Exploring the heterogeneous transcriptional response of the CNS to systemic LPS and Poly(I:C)

the CNS poses a significant challenge to the study of cell type and stimulus dependent responses of neural cells during acute inflammation. Here we utilized single nuclei RNA sequencing (snRNAseq), serum proteome profiling and primary cell culture methods to systematically compare the acute response of the mammalian brain to the bacterial PAMP lipopolysaccharide (LPS) and the viral PAMP polyinosinic:polycytidylic acid (Poly(I:C)), at single cell resolution. Our study unveiled convergent transcriptional cytokine and cellular stress responses in brain vascular and ependymal cells and a downregulation of several key mediators of directed blood brain barrier (BBB) transport. In contrast the neuronal response to PAMPs was limited in acute neuroinflammation. Moreover, our study highlighted the dominant role of IFN signalling upon Poly(I:C) challenge, particularly in cells of the oligodendrocyte lineage. Collectively our study unveils heterogeneous, shared and distinct cell type and stimulus dependent acute responses of the CNS to bacterial and viral PAMP challenges. Our findings highlight inflammation induced dysregulations of BBB-transporter gene expression, suggesting potential translational implications on drug pharmacokinetics variability during acute neuroinflammation. The pronounced dependency of oligodendrocytes on IFN stimulation during viral PAMP challenges, emphasizes their limited molecular viral response repertoire.

Single nucleus RNA sequencing (snRNAseq) Pathogen-associated molecular pattern (PAMP) Interferon signalling the CNS poses a significant challenge to the study of cell type and stimulus dependent responses of neural cells during acute inflammation.
Here we utilized single nuclei RNA sequencing (snRNAseq), serum proteome profiling and primary cell culture methods to systematically compare the acute response of the mammalian brain to the bacterial PAMP lipopolysaccharide (LPS) and the viral PAMP polyinosinic:polycytidylic acid (Poly(I:C)), at single cell resolution.
Our study unveiled convergent transcriptional cytokine and cellular stress responses in brain vascular and ependymal cells and a downregulation of several key mediators of directed blood brain barrier (BBB) transport.In contrast the neuronal response to PAMPs was limited in acute neuroinflammation.Moreover, our study highlighted the dominant role of IFN signalling upon Poly(I:C) challenge, particularly in cells of the oligodendrocyte lineage.
Collectively our study unveils heterogeneous, shared and distinct cell type and stimulus dependent acute responses of the CNS to bacterial and viral PAMP challenges.Our findings highlight inflammation induced dysregulations of BBB-transporter gene expression, suggesting potential translational implications on drug pharmacokinetics variability during acute neuroinflammation.The pronounced dependency of oligodendrocytes on IFN stimulation during viral PAMP challenges, emphasizes their limited molecular viral response repertoire.

Introduction
Upon pathogen contact in the periphery, recognition of PAMPs evokes a rapid innate immune response, characterized by the release of a plethora of inflammatory mediators (Li and Wu, 2021).In the brain this immune response triggers complex cellular interactions, culminating in various functional outcomes, such as altered synaptic plasticity (Golia et al., 2019), as well as behavioral alterations, such as sickness behaviour, fatigue (Foster et al., 2021) and cognitive impairments (Bossù et al., 2012).Moreover, mounting evidence implicates neuroinflammation in neurodegeneration (Batista et al., 2019), psychiatric disorders (Miller and Raison, 2016), as well as dysregulated neurodevelopment and altered maternal and offspring behaviour in MIA models (Ronovsky et al., 2017;Kalish et al., 2021).Previous work on the brain's response to peripheral immune challenges extensively utilized the PAMPs LPS and Poly(I:C) to mimic bacterial and viral immune challenges, respectively (Batista et al., 2019;Arsenault et al., 2014;Cunningham et al., 2007).Upon peripheral LPS or Poly(I:C) exposure, a sickness behaviour phenotype becomes evident within minutes to hours.The corresponding transcriptional perturbations of CNS cells peak between 2 and 24 h and resolve within days depending on the cell type (Duan et al., 2018;Hasel et al., 2021;Shemer et al., 2020).Thus, the acute to subacute time interval has emerged as a popular window of opportunity to study the molecular mechanisms of the brain's response to PAMPs.
The functional effects of LPS are primarily initiated by binding the PRR TLR4, while the effects of Poly(I:C) are primarily triggered by binding to TLR3, RIG-I and MDA5 (Li and Wu, 2021;Fitzgerald and Kagan, 2020).
Previous work documented divergent effects of different PAMPs on neural network function, peripheral and CNS cytokine profiles and behavioral outcomes in various rodent models (Foster et al., 2021;Arsenault et al., 2014;Hopwood et al., 2009;Borsody and Weiss, 2004;Bao et al., 2022).Likewise, in vitro studies have shown that different PAMPs evoke different, cell-type specific phenotypical changes in cultured microglia (He et al., 2021), brain endothelial cells (Johnson et al., 2018), astrocytes (Bsibsi et al., 2006) and oligodendrocytes (Bsibsi et al., 2012).Previous evidence indicated that multiple effects of PAMPs, for example on cell viability, depend on secondary cellular interactions between different neural cell populations, rather than on direct PRRresponses (Steelman and Li, 2011;Li et al., 2008).The cellular heterogeneity of the CNS poses a significant obstacle in untangling cell-type specific responses and cellular interactions in health and diseases in vivo, a challenge which is increasingly addressed through single-cell resolution transcriptomics technologies (Piwecka et al., 2023).
The response of microglia and other immune cells to peripheral TLRligand challenges were among the first to be studied at single cell resolution (Shemer et al., 2020;Sousa et al., 2018), while further studies expanded this approach to other neural cell populations (Duan et al., 2018;Hasel et al., 2021;Allen et al., 2023;Crowell et al., 2020).However, all of these studies used peripheral LPS administration to model neuroinflammation.While previous work characterized the single cell brain transcriptome of offspring in a model of Poly(I:C) induced MIA (Kalish et al., 2021), the acute transcriptional response of the adult mammalian CNS to Poly(I:C) is hitherto uncharacterized at single cell resolution.Moreover, there is currently no study available which directly contrasts the effects of peripheral LPS and Poly(I:C) exposure on acute to subacute transcriptional changes in the CNS in vivo at the single cell level.
Here we systematically analysed cell type and stimulus dependent acute transcriptional perturbations induced by LPS and Poly(I:C) in the mammalian CNS, leveraging snRNAseq, as well as auxiliary serum proteome profiling and primary cell culture methods.Our data unveiled substantially overlapping peripheral and brain vasculature cytokine responses, shared cellular stress responses and perturbations of BBBtransporter gene expressional profiles in vascular cells of the brain.Furthermore, we dissected key differences in the brain's response to LPS and Poly(I:C), most notably regarding the dominant role of IFN signalling upon Poly(I:C) treatment, in particular in cells of the oligodendrocyte lineage.

Ethics statement
All experimental procedures were reviewed and approved by the local ethics committee of the Medical University of Vienna,) and the Austrian Federal Ministry of Science and Research (GZ.: BMBWF 2021-0.644.864;GZ.: BMBWF 2022-0.351.596).All procedures were conducted in accordance with national regulations, EU Directive 2010/63/EU for animal experiments and are reported according to ARRIVE 2.0 guidelines.

Animal husbandry
Twelve week old, female Sprague Dawley Rats (Crl:CD(SD)) were obtained from Charles River (Charles River Laboratories, Sulzfeld, Germany).All animals were housed at the Department of Biomedical D. Bormann et al.Research at the Medical University of Vienna, in standard Makrolon type IV cages, with aspen wood bedding material and gnawing wood, in groups of two to three, with access to food and water ad libitum.Rats were housed at standard conditions (controlled temperature: 21 • C, humidity: 50%) with an automatically controlled 12 h light/dark cycle, with light periods starting at 7 am.

