Nature genetics最新文献

Machine learning models have recently generated excitement for their potential in a broad range of domain applications, including genomics. However, owing to their complexity, they are prohibitively expensive to train for the large genomic contexts of DNA language models, resulting in limited receptive fields and/or n-mer sequence tokenization. Nguyen et al. present a step forward for the field with Evo, a foundation model that applies the efficient, hybrid StripedHyena architecture trained on 80,000 prokaryotic and millions of phage and plasmid sequences at single-nucleotide resolution. In benchmarking, Evo shows equivalent or improved performance against state-of-the-art nucleotide and language models for variant fitness, promoter activity and protein expression prediction. Impressively, Evo can be used to generate novel, experimentally validated CRISPR–Cas and transposon systems and predict gene essentiality by premature stop codon insertion; it also shows some promise for generating synthetic whole genomes. Foundation DNA language models that are applicable to many tasks would be of broad utility and Evo underlines their promise. However, it must be noted that the training datasets are small compared to the genomes of eukaryotes, and the still-limited 131-kb context and next-token prediction will need to be further adapted for the increased complexity of multicellular life, showing there is still much to do.

Original reference: Science 386, eado9336 (2024)

Brain metastases are common and highly deadly occurrences that derive from primary cancer, especially in patients with lung adenocarcinoma (LUAD). However, it is challenging to predict if and when these metastases will appear. To better understand this issue, Zuccato et al. studied 402 LUAD tumor and plasma samples from patients with or without brain metastases. Specifically, they analyzed DNA methylation signatures and other clinical variables to derive a model that predicts the development of brain metastases. They found that promoters for some immune- and cell interaction-related genes were differentially methylated in brain metastases. Additionally, the abundance of certain immune cells was different in brain metastases versus LUAD. Importantly, they successfully leveraged the identification of liquid biomarkers — based on an analysis of methylated cell-free DNA obtained from plasma samples — to create and validate classifiers for early detection of brain metastases. The notion that LUAD methylomes can be used to predict the development of brain metastases in a noninvasive manner is a potentially exciting step toward personalized medicine. It will be interesting to investigate whether similar approaches can be used to predict the formation of brain metastases originating from other cancer types, and more broadly to predict different kinds of metastases in other organs.

Original reference: Nat. Med. https://doi.org/10.1038/s41591-024-03286-y (2024)

It is widely established that single-nucleotide variants — including those known to drive tumors — are pervasive across normal tissues, but whether more complex genetic lesions are present is less clear. Lin et al. addressed this issue by profiling breast tissue sampled from 49 healthy women by single-cell DNA sequencing and ATAC-seq. Overall, their analysis showed that copy number changes and aneuploidy were observed in nearly all women in the cohort and became more frequent with age. Common events associated with clonal expansions included loss of chromosome X, gain of chromosome 1q and losses of chromosomes 10q, 16q and 22. A subset of samples was profiled with spatial transcriptomics, which suggested that these aneuploid cells mapped to epithelial structures such as ducts, lobules and terminal ductal lobular units. The team noted that some of the copy number changes were characteristic of estrogen receptor-positive (ER⁺) or -negative (ER⁻) disease and, among these, some were enriched in different epithelial populations, supporting the idea of distinct cells of origin for ER⁺ and ER⁻ breast cancers. This will, of course, need to be confirmed in subsequent studies, but for now this study provides a compelling glimpse into how the state of the genome in healthy tissue might inform our understanding of disease etiology.

Original reference: Nature 636, 663–670 (2024)

Caenorhabditis elegans is a powerful model to study neuronal function, but despite the limited number of well-defined neurons in this worm, the roles that specific proteins and receptors have in individual neurons is not known. St. Ange et al. used single-nucleus RNA sequencing (snRNA-seq) to define the transcriptome of each neuron type in both wild-type and daf-2 worms, mutants in insulin signaling that are known to have improved memory. The authors show that neuronal gene expression changes significantly between the L4 larva and young adult (day 1) stages; this contributes to behaviors exhibited only in adults, such as butanone associative learning. Moreover, the authors identify candidate genes not previously linked with AWC-related roles that are upregulated specifically in AWC chemosensory neurons in daf-2 mutants and that are required for AWC-mediated learning and memory. Together, these data provide an atlas of adult wild-type and daf-2 neuronal gene expression, with functional behavioral insights, and highlight snRNA-seq as a powerful tool to identify genes that only change in a subset of neurons and thus would be masked in an analysis of whole animals or specific tissues.

Original reference: Cell Genom. 4, 100720 (2024)

Tandem repeat (TR) size variation is implicated in ~50 neurological disorders, yet its impact on gene regulation in the human brain remains largely unknown. In the present study, we quantified the impact of TR size variation on brain gene regulation across distinct molecular phenotypes, based on 4,412 multi-omics samples from 1,597 donors, including 1,586 newly sequenced ones. We identified ~2.2 million TR molecular quantitative trait loci (TR-xQTLs), linking ~139,000 unique TRs to nearby molecular phenotypes, including many known disease-risk TRs, such as the G₂C₄ expansion in C9orf72 associated with amyotrophic lateral sclerosis. Fine-mapping revealed ~18,700 TRs as potential causal variants. Our in vitro experiments further confirmed the causal and independent regulatory effects of three TRs. Additional colocalization analysis indicated the potential causal role of TR variation in brain-related phenotypes, highlighted by a 3ʹ-UTR TR in NUDT14 linked to cortical surface area and a TG repeat in PLEKHA1, associated with Alzheimer’s disease.

