Edward W. Roberts

Multiplex immunofluorescent imaging enables deep spatial profiling of protein expression in tissues but is often limited by reliance on proprietary reagents, dedicated hardware, and closed analysis ecosystems. Here we present CYCLOPS (Cyclic Open Platform for Spatial Proteomics), an end-to-end, open-source workflow for cyclic multiplex imaging and single-cell phenotyping using standard microscopy infrastructure. CYCLOPS integrates an Arduino-based automated fluidics system, an open-chamber stage insert, and antibody-oligonucleotide conjugation based entirely on published chemistries and off-the-shelf components. We demonstrate robust and reproducible antibody conjugation, high-quality multiplexed staining, and stable imaging across >10 cycles with minimal drift (<1 μm) and consistent fluorescence retention with low signal carry-over. The system supports efficient buffer exchange and consistent performance across multiple markers and imaging rounds. Using confocal microscopy, the workflow is compatible with three-dimensional imaging, enabling multiplexed analysis of volumetric tissue structures. To enable quantitative analysis, we establish an open-source image processing and analysis pipeline for single-cell feature extraction and phenotypic classification, avoiding reliance on proprietary software or black-box workflows. This framework integrates image registration, segmentation, and supervised classification to generate biologically interpretable single-cell data. Together, CYCLOPS provides a flexible and accessible platform for cyclic multiplex imaging, lowering barriers to adoption and enabling broader use of spatial proteomics across diverse research settings. This accessible framework democratizes high-plex imaging by enabling any laboratory with a standard confocal microscope to perform iterative multiplexing without reliance on proprietary reagents or hardware.

T cell metabolism increases upon activation, underpinning immune effector functions. Nicotinamide adenine dinucleotide (NAD/H) is an essential redox cofactor for glycolysis and mitochondrial substrate oxidation. It’s phosphorylation to NADP/H regulates reactive oxygen species (ROS) abundance. NAD/H levels increase upon T cell activation, but synthesis pathways and implications are not fully characterised. Here, we interrogate the role of the NAD/H-synthesis enzyme nicotinamide riboside kinase 1 (NRK1), the expression of which increases upon stimulation of both human and murine CD4 ⁺ T cells. Functionally, NRK1 activity restrains activation and cytokine production of CD4 ⁺ T cells while promoting survival. These activities are linked to increased NRK1 expression in the cytoplasm, where it locally raises NAD/H levels. This supports glycolysis, but more profoundly impacts cytoplasmic NADP/H generation, thereby controlling ROS abundance and nuclear NFAT translocation. During fungal and viral infection, T-cell-intrinsic NRK1 maintains effector CD4 ⁺ T cell abundance within affected tissues and draining lymph nodes, supporting infection control. Taken together, these data confirm that subcellular regulation of immune cell metabolism determines immune responses at the level of whole organism.

Avian influenza H5N1 clade 2.3.4.4b viruses caused a global panzootic and, unexpectedly, widespread outbreaks in dairy cattle, therefore representing a pandemic threat. To inform control strategies, it is critical to determine whether the potential to adapt to bovine cells is a general feature of H5N1 viruses, is specific to viruses of clade 2.3.4.4b, or narrowly restricted to some genotypes within this clade. Using a large panel of recombinant viruses representing >60 years of H5N1 history and other IAVs for comparison, we demonstrate replicative fitness in bovine cells is: (i) highly variable across 2.3.4.4b genotypes, (ii) limited in viruses predating the global expansion of this clade, (iii) determined by the internal gene cassette, and (iv) not restricted to udder epithelial cells. Mutations in the PB2 polymerase subunit emerge as key determinants of adaptation, although their phenotypic effects are context dependent. Bovine B3.13 and some avian genotypes exhibit enhanced modulation of bovine interferon-induced antiviral responses, determined by at least PB2, nucleoprotein, and the non-structural protein NS1. Our results highlight the polygenic nature of IAV host range, and reveal that the replication fitness in bovine cells, and likely their potential to adapt to cattle, varies greatly during the evolutionary trajectory of H5N1 viruses.

