Researchers have identified three cell-surface proteins with potential as therapeutic targets in cervical cancer, while demonstrating a strategy that could help match patients with antibody-based treatments based on the molecular characteristics of their tumors.
The study, published in Computational Biomedicine, combined gene expression data from hundreds of cervical tumors with laboratory testing of antibody-drug conjugates (ADCs) directed against three potential targets: mesothelin (MSLN), TROP-2, and LIV-1.
For principal investigator Stefan Barth, PhD, of the University of Cape Town, the study builds on a broader question his group has been exploring: What makes a cancer protein a genuinely useful target?
Tumor-associated antigens are proteins found on normal cells but expressed at higher levels on cancer cells. Barth and his colleagues previously developed a transcriptomic approach for comparing cell-surface targets in tumors with their expression in healthy tissues, initially demonstrating the strategy in breast cancer. The broader research program has examined the approach across 26 cancer types.
“We were looking for what we were calling ideal targets,” Barth told Inside Precision Oncology. “So, very low on normal, but very high on diseased.” The new study applies that idea specifically to cervical cancer.
Looking beyond tumor averages
The researchers analyzed gene-expression data from 304 primary cervical cancers and 7,597 healthy tissue samples, initially identifying 30 cell-surface proteins that were significantly more highly expressed in cervical cancer.
But they did not stop with a broad comparison between cancer and healthy tissue. They also compared tumors with the specific normal cervical tissue from which different forms of the cancer arise.
That distinction matters because cervical cancer is not a single, uniform disease. Major subtypes include squamous cell carcinoma and adenocarcinoma, which arise from different tissues and have distinct biological characteristics.
The closer analysis produced different pictures for each of the three targets.
MSLN showed the broadest potential, remaining strongly elevated when tumors were compared with both normal ectocervical and endocervical tissue.
TROP-2 initially also looked like a broadly promising target. Although it was detected across the cervical tumors, it was not elevated compared with normal ectocervical tissue. It was substantially higher than in normal endocervical tissue, suggesting it could be more useful as a target in cervical adenocarcinoma rather than cervical cancer overall.
LIV-1 provided perhaps the clearest example of why researchers may need to look beyond averages. It did not rank among the top 30 targets when the tumors were considered as a single group. When the investigators looked at individual tumors, however, they found 14—about 5% of the cohort—with particularly high LIV-1 expression.
“You might identify antigens which are upregulated in some patients but downregulated in others,” Barth said.
For Barth, the LIV-1 result was also a reminder of the limitations of any broad screening approach. A target that appears unremarkable when researchers average results across hundreds of tumors could still be highly relevant to a smaller group of patients.
Putting the targets to the test
The investigators next asked whether targeting MSLN, TROP-2, and LIV-1 could actually kill cervical cancer cells.
They created experimental ADCs against each protein. ADCs pair an antibody that recognizes a target on a cancer cell with a potent cancer-killing drug, with the goal of delivering the drug more directly to malignant cells.
All three experimental ADCs bound to cervical cancer cells and produced dose-dependent cell killing, although their activity varied among the cell lines tested. TROP-2-targeted ADCs, for example, were particularly active in the squamous cell carcinoma-derived CaSki and SiHa cell lines and also showed activity in adenocarcinoma-derived HeLa cells.
LIV-1-targeted treatment likewise showed stronger activity in some cell lines than others, broadly reflecting the variable expression seen in the computational analysis.
“The findings of our transcriptomic studies [are] confirming this for cervical cancer, and our in vitro activity studies are confirming that you can use them to selectively kill cervical cancer tumor cell lines,” Barth said.
Toward more personalized target selection
Together, the findings suggest a more refined way to search for therapeutic targets: identify proteins that are elevated in tumors, compare them with the appropriate normal tissue, and then look for smaller groups of patients whose tumors may express a particular target at especially high levels.
That last step could be particularly important for precision oncology. Patients diagnosed with the same type of cancer may have very different levels of a particular target—and therefore may not be equally likely to benefit from a therapy directed against it.
“If you…have a screening tool in place that would allow you to identify the patients with the best upregulated cell surface antigens, that’s the precision medicine approach you would need to identify the best patients responding to any type of immunotherapy,” Barth said.
The study remains preclinical. The computational analysis measured messenger RNA rather than the amount of target protein actually present on tumor cells, and the therapeutic experiments were performed in cell lines rather than patients. Barth said moving from transcriptomic analysis toward protein-level data will be important for making target selection more predictive.
Still, the cervical cancer study offers a proof of principle for looking beyond whether a tumor is simply “positive” or “negative” for a particular target. By considering how strongly a target is expressed, what normal tissue it is being compared with, and how expression varies from patient to patient, researchers may be able to more precisely identify who is most likely to benefit from targeted immunotherapies.
