Daily Endocrinology Research Analysis
Analyzed 34 papers and selected 3 impactful papers.
Summary
Three impactful endocrinology papers span basic, prognostic, and digital health advances: mechanistic dissection of PPARG2 enhancer networks in adipogenesis, a large prospective cohort showing Bone Strain Index (BSI) improves 10-year fracture risk prediction beyond BMD and FRAX, and a temporal retinal foundation model (RETFound Plus) that enhances 5-year systemic disease risk prediction including diabetes. Together, they refine molecular targets, risk stratification, and scalable screening pathways.
Research Themes
- Enhancer architecture governing adipogenesis (PPARG2)
- Fracture risk stratification beyond aBMD and FRAX
- Temporal retinal AI for systemic cardiometabolic risk
Selected Articles
1. Extensive enhancer crosstalk controls PPARG2 activation during adipogenesis.
Using systematic enhancer deletions across the PPARG locus, the study reveals cis enhancer crosstalk that stabilizes C/EBPβ recruitment prior to chromatin remodeling and identifies the super-enhancer element E+102 as obligate for PPARG2 activation. Cardiometabolic trait-associated noncoding variants map to E+102 and other essential enhancers, highlighting causal regulatory architecture in human adipogenesis.
Impact: Defines an obligate enhancer and inter-enhancer crosstalk driving PPARG2, the master regulator of adipogenesis, linking noncoding variation to function. This advances mechanistic understanding with direct relevance to cardiometabolic genetics.
Clinical Implications: Clarifying PPARG regulatory architecture could inform interpretation of GWAS signals, enable functional prioritization of noncoding variants, and guide development of therapies that modulate enhancer–TF interactions to treat obesity and insulin resistance.
Key Findings
- Systematic deletion of nine enhancers uncovered cis enhancer crosstalk stabilizing C/EBPβ before chromatin remodeling.
- The super-enhancer constituent E+102 is obligate for activating PPARG expression and mediates dual roles in crosstalk and feedback.
- Cardiometabolic trait-associated noncoding variants map to E+102 and other essential enhancers, supporting physiological relevance.
Methodological Strengths
- CRISPR-based systematic enhancer deletions across upstream, promoter-proximal, and downstream super-enhancer elements
- Integration of functional readouts with TF recruitment dynamics to delineate causal regulatory mechanisms
Limitations
- Findings derived from in vitro human mesenchymal stem cell adipogenesis without in vivo validation
- Direct causal links to human metabolic disease phenotypes were not tested
Future Directions: Functionally test human cardiometabolic GWAS variants at E+102 using base editing; validate enhancer dependencies in primary adipose depots and in vivo models; explore pharmacologic modulation of enhancer–TF interactions.
Transcriptional master regulators drive cell fate transitions. Peroxisome proliferator-activated receptor γ (PPARγ) is the master regulator of adipogenesis, and its expression must therefore be tightly regulated and efficiently induced in response to adipogenic cues. Here we decipher the regulatory mechanisms of the highly connected enhancer community driving activation of the PPARG locus during adipocyte differentiation of human mesenchymal stem cells. By systematically deleting nine individual enhancers, spanning upst
2. Bone strain index improves 10-year fracture risk prediction beyond bone mineral density and FRAX® in community-dwelling older women: A study of the FRISBEE cohort.
In 3,338 older women followed for 5 and 10 years, higher BSI independently predicted incident fragility and major osteoporotic fractures beyond age, aBMD, and FRAX. Total hip BSI performed best (AUC 0.63), and each 1-SD increase conferred higher fracture risk that remained significant after FRAX adjustment.
Impact: Provides large-scale, prospective evidence that a DXA-derived finite element metric (BSI) improves fracture prediction beyond current standards, supporting near-term integration into risk assessment workflows.
Clinical Implications: BSI could refine fracture risk stratification in older women and inform treatment thresholds, particularly when aBMD and FRAX provide borderline risk estimates.
Key Findings
- BSI independently predicted 5- and 10-year fragility and major osteoporotic fractures after adjusting for age, aBMD, and FRAX.
- Total hip BSI showed the strongest discrimination (AUC 0.63 at 5 and 10 years), outperforming lumbar spine BSI.
- Each 1-SD increase in total hip BSI raised fracture risk (HR 1.42 at 5 years; 1.41 at 10 years), remaining significant after FRAX adjustment (HR 1.24).
Methodological Strengths
- Prospective cohort with 5- and 10-year follow-up and radiographic validation of fractures
- Multivariable models adjusting for age, aBMD, and FRAX to establish incremental value
Limitations
- Moderate discrimination (AUC ~0.63) indicating room for improvement
- Generalizability may be limited to similar populations and DXA/analysis platforms
Future Directions: Prospective impact studies to test BSI-guided treatment decisions; integration with trabecular bone score and clinical biomarkers; external validation across devices and ethnicities.
OBJECTIVES: The Bone Strain Index (BSI), derived from finite element analysis of DXA, quantifies mechanical strain within bone. Its predictive value for fragility fractures in older women remains to be established in large prospective studies. METHODS: This prospective cohort study assessed the association between BSI, incident fragility and major osteoporotic fractures (MOF) in 3338 community-dwelling older women (age 64-77 years) during a 5-year and 10-year follow-up. BSI was calculated at the lumbar spi
3. Time and person sensitive foundation model for disease prediction and risk stratification.
RETFound Plus leverages temporal fundus image sequences from >300k individuals to improve 5-year risk prediction and calibration for systemic diseases (diabetes, hypertension, MI, stroke) and ocular conditions. Gains were consistent across external, multi-ethnic datasets, with improved hazard-ratio trends for systemic outcomes.
Impact: Demonstrates progression-aware retinal AI that enhances risk prediction for systemic cardiometabolic diseases at population scale, enabling low-cost, non-invasive stratification pathways that could augment traditional screening.
Clinical Implications: Eye-based risk stratification could identify individuals at elevated 5-year risk for diabetes and cardiovascular events during routine imaging, prioritizing preventive interventions and monitoring without additional tests.
Key Findings
- Temporal modeling on 1,304,292 fundus photographs from 304,345 participants improved 5-year risk prediction and calibration over RETFound.
- Larger c-index gains were observed for systemic diseases (stroke, MI, diabetes, hypertension: +4–10%) versus ocular outcomes (+3–7%).
- Improved risk stratification with 1.2–2.1-fold higher hazard-ratio trend for systemic diseases, consistent across multi-regional, multi-ethnic external datasets.
Methodological Strengths
- Massive multi-visit dataset enabling temporal representation learning and progression-aware modeling
- Extensive external validation across regions and ethnicities with calibration and risk stratification analyses
Limitations
- Observational, retrospective modeling without prospective clinical utility trials
- Potential confounding from systemic–ocular correlations and variable imaging quality across sites
Future Directions: Prospective implementation studies testing clinical decision impact; integration with EHR risk factors; fairness and robustness audits; evaluation of cost-effectiveness in population screening.
Foundation models (FMs) enable generalizable medical AI, but existing retinal FMs perform best on cross-sectional classification and detection and are less effective for predicting disease incidence and progression. We present RETFound Plus, a CFP-based FM trained with temporal modeling on 1,304,292 fundus photographs from 304,345 participants across multiple visits to learn progression-aware representations. Compared with RETFound, RETFound Plus improved calibration and 5-year risk prediction across syst