Cancer Researchers Share Cutting Edge Approaches to Combat Tumors

Cancer treatments have come a long way since the early days of surgery and radiation. While both are still routinely used as part of treatment regimens, today, physicians can target specific cancerous cells with engineered T cells and small molecules alike, improving tumor targeting and reducing patient side effects.

Alongside these advancements, though, many challenges remain in the field of cancer therapy. Researchers across institutions shared how their groups are tackling some of these obstacles, ranging from combatting therapy resistance to improving diagnostics, to overcome cancer in the plenary session “Precision, Partnership, Purpose: Advance Cancer Science to Save Lives Globally” at the American Association for Cancer Research (AACR) 2026 annual meeting.

Leveraging the Mechanism of Cell Lineage Shifts to Stop Tumor Therapy Evasion

As researchers started developing successful targeted therapies against cancers, they found that the tumors started evading the drugs through acquired resistance mechanisms. As Charles Sawyers, a physician scientist at the Memorial Sloan Kettering Cancer Center, explained, scientists investigated the underpinnings of these evasion strategies to develop new treatments.

One of these new tumor escape routes came in the form of the shifting expression of lineage markers, or genes that define cell populations. Sawyers and his group have explored this phenomenon in prostate cancers, which typically present as adenocarcinomas. They studied how androgen receptor inhibitor therapy shifted the tumor from this lineage to that resembling neuroendocrine cancer through changes in transcription factor expression.

Sawyers and his team developed mouse and organoid models to track down the molecular steps that the tumors take to influence this lineage shift.^1,2 The researchers found that some of the cancer cells are epigenetically poised to reduce integrin signaling with the extracellular matrix, which leads them to their lineage shift after loss of the gene for retinoblastoma 1, a tumor suppressor gene.³ In particular, Sawyers showed that one histone acetyl transferase, TIP60, identified by a CRISPR and coactivator screen, was a critical switch in this process. His team degraded this acetyl transferase or its target, the oncogene myc, to promote killing these lineage-shifted tumors in vitro.

In summary, Sawyers showed that by unraveling the molecular pathways that tumors use to evade chemotherapies, researchers could find new drug targets. He shared how with this information and pharmacological mediators researchers could shift these tumors toward one lineage or another to influence tumor treatment.

Improving CAR T Cell Therapy for Solid Tumors

Next, Carl June, a physician scientist at the University of Pennsylvania, shared his group’s work on improving CAR T cell therapy against solid tumors. While this immunotherapy has seen great success against blood cancers, June explained that CAR T cells are less effective against solid tissue cancers because these tumors have more established tumor microenvironments that suppress the therapeutic cells. He referenced one study that targeted CAR T cells against tight junction proteins in gastroesophageal cancer and had good response rates, but the therapy only extended survival by two months.⁴ “So, we have a foot in the door now in solid cancers with CAR cells, but we need durability to be improved,” June said.

To address this issue, June shared how his group has explored different strategies to improve CAR T cell activity against solid tumors. In trials against glioblastoma, June’s group developed CAR T cells that targeted a receptor for an epithelial growth factor and the receptor for interleukin-13.⁵ They also switched to delivering this therapy directly into the cerebrospinal fluid (CSF).

“We found, for the first time now, very encouraging findings on imaging,” June said about the team’s pilot study that showed the CSF delivery and dual-targeting decreased tumor volumes.⁶ Exploring the biological mechanisms of this effect, his team found an increase in natural killer cells in the tumor microenvironment (TME) and reduced regulatory T cells in patients who responded to the treatment.

Separately, June shared his group’s work targeting the tumor associated antigen mesothelin. They identified two mechanisms of resistance in pancreatic cancer: decreased TME infiltration and rapid CAR T cell exhaustion.

Continue reading below...

Webinar

Overcoming Immunotherapy Resistance in Liver Cancer

Explore emerging mechanisms of immune checkpoint inhibitor resistance in hepatocellular carcinoma (HCC) and how combination immunotherapy strategies may improve antitumor immune responses and patient-specific treatment approaches.

Read More

To tackle the problem of CAR T cell entry into the TME, June and his collaborators explored ways to decrease signaling from carcinoma-associated fibroblasts (CAFs) in the TME that promote tumor growth. Sharing one colleague’s hypothesis, June said, “If we can de-CAF-einate, if you will, the tumor microenvironment, we may then be able to improve trafficking of CAR T cells into the tumor microenvironment.” The researchers developed CAR T cells that targeted a protein on these CAFs, and they showed that eliminating CAFs with these CAR T cells prior to administering their mesothelin-specific CAR T cells improved the therapeutic outcome in mice.^7,8

Next, June and his team turned their attention to CAR T cells exhaustion. Using mesothelin-specific CAR T cells, the group found that continuously exposing these cells to their target antigen shifted their gene expression to inactive, with similar responses in patients treated with CAR T cells.^9,10 Although previous studies, including from June’s group, showed improved responses in mice and patients by deleting one or two genes related to this phenomenon, June said that no one had yet explored the impact of using synergistic gene disruptions.^11-14 His team conducted CRISPR screens that could knock out two genes at once with the goal of finding synergistic pairs. They showed some preliminary, unpublished findings of these cooperative gene knockouts in mesothelin-specific CAR T cells, which demonstrated improved tumor killing in pancreatic tumors in vivo.

June’s team plans to continue exploring how to improve CAR T cell responses in solid tumors through these multiple approaches.

