Biology subjects

Koehler, R.

Publications and source records attributed to Koehler, R..

7 recordsLinked to original sources

bioRxiv2025

A single-cell cytokine dictionary of human peripheral blood

Cytokines orchestrate immune responses, yet we still lack a comprehensive understanding of their specific effects across human immune cells due to their pleiotropy, context dependence and extensive functional redundancy. Here, we present a Human Cytokine Dictionary, created from high-resolution single-cell transcriptomes of 9,697,974 human peripheral blood mononuclear cells (PBMC) from 12 donors stimulated in vitro with 90 different cytokines. We describe donor-specific response variation and uncover robust consensus cytokine signatures across individuals. We then delineate similarities between cytokine response profiles, and derive cytokine-induced immune programs that organize responsive genes into data-driven, biologically interpretable functional modules. By integrating cell type-specific responses with expression of cytokines, we infer higher-order cell-to-cell and cytokine-to-cytokine communication networks exemplified by an IL-32-{beta}-initiated signaling cascade, which rewires myeloid programs by inducing neutrophil-recruiting factors while suppressing Th1-responses and promoting IL-10-family cytokines. Finally, we show how the Human Cytokine Dictionary enables the interpretation of cytokine-driven immune responses in other studies and disease contexts, including systemic lupus erythematosus, multiple sclerosis, and non-small cell lung carcinoma. Together, the Human Cytokine Dictionary constitutes the first comprehensive cell type-resolved transcriptional screen of human cytokine responses and provides an essential open-access, easy-to-use community resource with accompanying software package to advance our understanding of cytokine biology in human disease and guide therapeutic discovery.

immunology↗

bioRxiv2025

Mapping EGFR1 sorting domains in endosomes with a calibrated 3D expansion microscopy toolkit

Endosomes are nanoscale intracellular compartments that sort and recycle cell surface receptors such as epidermal growth factor receptor 1 (EGFR1). Nanometre-scale interactions and co-clustering of signalling proteins, cargo, and the membrane are critical to this process. Direct visualisation of these interactions has been hindered by the limited 3D resolution achievable with conventional and super-resolution microscopies. Here, we present the adaptation of expansion microscopy (ExM) to visualise, and quantify nanoclusters of endosomal proteins of human retinal pigment epithelial (RPE-1) cells. A 3D distortion analysis was developed leveraging the Farneback optical flow principle for detecting anisotropies in the hydrogel expansion. Analysis of pre- and post- ExM image volumes for 3D anisotropies revealed under-expansion of cytoplasmic regions within ExM hydrogels, often leading to over-estimation of size and distance measurements of small compartments such as endosomes. A self-assembling protein nanocage that reports the true local and nanoscale expansion factor was genetically introduced into the cells to calibrate ExM images of cytoplasmic regions containing endosomes. To stimulate and visualise the internalisation and sorting of EGFR1 in mammalian cells, a pulse-chase protocol was carried out with fluorescently-tagged EGF. The cells were subsequently fixed at 15- and 30- minute time points and subjected to 10-fold ExM and multiplexed 3D Airyscan microscopy to map cargo and EGFR1 vs other endosomal proteins. A volume tracing pipeline was developed to visualise the changes in the labelled EGF and EGFR1 densities at the limiting membrane of the endosomes. With multiplexed 3D ExM image volumes, we observed the enrichment of both EGF and EGFR1 in the endosomal interior and the accumulation of endosomal protein Rab5a near the limiting membrane during this early maturation of the endosomes. Taken together, the multiplexed 3D ExM toolkit offers a quantitative framework for visualising and measuring the intrinsic biology of small sub-cellular organelles like endosomes at true molecular-scale resolution.

biophysics↗

bioRxiv2025

A Generalizable Machine Learning Model for Early Detection of Hepatocellular Carcinoma Using Bisulfite Sequencing Data