Induction of peripheral inflammation and sampling of biological material
The general in vivo study design is depicted in Fig. 1A.LPS from Escherichia coli O111:B4 (cat#: L2630, batch: 0000126448, Sigma Aldrich, St. Louis, MO, USA), was resuspended in sterile saline and aliquoted to stock solutions of 2 mg/ml.Low molecular weight Poly(I:C) (cat#: tlrl-picw, Lot: 5936-43-02, InvivoGen, San Diego, CA, USA) was resuspended in sterile 0.9% saline and aliquoted to stock solutions of 20 mg/ml.Stocks were stored at − 20 • C until use.The same batches were used for all in vivo and in vitro experiments.
Animals were weighed before treatment, thereafter 1 ml LPS or Poly (I:C) solution was applied intra peritoneally (i.p.), at dosages of 5 mg/kg bodyweight and 12 mg/kg bodyweight, respectively.Control animals received 1 ml of a 0,9% sterile saline solution i.p. 7 h post injection sickness behaviour phenotype was documented by placing animals individually in empty standard Makrolon type IV cages.Cages were wiped with ethanol and ddH 2 O in between placing of animals.The frequency of rearing, climbing, and grooming bouts within 5 min was counted and animals were weighed immediately before terminal anaesthesia, which was induced by i.p. application of 100 mg/kg Ketamine (Ketasol-100, Livisto, Senden-Bösensell, Germany) and 40 mg/ kg Xylazine (Rompun 20 mg/ml, Bayer, Leverkusen, Germany).Mice were decapitated and trunk blood was collected.After coagulation blood tubes were centrifuged at 3000g, for 10 min, at 4 • C on a swing bucket centrifuge (Allegra X-12R, Beckman Coulter, Brea, CA, USA), serum was collected, aliquoted and stored at − 20 • C until further analyses.Whole brains were swiftly harvested, without removal of meninges and further dissected into standardized coronal sections using an adult rat brain slicer matrix (BSRAS003-1, with 3 mm coronal section intervals, Zivic Instruments, Pittsburgh, PA, USA).For nuclei purification, we used two coronal sections per animal (coronal sections 3 and 4, counting from the rostral edge), as shown in Fig. 1A to assure comparability of the sampled brain regions.The coronal sections were removed from the brain matrix after cutting, without any further dissection and immediately snap frozen in liquid nitrogen and stored at − 80 • C until nuclei purification.

Cytokine array and enzyme-linked immunosorbent assay based determination of serum cytokine levels
To assess serum cytokine signatures, samples from four animals per treatment group were pooled and subjected to the Proteome Profiler Rat XL Cytokine Array (cat#: ARY030, R&D Systems, Minneapolis, MN, USA), according to manufacturer's instructions.Signals were developed on a Gel Doc XR+ device (Bio-Rad Laboratories Inc., Hercules, CA, USA), followed by calculation of dot densities for each cytokine duplicate, utilizing the volume tool implicated in ImageLab 6.0.1 (Bio-Rad Laboratories Inc.), adjusting for background signal, as indicated by the manufacturer.Mean background adjusted dot volumes for each cytokine were presented as fold changes of treatment groups (LPS and Poly (I:C)) relative to saline controls, in GraphPad Prism v.8.01 (GraphPad Software, San Diego, CA, USA) generated bar plots.Only Cytokines with ≥1.5-fold change regulation relative to saline controls were reported.Overlapping and distinctly regulated cytokines are reported as a venn diagram, constructed using the R package eulerr v7.0.0 (Larsson, 2022).Serum Interferon beta (IFN-β) levels were measured using Enzyme-Linked Immunosorbent Assay (ELISA) (cat#: NBP3-06753, Novus Biologicals, R&D Systems), as per manufacturer's instructions.

Single nucleus processing and library preparation
All samples were processed using the Chromium™ Next GEM Single Cell 5' Kit v2 (PN-1000263, 10 × Genomics, Pleasanton, CA, USA), following manufacturer's instructions (CG000331 Rev. D, 10 × Genomics).In brief, nuclei were loaded onto Chromium™ Next GEM Chip K (PN-1000286,10 × Genomics) aiming at a recovery of 10-12 × 10 D. Bormann et al. (caption on next page) D. Bormann et al. nuclei per lane, overloading the lanes by 20% following previous recommendations (Maitra et al., 2021).After Gel Beads-in-Emulsion (GEMs) generation, GEM reverse transcription (GEM-RT) and clean up, the resulting cDNA was amplified for 14 cycles, followed by transcriptome library construction as per the 10 × 5′ v2 protocol.For cleanup procedures SPRIselect Reagent Kit (cat#: B23318, Beckman Coulter) beads were used as per manufactures instructions.Library quality was examined using a DNA screen tape D5000 on a TapeStation 4150 (Agilent Technologies, Santa Clara, CA, USA), cDNA quantification was performed using a Qubit 1xdsDNA HS assay kit (cat#: Q33231) on a QuBit 4.0 fluorometer (Invitrogen, ThermoFisher Scientific).Libraries with unique indices were pooled in equimolar ratio before sequencing.

Sequencing, pre-processing and quality control
Paired end, dual indexed (read length 50 bp) sequencing was performed on a NovaSeq 6000 (Illumina, San Diego, CA, USA), processing all samples on the same flow cell.To obtain raw gene counts, raw reads were demultiplexed and aligned to the most current rattus norvegicus reference genome mRatBN7.2, using the Cellranger v.7.0.0 pipeline.Intronic reads were retained in the count matrix to account for unspliced nuclear transcripts, as per developer's recommendations.
All further computational snRNAseq analyses were performed using R and R Studio (R version 4.2.2,The R Foundation, Vienna, Austria), within the environment of the Seurat package v.4.3.0 (Hao et al., 2021), as per developer's vignettes, unless otherwise stated.
Exploratory data analysis and quality control was conducted according to the standard Seurat workflow.Briefly, nuclei with <500 UMI counts, <200 expressed genes and > 5% percent mitochondrial reads were removed from downstream analysis.All mitochondrial genes and all genes with <3 UMI counts per feature were removed from UMI count matrices before further processing.To estimate and remove multiplets we utilized the DoubletFinder v2.0.3 package (McGinnis et al., 2019), as per developer's instructions.

Data integration
Normalization and variance stabilization was performed for all datasets individually, using SCTransform, with v2 regularization, while passing the percentage of mitochondrial reads "percent.mt"to the "vars.to.regress" argument.Thereafter, datasets were integrated utilizing a reciprocal PCA (RPCA) based approach.Briefly, the top 3000 highly variable genes were selected using the "SelectIntegrationFeatures" function, followed by preparation for integration using the "Pre-pSCTIntegration" function.The "RunPCA" function was used for dimensionality reduction and integration anchors were established using the "FindIntegrationAnchors" function.The first 30 significant principal components were used for RPCA reduction, with the "k.anchor" argument set to 15, followed by the execution of the "Integra-teData" function.The resultant single, integrated, batch-corrected expression matrix was used for all further downstream analyses.

