Faculty Bibliography

The rapid evolution of generative artificial intelligence (AI) is poised to transform medicine and medical education. Large language models (LLMs) have begun to demonstrate capabilities in reasoning, diagnosis, documentation, and patient communication that can rival or exceed those of clinicians. In medical education, AI is reshaping how students learn and how faculty teach-offering individualized, context-sensitive guidance at scale. This article outlines the current state of AI integration in health care, examines how systems can responsibly implement it to enhance patient care and education, and raises critical questions about ethics and safety as we harness its transformative potential.

Lymphatic vessels activate anti-tumor immune surveillance and support metastasis. Whether there are distinct lymphatic phenotypes that govern immunity and metastasis remains unclear. Here we reveal that cytotoxic immunity normalizes lymphatic function and uncouples immune and metastatic potential. We demonstrate that intratumoral lymphatic vessel density negatively correlates with cytotoxic immunity and that IFNγ reprograms the intratumoral lymphatic state. Lymphatic deletion of Ifngr1 expanded the intratumoral lymphatic network and drove the emergence of a tip-like state that promotes lymph node metastasis but not dendritic cell migration or response to immune checkpoint blockade (ICB). Mechanistically, IFNγ restrains proliferation and cell state programs through inhibition of mitochondrial respiration. Lymphatic-specific inhibition of mitochondrial complex III restrained the intratumoral tip-like state, blocked metastasis, and enhanced the response to ICB. Our data reveal that IFNγ induces a metabolic and phenotypic switch in tumor-associated lymphatic vessels that blocks regional metastasis and reinforces immune surveillance.

Chronic wounds, especially in diabetic patients, pose a significant clinical challenge due to impaired microvasculature and delayed healing. This study presents Exo-Q, a novel thermoresponsive hydrogel formed by co-gelation of engineered Q protein nanofibers with exosomes, a class of vesicular intercellular communication mediators. Exo-Q transitions from a gel to a viscoelastic solution at physiological temperature, enabling localized, topical delivery of exosomes with an initial burst release followed by sustained release. In a diabetic mouse wound model, Exo-Q effectively delivered human bone marrow multipotent stromal cell-derived exosomes directly to the wound bed, where they accumulated in endothelial cells of granulation tissue without detectable systemic distribution. Exosomes produced under stringent and replicable cell culture conditions consistently carried biomacromolecular cargo enriched for miRNAs with validated targets in angiogenesis-associated genes, indicative of their therapeutic potential. Topical application of Exo-Q resulted in extensive neovascularized granulation tissue, significantly accelerating wound closure to levels comparable to non-diabetic wounds. Importantly, the hydrogel's modular design maintained the functional integrity of Q protein nanofibers and exosomes, demonstrating compatibility with full-thickness human wounds. This platform allows for tailored customization to address critical stages of diabetic wound healing while ensuring efficacy at low dosages, potentially enabling patient-administered treatment. By leveraging advanced biomaterials, Exo-Q advances the therapeutic efficacy of exosome-based interventions for diabetic wounds, offering a localized, non-invasive solution to chronic, non-healing wounds. This innovative hydrogel platform represents a modular therapeutic strategy with significant potential for clinical applications in regenerative medicine.