PURPOSEEarly detection of hepatocellular carcinoma (HCC) is critical for improving patient outcomes, yet current screening tools lack sensitivity and specificity. We demonstrate a flexible machine learning framework for HCC detection using methylation profiles from bisulfite sequencing across multiple assay platforms and sample types. The framework supports a "split-and-filter" approach that routes each sequenced sample to an assay-matched classifier without requiring cross-assay feature compatibility. PATIENTS AND METHODSWe constructed assay-specific classifiers using four independent public methylation datasets ([~]2,500 total samples) representing distinct bisulfite-sequencing technologies: GSE93203 (MCB-targeted hypermethylation), GSE63775 (MCTA-Seq tandem-repeat hypermethylation), PRJCA001372 (HBV-integration-associated hypomethylation), and the HCC subset of GSE149438 (EpiPanGIDX - DMR-level hyper- and hypomethylation). Separate models were trained using using the biologically relevant features for each dataset and evaluated in two independent blind validation datasets: published tissue WGBS (24 samples: 12 early-stage HCC, 12 matched controls; PRJNA984754) and a new preclinical plasma cfDNA WGBS dataset (12 samples) generated at a commercial sequencing laboratory. Limited feature overlap among assays precluded a single unified model. Instead, overlapping features enabled construction of a proof-of-concept meta-classifier for sample routing across assay-specific models. RESULTSAssay-specific cfDNA models, trained independently on CpG sites from original publications, were evaluated using the biopsy(tissue) dataset and a new plasma dataset as blind validation. All four assay-specific models generalized well to the validation data, with accuracies of 83.5%-100%. In the validation with the plasma cfDNA samples, the best-performing classifier (among XGBoost, Random Forest, and Logistic Regression) for each public dataset achieved 80-100% sensitivity and 86-100% specificity, with all Stage 2 cases correctly detected across models. The single Stage 1A case showed methylation levels overlapping with cirrhotic controls, consistent with biological expectations. Despite this, a couple of the models predicted this correctly, showing greater sensitivity to Stage 1 cancer. CONCLUSIONA generalizable framework for early detection of HCC composed of assay-specific classifiers and a meta-classifier is described. This architecture readily accommodates addition of new assays via feature-matched models and meta-classifications. Larger, prospectively collected studies are necessary to confirm performance and enable clinical translation.

bioinformatics↗

bioRxiv2025

Tahoe-100M: A Giga-Scale Single-Cell Perturbation Atlas for Context-Dependent Gene Function and Cellular Modeling

Building predictive models of the cell requires systematically mapping how perturbations reshape each cells state, function, and behavior. Here, we present Tahoe-100M, a giga-scale single-cell atlas of 100 million transcriptomic profiles measuring how each of 1,100 small-molecule perturbations impact cells across 50 cancer cell lines. Our high-throughput Mosaic platform, composed of a highly diverse and optimally balanced "cell village", reduces batch effects and enables parallel profiling of thousands of conditions at single-cell resolution at an unprecedented scale. As the largest single-cell dataset to date, Tahoe-100M enables artificial-intelligence (AI)-driven models to learn context-dependent functions, capturing fundamental principles of gene regulation and network dynamics. Although we leverage cancer models and pharmacological compounds to create this resource, Tahoe-100M is fundamentally designed as a broadly applicable perturbation atlas and supports deeper insights into cell biology across multiple tissues and contexts. By publicly releasing this atlas, we aim to accelerate the creation and development of robust AI frameworks for systems biology, ultimately improving our ability to predict and manipulate cellular behaviors across a wide range of applications.

genomics↗

bioRxiv2023

In vivo optimized three-photon imaging of intact mouse tibia links plasma cell motility to functional states in the bone marrow.

