Biology subjects

Guo, L.

Publications and source records attributed to Guo, L..

At least 91 records · Page 5Linked to original sources

bioRxiv2025

SpaFun: Discovering Domain-specific Spatial Expression Patterns and New Disease-Relevant Genes using Functional Principal Component Analysis

SpaFun is a novel, non-model-based method developed to address limitations in existing spatially variable gene (SVG) detection techniques, particularly for large-scale spatially resolved transcriptomics (SRT) datasets. These limitations include computational inefficiency, limited statistical power with increasing data size, and the inability to capture spatial heterogeneity and co-expression patterns among genes. Built on functional principal component analysis (fPCA), SpaFun identifies domain-representative genes (DRGs) with significantly better computational efficiency and greater statistical power while accounting for spatial heterogeneity and co-expression patterns among genes. We applied SpaFun to three SRT datasets and demonstrated that SpaFun outperformed state-of-the-art algorithms for identifying representative genes for tumor regions (e.g., DESeq, edgeR, and limma), as well as recently developed novel algorithms designed for spatial omics to identify the representative genes (e.g., SPARK and CSIDE). This highlights SpaFuns ability to accurately identify genes most representative of each spatial domain (e.g., tumor, immune, or stroma regions). By uncovering novel disease-relevant genes overlooked by existing algorithms, SpaFun could provide insights into new molecular mechanisms and propose innovative therapeutic strategies to improve patient outcomes. Key PointsO_LISpaFun is the first method dedicated to identifying DRGs, capturing spatially representative expression patterns within annotated tissue regions, setting it apart from all traditional SVG detection methods. C_LIO_LIIt leverages fPCA to model gene expression as a function of spatial location, avoiding reliance on predefined spatial distribution assumptions. C_LIO_LIThe non-model-based framework ensures compatibility with different SRT platforms and experimental designs, making it a scalable and widely applicable tool for SRT research. C_LI

genomics↗

bioRxiv2025

Automatically Defining Protein Words for Diverse Functional Predictions Based on Attention Analysis of a Protein Language Model

Understanding the relationship between protein sequence and function remains a longstanding challenge in bioinformatics, and to date the lions share of related tools parse proteins at the domain or motif levels. Here, we define "protein words" as an alternative to "motif" for studying proteins and functional prediction applications. We first developed an unsupervised tool we term Protein Wordwise, which parses analyte protein sequences into protein words by analyzing attention matrices from a protein language model (PLM) through a community detection algorithm. We then developed a supervised sequence-function prediction model called Word2Function, for mapping protein words to GO terms through feature importance analysis. We compared the prediction performance of our protein word-based toolkit with a motif-based method (PROSITE) for multiple protein function datasets. We also assembled a functionally diverse data resource we term PWNet to support evaluation of protein words for predicting functional residues across 10 tasks (e.g., diverse biomolecular binding, catalysis, and ion-channel activity). Our toolkit outperforms PROSITE in all the examined datasets and tasks. By abandoning domains and instead using attention matrices from a PLM for automatic, systematic, and annotation-agnostic parsing of proteins, our toolkit both outperforms currently available tools for functional annotations at the residue and whole-protein levels and suggests innovative forms of protein analysis well-suited to the post-AlphaFold era of biochemistry.

bioinformatics↗

bioRxiv2025

Advanced Manufacturing of Coil-Reinforced Multilayer Vascular Grafts to Optimize Biomechanical Performance

Small diameter vascular grafts require a complex balance of biomechanical properties to achieve target burst pressure, arterial compliance-matching, and kink resistance to prevent failure. Iterative design of our multilayer vascular was previously used to achieve high compliance while retaining the requisite burst pressure and suture retention strength for clinical use. To impart kink resistance, a custom 3D solution printer was used to add a polymeric coil to the electrospun polyurethane graft to support the graft during bending. The addition of this reinforcing coil increased kink resistance but reduced compliance. A matrix of grafts were fabricated and tested to establish key structure-property relationships between coil parameters (spacing, diameter, modulus) and biomechanical properties (compliance, kink radius). A successful graft design was identified with a compliance similar to saphenous vein grafts (4.1 {+/-} 0.4 %/mmHgx10-2) while maintaining comparable kink resistance to grafts used currently in the clinic. To explore graft combinations that could increase graft compliance to match arterial values while retaining this kink resistance, we utilized finite element (FE) models of compliance and kink radius that simulated experimental testing. The FE-predicted graft compliance agreed well with experimental values. Although the kink model over-predicted the experimental kink radius values, key trends between graft parameters and kink resistance were reproduced. As an initial proof-of-concept, the validated models were then utilized to parse through a targeted graft design space. Although this initial parameter range tested did not yield a graft that improved upon the previous balance of graft properties, this combination of advanced manufacturing and computational framework paves the way for future model-driven design to further optimize graft performance. TOC O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=97 SRC="FIGDIR/small/633374v1_ufig1.gif" ALT="Figure 1"> View larger version (45K): org.highwire.dtl.DTLVardef@faca56org.highwire.dtl.DTLVardef@f9dccforg.highwire.dtl.DTLVardef@fbb5corg.highwire.dtl.DTLVardef@d5c021_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2025