Clustering and differentially expressed gene (DEG) analyses
Unbiased clustering analyses followed the standard Seurat workflow.Briefly, following integration, principal component analysis (PCA) was applied for primary dimensionality reduction, using the "RunPCA" function with default parameters.UMAP (Uniform Manifold Approximation and Projection) dimensionality reduction and Louvain clustering was performed, using the "RunUMAP" "FindNeighbors" and "FindClusters" functions, using the first 28 principal components (PCs) and a clustering resolution of 0.5.
Differentially expressed gene (DEG) calculations for the identification of cluster markers and differences in gene expression between experimental groups, were performed using the MAST statistical framework (Finak et al., 2015) within the Seurat "FindMarkers" and "FindAllMarkers" functions.DEGs were determined on log-normalized RNA-counts and only genes expressed in a minimum of 10% of nuclei in one of the tested groups were considered for DEG analyses.For between group comparisons only DEGs with a regulation of log2foldchange ≥ 0.58 (approximately 1.5 fold) and Bonferroni-adjusted pvalues <0.05 were considered.All non-subset, raw MAST outputs of cluster marker and between group comparison DEG calculations are reported as excel files (Supplementary data files 1-4).Upon initial exploratory cluster analysis, we identified 2 clusters co-expressing pan neuronal and markers of myelinating and mature oligodendrocytes (N = Frequency of rearing, climbing attempts and grooming bouts observed within a time interval of 5 min, 7 h after i.p. application of treatments.One-way ANOVA indicated a statistically significant main effect of treatment on the frequency of rearing (F (2, 9) = 58.86),p < 0.0001), climbing attempts (F (2, 9) = 13.26,p = 0.0021) and grooming bouts (F (2, 9) = 8.357, p = 0.0089).Tukey post hoc comparison tests indicated a significant reduction of rearing frequencies in LPS (q = 13.95,df = 9, p < 0.0001) and Poly(I:C) (q = 12.51, df = 9, p < 0.0001) treated animals, as compared to saline treated controls.Likewise, climbing frequencies were significantly reduced in both treatment groups (LPS vs Saline: q = 6.882, df = 9, p < 0.0023; Poly(I:C) vs Saline: q = 5.506, df = 9, p < 0.0092), as were grooming bouts (LPS vs Saline: q = 5.612, df = 9, p < 0.0082; Poly(I:C) vs Saline: q = 4.009, df = 9, p < 0.0468).Bodyweight (BW) in grams of N = 4 animals per treatment group is reported before treatment (baseline) and 7 h after i.p. injection of saline, LPS, or Poly(I: C).Two-way repeated measures ANOVA revealed a statistically significant main effect of time-point (F (1, 9) = 122.0,p < 0.0001), but not treatment group (F (2, 9) = 1.467, p = 0.2810), with a statistically significant time-point x treatment group interaction (F (2, 9) = 16.60,p = 0.0010).Bonferroni's post hoc tests indicated a significant decrease in bodyweight 7 h after LPS (t = 8.439, df = 9, p < 0.0001) and Poly(I:C) (t = 9.011, df = 9, p < 0.0001), but not saline (t = 1.685, df = 9, p = 0.3790) treatment.Data is presented as box plots, depicting medians, 25th to 75th percentiles as hinges and minimal and maximal values as One-way ANOVA, followed by Tukey post hoc pair-wise comparisons, revealed significant differences between the Poly(I:C) and saline (q = 6.63, df = 9, p = 0.0029), as well as LPS conditions (q = 4.97, df = 9, p = 0.0164).The difference between LPS and Saline group has not reached the threshold of statistical significance (q = 1.663, df = 9, p = 0.4954).Data is presented as box plots, depicting medians, 25th to 75th percentiles as hinges and

Gene set enrichment analysis
For gene set enrichment analyses, DEGs with a log2 fold change ≥1 were put into Enrichr (Kuleshov et al., 2016), querying the "Reactome 2022" and "GO Biological Process 2021" gene set databases.As the majority of databases are tailored to human genes, rat gene names were first converted to their respective human gene orthologs, using the gprofiler2 gorth tool (Kolberg et al., 2020), in analogy to previous scRNAseq studies analyzing rat tissue (Fletcher et al., 2023).Enrichr outputs were subset to enriched terms with a Benjamini-Hochberg method adjusted p value cut off of <0.05 and ranked by their respective "Combined Score", which combines p-values from Fisher's exact test and the respective term's deviation to its expected rank (Kuleshov et al., 2016).Enriched terms are presented as dotplots, depicting Combined Score and gene ratios of the respective enriched term.

Cell-cell communication inference analysis
Potential cell-cell communication (CCC) events between cell-types were inferred using the LIgand-receptor ANalysis frAmework (LIANA) v.0.1.12,following developer's vignettes (Türei et al., 2021;Dimitrov et al., 2022).We first used the "generate_homologs" function to convert the entries in LIANA's consensus CCC resource to rattus norvegicus ortholog gene symbols for downstream analysis.We then implemented the "liana_wrap" and "liana aggregate" functions, at default settings to infer ligand receptor pairs and obtain consensus ranks across all methods using Robust Rank Aggregation (RRA) for each dataset (Saline, LPS and Poly(I:C) treatment groups) separately.LIANA outputs were then subset to predicted ligand receptor interactions with aggregate rank scores ≤0.05.
Microscopy was performed on an OLYMPUS BX63 fluorescence microscope using Olympus cellSens software (Olympus, Shinjuku, Tokyo, Japan).Quantification was performed using QuPath software (Bankhead et al., 2017).Images from cover slips of two independent experiments, in total 4 replicates per staining and cell type (OPC, mOLIGO, microglia), were used for quantification.For each independent 20× magnification field a total cell count was obtained based on DAPI positive nuclei, cells positive for the respective marker of interest were counted and positive cell counts were normalized to the total cell count.Phase contrast images of cells in culture were taken at an Evos XL Cell Imaging System AMEX3300 (ThermoFisher Scientific).

RT-qPCR of mRNA
All primers were designed using NCBI-Primer Blast (Ye et al., 2012), synthesized by Microsynth AG (Balgach, Switzerland) and are summarized in Table 1.Real time quantitative polymerase chain reaction (RT-qPCR) was performed on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad), with CFX Maestro software v1.1 (Bio-Rad), using SYBR Green I Master Kit (cat#:04887352001, Roche Applied Science, Penzberg, Germany) according to the manufacturer's protocols.Gene expressions levels of the indicated targets of interest were normalized to Rpl13a, as in previous work utilizing in vitro neuroinflammation models (Boccazzi et al., 2021), according to the ΔΔCt method, as previously described (Pfaffl et al., 2002).

Statistical analysis
Frequencies of rearing, climbing and grooming bouts were compared using one-way ANOVA, followed by Tukey post hoc comparisons.Changes in bodyweight in the different treatment groups were analysed using two-way repeated measures ANOVA, followed by Bonferroni's multiple comparisons tests.Between group comparisons of Serum IFN-β levels were analysed using One-way ANOVA, followed by Tukey post hoc pair-wise comparisons.For RT-qPCR data relative quantification of target genes was performed according to the ΔΔCt method (Pfaffl et al., 2002), using Rpl13a as a reference gene.Gene expression values were reported as fold changes of gene expression levels in treatment groups relative to the mean expression levels in respective untreated controls.Outliers were removed using Grubbs' Test for Outliers before statistical analyses of between group differences.Fold changes of the expression of indicated genes of interest relative to untreated controls were analysed using Kruskal-Wallis-H-Tests, followed by Dunn's post hoc tests.A pvalue of <0.05 was set as threshold for statistical significance.All statistical analyses were carried out using GraphPad Prism v.8.01 (GraphPad Software).

