Biology subjects

Sahu, A.

Publications and source records attributed to Sahu, A..

At least 19 recordsLinked to original sources

bioRxiv2026

A CTCF-LDHB axis converts nuclear lactate availability into an H3K18 acylation switch that directs alternative splicing in hypoxic breast cancer

Lactate accumulating in hypoxic tumours lactylates histones, coupling metabolic state to chromatin. Why it differs between tumours of comparable glycolytic rate has remained unresolved. Here we show that nuclear, rather than total intracellular, lactate availability determines histone H3K18 lactylation in breast cancer, and that CCCTC-binding factor (CTCF) and lactate dehydrogenase B (LDHB) jointly set it. CTCF sustains glycolytic gene expression under hypoxia and drives LDHB in triple-negative breast cancer (TNBC), whereas a CTCF-ER complex represses LDHB in non-TNBC. LDHB consumes lactate favouring H3K18 acetylation; where LDHB is low, CTCF-dependent MCT1 delivers lactate into the nucleus and sustains lactylation. At a CTCF site downstream of TACC3 exon 3, an invariant polymerase II pause couples kinetic and adaptor control of splicing, with the mark deciding the outcome: H3K18la recruits SMARCA4 and hnRNPA1 to skip the exon, generating an isoform that drives EMT, whereas H3K18ac recruits TRIM33 and SRSF10 to include it, generating one that confers stemness. Inhibiting LDHB restores exon skipping and suppresses tumour growth in vivo. Nuclear lactate availability therefore acts as a metabolic switch that converts hypoxic stress into subtype-specific splicing decisions and identifies LDHB as a tractable target in TNBC.

cancer biology↗

bioRxiv2026

MutQA: A Cross-Validated Q&A Dataset for Genetic Mutations

We introduce MutQA, a dataset of 330,387 citation-grounded question-answer (QA) pairs linking genetic mutations to their functional consequences as described in the biomedical literature. Each record pairs a protein-coding mutation with a natural-language question and an evidence-backed answer anchored to sentences in one of 27,442 open-access PubMed Central articles, spanning 6,620 genes and 125,593 gene-variant pairs. The dataset is constructed using CrossValQA, a framework that enforces information isolation between independent LLMs: one model generates QA pairs from an article, a second model answers the same question without access to the first model's response, and a third model judges whether each model's rationale supports the other's answer, so that retained QA pairs are reproducible from the source text rather than hallucinated. Of 401,406 audited QA pairs, 82.7% pass both directions of this cross-validation; those with a recoverable reference protein sequence form the released benchmark. In a 999-pair stratified audit, two expert annotators agree exactly on 86.4% of cross-validated QA pairs (95% CI 83.7-89.0; n = 638), a tier in which about 90% of ratings are "correct". Individual LLM judges agree with individual annotators at a comparable 87.0% on this tier but far less on QA pairs that fail cross-validation. Fine-tuning MutaPLM on MutQA nearly doubles its mean LLM-judge score for question-conditioned mutation-effect answering (1.45 to 2.82 on a 0-5 scale) and improves mutation-recovery accuracy from effect descriptions from 19.4% to 32.1%. Providing variant-matched MutQA retrieval context, without fine-tuning, improves strict answer accuracy across four LLM families by 20.0-28.6 percentage points, demonstrating the framework's ability to improve the reliability of LLM answers. We also apply CrossValQA to proteomics, providing a proof of concept for its use beyond mutation-specific QA. We release the dataset, construction framework, and evaluation code.

genomics↗

bioRxiv2026

Leucine Regulates Neuronal Health During Nutrient Deprivation in Caenorhabditis elegans

Alterations in nutrient availability induce changes in neuronal structure and connectivity across species, ultimately leading to nervous system plasticity and behavioural changes. However, it remains unclear how prolonged acute nutrient deprivation affects the integrity and function of the nervous system. Using the nematode Caenorhabditis elegans, we uncovered that prolonged macronutrient deprivation profoundly impacts neuronal health. Across developmental stages, multiple sensory neuron classes in the C. elegans nervous system, including polymodal IL2, volatile odor-sensing AWB and AWC, and CO2-sensing BAG neurons, display extensive dendritic and axonal blebbing, severely disorganized dendritic, axonal, and cell body morphologies, and severely disrupted chemical and electrical synaptic organization during prolonged nutrient deprivation. The severity of these defects increases progressively with the duration of nutrient deprivation, ultimately affecting animal behaviours. Our results show that the availability of a particular macronutrient, specifically the branched-chain amino acid leucine alone, in the absence of any other macronutrients, although insufficient to support animal growth, is sufficient to completely prevent nutrient-deprivation-mediated neuronal damage and the decline in animal behavioural responses. Moreover, leucine supplementation alone is sufficient to reverse nutrient-deprivation-mediated neuronal abnormalities. These findings suggest that the availability of leucine, rather than absolute caloric intake, is crucial for maintaining neuronal health and function under adverse conditions.

