predictive of in vivo tumor response. We have recently updated the assay to use acoustic dispensation of compounds and an integrated informatics workflow. The number of compounds assayed is limited by cell yield, and we are currently developing a more sensitive end-point and methods for expanding primary patient cells. The predictive strength of the assay lies in predicting resistance.
Emerging knowledge has demonstrated the importance of the tumor microenvironment as a determinant of chemoresponse. To incorporate this component in the assay we are investigating stroma cell co-cultures to provide clinically relevant chemoresponses.
Our continuously growing database with over 10,000 patient samples analyzed across a variety of diagnoses is a valuable resource for data-mining of patterns of chemoresistance. In addition, as a part of drug candidate characterization, we routinely assay novel drug candidates on patient primary cells. This drug activity profiling provides valuable information on specificity and efficacy. Patient cell super-responders and highly resistant cells can be further characterized by systems pathology to elucidate mechanisms of action and resistance.
We also screen small to medium sized compound libraries on primary patient cells. Using bioactive compound libraries of registered drugs we are able to discover novel indications as cancer therapy and reposition the drug for novel treatment, such as e.g. mebendazole.
Contact Peter Nygren (email@example.com), Kristin Blom (firstname.lastname@example.org), Claes Andersson (email@example.com) or Rolf Larsson (firstname.lastname@example.org) for more information.
Cancer pharmacology - Targeting aminopeptidases for cancer therapy
Many tumors express hydrolytic enzymes to modulate the microenvironment and maintain malignant growth. For example, aminopeptidase N is an ubiquitous enzyme with strong association with the characteristics of malignancy, e.g. angiogenses, cell motility and aggressive growth. These enzymes may be utilized as targets for therapy. In this project we have synthesized derivatives of cytotoxic compounds to be activated or potentiated by these enzymes in order to develop anti-cancerous therapy with higher therapeutic index. The project has led to the development of one candidate drug, melflufen, currently evaluated in clinical trials in multiple myeloma. Current research will further investigate the spectrum of preclinical activity of melflufen in different tumor types, screening for promising combination partners and explore the effects on the immune system.
For more information contact Joachim Gullbo (email@example.com).
CANCER PHARMACOLOGY – SENSITIZATION OF RESISTANT GLIOBLASTOMA CLONES
Gliobastoma multiforme is the most malignant primary brain tumor and treatment is essentially lacking. The cancer cells vary between more treatment-resistant and more treatment sensitive cell-states and the high relapse rates may be due to drug treatment affecting only part of the tumour cell population.
By raising cell cultures from single tumor cells originating from the same patient tumor (clones) we partly capture this variation and have linked drug and radiation resistance to a cell-state of mesenchymal (MES) character. Importantly, this cell-state appears epigenetically regulated and reversible. Current studies aim to uncover forces that drives changes in treatment resistance/cell-state and based on this knowledge make all cells vulnerable to therapy.
The general approach is to, by various means, stimulate and/or antagonize pathways that appear to be more active in sensitive and resistant clones, respectively. Drugs, natural ligands and possibly genetic techniques are used and effectiveness of sensitization towards conventional treatment evaluated. To link drug response to biomarkers the project includes molecular analysis of cell-state transitions for selected sensitizing treatments.
Mesenchymal transition of cancer cells is a generally acknowledged resistance mechanism. The studies are concentrated on GBM but the concept might well be applicable on other tumour types.
Computational medicine - Drug combination discovery and optimization
For complex disease like cancer, theoretical as historical facts clearly suggest that successful clinical results can only be obtained by means pharmacotherapies that consists of more than one drug. Drug combinations make it possible to target multiple biological processes and pathways, including compensatory feedback loops that often make single compound treatments ineffective or only successful for short periods of time.
