Biology subjects

Xu, H.

Publications and source records attributed to Xu, H..

At least 361 records · Page 20Linked to original sources

bioRxiv2021

How alginate lyase produces quasi-monodisperse oligosaccharides: a normal mode analysis-based docking and molecular dynamics simulation study

Polysaccharide degradation products are widely used in medicine, health food, textile and other industries. The preparation of monosaccharides by enzymatic degradation is a key technology in bio industrial production. Unfortunately, most of the known digested products are complex oligosaccharide mixtures, which limit their industrial processing and application. In this study, we explored a docking technique based on normal mode analysis to examine the possible cleavage mechanism of an alginate lyase (AlyB) from Birio Splendidus, which contains the catalytic domain of polysaccharide lyase family 7 (PL7) and a CBM32 sugar binding module, and was observed to produce trisaccharide products with quasi-monosaccharide distribution. We compared the molecular interactions of the enzyme with the natural alginates, the polyMG whose products has the quasi-monodisperse distribution of tri-saccharide and two synthetic polysaccharides, the polyM and polyG whose products has a wider distribution of oligosaccharides. Our calculations quantitatively show that there are a series of deterministic conformational changes in the catalytic pocket, which control the specificity of the substrate; at the same time, it determines the uniformity of the final product together with the spatial position of the key catalytic sites. The dynamic simulations revealed that CBM domain plays a key role in assisting the release of tri-saccharides. Our data highlights the important role of enzyme flexibility in determining product uniformity, which may provide new insight into design of enzymes for the production of high-value mono-distributed oligosaccharides.

biochemistry↗

bioRxiv2021

Defining the Sensitivity Landscape of 74,389 EGFR Variants to Tyrosine Kinase Inhibitors

BackgroundTyrosine kinase inhibitors (TKIs) therapy is a standard treatment for patients with advanced non-small-cell lung carcinoma (NSCLC) when activating epidermal growth factor receptor (EGFR) mutations are detected. However, except for the well-studied EGFR mutations, most EGFR mutations lack treatment regimens. MethodsWe constructed two EGFR variant libraries containing substitutions, deletions, or insertions using the saturation mutagenesis method. All the variants were located in the EGFR mutation hotspot (exons 18-21). The sensitivity of these variants to afatinib, erlotinib, gefitinib, icotinib, and osimertinib was systematically studied by determining their enrichment in massively parallel cytotoxicity assays using an endogenous EGFR-depleted cell line, PC9. ResultsA total of 3,914 and 70,475 variants were detected in the constructed EGFR Substitution-Deletion (Sub-Del) and exon 20 Insertion (Ins) libraries, accounting for 99.3% and 55.8% of the designed variants, respectively. Of the 3,914 Sub-Del variants, 813 were highly enriched in the reversible TKI (erlotinib, gefitinib, icotinib) cytotoxicity assays and 51 were enriched in the irreversible TKI (afatinib, osimertinib) cytotoxicity assays. For the 70,475 Ins variants, insertions at amino acid positions 770-774 were highly enriched in all the five TKI cytotoxicity assays. Moreover, the top 5% of the enriched insertion variants included a glycine or serine insertion at high frequency. ConclusionsWe present a comprehensive reference for the sensitivity of EGFR variants to five commonly used TKIs. The approach used here should be applicable to other genes and targeted drugs.

genetics↗

bioRxiv2021

Making Strides in Doctoral-Level Career Outcomes Reporting: Surveying the Landscape of Classification and Visualization Methodologies and Creating a Crosswalk Tool

The recent movement underscoring the importance of career taxonomies has helped usher in a new era of transparency in PhD career outcomes. The convergence of discipline-specific organizational movements, interdisciplinary collaborations, and federal initiatives have all helped to increase PhD career outcomes tracking and reporting. Transparent and publicly available PhD career outcomes are being used by institutions to attract top applicants, as prospective graduate students are factoring these in when deciding on the program and institution in which to enroll for their PhD studies. Given the increasing trend to track PhD career outcomes, the number of institutional efforts and supporting offices for these studies have increased, as has the variety of methods being used to classify and report/visualize outcomes. This report identifies and summarizes currently available PhD career taxonomy tools, resources, and visualization options to help catalyze and empower institutions to develop and publish their own PhD career outcomes. Similar fields between taxonomies were mapped to create a new crosswalk tool. This work serves as an empirical review of the career outcome tracking systems available and highlights organizations, consortia, and funding agencies that are impacting policy change toward greater transparency in PhD career outcomes reporting.