Chronic wounds, especially in diabetic patients, pose a significant clinical challenge due to impaired microvasculature and delayed healing. This study presents Exo-Q, a novel thermoresponsive hydrogel formed by co-gelation of engineered Q protein nanofibers with exosomes, a class of vesicular intercellular communication mediators. Exo-Q transitions from a gel to a viscoelastic solution at physiological temperature, enabling localized, topical delivery of exosomes with an initial burst release followed by sustained release. In a diabetic mouse wound model, Exo-Q effectively delivered human bone marrow multipotent stromal cell-derived exosomes directly to the wound bed, where they accumulated in endothelial cells of granulation tissue without detectable systemic distribution. Exosomes produced under stringent and replicable cell culture conditions consistently carried biomacromolecular cargo enriched for miRNAs with validated targets in angiogenesis-associated genes, indicative of their therapeutic potential. Topical application of Exo-Q resulted in extensive neovascularized granulation tissue, significantly accelerating wound closure to levels comparable to non-diabetic wounds. Importantly, the hydrogel's modular design maintained the functional integrity of Q protein nanofibers and exosomes, demonstrating compatibility with full-thickness human wounds. This platform allows for tailored customization to address critical stages of diabetic wound healing while ensuring efficacy at low dosages, potentially enabling patient-administered treatment. By leveraging advanced biomaterials, Exo-Q advances the therapeutic efficacy of exosome-based interventions for diabetic wounds, offering a localized, non-invasive solution to chronic, non-healing wounds. This innovative hydrogel platform represents a modular therapeutic strategy with significant potential for clinical applications in regenerative medicine.

BACKGROUNDPlatelets are increasingly recognized as active participants in immune signaling and systemic inflammation. Upon activation, platelets form monocyte platelet aggregates (MPA) representing the crossroads of thrombosis and inflammation. We hypothesized that platelet transcriptomics could capture this thromboinflammatory axis and identify individuals at elevated cardiovascular risk.METHODS: MPA levels, defined as CD14+CD61+ cells, were measured using flow cytometry at 2 time points, 4 weeks apart, in healthy individualsPlatelets were isolated and sequenced. Individuals were categorized as MPAhi or MPAlo based on consistently high or low MPA levels across time points.RESULTSAmong 149 participants (median age 52 years, 57% female, 50% non-White), MPAhi individuals exhibited increased expression of platelet activation markers P-selectin (P < 0.001), PAC-1 (P = 0.021), and CD40L (P < 0.001) and enriched immune signaling pathways. Informed by MPA levels and derived from the platelet transcriptome, we developed a 42-gene thromboinflammation platelet signature (TIPS), which correlated with MPA levels in multiple cohorts and was reproducible over time. TIPS was elevated in patients with COVID-19 (P = 0.0002) and myocardial infarction (Padj = 0.008), and as in predicted future cardiovascular events in patients who underwent lower extremity revascularization after a median follow-up of 18 months (adjusted for age, sex, race, and ethnicity [adjHR] 1.55, P = 0.006). Notably, TIPS was modifiable by ticagrelor (P = 0.002) but not aspirin.CONCLUSIONThese findings establish MPA as a biomarker of thromboinflammation and introduce TIPS, a platelet RNA signature, that captures thromboinflammation and provides a promising tool for cardiovascular risk stratification and a potential therapeutic target.TRIAL REGISTRATIONNCT04369664FUNDINGNIH R35HL144993, NIH R01HL139909, and AHA 16SFRN2873002 to JSB, DFG Walter-Benjamin-Programme 537070747 to AB.

Systemic lupus erythematosus (SLE) is a complex autoimmune disease with an increased risk of vascular dysfunction and cardiovascular disease. We validate our previously developed Systemic Lupus Erythematosus Activity Platelet-Gene Expression Signature (SLAP-GES) score and investigate its relationship with platelet activity and vascular health. SLAP-GES was associated with the SLE Disease Activity Index (Padj < 0.001) and consistent over time (r = 0.76; P = 9 × 10^-5). Moreover, SLAP-GES was associated increased platelet aggregation in response to submaximal epinephrine (P = 0.084), leukocyte platelet aggregates (P = 0.014), and neutrophil platelet aggregates (P = 0.043). SLAP-GES was also associated with impaired glycocalyx (P = 0.011) and brachial artery flow-mediated dilation (P = 0.045). Altogether, SLAP-GES is associated with SLE disease activity, platelet activity, and impaired vascular health.

There is limited data on which hospitalist switch day is optimal for hospital operations and throughput. A quality improvement intervention was implemented, changing the hospitalist switch day from Monday to Tuesday. Retrospective observational analysis revealed an increase in Monday discharges (1.3%, p = .01), a decrease in Tuesday discharges (-1.6%, p < .005), and a significant reduction in 30-day unplanned readmission rates (-1.5%, p = .003), with no significant changes in the average length of stay. Additional studies are needed to further verify these findings in different hospital settings and to consider other switch day patterns.