LPS and Poly(I:C) treatments induce overlapping systemic cytokine responses
We induced systemic inflammation by i.p. application of the bacterial mimic LPS (5 mg/kg bodyweight), or the viral mimic Poly(I:C) (12 mg/kg bodyweight).Previous work documented pronounced and lasting effects of the indicated treatment regimens on brain function and CNS cytokine milieu (Bossù et al., 2012;Cunningham et al., 2007) and previous RNAseq analyses indicated robust changes to the transcriptome of several CNS cell populations 6 h after LPS application (Shemer et al., 2020;Crowell et al., 2020).
A reduction in locomotor activity, such as rearing, climbing and grooming, in both PAMP treatment groups was evident at the time point of serum and brain sampling (Fig. 1.B), in the absence of mortality and

Table 1
List of all primers used for qPCR assays, depicting target genes, primer sequences product length of amplicon and NCBI Accession numbers.  .However, we also detected a 16.28 fold increase of the mean IFN-β concentration upon LPS treatment relative to saline control levels, although not reaching the threshold of statistical significance (Fig. 1F).

Limited neuronal transcriptional response after systemic LPS and poly(I:C) treatment
After quality control and filtering, the integrated dataset contained 43,017 single nuclear transcriptomes of which 19,519 were derived from saline controls, 10,468 from LPS, and 13,030 from Poly(I:C) treated animals.Distribution of UMI counts, gene counts, percentages of contaminating mitochondrial RNA and gene count to UMI count ratios are reported in To identify shared and distinct cell type and stimulus dependent changes in gene expression upon systemic TLR-ligand challenge, DEGs were calculated for each annotated cell type, comparing clusters from LPS, or Poly(I:C) datasets with their corresponding clusters in the saline control dataset.As neurons represented the most abundant cell type in our dataset (Fig. 1G), we performed further sub-clustering of cortical glutamatergic neurons (GLU_Sv2b+_Satb2+) (Fig. 2A) and GABAergic interneuron populations (GABA_IN) (Fig. 2 B) to facilitate the detection of potential subtle, neuronal subtype specific transcriptional responses to systemic LPS and Poly(I:C) challenges.Overall we detected a total of 40 and 55 DEGs in glutamatergic and mixed clusters of LPS and Poly(I:C) treated animals, respectively (Fig. 3A,B).In GABAergic and cholinergic clusters 26 DEGs were induced by LPS and 3 DEGs by Poly(I:C) treatment (Fig. 3C,D), portraying a limited transcriptional response of neurons to peripheral PAMP challenges.Overall, we observed a slightly higher responsiveness of glutamatergic neuronal populations, particularly in the dataset derived from Poly(I:C) treated animals (Fig. 3B).Notably, the majority of DEGs induced by systemic application of Poly(I: C) represented in at least two neuronal clusters (N = 9) consisted of known anti-viral ISGs, such as Stat4, Samd9 (Peng et al., 2022), Eif2ak2 (Gal-Ben-Ari et al., 2018), Rnf213 (Kobayashi et al., 2015), Herc6 (Oudshoorn et al., 2012), or Usp18 (Honke et al., 2016).Several of these genes were also notably upregulated in non-glutamatergic neuronal populations (Fig. 3E-H), although not reaching the threshold of statistical significance.

LPS and poly(I:C) challenge induced pronounced shared and divergent transcriptional perturbations in vascular and ependymal cells
With a total of 308 DEGs (207 upregulated, 101 downregulated) upon LPS and 384 DEGs (238 upregulated, 146 downregulated) upon Poly(I:C) treatment, vascular cells were primarily the most responsive cell population in our dataset, followed by ependymal cells with a total of 123 LPS induced DEGs (43 upregulated, 80 downregulated) and 40 Poly(I:C) induced DEGs (all upregulated) (Fig. 4A-F).Genes upregulated upon systemic LPS and Poly(I:C) treatment overlapped substantially in vascular and to a lesser degree in ependymal cells (Fig. 4C).Vascular cells responded to both LPS and Poly(I:C) treatments with an upregulation of genes encoding pro-inflammatory cytokines, such as the CXCR3 ligands Cxcl9, 10 and 11 (Groom and Luster, 2011), key players in type I cytokine signalling such as Il6st, Irak2, Osmr, as well as JAK-STAT protein family members such as Stat3 and Stat4 (Morris et al., 2018).Further commonly upregulated genes included Cfb, and several metalloproteases, such as Adamts9.Furthermore, we observed a convergent cellular stress response to both LPS and Poly(I:C) challenge in vascular cells, evidenced by the common upregulation of established responders to ROS-stress, such as Sod2, Hif1a, Nfe2l2, encoding NRF2 and the antioxidant Txn1, as well as a host of heat shock protein-encoding genes, including Hsp90aa1, Hsp90b1 and Hspbh1 (Fulda et al., 2010) (Fig. 4G).
Likewise, the majority of DEGs downregulated by LPS and Poly(I:C) treatment in vascular cells were shared among both TLR-ligand treatment conditions (Fig. 4F) and belonged to functionally related, possibly co-regulated gene clusters.This was most evident for genes with established roles in maintaining endothelial blood brain barrier integrity and function (Fig. 4H).We detected a subtle downregulation of brain endothelial tight junction (TJ) complex associated genes such as Tjp1 and Cgnl1 (Stamatovic et al., 2016) upon LPS and Poly(I:C) treatment.More strikingly, both TLR-ligand challenges robustly downregulated genes encoding major mediators of directed substrate import and export across the blood brain barrier (Fig. 4H).These downregulated DEGs included multiple ATP-binding cassette (ABC) transporter superfamily members, such as Abcb1b, encoding multidrug resistance (MDR) P-glycoprotein1 (MDR1/Pgp), Abcc4 (MDR2/3) and Abcg2, also known as BCRP, as well as SLC-transporters such as the organic anion transporter Slco1c1, and several amino acid importers such as Slc7a5, Slc7a1 and Slc38a3 (Cressman et al., 2012).
Results of gene set enrichment analyses were congruent with this pronounced overlap of LPS and Poly(I:C) induced DEGs in vascular cells (Fig. A.4).The majority of enriched GO terms derived from DEGs upregulated in vascular cells in response to LPS and Poly(I:C) were shared, and pertained to a broad response to various cytokines and related intracellular signalling pathways, including JAK-STAT signalling, as well as cellular stress responses such as unfolded protein response and oxygen homeostasis (    Indeed, several established anti-viral ISGs such as the dsNA sensors Rig1 (Thoresen et al., 2021) andEik2ak2 (Gal-Ben-Ari et al., 2018), or pleiotropic inhibitors of viral replication and release such as Mx1 (Verhelst et al., 2013), Isg15 (Perng and Lenschow, 2018) and Rsad2 (Kurokawa et al., 2019), were uniquely upregulated by Poly(I:C), but not LPS in both vascular and ependymal cells (Fig. 4I).Other ISGs including Ifi44 (Busse et al., 2020), E3-ubiqutin ligases, such as Rnf213 (Ahel et al., 2020), genes encoding antiviral oligoadenylate synthetaseslike proteins such as Oasl2 (Zhu et al., 2014), or Cd274, encoding the IFN-inducible inhibitory immune receptor ligand PDL-1 (Copic et al., 2023) were induced substantially by Poly(I:C), but only subtly by LPS in vascular cells (Fig. 4).Some ISGs, such as the IFN-induced transmembrane family member Ifitm3 (Bailey et al., 2014), were however upregulated in vascular cells of both LPS and Poly(I:C) treated animals to similar degree (Fig. 4G).Taken together, while a limited induction of IFN-mediated signalling was detectable upon LPS treatment, it was profoundly induced by Poly(I:C) in vascular cells.
This dominant role of IFN and viral response pathways in response to Poly(I:C) was even more evident in ependymal cells, where the vast majority of Poly(I:C) induced DEGs consisted of ISGs (Fig. 4E,I).Congruently, more than half of the TOP 20 GO-terms derived from systemic Poly(I:C) induced upregulated DEGs in ependymal cells, were related to either type I IFN-mediated signalling, or anti-viral host responses (Fig.A.4B).In contrast, systemic LPS stimulation only moderately induced a small number of ISGs such as Rnf213, Oasl2 and Nlrc5 in ependymal cells, but led to a prominent upregulation of canonical innate immune response mediators such as Cp, Csf1 and Cfb.Moreover, several DEGs derived from ependymal cells of LPS treated animals were related to cilliogenesis and ciliar movement such as the cilliogenic transcription factor Rfx3, dynein axonemal heavy chain family members such as Dnah9, or the flagellar protein Spef2 (Fig. 4D), corroborated by gen set enrichment analysis (Fig.A.4B).Interestingly, while none of the genes downregulated by systemic Poly(I:C) met our DEG criteria, several genes were notably downregulated upon LPS treatment in ependymal cells.These included several genes also downregulated in vascular cells, of the LPS dataset such as Cxcl12, Bsg and Itm2a, but also unique DEGs, including neuromodulatory peptides, such as Ptn (González-Castillo et al., 2014) (Fig. 4 D,H, Fig.A.4 D).Overall the overlap between DEGs induced by systemic LPS and Poly(I:C) treatment was less pronounced in ependymal compared to vasculature cells, with a dominant role of ISGs among systemic Poly(I:C) treatment induced DEGs.
Gene set enrichment analyses, where the DEGs depicted in Fig. 5E were entered into Enrichr GO and Reactome gene set database queries, corroborated these findings.The majority of the TOP 10 enriched biological processes induced in oligodendrocyte lineage cells upon Poly(I: C) treatment related to type I and II IFN-mediated signalling pathways, negative regulation of viral genome replication, and related anti-viral host defences such as ADP ribosylation, mediated by the Poly(ADPribose) polymerases Parp9, Parp14 and Zc3hav1 (Todorova et al., 2015;Xing et al., 2021) (Fig. 5 F,G).
Only few regulated genes in astrocytes of Poly(I:C) and LPS treated animals met all defined DEG criteria ( We next analysed DEGs induced by systemic LPS treatment in cells of the oligodendrocyte lineage to determine whether the above described Poly(I:C) induced ISG response (Fig. 5E) was stimulus dependent.Indeed, with the exception of Ifi44, none of the core Poly(I:C) induced ISGs were detected among the LPS challenge induced DEGs in OPC, imOLIGO, or mOLIGO clusters (Fig. 5H,J).Few DEGs were upregulated in the LPS treatment group, while the downregulated DEGs did not group into co-regulated, functional categories similar to the ones observed in oligodendrocyte clusters upon Poly(I:C) treatment.Taken together, the acute response of oligodendrocytes to a peripheral Poly(I: C) challenge was characterized by the induction of a highly stimulus dependent IFN-driven, anti-viral gene signature.

