People with acute myeloid leukaemia have a bleak prognosis, with only 25% surviving 5 years. Chemotherapy seems effective for many, but cancer cells often develop resistance. Researchers can now combine single-agent screening with artificial intelligence and determine which combination of chemotherapy an individual should have. Once the system undergoes clinical trials, it should be able to help people with leukaemia at different stages and potentially people with other types of cancer.
A vaccine can target one specific mutated version of SARS-CoV-2 to combat COVID-19. However, cancer poses another challenge because the pathogen is not external but cells in a person’s body that have begun to grow uncontrollably. This means that the cells to be destroyed are similar to one’s own and that the same type of cancer must be treated differently from person to person. Cancer researchers, together with experts in artificial intelligence and machine learning, have now found a possible solution to this problem.
“The challenge for people with acute myeloid leukaemia is that the cancer cells gradually become resistant to any specific type of chemotherapy, but if you combine two types of therapies, resistance develops less often and more slowly. We have trained computers to analyse the single-cell gene expression in both cancer cells and healthy cells and, based on that, we can choose the right combination of chemotherapy at a given time to combat the leukaemia. Once a fully developed system has been tested in animal models, we hope that we can effectively treat the many people who have had to give up treatment because the usual chemotherapy no longer worked,” explains Krister Wennerberg, Professor, Biotech Research and Innovation Centre (BRIC), University of Copenhagen.
Chemotherapy can be unusable
Acute myeloid leukaemia starts in the bone marrow, where new blood cells are formed, but it often also moves quickly into the bloodstream, make it difficult to stop in time. It is also a complex type of cancer characterized by a broad spectrum of molecular changes. Although an increasing number of molecularly targeted treatment options have been developed, resistance to treatment is still the stumbling block to arrest the disease.
“The new drugs that are used to treat people with leukaemia with specific mutations give them better treatment options. They are more effective at pushing back the cancer than older drugs and often cause fewer side-effects. However, the unfortunate reality is that, in most individual cases, the cancer is either resistant to the drug from the start or the cancer comes back in a resistant form after only a few months of treatment. The reason is the development of other mutations or changes in the leukaemic cells that activate signals so that the targeted drug can no longer kill them. We therefore wanted to be able to measure the state of the cells of the individual patient, so that at a specific time we can determine which drug or, preferably, drugs can eliminate the cancer cells,” says Krister Wennerberg.
Cancer cells develop resistance considerably more easily to one rather than two chemotherapy drugs, and the researchers therefore tried to find combinations of drugs that can effectively kill the diseased cells and, just as importantly, do not kill the healthy cells. This can be a challenge, especially with combination therapy. Combining two types of chemotherapy makes adapting difficult for the cells because they have to evolutionarily change two mechanisms simultaneously.
“However, this applies to both cancer cells and healthy cells, so even if a combination of drugs is extremely effective in killing cancer cells, it can prove useless because it also kills a lot of healthy cells. We have therefore linked our testing of the combination chemotherapies with a flow cytometric separation method. This enables us to separate the healthy cells from the diseased cells from a tissue sample and see how effective the drugs are, so we can find the combination that is maximally effective on the cancer cells and minimally effective on the healthy cells,” explains Krister Wennerberg.
Success rate 10 times greater
Flow cytometry detects, counts and distinguishes specific cells by using a laser: for example, by physical properties or markers on the cell surfaces. The researchers also examined the individual cells using single-cell RNA sequencing, a technique that measures which of the cells’ genes are translated into RNA molecules, which function as a template for protein production in the cell.
“The RNA profile provides a snapshot of the condition of the cells. By combining that RNA profile with our cytometry test of different types and combinations of chemotherapy, we trained a computer to be able to predict which types of chemotherapy will be effective. We have therefore succeeded in developing a system that very accurately selects drug combinations that both combat the leukaemia cells and protects the body’s healthy cells,” says Krister Wennerberg.
The new machine learning method proved to be extremely effective because finding synergy between drugs is usually difficult. In existing screening systems, only about 5% of the combinations found are effective. In the new system, the success rate was more than 50%, 10 times higher.
“The existing screening systems often work with predefined combinations and fixed concentrations to limit the scope of screening. This means that we can easily miss new synergies. Our more data-driven approaches to systematically examining the most potent combinations for each individual enable us to find combinations that would otherwise be overlooked,” explains Krister Wennerberg.
Attempting to understand resistance
The new system has so far only been tested on cells in the laboratory, so the researchers still have a lot of work ahead: optimizing the system, testing in animal models and clinical trials involving people. These trials will probably initially focus on the people with the most severe leukaemia, for whom treatment options seem exhausted.
“Leukaemia is a great challenge. One type is chronic and develops more slowly, and we have good experience with that. But acute leukaemia quickly spreads to other parts of the body such as the lymph nodes, liver, spleen and brain, so rapid and effective treatment and continually being able to screen when resistance occurs are important. This is an option with the new system, and also that we can continually adapt the treatment to the mutations that occur in the cancer cells,” says Krister Wennerberg.
The researchers do not yet understand exactly how resistance occurs. It is probably a dynamic process that is influenced by both genetic and molecular factors along with the selective pressure that chemotherapy puts on the cells and their environment – all factors that continue to affect how well chemotherapy works.
“We are trying to understand the cellular conditions and phenotypes that can confer resistance and ultimately how these cellular conditions can be altered to completely prevent resistance from developing in cancer,” explains Krister Wennerberg.
According to Krister Wennerberg, there is great clinical need for developing more rational and systematic strategies for combination chemotherapy and not just for treating people with acute myeloid leukaemia.
“When we developed this method, our research focused on acute myeloid leukaemia, but this method can certainly be used for other types of cancer such as ovarian cancer. However, many types of cancer have heterogeneity among the cells, both because there are often different types of cells but also because they are at different stages. We hope that our system will be able to handle this in future cancer treatment,” concludes Krister Wennerberg.