Unveiling the Evolution of Antimicrobial Peptides in Gut Microbes via Foundation Model-Powered Framework

Antimicrobial resistance poses a growing threat to public health, emphasizing the urgent need for novel therapeutic strategies. Antimicrobial peptides (AMPs), short peptide sequences with diverse mechanisms of action, offer a promising alternative due to their broad-spectrum activity against pathogens. Recent advances in protein language models (PLMs) have revolutionized protein structure prediction and functional annotation, highlighting their potential for AMP discovery and therapeutic development. In this context, we present AMP-SEMiner (Antimicrobial Peptide Structural Evolution Miner), an AI-driven framework designed to identify AMPs from metagenome-assembled genomes (MAGs). By integrating PLMs, structural clustering and evolutionary analysis into the framework, AMP-SEMiner can identify AMPs encoded by small open reading frames (smORFs) and encrypted peptides (EPs), significantly expanding the discovery space. Using this approach, we identified 1,670,600 AMP candidates from diverse habitats. Experimental validation of 29 candidates revealed antimicrobial activity in 18, with 13 surpassing antibiotics in effectiveness. Further analysis of AMPs from human gut microbiomes demonstrated both conserved and adaptive evolutionary strategies, ensuring their functional efficacy in the dynamic gut environment. These findings position AMP-SEMiner as a powerful tool for the discovery and characterization of novel AMPs, with significant potential to drive the development of new antimicrobial therapies.

microbiology↗

bioRxiv2025

Looking for the optimal pathway to stimulate hippocampus in human participants using transcranial ultrasound stimulation (TUS): a simulation study

At present, very limited methods can directly modulate the neural activities in deep brain structures in human participants. Low-intensity transcranial ultrasound stimulation (TUS), as an emerging and advanced modality of non-invasive brain stimulation (NIBS), has great potential for focally stimulating the subcortical structures that are related to sleep, emotion and the functions of motor of cognition. Due to the focality, the integration of various methods, including magnetic resonance imaging and possible pathways, is important to perform TUS intervention for precise targeting and dosing. Based on structural MRI scans, we constructed a simulation model of low-intensity TUS, with a particular focus on the optimal pathways of targeting hippocampus in human participants. Finally, we outlooked the future perspectives of TUS in clinical applications in neurological and psychiatric fields.

neuroscience↗

bioRxiv2025

Global expression profile of Enterohemorrhagic Escherichia coli O157:H7 in phagosome of murine macrophages

EHEC O157:H7 is responsible for Hemolytic Uremic Syndrome (HUS) outbreaks in humans and its virulence is associated with Shiga toxin (Stx). During the pathogenesis of HUS, EHEC O157:H7 is phagocytized by the intestinal host macrophage into the phagosome. According to our study, a low percentage of phagocytized bacteria survived in the murine macrophage phagosome. To understand the initials mechanisms involved in bacterial persistence, we performed a global profile of bacterial RNAseq analysis in the early phagosome of murine macrophage and in vitro assays with more extreme stress conditions. The defense strategy of an early murine phagosome consisted on producing DNA damage, membrane damage, acid pH and nitric oxide (NO) agents. These hostile conditions triggering the bacterial SOS response, lipid biosynthesis and NO detoxification enzymes in response to DNA damage, membrane damage and nitric oxide agents, respectively. In turn, to survive, the bacterium conserves energy by downregulating flagellar biosynthesis, T3SS and T6SS virulence mechanisms. In contrast, stx2a expression is upregulated. At the same time, it increases the ribosomal levels and amino acid synthesis to react more effectively against adverse conditions. Under more extreme stress conditions, EHEC O157:H7 expressed genes related to an acidic environment (pH 4.5) upregulating the acid stress response pathways AR2-5, whereas, upon high concentrations of hydrogen peroxide, it transcribed higher levels of genes of the OxyR pathway and subsequently hydrogen peroxide oxidative stress-related genes.