Minor cellular communication network perturbations between vascular, ependymal and oligodendrocyte lineage cells after systemic PAMP immune challenges
We used LIANA to infer potential effects of systemic LPS and Poly(I: C) immune challenges on cell-cell communication processes.As vascular, ependymal and oligodendrocyte lineage cells emerged as the most responsive cell populations within our dataset, we focused on potential PAMP treatment induced changes on ligand receptor signalling between those cell-types.To focus on the most robustly predicted ligandreceptor complexes, we subset all LIANA predicted ligand receptor interactions to aggregate rank scores ≤0.05 (supplementary data file 5).The top 50 predicted ligand receptor interactions between vascular, ependymal cell and oligodendrocyte lineage cells based on the -log 10 of the aggregate rank scores are depicted in Fig. A.6.The majority of predicted ligand receptor pairs was comparable between all conditions, although we observed a slight decrease in the overall number of predicted ligand receptor interactions from oligodendrocyte lineage cells and ependymal cells towards other cell populations in the LPS treatment group dataset, as compared to datasets from saline and Poly(I:C) treated animals (Fig. A.6). Ligand receptor pairs unique to either saline control or one of the PAMP treatment groups were sparse and consisted of genes, which were up or downregulated by systemic PAMP treatment.For example, in line with the downregulation of Tfrc and Igf1r upon systemic PAMP treatment in vascular cells, predicted B2m -Tfrc, and Gnai2 -Igf1r ligand receptor signalling was unique to the Saline dataset, while Angpt2 -Tek, Cxcl11-Grm7, or Tgm2 -Itga9 ligand receptor signalling was predicted exclusively in the LPS and Poly(I:C) datasets.

The induction of an anti-viral response gene signature in oligodendrocytes is dependent on IFN stimulation
ISGs are directly inducible through both dsNA sensing, or IFN stimulation (Ashley et al., 2019).Therefore, we next investigated whether the ISG signature, observed in oligodendrocyte lineage cells upon Poly(I: C) treatment in vivo, was induced directly by Poly(I:C) treatment or indirectly via Poly(I:C)-induced IFN production.To this end we purified and cultured OPCs, and maturated them to mOLIGOs to allow for the detection of possible developmental stage-dependent treatment effects.Microglia, which are highly sensitive to both TLR-ligands and IFN stimulation (He et al., 2021;Colton et al., 1992), served as a positive control.
We investigated the effect of the indicated treatments on the gene expression of 9 representative targets, derived from the oligodendrocyte ISG signature reported in Fig. 5E.TLR-ligand treatment induced only moderate ISG upregulation in OPCs and mOLIGOs (Fig. 7).Only the increase in Herc6 (Fig. 7B), Mx1 (Fig. 7G), and Ifi44 (Fig. 7H) gene expression in Poly(I:C) treated OPCs relative to untreated OPCs reached statistical significance.In contrast, virtually all screened ISGs were substantially and significantly upregulated by either IFN-β or IFN-γ in both OPCs and mOLIGOs (Fig. 7).These findings indicate, that Poly(I:C) stimulation alone is insufficient to induce the full ISG signature observed in vivo, while in vitro treatment with IFN-β and IFN-γ fully mimicked the core ISG signature (Fig. 5E) in oligodendrocyte lineage cells, at various developmental stages.
Contrasting the effects of TLR-ligand and IFN treatment on the ISG response in oligodendrocyte lineage cells, the gene expression of virtually all targets of interest was significantly upregulated by LPS, Poly(I:C) and IFN-β treatments, relative to untreated controls in microglia (Fig. 8).In IFN-γ treated microglia, none of the analysed ISGs were significantly upregulated.However, the upregulation of Rnf213 and Herc6, in IFN-γ treated microglia relative to untreated controls approached, but not reached statistical significance (Fig. 8).