Drug combinations also open for the use of low concentration treatments that can reduce problems with adverse side effects. During the last years we have therefore stated to develop novel theoretical, computational and experimental tools that can accelerate the identification of promising drug combinations for anti-cancer treatment. The focus disease areas are colorectal carcinoma, glioblastoma, and acute myelocytic leukemia. These efforts include:
- High throughput pairwise dose titration analysis
- Cancer cell reprogramming by drug mixture pre-treatments
- Iterative search strategies for combinations of arbitrary size
- Exhaustive profiling of all combination sizes from pairs of 2 up to 8 drugs
- In vitro systems pharmacology based identification of drug combinations
- Joint Bliss and Loewe synergy analysis combined with statistical resampling
Computational medicine - Cell level molecular systems pharmacology and pathology
By means of molecular profiling of human normal and cancer cells at different molecular levels (DNA-methylation, DNA, mRNA, protein, metabolite) we are performing systems pharmacology and pathology analyses and perform associated development of tailor made computational/algorithmic tools.
Performing systems pharmacology mainly at the mRNA and metabolite levels we are comparing how how different chemical and genetic perturbations are affecting the genome wide mRNA gene expression or metabolite patterns with applications to mechanism of action determination and for discovery of new promising drug and drug combinations.
Performing systems pathology at several molecular levels we are or have been studying different cancer forms such as glioblastoma, thyroid and colon carcinoma, as well as different types of leukemia. The general aim is to find systemic biomarkers that can predict and explain clinical outcomes and/or in vitro drug resistance patterns. These attempts to find single or systemic biomarkers are performed by means of different statistical machine learning (multivariate data analysis) approaches.
In term of computational tools we develop statistical machine learning methods that are free from information leakages that otherwise may cause over optimistic performance estimates and associated misleading biomedical interpretations. We also develop machine learning methods producing predictive models which have relatively simple and compact geometrical interpretations, for example based on modeling latent (hidden) variables/relationships. This includes development of automated algorithms for computational construction and analysis of static and dynamic molecular interaction network models based on collected experimental data. For example we compare different drugs in terms of the protein-protein networks they are interacting with, as an alternative to the comparing them in terms of their actual protein targets or in terms of their chemical structures. This confirms that very different drugs, in terms of their chemical structures and protein targets, may target the same protein-protein network and therefore be suitable for treatment of the same disease.
Computational medicine - Multivariate models of molecular structure-activity relationships
Improved understanding about how the chemical structure of small molecules, peptides and proteins are related to the biological effect they induced when exposed to human cells is of fundamental importance, for example in the context of improved pharmacological understanding of drug-target interactions, drug structure optimization, and improved understanding of allergy inducing mechanisms. By means of different statistical machine learning based procedures we model multivariate relationships between the structure of chemicals (small molecules, peptides as well as proteins) and their biological effects.
Computational medicine - Label-free Phenotypic Profiling Using Time-lapse (Video) Microscopy
As a replacement for or an addition to conventional cell viability characterization of cell cultures growing in vitro, a phenotypic profiling using time-lapse microscopy is a very attractive alternative. One obvious strength is that such a profiling may cover several days/weeks whereas a cell viability assay is an endpoint assay. Another strength is that as it may generate a lot of information relative to the experimental human manual efforts required, provided that adequate information extracting image analysis algorithms can be developed. A fundamental problem we address in this context is that in densely growing human in vitro cell cultures, it is not possible to segment out individual cells from label-free (gray scale) images. This means that it is not possible to quantify the cell properties individually, unless molecular labeling techniques are used, or the cells are grown sparsely enough to allow reliable image segmentation. As expected and shown from our research, it is possible to quantify very informative global morphological changes from label-free images without isolating and characterizing the individual cells. Often there are morphological changes in the cell cultures induced by chemical perturbations that are not reflected in the cell viability readout.
Based on two different types of time-lapse microscopy systems and ideas from statistical signal processing and machine learning we are developing very fast novel image processing methods designed to process label-free (gray scale) movies for phenotypic profiling of cancer and normal cell cultures growing in vitro in a semi high-throughput format (384-well microtiter plates). The main areas of development and applications are:
- Detection of differences in time evolving morphological patterns.
- Matched filters for detection of specific phenotypes like vesicles and apoptosis.
- Ultra-fast and memory efficient image processing based on modern computational parallelization strategies.
- Parameter optimization of the image processing enabled by the parallelization.
- Drug and drug combination screening for colorectal carcinoma and glioblastoma.
- Phenotypic characterization of glioblastoma subclone diversity.
- Drug combination analysis for treatment of anti-biotic resistant bacteria.
Contact Mats Gustafsson (firstname.lastname@example.org) for more information.