microbiology↗

bioRxiv2024

A Wearable Spiral Ultrasound Patch for Focused Ultrasound Peripheral Neuromodulation of Diabetic Neuropathic Pain

Diabetic foot disease, a severe complication of diabetes, frequently results in diabetic neuropathic pain, which diminishes quality of life and increases the risk of amputation. Current treatment options are limited in their effectiveness and can present adverse effects. We have demonstrated a first-of-its-kind wearable spiral ultrasound patch for focused ultrasound peripheral neuromodulation to alleviate diabetic neuropathic pain. This innovative, stretchable ultrasound device reliably delivers focused ultrasound to peripheral nerves. The devices ultrasound elements are arranged in a phyllotaxis spiral pattern, enhancing transmitted acoustic focus performance and minimizing grating and side lobes. Using a diabetic foot rat model, we confirmed peripheral neuropathy through behavioral assessments of mechanical allodynia and heat hyperalgesia, which showed significant pain relief after two weeks of ultrasound treatment. Additionally, nerve conduction velocity improved. Biological safety tests indicated no adverse effects on nerve tissues or cells. Our approach offers a non-invasive solution for managing diabetic neuropathic pain and provides critical insights into its underlying mechanisms, paving the way for more effective therapies.

bioengineering↗

bioRxiv2024

Instantaneous Frequency: A New Functional Biomarker for Dynamic Brain Causal Networks

This study introduces instantaneous frequency (IF) analysis as a novel method for characterizing dynamic brain causal networks from fMRI blood-oxygen-level-dependent (BOLD) signals. Effective connectivity, estimated using dynamic causal modeling (DCM), is analyzed to derive IF sequences, with the average IF across brain regions serving as a potential biomarker for global network oscillatory behavior. Analysis of data from the Alzheimers Disease Neuroimaging Initiative (ADNI), Open Access Series of Imaging Studies (OASIS), and Human Connectome Project (HCP) demonstrates the methods efficacy in distinguishing between clinical and demographic groups, such as cognitive decline stages, sex differences, and sleep quality levels. Statistical analyses reveal significant group differences in IF metrics, highlighting its potential as a sensitive indicator for early diagnosis and monitoring of neurodegenerative and cognitive conditions. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=100 SRC="FIGDIR/small/628965v1_ufig1.gif" ALT="Figure 1"> View larger version (35K): org.highwire.dtl.DTLVardef@10da58org.highwire.dtl.DTLVardef@6eee6borg.highwire.dtl.DTLVardef@e6e756org.highwire.dtl.DTLVardef@15dff2a_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsO_LIThe study introduces instantaneous frequency (IF) as a novel biomarker derived from dynamic brain effective connectivity, capturing temporal fluctuations in brain networks. C_LIO_LIThe proposed IF biomarker effectively differentiates between various clinical stages, such as Mild Cognitive Impairment (MCI) and Alzheimers Disease (AD), and demographic factors, including sex and sleep quality. C_LIO_LIThe robustness and clinical relevance of the IF biomarker are validated using three independent datasets: ADNI, OASIS, and HCP, demonstrating its potential in cognitive and neurological research. C_LI

neuroscience↗

bioRxiv2024

Structure and organization of full-length Epidermal Growth Factor Receptor in extracellular vesicles by cryo-electron tomography

