About

Marcus Ramsay Clark is a Principal Investigator at the University of Chicago, leading research focused on understanding immune cell behavior and disease mechanisms, particularly in the context of lupus nephritis. His lab investigates the signaling pathways, cellular interactions, and molecular features that underpin immune responses and inflammation, utilizing advanced imaging and computational techniques to analyze tissue biopsies and immune cell populations. His work aims to elucidate the molecular and cellular underpinnings of immune-mediated diseases, contributing to the development of targeted therapeutic strategies.

Research topics

• Biology
• Cell biology
• Chemistry
• Immunology
• Medicine
• Internal medicine
• Genetics
• Computational biology
• Computer Science
• Cancer research
• Molecular biology
• Geology
• Pathology
• Endocrinology
• Evolutionary biology

Selected publications

• High-fidelity multiclass instance segmentation of cells for spatial proteomics

  2026-03-04

  article

  High quality cell segmentations across cell classes remain elusive in spatial proteomics. Existing methods rely on expansion of cell nuclei, or on “pan-membrane” markers that are not actually uniformly expressed across all cells and tissues. Here, we present an extension of pseudo-spectral angle mapping (pSAM) for multiclass instance segmentation of cells. By predicting segmentations on class maps rather than images, cell shape is better captured. Improved representation of cell shape could help better infer interaction of cells in static images.

  PublisherDOI

• Foundation-model‑assisted characterization of mixed rejection of renal allograft

  2026-04-02

  article

  Kidney allograft rejection is categorized by pathologic changes driven by the innate and adaptive immune system recognition of the non-self-antigen of the allograft, often leading to allograft dysfunction. Using histopathological and immunological characteristics, acute transplant rejections are broadly categorized into antibody-mediated rejection (ABMR), which typically demonstrates glomerular and peritubular capillary injury, and T-cell mediated rejection (TCMR), which is categorized by lymphocytic infiltration of the tubules and interstitium. Mixed rejection (MR) is a third category that is currently understood as the co-occurrence of ABMR and TCMR, yet it remains difficult to assess the contributions of each mechanism. Here, we leverage deep learning on foundation-model extracted features of ABMR and TCMR to describe the pathology of MR. Our dataset consists of 32 kidney biopsies from patients diagnosed with ABMR, TCMR and MR. All biopsies were imaged with a 43-marker immunofluorescence panel using high-plex immunofluorescence imaging, plus label-free channels for tissue autofluorescence (TAF). Hematoxylin and eosin (H&E)- like images were generated by false coloring the nuclear stain channel and TAF images. Using the foundation model UNI, we extracted tile-level features from the whole slide images (WSIs) of all biopsies. We demonstrate that UNI embeddings of MR images overlap with both ABMR and TCMR. Furthermore, we trained a multi-layer perceptron on these embeddings to discriminate between ABMR and TCMR (AUC [95CI] = 0.714 [0.707, 0.721]). This classifier was used to predict which image patches from MR biopsies were most similar to ABMR and TCMR. Overall, we demonstrate a framework that leverages pre-trained foundation models to improve histopathological characterization of MR biopsies.

  PublisherDOI

• Deep profiling of lupus nephritis kidneys reveals dynamic changes in myeloid cells associated with disease progression