Discussion
The relative contributions of the different cell populations in the brain to the overall CNS response to different immune challenges in vivo is complex and not well-understood.The cellular heterogeneity of the CNS poses a significant obstacle in understanding the molecular basis of overlapping and divergent responses to different PAMPs, as well as the underlying mechanisms of the brain's response to innate immune challenges.Addressing this challenge, single cell and single nuclei RNA sequencing approaches have considerably advanced our understanding of cell type specific innate immune responses (Duan et al., 2018;Shemer et al., 2020;Sousa et al., 2018;Allen et al., 2023;Crowell et al., 2020).In the present study we describe several cell type-and stimulus-dependent acute transcriptional responses of the mammalian brain to peripheral bacterial and viral like stimuli at single cell resolution.In agreement with previous studies, we observed a relative transcriptional quiescence of neurons after systemic LPS and Poly(I:C) challenges, compared to non-neuronal cell populations (Allen et al., 2023;Crowell et al., 2020).
Although glutamatergic neurons were slightly more transcriptionally responsive, particularly after systemic Poly(I:C) stimulation, no robust neuronal subtype or stimulus specific gene signature emerged from our dataset.Nevertheless, as a result of our standardized brain tissue sampling method, which included several but not all brain regions, we acknowledge that our analysis may have overlooked important transcriptional responses of minor neuronal populations with limited spatial distribution.In line with this assumption, work by Ilanges et al. showed that the sickness behaviour response induced by systemic LPS application can be mainly traced to specific subgroups of neurons in the brainstem, indicating that the reactions of neurons to peripheral inflammation vary significantly across spatially confined brain areas (Ilanges et al., 2022).
In contrast to the observed low number of DEGs after systemic TLRligand treatment in neuronal cells, we identified a robust upregulation of common gene signatures in vascular cells in response to both stimuli.In line with previously described transcriptional profiles of endothelial cells derived from the CNS of rodents subjected to acute LPS treatment, both TLR-ligands elicited a similar transcriptional cytokine response in vascular cells in our data set (Crowell et al., 2020;Jambusaria et al., 2020).These findings suggest a pronounced convergence of intracellular signalling pathways downstream of TLR3 and 4 in vascular cells.In fact, the interaction between LPS and TLR4 primarily triggers proinflammatory NfκB signalling, but it can also lead to internalization of TLR4 and activation of the TRAF3/IRF3/IFN pathway within the endosomes (Ciesielska et al., 2021).This signalling pathway has been found to rely on the co-activation of CD14, which is also expressed in brain vascular cells in our data set (data not shown) (Ciesielska et al., 2021;Varatharaj and Galea, 2017).However, our study specifically focused on comparing the inflammatory response of the brain to TLR3 and TLR4 activation, and thus, the detailed signalling cascades following TLR activation were not analysed and require further investigation.
Interestingly, the upregulation of immunomodulators, such as the suppressor of cytokine signalling Socs3 (Carow and Rottenberg, 2014), inhibitor of IFN signalling Usp18 (Honke et al., 2016) and immune checkpoint molecules Cd274 and Cd47 (Marin-Acevedo et al., 2021), suggests a remarkably early immunomodulatory response, presumably limiting acute inflammation.Moreover, the upregulation of various cellular stress response genes is indicative of a protective response to ROS stress, in brain vascular cells during acute PAMP immune challenges.Given that ROS and inflammatory response pathways are tightly interlocked (Mittal et al., 2014), the upregulation of pathways that ameliorate ROS stress might serve as a crucial safeguard to prevent the buildup of potentially detrimental positive feedback loops and subsequent tissue injury in the brain vasculature, during acute inflammatory challenges.
The extensive downregulation of genes encoding important BBB transporters represented a further shared acute transcriptional response of vascular cells to systemic administration of both TLR-ligands in our data.Our findings are in agreement with previous work documenting a decreased expression of ABC-Transporters, such as P-gp in whole brain lysates of LPS treated WT rats (Goralski et al., 2003;Wang et al., 2005) and brain cortex crude membrane fractions of LPS treated Alzheimer's diseases model APdE9 mice (Puris et al., 2021).In line with the role of many luminal ABC-Transporters as drug exporters of brain endothelial cells, the LPS induced downregulation of P-gp coincided with the brain retention of exogenous substrates such as H 3 -digoxin (Goralski et al., 2003), or 99 m Tc-sestamibi (Wang et al., 2005).Indeed, the influence of various states of inflammation on the pharmacokinetics of multiple drugs has been documented in rodent models and humans (Cressman et al., 2012;Saib and Delavenne, 2021).Of note, these studies on the expression and functionality of ABC-transporters in vivo predominantly utilized whole tissue samples, impeding the study of cell type specific effects.Thus, our in vivo single cell omics approach allowed the study of cell type-dependent effects of inflammation on BBB-transporter D. Bormann et al. (caption on next page) D. Bormann et al. expression in detail.In contrast to our observation, previous in vitro studies documented variable effects of different PAMP and cytokine treatment regimens on ABC-Transporter expression and function, which largely depended on the endothelial cell culture model and inflammatory stimulus (Saib and Delavenne, 2021;Lye et al., 2023).In addition to the observed downregulation of ABC-family members, we also found a significant downregulation of other BBB transporters such as Tfrc, lipid transfer proteins such as Pltp, which was linked to BBB integrity (Zhou et al., 2014), and several members of the SLC-family.These findings are in agreement with previous work showing an acute LPS induced downregulation of brain endothelial Slc7a5 (Wittmann et al., 2015).Collectively, the widespread downregulation of diverse BBB-transport genes might implicate a far reaching dysregulation of both directed influx and efflux processes at the BBB during acute inflammation.
In contrast to the observed shared responses to systemic LPS and Poly (I:C) challenges, the upregulation of anti-viral, IFN-associated genes was substantially more pronounced upon treatment with the viral RNA mimic Poly(I:C).Recent sc and snRNAseq studies of neurotropic JHMV (Syage et al., 2020) and LGTV (Chotiwan et al., 2023) infected rodent brains reported gene signatures in various CNS cell types, reminiscent of the herein observed transcriptional response to systemic Poly(I:C) application.A striking induction of both Type I and Type II IFN signalling pathways in various immune and brain parenchymal cell populations was evident in both RNA virus infection models, as well as in the brains of Poly(I:C) treated animals in our dataset (Syage et al., 2020;Chotiwan et al., 2023).Moreover, recent studies causally implicated IFN-I dependent signalling in restricting viral replication and dissemination (Chotiwan et al., 2023) and reducing mortality (Bühler et al., 2023) in rodent models of viral encephalitis.Likewise, abrogation of IFN-γ signalling in oligodendrocytes during viral encephalitis was associated with increased viral load and loss of oligodendrocytes, as well as demyelination, neurite damage and increased mortality (Parra et al., 2010).The involvement of Type I and II IFN signalling in numerous physiological processes and diseases, in addition to viral host defence mechanisms, further highlights the exceptionally widespread responsiveness to IFN across the heterogeneous cell population of the mammalian CNS (Viengkhou and Hofer, 2023;Dafny and Yang, 2005).
In cells of the oligodendrocyte lineage we observed a specific antiviral transcriptional response to Poly(I:C) in vivo, which was rather driven by IFN stimulation than direct Poly(I:C)-PRR mediated mechanisms, as indicated by our in vitro validation assays.Indeed, while stimulation with IFNβ and IFNγ in vitro induced a comparable ISG signature in cultivated oligodendrocytes, stimulation with TLR-ligands was not as effective.In contrast, microglia cell cultures responded significantly stronger to direct TLR than to IFN stimulation.Our findings expand upon previous work, documenting a lower baseline level expression of dsNA sensors, a more limited and delayed ISG response and higher viral RNA titers in oligodendrocytes as compared to microglia of JHMV infected and Poly(I:C) treated mice (Kapil et al., 2012).Importantly, the same study showed that the in vivo upregulation of Mda5 and Rig1 in brain tissue of JHMV infected mice depended on type I IFN signalling (Kapil et al., 2012).Type I IFNs are upregulated in brain tissue as early as 2 h after Poly(I:C) application and 3 to 5 days after JHMV infection, while IFN-γ is notably upregulated 5 to 7 days after viral infection (Kapil et al., 2012).In our study, peripheral upregulation of both IFN-β and IFN-γ, along with BBB permeability disrupting chemokines such as CCL2 (Yao and Tsirka, 2014) was evident 7 h after Poly (I:C) administration.Previous research has shown a profound increase in BBB permeability 24 h after i.p. injection of Poly(I:C) (Wang et al., 2004), allowing for the speculation that the entry of interferon from the systemic circulation into the brain parenchyma might be facilitated after systemic PAMP challenge.As LPS and Poly(I:C) are most likely not transported across the intact BBB to a degree sufficient enough to elicit profound PRR signalling (Mallard, 2012), our data suggests that secondary processes are instead responsible for the observed effects.Our work is in line with previous studies showing that multiple behavioral outcomes and perturbations of the CNS cytokine milieu, upon systemic Poly(I:C) application are mediated largely by IFNs produced outside of the CNS (McGarry et al., 2021).Furthermore, previous work defined multiple mechanisms by which IFN signalling is relayed to the brain parenchyma via brain endothelial and epithelial IFN-receptor chain 1 (IFNAR), which in turn signal to other CNS cell populations via tertiary signalling molecules, such as CXCL10 (Blank et al., 2016).Importantly, these studies indicate mechanisms by which peripherally induced IFN triggers IFN-dependent signalling in the brain parenchyma at the BBB interface, which do not depend on IFN transport from the periphery across the BBB.However, the exact underlying mechanisms, by which interferon signalling might be relayed to the CNS, in particular to oligodendrocytes in our study remain to be fully elucidated and require further investigation.