We report here transport of the Epidermal Growth Factor Receptor (EGFR), Insulin Receptor, 7-pass transmembrane receptor Smoothened, and 13-pass Sodium-iodide symporter to extracellular vesicles (EVs) for structural and functional studies. Mass spectrometry confirmed the transported proteins as the most abundant in EV membranes, and the presence of many receptor-interacting proteins demonstrates the utility of EVs for characterizing membrane protein interactomes. Cryo-electron tomography of EGFR-containing EVs reveals that EGFR forms clusters in the presence of EGF with a [~]3 nm gap between the inner membrane and cytoplasmic density. EGFR extracellular regions do not form regular arrays, suggesting that clustering is mediated by the intracellular region. Subtomogram averaging of the EGFR extracellular region (ECR) yielded a 15 [A] map into which the crystal structure of the ligand-bound EGFR ECR dimer fits well. These findings refine our understanding of EGFR activation, clustering, and signaling, and they establish EVs as a versatile platform for structural and functional characterization of human membrane proteins in a native-like environment. Significance StatementAtomic or near-atomic resolution structural studies of proteins embedded in cell membranes have proven challenging. We show that transporting integral membrane proteins to cell-derived extracellular vesicles enables structural and functional studies of human membrane proteins in a native membrane environment. We have used this approach to visualize an active form of full-length Epidermal Growth Factor Receptor (EGFR) and show that it forms clusters in the membrane and projects its cytoplasmic signaling domains [~]3 nm away from the membrane surface. EGFR is essential for normal development, but abnormal EGFR activity is associated with several human cancers and is the target of many anticancer therapies. Our studies refine current models of how ligand binding to EGFR transmits signals across cell membranes.

biophysics↗

bioRxiv2024

Astrocyte-specific secretome profiling reveals its correlation with neurological disorders

Secreted proteins mediate intercellular communication throughout the lives of multicellular organisms. However, due to the lack of new technology for secreted protein capturing, the progress of "secretomics lags behind. Here, we report a two-step secretome enrichment method (tsSEM) combining unnatural amino acid labeling and click chemistry-based biorthogonal reaction, which enables in vitro secretome profiling in the presence of serum. Using this novel method, we systematically investigated the secretome of human iPSCs-derived astrocytes (iAst) in different disease models and identified a panel of astrocyte-secreted proteins that are responsible for its non-cell autonomous toxicity under disease conditions. Furthermore, we validated two astrocytes-derived novel neurotrophic proteins, FAM3C and KITLG, which we identified from disease models, and found that they could boost neurite outgrowth, protect neurons, and promote neural progenitor proliferation. Our study highlights the utility of secretome profiling of iAst and demonstrates its applications in disease study and target identification and validation in drug development.

neuroscience↗

bioRxiv2024

Substrate Stiffness Dictates Unique Doxorubicin-induced Senescence-associated Secretory Phenotypes and Transcriptomic Signatures in Human Pulmonary Fibroblasts

Cells are subjected to dynamic mechanical environments which impart forces and induce cellular responses. In age-related conditions like pulmonary fibrosis, there is both an increase in tissue stiffness and an accumulation of senescent cells. While senescent cells produce a senescence-associated secretory phenotype (SASP), the impact of physical stimuli on both cellular senescence and the SASP is not well understood. Here, we show that mechanical tension, modeled using cell culture substrate rigidity, influences senescent cell markers like SA-{beta}-gal and secretory phenotypes. Comparing human primary pulmonary fibroblasts (IMR-90) cultured on physiological (2 kPa), fibrotic (50 kPa), and plastic (approximately 3 GPa) substrates, followed by senescence induction using doxorubicin, we identified unique high-stiffness-driven secretory protein profiles using mass spectrometry and transcriptomic signatures, both showing an enrichment in collagen proteins. Consistently, clusters of p21+ cells are seen in fibrotic regions of bleomycin induced pulmonary fibrosis in mice. Computational meta-analysis of single-cell RNA sequencing datasets from human interstitial lung disease confirmed these stiffness SASP genes are highly expressed in disease fibroblasts and strongly correlate with mechanotransduction and senescence-related pathways. Thus, mechanical forces shape cell senescence and their secretory phenotypes.

cell biology↗

bioRxiv2024

Elovl1 inhibition reduced very long chain fatty acids in a mouse model of adrenoleukodystrophy

Adrenoleukodystrophy (ALD) is a rare neurometabolic disease caused by mutations in the ABCD1 gene, which encodes for the peroxisomal very long chain fatty acid (VLCFAs) transporter. It is a debilitating disorder, which has a spectrum of clinical presentations. The most severe form is a rapidly progressing demyelinating disease called cerebral ALD or CALD. Patients with cALD have a life expectancy of 2-4 years after onset and symptoms often manifest in childhood. The other forms are adrenomyeloneuropathy or AMN, which is a slower progressing degeneration of the spinal cord, and adrenal insufficiency (Addison disease). Since the accumulation of VLCFAs are a common factor in all ALD pathologies, we identified therapeutic approach that could correct this metabolic defect. We developed a substrate reduction therapy (SRT) for ALD in the form of an inhibitor of the lipid elongase principally responsible for the generation of VLCFAs, Elovl1. This small molecule was able to successfully reduce the accumulation of VLCFA in the brain and spinal cord of ABCD1-/y mice. We used single nuclei RNA seq to identify the pathways altered in the ABCD1-/y mouse and corrected with Elovl1 inhibition. Though many lipid metabolism genes and pathways were indeed corrected, treatment with the Elovl1 inhibitor unexpectedly led to profound transcriptional changes beyond correction of pathways altered by loss of ABCD1. These data suggest that Elovl1 inhibition may have broader consequences in ABCD1-/y mice than correction of lipid homeostasis.