neuroscience↗

bioRxiv2026

High-level ciprofloxacin resistance in Pseudomonas aeruginosa is associated with genomic adaptation, collateral antibiotic susceptibility, and physiological trade-offs

Antimicrobial resistance in Pseudomonas aeruginosa has emerged as a major clinical concern. Ciprofloxacin has potent intrinsic activity against P. aeruginosa, but resistance to this antibiotic is increasingly reported in clinical settings. In this study, P. aeruginosa ATCC 27853 strain was exposed to gradually increasing ciprofloxacin concentrations for 50 passages ([~]100 days). The final selected resistant strain (CIP-R) with 256-fold of minimum inhibitory concentration was studied further. Whole-genome sequencing revealed 12 genomic alterations, including known mutations in the quinolone resistance-determining genes gyrA (Thr83Ile), parC (Thr177Asn), parE (Glu459Lys). A duplication mutation in nfxB (Tyr153_Gly154dup) was also observed. In addition to these genes, we observed mutations in pilA, tadB, psdR, NP446_RS32055 (TRAP transporter permease), multiple dppA3 variants, and a prophage-associated hypothetical gene (NP446_RS24255), which have not been reported previously. The gyrA Thr83Ile substitution was conserved in 86.38% of ciprofloxacin-resistant clinical isolates retrieved from the NCBI database. Resistance acquisition was accompanied by slower growth, impaired swimming and swarming motility, diminished surface attachment, reduced biofilm formation. The resistant strain has enhanced {beta}-lactamase activity, and resistance to levofloxacin, cefepime, and meropenem together with sensitivity to piperacillin-tazobactam and aztreonam. This study highlights the gyrA Thr83Ile mutation as a key genomic marker for molecular screening of ciprofloxacin resistance and reveals secondary adaptive trade-offs that can be targeted for clinical diagnostic and therapeutic decision-making. In conclusion, achieving high-level ciprofloxacin resistance in P. aeruginosa involves non-target-site genomic adaptations and physiological trade-offs beyond classical target mutations.

microbiology↗

bioRxiv2026

BioWorldModel: a single architecture predictsphenotype from genotype across four kingdoms of life

The same genome produces different phenotypes in different conditions--yet predictive models encode genotype once and treat each trait independently. Here we show that representing phenotype generation as a dynamic biological process transforms predictive accuracy across bacteria, fungi, animals and plants. BioWorldModel learns how organisms interpret their genome: frozen gene embeddings (species context) modulated by individual variation pass through four biological process layers (regulation [->] expression [->] pathway [->] cellular) that respond to environment and time. A state-conditioned attention mechanism rereads this dynamic representation, predicting full multivariate trait distributions. Without modification, the architecture achieves mean correlation r = 0.678 on 214 bacterial growth traits (207% better than ridge regression), r = 0.915 on 35 yeast fitness traits (167% better), r = 0.499 on 199 fly phenotypes in small-sample regime (760% better), and r = 0.995 on 36 rice traits (49% better). Ablations confirm that modeling biological process--not model size--drives performance. When neural architectures represent how biology generates phenotype rather than merely associating genotype with outcome, they capture what static methods miss.

genomics↗

bioRxiv2025

Linking MSMEG_1353 to lipid metabolism and envelope integrity in Mycolicibacterium smegmatis