scientific communication and education↗

bioRxiv2021

APRILE: Exploring the Molecular Mechanisms of Drug Side Effects with Explainable Graph Neural Networks

The majority of people over the age of 65 take two or more medications. While many individual drug side effects are known, polypharmacy side effects due to novel drug combinations poses great risk. Here, we present APRILE: an explainable artificial intelligence (XAI) framework that uses graph neural networks to explore the molecular mechanisms underlying polypharmacy side effects. Given a list of side effects and the pairs of drugs causing them, APRILE identifies a set of proteins (drug targets or non-targets) and associated Gene Ontology (GO) terms as mechanistic explanations of associated side effects. Using APRILE, we generate such explanations for 843,318 (learned) and 93,966 (novel) side effect-drug pair events, spanning 861 side effects (472 diseases, 485 symptoms and 9 mental disorders) and 20 disease cate-gories. We show that our two new metrics--pharmacogenomic information utilization and protein-protein interaction information utilization--provide quantitative estimates of mechanism complexity. Explanations were significantly consistent with state of the art disease-gene associations for 232/239 (97%) side effects. Further, APRILE generated new insights into molecular mechanisms of four diverse categories of adverse drug reactions: infection, metabolic diseases, gastrointestinal diseases, and mental disorders, including paradoxical side effects. We demonstrate the viability of discovering polypharmacy side effect mechanisms by training an XAI framework on massive biomedical data. Consequently, it facilitates wider and more reliable use of AI in healthcare.

bioinformatics↗

bioRxiv2021

Transfer learning improves outcome predictions for ASD from gene expression in blood

BackgroundPredicting outcomes on human genetic studies is difficult because the number of variables (genes) is often much larger than the number of observations (human subject tissue samples). We investigated means for improving model performance on the types of under-constrained problems that are typical in human genetics, where the number of strongly correlated genes (features) may exceed 10,000, and the number of study participants (observations) may be limited to under 1,000. MethodsWe created train, validate and test datasets from 240 microarray observations from 127 subjects diagnosed with autism spectrum disorder (ASD) and 113 typically developing (TD) subjects. We trained a neural network model (a.k.a., the naive model) on 10,422 genes using the train dataset, composed of 70 ASD and 65 TD subjects, and we restricted the model to one, fully-connected hidden layer to minimize the number of trainable parameters, including a dropout layer to help prevent overfitting. We experimented with alternative network architectures and tuned the hyperparameters using the validate dataset, and performed a single, final evaluation using the holdout test dataset. Next, we trained a neural network model using the identical architecture and identical genes to predict tissue type in GTEx data. We transferred that learning by replacing the top layer of the GTEx model with a layer to predict ASD outcome and we retrained the new layer on the ASD dataset, again using the identical 10,422 genes. FindingsThe naive neural network model had AUROC=0.58 for the task of predicting ASD outcomes, which saw a statistically significant 7.8% improvement from transfer learning. InterpretationWe demonstrated that neural network learning could be transferred from models trained on large RNA-Seq gene expression to a model trained on a small, microarray gene expression dataset with clinical utility for mitigating over-training on small sample sizes. Incidentally, we built a highly accurate classifier of tissue type with which to perform the transfer learning. FundingThis work was supported in part by NIMH R01-MH110558 (E.C., N.E.L.) Author SummaryImage recognition and natural language processing have enjoyed great success in reusing the computational efforts and data sources to overcome the problem of over-training a neural network on a limited dataset. Other domains using deep learning, including genomics and clinical applications, have been slower to benefit from transfer learning. Here we demonstrate data preparation and modeling techniques that allow genomics researchers to take advantage of transfer learning in order to increase the utility of limited clinical datasets. We show that a non-pre-trained, naive model performance can be improved by 7.8% by transferring learning from a highly performant model trained on GTEx data to solve a similar problem.

bioinformatics↗

bioRxiv2021

Laminin-binding Integrins Regulate Angiogenesis by Distinct and Overlapping Mechanisms in Organotypic Cell Culture Models