pharmacology and toxicology↗

bioRxiv2024

Alternative splicing modulates chromatin interactome and phase separation of the RIF1 C-terminal domain

How RIF1 (RAP1 interacting factor) fulfills its diverse roles in DNA double-strand break (DSB) repair, DNA replication, and nuclear organization remains elusive. Here we show that alternative splicing (AS) of a cassette exon (Ex32) encoding a Ser/Lys-rich (S/K) cassette in the RIF1 C-terminal domain (CTD) gives rise to RIF1-Long (RIF1-L) and RIF1-Short (RIF1-S) isoforms with different functional characteristics. We demonstrate that RIF1-Ex32 splice-in is mediated by an exonic splicing enhancer that is recognized by the splicing factor SRSF1 and antagonized by splicing inhibitors SRSF3 and SRSF7. Exposure to DNA damage inhibited Ex32 splice-in, potentiated the association of SRSF3 and SRSF7 with RIF1 pre-mRNA, and caused an increase in RIF1-S protein expression, which was also observed across a diverse set of primary cancers. Isoform-specific proteomic analyses revealed RIF1-L preferentially associated with mediator of DNA damage checkpoint 1 (MDC1) and sustained MDC1 focus formation to a greater extent than RIF1-S. We further show that the S/K cassette stabilized a novel phase separation activity of the RIF1 CTD and enhanced RIF1-L chromatin retention, which was reversed by CDK1-dependent phosphorylation of the RIF1 CTD in response to G2 DNA damage checkpoint inhibition. These combined findings suggest DNA damage-dependent RIF1 AS contributes to RIF1 functional diversification in genome protection.

molecular biology↗

bioRxiv2024

A potently neutralizing and protective human antibody targeting antigenic site V on RSV and hMPV fusion glycoprotein

Human respiratory syncytial virus (RSV) and human metapneumovirus (hMPV) are frequent drivers of morbidity and mortality in susceptible populations, most often infantile, older adults, and immunocompromised. The primary target of neutralizing antibodies is the fusion (F) glycoprotein on the surface of the RSV and hMPV virion. As a result of the structural conservation between RSV and hMPV F, three antigenic regions are known to induce cross-neutralizing responses: sites III, IV, and V. Leveraging LIBRA-seq, we identify five RSV/hMPV cross-reactive human antibodies. One antibody, 5-1, potently neutralizes all tested viruses from the major subgroups of RSV and hMPV and provides protection against RSV and hMPV in a mouse challenge model. Structural analysis reveals that 5-1 utilizes an uncommon genetic signature to bind an epitope that spans sites O, II and V, defining a new mode of antibody cross-reactivity between RSV and hMPV F. These findings highlight the molecular and structural elements influencing RSV and hMPV cross-reactivity as well as the potential of antibody 5-1 for translational development.

immunology↗

bioRxiv2024

PAM-flexible adenine base editing rescues hearing loss in a humanized MPZL2 mouse model harboring an East Asian founder mutation

Hearing loss is one of the most prevalent sensory disorders, but no commercial biological treatments are currently available. Here, we identified an East Asia-specific founder mutation, the homozygous c.220C>T mutation in MPZL2, that contributes to a significant proportion of hereditary deafness cases in our cohort study. We found that the disease-causing mutation could be targetable by adenine base editors (ABEs) that enable A{middle dot}T-to-G{middle dot}C base corrections without DNA double-strand breaks. To demonstrate this, we developed a humanized mouse model (hMPZL2Q74X/Q74X) that recapitulates human MPZL2 deafness and leads to progressive hearing loss. A PAM-flexible ABE variant with reduced bystander and off-target effects (ABE8eWQ-SpRY:sgRNA3) was packaged in dual adeno-associated viruses (AAVs) and injected into the inner ear of hMPZL2Q74X/Q74X mice and effectively corrected the mutation. This treatment significantly restored hearing function, improved inner ear structural integrity, and reversed altered gene expression. Base editing may hold therapeutic potential for hereditary deafness, including most cases of MPZL2 deafness.