Although a constitutive type I IFN production in various cell populations of the brain has been previously discussed (Blank and Prinz, 2017), recent work implicated choroid plexus epithelial cells and to a lesser degree astrocytes and microglia as the main CNS endogenous sources of IFN-β, and CD8 T and NK cells as the main IFN-γ producers during viral encephalitis (Chotiwan et al., 2023).In line with a previous publication (Nagy et al., 2020), these cell populations are either absent or strongly underrepresented in snRNAseq datasets using a similar nuclei purification protocol, explaining the absence of clusters with a robust IFN production signature (data not shown).As a result of this constraint, we are currently unable to specify the roles played by peripheral cells and cells of the CNS in the observed IFN response signature described in this study.Previous work provided detailed characterizations of the effects of peripheral LPS challenges on the transcriptome of microglia and astrocytes at single cell resolution (Hasel et al., 2021;Sousa et al., 2018), while comparable analyses of the effects of peripheral Poly(I:C) administration on these populations are still missing.Here our dataset leaves a critical gap, that needs to be addressed by future research to more comprehensively elucidate the complex intercellular interactions in the CNS upon Poly(I:C) challenge.
Several of the herein reported ISGs such as Rnf213, Herc6, or the sparsely researched Schlafen family member Slfn4, have currently undefined regulatory roles in oligodendrocyte lineage cells.We propose that additional investigations into the role of these ISGs in oligodendrocytes could provide valuable insights extending beyond the field of viral host defence research.For instance, previous work showed that IFNγ abrogates the differentiation of OPCs into myelinating oligodendrocytes (Tanner et al., 2011;Kirby et al., 2019).Interestingly, independent research demonstrated that Nfat/calcineurin signalling promotes oligodendrocyte differentiation and myelination (Weider et al., 2018).Given that RNF213 has been implicated in the ubiquitination and thus proteosomal degradation of NFAT1 (Mineharu and Miyamoto, 2021), it might represent a promising candidate target for future studies on the molecular mechanisms underlying the suppressive effects of IFNγ on oligodendrocyte differentiation.
Collectively, our study unveils overlapping and distinct neural cell type dependent transcriptional perturbations in response to bacterial and viral PAMP immune challenges, highlighting shared multifaceted cytokine, cellular stress and BBB transporter related gene signatures.However, we denote several limitations of our approach.Although, we observed considerable overlaps in the serum cytokine profiles and transcriptional responses in vascular and ependymal cells to systemic  LPS and Poly(I:C) challenges, the overall magnitude of systemic and CNS inflammation elicited by these stimuli might still differ.Indeed, previous work showed a more potent effect of LPS on anorexia, fever and locomotor activity, as compared to Poly(I:C) (Hopwood et al., 2009).Although in our study both treatments reduced locomotor activity and body weight, more sensitive measurements of sickness behaviour over a prolonged period of time will be necessary to identify differences on behavioral outcomes, between the herein used PAMP treatment regiments.Furthermore, several studies documented a more robust induction of sickness behaviour and upregulation of pro-inflammatory cytokines, including IFNβ, in response to systemic application of high molecular weight (HMW) as compared to low molecular weight (LMW) Poly(I:C) challenges (McGarry et al., 2021;Mueller et al., 2019;Kowash et al., 2019).It is thus well conceivable that HMW Poly(I:C) might elicit an even more pronounced IFN response in the brain.Importantly, previous scRNAseq studies which deliberately purified brain vascular cells, unveiled a remarkable heterogeneity of endothelial, pericyte and mural cell populations (Jeong et al., 2022;Vanlandewijck et al., 2018), as well as an endothelial subtype-dependent responsiveness to neuroinflammation in EAE models (Jeong et al., 2022;Fournier et al., 2023).We therefore concede that the relatively low abundance of vascular cells in our dataset, which at the given sequencing depth impeded a finer resolution of endothelial and mural subpopulations, poses a significant limitation of this study.Furthermore, given that many effects of systemic PAMP challenges on the CNS are induced via secondarily released mediators (Blank et al., 2016), the precise anatomical localization of reactive vascular cells, as well as the spatial relationship between reactive vascular cells and surrounding reactive cells carries important information, that is not captured by our snRNAseq approach.
Notably, our work elucidates the acute transcriptional response of several CNS cell populations to moderately severe systemic TLR-ligand challenges in a priori healthy young rodents.While important overlaps of the herein observed gene signatures and data from viral encephalitis models (Chotiwan et al., 2023) are evident these approaches as well as the modelled neuropathologies differ substantially.Moreover, previous work documented several age dependent effects of systemic PAMP challenges (McGarry et al., 2021), as well as exacerbated PAMP responses in a rodent model of prion disease (Field et al., 2010).Our work thus invites for further investigations into the acute and subacute response of diseased and aged brains to immune challenges, the CNS responses to severe sepsis and the effect of chronic and reoccurring PAMP challenges, at single cell resolution.Moreover, our dataset documents the effect of systemic LPS and Poly(I:C) challenges at a single time point.As several previous studies documented dynamic changes in systemic and brain cytokine expression, with some peaking already within 2-3 h after PAMP challenge (Duan et al., 2018;McGarry et al., 2021;Qin et al., 2007) longitudinal snRNAseq analyses across multiple time points, with concomitant sampling of sera, peripheral immune cells and CNS tissue offer an opportunity for further mechanistic insights.
Our findings might bare important translational implications.In line with previous work (Saib and Delavenne, 2021), we propose that the downregulation of BBB-transporters might promote the brain retention of a variety of drugs in patients with inflammatory events, as well as impede the SLC-transport mediated import of essential amino acids into the brain.However, further investigations are required to assess whether the herein reported dysregulations of BBB transporter expression extend to reoccurring or chronic inflammatory challenges and translate to clinically relevant BBB transporter dysfunctions in patients.Our study further identified a dominant role of IFN signalling in response to viral PAMPs, particularly in oligodendrocytes.In agreement with previous findings (Parra et al., 2010) our study highlights the pronounced dependency of oligodendrocytes on IFN stimulation in order to mount a full anti-viral transcriptional response.Collectively, these findings suggest a particular vulnerability of oligodendrocytes to viral challenges, in the absence of primary immunocompetent cells which provide a secondary IFN response.
Overall our dataset provides the first direct comparison of the mammalian CNS transcriptome in response to acute LPS and Poly(I:C) challenges, at single cell resolution.We propose that the dissection of the heterogeneous, cell type and stimulus dependent transcriptional responses to inflammation will help to untangle the molecular basis of the detrimental effects of inflammation on the CNS, its physiological homeostatic responses to it and ultimately aid in improving treatment strategies in patients undergoing inflammatory events.
all animals (N = 4 per group) were included.Although not reaching the threshold of statistical significance the mean frequencies of rearing (LPS: 5, Poly(I:C): 7), climbing attempts (LPS: 1.5, Poly(I:C): 2.5) and grooming bouts (LPS: 0.25, Poly(I:C): 0.75) was lower in the LPS compared to the Poly(I:C) treatment group (Fig.1.B).Bodyweight decreased slightly but significantly during the 7 h period in both TLRligand treatment conditions (Fig.1.B).To validate the successful induction of peripheral inflammation and to systematically compare changes in the peripheral cytokine profile in response to LPS and Poly(I: C) we assayed sera from LPS and Poly(I:C) treated animals, as well as saline controls using a comprehensive panel of 79 cytokines and related targets (Fig.1C-E).Relative to saline controls, 27 proteins were upregulated above 1.5 fold in LPS and 22 proteins in Poly(I:C) group sera 7 h after i.p. application (Fig.1C,D), while only 7 and 4 proteins were downregulated by LPS and Poly(I:C), respectively.Notably, the majority of proteins upregulated above 1.5 fold in either of the TLR-ligand treatment groups (N total = 31), were also enriched in the sera of both LPS and Poly(I:C) treated animals (N shared = 18, [58,06%]) (Fig.1E).This included robustly enriched chemokines such as the CC-chemokine ligands (CCL) CCL2, CCL5 and CCL20, while CCL21 and CXCL2 were close to uniquely upregulated by LPS.Likewise, the matrix metalloproteases MMP3 and MMP9, the serine proteinase inhibitor Serpin E1 and the pro-inflammatory interleukins IL-1β and IL-6 were enriched in the sera of both LPS and Poly(I:C) treated animals, while IL-1α, IL-1ra and IL-3 were restricted to sera of LPS, and IL4 to sera of Poly(I:C) treated animals.Enrichment of the growth factors GDF-15 and G-CSF was observed in both TLR-ligand treatment groups, while FGF-7, CNTF and EG-VEGF were distinctly enriched in sera of LPS treated animals and VEGF was only notably upregulated upon Poly(I:C) treatment.Interestingly, IFN-γ was upregulated to similar extent in both LPS (4.35-fold relative to saline controls) and Poly(I:C) (4.85 fold relative to saline controls) treated animals.Due to our sample size IFN-β serum levels were only significantly elevated in sera of Poly(I:C) treated animals as compared to both saline control (61.94 fold increase of mean concentration), and LPS treatment groups (3.80 fold increase of mean concentration) Fig.A.1C.Cell clusters were annotated using well established cell-type specific marker genes.Cell cluster annotations are reported in Fig. 1G and Fig. 2, curation of marker genes for cluster annotation is described in detail in the supplementary notes in Appendix A and Fig.A.2 and Fig.A.3.
Fig.A.4 A).Enrichment analysis inputting downregulated DEGs implied a downregulation of biological processes related to the transport of various amino acids, proteins and lipids across the blood brain barrier vasculature in response to both LPS and Poly(I:C) (Fig.A.4C).In contrast, GO-terms associated to Type I IFN signalling and antiviral responses were enriched more robustly upon Poly(I:C) stimulation,

