Biology subjects

Lowe, S.

Publications and source records attributed to Lowe, S..

17 recordsLinked to original sources

bioRxiv2026

Beyond Segmentation Accuracy: Evaluating Biological Fidelity in Spatial Pathology

Multiplex immunofluorescence (MxIF) resolves cellular identity, state, and spatial organization within tumor tissue, enabling higher-order representations of cancer ecosystems including immune infiltration, multicellular neighborhoods, and tissue niches. Yet these analyses depend on a deceptively simple first step: accurately identifying individual cells. Segmentation errors can alter marker quantification, cellular phenotypes, cell-cell relationships, and ultimately the spatial organization inferred from an image. Consequently, improvements in conventional segmentation benchmarks matter biologically only if they propagate to more faithful representations of the tumor microenvironment. We test this relationship directly using an expert-curated benchmark of 31,924 expert-reviewed cells on three slides under a leave-slide-out protocol. We compare a domain-specific self-supervised encoder, adapted to multiplex tissue imaging, with supervised, microscopy-specific, prompted generalist, and general-purpose vision models, and pass every method's predicted cells through an identical downstream pipeline. We evaluate fidelity across four levels: cell segmentation, phenotype assignment, spatial neighborhood structure, and tissue organization, comprising niches and tumor-immune interactions. The resulting hierarchy reveals that these levels do not improve in lockstep. Cellular phenotype assignment is comparatively robust to segmentation differences, whereas missed cells increasingly disrupt neighborhood structure and tumor-immune interactions. As a stringent spatial-interaction test, tumor-CD8+ contact recovery is poorly predicted by conventional segmentation quality (Panoptic Quality: {rho} = 0.18), while errors specifically at tumor-immune interfaces are strongly associated with contact fidelity ({rho} = -0.80). Methods with nearly identical segmentation accuracy can therefore produce substantially different biological representations. These results argue that cell segmentation for spatial pathology should be evaluated not only by boundary and instance accuracy, but also by preservation of the biological relationships that downstream analyses are intended to measure.

bioinformatics↗

bioRxiv2026

Discovering Latent Scientific Concepts through Discrete Representation Learning

Scientific imaging is undergoing a fundamental transformation. Advances in multiplex immunofluorescence, computational pathology, and electron microscopy now enable observation of biological and physical systems at unprecedented spatial resolution. Rather than simply classifying known objects, these data offer the opportunity to discover previously unknown organizational principles. However, scientific image datasets are typically highly curated, information-rich, and limited in size, making reliable representation learning particularly challenging. We show that standard continuous self-supervised adaptation can become unstable in this regime, allowing teacher representations to drift and degrading downstream transfer. We introduce Kodiak, a discrete representation learning architecture that discovers latent scientific concepts through Sinkhorn-balanced codebook assignments. Instead of learning unconstrained continuous embeddings, Kodiak constructs a reusable vocabulary of discrete concepts that provides stable supervision for both masked-patch prediction and cross-view representation learning within a single-stage framework. Across five scientific imaging domains, including multiplex immunofluorescence for pancreatic cancer pathology and scanning electron microscopy, Kodiak consistently improves representation quality under both continued foundation-model adaptation and training from scratch. Beyond improved classification and transfer,the learned concepts are interpretable, reusable, and produce spatially coherent concept maps that support applications such as cell phenotyping, segmentation-free patch phenotyping, spatial niche discovery, tissue organization, and materials characterization. We view Kodiak as the representation-learning engine of our foundation model for spatial biology. More broadly, our results suggest that discrete representation learning provides a principled foundation for discovering reusable latent scientific concepts that can accelerate scientific interpretation across imaging domains.

cancer biology↗

bioRxiv2026

Systemic iron sequestration restricts the intratumoral CD8+ T cell landscape in pancreas cancer

Systemic iron sequestration occurs frequently in cancer due to inflammation-driven expression of the iron-regulatory hormone hepcidin. The impact of systemic iron availability on tumoral immunity is unclear. Here, we show that elevated serum hepcidin is associated with reduced survival and decreased intratumoral CD8+ T cells in patients with pancreas cancer. While hepcidin is not induced in murine tumor models, administration of a hepcidin mimetic phenocopies the T-cell-depleted tumor microenvironment seen in patients. Mechanistically, chronic antigen-driven mitochondrial dysfunction disrupts iron metabolism and selectively depletes high avidity CD8+ T cells during iron restriction. These findings establish a direct link between hepcidin-mediated iron sequestration and tumoral immunity and nominate systemic iron dysregulation as a therapeutic target to enhance anti-tumoral CD8+ T cell responses.