Endothelial cells engage extracellular matrix and basement membrane components through integrin-mediated adhesion to promote angiogenesis. Our previous studies demonstrated that endothelial expression of laminin-411 and laminin-511 as well as 6 integrins is required for endothelial sprouting and tube formation in organotypic angiogenesis assays. These studies demonstrated that 6 integrins promote migration and regulate the expression of ANGPT2 and CXCR4 and that 6-dependent regulation of CXCR4 contributes to endothelial morphogenesis in our assays. However, these studies did not identify specific roles for the 6{beta}1, 6{beta}4, or 3{beta}1 laminin-binding integrins. Here, we employ RNAi technology to parse the contributions of these integrins. We demonstrate that 6{beta}4 promotes migration, sprouting, and tube formation, and also positively regulates the expression of ANGPT2, but does not promote CXCR4 expression, suggesting that 6{beta}1 functions in this regulation. Additionally, we show that 3{beta}1 regulates endothelial sprouting and tube formation, but is not required for migration in our assays or for the expression of ANGPT2 or CXCR4. Integrin 3{beta}1 promotes the expression of NRP1 and ID1 RNAs, both of which are known to promote angiogenesis. Taken together, our results indicate that laminin-binding integrins play distinct roles during endothelial morphogenesis and do not compensate for one another in organotypic culture. Summary StatementThe laminin-binding integrins 3{beta}1, 6{beta}1, and 6{beta}4 contribute to endothelial sprouting and tube formation in organotypic angiogenesis assays.

cell biology↗

bioRxiv2021

How does the brain represent the semantic content of an image?

Using deep neural networks (DNNs) as models to explore the biological brain is controversial, which is mainly due to the impenetrability of DNNs. Inspired by neural style transfer, we circumvented this problem by using deep features that were given a clear meaning--the representation of the semantic content of an image. Using encoding models and the representational similarity analysis, we quantitatively showed that the deep features which represented the semantic content of an image mainly predicted the activity of voxels in the early visual areas (V1, V2, and V3) and these features were essentially depictive but also propositional. This result is in line with the core viewpoint of the grounded cognition to some extent, which suggested that the representation of information in our brain is essentially depictive and can implement symbolic functions naturally.

neuroscience↗

bioRxiv2021

CRISPRa screen on a genetic risk locus shared by multiple autoimmune diseases identifies a dysfunctional enhancer that affects IRF8 expression through cooperative lncRNA and DNA methylation machinery

Dysregulated transcription factors represent a major class of drug targets that mediate the abnormal expression of many critical genes involved in SLE and other autoimmune diseases. Although strong evidence suggests that natural human genetic variation affects basal and inducible gene expression, it is still a considerable challenge to establish a biological link between GWAS-identified non-coding genetic risk variants and their regulated gene targets. Here, we combine genetic data, epigenomic data, and CRISPR activation (CRISPRa) assays to screen for functional variants regulating IRF8 expression. Using CRISPR-mediated deletion and 3D chromatin structure analysis, we demonstrate that the locus containing rs2280381 is a cell-type-specific distal enhancer for IRF8 that spatially interacts with the IRF8 promoter. Further, rs2280381 mediates IRF8 expression through enhancer RNA AC092723.1, which recruits TET1 to the IRF8 promoter to modulate IRF8 expression by affecting methylation levels. The alleles of rs2280381 modulate PU.1 binding and chromatin state to differentially regulate AC092723.1 and IRF8 expression. Our work illustrates a strategy to define the functional genetic variants modulating transcription factor gene expression levels and identifies the biologic mechanism by which autoimmune diseases risk genetic variants contribute to the pathogenesis of disease.

genetics↗

bioRxiv2021

Thin-Film Freeze-Drying of a Norovirus Vaccine Candidate

A bivalent Norovirus vaccine candidate has been developed that contains Norovirus strain GI.1 Norwalk-virus like particles (VLP) and strain GII.4 Consensus VLP adsorbed onto aluminum (oxy)hydroxide. In the present study, we tested the feasibility of converting the vaccine from a liquid suspension into dry powder by thin-film freeze-drying (TFFD). With the proper amount of trehalose and/or sucrose as cryoprotectant, TFFD can be applied to transform the Norovirus vaccine candidate into dry powders without causing antigen loss or particle aggregation, while maintaining the potency of the antigens within a specified acceptable range. In an accelerated stability study, the potency of the antigens was also maintained in the specified acceptable range after the dry powders were stored for eight weeks at 40{degrees}C, 75% relative humidity. The dry powder Norovirus vaccine offers the potential to eliminate the cold chain requirement for transport and/or storage of the vaccine.