Mycobacterium tuberculosis (Mtb) poses a significant global health burden. Rv0647c, an essential Mtb cell wall protein, is a potential drug target. We studied its Mycolicibacterium smegmatis (MSMEG) homologue, MSMEG_1353, using a CRISPRi conditional knockdown. MSMEG_1353 depletion increased cell width and volume, delayed log-phase initiation, slowed aggregation, reduced biofilm formation, and heightened susceptibility to antibiotics and sodium dodecyl sulfate (SDS). AlphaFold, sequence alignment, and UNIPROT analyses suggest MSMEG_1353 functions as a protein kinase, with conserved residues in intermediate high-confidence regions potentially forming an ATP-binding site. Investigating its role in cell envelope biosynthesis, mass spectrometry revealed elevated levels of mycolic acid biosynthesis proteins upon MSMEG_1353 knockdown. RT-qPCR confirmed upregulation of the fabD-acpM-kasA-KasB-accD6 operon, encoding key mycolic acid synthesis enzymes. Notably, a strong negative correlation with MSMEG_0911, the predominant isocitrate lyase, important in the glyoxylate cycle, was observed via both MS and RT-qPCR. Collectively, MSMEG_1353 deficiency compromises cell wall integrity, likely due to altered lipid composition resulting from dysregulated mycolic acid biosynthesis and lipid metabolism. These findings support the development of models explaining MSMEG_1353s involvement in these pathways.

microbiology↗

bioRxiv2025

The claudin-like molecule CLC-3 regulates neuromuscular function in Caenorhabditis elegans by modulating cholinergic signalling

Cell adhesion molecules (CAMs play important roles in neurons, contributing to nervous system development, synapse formation, and activity-dependent plasticity. Claudins, the cell-adhesion molecules known for their roles at tight junctions in epithelial and endothelial cells, remain underexplored in neurons, particularly in vertebrates. In contrast, emerging studies in Caenorhabditis elegans have begun to reveal neuronal functions of claudin-like proteins. However, a systematic analysis of their neuronal expression has not been performed. We conducted a transcriptional reporter screen of all claudin-like genes in C. elegans and identified several candidates with previously unreported neuronal expression, highlighting a broader role of this family in the nervous system. One candidate, clc-3, showed robust expression in head, tail, and ventral cord neurons, with no detectable expression in non-neuronal tissues. Functional analyses of clc-3 mutants revealed increased body-bend amplitude and elevated evoked postsynaptic currents at the cholinergic neuromuscular synapses. Imaging and molecular interaction studies demonstrated that CLC-3 localises to the presynaptic membranes in cholinergic neurons, where it interacts with the actin-binding protein NAB-1 and regulates cholinergic signalling. This presynaptic role of CLC-3 likely contributes to the regulation of sinusoidal movement in C. elegans. Our findings identify CLC-3 as a neuronally expressed claudin that regulates motor system output by influencing synaptic vesicle organization and illustrate how changes in synaptic organization are coupled to whole-animal behaviour.

neuroscience↗

bioRxiv2025

panomiX: Investigating Mechanisms Of Trait Emergence Through Multi-Omics Data Integration

Complex omics approaches and high-throughput phenotyping generate large, heterogeneous datasets that make linking molecular signatures to plant traits challenging. To address this challenge, here we introduce panomiX, a user-friendly toolbox for multi-omics integration, designed to enable non-experts to apply advanced computational methods with ease. panomiX automates data preprocessing, variance analysis, multi-omics prediction, and interaction modeling through machine learning, revealing meaningful molecular interactions and synergies. We applied panomiX to a tomato heat-stress experiment combining image-based phenotyping, transcriptomics, and Fourier-transform infrared spectroscopy data, with the aim of identification of condition-specific, cross-domain relationships between gene expression, metabolite levels, and phenotypic traits. Our approach identified a network of such connections, with those linking photosynthesis traits with stress-responsive kinases in elevated temperatures among most significant ones. By simplifying complex analyses and improving interpretability, panomiX offers a platform to accelerate the discovery of trait emergence in plants and select specific candidate genes based on multi-omics analyses.

bioinformatics↗

bioRxiv2025

Weighted gene co-expression network analysis reveals key genes and lncRNAs in desi cattle chronically infected with Johne's disease

Johnes disease (JD), caused by Mycobacterium avium subsp. paratuberculosis (MAP), is a chronic enteric disease with significant economic impacts on the dairy industry. To elucidate molecular mechanisms underlying MAP infection in desi cattle (Bos indicus), we conducted a comprehensive transcriptomic study. This revealed 1,905 protein-coding and 3,123 lncRNA genes as differentially expressed in MAP-infected samples, alongside the first comprehensive annotation of 45,947 lncRNAs in desi cattle. Weighted Gene Co-expression Network Analysis identified 11 co-expressed gene modules, with the turquoise module comprising 870 protein- coding genes and 934 lncRNAs, highly correlated with clinical traits. Functional enrichment analysis of this key module revealed significant involvement in defense response, inflammatory processes, and NK cell-mediated immunity and cytotoxicity. GSE analysis revealed suppressed pathways, including NK cell immunity and lectin response, facilitating bacterial persistence in chronic infection, while activated pathways included G-protein-coupled receptor signaling and metabolic pathways. Network approach identified 12 hub genes (IL7R, TLR4, KLRK1, IFNG, TGFB1, CD68, CXCR6, GZMB, KLRG1, MMP9, GZMA, SELL) associated with immune suppression, inflammation, and tissue remodeling. Additionally, 22 lncRNAs co-expressed with hub genes suggesting roles in modulating immune regulation. By identifying critical genes and lncRNAs, our study offers potential targets for developing innovative diagnostic and therapeutic strategies for JD.