genetics↗

bioRxiv2024

Ten-Fold Expansion MALDI Mass Spectrometry Imaging of Tissues and Cells at 500 nm Resolution

Achieving high spatial resolution in matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) is crucial for detailed molecular mapping in biological tissues and cells. However, conventional MALDI-MSI platforms are typically limited to spatial resolutions of 5-20 m, restricting their ability to visualize fine subcellular structures. Here, we present a novel ten-fold expansion MALDI-MSI (10X ExMSI) method that attains 500 nm spatial resolution without the need for specialized optics or customized instrumentation. By physically expanding biological samples using a swellable hydrogel matrix, our method maintains high retention and broadens the detection range of lipids, ensuring comprehensive lipid profiling. We demonstrated the effectiveness of 10X ExMSI by visualizing intricate subcellular structures within mouse brain tissues, including individual nuclei, astrocytes, Purkinje cells, and, notably, dendritic arborizations-features previously unresolvable using mass spectrometry imaging. Applying 10X ExMSI to single-cell analysis of cultured cells revealed detailed spatial distributions of lipids at the subcellular to organelle level. Notably, the method is fully compatible with standard commercial MALDI-MSI platforms, enabling widespread adoption without significant additional investment. The 10X ExMSI offers a transformative tool for high-resolution molecular imaging, opening new avenues for molecular analysis in diverse biological and biomedical fields, including neuroscience, pathology, and cellular biology.

biochemistry↗

bioRxiv2024

Generative machine learning of ADAR substrates for precise and efficient RNA editing

Adenosine Deaminase Acting on RNA (ADAR) converts adenosine to inosine within certain double-stranded RNA structures. However, ADARs promiscuous editing and poorly understood specificity hinder therapeutic applications. We present an integrated approach combining high-throughput screening (HTS) with generative deep learning to rapidly engineer efficient and specific guide RNAs (gRNAs) to direct ADARs activity to any target. Our HTS quantified ADAR-mediated editing across millions of unique gRNA sequences and structures, identifying key determinants of editing outcomes. We leveraged these data to develop DeepREAD (Deep learning for RNA Editing by ADAR Design), a diffusion-based model that elucidates complex design rules to generate novel gRNAs outperforming existing design heuristics. DeepREADs gRNAs achieve highly efficient and specific editing, including challenging multi-site edits. We demonstrate DeepREADs therapeutic potential by designing gRNAs targeting the MECP2R168X mutation associated with Rett syndrome, achieving both allelic specificity and species cross-reactivity. This approach significantly accelerates the development of ADAR-based RNA therapeutics for diverse genetic diseases.

bioengineering↗

bioRxiv2024

Functional consequences of fast-spiking interneurons in striatum

The striatum features a distinct network characterized by a high degree of shared feedforward inhibition (FFI) from a mere 1% of fast-spiking interneurons (FSI). We investigate the potential roles of this extensively shared FFI in striatal function beyond inducing synchrony. Our findings reveal that FSIs increase the acrosstrial variability of striatal responses to cortical stimuli and, combined with recurrent inhibition, lead to a correlation attractor of striatal activities, i.e., weakly correlated inputs result in more correlated responses and vice versa. Thus, we uncover a mechanism by which input correlation can be bidirectionally modulated, which is possible only because of high sharing of FSI inputs. We posit that the emergence of a correlation attractor leads to non-zero correlation level and variable rate trajectories of striatal responses across trials, hence beneficial for exploration in learning. However, given their role in across-trial variability, we argue that FSIs should be disengaged from the MSNs during performance where stability across trials is required. Significance StatementStriatum is a network of inhibitory neurons. Fast spiking interneurons constitute about 1% of the striatal neural population and provide feedforward inhibition (FSI). Here, we unravel two novel ways in which FSIs may shape striatal function. Given the recurrent inhibition, it is assumed that striatum can only de-correlate inputs. We show that high sharing of FSI also renders the striatum an ability to correlate inputs. Thus, recurrent and shared FSI create a correlation attractor. Besides, we show that shared FSIs give rise to high across-trial variability. Therefore, we argue that FSIs are more crucial in learning as they provide the neural basis of exploration, but they may impair learned behavior due to high across-trial variability.

neuroscience↗