  Annals of the Rheumatic Diseases · 2026-04-01

  article
  PublisherDOI

• Jk DNA GAGA MOTIFS ARE REQUIRED FOR LOCAL NUCLEOSOME REMODELING AND Vk-Jk RECOMBINATION

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-01-23

  preprintSenior authorCorresponding

  Abstract Immunoreceptor gene recombination requires complementary 12 bp and 23 bp recombination signal sequences (RSSs). In addition, the RSSs that assemble the RAG proteins, recombination centers, must be accessible yet flanked by a 5’ nucleosome decorated with H3K4me3. In Drosophila , DNA GAGA motifs play an important role in nucleosome positioning. Herein, we report that 5’ to each functional Jk RSS is a DNA GAGA motif conserved across mammalian species. In mice, the GAGA motif 5’ to Jk1 regulated local RSS accessibility and 5’ nucleosome placement. Furthermore, it was required for Vk-Jk1 recombination. Murine Jk3 is nonfunctional, having mutations in both RSS and GAGA motifs. Restoring both GAGA and RSS motifs rescued Vk-Jk3 recombination. In contrast, restoring the RSS alone did not. Genome-wide, strong cryptic 23 RSSs were preferentially bound to nucleosomes. Furthermore, evolutionary selection against cRSS only occurred in the A Compartment of B lymphocytes, not embryonic stem cells. These data indicate that in developing B cells, nucleosome positioning both enables and restricts recombination to Jk. Furthermore, our data suggest an expanded definition of recombination center-associated RSSs to include a 5’ GAGA sequence that dictates the local epigenetic state required for gene recombination. Summary Recombination center assembly requires a specific epigenetic topology at recombination signal sequences. Herein, we report that conserved GAGA motifs 5’ to each Jk segment are required for establishing this epigenetic topology and subsequent local gene recombination.

  PublisherDOI

• Virtual multiplex immunofluorescence identifies lymphocyte subsets predictive of response to neoadjuvant therapy

  Therapeutic Advances in Medical Oncology · 2025-09-01

  articleOpen access

  Background: Hematoxylin and eosin (H&E) staining is routine in pathology but lacks cellular specificity. Multiplex immunofluorescence (mIF) captures spatial immune relationships in tumors, but cost and complexity limit clinical application. Novel approaches to yield similar information from readily available tumor histology are needed. Objectives: Develop and validate a novel deep learning tool capable of translating standard H&E-stained histopathology images into high-fidelity synthetic mIF images that preserve immune cell information predictive of treatment response in breast cancer. Design: Comparative model evaluation and predictive modeling in a retrospective breast cancer cohort. Methods: Core-needle biopsies from 17 triple-negative breast cancer cases underwent mIF imaging. Hematoxylin and eosin and mIF images for DAPI (nuclei), pan-CK (tumor), CD3/CD4/CD8 (T-cells), and CD20 (B cells) were aligned. A pipeline outperforming standard Pix2Pix and CycleGAN image translation networks was developed, "multiplex Synthetic Immunofluoresence Generated through H&E Translation" (mSIGHT), which integrates a registration network to overcome misalignment between the input and target images. Generated images were evaluated with pixel-level metrics and biological metrics, including cell density and cell-to-cell adjacency. The pipeline was then applied to an external cohort to assess associations between predicted immune features and pathologic response to neoadjuvant chemotherapy. Results: = 0.002), independent of receptor status, grade, and pathologist TIL annotations. Conclusion: The mSIGHT pipeline enables translation of routine H&E slides into virtual mIF images with interpretable immune biomarkers, offering a scalable and affordable alternative to multiplex imaging. It also identifies immune features predictive of therapeutic response and has the potential to assist in the personalization of neoadjuvant therapy.

  PublisherDOI

• Spatial Proximity Sequencing Maps Developmental Dynamics in the Germinal Center

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-10-28 · 1 citations

  preprintOpen access

  Spatial profiling of proteins and protein interactions is essential for immunology, signaling, development, and cancer. We present Spatial Proximity-Sequencing (Sprox-seq), a multi-omic technique that simultaneously measures proteins, protein complexes and mRNAs, where location of each molecule is also recorded. Sprox-seq profiled 32 proteins, 528 pairwise protein interactions and thousands of mRNAs across human tonsil tissues and germinal centers. Mapping protein interactions recapitulated RNA-defined tissue architecture in germinal centers, but also revealed much higher interaction complexity in the Light zone. Developmental trajectories inferred from protein interactions uncovered a B cell maturation pathway distinct from that inferred by RNA. Integrated protein-complex and mRNA analysis related spatially-enriched complexes with immune regulation and mitotic gene-expression pathways. Furthermore, Sprox-seq captured B cell-Follicular Dendritic Cell interactions mediated by the protein complex VLA-4-VCAM1 in the Light zone. Sprox-seq provides a multi-modal view of cell states and a powerful tool for studying protein and cellular interactions across tissues.