Modern spatial proteomic methods, such as multiplex immunofluorescence (mIF) imaging, offer a data-rich view of spatial biology in intact tissues. However, interpreting its complexity is a major bottleneck, limiting its potential for biological discovery and clinical translation. Current computational methods often rely solely on segmentation-based analysis that discards crucial morphological information and are limited to testing pre-defined hypotheses. Here, we introduce a self-supervised learning (SSL) framework, HyPERSTAC, which integrates SSL and segmentation-based analysis to enable hypothesis-agnostic, context-aware discovery of biomarkers directly from mIF images. Our approach extracts rich feature representations that capture holistic architectural patterns, which integrate cellular morphology, marker interactions, and microenvironmental context without human supervision. Applying HyPERSTAC to close to 400,000 mIF tissue images from two cancer cohorts totalling over 1,800 patients, we show that SSL-derived features complement existing spatial analyses. Using HyPERSTAC to integrate SSL-derived features with segmentation-based neighbourhood analysis or the established spatial frameworks TRACERx-PHLEX and MuSpAn increased prognostic concordance in held-out patients. HyPERSTAC improves survival prediction for four SSL training methods, and two state-of-the-art foundation models. The method autonomously identified previously unknown and potentially clinically actionable biological patterns. In lung adenocarcinoma, these include a Ki67-mediated immune evasion phenotype, a subcellular pattern of GLB1 expression which aligns with low-grade EGFR -driven tumours, and distinct modes of tumour-immune interaction in PD-L1 ⁺ patients. We also find a regulatory T-cell-mediated immuno-suppressive environment promoting tumour budding in colorectal carcinoma. Our work shows that SSL can complement segmentation-based analysis, establishing HyPERSTAC as a powerful, scalable, explainable, and unbiased platform to decode tissue ecosystems without pre-defined hypotheses. This transforms high-plex imaging from a hypothesis-testing tool into a hypothesis-generating engine that can accelerate the discovery of next-generation spatial biomarkers. We release two large image datasets from LATTICe-A , creating the largest publicly available multiplex immunofluorescence patient cohort, with two panels totalling almost 6,000 images from 1,000 patients and accompanying survival data. This creates a new benchmark for the community and enables further biomarker discovery from these data.

Background: Viral lower respiratory tract infections (LRTIs) can reduce the severity of subsequent LRTIs but have also been linked to respiratory allergy development and exacerbation. Here, we show that viral LRTI can imprint lung epithelial cells (LECs), leading to prolonged phenotypic and functional changes.Methods: Mice were infected via intranasal administration of respiratory syncytial virus (RSV). After 28 days, LECs were isolated using cold dispase digestion followed by magnetic‐activated cell sorting. Epigenetic changes were assessed using Cleavage Under Targets and Release Using Nuclease (CUT&RUN), while transcriptional changes were evaluated using NanoString and qPCR. Flow cytometry was employed to measure cell surface major histocompatibility complex (MHC) levels, antigen uptake and processing rates, and OT‐I cell proliferation after antigen presentation.Results: We identified epigenetic and transcriptomic changes in murine LECs 28 days after RSV infection, especially impacting genes associated with MHC. Lasting upregulation of MHC‐I and MHC‐II was further increased following in vivo LPS stimulation. Importantly, MHC upregulation was associated with increased antigen uptake and processing, as well as increased antigen presentation to T cells.Conclusions: Our data demonstrate that RSV can induce prolonged upregulation of antigen presentation by LECs, with the potential to facilitate local T cell responses to microbial antigens and allergens and to enhance immunity or in susceptible hosts respiratory allergy.

Understanding mechanisms that facilitate early events in metastatic seeding is key to developing therapeutic approaches to reduce metastasis. Here we identify uracil as a metastasis-associated metabolite in genetically engineered mouse models of cancer and in patients with metastatic breast cancer. Uracil is generated by the enzyme uridine phosphorylase-1 (UPP1), and we find that neutrophils are a significant source of UPP1 in metastatic cancer. Mammary tumours increase expression of adhesion molecules on the neutrophil surface, in a UPP1-dependent manner, leading to decreased neutrophil motility in the pre-metastatic lung. UPP1-expressing neutrophils suppress T-cell proliferation, and the UPP1 product uracil increases fibronectin deposition in the extracellular microenvironment. Knockout or inhibition of UPP1 in mice with mammary tumours increases T-cell numbers and reduces fibronectin content in the lung, and decreases the proportion of mice that develop lung metastasis. These data indicate that UPP1 influences neutrophil behaviour and extracellular matrix deposition in the lung, and suggest that circulating uracil could be a marker of metastasis, and that pharmacological inhibition of UPP1 could be a strategy to reduce recurrence.