systems biology↗

bioRxiv2025

Genome assemblies of Indian desi cattle reveals hotspots of rearrangements and immune-related genetic diversity

India, home to the worlds largest cattle population, hosts native dairy breeds essential to its agricultural economy because of their adaptability and resilience. This study characterizes the genomes of five prominent breeds Gir, Kankrej, Red Sindhi, Sahiwal and Tharparkar, highlighting their unique genomic characteristics. The de novo assemblies ranged from 2.70-2.78 Gb in size, with 90% of the genomes assembled in just 56 to 1,663 scaffolds. The use of reference-guided scaffolding further enhanced these genomes, resulting in 93.3-96.7% pseudomolecule coverage with strong BUSCO scores (94.1-95.5%). Comparative analyses revealed 87-95% synteny with the Brahman genome and identified 19.84-153.16 Mb of structural rearrangements per genome, including inversions, translocations, and duplications. Synteny diversity analysis uncovered 10,643 perfectly collinear regions spanning 87.3 Mb and 6,622 hotspots of rearrangement (HOT regions) covering 55.18 Mb. These HOT regions, characterized by high synteny diversity, were significantly enriched with immune-related genes. Moreover, immune-related gene clusters, including MHC, NKC, and LRC, were identified within HOT regions in the desi reference genome. Our findings provide valuable insights into the genetic diversity of desi cattle breeds. The high-quality genome assemblies generated in this study will serve as valuable resources for future research in genetic improvement, disease resistance, and environmental adaptation.

genomics↗

bioRxiv2024

Isoprene deters insect herbivory by priming plant hormone responses

Isoprene, emitted by some plants, enhances plant abiotic resilience but its role in biotic stress resilience remains elusive. We used tobacco plants engineered to emit isoprene (IE) and the corresponding azygous non-emitting control (NE) to investigate isoprene emission and biotic stress resilience. IE plants were more resistant to insect herbivory than NE plants. Worms preferred to feed on NE rather than IE leaves. IE plants showed less decline in photosynthesis during worm feeding. Insect feeding increased jasmonate levels in IE leaves, suggesting isoprene-mediated priming of the jasmonic acid response. Wound-induced increase in isoprene emission corresponded with elevation of methyl-D-erythritol-4-phosphate pathway and Calvin-Benson cycle metabolites. The results highlight interactive functions of isoprene and jasmonic acid and advance our understanding of how isoprene emission enhances plant resilience.

plant biology↗

bioRxiv2024

DeepVul: A Multi-Task Transformer Model for Joint Prediction of Gene Essentiality and Drug Response

Despite their potential, current precision oncology approaches benefit only a small fraction of patients due to their limited focus on actionable genomic alterations. To expand its applicability, we propose DeepVul, a multi-task transformer-based model designed to predict gene essentiality and drug response from cancer transcriptome data. DeepVul aligns gene expressions, gene perturbations, and drug perturbations into a latent space, enabling simultaneous and accurate prediction of cancer cell vulnerabilities to numerous genes and drugs. Benchmarking against existing precision oncology approaches revealed that Deep-Vul not only matches but also complements oncogene-defined precision methods. Through interpretability analyses, DeepVul identifies underlying mechanisms of treatment response and resistance, as demonstrated with BRAF vulnerability prediction. By leveraging whole-genome transcriptome data, DeepVul enhances the clinical actionability of precision oncology, aiding in the identification of optimal treatments across a broader range of cancer patients. DeepVul is publicly available at https://github.com/alaaj27/DeepVul.git.

bioinformatics↗

bioRxiv2024

MedGraphNet: Leveraging Multi-Relational Graph Neural Networks and Text Knowledge for Biomedical Predictions