pharmacology and toxicology↗

bioRxiv2021

Differential Bcd activation of two hunchback promoters emerges from unified kinetics of enhancer-promoter interaction

Many eukaryotic genes contain alternative promoters with distinct expression patterns. How these promoters are differentially regulated remains elusive. Here, we apply single-molecule imaging to quantify the transcriptional regulation of two alternative promoters (P1 and P2) of the Bicoid (Bcd) target gene hunchback in syncytial blastoderm Drosophila embryos. Contrary to the previous notion that Bcd only activates P2, we find that Bcd activates both promoters via the same two enhancers. P1 activation is less frequent and requires binding of more Bcd molecules than P2 activation. Using a theoretical model to relate promoter activity to enhancer states, we show that the two promoters follow common transcription kinetics driven by sequential Bcd binding at the two enhancers. Bcd binding at either enhancer primarily activates P2, while P1 activation relies more on Bcd binding at both enhancers. These results provide a quantitative framework for understanding the kinetic mechanisms of complex eukaryotic gene regulation.

biophysics↗

bioRxiv2021

The impact of mutational processes on structural genomic plasticity in cancer cells

Structural genome alterations are determinants of cancer ontogeny and therapeutic response. While bulk genome sequencing has enabled delineation of structural variation (SV) mutational processes which generate patterns of DNA damage, we have little understanding of how these processes lead to cell-to-cell variations which underlie selection and rates of accrual of different genomic lesions. We analysed 309 high grade serous ovarian and triple negative breast cancer genomes to determine their mutational processes, selecting 22 from which we sequenced >22,000 single cell whole genomes across a spectrum of mutational processes. We show that distinct patterns of cell-to-cell variation in aneuploidy, copy number alteration (CNA) and segment length occur in homologous recombination deficiency (HRD) and fold-back inversion (FBI) phenotypes. Widespread aneuploidy through induction of HRD through BRCA1 and BRCA2 inactivation was mirrored by continuous whole genome duplication in HRD tumours, contrasted with early ploidy fixation in FBI. FBI tumours exhibited copy number distributions skewed towards gains, widespread clone-specific variation in amplitude of high-level amplifications, often impacting oncogenes, and break-point variability consistent with progressive genomic diversification, which we termed serriform structural variation (SSV). SSVs were consistent with a CNA-based molecular clock reflecting a continual and distributed process across clones within tumours. These observations reveal previously obscured genome plasticity and evolutionary properties with implications for cancer evolution, therapeutic targeting and response.

cancer biology↗

bioRxiv2021

CNS antigen-specific control of early B-lineage development in the meninges

The random V(D)J recombination of immunoglobulins (Ig) loci often creates autoreactive B cell progenitors expressing self-recognized B-cell receptors (BCRs)1, which are eliminated or inactivated through an autoantigen-dependent central tolerance checkpoint to prevent autoimmune reactions2,3, a process thought to be restricted to the bone marrow (BM) in the adult mammals4. Here we report that early developing B cells are also present in the meninges of mice at all ages. Single cell RNA-sequencing (scRNA-seq) analysis revealed a consecutive trajectory of meningeal developing B cells in mice and non-human primates (NHPs). Parabiosis together with lineage tracing of hematopoietic stem cells (HSCs) showed that meningeal developing B cells are continuously replenished from the HSC-derived progenitors via a circulationindependent route. Importantly, autoreactive immature B cells which recognize myelin oligodendrocyte glycoprotein (MOG)5, a central nervous system (CNS)-specific antigen, are eliminated from the meninges but not BM. Furthermore, genetic deletion of MOG restored the self-reactive B cells in the meninges. Thus, we propose that meninges function as a unique reservoir for B cell development, allowing in situ negative selection of CNS-antigen-autoreactive B cells to ensure a local non-self-reactive immune repertoire.

immunology↗

bioRxiv2021

Differential Activity-Dependent Scaling of Synaptic Inhibition and Parvalbumin Interneuron Recruitment in Dentate Projection Neuron Subtypes

