Biology subjects

Khosla, A.

Publications and source records attributed to Khosla, A..

9 recordsLinked to original sources

bioRxiv2024

Spatial mapping of immunosuppressive cancer-associated fibroblast gene signatures in H&E-stained images using additive multiple instance learning

The relative abundance of cancer-associated fibroblast (CAF) subtypes influences a tumors response to treatment, especially immunotherapy. However, the extent to which the underlying tumor composition associates with CAF subtype-specific gene expression is unclear. Here, we describe an interpretable machine learning (ML) approach, additive multiple instance learning (aMIL), to predict bulk gene expression signatures from H&E-stained whole slide images (WSI), focusing on an immunosuppressive LRRC15+ CAF-enriched TGF{beta}-CAF signature. aMIL models accurately predicted TGF{beta}-CAF across various cancer types. Tissue regions contributing most highly to slide-level predictions of TGF{beta}-CAF were evaluated by ML models characterizing spatial distributions of diverse cell and tissue types, stromal subtypes, and nuclear morphology. In breast cancer, regions contributing most to TGF{beta}-CAF-high predictions ("excitatory") were localized to cancer stroma with high fibroblast density and mature collagen fibers. Regions contributing most to TGF{beta}-CAF-low predictions ("inhibitory") were localized to cancer epithelium and densely inflamed stroma. Fibroblast and lymphocyte nuclear morphology also differed between excitatory and inhibitory regions. Thus, aMIL enables a data-driven link between tissue phenotype and transcription, offering biological interpretability beyond typical black-box models.

cancer biology↗

bioRxiv2024

The transcription factor ATML1 maintains giant cell identity by inducing synthesis of its own (very) long-chain fatty acid-containing ligands

During development, cells not only adopt specialized identities but also maintain those identities. Endoreduplication is thought to maintain cell identity. High concentrations of ARABIDOPSIS THALIANA MERISTEM LAYER1 (ATML1) specify giant cell identity and induce endoreduplication in sepals. How different concentrations of ATML1 can specify different identities remains unclear. Here, we show that high concentrations of ATML1 induce the biosynthesis of both long-chain and very long-chain fatty acids (LCFAs/VLCFAs), and these fatty acids are required for the maintenance of giant cell identity. Inhibition of VLCFA biosynthesis causes endoreduplicated giant cells to resume division and lose their identity, indicating that endoreduplication is not sufficient to maintain cell identity. Structural predictions suggest that LCFA-containing lipids bind to the START domain 2 of ATML1, causing ATML1 dimerization and its auto-activation. Our data and modeling imply that ATML1 induces biosynthesis of its own lipid ligands in a positive feedback loop, shedding light on the intricate network dynamics that specify and maintain giant cell identity. Teaser: Endoreduplicated cells in Arabidopsis thaliana sepals divide and de-differentiate in the absence of VLCFA biosynthesis.

plant biology↗

bioRxiv2023

Cell-type-specific nuclear morphology predicts genomic instability and prognosis in multiple cancer types

While alterations in nucleus size, shape, and color are ubiquitous in cancer, comprehensive quantification of nuclear morphology across a whole-slide histologic image remains a challenge. Here, we describe the development of a pan-tissue, deep learning-based digital pathology pipeline for exhaustive nucleus detection, segmentation, and classification and the utility of this pipeline for nuclear morphologic biomarker discovery. Manually-collected nucleus annotations were used to train an object detection and segmentation model for identifying nuclei, which was deployed to segment nuclei in H&E-stained slides from the BRCA, LUAD, and PRAD TCGA cohorts. Interpretable features describing the shape, size, color, and texture of each nucleus were extracted from segmented nuclei and compared to measurements of genomic instability, gene expression, and prognosis. The nuclear segmentation and classification model trained herein performed comparably to previously reported models. Features extracted from the model revealed differences sufficient to distinguish between BRCA, LUAD, and PRAD. Furthermore, cancer cell nuclear area was associated with increased aneuploidy score and homologous recombination deficiency. In BRCA, increased fibroblast nuclear area was indicative of poor progression-free and overall survival and was associated with gene expression signatures related to extracellular matrix remodeling and anti-tumor immunity. Thus, we developed a powerful pan-tissue approach for nucleus segmentation and featurization, enabling the construction of predictive models and the identification of features linking nuclear morphology with clinically-relevant prognostic biomarkers across multiple cancer types.

