Biology subjects

Mehta, R.

Publications and source records attributed to Mehta, R..

10 recordsLinked to original sources

bioRxiv2026

Cell cycling throttles epithelial--mesenchymal transition

Cell cycle phase is an understudied source of variation in gene expression. EMT is a case in point: this cell state transition that drives metastasis has been studied widely across tissue, stimulus, and microenvironmental conditions, yet cell cycle effects are routinely regressed out of single-cell EMT analyses. Here, we compare EMT between G1 and cycling (S/G2/M) cell states in two independent scRNA-seq datasets of TGF-{beta}-stimulated MCF10A mammary epithelial cells (varying dose and timepoint). G1 cells consistently show higher expression of EMT genes and upregulated RNA velocities than cycling cells, with the strongest upregulation in G1 EMT intermediate states. Consistent with the go-or-grow hypothesis, EMT and cell cycle programs inversely correlate; moreover, this anticorrelation extends to RNA velocity and strengthens with TGF-{beta} dose. We identify a 77-gene G1-specific EMT signature enriched for actin cytoskeletal and extracellular matrix genes that would otherwise be lost by regressing out cell cycle effects. This gene set prompts the hypothesis that, unconstrained by the cytoskeletal demands of mitosis, G1 cells have greater capacity to execute EMT. Further, we developed an EMT gene regulatory model coupled with a cell cycle module, and found through simulation that the model recapitulated this inverse EMT--cell cycle relationship. Using this model, we predict that cycle-selective chemotherapy enriches for the G1 intermediate state, which has high metastatic potential. Cell cycle phase thus ought not be regressed out of EMT states, given its influence on EMT.

systems biology↗

bioRxiv2026

Development and design rules for a modular human peptide-tag adaptor CAR T cell system

CAR T cells are highly effective in haematological malignancies, but limited post-infusion control and tumour antigen heterogeneity remain major challenges. Modular CAR systems address these limitations using interchangeable adaptor molecules (AM) to control CAR T cell engagement and redirect targeting toward different antigens, yet the design principles underlying effective modular function remain undefined. We present our modular platform, BRiDGE CAR, and systematically examine how CAR and adaptor configuration influence activity. BRiDGE CAR activity was influenced by CAR hinge length and adaptor format and valency; a medium-length hinge combined with a full-length, bivalent adaptor supported potent activity across multiple tumour antigens and consistently controlled tumour growth across the tested xenograft models. During repeated antigen challenge, CD33-directed BRiDGE CAR T cells retained less-differentiated phenotypes than direct CD33-targeting CAR T cells. In vivo, BRiDGE CAR T cells persisted during adaptor withdrawal and regained antitumour activity upon adaptor re-administration, while IL-18 armouring enhanced tumour control at a reduced CAR T-cell dose. We identified key design parameters governing rules for adaptor-mediated CAR activity, with antigen targeting and CAR T cell function remaining independently modifiable, establishing BRiDGE CAR as a flexible and controllable modular platform.

cancer biology↗

bioRxiv2025

Solid-state nanopore sensing reveals conformational changes induced by a mutation in a neuron-specific tRNAArg

We demonstrate that solid-state nanopore sensing is a powerful single-molecule method for analyzing RNA conformational ensembles. As a model, we employed n-Tr20, a neuron-specific cytoplasmic tRNAArgUCU, whose C50U mutation is associated with neurodegeneration in C57BL/6J mice. Maturation of the Tr20C50U precursor is impaired as the mutation stabilizes a conformational ensemble different from the wild-type. To gain insights into how this mutation engenders structural differences, we used solid-state nanopore sensing for the real-time identification of metastable conformers that are not easily observable by ensemble methods. Ion-current traces recorded using an 8-nm nanopore revealed broad contours of the conformational landscape of n-Tr20/n-Tr20C50U {+/-} Mg2+. Additionally, cryo-EM analysis and small-angle X-ray scattering studies revealed structural plasticity even more than predicted from the nanopore-sensing data. Since dynamics undergird RNA (dys)function in cellular physiology and pathology, nanopore sensing to determine RNA conformational sampling is a valuable addition to the growing RNA structural analysis toolkit. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=79 SRC="FIGDIR/small/647894v2_ufig1.gif" ALT="Figure 1"> View larger version (22K): org.highwire.dtl.DTLVardef@45bforg.highwire.dtl.DTLVardef@1e31110org.highwire.dtl.DTLVardef@75388forg.highwire.dtl.DTLVardef@1d45037_HPS_FORMAT_FIGEXP M_FIG C_FIG