Intravital multi-photon imaging of the bone marrow is crucial to the study of cellular dynamics, communication with the microenvironment and functions, however, imaging of deep tissue areas is challenging and minimally invasive methods for deep-marrow imaging in intact long bones are needed. We developed a high pulse energy 1650 nm laser prototype, which permits to surpass >100 {micro}m thick cortical bone and to perform three-photon microscopy (3PM) in more than 400 {micro}m depth in the marrow cavity of intact mouse tibia in vivo. Its unique 3 and 4 MHz laser repetition rates allowed us to analyze motility patterns of rare cells over large fields of view deep within the unperturbed marrow. In this way, we found a bi-modal migratory behavior of marrow plasma cells. Besides, the analysis of third harmonics generation (THG) in the tibia identified this signal to be a label-free indicator of the abundance of cellular organelles, in particular the endoplasmic reticulum, reflecting protein biosynthesis capacity. We found that only one third of the plasma cells in the tibia marrow of adult mice have a strong THG signal and, thus, a high protein synthesis capacity, while the other two thirds of plasma cells display a low THG signal. Finally, we identified an inverse link between migratory behavior and THG signal strength in marrow plasma cells. As in these cells, the protein biosynthesis capacity indicated by a strong THG signal is mainly associated with antibody secretion, we could relate motility to functional states of plasma cells in vivo. Our 3PM method retains the ability to connect cellular dynamics to protein biosynthesis capacity in various marrow cell types beyond plasma cells, as THG is a ubiquitous signal, opening new perspectives on understanding how tissue microenvironment impacts on cellular functions in the bone marrow.

biophysics↗

bioRxiv2022

High sensitivity single cell RNA sequencing with split pool barcoding

Single cell RNA sequencing (scRNA-seq) has become a core tool for researchers to understand biology. As scRNA-seq has become more ubiquitous, many applications demand higher scalability and sensitivity. Split-pool combinatorial barcoding makes it possible to scale projects to hundreds of samples and millions of cells, overcoming limitations of previous droplet based technologies. However, there is still a need for increased sensitivity for both droplet and combinatorial barcoding based scRNA-seq technologies. To meet this need, here we introduce an updated combinatorial barcoding method for scRNA-seq with dramatically improved sensitivity. To assess performance, we profile a variety of sample types, including cell lines, human peripheral blood mononuclear cells (PBMCs), mouse brain nuclei, and mouse liver nuclei. When compared to the previously best performing approach, we find up to a 2.6-fold increase in unique transcripts detected per cell and up to a 1.8-fold increase in genes detected per cell. These improvements to transcript and gene detection increase the resolution of the resulting data, making it easier to distinguish cell types and states in heterogeneous samples. Split-pool combinatorial barcoding already enables scaling to millions of cells, the ability to perform scRNA-seq on previously fixed and frozen samples, and access to scRNA-seq without the need to purchase specialized lab equipment. Our hope is that by combining these previous advantages with the dramatic improvements to sensitivity presented here, we will elevate the standards and capabilities of scRNA-seq for the broader community.

genomics↗

bioRxiv2017

A Robust Targeted Sequencing Approach For Low Input And Variable Quality DNA From Clinical Samples

Next-generation deep sequencing of gene panels is being adopted as a diagnostic test to identify actionable mutations in cancer patient samples. However, clinical samples, such as formalin-fixed, paraffin-embedded specimens, frequently provide low quantities of degraded, poor quality DNA. To overcome these issues, many sequencing assays rely on extensive PCR amplification leading to an accumulation of bias and artifacts. Thus, there is a need for a targeted sequencing assay that performs well with DNA of low quality and quantity without relying on extensive PCR amplification. We evaluate the performance of a targeted sequencing assay based on Oligonucleotide Selective Sequencing, which permits the enrichment of genes and regions of interest and the identification of sequence variants from low amounts of damaged DNA. This assay utilizes a repair process adapted to clinical FFPE samples, followed by adaptor ligation to single stranded DNA and a primer-based capture technique. Our approach generates sequence libraries of high fidelity with reduced reliance on extensive PCR amplification - this facilitates the accurate assessment of copy number alterations in addition to delivering accurate SNV and indel detection. We apply this method to capture and sequence the exons of a panel of 130 cancer-related genes, from which we obtain high read coverage uniformity across the targeted regions at starting input DNA amounts as low as 10 ng per sample. We further demonstrate the performance of this assay using a series of reference DNA samples, and by identifying sequence variants in DNA from matched clinical samples originating from different tissue types.

genomics↗