Loss of the tumor suppressor LKB1 (STK11) alters tumor metabolism and immune interactions, but clinical identification of LKB1-deficient tumors has largely relied on detecting somatic mutation or deletion. Emerging therapies intend to restore LKB1 pathway activity in tumors that are structurally wild-type but functionally deficient, so a method to identify functional LKB1 loss beyond genomic alteration is needed.
The authors aimed to define a transcriptional readout of LKB1 functional loss that would capture tumors silenced by non-genomic mechanisms and thereby enable broader patient selection for LKB1-directed strategies.
Using lung adenocarcinoma (LUAD) samples with genomically-defined STK11 loss as the training set, the investigators derived a 30-gene transcriptional signature intended to report functional inactivation of LKB1. The signature was validated in an independent squamous cohort, achieving an AUROC of 0.926.
The signature also transferred to four independent non-TCGA LUAD cohorts with AUROC values greater than 0.92. In comparative analyses the 30-gene signature outperformed a previously published 16-gene classifier (Kaufman), indicating improved performance in identifying LKB1 functional loss in these lung cancer data sets.
To test whether the signature reports LKB1 function and not merely mutation presence, the authors performed functional experiments in non–small cell lung cancer (NSCLC) cell lines that harbor LKB1 mutations. Restoring wild-type LKB1 expression in these mutant lines reversed the transcriptional signature, whereas expression of a kinase-dead LKB1 mutant did not reverse it. These results support that the signature reflects LKB1 kinase activity rather than only the underlying genomic status.
When applied across cancer types (pan-cancer), the transcriptional signature identified functional LKB1 loss in tumors that are genomically wild-type at a rate of 7.2%. Combining genomic loss with signature-positive wild-type cases expanded the total LKB1-loss prevalence from 2.7% (genomic loss alone) to 9.9% overall — a 3.7-fold increase in the population considered LKB1-deficient by function.
The largest absolute expansions in signature-positive, genomically wild-type tumors occurred in esophageal, colorectal, endometrial and cutaneous cancers, indicating that functional LKB1 loss is not confined to tumor types with high STK11 mutation frequency.
The signature-positive population did not simply mirror LKB1 mutation or deletion frequency; rather, it exhibited selective enrichment within specific molecular subgroups. Examples highlighted by the authors include BRAF-mutant colorectal cancers, where 25% were signature-positive, and HER2-amplified breast cancers, where 21% were signature-positive. The authors propose these readily-testable subgroups as practical screening targets.
In lung adenocarcinoma, signature positivity increased across a co-mutation gradient. The reported rates were a 6% background prevalence, 15% in KRAS-mutant tumors, 35% in tumors with NRF2-pathway mutations (KEAP1 or NFE2L2), and 64% in tumors bearing both KRAS and NRF2-pathway mutations. This gradient resembles known biology of KRAS/LKB1 co-mutant tumors and suggests functional LKB1 loss often co-occurs with specific oncogenic contexts.
The study reports that LKB1 wild-type but signature-high tumors in lung adenocarcinoma are immune-cold. Across multiple independent lung cancer cohorts, signature positivity correlated negatively with antigen-presentation programs, interferon-γ response, and T-cell–inflamed gene programs. This immune-cold profile parallels the phenotype described for KRAS/LKB1 co-mutant tumors and may have implications for responsiveness to immune-based therapies.
By identifying a broader, functionally defined population of LKB1-deficient tumors, the transcriptional signature could expand the set of patients eligible for therapies aimed at restoring LKB1 pathway activity or otherwise targeting vulnerabilities associated with LKB1 loss. The signature’s enrichment in practical subgroups (for example, BRAF-mutant colorectal and HER2-amplified breast cancers) suggests immediate opportunities for targeted screening.
Because the signature reports kinase function (as shown by reversal with wild-type but not kinase-dead LKB1), it may help distinguish tumors likely to respond to therapies that depend on intact LKB1 kinase reactivation.
The authors reference a public repository for code and data (URL provided in the preprint footnotes). This work is presented as a preprint and has not undergone peer review. A competing interest is declared: one author (JDG) holds a patent for therapeutics relevant to LKB1 restoration.
Note: All details above are taken from the preprint; methodological specifics beyond what is reported in the abstract and author notes (for example, exact gene list, statistical model particulars, cohort sizes beyond those summarized, or functional assay protocols) were not reproduced here because full details are provided in the source preprint and associated data/code links.