Race, Ethnicity, and Ancestry in Clinical Pathways: A Framework for Evaluation.

Clinical algorithms, or "pathways," promote the delivery of medical care that is consistent and equitable. Race, ethnicity, and/or ancestry terms are sometimes included in these types of guidelines, but it is unclear if this is appropriate for clinical decision-making. At our institution, we developed and applied a structured framework to determine whether race, ethnicity, or ancestry terms identified in our clinical pathways library should be retained, modified, or removed. First, we reviewed all text and associated reference documents for 132 institutionally-developed clinical pathways and identified 8 pathways that included race, ethnicity, or ancestry terms. Five pathways had clear evidence or a change in institutional policy that supported removal of the term. Multispecialty teams conducted additional in-depth evaluation of the 3 remaining pathways (Acute Viral Illness, Hyperbilirubinemia, and Weight Management) by applying the framework. In total, based on these reviews, race, ethnicity, or ancestry terms were removed (n = 6) or modified (n = 2) in all 8 pathways. Application of the framework established several recommended practices, including: (1) define race, ethnicity, and ancestry rigorously; (2) assess the most likely mechanisms underlying epidemiologic associations; (3) consider whether inclusion of the term is likely to mitigate or exacerbate existing inequities; and (4) exercise caution when applying population-level data to individual patient encounters. This process and framework may be useful to other institutional programs and national organizations in evaluating the inclusion of race, ethnicity, and ancestry in clinical guidelines.

Transfer learning enables predictions in network biology.

Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues. Recently, transfer learning has revolutionized fields such as natural language understanding1,2 and computer vision3 by leveraging deep learning models pretrained on large-scale general datasets that can then be fine-tuned towards a vast array of downstream tasks with limited task-specific data. Here, we developed a context-aware, attention-based deep learning model, Geneformer, pretrained on a large-scale corpus of about 30 million single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology. During pretraining, Geneformer gained a fundamental understanding of network dynamics, encoding network hierarchy in the attention weights of the model in a completely self-supervised manner. Fine-tuning towards a diverse panel of downstream tasks relevant to chromatin and network dynamics using limited task-specific data demonstrated that Geneformer consistently boosted predictive accuracy. Applied to disease modelling with limited patient data, Geneformer identified candidate therapeutic targets for cardiomyopathy. Overall, Geneformer represents a pretrained deep learning model from which fine-tuning towards a broad range of downstream applications can be pursued to accelerate discovery of key network regulators and candidate therapeutic targets.

MIRA: joint regulatory modeling of multimodal expression and chromatin accessibility in single cells.

Rigorously comparing gene expression and chromatin accessibility in the same single cells could illuminate the logic of how coupling or decoupling of these mechanisms regulates fate commitment. Here we present MIRA, probabilistic multimodal models for integrated regulatory analysis, a comprehensive methodology that systematically contrasts transcription and accessibility to infer the regulatory circuitry driving cells along cell state trajectories. MIRA leverages topic modeling of cell states and regulatory potential modeling of individual gene loci. MIRA thereby represents cell states in an efficient and interpretable latent space, infers high-fidelity cell state trees, determines key regulators of fate decisions at branch points and exposes the variable influence of local accessibility on transcription at distinct loci. Applied to epidermal differentiation and embryonic brain development from two different multimodal platforms, MIRA revealed that early developmental genes were tightly regulated by local chromatin landscape whereas terminal fate genes were titrated without requiring extensive chromatin remodeling.

Network-based screen in iPSC-derived cells reveals therapeutic candidate for heart valve disease.

Mapping the gene-regulatory networks dysregulated in human disease would allow the design of network-correcting therapies that treat the core disease mechanism. However, small molecules are traditionally screened for their effects on one to several outputs at most, biasing discovery and limiting the likelihood of true disease-modifying drug candidates. Here, we developed a machine-learning approach to identify small molecules that broadly correct gene networks dysregulated in a human induced pluripotent stem cell (iPSC) disease model of a common form of heart disease involving the aortic valve (AV). Gene network correction by the most efficacious therapeutic candidate, XCT790, generalized to patient-derived primary AV cells and was sufficient to prevent and treat AV disease in vivo in a mouse model. This strategy, made feasible by human iPSC technology, network analysis, and machine learning, may represent an effective path for drug discovery.

Long telomeres protect against age-dependent cardiac disease caused by NOTCH1 haploinsufficiency.

Diseases caused by gene haploinsufficiency in humans commonly lack a phenotype in mice that are heterozygous for the orthologous factor, impeding the study of complex phenotypes and critically limiting the discovery of therapeutics. Laboratory mice have longer telomeres relative to humans, potentially protecting against age-related disease caused by haploinsufficiency. Here, we demonstrate that telomere shortening in NOTCH1-haploinsufficient mice is sufficient to elicit age-dependent cardiovascular disease involving premature calcification of the aortic valve, a phenotype that closely mimics human disease caused by NOTCH1 haploinsufficiency. Furthermore, progressive telomere shortening correlated with severity of disease, causing cardiac valve and septal disease in the neonate that was similar to the range of valve disease observed within human families. Genes that were dysregulated due to NOTCH1 haploinsufficiency in mice with shortened telomeres were concordant with proosteoblast and proinflammatory gene network alterations in human NOTCH1 heterozygous endothelial cells. These dysregulated genes were enriched for telomere-contacting promoters, suggesting a potential mechanism for telomere-dependent regulation of homeostatic gene expression. These findings reveal a critical role for telomere length in a mouse model of age-dependent human disease and provide an in vivo model in which to test therapeutic candidates targeting the progression of aortic valve disease.

