ABSTRACT
Image-based spatial transcriptomics methods enable transcriptome-scale gene expression measurements with spatial information but require complex, manually tuned analysis pipelines. We present Polaris, an analysis pipeline for image-based spatial transcriptomics that combines deep-learning models for cell segmentation and spot detection with a probabilistic gene decoder to quantify single-cell gene expression accurately. Polaris offers a unifying, turnkey solution for analyzing spatial transcriptomics data from multiplexed error-robust FISH (MERFISH), sequential fluorescence in situ hybridization (seqFISH), or in situ RNA sequencing (ISS) experiments. Polaris is available through the DeepCell software library (https://github.com/vanvalenlab/deepcell-spots) and https://www.deepcell.org.
Subject(s)
Deep Learning , Gene Expression Profiling , In Situ Hybridization, Fluorescence , Transcriptome , In Situ Hybridization, Fluorescence/methods , Transcriptome/genetics , Gene Expression Profiling/methods , Software , Humans , Single-Cell Analysis/methods , Image Processing, Computer-Assisted/methods , Single Molecule Imaging/methods , Animals , Supervised Machine LearningABSTRACT
Gut inflammation involves contributions from immune and non-immune cells, whose interactions are shaped by the spatial organization of the healthy gut and its remodeling during inflammation. The crosstalk between fibroblasts and immune cells is an important axis in this process, but our understanding has been challenged by incomplete cell-type definition and biogeography. To address this challenge, we used multiplexed error-robust fluorescence in situ hybridization (MERFISH) to profile the expression of 940 genes in 1.35 million cells imaged across the onset and recovery from a mouse colitis model. We identified diverse cell populations, charted their spatial organization, and revealed their polarization or recruitment in inflammation. We found a staged progression of inflammation-associated tissue neighborhoods defined, in part, by multiple inflammation-associated fibroblasts, with unique expression profiles, spatial localization, cell-cell interactions, and healthy fibroblast origins. Similar signatures in ulcerative colitis suggest conserved human processes. Broadly, we provide a framework for understanding inflammation-induced remodeling in the gut and other tissues.
Subject(s)
Colitis, Ulcerative , Colitis , Animals , Humans , Mice , Colitis/metabolism , Colitis/pathology , Colitis, Ulcerative/metabolism , Colitis, Ulcerative/pathology , Fibroblasts/metabolism , Fibroblasts/pathology , In Situ Hybridization, Fluorescence/methods , Inflammation/metabolism , Inflammation/pathology , Cell Communication , Gastrointestinal Tract/metabolism , Gastrointestinal Tract/pathologyABSTRACT
Image-based spatial transcriptomics methods enable transcriptome-scale gene expression measurements with spatial information but require complex, manually-tuned analysis pipelines. We present Polaris, an analysis pipeline for image-based spatial transcriptomics that combines deep learning models for cell segmentation and spot detection with a probabilistic gene decoder to quantify single-cell gene expression accurately. Polaris offers a unifying, turnkey solution for analyzing spatial transcriptomics data from MERFSIH, seqFISH, or ISS experiments. Polaris is available through the DeepCell software library (https://github.com/vanvalenlab/deepcell-spots) and https://www.deepcell.org.
ABSTRACT
Gut inflammation involves contributions from immune and non-immune cells, whose interactions are shaped by the spatial organization of the healthy gut and its remodeling during inflammation. The crosstalk between fibroblasts and immune cells is an important axis in this process, but our understanding has been challenged by incomplete cell-type definition and biogeography. To address this challenge, we used MERFISH to profile the expression of 940 genes in 1.35 million cells imaged across the onset and recovery from a mouse colitis model. We identified diverse cell populations; charted their spatial organization; and revealed their polarization or recruitment in inflammation. We found a staged progression of inflammation-associated tissue neighborhoods defined, in part, by multiple inflammation-associated fibroblasts, with unique expression profiles, spatial localization, cell-cell interactions, and healthy fibroblast origins. Similar signatures in ulcerative colitis suggest conserved human processes. Broadly, we provide a framework for understanding inflammation-induced remodeling in the gut and other tissues.
ABSTRACT
Single-molecule spatial transcriptomics protocols based on in situ sequencing or multiplexed RNA fluorescent hybridization can reveal detailed tissue organization. However, distinguishing the boundaries of individual cells in such data is challenging and can hamper downstream analysis. Current methods generally approximate cells positions using nuclei stains. We describe a segmentation method, Baysor, that optimizes two-dimensional (2D) or three-dimensional (3D) cell boundaries considering joint likelihood of transcriptional composition and cell morphology. While Baysor can take into account segmentation based on co-stains, it can also perform segmentation based on the detected transcripts alone. To evaluate performance, we extend multiplexed error-robust fluorescence in situ hybridization (MERFISH) to incorporate immunostaining of cell boundaries. Using this and other benchmarks, we show that Baysor segmentation can, in some cases, nearly double the number of cells compared to existing tools while reducing segmentation artifacts. We demonstrate that Baysor performs well on data acquired using five different protocols, making it a useful general tool for analysis of imaging-based spatial transcriptomics.