Strong inhibitory synaptic gating of dentate gyrus granule cells (GCs), attributed largely to fast-spiking parvalbumin interneurons (PV-INs), is essential to maintain sparse network activity needed for dentate dependent behaviors. However, the contribution of PV-INs to basal and input driven sustained synaptic inhibition in GCs and semilunar granule cells (SGCs), a sparse morphologically distinct dentate projection neuron subtype are currently unknown. In studies conducted in hippocampal slices from mice, we find that although basal inhibitory postsynaptic currents (IPSCs) are more frequent in SGCs and optical activation of PV-INs elicited IPSCs in both GCs and SGCs, optical suppression of PV-INs failed to reduce IPSC frequency in either cell type. Amplitude and kinetics of IPSCs evoked by perforant path activation were not different between GCs and SGCs. However, the robust increase in sustained polysynaptic IPSCs elicited by paired afferent stimulation was lower in SGCs than in simultaneously recorded GCs. Optical suppression of PV-IN selectively reduced sustained IPSCs in SGCs but not in GCs. These results demonstrate that PV-INs, while contributing minimally to basal synaptic inhibition in both GCs and SGCs in slices, mediate sustained feedback inhibition selectively in SGCs. The temporally selective blunting of activity-driven sustained inhibitory gating of SGCs could support their preferential and persistent recruitment during behavioral tasks. Significance StatementOur study identifies that feedback inhibitory regulation of dentate semilunar granule cells, a sparse and functionally distinct class of projection neurons, differs from that of the classical projection neurons, granule cells. Notably, we demonstrate relatively lower activity dependent increase in sustained feedback inhibitory synaptic inputs to semilunar granule cells when compared to granule cells which would facilitate their persistent activity and preferential recruitment as part of memory ensembles. Since dentate granule cell activity levels during memory processing are heavily shaped by basal and feedback inhibition, the fundamental differences in basal and evoked sustained inhibition between semilunar granule cells and granule cells characterized here provide a framework to reorganize current understanding of the dentate circuit processing.

neuroscience↗

bioRxiv2021

Interferon-armed RBD dimer enhances the immunogenicity of RBD for sterilizing immunity against SARS-CoV-2

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a global crisis, urgently necessitating the development of safe, efficacious, convenient-to-store, and low-cost vaccine options. A major challenge is that the receptor-binding domain (RBD)-only vaccine fails to trigger long-lasting protective immunity if used solely for vaccination. To enhance antigen processing and cross-presentation in draining lymph nodes (DLNs), we developed an interferon (IFN)-armed RBD dimerized by immunoglobulin fragment (I-R-F). I-R-F efficiently directs immunity against RBD to DLN. A low dose of I-R-F induces not only high titer long-lasting neutralizing antibodies but also comprehensive T cell responses than RBD, and even provides comprehensive protection in one dose without adjuvant. This study shows that the I-R-F vaccine provides rapid and complete protection throughout upper and lower respiratory tracts against high dose SARS-CoV-2 challenge in rhesus macaques. Due to its potency and safety, this engineered vaccine may become one of the next-generation vaccine candidates in the global race to defeat COVID-19.

immunology↗

bioRxiv2021

Protein phosphatase 1 regulatory inhibitor subunit 14C promotes triple-negative breast cancer progression via sustaining inactive glycogen synthase kinase 3 beta

The majority of clinical deaths among patients with triple-negative breast cancer (TNBC) are caused by uncontrolled cell proliferation and aggressive metastases, which is regulated by hyperactive glycogen synthase kinase 3 beta (GSK3{beta}); however, the underlying mechanisms remain largely unknown. In the present study, we found that protein phosphatase 1 regulatory inhibitor subunit 14C (PPP1R14C) was specifically upregulated in TNBC, compared with that in normal tissues and non-TNBC. High PPP1R14C expression correlated significantly with shorter 5-year relapse-free survival and overall survival in patients with TNBC. Overexpressing PPP1R14C promoted, while suppression of PPP1R14C decreased the cell proliferation and the aggressive phenotype of TNBC cells, in vitro and in vivo. Importantly, we revealed that PPP1R14C interacted with and inactivated type 1 Ser/Thr protein phosphatase (PP1) to sustain GSK3{beta} phosphorylation at S9, and induced the ubiquitylation and degradation of non-phosphorylated GSK3{beta} (S9A) via recruiting E3 ligase, TRIM25. Furthermore, treating with C2 ceramide (C2), which recovered kinase activity of GSK3{beta}, resulted in tumor growth inhibition in PPP1R14C-overexpressing TNBC cells. Our findings reveal a novel mechanism whereby PPP1R14C sustains inactive GSK3{beta}, which might lead to targeted therapy for TNBC.

