Transforming growth factor-β (TGF-β) signaling members, TGF-β receptor type II (TBRII), Smad2, Smad4 and Smad adaptor, embryonic liver fodrin (ELF), are prominent tumor suppressors in gastrointestinal cancers. Here, we show that 40% of elf^+/− mice spontaneously develop hepatocellular cancer (HCC) with markedly increased cyclin D1, cyclin-dependent kinase 4 (Cdk4), c-Myc and MDM2 expression. Reduced ELF but not TBRII, or Smad4 was observed in 8 of 9 human HCCs (P<0.017). ELF and TBRII are also markedly decreased in human HCC cell lines SNU-398 and SNU-475. Restoration of ELF and TBRII in SNU-398 cells markedly decreases cyclin D1 as well as hyperphosphorylated-retinoblastoma (hyperphosphorylated-pRb). Thus, we show that TGF-β signaling and Smad adaptor ELF suppress human hepatocarcinogenesis, potentially through cyclin D1 deregulation. Loss of ELF could serve as a primary event in progression toward a fully transformed phenotype and could hold promise for new therapeutic approaches in human HCCs.

MLL2, an epigenetic regulator in mammalian cells, mediates histone 3 lysine 4 tri-methylation (H3K4me3) through the formation of a multiprotein complex. MLL2 shares a high degree of structural similarity with MLL, which is frequently disrupted in leukemias via chromosomal translocations. However, this structural similarity is not accompanied by functional equivalence. In light of this difference, and previous reports on involvement of epigenetic regulators in malignancies, we investigated MLL2 expression in established cell lines from breast and colon tissues. We then investigated MLL2 in solid tumors of breast and colon by immunohistochemistry, and evaluated potential associations with established clinicopathologic variables.

Introduction: The growing presence of artificial intelligence (AI) in health professions has created a need to investigate its potential benefits and challenges in medical education. This article presents findings from an AI learner training needs analysis survey at a U.S. medical school. It compares faculty and student experiences and perspectives on using generative AI (GAI) and other AI tools for undergraduate medical education, focusing on their respective knowledge and learning preferences. Methods: Faculty and students were surveyed using an online cross-sectional survey design to assess their GAI experience, AI patterns of use, adoption readiness, and training preferences. Surveys contained 14 to 15 multiple-choice items, with 8 items including a write-in option. A total of 68 faculty and 506 students responded to the survey, with a 50% response rate for faculty and 30% for students. Statistical tests were used to determine whether students and faculty differed significantly in their GAI experience. Results: We found that students were significantly more familiar with GAI than faculty ( P < .001) but not significantly more experienced with GAI tools. There were no significant differences in frequency of use. Both groups considered AI tools and technology useful for personal, academic, research, and clinical applications. More than half of both groups were using AI for academic tasks. Both groups expressed concerns about the reliability of AI output, with faculty showing a much greater level of concern. Both groups identified several training formats as beneficial, with faculty preferring formal training (either online or in-person), followed by peer tutorials and self-study. On the other hand, students showed slightly greater interest in self-study than other formats. Conclusion: Our findings will inform the design of two parallel structured AI training programs, focusing on faculty and student priorities, including hands-on skills practice, and emphasizing AI's ethical use, reliability, and limitations.

Several different cancer studies have indicated that lymphocyte mutagen sensitivity is a marker of DNA repair deficiency and increased cancer risk. We have used a mutagen sensitivity assay (MSA) measuring γ-radiation-induced chromosomal aberrations in freshly cultured lymphocytes and assessed breast cancer risk in African-American women. Concurrently, we conducted duplicate cultures in the presence of caffeine, which overrides G₂ arrest in cultured cells, decreases time to DNA repair, and hence increases the aberration rate. In comparison with the non–caffeine-treated cells, we are conceptually segregating the contribution of DNA repair and time for DNA repair as individual susceptibility phenotypes. Blood samples were obtained from 61 cases and 86 controls at Howard University Hospital. Two sets of whole-blood cultures were established and γ-irradiated (1 Gy) at 67 hours, one of which was treated with caffeine (1 mg/mL). Thereafter, cultures were processed for obtaining metaphase spreads. Fifty metaphases were screened for chromatid breaks. The mean breaks per cell (MBPC) for cases (0.34 ± 0.15) was significantly greater than for controls (0.24 ± 0.12; P < 0.0001). Using the 75th percentile value of controls as a cutoff to define mutagen sensitivity, the sensitive individuals had an odds ratio of 4.5 (95% confidence intervals, 2.2-9.1) for breast cancer compared with individuals that were not sensitive. The adjusted odds ratio was 3.3 (95% confidence intervals, 0.147-73.917), which was statistically significant but was limited by the small number of subjects. The results for caffeine co-culture were not predictive of breast cancer (MBPC: cases, 1.6 ± 0.9 versus controls, 1.5 ± 0.8; P = 0.8663). Comparing the MBPC for caffeine and noncaffeine cultures, there was a correlation in controls (n = 79; Spearman r = 0.4286; P < 0.0001), but not in cases (n = 58; Spearman r = 0.06609; P = 0.6221). This study indicates that the MSA phenotype is a risk factor for breast cancer in African-American women, with a significant effect observable even in small studies. The use of caffeine did not enhance the predictivity of MSA, but the correlation with non-caffeine cultures in controls indicates that the MSA phenotype is due to both DNA repair and G₂ arrest capacity. (Cancer Epidemiol Biomarkers Prev 2006;15(3):437–42)