cancer biology↗

bioRxiv2022

A robust modular automated neuroimaging pipeline for model inputs to TheVirtualBrain

TheVirtualBrain, an open-source platform for large-scale network modelling, can be personalized to an individual using a wide range of neuroimaging modalities. With the growing number and scale of neuroimaging data sharing initiatives of both healthy and clinical populations comes an opportunity to create large and heterogeneous sets of dynamic network models to better understand individual differences in network dynamics and their impact on brain health. Here we present TheVirtualBrain-UK Biobank pipeline, a robust, automated and open-source brain image processing solution to address the expanding scope of TheVirtualBrain project. Our pipeline generates connectome-based modelling inputs compatible for use with TheVirtualBrain. We leverage the existing multimodal MRI processing pipeline from the UK Biobank made for use with a variety of brain imaging modalities. We add various features and changes to the original UK Biobank implementation specifically for informing large-scale network models, including user-defined parcellations for the construction of matching whole-brain functional and structural connectomes. Changes also include detailed reports for quality control of all modalities, a streamlined installation process, modular software packaging, updated software versions, and support for various publicly available datasets. The pipeline has been tested on both healthy and clinical populations and is robust to the morphological changes observed in aging and dementia. In this paper, we describe these and other pipeline additions and modifications in detail, as well as how this pipeline fits into the TheVirtualBrain ecosystem.

neuroscience↗

bioRxiv2021

START domain mediates Arabidopsis GLABRA2 transcription factor dimerization and turnover independently of homeodomain DNA binding

Class IV homeodomain leucine-zipper transcription factors (HD-Zip IV TFs) are key regulators of epidermal differentiation that are characterized by a DNA-binding homeodomain (HD) in conjunction with a lipid-binding domain termed START (Steroidogenic Acute Regulatory (StAR)-related lipid Transfer). Previous work established that the START domain of GLABRA2 (GL2), a HD-Zip IV member from Arabidopsis, is required for transcription factor activity. Here, we address the functions and possible interactions of START and the HD in DNA binding, dimerization, and protein turnover. Deletion analysis of the HD and missense mutations of a conserved lysine (K146) result in phenotypic defects in leaf trichomes, root hairs and seed mucilage, similar to those observed for START domain mutants, despite nuclear localization of the respective proteins. In vitro and in vivo experiments demonstrate that while HD mutations impair binding to target DNA, the START domain is dispensable for DNA binding. Vice versa, protein interaction assays reveal impaired GL2 dimerization for multiple alleles of START mutants, but not HD mutants. Using in vivo cycloheximide chase experiments, we provide evidence for the role of START, but not HD, in maintaining protein stability. This work advances our mechanistic understanding of HD-Zip TFs as multidomain regulators of epidermal development in plants.

plant biology↗

bioRxiv2021

Arabidopsis PROTODERMAL FACTOR2 binds lysophosphatidylcholines and transcriptionally regulates phospholipid metabolism

Plant homeodomain leucine-zipper IV (HD-Zip IV) transcription factors (TFs) contain an evolutionarily conserved steroidogenic acute regulatory protein (StAR)-related lipid transfer (START) domain. The START domain is required for TF activity; however, its presumed role as a lipid sensor is not well understood. Here we used tandem affinity purification from Arabidopsis cell cultures to demonstrate that PROTODERMAL FACTOR2 (PDF2), a representative family member which controls epidermal differentiation, recruits lysophosphatidylcholines in a START-dependent manner. In vitro assays with recombinant protein verified that a missense mutation in a predicted ligand contact site reduces lysophospholipid binding. We additionally uncovered that PDF2 controls the expression of phospholipid-related target genes by binding to a palindromic octamer with consensus to a phosphate (Pi) response element. Phospholipid homeostasis and elongation growth were altered in pdf2 mutants according to Pi availability. Cycloheximide chase experiments further revealed a role for START in maintaining protein levels, and Pi limitation resulted in enhanced protein destabilization, suggesting a mechanism by which lipid binding controls TF activity. We propose that the START domain serves as a molecular sensor for membrane phospholipid status in the epidermis. Overall our data provide insights towards understanding how the lipid metabolome integrates Pi availability with gene expression.