  PublisherOA PDFDOI

• Active Antigen-Specific Adaptive Immune Responses Are Shared among Patients with Progressive Fibrotic Interstitial Lung Disease

  American Journal of Respiratory and Critical Care Medicine · 2025-08-15 · 1 citations

  article

  Abstract Rationale Enlargement of lung-associated lymph nodes (LNs) predicts worse survival in all patients with interstitial lung disease (ILD). This phenomenon occurs in both connective tissue disease–associated ILD and, surprisingly, idiopathic pulmonary fibrosis (IPF), where immune-driven pathogenesis is controversial. Objectives To determine whether immune responses in the lung LNs of patients with ILD are antigen-specific and significant to pathology and etiology. Methods ILD lung LNs excised at transplant (30 IPF, 7 interstitial pneumonia with autoimmune features, 4 hypersensitivity pneumonitis, 5 connective tissue disease–associated ILD, 5 other ILD) and 36 donor control lung LNs were analyzed by spectral flow cytometry. Formalin-fixed lung LNs and OCT-fixed lung samples of patients with IPF were used to determine germinal center (GC) and antigen-specific responses. Serum autoantibody responses were measured by radioligand binding assay. Measurements and Main Results All patients with ILD revealed a common adaptive immune landscape of antigen responses in lung LNs characterized by the presence of GC B cells, T follicular helper cells, and activated T cells. Immunological synapses identified in the lung LNs demonstrated that antigen stimulation is ongoing in patients with ILD. Lung LN frequencies of T follicular helper and T regulatory cells correlated with circulating antibody concentrations to ABLIM1, a recently identified autoantigen expressed widely, including in aberrant basaloid cells that are uniquely found in fibrotic lungs. Conclusions Antigen-induced activation and development of GC in enlarged lung LNs represents a shared immunopathologic mechanism associated with progressive pulmonary fibrosis among patients with ILD, regardless of etiology. Autoantigens overexpressed in progressive pulmonary fibrosis may be key drivers of these GC responses.

  PublisherDOI

• Immune cell quantification of in situ inflammation partitions human lupus nephritis into mechanistic subtypes

  Journal of Clinical Investigation · 2025-09-04 · 5 citations

  articleOpen accessSenior author

  BACKGROUNDIn human lupus nephritis (LuN), tubulointerstitial inflammation (TII) is prognostically more important than glomerular inflammation. However, a comprehensive understanding of both TII complexity and heterogeneity is lacking.METHODSHerein, we used high-dimensional confocal microscopy, spatial transcriptomics, and specialized computer vision techniques to quantify immune cell populations and localize these within normal and diseased renal cortex structures. With these tools, we compared LuN to renal allograft rejection (RAR) and normal kidney tissues on 54 deidentified biopsies.RESULTSIn both LuN and RAR, the 33 characterized immune cell populations formed discrete subgroups whose constituents covaried in prevalence across biopsies. In both diseases, these covariant immune cell subgroups organized into the same unique niches. Therefore, inflammation could be resolved into trajectories representing the relative prevalence and density of cardinal immune cell members of each covariant subgroup. Indeed, in any one biopsy, the inflammatory state could be characterized by quantifying constituent immune cell trajectories. Remarkably, LuN heterogeneity could be captured by quantifying a few myeloid immune cell trajectories, while RAR was more complex with additional T cell trajectories.CONCLUSIONSOur studies identify rules governing renal inflammation and thus provide an approach for resolving LuN into discrete mechanistic categories.FUNDINGNIH (U19 AI 082724 [MRC], R01 AI148705 [MRC and ASC]), Chan Zuckerberg Biohub (MRC), and Lupus Research Alliance (MRC).