biophysics↗

bioRxiv2025

Genome modeling and design across all domains of life with Evo 2

All of life encodes information with DNA. While tools for sequencing, synthesis, and editing of genomic code have transformed biological research, intelligently composing new biological systems would also require a deep understanding of the immense complexity encoded by genomes. We introduce Evo 2, a biological foundation model trained on 9.3 trillion DNA base pairs from a highly curated genomic atlas spanning all domains of life. We train Evo 2 with 7B and 40B parameters to have an unprecedented 1 million token context window with single-nucleotide resolution. Evo 2 learns from DNA sequence alone to accurately predict the functional impacts of genetic variation--from noncoding pathogenic mutations to clinically significant BRCA1 variants--without task-specific finetuning. Applying mechanistic interpretability analyses, we reveal that Evo 2 autonomously learns a breadth of biological features, including exon-intron boundaries, transcription factor binding sites, protein structural elements, and prophage genomic regions. Beyond its predictive capabilities, Evo 2 generates mitochondrial, prokaryotic, and eukaryotic sequences at genome scale with greater naturalness and coherence than previous methods. Guiding Evo 2 via inference-time search enables controllable generation of epigenomic structure, for which we demonstrate the first inference-time scaling results in biology. We make Evo 2 fully open, including model parameters, training code, inference code, and the OpenGenome2 dataset, to accelerate the exploration and design of biological complexity.

genomics↗

bioRxiv2024

Leveraging machine learning and accelerometry to classify animal behaviours with uncertainty

O_LIAnimal-worn sensors have revolutionised the study of animal behaviour and ecology. Accelerometers, which measure changes in acceleration across planes of movement, are increasingly being used in conjunction with machine learning models to classify animal behaviours across taxa and research questions. However, the widespread adoption of these methods faces challenges from imbalanced training data, unquantified uncertainties in model outputs, shifts in model performance across contexts, and noisy classifications in continuous data streams, where predicted behaviours change abruptly within a sequence. C_LIO_LITo address these challenges, we introduce an open-source approach for classifying animal behaviour from raw acceleration data. Our approach integrates machine learning and statistical inference techniques to evaluate and mitigate class imbalances, changes in model performance across ecological settings, and noisy classifications. Importantly, we extend predictions from single behaviour classifications to prediction sets: sets of behaviour labels guaranteed to contain the true behaviour with a pre-specified probability, in a framework analogous to the use of prediction intervals in statistical analyses. C_LIO_LIWe evaluate our approach via simulation and highlight its utility using data collected from a free-ranging large carnivore, African wild dogs (Lycaon pictus), in the Okavango Delta, Botswana. We demonstrate significantly improved predictions along with associated uncertainty metrics in African wild dog behaviour classification, particularly for rare and ecologically important behaviours such as feeding, where correct classifications more than doubled following quality checks and data rebalancing introduced in our pipeline. C_LIO_LIOur approach is applicable across taxa and represents a key step towards advancing the burgeoning use of machine learning to remotely observe around-the-clock behaviours of free-ranging animals. Future work could include the integration of multiple data streams, such as accelerometer, audio, and GPS data, for model training and could be incorporated directly into our pipeline. C_LI

ecology↗

bioRxiv2024

Modeling the dynamics of EMT reveals genes associated with pan-cancer intermediate states and plasticity