cancer biology↗

bioRxiv2021

Retention Time Standardization and Registration (RTStaR): An algorithm that matches corresponding and identifies unique species in nanoliquid chromatography-nanoelectrospray ionization-mass spectrometry lipidomic datasets

Bioinformatic tools capable of registering, rapidly and reproducibly, large numbers of nanoliquid chromatography-nanoelectrospray ionization-tandem mass spectrometry (nLC-nESI-MS/MS) lipidomic datasets are lacking. We provide here a freely available Retention Time Standardization and Registration (RTStaR) algorithm that aligns nLC-nESI-MS/MS spectra within a single dataset and compares these aligned retention times across multiple datasets. This two-step calibration matches corresponding and identifies unique lipid species in different lipidomes from different matrices and organisms. RTStaR was developed using a population-based study of 1001 human serum samples composed of 71 distinct glycerophosphocholine metabolites comprising a total of 68,572 analytes. Platform and matrix independence were validated using different MS instruments, nLC methodologies, and mammalian lipidomes. The complete algorithm is packaged in two modular ExcelTM workbook templates for easy implementation. RTStaR is freely available from the India Taylor Lipidomics Research Platform http://www.neurolipidomics.ca/rtstar/rtstar.html. Technical support is provided through ldomic@uottawa.ca

bioinformatics↗

bioRxiv2021

Spatial analysis of tumor infiltrating lymphocytes based on deep learning using histopathology image to predict progression-free survival in colorectal cancer

PurposeThis study aimed to explore the prognostic impact of spatial distribution of tumor infiltrating lymphocytes (TILs) quantified by deep learning (DL) approaches based on digitalized whole slide images stained with hematoxylin and eosin in patients with colorectal cancer (CRC). MethodsThe prognostic impact of spatial distributions of TILs in patients with CRC was explored in the Yonsei cohort (n=180) and validated in the TCGA cohort (n=268). Concurrently, two experienced pathologists manually measured TILs at the most invasive margin as 0-3 by the Klintrup-Makinen (KM) grading method and compared to DL approaches. Interobserver agreement for TILs was measured using Cohens kappa coefficient. ResultsOn multivariate analysis of spatial TILs features derived by DL approaches and clinicopathological variables including tumor stage, Microsatellite instability, and KRAS mutations, TILs densities within 200 m of the invasive margin (f_im200) was remained as the most significant prognostic factor for progression-free survival (PFS) (HR 0.004 [95% CI, 0.0001-0.1502], p=.002) in the Yonsei cohort. On multivariate analysis using the TCGA dataset, f_im200 retained prognostic significance for PFS (HR 0.031, [95% CI 0.001-0.645], p=.024). Interobserver agreement of manual KM grading based on Cohens kappa coefficient was insignificant in the Yonsei ({kappa}=.109) and the TCGA ({kappa}=.121), respectively. The survival analysis based on KM grading showed statistically significant different PFS from the TCGA cohort, but not the Yonsei cohort. ConclusionsAutomatic quantification of TILs at the invasive margin based on DL approaches showed a prognostic utility to predict PFS, and could provide robust and reproducible TILs density measurement in patients with CRC. Data and Code AvailabilitySource code and data used for this study is available at the following link: https://github.com/hwanglab/TILs_Analysis

bioinformatics↗

bioRxiv2021

Bistable host-pathogen interaction explains varied infection outcomes

Death or survival are the two outcomes of an infection. The development of an infection depends on multiple interplaying factors, including the temporal dynamic of host nutrition, immune response and pathogen proliferation, but how these factors collectively determine the outcome of an infection is less understood. To understand the underlying principle of infection, here, we took a systems approach by developing a simple mathematical model to understand multi-factored host-pathogen interaction. We find that nonlinear interactions feedback between host and pathogens resulting in two distinct trajectories of disease progression. We show that disease progression, measured by the pathogen load, responds in a bistable manner, by either proliferating or eventually being eliminated. We identified key processes accounting for bistability, including hosts resource assimilation, immune response and intrinsic noise. In addition, resource availability, immunity, and pathogen load at the onset of infection are also critical in determining the bistability. We then highlight, by discussing in the light of previous theoretical and experimental work, how our framework provides testable theories for host-pathogen interaction inside the host.

ecology↗