Science and Cell Sub-Journal: Developing "smart" cell therapies for cancer with big data

Nov. 29, 2020 // --- The search for drugs that kill cancer cells and keep normal tissues unsalted is the top goal of oncology research.
in two new papers, researchers from the University of California, San Francisco, and Princeton University suggest complementary strategies for solving this conundrum with "smart" cell therapy: these live cell drugs remain inert unless activated by a group of proteins that appear only in cancer cells at the same time.
the biology of this universal approach has been explored for several years by Dr. Wendell Lim and his colleagues in the labs of the University of California, San Francisco's Cell Design Program and the National Cancer Institute-sponsored Synthetic Immunology Center.
, however, adds a powerful new dimension to this by combining cutting-edge therapeutic cell engineering with advanced computational methods.
in the first paper, published In the September 23, 2020 issue of cell Systems, entitled "Grady Power of Combinatorial Antigen Acknowledge in Cancer T Cell Therapies," members of Lim Labs teamed up with Dr. Olga G. Troyanskaya, a computer scientist at Princeton University's Louis-Sigler Institute for Integrated Genomics.
using machine learning, they analyzed a large database of thousands of proteins found in cancer and normal cells.
, they screened millions of possible protein combinations to create a list of protein combinations that could be used to precisely target only cancer cells, not normal cells.
from Cell Systems, 2020, doi:10.1016/j.cels.2020.08.002.
In a second paper, published in the November 27, 2020 issue of Science, entitled "Precise T cell system programs designed by transcriptionally linking multiple receptors," Lim and his colleagues then demonstrated how these calculated protein data can be used to drive the design of effective and highly selective cancer cell therapies.
, most cancer treatments, including cell therapy, are told to 'stop this' or kill 'this', " says Lim.
want to increase the nuances and complexity of therapeutic cell decisions.
"Over the past decade, chisellular antigens (CAR) T cells (CAR-T) have been a powerful treatment for cancer.
in CAR-T cell therapy, T cells in the immune system are taken from the patient's blood and genetically manipulated in the lab to express a specific subject that identifies very specific markers or antigens on the surface of cancer cells.
Although scientists have shown that CAR-T cells are quite effective and sometimes even curable for blood cancers such as leukemia and lymphoma, so far this method has not worked well in solid tumors, such as breast, lung or liver cancer.
cells in these solid cancers tend to have the same antigens as normal cells in other tissues, which carries the risk that CAR-T cells will have off-target effects on healthy organs.
, solid tumors often produce inhibitory micro-environments that limit the efficacy of CAR-T cells.
images from CC0 Public Domain.
In Lim's view, cells are similar to molecular computers, sensing their environment and then integrating that information to make decisions.
, he says, given that solid tumors are more complex than blood cancers, "you have to make more complex products" to fight them.
in the first paper, led by Dr. Ruth Dannenfelser, a former graduate student on the Troyanskaya team, and Dr. Gregory Allen, a clinical researcher at Lim Labs, the researchers searched public databases and explored the expression of more than 2,300 genes in normal and tumor cells to see which antigens could help distinguish them.
they use machine learning techniques to arrive at possible hit rates and see which antigen clusters are grouped together.
based on this gene expression analysis, Lim, Troyanskaya and their colleagues applied Boolean logic to antigen combinations to determine whether they could significantly improve T-cell identification of tumors while ignoring normal tissue.
For example, by using boolean operators AND, OR, or NOT, such as "A" OR"B"NOT "C", tumor cells may be distinguished from normal tissues, where antigens A and B are found only in tumor cells, while antigen C is found only in normal tissues.
to program these instructions into T-cells, they used a system called synNotch, a customizable molecular sensor that allows synthetic biologists to fine-tune cell programming.
in Lim Labs in 2016, SynNotch is a subject that can be modified to identify numerous target antigens.
the output reaction of synNotch can also be programmed, so once the antigen is identified, the cell performs any one of a series of reactions.
to demonstrate the potential power of the data they accumulated, the researchers used synNotch to program T cells to kill kidney cancer cells that express unique antigen combinations ---CD70 and AXL---
While CD70 is also found in healthy immune cells and AXL is found in healthy lung cells, T cells carrying the modified synNotch AND logic gate only kill cancer cells and release healthy cells.
big data analytics and cell engineering for cancer have made explosive advances over the past few years, but they haven't come together," said Troyanskaya, a cancer analyst.
the computing power of therapeutic cells and machine learning methods, it is available to make operational use of the growing wealth of cancer genome and proteomic data.
in a second paper, led by former UC San Francisco graduate student Jasper Williams, the researchers showed how multiple synNotch subjects are daisy-chained together to build a complex set of cancer identification circuits.
Since synNotch can activate the expression of selected genes in a "plug and play" manner, these components can be connected in different ways to build circuits with different Boolean functions, allowing for accurate identification of diseased cells and a series of reactions when they are identified.
Lim said, "This study is essentially a cell engineering manual that provides us with a blueprint for how to build different types of therapeutic T-cells that can identify almost any combination of antigen patterns that may exist on the surface of cancer cells."
"For example, a synNotch subject is modified to recognize antigen A, and when it identifies antigen A, the cell produces a second synNotch to identify the antigen B, which in turn induces the identification of the CAR expression of the antigen C, with the result that T cells need the presence of all three antigens to trigger the killing.
In another example, if T cells encounter antigens that are present in normal tissue but not in cancer, the nott-functioning synNotch subject can be programmed, which causes the T cells carrying it to die, thus protecting normal cells from attack and possible toxicity.
paper, using such a complex synNotch configuration, Lim and his colleagues showed that they could selectively kill cells carrying different combinations of melanoma and breast cancer.
In addition, when T-cells containing synNotch were injected into mice carrying two tumor-like combinations with different antigen combinations, they were able to efficiently and accurately locate tumors that they had modified to identify and reliably perform the cell procedures they designed.
Lim team is now exploring how to use these circuits in CAR-T cells to treat glioblastoma.
glioblastoma is an invasive form of brain cancer, and traditional treatments are almost always ineffective.
Lim, "You're not just looking for magic cures.
you're trying to use all the data.
we need to comb through all available cancer data to find a clear signal of cancer combinations.
if we can do this, we can use these intelligent cells to really harness complex biological calculations and have a real impact on the fight against cancer, " he said.
" (Bioon.com) Reference: 1. Ruth Dannenfelser et al. Discriminatory Power of Combinatorial Antigen Recognition in Cancer T Cell Therapies. Cell Systems, 2020, doi:10.1016/j.cels.2020.08.002.2.Jasper Z. Williams et al. Precise T cell recognition programs designed by transcriptionally linking multiple receptors. Science, 2020, doi:10.1126/science.abc6270.3.Big data powers design of 'smart' cell therapies for cancer。