Genetic, molecular, and environmental factors influence diseases through complex interactions with genes, phenotypes, and drugs. Current methods often fail to integrate diverse multi-relational biological data meaningfully, limiting the discovery of novel risk genes and drugs. To address this, we present MedGraphNet, a multi-relational Graph Neural Network (GNN) model designed to infer relationships among drugs, genes, diseases, and phenotypes. MedGraphNet initializes nodes using informative embeddings from existing text knowledge, allowing for robust integration of various data types and improved generalizability. Our results demonstrate that MedGraphNet matches and often outperforms traditional single-relation approaches, particularly in scenarios with isolated or sparsely connected nodes. The model shows generalizability to external datasets, achieving high accuracy in identifying disease-gene associations and drug-phenotype relationships. Notably, MedGraphNet accurately inferred drug side effects without direct training on such data. Using Alzheimers disease as a case study, MedGraphNet successfully identified relevant phenotypes, genes, and drugs, corroborated by existing literature. These findings demonstrate the potential of integrating multi-relational data with text knowledge to enhance biomedical predictions and drug repurposing for diseases.MedGraphNet code is available at https://github.com/vinash85/MedGraphNet

bioinformatics↗

bioRxiv2024

Craters on the melanoma surface facilitate tumor-immune interactions and demonstrate pathologic response to checkpoint blockade in humans

Immunotherapy leads to cancer eradication despite the tumors immunosuppressive environment. Here, we used extended long-term in-vivo imaging and high-resolution spatial transcriptomics of endogenous melanoma in zebrafish, and multiplex imaging of human melanoma, to identify domains that facilitate immune response during immunotherapy. We identified crater-shaped pockets at the margins of zebrafish and human melanoma, rich with beta-2 microglobulin (B2M) and antigen recognition molecules. The craters harbor the highest density of CD8+ T cells in the tumor. In zebrafish, CD8+ T cells formed prolonged interactions with melanoma cells within craters, characteristic of antigen recognition. Following immunostimulatory treatment, the craters enlarged and became the major site of activated CD8+ T cell accumulation and tumor killing that was B2M dependent. In humans, craters predicted immune response to ICB therapy, showing response better than high T cell infiltration. This marks craters as potential new diagnostic tool for immunotherapy success and targets to enhance ICB response.

cancer biology↗

bioRxiv2024

Constructing a Draft Indian Cattle Pangenome Using Short-Read Sequencing

BackgroundIndian cattle known as desi cattle, renowned for their adaptability to harsh environments and diverse phenotypic traits, represent a valuable genetic resource. While reference genome assemblies have been instrumental in advancing cattle genomics, they often fail to capture the full spectrum of genetic variation present within diverse populations. To address this limitation, we aimed to construct a pangenome for desi cattle by identifying and characterizing Non-Reference Novel Sequences (NRNS). FindingsWe sequenced 68 desi cattle genomes representing seven distinct breeds, generating 48.35 billion short reads. A PanGenome Analysis (PanGA) pipeline was developed in Bash scripts to process these data to identify NRNS missing in the reference genome. A total of 13,065 NRNS with a cumulative length of [~]41 Mbp were identified that exhibited substantial variation across the population. These NRNS were found to be exclusive to Indian desi cattle, matching only 4.1% with the Chinese indicine pangenome. However, a significant proportion ([~]40%) of NRNS displayed ancestral origins within the Bos genus. These sequences were enriched in genic regions, suggesting functional roles, and were associated with quantitative trait loci (QTLs), particularly for milk production. Compared to a single reference genome, the pangenome approach significantly enhanced read mapping accuracy, reduced spurious SNP calls, and facilitated the discovery of novel genetic variants. ConclusionsThis study has successfully established a within-species cattle pangenome specifically focused on desi cattle breeds from India. Our findings highlight the importance of pangenome based analyses for understanding the complex genetic architecture of desi cattle.

genomics↗

bioRxiv2024

LitGene: a transformer-based model that uses contrastive learning to integrate textual information into gene representations

Representation learning approaches leverage sequence, expression, and network data, but utilize only a fraction of the rich textual knowledge accumulated in the scientific literature. We present LitGene, an interpretable transformer-based model that refines gene representations by integrating textual information. The model is enhanced through a Contrastive Learning (CL) approach that identifies semantically similar genes sharing a Gene Ontology (GO) term. LitGene demonstrates accuracy across eight benchmark predictions of protein properties and robust zero-shot learning capabilities, enabling the prediction of new potential disease risk genes in obesity, asthma, hypertension, and schizophrenia. LitGenes SHAP-based interpretability tool illuminates the basis for identified disease-gene associations. An automated statistical framework gauges literature support for AI biomedical predictions, providing validation and improving reliability. LitGenes integration of textual and genetic information mitigates data biases, enhances biomedical predictions, and promotes ethical AI practices by ensuring transparent, equitable, open, and evidence-based insights. LitGene code is open source and also available for use via a public web interface at litgene.avisahuai.com.