immunology↗

bioRxiv2026

GATA4 loss promotes mutant Kras-driven pancreatic ductal adenocarcinoma in the absence of canonical precursor lesions

ABSTRACTGATA6 and GATA4 play key roles in pancreatic development and are essential to maintain the classical transcriptional program in pancreatic ductal adenocarcinoma (PDAC). Using genetic mouse models we show that, in contrast to GATA6, GATA4 is dispensable for the maintenance of acinar homeostasis in the adult pancreas. Deletion of Gata4 in mice expressing mutant Kras in the embryonic pancreas (KG4C) leads to PDAC development in the absence of tissue remodeling, pancreatic intraepithelial neoplasia (PanIN), or other canonical precursor lesions present in Gata4-proficient (KC) mice. Similar observations were made when Gata4 was selectively inactivated in adult, Kras-mutant, acinar cells. We identify Pale Acinar Lesions (PALes) as a previously unrecognized pancreatic lesion, distinct from acino-ductal metaplasia (ADM) and PanINs, present in KC and KG4C mice but not in wild type mice. PALes display weak expression of acinar and ductal markers and lack mucins; they have lower proliferation rates than PanINs. RNA-seq and ChIP-seq reveal that GATA4 and GATA6 partially share genomic binding sites and transcriptomic effects, but they exert opposing influences on mutant Kras-induced, haematopoietic cell-dependent, transcriptional inflammatory programs. Adenoviral-mediated pancreatic expression of IL17 restored the formation of ductal lesions in KG4C mice but failed to rescue PanIN development. Our data indicate that GATA4 functions through the coordinated action of multiple inflammatory factors that are required for ADM/PanIN formation but are dispensable for PDAC development. Collectively, these findings challenge current paradigms of PDAC initiation and progression.

cancer biology↗

bioRxiv2025

Functional interrogation uncovers a critical role for a high-plasticity cell state in lung adenocarcinoma

Plasticity--the ability of cells to undergo phenotypic transitions--drives cancer progression and therapy resistance1-3. To date, strategies targeting cancer plasticity have not advanced to the clinic due to a lack of fundamental understanding of the underlying mechanisms. Recent studies have suggested that plasticity in solid tumors is concentrated in a minority subset of cancer cells4-6, yet functional studies interrogating this high plasticity cell state (HPCS) in situ are lacking. Here, we developed mouse models enabling detection, longitudinal lineage tracing, and ablation of the HPCS in autochthonous lung tumors in vivo. Using lineage tracing, we uncover the HPCS cells are dedifferentiated but possess high capacity for cell state transitions, giving rise to both early neoplastic (differentiated) and advanced lung cancer cell states in situ. Longitudinal lineage tracing using secreted luciferases reveals HPCS-derived cells harbor high capacity for growth when compared to bulk cancer cells or another defined cancer cell state with features of differentiated lung epithelium. Suicide gene-mediated ablation of the HPCS in early neoplasias abrogates tumor progression. Ablating HPCS cells in established tumors by suicide gene or HPCS-directed CAR T cells robustly reduces tumor burden, whereas ablation of a differentiated lung cancer cell state had no effect. We further demonstrate that the HPCS gives rise to therapy-resistant cell states, whereas ablation of the HPCS abrogates resistance to chemotherapy and oncoprotein-targeted therapy. Interestingly, an HPCS-like state is ubiquitous in regenerating epithelia and in carcinomas of multiple other tissues, revealing a convergence of plasticity programs. Our work establishes the HPCS as a critical hub enabling reciprocal transitions between cancer cell states, including acquisition of states adapted to cancer therapies. Targeting the HPCS in lung cancer and in other carcinomas may suppress cancer progression and eradicate treatment resistance.

cancer biology↗

bioRxiv2025

A data-driven model of macrophage polarization states reveals an IFN macrophage signature in active Crohn's disease

