Computer model simplifies immune cell identification for lung cancer treatment

Researchers at the Johns Hopkins Kimmel Cancer Center and Bloomberg-Kimmel Institute for Cancer Immunotherapy have developed a computer model to help scientists identify tumor-fighting immune cells in patients with lung cancer treated with immune checkpoint inhibitors.

In their study published Feb. 3 in Nature Communications, the team, including first author Zhen Zeng, Ph.D., a bioinformatics research associate at the Kimmel Cancer Center, demonstrated that their three-gene “MANAscore” computer model can identify the immune-cells targeted by immune checkpoint inhibitor therapies. It also helped the team identify differences associated with patient response to immunotherapy.

“We have developed a way to identify the cells directly targeted by immune checkpoint inhibitors, and if we can identify them, we can study them,” says the study’s senior author, Kellie Smith, Ph.D., an associate professor of oncology at Johns Hopkins. “If we can study them, that means we can identify better biomarkers and better targets for combination immunotherapy.”

Immune checkpoint inhibitors like PD-1 inhibitors are available to treat dozens of cancer types. These revolutionary therapies work by unleashing tumor-killing immune cells, called T cells, that are switched off by the protein PD-1.

PD-1 inhibitors turn the patient’s T cells back on, allowing patients’ immune systems to fight cancer more effectively. But not all patients respond to these therapies, and scientists need to know why in order to develop improved therapies that help nonresponders.

“Tumor-active T cells are very important to a patient’s response to therapy, but they are difficult to find,” Zeng says.

Smith helped develop the MANAFEST technology (Mutation-Associated NeoAntigen Functional Expansion of Specific T Cells), which she and colleagues first described in the journal Cancer Immunology Research in 2018.

Their approach combined MANAFEST with single-cell sequencing to identify these rare immune cells in six patients with lung cancer, a laborious process that took several years and cost millions of dollars.

The original study showed that the immunotherapy-activated immune cells share a common gene expression profile. In the new study, Zeng, Smith and their colleagues built on those discoveries to develop MANAscore.

“Our model allows us to skip a time-consuming and expensive process to identify the cells targeted by immunotherapy, and will help us identify what distinguishes who will respond to these therapies,” Smith says.

“We’re not the first to come up with one of these models, but what sets ours apart is that it uses only three genes, while the most commonly used model requires more than 200 genes. Ours is simpler and easier to use.”

The team also found key differences in the T cells activated in the tumors of patients who respond to immune checkpoint therapy compared with those who don’t. Responding patients exhibit a higher proportion of stem-like memory T cells, which act as a reservoir for new cells and can develop into many effective anti-tumor cells, Zeng says.

This observation may help explain why patients are more able to respond; the stem-like characteristics may make it easier for the T cells to multiply into an army of tumor-fighting cells. More studies are needed to confirm these observations.

“The stem-like characteristics of T cells are critical because they enable self-renewal and long-term persistence,” Zeng says. “This allows for sustained immune responses and the ability to expand into a robust population of effector T cells when needed.”

In the meantime, the team is working to develop a clinical test that uses multispectral immunofluorescence panels to identify the three-gene signature of immune therapy responding T cells.

“We hope to translate our three-gene signature into a biomarker that clinicians can use to guide cancer care,” Smith says.

Zeng is also using their new model to determine whether the proximity of T cells with the three-gene signature to other types of immune cells, like regulatory T cells, helps control the immune response.

“We want to apply our model to spatial data to learn whether cell-to-cell interactions among tumor-targeting T cells and other cell types affect clinical outcomes,” Zeng says.

She is also collaborating with other laboratories across the country to determine if MANAscore can be used in patients with different types of cancer. They’ve created a database of single-cell sequencing data across cancer types and will use the score to help identify cancer-type-specific responder T cell characteristics.

Researchers at the Johns Hopkins Kimmel Cancer Center and Bloomberg-Kimmel Institute for Cancer Immunotherapy have developed a computer model to help scientists identify tumor-fighting immune cells in patients with lung cancer treated with immune checkpoint inhibitors.

In their study published Feb. 3 in Nature Communications, the team, including first author Zhen Zeng, Ph.D., a bioinformatics research associate at the Kimmel Cancer Center, demonstrated that their three-gene “MANAscore” computer model can identify the immune-cells targeted by immune checkpoint inhibitor therapies. It also helped the team identify differences associated with patient response to immunotherapy.

“We have developed a way to identify the cells directly targeted by immune checkpoint inhibitors, and if we can identify them, we can study them,” says the study’s senior author, Kellie Smith, Ph.D., an associate professor of oncology at Johns Hopkins. “If we can study them, that means we can identify better biomarkers and better targets for combination immunotherapy.”

Immune checkpoint inhibitors like PD-1 inhibitors are available to treat dozens of cancer types. These revolutionary therapies work by unleashing tumor-killing immune cells, called T cells, that are switched off by the protein PD-1.

PD-1 inhibitors turn the patient’s T cells back on, allowing patients’ immune systems to fight cancer more effectively. But not all patients respond to these therapies, and scientists need to know why in order to develop improved therapies that help nonresponders.

“Tumor-active T cells are very important to a patient’s response to therapy, but they are difficult to find,” Zeng says.

Smith helped develop the MANAFEST technology (Mutation-Associated NeoAntigen Functional Expansion of Specific T Cells), which she and colleagues first described in the journal Cancer Immunology Research in 2018.

Their approach combined MANAFEST with single-cell sequencing to identify these rare immune cells in six patients with lung cancer, a laborious process that took several years and cost millions of dollars.

The original study showed that the immunotherapy-activated immune cells share a common gene expression profile. In the new study, Zeng, Smith and their colleagues built on those discoveries to develop MANAscore.

“Our model allows us to skip a time-consuming and expensive process to identify the cells targeted by immunotherapy, and will help us identify what distinguishes who will respond to these therapies,” Smith says.

“We’re not the first to come up with one of these models, but what sets ours apart is that it uses only three genes, while the most commonly used model requires more than 200 genes. Ours is simpler and easier to use.”

The team also found key differences in the T cells activated in the tumors of patients who respond to immune checkpoint therapy compared with those who don’t. Responding patients exhibit a higher proportion of stem-like memory T cells, which act as a reservoir for new cells and can develop into many effective anti-tumor cells, Zeng says.

This observation may help explain why patients are more able to respond; the stem-like characteristics may make it easier for the T cells to multiply into an army of tumor-fighting cells. More studies are needed to confirm these observations.

“The stem-like characteristics of T cells are critical because they enable self-renewal and long-term persistence,” Zeng says. “This allows for sustained immune responses and the ability to expand into a robust population of effector T cells when needed.”

In the meantime, the team is working to develop a clinical test that uses multispectral immunofluorescence panels to identify the three-gene signature of immune therapy responding T cells.

“We hope to translate our three-gene signature into a biomarker that clinicians can use to guide cancer care,” Smith says.

Zeng is also using their new model to determine whether the proximity of T cells with the three-gene signature to other types of immune cells, like regulatory T cells, helps control the immune response.

“We want to apply our model to spatial data to learn whether cell-to-cell interactions among tumor-targeting T cells and other cell types affect clinical outcomes,” Zeng says.

She is also collaborating with other laboratories across the country to determine if MANAscore can be used in patients with different types of cancer. They’ve created a database of single-cell sequencing data across cancer types and will use the score to help identify cancer-type-specific responder T cell characteristics.