Manipulating Protein Interactions to Enhance Targeting

In cancer therapy, most strategies rely on targeting specific proteins. However, Georg Winter, a chemical biologist at AITHYRA, Research Institute for Biomedical Artificial Intelligence of the Austrian Academy of Sciences, outlined that small molecules can target only about 20 percent of human proteins. He said that part of the reason for this limit is due to targeting strategies that most drug developers take, which rely upon the protein having specific features that can be bound and inhibited by a target, leaving the majority of the proteome “undruggable” by this approach.

Winter outlined how his group seeks to use small molecules that will bind to their target and, instead of outright inhibiting the protein, alter its function via a protein-protein interaction for a therapeutic goal. One of these strategies relies upon using a small molecule to bring a target protein into proximity of a ubiquitin ligase to promote its degradation.¹⁵ Winter and his team showed how they previously used one such molecule to promote degradation of a target protein in leukemia tumors.

However, it is difficult to bind a target with intrinsically disordered regions using this approach. In response, other groups have used so-called “molecular glue degraders,” which bind to a ligase and alter its surface structure to promote protein-protein interactions with a previously “undruggable” protein.^{16, 17}

“With those studies and related studies, the field was really presented with almost a holy grail blueprint,” Winter said. He and his group have focused on identifying these molecular glue degraders. In one study, after mapping genes required for these degraders to work, they found one such molecule that bound the protein cyclin K, another undruggable target to promote its degradation.^{18, 19} In another set of experiments, they explored a different molecular glue degrader that bound BRD4, a transcription factor regulator in some cancers. However, instead of binding directly to an active ligase to promote BRD4 degradation, they found that this molecule bridged two domains of BRD4 that promoted its binding to the protein’s ligase.²⁰ This demonstrated a novel type of function for these molecular glue degraders that the team also found worked with other enzymes as well.

Winter shared another example of identifying a novel function of molecular glue degraders with a study that came from an observation that many small molecules designed as protein inhibitors also lead to their targets’ degradation. His team showed that these compounds achieve this by locking a target protein into a structure that will be recognized by its normal ligase to promote its degradation, as opposed to bringing the protein into proximity of a ligase bound to the molecular glue degrader.^{21, 22}

Next, Winter discussed how his group is exploring the mechanisms of action of currently used compounds that scientists thought worked as molecular glue degraders but that actually function in distinct ways, opening up the potential for using this knowledge to create new drugs. In unpublished experiments using genome-wide screens and transcriptional profiling, Winter’s team showed that an FDA-approved degrader against estrogen receptors actually worked by promoting the post-translational modification of the target protein that recruited transcriptional repressors. They found that several other cancer drugs work by a similar principle.

Lastly, Winter introduced how most cancer drugs act by disconnecting pathways in cancer cells so that proteins lose their functions. “What could we do if, instead of rewiring an effector to eliminate the target, we could just directly rewire the target,” Winter proposed.

To pursue this idea, he and his team created a bifunctional molecule that bound a transcriptional repressor of pro-apoptotic genes on one end and a chromatin remodeling complex, BAF, on the other. They showed that this molecule recruited the BAF complex to where the transcriptional repressor was bound, leading to removal of the repressor from the apoptotic genes and the addition of active chromatin markers to the area.²³ Subsequently, the researchers saw the expression of these genes. In a tumor model, this approach killed B cell lymphoma cells more effectively than inhibiting the repressor alone.

Continue reading below...

Webinar

Mapping Clonal Mosaicism in Aging Tissues

Explore how single-cell genotype-to-phenotype mapping can reveal clonal architecture and the functional consequences of somatic mutations in aging human tissues.

Read More

Looking ahead, Winter said, “I think really the rate limiting step in the field here is our imagination and our creativity.”

Exploring the Potential for AI in Cancer Therapy and Detection

In the final talk of the session, computer scientist Regina Barzilay from the Massachusetts Institute of Technology discussed how AI can improve cancer diagnostics and treatment decisions. She explained how current methods of determining treatment efficacy cannot predict who a therapy will be most suited for. The data typically needed for these evaluations—including spatial transcriptomics, single-cell sequencing, and more—though, isn’t typically collected in the clinical setting.²⁴ Additionally, while AI tools trained on current clinical data, like pathology slides, can predict an outcome, they aren’t able to determine what the biological mechanism of action is for a given intervention.²⁵

To address this limitation, Barzilay trained a machine learning model with imaging data from pathology slides and spatial transcriptomic data from the same images.²⁶ The model, called SPARC, could not only predict patient outcomes but also determine the molecular pathways active at different areas of the tissue sample after looking at just the pathology slide. Barzilay showed how spatial transcriptomics data improved AI predictions compared to models that were only trained on imaging data. She said that this dual-trained model could aid in determining what cellular programs are causing a specific response and help clinicians decide on intervention strategies for patients.

Barzilay then showed a model, Boltz, that her group developed. It evaluates the most effective small molecule inhibitor for a patient based on their specific protein sequence and explores possible off-target effects of drug candidates, potentially aiding in clinical diagnostics.²⁷ “The idea here is not to eliminate experiments. It just pushes your experiments to the most promising part of the space,” she said.

Finally, she showed how another AI tool that her group developed, Mirai, can detect breast cancer sooner than standard evaluation by a radiologist.^28,29 She and her team trained this model on data from mammogram images leading up to a breast cancer diagnosis, where the human eye can often miss early evidence of cancer. “The machine, which is trained on the right data, can actually detect these very subtle signs,” Barzilay said.

Barzilay shared that Mirai is currently being used in several clinical trials around the world. She also showed a new, yet unpublished, tool called MammoGen, that generates a potential future mammogram image from a current one, comparing it to aging simulators. Talking about its potential, Barzilay said, “It can help us because it can tell us how aggressive likely the cancer is going to be.”

Collectively, the talks showcased how research across cancer biology and drug discovery aims to improve cancer diagnosis and therapy.