Cervical cancer may need more than a single therapeutic strategy, according to a new study that combines large-scale gene-expression analysis with laboratory testing to identify cell-surface proteins suitable for more precisely targeted antibody-drug conjugates. The work suggests that different cervical cancer subtypes—and even smaller molecularly defined groups within those subtypes—may require different targets. The researchers identified mesothelin, or MSLN, as a broadly applicable candidate across cervical cancers, while TROP-2 appeared more closely associated with adenocarcinoma and LIV-1 emerged as a potential target for a much smaller high-expression subgroup. The findings offer a framework for moving beyond the traditional model of selecting one target for an entire cancer type and instead matching recombinant immunotherapeutics to tumor biology.
Cervical cancer remains a major health burden among women of reproductive age, despite improvements in screening, vaccination, chemotherapy, radiotherapy, targeted treatments, and immunotherapy. Patients with metastatic, recurrent, or treatment-resistant disease still face limited options, in part because cervical tumors are not biologically uniform. Squamous cell carcinoma, which develops from squamous epithelial cells, and adenocarcinoma, which arises from glandular tissue, differ in their cellular origins, molecular features, and patterns of protein expression. This heterogeneity can make therapeutic target discovery difficult: a protein that appears abundant in a mixed tumor cohort may be highly expressed in only one subtype, while a promising target may be missed entirely if it is present in a relatively small patient subgroup.
To address this problem, the researchers developed a multi-stage computational pipeline that reanalyzed transcriptomic data from 304 primary cervical cancer tumors and 7,597 healthy tissue samples representing 52 tissue types. Transcriptomics measures the RNA molecules produced by cells and provides an indirect view of which genes are active. The team focused on 259 high-confidence genes encoding proteins located at the cell surface, where they can be reached by antibody-based drugs. The first comparison examined cervical tumors against a broad collection of normal tissues and identified 30 significantly overexpressed candidates. This initial step provided a wide screening net, but the investigators then tested whether the apparent tumor-associated expression remained selective when each cancer subtype was compared with the normal tissue from which it originated.
That tissue-of-origin analysis changed the status of several candidates. The researchers found that MSLN was the most consistent candidate across the cervical cancer samples, retaining strong overexpression when tumors were compared separately with normal ectocervical and endocervical tissue. This consistency makes MSLN a potential pan-cervical cancer target, although broad expression in a tumor does not automatically guarantee a safe or effective treatment window. TROP-2 showed a more specialized pattern. It had the second-highest global fold change and was detected in every tumor examined, but it was not selectively elevated relative to normal ectocervix. Against normal endocervix, however, TROP-2 displayed strong differential expression, supporting its possible use as an adenocarcinoma-selective target rather than a universal target for all cervical cancers.
The analysis also demonstrated how average expression can conceal a clinically relevant subgroup. LIV-1 did not appear among the top 30 candidates in the global comparison, and its overall fold change was only 0.97. If the researchers had relied solely on cohort-wide averages, the protein might have been discarded. Instead, they examined the distribution of expression among individual tumors and identified 14 high-expressing cases, representing approximately 4.6 percent of the cohort. These tumors had a median LIV-1 expression of 122.7 transcripts per million, or TPM, and a fold change of 3.07. TPM is a normalized measure used to compare transcript abundance between samples. The result suggests that LIV-1 may be valuable not as a broadly applicable cervical cancer target, but as a precision-medicine option for patients whose tumors belong to a distinct LIV-1-high subgroup.
The investigators then tested whether the computational predictions translated into functional activity. They produced recombinant single-chain variable fragment, or scFv, fusion proteins directed against MSLN, TROP-2, and LIV-1. An scFv is a compact antibody-derived binding unit that retains antigen recognition without requiring the full antibody structure. The constructs were linked to SNAP-tag technology, which allows a protein to be chemically coupled to selected molecular payloads. After conjugation with auristatin F, a potent microtubule-disrupting cytotoxic compound, the resulting antibody-drug conjugate-like constructs were evaluated for binding and cell-killing activity in cervical cancer cell lines. The strategy is designed to concentrate the toxin in antigen-bearing cells, reducing exposure to cells that lack the target, although the selectivity observed in cell culture must ultimately be confirmed in more complex biological systems.