bioinformatics↗

bioRxiv2024

Advancing the Indian Cattle Pangenome: Characterizing Non-Reference Sequences in Bos indicus

BackgroundIndia, with the worlds largest cattle population and more than 50 registered breeds of Bos indicus, stands as a vital reservoir of genetic diversity. However, the abundant diversity among Indian cattle breeds highlights the inadequacy of a single reference sequence to represent the entire genomic content of desi cattle. We recognize the need to capture the genomic differences within the Bos indicus population as a whole, and specifically within the dairy cattle subset by identifying non-reference sequences and constructing a pangenome. FindingFive representative genomes of prominent dairy breeds, including Gir, Kankrej, Tharparkar, Sahiwal, and Red Sindhi, were sequenced using 10X Genomics Linked-Read technology. Assemblies generated from these linked-reads ranged from 2.70 Gb to 2.77 Gb,comparable to the Bos indicus Brahman reference genome. A pangenome of Bos indicus cattle was constructed by comparing the newly assembled genomes with the reference using alignment and graph-based methods, revealing 8 Mb and 17.7 Mb of novel sequence respectively. A confident set of 6,844 Non-Reference Unique Insertions (NUIs) spanning 7.57 Mbs was identified through both methods, representing the pangenome of Indian Bos indicus breeds. Comparative analysis with previously published pangenomes unveiled 2.8 Mb (37%) commonality with the Chinese indicine pangenome and only 1% commonality with the Bos taurus pangenome. Among these, 2,312 NUIs - encompassing [~]2 Mb, were commonly found in 98 samples of the 5 breeds and designated as Bos indicus Common Insertions (BICIs) in the population. Furthermore, 926 BICIs were identified within 682 protein-coding genes, 54 long non-coding RNAs (LncRNA), and 18 pseudogenes. These protein-coding genes were enriched for functions such as chemical synaptic transmission, cell junction organization, cell-cell adhesion, and cell morphogenesis. The protein-coding genes were found in various prominent Quantitative Trait Loci (QTL) regions, suggesting potential roles of BICIs in traits related to milk production, reproduction, exterior, health, meat, and carcass. Notably, 63.21% of the bases within the BICIs call set contained interspersed repeats, predominantly LINEs. Additionally,70.28% of BICIs are shared with other domesticated and wild species, highlighting their evolutionary significance. ConclusionThis is the first report unveiling a robust set of NUIs defining the pangenome of Bos indicus breeds of India. The analyses contribute valuable insights into the genomic landscape of desi cattle breeds.

genomics↗

bioRxiv2024

The majority of somatic mutations in fruit trees are layer-specific

BackgroundAll plant tissues and organs develop from meristems. Plant meristems are structured organs consisting of distinct layers of stem cells. Somatic mutations occurring in one of these layers can propagate into large sectors of the plant. However, the frequency and characteristics of meristematic mutations that form the basis of somaclonal phenotypic variation remain unclear. ResultsHere, we analysed the frequency and distribution of somatic mutations in an individual Apricot tree. For this, we sequenced the genomes of fruit samples corresponding to distinct meristematic cell layers selected across the entire tree. Most somatic mutations (>90%) were specific to individual layers. Genotyping the somatic mutations in leaves sampled next to the fruits confirmed their meristematic origin. Interestingly, layer 1 (epidermis) had a higher mutation load than layer 2 (mesocarp), implying differential mutational dynamics between the layers. The somatic mutations followed the branching pattern of the tree. These factors led to the unexpected observation that the layer 1 samples from different branches were more similar to each other than to layer 2 samples of the same branch. Further, using single-cell RNA sequencing, we demonstrated that the layer-specific mutant alleles could only be found in the transcripts of the respective, layer-specific cell clusters and could form the basis for somaclonal phenotypic variation. ConclusionsHere, we analyzed the prevalence and distribution of somatic mutations with meristematic origin. Our insights into the yet unexplored layer-specificity of such somatic mutations outlined how they can be identified and how they impact the breeding of clonally propagated crops.

genomics↗