The ACVR1 R206H mutation found in fibrodysplasia ossificans progressiva increases human induced pluripotent stem cell-derived endothelial cell formation and collagen production through BMP-mediated SMAD1/5/8 signaling.

BACKGROUND: The Activin A and bone morphogenetic protein (BMP) pathways are critical regulators of the immune system and of bone formation. Inappropriate activation of these pathways, as in conditions of congenital heterotopic ossification, are thought to activate an osteogenic program in endothelial cells. However, if and how this occurs in human endothelial cells remains unclear. METHODS: We used a new directed differentiation protocol to create human induced pluripotent stem cell (hiPSC)-derived endothelial cells (iECs) from patients with fibrodysplasia ossificans progressiva (FOP), a congenital disease of heterotopic ossification caused by an activating R206H mutation in the Activin A type I receptor (ACVR1). This strategy allowed the direct assay of the cell-autonomous effects of ACVR1 R206H in the endogenous locus without the use of transgenic expression. These cells were challenged with BMP or Activin A ligand, and tested for their ability to activate osteogenesis, extracellular matrix production, and differential downstream signaling in the BMP/Activin A pathways. RESULTS: We found that FOP iECs could form in conditions with low or absent BMP4. These conditions are not normally permissive in control cells. FOP iECs cultured in mineralization media showed increased alkaline phosphatase staining, suggesting formation of immature osteoblasts, but failed to show mature osteoblastic features. However, FOP iECs expressed more fibroblastic genes and Collagen 1/2 compared to control iECs, suggesting a mechanism for the tissue fibrosis seen in early heterotopic lesions. Finally, FOP iECs showed increased SMAD1/5/8 signaling upon BMP4 stimulation. Contrary to FOP hiPSCs, FOP iECs did not show a significant increase in SMAD1/5/8 phosphorylation upon Activin A stimulation, suggesting that the ACVR1 R206H mutation has a cell type-specific effect. In addition, we found that the expression of ACVR1 and type II receptors were different in hiPSCs and iECs, which could explain the cell type-specific SMAD signaling. CONCLUSIONS: Our results suggest that the ACVR1 R206H mutation may not directly increase the formation of mature chondrogenic or osteogenic cells by FOP iECs. Our results also show that BMP can induce endothelial cell dysfunction, increase expression of fibrogenic matrix proteins, and cause differential downstream signaling of the ACVR1 R206H mutation. This iPSC model provides new insight into how human endothelial cells may contribute to the pathogenesis of heterotopic ossification.

NOTCH1 regulates matrix gla protein and calcification gene networks in human valve endothelium.

Valvular and vascular calcification are common causes of cardiovascular morbidity and mortality. Developing effective treatments requires understanding the molecular underpinnings of these processes. Shear stress is thought to play a role in inhibiting calcification. Furthermore, NOTCH1 regulates vascular and valvular endothelium, and human mutations in NOTCH1 can cause calcific aortic valve disease. Here, we determined the genome-wide impact of altering shear stress and NOTCH signaling on human aortic valve endothelium. mRNA-sequencing of primary human aortic valve endothelial cells (HAVECs) with or without knockdown of NOTCH1, in the presence or absence of shear stress, revealed NOTCH1-dependency of the atherosclerosis-related gene connexin 40 (GJA5), and numerous repressors of endochondral ossification. Among these, matrix gla protein (MGP) is highly expressed in aortic valve and vasculature, and inhibits soft tissue calcification by sequestering bone morphogenetic proteins (BMPs). Altering NOTCH1 levels affected MGP mRNA and protein in HAVECs. Furthermore, shear stress activated NOTCH signaling and MGP in a NOTCH1-dependent manner. NOTCH1 positively regulated endothelial MGP in vivo through specific binding motifs upstream of MGP. Our studies suggest that shear stress activates NOTCH1 in primary human aortic valve endothelial cells leading to downregulation of osteoblast-like gene networks that play a role in tissue calcification.

Human disease modeling reveals integrated transcriptional and epigenetic mechanisms of NOTCH1 haploinsufficiency.

The mechanisms by which transcription factor haploinsufficiency alters the epigenetic and transcriptional landscape in human cells to cause disease are unknown. Here, we utilized human induced pluripotent stem cell (iPSC)-derived endothelial cells (ECs) to show that heterozygous nonsense mutations in NOTCH1 that cause aortic valve calcification disrupt the epigenetic architecture, resulting in derepression of latent pro-osteogenic and -inflammatory gene networks. Hemodynamic shear stress, which protects valves from calcification in vivo, activated anti-osteogenic and anti-inflammatory networks in NOTCH1(+/+), but not NOTCH1(+/-), iPSC-derived ECs. NOTCH1 haploinsufficiency altered H3K27ac at NOTCH1-bound enhancers, dysregulating downstream transcription of more than 1,000 genes involved in osteogenesis, inflammation, and oxidative stress. Computational predictions of the disrupted NOTCH1-dependent gene network revealed regulatory nodes that, when modulated, restored the network toward the NOTCH1(+/+) state. Our results highlight how alterations in transcription factor dosage affect gene networks leading to human disease and reveal nodes for potential therapeutic intervention.