Epithelial-mesenchymal transition (EMT) is a developmental cell state transition co-opted by cancer to drive metastasis and resistance. Stable EMT intermediate states play a particularly important role in cell state plasticity and confer metastatic potential. To explore the dynamics of EMT and identify marker genes of highly metastatic intermediate cells, we analyzed EMT across multiple tumor types and stimuli via mathematical modeling with single-cell RNA-sequencing (scRNA-seq) data. We identified pan-cancer genes consistently expressed or upregulated in EMT intermediate states, most of which were not previously annotated as markers of EMT. Using Bayesian parameter inference, we fit a simple mathematical model to scRNA-seq data, revealing tumor-specific transition rates. This mathematical model offers a framework to quantify EMT progression. A consensus analysis of differential intermediate expression, regulation, and model-derived dynamics identified marker genes associated with persistence of the intermediate EMT state. SFN and NRG1 emerged as genes with the strongest evidence for their role influencing intermediate EMT dynamics. Through analysis of an independent cell line, we verified the role of SFN as a marker intermediate EMT transition. Modeling and inference of genes associated with EMT dynamics offer means to find biomarkers and to identify therapeutic approaches to harness or reverse tumor-promoting cell state transitions driven by EMT.

systems biology↗

bioRxiv2024

Mutation of negative regulatory gene CEHC1 encoding an FBXO3 protein results in normoxic expression of HYDA genes in Chlamydomonas reinhardtii

Oxygen is known to prevent hydrogen production in Chlamydomonas, both by inhibiting the hydrogenase enzyme and by preventing the accumulation of HYDA-encoding transcripts. We developed a screen for mutants showing constitutive accumulation of HYDA1 transcripts in the presence of oxygen. A reporter gene required for ciliary motility, placed under the control of the HYDA1 promoter, conferred motility only in hypoxic conditions. By selecting for mutants able to swim even in the presence of oxygen we obtained strains that express the reporter gene constitutively. One mutant identified a gene encoding an F-box only protein 3 (FBXO3), known to participate in ubiquitylation and proteasomal degradation pathways in other eukaryotes. Transcriptome profiles revealed that the mutation, termed cehc1-1, leads to constitutive expression of HYDA1 and other genes regulated by hypoxia, and of many genes known to be targets of CRR1, a transcription factor in the nutritional copper signaling pathway. CRR1 was required for the constitutive expression of the HYDA1 reporter gene in cehc1-1 mutants. The CRR1 protein, which is normally degraded in Cu-supplemented cells, was stabilized in cehc1-1 cells, supporting the conclusion that CEHC1 acts to facilitate the degradation of CRR1. Our results reveal a novel negative regulator in the CRR1 pathway and possibly other pathways leading to complex metabolic changes associated with response to hypoxia.

plant biology↗

bioRxiv2024

Average Nucleotide Identity based Staphylococcus aureus strain grouping allows identification of strain-specific genes in the pangenome

Staphylococcus aureus causes both hospital and community acquired infections in humans worldwide. Due to the high incidence of infection S. aureus is also one of the most sampled and sequenced pathogens today, providing an outstanding resource to understand variation at the bacterial subspecies level. We processed and downsampled 83,383 public S. aureus Illumina whole genome shotgun sequences and 1,263 complete genomes to produce 7,954 representative substrains. Pairwise comparison of core gene Average Nucleotide Identity (ANI) revealed a natural boundary of 99.5% that could be used to define 145 distinct strains within the species. We found that intermediate frequency genes in the pangenome (present in 10-95% of genomes) could be divided into those closely linked to strain background ("strain-concentrated") and those highly variable within strains ("strain-diffuse"). Non-core genes had different patterns of chromosome location; notably, strain-diffuse associated with prophages, strain-concentrated with the vSa{beta} genome island and rare genes (<10% frequency) concentrated near the origin of replication. Antibiotic genes were enriched in the strain-diffuse class, while virulence genes were distributed between strain-diffuse, strain-concentrated, core and rare classes. This study shows how different patterns of gene movement help create strains as distinct subspecies entities and provide insight into the diverse histories of important S. aureus functions. ImportanceWe analyzed the genomic diversity of Staphylococcus aureus, a globally prevalent bacterial species that causes serious infections in humans. Our goal was to build a genetic picture of the different strains of S. aureus and which genes may be associated with them. We used a large public dataset (>84,000 genomes) that was re-processed and subsampled to remove redundancy. We found that individual genomes could be grouped into strains by sharing > 99.5% identical nucleotide sequence of the core part of their genome. We also showed that a portion of genes that are present in intermediate frequency in the species are strongly associated with some strains but completely absent from others, suggesting a role in strain-specificity. This work lays the foundation for understanding individual gene histories of the S. aureus species and also outlines strategies for processing large bacterial genomic datasets.