Excessive accumulation of lipids within cardiomyocytes can sometimes initiate cardiomyopathy, while in other situations excess lipids do not cause harm. To understand how pathologic and non-pathologic lipid accumulation differ, we isolated lipid droplets (LDs) from two genetically altered mouse lines and from wild-type (WT) mice after an overnight fast. The LDs from MHC-peroxisomal proliferator-activated receptor γ1(MHC-Pparg1) transgenic mice were threefold larger than those from either fasted WT or non-cardiomyopathy MHC-diacylglycerol acyl transferase 1 (MHC-Dgat1) transgenic mice. Proteomic analysis of the LD-associated membrane proteins (LDAMPs) showed that MHC-Pparg1 LDs had less perilipin (PLIN). Proteins associated with lipolysis and LD formation (CIDEs and MTP), lipid synthesis, and Pparg signaling pathways were increased in MHC-Pparg1 LDAMPs. Unlike in MHC-Pparg1, MHC-Dgat1 LDAMPs exhibited increased mitochondrial peroxidative proteins with reduced adipose triglyceride lipase (Pnpla2), and Pparg coactivator 1 alpha (Pgc1A). Cardiomyocytes from MHC-Pparg1 hearts had transmission electron microscopy (TEM) images of ongoing lipolysis and greater amounts of lipolytic proteins. In contrast, images from MHC-Dgat1 cardiomyocytes showed more lipophagy. Consistent with the proteomic study and EM images, cardiac immunofluorescence staining showed that PLIN5 protein, thought to block LD lipolysis, was markedly reduced with MHC-Pparg1 overexpression, while hormone-sensitive lipase was increased. The autophagosome marker protein LC3B was increased in MHC-Dgat1 but not in MHC-Pparg1 hearts. Potentially toxic lipids like diacylglycerols and ceramides were increased in hearts but not LDs from MHC-Pparg1 mice. Our data indicate that cardiomyocyte LDs vary in size, composition, and metabolism. Cardiotoxicity was associated with greater LD lipolysis, which we postulate leads to intracellular release of toxic lipids.

BACKGROUND:Delayed diagnosis of venous thromboembolism (VTE) is prevalent among hospitalized patients, yet case identification is challenging and feedback limited. OBJECTIVE:To develop a large language model (LLM)-based electronic-trigger to identify VTE diagnostic delays. METHODS:All admissions to internal medicine (IM) residents at NYU Langone Health between January 2022 and December 2023 (n = 20,843) were included. Using an open-source LLM, prompts were validated to detect (1) residents considering VTE in admission notes and (2) VTE confirmation in five types of imaging reports (n = 100 for each prompt validation set). The validated prompts were applied to determine discordance between admission note differential omitting VTE and imaging report confirming VTE. Two hospitalists reviewed discordant cases using a validated tool to identify diagnostic delays. Hospitalizations were labeled as diagnostic delays, in-hospital complication, or false-positive. Based on in-hospital complication and false-positive patterns, exclusion criteria were implemented. Positive predictive value (PPV) and negative predictive value (NPV) were calculated. RESULTS:The LLM prompts correctly classified admission notes and VTE imaging studies with high accuracy (range 98%-100%, n = 699 VTE cases identified). Of the 137 diagnostic delays the LLM-based electronic-trigger identified, 31 were true-positives, 60 in-hospital complications, and 46 false-positives. 4.4% of all VTE hospitalizations had a diagnostic delay. With the exclusion criteria, the PPV was 48% (95% confidence interval [CI], 35%-62%) and NPV was 95% (95% CI, 87%-98%). CONCLUSIONS:We developed the first LLM-based electronic-trigger to identify VTE diagnostic delays, with higher performance than existing non-LLM electronic-triggers. LLM-based approaches can facilitate diagnostic performance feedback and are scalable to other conditions and institutions.