Subject(s)
Single-Cell Analysis , Transcriptome , Gene Expression Profiling/methods , In Situ Hybridization, Fluorescence/methods , RNA/analysis , Single-Cell Analysis/methods , Transcriptome/geneticsABSTRACT
To investigate the cyanylated cysteine vibrational probe group's ability to report on binding-induced changes along a protein-protein interface, the probe group was incorporated at several sites in a peptide of the calmodulin (CaM)-binding domain of skeletal muscle myosin light chain kinase. Isothermal titration calorimetry was used to determine the binding thermodynamics between calmodulin and each peptide. For all probe positions, the binding affinity was nearly identical to that of the unlabeled peptide. The CN stretching infrared band was collected for each peptide free in solution and bound to calmodulin. Binding-induced shifts in the IR spectral frequencies were correlated with estimated solvent accessibility based on molecular dynamics simulations. This work generally suggests (1) that site-specific incorporation of this vibrational probe group does not cause major perturbations to its local structural environment and (2) that this small probe group might be used quite broadly to map dynamic protein-binding interfaces. However, site-specific perturbations due to artificial labeling groups can be somewhat unpredictable and should be evaluated on a site-by-site basis through complementary measurements. A fully quantitative, simulation-based interpretation of the rich probe IR spectra is still needed but appears to be possible given recent advances in simulation techniques.
Subject(s)
Calmodulin/metabolism , Cysteine/metabolism , Myosin-Light-Chain Kinase/metabolism , Peptides/metabolism , Amino Acid Sequence , Animals , Binding Sites , Calmodulin/chemistry , Cysteine/chemistry , Drosophila melanogaster , Models, Molecular , Myosin-Light-Chain Kinase/chemistry , Nitriles/analysis , Nitriles/metabolism , Peptides/chemistry , Protein Binding , Protein Interaction Domains and Motifs , Protein Interaction Maps , Rabbits , Spectrophotometry, Infrared/methods , ThermodynamicsABSTRACT
A quantitative connection between molecular dynamics simulations and vibrational spectroscopy of probe-labeled systems would enable direct translation of experimental data into structural and dynamical information. To constitute this connection, all-atom molecular dynamics (MD) simulations were performed for two SCN probe sites (solvent-exposed and buried) in a calmodulin-target peptide complex. Two frequency calculation approaches with substantial nonelectrostatic components, a quantum mechanics/molecular mechanics (QM/MM)-based technique and a solvatochromic fragment potential (SolEFP) approach, were used to simulate the infrared probe line shapes. While QM/MM results disagreed with experiment, SolEFP results matched experimental frequencies and line shapes and revealed the physical and dynamic bases for the observed spectroscopic behavior. The main determinant of the CN probe frequency is the exchange repulsion between the probe and its local structural neighbors, and there is a clear dynamic explanation for the relatively broad probe line shape observed at the "buried" probe site. This methodology should be widely applicable to vibrational probes in many environments.
ABSTRACT
Seven native residues on the regulatory protein calmodulin, including three key methionine residues, were replaced (one by one) by the vibrational probe amino acid cyanylated cysteine, which has a unique CN stretching vibration that reports on its local environment. Almost no perturbation was caused by this probe at any of the seven sites, as reported by CD spectra of calcium-bound and apo calmodulin and binding thermodynamics for the formation of a complex between calmodulin and a canonical target peptide from skeletal muscle myosin light chain kinase measured by isothermal titration. The surprising lack of perturbation suggests that this probe group could be applied directly in many protein-protein binding interfaces. The infrared absorption bands for the probe groups reported many dramatic changes in the probes' local environments as CaM went from apo- to calcium-saturated to target peptide-bound conditions, including large frequency shifts and a variety of line shapes from narrow (interpreted as a rigid and invariant local environment) to symmetric to broad and asymmetric (likely from multiple coexisting and dynamically exchanging structures). The fast intrinsic time scale of infrared spectroscopy means that the line shapes report directly on site-specific details of calmodulin's variable structural distribution. Though quantitative interpretation of the probe line shapes depends on a direct connection between simulated ensembles and experimental data that does not yet exist, formation of such a connection to data such as that reported here would provide a new way to evaluate conformational ensembles from data that directly contains the structural distribution. The calmodulin probe sites developed here will also be useful in evaluating the binding mode of calmodulin with many uncharacterized regulatory targets.