Macrophages are key innate immune cells responsible for initiating and coordinating immune responses. A major determinant of macrophage function is their tissue-context-dependent polarization state, regulated by the cytokines present in the tissue microenvironment. Yet, the capability of characterizing macrophage polarization states in clinical studies remains limited. Here, we used a defined set of cytokines to polarize human PBMC derived macrophages and determine their transcriptional signatures and stimulus responsiveness. The resultant atlas of transcriptional signatures for human macrophage polarization states was applied to a dataset of intestinal biopsies from Crohns disease patients and healthy controls. Our analysis identified a dominant population of IFN{gamma}-polarized macrophages in areas of active Crohns intestinal inflammation and a loss of wound healing IL-4-, IL-10- and IL-13-polarized macrophages. This study demonstrates that in vitro datasets of macrophages in defined conditions can be leveraged to interpret the functionality of cells transcriptional profiled in clinical studies.

immunology↗

bioRxiv2025

An immunocompetent murine model of virus-elicited liver fibrosis and hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is the third deadliest cancer worldwide. Over 75% of HCC cases are associated with chronic viral infections. Mechanistic studies and preclinical therapeutic development for virus-associated HCC have been limited by a paucity of small animal models of chronic hepatotropic virus infection that faithfully recapitulate human disease. Here we demonstrate the induction of chronic hepatitis, progressive liver fibrosis, and HCC in immunocompetent laboratory mice upon chronic viral infection with Norway rat hepacivirus (NrHV) - a virus closely related to hepatitis C virus (HCV). NrHV-elicited tumors resemble HCV-associated tumors and liver transcriptome analyses reveal numerous similarities between chronic NrHV and HCV. These findings establish an experimentally tractable, physiologically relevant, and immunocompetent mouse model of virus-elicited progressive liver fibrosis and oncogenesis.

pathology↗

bioRxiv2024

Epg5 links proteotoxic stress due to defective autophagic clearance and epileptogenesis in Drosophila and Vici Syndrome patients

Epilepsy is a common neurological condition that arises from dysfunctional neuronal circuit control due to either acquired or innate disorders. Autophagy is an essential neuronal housekeeping mechanism, which causes severe proteotoxic stress when impaired. Autophagy impairment has been associated to epileptogenesis through a variety of molecular mechanisms. Vici Syndrome (VS) is the paradigmatic congenital autophagy disorder in humans due to recessive variants in the ectopic P-granules autophagy tethering factor 5 (EPG5) gene that is crucial for autophagosome-lysosome fusion and ultimately for effective autophagic clearance. VS is characterized by a wide range of neurodevelopmental, neurodegenerative, and neurological features, including epilepsy. Here, we used Drosophila melanogaster to study the importance of epg5 in development, ageing, and seizures. Our data indicate that proteotoxic stress due to impaired autophagic clearance and seizure-like behaviors correlate and are commonly regulated, suggesting that seizures occur as a direct consequence of proteotoxic stress and age-dependent neurodegenerative progression in epg5 Drosophila mutants, in the absence of evident neurodevelopmental abnormalities. We provide complementary evidence from EPG5-mutated patients demonstrating an epilepsy phenotype consistent with Drosophila predictions and propose autophagy stimulating diets as a feasible approach to control EPG5-related pharmacoresistant seizures.