Rice production is facing substantial threats from global warming associated with extreme temperatures. Here we report that modifying a heat stress-induced negative regulator, a negative regulator of thermotolerance 1 (NAT1), increases wax deposition and enhances thermotolerance in rice. We demonstrated that the C2H2 family transcription factor NAT1 directly inhibits bHLH110 expression, and bHLH110 directly promotes the expression of wax biosynthetic genes CER1/CER1L under heat stress conditions. In situ hybridization revealed that both NAT1 and bHLH110 are predominantly expressed in epidermal layers. By using gene-editing technology, we successfully mutated NAT1 to eliminate its inhibitory effects on wax biosynthesis and improved thermotolerance without yield penalty under normal temperature conditions. Field trials further confirmed the potential of NAT1-edited rice to increase seed-setting rate and grain yield. Therefore, our findings shed light on the regulatory mechanisms governing wax biosynthesis under heat stress conditions in rice and provide a strategy to enhance heat resilience through the modification of NAT1.

Ambiguity about whether the histidine-rich calcium-binding protein-coding gene (His^R) or the pore-forming toxin-like gene (PFT) or both are responsible for Fusarium head blight (FHB) resistance conferred by the Fhb1 quantitative trait locus hinders progress toward elucidating Fhb1 resistance mechanisms. Here, with a series of developed lines with or without PFT but all possessing His^R and five His^R-carrying PFT mutant lines created via gene editing, we show that PFT does not confer FHB resistance and that the His^R resistance effect does not require PFT in the tested conditions. We also show that PFT mutations are not associated with morphological and phenological characteristics that often affect FHB severity.

Members of the KMT2C/D–KDM6A complex are recurrently mutated in urothelial carcinoma and in histologically normal urothelium. Here, using genetically engineered mouse models, we demonstrate that Kmt2c/d knockout in the urothelium led to impaired differentiation, augmented responses to growth and inflammatory stimuli and sensitization to oncogenic transformation by carcinogen and oncogenes. Mechanistically, KMT2D localized to active enhancers and CpG-poor promoters that preferentially regulate the urothelial lineage program and Kmt2c/d knockout led to diminished H3K4me1, H3K27ac and nascent RNA transcription at these sites, which leads to impaired differentiation. Kmt2c/d knockout further led to KMT2A–menin redistribution from KMT2D localized enhancers to CpG-high and bivalent promoters, resulting in derepression of signal-induced immediate early genes. Therapeutically, Kmt2c/d knockout upregulated epidermal growth factor receptor signaling and conferred vulnerability to epidermal growth factor receptor inhibitors. Together, our data posit that functional loss of Kmt2c/d licenses a molecular ‘field effect’ priming histologically normal urothelium for oncogenic transformation and presents therapeutic vulnerabilities.

Gene expression quantitative trait loci are widely used to infer relationships between genes and central nervous system (CNS) phenotypes; however, the effect of brain disease on these inferences is unclear. Using 2,348,438 single-nuclei profiles from 391 disease-case and control brains, we report 13,939 genes whose expression correlated with genetic variation, of which 16.7–40.8% (depending on cell type) showed disease-dependent allelic effects. Across 501 colocalizations for 30 CNS traits, 23.6% had a disease dependency, even after adjusting for disease status. To estimate the unconfounded effect of genes on outcomes, we repeated the analysis using nondiseased brains (n = 183) and reported an additional 91 colocalizations not present in the larger mixed disease and control dataset, demonstrating enhanced interpretation of disease-associated variants. Principled implementation of single-cell Mendelian randomization in control-only brains identified 140 putatively causal gene–trait associations, of which 11 were replicated in the UK Biobank, prioritizing candidate peripheral biomarkers predictive of CNS outcomes.

The rapid growth of modern biobanks is creating new opportunities for large-scale genome-wide association studies (GWASs) and the analysis of complex traits. However, performing GWASs on millions of samples often leads to trade-offs between computational efficiency and statistical power, reducing the benefits of large-scale data collection efforts. We developed Quickdraws, a method that increases association power in quantitative and binary traits without sacrificing computational efficiency, leveraging a spike-and-slab prior on variant effects, stochastic variational inference and graphics processing unit acceleration. We applied Quickdraws to 79 quantitative and 50 binary traits in 405,088 UK Biobank samples, identifying 4.97% and 3.25% more associations than REGENIE and 22.71% and 7.07% more than FastGWA. Quickdraws had costs comparable to REGENIE, FastGWA and SAIGE on the UK Biobank Research Analysis Platform service, while being substantially faster than BOLT-LMM. These results highlight the promise of leveraging machine learning techniques for scalable GWASs without sacrificing power or robustness.