TROP-2-directed constructs produced particularly strong activity in the squamous cell carcinoma-derived CaSki and SiHa cell lines, with half-maximal inhibitory concentration, or IC50, values of 7.0 nanomolar and 17.9 nanomolar, respectively. The IC50 is the concentration required to reduce measured cell viability by 50 percent; lower values generally indicate greater potency under the conditions of the experiment. The same construct also affected the adenocarcinoma-derived HeLa line, but required a higher concentration, consistent with the possibility that target abundance, internalization, or other cellular factors influence sensitivity. MSLN-directed constructs bound the tested cervical cancer cell lines at levels ranging from 43.6 percent to 99.4 percent and showed selective nanomolar cytotoxicity. These observations support the idea that surface abundance and drug response can be related, but they also show that expression alone is unlikely to explain every difference between cell lines.
The LIV-1 experiments provided a notable link between the subgroup analysis and the laboratory results. CaSki and SiHa cells showed stronger responses to the LIV-1-targeted constructs, whereas activity was substantially lower in HeLa and ME180 cells. This pattern was consistent with the computational proposal that only a minority of cervical tumors may express LIV-1 at sufficiently high levels to support targeted treatment. The result is important because it illustrates a central challenge in precision oncology: a target can be clinically meaningful even when it is uncommon, provided that patients can be reliably identified and the target provides enough tumor selectivity. In practice, that would require a validated diagnostic test, likely based on protein-level measurements or another assay capable of determining whether LIV-1 is present on the surface of an individual patient’s tumor cells.
Together, the findings establish a three-layered approach to target prioritization: broad screening of cell-surface genes, refinement using the appropriate tissue-of-origin comparison, and analysis of expression variation within the tumor population. MSLN emerged from this process as a candidate with broad cervical cancer coverage, TROP-2 showed a stronger association with an adenocarcinoma-related context, and LIV-1 represented a possible target for a smaller biologically defined population. The researchers caution that the work remains preclinical. RNA abundance does not always predict the amount of protein displayed on a cell surface, and cell lines do not reproduce the architecture, immune environment, stromal interactions, or treatment history of human tumors. Future studies using patient-derived organoids, single-cell RNA sequencing, surface proteomics, and patient-derived samples will be needed to establish how frequently these targets occur, which tumor cells express them, and whether the proposed therapeutic windows persist in realistic models. Even with those limitations, the study presents a potentially influential blueprint for developing recombinant immunotherapeutics that recognize the diversity of cervical cancer rather than treating the disease as a single molecular entity.
Subject of Research: Cells
Article Title: Multi-database transcriptomic screening to identify subtype-selective cell surface targets for development of SNAP-tag based immunotherapeutics
Web References: https://doi.org/10.70401/10.70401/cbm.2026.0023; Computational Biomedicine
References: Matshoba T, Bolotnikov V, Henry M, Lekena N, Hunter R, Barth S. “Multi-database transcriptomic screening to identify subtype-selective cell surface targets for development of SNAP-tag based immunotherapeutics.” Computational Biomedicine. DOI: 10.70401/10.70401/cbm.2026.0023.
Image Credits: © Thabo Matshoba, Viacheslav Bolotnikov, Marc Henry, Nkhasi Lekena, Roger Hunter, Stefan Barth, 2026. Licensed under a Creative Commons Attribution 4.0 International License.
Keywords: cervical cancer, antibody-drug conjugates, ADCs, precision medicine, MSLN, TROP-2, LIV-1, transcriptomics, scFv-SNAP fusion proteins, auristatin F, tumor heterogeneity, recombinant immunotherapeutics