neuroscience↗

bioRxiv2023

Reliability of dynamic causal modelling of resting state magnetoencephalography

This study assesses the reliability of resting-state dynamic causal modelling (DCM) of magneto-electroencephalography under conductance-based canonical microcircuit models, in terms of both posterior parameter estimates and model evidence. We use resting state magneto-electroencephalography (MEG) data from two sessions, acquired two weeks apart, from a cohort with high between-subject variance arising from Alzheimers disease. Our focus is not on the effect of disease, but on the predictive validity of the methods implicit in their reliability, which is crucial for future studies of disease progression and drug intervention. To assess the predictive validity of first-level DCMs, we compare model evidence associated with the covariance among subject-specific free energies (i.e., the quality of the models) with vs. without interclass correlations. We then used parametric empirical Bayes (PEB) to investigate the predictive validity of DCM parameters at the between subject level. Specifically, we examined the evidence for or against parameter differences (i) within-subject, within-session, between-epochs; (ii) within-subject between-session and (iii) within-site between-subjects, accommodating the conditional dependency among parameter estimates. We show that for data acquired close in time, and under similar circumstances, more than 95% of inferred DCM parameters are unlikely to differ, speaking to mutual predictability over sessions. Using PEB, we show a reciprocal relationship between a conventional definition of reliability and the conditional dependency among inferred model parameters. Our analyses confirm the predictive validity and reliability of the conductance-based DCMs for resting-state neurophysiological data. In this respect, the implicit generative modelling is suitable for interventional and longitudinal studies of neurological and psychiatric disorders. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=147 SRC="FIGDIR/small/562379v1_ufig1.gif" ALT="Figure 1"> View larger version (41K): org.highwire.dtl.DTLVardef@135c964org.highwire.dtl.DTLVardef@38e4a7org.highwire.dtl.DTLVardef@12bc358org.highwire.dtl.DTLVardef@25f110_HPS_FORMAT_FIGEXP M_FIG C_FIG

biophysics↗

bioRxiv2022

Chromosome 9p21.3 Coordinates Cell Intrinsic and Extrinsic Tumor Suppression

Somatic chromosomal deletions are prevalent in cancer, yet their functional contributions remain ill-defined. Among the most prominent of these events are deletions of chromosome 9p21.3, which disable a cell intrinsic barrier to tumorigenesis by eliminating the CDKN2A/B tumor suppressor genes. However, half of 9p21.3 deletions encompass a cluster of 16 type I interferons (IFNs) whose co-deletions have not been functionally characterized. To dissect how 9p21.3 and other genomic deletions impact cancer, we developed MACHETE (Molecular Alteration of Chromosomes with Engineered Tandem Elements), a genome engineering strategy that enables flexible modeling of megabase-sized deletions. Generation of 9p21.3-syntenic deletions in a mouse model of pancreatic cancer revealed that concomitant loss of Cdkn2a/b and the IFN cluster led to immune evasion and metastasis compared to Cdkn2a/b-only deletions. Mechanistically, IFN co-deletion disrupted type I IFN signaling, altered antigen-presenting cells, and facilitated escape from CD8+ T cell surveillance in a cell extrinsic manner requiring loss of interferon epsilon (Ifne). Our results establish co-deletions of the IFN cluster as a pervasive route to tumor immune evasion and metastasis, revealing how deletions can disable physically linked cell intrinsic and extrinsic tumor suppression. Our study establishes a framework to dissect the functions of genomic deletions in cancer and beyond.

cancer biology↗

bioRxiv2021

Netrins and receptors control Drosophila optic lobe organization and transmedullary neuron axon targeting

During development, integration of temporal patterning and spatial patterning of neural progenitors as well as Notch-dependent binary fate choice between sister neurons contribute to generation of neural diversity. How these upstream neural fate specification programs regulate downstream effector genes to control axon targeting and neuropil assembly remains less well-understood. Here we show that Notch-dependent binary fate choice in Drosophila medulla neurons regulates the expression of Netrin, and that Netrin pathway controls axon guidance of transmedullary (Tm) neurons and contributes to the organization of optic lobe neuropils. Netrins are enriched in the lobula where Tm axons target, and the attractive receptor Frazzled is expressed broadly in medulla neurons, while the repulsive receptor Unc-5 is excluded from Tm neurons and this is necessary for their correct targeting to the lobula. Frazzled is required collectively in a group of early-born Tm neurons to establish the inner optic chiasm (IOC) through which Tm axons target lobula. In addition, Frazzled acts in the layer-specific targeting step of Tm3 and Tm4 cell-autonomously, and is also required for the formation of the lobula branch of TmY3. Moreover, we show that the diffusibility of Netrins is necessary for Netrin enrichment in the lobula, the IOC formation and layer-specific targeting of Tm3 and Tm4. Netrin enrichment in the lobula is promoted by Frazzled expressed in Tm neurons, while Unc-5 appears to have an opposite role in Netrin distribution. Furthermore, we show that Netrin B is expressed in the Notch-on hemilineage of medulla neurons including most Tm and TmY neurons that target lobula, and loss of Su(H) abolished NetB expression in the medulla. Without medulla-originated NetB, Tm axons from late-born medulla columns cannot join the IOC. Therefore, the Notch-dependent binary fate choice regulates the assembly of the optic lobe neuropils by controlling the expression of Netrin.