  PublisherOA PDFDOI

• Comparative analysis of multiplex microscopy images of human renal biopsy tissue

  2025-04-10

  articleSenior author

  In recent years, many significant advances in computer vision methods have led to an increasing interest in using biomedical images for precision medicine and biological discovery. This work aims to develop computational and statistical approaches to probe the in situ differences between two common forms of renal inflammation as captured using multiplex microscopy. Using our computational pipeline we collected and analyzed: 25 Lupus nephritis (LuN), 23 renal allograft rejection (RAR), and six kidney (KC) control samples. We segmented and annotated approximately 2.19 million cells using Cellpose2.0 for nuclear detection, and a decision-tree classifier for the multiclass annotation of renal cells, analogous to flow cytometry-based immunophenotyping. Using an MWU test, we observed a statistically significant enrichment of myeloid and T-cell lineages comparing LuN and RAR to the KC control. When we further use DBSCAN to analyze in situ cellular aggregates, we find two prominent CD14+MerTk+ macrophage-enriched and CD14+CD163+ macrophage-enriched cell "neighborhoods". Moreover, we apply our computational tool to capture Microtubule Organizing Center (mTOC) polarization between pairs of cells, which we leverage to capture cell-cell cognate immunity. In doing so, we find an enrichment of polarized B-cells: CD4+PD1+ICOS+ T-cells in the Mixed Rejection cohort in agreement with prior published literature. An expert ground truth reader had a 52.5% agreement with the mTOC polarization tool, above the random chance of 33.3%. This work paves the way for future research on targeted therapies addressing patients' unique immunological profiles, focusing on the myeloid cell compartment in LuN and myeloid:T-cell interactions in RAR.

  PublisherDOI

• Resolution of <i>in situ</i> inflammation in human lupus nephritis into principal immune cell trajectories

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-02-21

  preprintOpen accessSenior authorCorresponding

  ABSTRACT In human lupus nephritis (LuN), tubulointerstitial inflammation (TII) is prognostically more important than glomerular inflammation. However, a comprehensive understanding of both TII complexity and heterogeneity is lacking. Herein, we used high-dimensional confocal microscopy and specialized computer vision techniques to quantify immune cell populations and localize these within normal and diseased renal cortex structures. With these tools, we compared LuN to renal allograft rejection (RAR) and normal kidney. In both LuN and RAR, the 33 characterized immune cell populations formed discrete subgroups whose constituents co-varied in prevalence across biopsies. In both diseases, these co-variant immune cell subgroups organized into the same unique niches. Therefore, inflammation could be resolved into trajectories representing the relative prevalence and density of cardinal immune cell members of each co-variant subgroup. Indeed, in any one biopsy, the inflammatory state could be characterized by quantifying constituent immune cell trajectories. Remarkably, LuN heterogeneity could be captured by quantifying a few myeloid immune cell trajectories while RAR was more complex with additional T cell trajectories. Our studies identify rules governing renal inflammation and thus provide an approach for resolving LuN into discrete mechanistic categories. Abstract Figure Graphical Abstract

  PublisherOA PDFDOI

Recent grants

• The epigenetic reader BRWD1 in peripheral adaptive immunity

  NIH · $2.4M · 2019–2024

• Regulation of Ig-kappa recombination during B lymphopoiesis

  NIH · $2.3M · 2015–2020

• In vivo functions of Ig-beta ubiquitinylation

  NIH · $1.3M · 2012–2016

• NIH Grant R21GM076601

  NIH · $227k · 2008

• Role of CXCR4 in immunoglobulin light chain recombination

  NIH · $2.9M · 2021–2026

Frequent coauthors

• Maryellen L. Giger

  University of Chicago

  37 shared

• John C. Cambier

  33 shared

• Junting Ai

  University of Chicago

  31 shared

• Madeleine S. Durkee

  University of Chicago

  29 shared

• Malay Mandal

  28 shared

• Anthony Chang

  Délégation Paris 6

  27 shared

• Margaret Veselits

  26 shared

• Vladimir M. Liarski

  University of Chicago

  22 shared

Labs

• Clark LabPI

Awards & honors

• Translational Bridge Award from the Lupus Research Alliance…