plant biology↗

bioRxiv2021

Rapid analysis of strigolactone receptor activity in a Nicotiana benthamiana dwarf14 mutant

DWARF14 (D14) is an [a]/{beta}-hydrolase and receptor for the plant hormone strigolactone (SL) in angiosperms. Upon SL perception, D14 works with MORE AXILLARY GROWTH2 (MAX2) to trigger polyubiquitination and degradation of DWARF53(D53)-type proteins in the SUPPRESSOR OF MAX2 1-LIKE (SMXL) family. We used CRISPR-Cas9 to generate knockout alleles of the two homoeologous D14 genes in the Nicotiana benthamiana genome. The Nbd14a,b double mutant had several phenotypes that are consistent with the loss of SL perception in other plants, including increased axillary bud outgrowth, reduced height, shortened petioles, and smaller leaves. A ratiometric fluorescent reporter system was used to monitor degradation of SMXL7 from Arabidopsis thaliana (AtSMXL7) after transient expression in N. benthamiana and treatment with the strigolactone analog GR24. AtSMXL7 was degraded after treatment with GR245DS, which has the stereochemical configuration of SLs, as well as its enantiomer GR24ent-5DS . In Nbd14a,b leaves, AtSMXL7 abundance was unaffected by GR24. Transient coexpression of AtD14 with the AtSMXL7 reporter in Nbd14a,b restored the degradation response to GR24, but required an active catalytic triad. With this platform, we evaluated the ability of several AtD14 mutants that had not been characterized in plants to target AtSMXL7 for degradation.

plant biology↗

bioRxiv2020

Dense, high-resolution mapping of cells and tissues from pathology images for the interpretable prediction of molecular phenotypes in cancer

While computational methods have made substantial progress in improving the accuracy and throughput of pathology workflows for diagnostic, prognostic, and genomic prediction, lack of interpretability remains a significant barrier to clinical integration. In this study, we present a novel approach for predicting clinically-relevant molecular phenotypes from histopathology whole-slide images (WSIs) using human-interpretable image features (HIFs). Our method leverages >1.6 million annotations from board-certified pathologists across >5,700 WSIs to train deep learning models for high-resolution tissue classification and cell detection across entire WSIs in five cancer types. Combining cell- and tissue-type models enables computation of 607 HIFs that comprehensively capture specific and biologically-relevant characteristics of multiple tumors. We demonstrate that these HIFs correlate with well-known markers of the tumor microenvironment (TME) and can predict diverse molecular signatures, including immune checkpoint protein expression and homologous recombination deficiency (HRD). Our HIF-based approach provides a novel, quantitative, and interpretable window into the composition and spatial architecture of the TME.

bioinformatics↗

bioRxiv2020

A series of dual-reporter vectors for ratiometric analysis of protein abundance in plants

Ratiometric reporter systems enable comparisons of the abundance of a protein of interest, or "target," relative to a reference protein. Both proteins are encoded on a single transcript but are separated during translation. This arrangement bypasses the potential for discordant expression that can arise when the target and reference proteins are encoded by separate genes. We generated a set of 18 Gateway-compatible vectors termed pRATIO that combine a variety of promoters, fluorescent and bioluminescent reporters, and 2A "self-cleaving" peptides. These constructs are easily modified to produce additional combinations or introduce new reporter proteins. We found that mScarlet-I provides the best signal-to-noise ratio among several fluorescent reporter proteins during transient expression experiments in Nicotiana benthamiana. Firefly and Gaussia luciferase also produce high signal-to-noise in N. benthamiana. As proof of concept, we used this system to investigate whether degradation of the receptor KAI2 after karrikin treatment is influenced by its subcellular localization. KAI2 is normally found in the cytoplasm and the nucleus of plant cells. In N. benthamiana, karrikin-induced degradation of KAI2 was only observed when it was retained in the nucleus. These vectors are tools to easily monitor in vivo the abundance of a protein that is transiently expressed in plants, and will be particularly useful for investigating protein turnover in response to different stimuli.

plant biology↗