Self-supervised representation learning methods often fail to learn subtle or complex features, which can be dominated by simpler patterns which are much easier to learn. This limitation is particularly problematic in applications to science and engineering, as complex features can be critical for discovery and analysis. To address this, we introduce Split Component Embedding Registration (SpliCER), a novel architecture which splits the image into sections and distils information from each section to guide the model to learn more subtle and complex features without compromising on simpler features. SpliCER is compatible with any self-supervised loss function and can be integrated into existing methods without modification. The primary contributions of this work are as follows: i) we demonstrate that existing self-supervised methods can learn shortcut solutions when simple and complex features are both present; ii) we introduce a novel self-supervised training method, SpliCER, to overcome the limitations of existing methods, and achieve significant downstream performance improvements; iii) we demonstrate the effectiveness of SpliCER in cutting-edge medical and geospatial imaging settings. SpliCER offers a powerful new tool for representation learning, enabling models to uncover complex features which could be overlooked by other methods.

T-cell receptor recognition of cognate peptide–MHC leads to the formation of signalling domains and the immunological synapse. Because of the close membrane apposition, there is rapid exclusion of CD45, and therefore LCK activation. Much less is known about whether spatial regulation of the intracellular face dictates LCK activity and TCR signal transduction. Moreover, as LCK is a driver in T acute lymphocytic leukaemia, it is important to understand its regulation. Here, we demonstrate a direct role of the ciliary protein UNC119 in trafficking LCK to the immunological synapse. Inhibiting UNC119 reduces localisation of LCK without impairing LCK phosphorylation and reduces T-cell receptor signal transduction. Although important for initial LCK reorganisation, activated CD8⁺T cells retained their ability to kill target tumour cells when UNC119 was inhibited. UNC119 was also needed to sustain proliferation in patient-derived T-ALL cells. UNC119 may therefore represent a novel therapeutic target in T acute lymphocytic leukaemia, which alters the subcellular localisation of LCK in T acute lymphocytic leukaemia cells but preserves the function of existing cytotoxic lymphocytes.

Loss of IFNγ-sensitivity by tumours is thought to be a mechanism enabling evasion, but recent studies suggest that IFNγ-resistant tumours can be sensitised for immunotherapy, yet the underlying mechanism remains unclear. Here, we show that IFNγ receptor-deficient B16-F10 mouse melanoma tumours are controlled as efficiently as WT tumours despite their lower MHC class I expression. Mechanistically, IFNγ receptor deletion in B16-F10 tumours increases IFNγ availability, triggering a remodelling of the immune landscape characterised by inflammatory monocyte infiltration and the generation of ‘mono-macs’. This altered myeloid compartment synergises with an increase in antigen-specific CD8⁺ T cells to promote anti-tumour immunity against IFNγ receptor-deficient tumours, with such an immune crosstalk observed around blood vessels. Importantly, analysis of transcriptomic datasets suggests that similar immune remodelling occurs in human tumours carrying mutations in the IFNγ pathway. Our work thus serves mechanistic insight for the crosstalk between tumour IFNγ resistance and anti-tumour immunity, and implicates this regulation for future cancer therapy.

As well as spreading through virions, influenza A viruses (IAVs) can evade antiviral drugs and neutralising antibodies by spreading directly from cell to cell. In cell culture this can occur by the induction of intercellular membrane connections known as tunnelling nanotube-like structures (TLSs), which are capable of trafficking the viral genome between cells. Here, we showed that TLSs are present at the site of IAV infections in vivo , and then used in vitro models to ask how IAVs induce their formation. We found that TLS induction cannot be induced by cytokine signalling from infected to uninfected cells, but requires IAV replication within cells. IAV replication can form filamentous virions with structural similarities to TLSs, but we found that TLS induction is independent of virion morphology. We therefore looked at the intracellular responses to infection. Using a pan-caspase inhibitor, we found that TLS induction by IAVs requires the onset of apoptosis. Our results, which suggest that IAVs control their ability to spread directly from cell to cell by driving infected cells into apoptosis, identifies a new way in which a virus can manipulate its host to evade antiviral immune responses. Author Summary: Influenza A viruses (IAVs) spread efficiently through the respiratory tract in the form of extracellular virus particles, but can be restricted by neutralising antibodies and antiviral drugs. IAVs can avoid this restriction by transporting viral genomes directly from one cell to the next. They can do this by inducing the formation of long, thin intercellular connections known as tunnelling nanotube-like structures, which are capable of trafficking viral genomes. In this study, we demonstrate for the first time that tunnelling-nanotube like structures form within IAV infected lungs. We then asked how IAVs induce these structures. We found that cell death pathways triggered by IAV replication induce the formation of tunnelling nanotube-like structures, thereby establishing routes of infection spread to other cells. In this way, the virus exploits the cell death response of its host to ensure that its infection can continue to spread even within the restrictive environment of the respiratory tract.