microbiology↗

bioRxiv2023

SOX4-SMARCA4 complex promotes glycolysis-dependent TNBC cell growth through transcriptional regulation of Hexokinase 2

Tumor cells rely on increased glycolytic capacity to promote cell growth and progression. While glycolysis is known to be upregulated in the majority of triple negative (TNBC) or basal-like subtype breast cancers, the mechanism remains unclear. Here, we used integrative genomic analyses to identify a subset of basal-like tumors characterized by increased expression of the oncogenic transcription factor SOX4 and its co-factor the SWI/SNF ATPase SMARCA4. These tumors are defined by unique gene expression programs that correspond with increased tumor proliferation and activation of key metabolic pathways, including glycolysis. Mechanistically, we demonstrate that the SOX4-SMARCA4 complex mediates glycolysis through direct transcriptional regulation of Hexokinase 2 (HK2) and that aberrant HK2 expression and altered glycolytic capacity are required to mediate SOX4-SMARCA4-dependent cell growth. Collectively, we have defined the SOX4-SMARCA4-HK2 signaling axis in basal-like breast tumors and established that this axis promotes metabolic reprogramming which is required to maintain tumor cell growth.

cancer biology↗

bioRxiv2023

In Vivo Antiviral Efficacy of LCTG-002, a Pooled, Purified Human Milk Secretory IgA product, Against SARS-CoV-2 in a Murine Model of COVID-19

Immunoglobulin A (IgA) is the most abundant antibody (Ab) in human mucosal compartments including the respiratory tract, with the secretory form of IgA (sIgA) being dominant and uniquely stable in these environments. sIgA is naturally found in human milk, which could be considered a global resource for this biologic, justifying the development of human milk sIgA as a dedicated airway therapeutic for respiratory infections such as SARS-CoV-2. In the present study, methods were therefore developed to efficiently extract human milk sIgA from donors who were either immunologically naive to SARS-CoV-2 (pooled as a control IgA) or had recovered from a PCR-confirmed SARS-CoV-2 infection that elicited high-titer anti-SARS-CoV-2 Spike sIgA Abs in their milk (pooled together to make LCTG-002). Mass spectrometry determined that proteins with a relative abundance of 1.0% or greater were all associated with sIgA. None of the proteins exhibited statistically significant differences between batches. Western blot demonstrated all batches consisted predominantly of sIgA. Compared to control IgA, LCTG-002 demonstrated significantly higher binding to Spike, and was also capable of blocking the Spike - ACE2 interaction in vitro with 6.3x greater potency compared to control IgA (58% inhibition at [~]240ug/mL). LCTG-002 was then tested in vivo for its capacity to reduce viral burden in the lungs of K18+hACE2 transgenic mice inoculated with SARS-CoV-2. LCTG-002 was demonstrated to significantly reduce SARS-CoV-2 titers in the lungs compared to control IgA when administered at either 250ug/day or 1 mg/day, as measured by TCID50, plaque forming units (PFU), and qRT-PCR, with a maximum reduction of 4.9 logs. This innovative study demonstrates that LCTG-002 is highly pure, efficacious, and well tolerated in vivo, supporting further development of milk-derived, polyclonal sIgA therapeutics against SARS-CoV-2 and other mucosal infections.

immunology↗