developmental biology↗

bioRxiv2021

MYC controls metastatic heterogeneity in pancreatic cancer

The degree of metastatic disease varies widely amongst cancer patients and impacts clinical outcomes. However, the biological and functional differences that drive the extent of metastasis are poorly understood. We analyzed primary tumors and paired metastases using a multi-fluorescent lineage-labeled mouse model of pancreatic ductal adenocarcinoma (PDAC) - a tumor type where most patients present with metastases. Genomic and transcriptomic analysis revealed an association between metastatic burden and gene amplification or transcriptional upregulation of MYC and its downstream targets. Functional experiments showed that MYC promotes metastasis by recruiting tumor associated macrophages (TAMs), leading to greater bloodstream intravasation. Consistent with these findings, metastatic progression in human PDAC was associated with activation of MYC signaling pathways and enrichment for MYC amplifications specifically in metastatic patients. Collectively, these results implicate MYC activity as a major determinant of metastatic burden in advanced PDAC.

cancer biology↗

bioRxiv2020

Evolutionary Dynamics of Non-Coding Regions in Pancreatic Ductal Adenocarcinoma

While the non-coding genome appears to play a role, the dynamic nature of noncoding alterations with respect to clonal progression of solid tumors remains unexplored. To address this gap in knowledge we performed multiregional whole genome sequencing and clonal analysis to elucidate the evolutionary dynamics of non-coding regions in pancreatic cancer relative to those of the coding genome. We find that the mutational burden of noncoding DNA is higher than coding DNA. However, when noncoding DNA was segregated into enhancer and non-enhancer regions, enhancers were more similar to coding DNA. Mutational signatures of noncoding and coding DNA further revealed the similar mutational spectra of enhancers to coding DNA whereas the mutational spectra of non-enhancer, noncoding DNA had an entirely different pattern. These findings shed light on the role of noncoding DNA in pancreatic cancer.

cancer biology↗

bioRxiv2020

A knock-in Drosophila model supports a conserved link between potassium channelopathy and involuntary movement

BackgroundGenetic and in vitro studies have linked a heterozygous gain-of-function mutation (D434G) in the hSlo1 BK (Big potassium) channel to paroxysmal dyskinesia. However, support for this linkage from in vivo models has been lacking. ObjectivesWe aimed to re-create the equivalent mutation to hSlo1 D434G in the fruit fly, Drosophila, and examine how this mutation altered movement and action potential waveforms. MethodsWe generated a knock-in Drosophila model of hSlo1 D434G. We used video-tracking and infra-red beam-break systems to test whether locomotion was altered in this model, and patch-clamp electrophysiology to determine how the mutation affected action potential waveforms. ResultsWe identified profound motor dysfunction and sporadic leg twitches, as well as a reduced width and an enhancement of the afterhyperpolarization phase of action potentials, in the model background. ConclusionOur results support a conserved relationship between enhanced BK channel function and disrupted motor control across distantly related species.

neuroscience↗

bioRxiv2020

Altered synaptic plasticity and central pattern generator dysfunction in a Drosophila model of PNKD3 paroxysmal dyskinesia

BackgroundParoxysmal non-kinesigenic dyskinesia type-3 (PNKD3) has been linked to gain-of-function (GOF) mutations in the hSlo1 BK potassium channel, in particular a dominant mutation (D434G) that enhances Ca2+-sensitivity. However, while BK channels play well-known roles in regulating neurotransmitter release, it is unclear whether the D434G mutation alters neurotransmission and synaptic plasticity in vivo. Furthermore, the subtypes of movement-regulating circuits impacted by this mutation are unknown. ObjectivesWe aimed to use a larval Drosophila model of PNKD3 (sloE366G/+) to examine how BK channel GOF in dyskinesia alters synaptic properties and motor circuit function. MethodsWe used video-tracking to test for movement defects in sloE366G/+ larvae, and sharp-electrode recordings to assess the fidelity of Ca2+-dependent neurotransmitter release and short-term plasticity at the neuromuscular junction. We then combined sharp-electrode recording with ex vivo Ca2+-imaging to investigate the functionality of the central pattern generator (CPG) driving foraging behavior in sloE366G/+ larvae. ResultsWe show that the PNKD3 mutation leads to Ca2+-dependent alterations in synaptic release and paired-pulse facilitation. Furthermore, we identify robust alterations in locomotor behaviors in sloE366G/+ larvae which were mirrored by dysfunction of the upstream, movement-generating CPG in the larval ventral nerve cord. ConclusionOur results demonstrate that a BK channel GOF mutation can alter neurotransmitter release and short-term synaptic plasticity, and result in CPG dysfunction, in Drosophila larvae. These data add to a growing body of work linking paroxysmal dyskinesias to aberrant neuronal excitability and synaptic plasticity in pre-motor circuits.

