Matteo Di Bernardo

Directed protein evolution is central to biomedical applications but faces challenges such as experimental complexity, inefficient multiproperty optimization, and local maxima traps. Although in silico methods that use protein language models (PLMs) can provide modeled fitness landscape guidance, they struggle to generalize across diverse protein families and map to protein activity. We present EVOLVEpro, a few-shot active learning framework that combines PLMs and regression models to rapidly improve protein activity. EVOLVEpro surpasses current methods, yielding up to 100-fold improvements in desired properties. We demonstrate its effectiveness across six proteins in RNA production, genome editing, and antibody binding applications. These results highlight the advantages of few-shot active learning with minimal experimental data over zero-shot predictions. EVOLVEpro opens new possibilities for artificial intelligence-guided protein engineering in biology and medicine.

Abstract Mitochondria provide the first line of defense against the tumor-promoting effects of oxidative stress. Here we show that the prostate-specific homeoprotein NKX3.1 suppresses prostate cancer initiation by protecting mitochondria from oxidative stress. Integrating analyses of genetically engineered mouse models, human prostate cancer cells, and human prostate cancer organotypic cultures, we find that, in response to oxidative stress, NKX3.1 is imported to mitochondria via the chaperone protein HSPA9, where it regulates transcription of mitochondrial-encoded electron transport chain (ETC) genes, thereby restoring oxidative phosphorylation and preventing cancer initiation. Germline polymorphisms of NKX3.1 associated with increased cancer risk fail to protect from oxidative stress or suppress tumorigenicity. Low expression levels of NKX3.1 combined with low expression of mitochondrial ETC genes are associated with adverse clinical outcome, whereas high levels of mitochondrial NKX3.1 protein are associated with favorable outcome. This work reveals an extranuclear role for NKX3.1 in suppression of prostate cancer by protecting mitochondrial function. Significance: Our findings uncover a nonnuclear function for NKX3.1 that is a key mechanism for suppression of prostate cancer. Analyses of the expression levels and subcellular localization of NKX3.1 in patients at risk of cancer progression may improve risk assessment in a precision prevention paradigm, particularly for men undergoing active surveillance. See related commentary by Finch and Baena, p. 2132. This article is highlighted in the In This Issue feature, p. 2113

BackgroundThe estimated incidence of colorectal cancer is rising in Nigeria, where most patients present with advanced disease. Earlier detection of colorectal cancer is a goal of the Nigerian National Cancer Control Plan, but the utility of fecal-based screening is unclear. This study aimed to assess the fecal immunochemical test as a colorectal cancer screening modality in average-risk individualS in Nigeria.MethodsA population-based, cross-sectional study of qualitative fecal immunochemical test-based colorectal cancer screening was done in asymptomatic, average-risk participants aged 45–75 years in three states in Nigeria (Osun, Kwara, and Lagos). Participants were invited to enrol using age-stratified and sex-stratified convenience sampling following community outreach. Exclusion criteria included a personal history of colorectal cancer or rectal bleeding in the previous 6 months, a first-degree relative with a known diagnosis of colorectal cancer, or a comorbidity that would preclude conscious sedation or general anesthesia. Participants with positive fecal immunochemical test results underwent colonoscopy, and the positive predictive value of fecal immunochemical testing for colorectal cancer and advanced adenomas (≥10 mm, tubulovillous or villous or high-grade dysplasia) was calculated. Data on demographics and acceptability of fecal immunochemical testing and colonoscopy were collected.FindingsBetween January and April 2021, 2330 participants were enrolled in the study and received a fecal immunochemical test, which was returned by 2109 participants. 1677 participants tested negative and 432 tested positive. Of these 432 participants, 285 underwent a colonoscopy (235 showed no polyps or cancer, 47 had polyps identified, and three had colorectal cancer identified). Of the 47 participants who had polyps identified, 20 had advanced adenomas diagnosed. The median age was 57 years (IQR 50–63), 958 (41%) were male and 1372 (59%) were female, and 68% had at least a secondary-level education. Participants were evenly spread across wealth quintiles. The positivity rate of the fecal immunochemical test was 21% overall (432 of 2109; 95% CI 20–21%), 11% (51 of 455; 95% CI 10–12) in Lagos, 20% (215 of 1052; 95% CI 20–21) in Osun, and 28% (166 of 597; 95% CI 27–29) in Kwara. Among the patients with a positive fecal immunochemical test who completed colonoscopy, the positive predictive value for invasive colorectal cancer was 1·1% (95% CI 0·3–3·3), and 7·0% (4·5–10·8) for advanced adenoma. The acceptability of fecal immunochemical screening among participants was very high.InterpretationColorectal cancer screening with qualitative fecal immunochemical tests in Nigeria is feasible and acceptable to average-risk asymptomatic participants. However, the low positive predictive value for advanced neoplasia and high endoscopy burden investigating false positives suggests it might not be an appropriate screening tool in this setting.FundingThompson Family Foundation, Prevent Cancer Foundation, National Institutes of Health/National Cancer Institute Program Cancer Center.

Abstract To study the progression of bladder cancer from non–muscle-invasive to muscle-invasive disease, we have developed a novel toolkit that uses complementary approaches to achieve gene recombination in specific cell populations in the bladder urothelium in vivo, thereby allowing us to generate a new series of genetically engineered mouse models (GEMM) of bladder cancer. One method is based on the delivery of adenoviruses that express Cre recombinase in selected cell types in the urothelium, and a second uses transgenic drivers in which activation of inducible Cre alleles can be limited to the bladder urothelium by intravesicular delivery of tamoxifen. Using both approaches, targeted deletion of the Pten and p53 tumor suppressor genes specifically in basal urothelial cells gave rise to muscle-invasive bladder tumors. Furthermore, preinvasive lesions arising in basal cells displayed upregulation of molecular pathways related to bladder tumorigenesis, including proinflammatory pathways. Cross-species analyses comparing a mouse gene signature of early bladder cancer with a human signature of bladder cancer progression identified a conserved 28-gene signature of early bladder cancer that is associated with poor prognosis for human bladder cancer and that outperforms comparable gene signatures. These findings demonstrate the relevance of these GEMMs for studying the biology of human bladder cancer and introduce a prognostic gene signature that may help to stratify patients at risk for progression to potentially lethal muscle-invasive disease. Significance: Analyses of bladder cancer progression in a new series of genetically engineered mouse models has identified a gene signature of poor prognosis in human bladder cancer.