neuroscience↗

bioRxiv2018

Neuronal overexpression of DYRK1A/minibrain alters motor decline, neurodegeneration and synaptic plasticity in Drosophila

Down syndrome (DS) is characterised by abnormal cognitive and motor development, and later in life by progressive Alzheimers disease (AD)-like dementia, neuropathology, declining motor function and shorter life expectancy. It is caused by trisomy of chromosome 21 (Hsa21), but how individual Hsa21 genes contribute to various aspects of the disorder is incompletely understood. Previous work has demonstrated a role for triplication of the Hsa21 gene DYRK1A in cognitive and motor deficits, as well as in altered neurogenesis and neurofibrillary degeneration in the DS brain, but its contribution to other DS phenotypes is unclear. Here we demonstrate that overexpression of minibrain (mnb), the Drosophila ortholog of DYRK1A, in the Drosophila nervous system accelerated age-dependent decline in motor performance and shortened lifespan. Overexpression of mnb in the eye was neurotoxic and overexpression in ellipsoid body neurons in the brain caused age-dependent neurodegeneration. At the larval neuromuscular junction, an established model for mammalian central glutamatergic synapses, neuronal mnb overexpression enhanced spontaneous vesicular transmitter release. It also slowed recovery from short-term depression of evoked transmitter release induced by high-frequency nerve stimulation and increased the number of boutons in one of the two glutamatergic motor neurons innervating the muscle. These results provide further insight into the roles of DYRK1A triplication in abnormal aging and synaptic dysfunction in DS.\n\nAuthor summaryDown syndrome (DS) is caused by three copies of chromosome 21 instead of the usual two. It is characterised by cognitive and motor deficits, which worsen with age resulting in Alzheimers disease (AD). Which genes on chromosome 21 cause these phenotypes is incompletely understood. Here we demonstrate that neuronal overexpression of minibrain, the Drosophila ortholog of the chromosome 21 gene DYRK1A, causes age-dependent degeneration of brain neurons, accelerates age-dependent decline in motor performance and shortens lifespan. It also modifies presynaptic structure, enhances spontaneous transmitter release and slows recovery from short-term depression of synaptic transmission at a model glutamatergic synapse. These findings give insight into the role of DYRK1A overexpression in aberrant aging and altered information processing in DS and AD.

neuroscience↗

bioRxiv2017

Mutations In Membrin/GOSR2 Reveal Stringent Secretory Pathway Demands Of Dendritic Growth And Synaptic Integrity

Mutations in the Golgi SNARE protein Membrin (encoded by the GOSR2 gene) cause progressive myoclonus epilepsy (PME). Membrin is a ubiquitously important protein mediating ER-to-Golgi membrane fusion, and hence it is unclear how these mutations result in a disorder restricted to the nervous system. Here we use a multi-layered strategy to elucidate the consequences of Membrin mutations from protein to neuron. We show that the pathogenic mutations cause partial reductions in SNARE-mediated membrane fusion. Importantly, these alterations were sufficient to profoundly impair dendritic growth in novel Drosophila models of GOSR2-PME. We also observed axonal trafficking abnormalities in this model, as well as synaptic malformations, trans-synaptic instability and hyperactive synaptic transmission. Our study highlights how dendritic growth is vulnerable even to subtle secretory pathway deficits, uncovers a previously uncharacterized role for Membrin in synaptic function, and provides a comprehensive explanatory basis for genotype-phenotype relationships in GOSR2-PME.

neuroscience↗