Fig. 2 .
Fig. 2. snRNAseq sub clustering of cortical glutamatergic neurons and interneurons.(A) Sub clustering analysis of glutamatergic, Sv2b/Satb2 double positive neurons.UMAP plot is shown in the left panel.Relative abundances of each sub cluster within each treatment group are presented as pieplots.Curated cluster markers are presented as dotplot in the right panel.Estimated cortical, laminar distributions are depicted as colored bars (L = layer 2 to 6, C = Claustrum).(B) Sub clustering analysis of GABAergic interneurons.UMAP plots are shown in the left panel, relative abundances of each sub cluster within each condition are presented as pieplots.Curated cluster markers are presented as dotplot in the right panel.Color code showing average gene expression and indicating percentages of nuclei expressing the respective gene.Cluster Abbreviations: Glutamatergic, Sv2b/Satb2 double positive (GLU_Sv2b+_Satb2+_1 to 11), Adarb2 positive GABAergic interneuron (GABA_IN_Adarb2+_1 to 3) and Adarb2 negative GABAergic interneuron (GABA_IN_Adarb2-) sub clusters are numbered consecutively, GABA_IN_THAL = GABAergic thalamic interneurons, CHOL_IN = Cholinergic interneurons.

Fig. 3 .
Fig. 3. Differentially expressed genes in neuronal cells of LPS and Poly(I:C) treated animals.Strip plots showing DEGs in glutamatergic and mixed neuronal clusters and sub clusters of (A) LPS treated and (B) Poly(I:C) treated animals, as well as GABAergic and cholinergic clusters and sub clusters of (C) LPS and (D) Poly(I:C) treated animals.DEG thresholds in all strip plots are color coded, only genes meeting all thresholds (adjusted p-values <0.05 and log2(FC) ≥ 0.58) are labelled.(E-H) Average expression levels of selected genes meeting all DEG thresholds (adjusted p-values <0.05 and log2(FC) ≥ 0.58) in ≥3 neuronal clusters (E: Rnf213, F: Eif2ak2, G: Usp18, H: Stat4) are depicted as feature plots, split by treatment group.

Fig. 4 .
Fig. 4. Differentially expressed genes in vascular and ependymal cells of LPS and Poly(I:C) treated animals.Volcano plots showing DEGs in vascular cells (VASC) of (A) LPS and (B) Poly(I:C) treated animals, as well as in ependymal cells (EPC) of (D) LPS and (E) Poly(I:C) treated animals.UpSet plots depicting numbers of total, unique and overlapping (C) upregulated and (F) downregulated DEGs (adjusted p-values <0.05 and log 2 (FC) ≥ 0.58) upon LPS and Poly(I:C) treatment in vascular and ependymal cells.Average expression levels of representative DEGs (G) commonly upregulated or (H) downregulated in vascular cells of both LPS and Poly(I:C) treated animals and (I) DEGs more strongly regulated in either treatment group are presented as violin plots, split by cluster and condition.Colored bars denote functional categories related to presented DEGs.
Fig.A.5A,B).Although the majority of the core upregulated ISGs identified in oligodendrocyte linage clusters of Poly(I:C) treated animals were also upregulated in astrocytes (Fig.A.5C), only Rnf213, Parp9 and Mx1 met the threshold of statistical significance, likely due to the sparse representation of astrocytes in our dataset.

Fig. 5 .
Fig. 5. Differentially expressed genes in oligodendrocyte lineage cells of LPS and Poly(I:C) treated animals.Volcano plots showing DEGs in (A) oligodendrocyte precursor cell (OPC), (B) immature oligodendrocyte (imOLIGO) (C) and myelinating, mature oligodendrocyte (mOLIGO) of Poly(I:C) treated animals.(D) UpSet plot depicting numbers of total, unique and overlapping upregulated DEGs in oligodendrocyte lineage clusters, upon Poly(I:C) treatment.(E) All genes meeting DEG criteria (adjusted p-values <0.05 and log 2 (FC) ≥0.58) in at least 2 oligodendrocyte lineage clusters, upon Poly(I: C) stimulation (N = 22) are presented as violin plots, split by treatment group and oligodendrocyte lineage cluster, with gene expression values plotted on y axes.(F) Dot plots of top 10 enriched GO and (G) REACTOME database terms of Poly(I:C) upregulated genes depicted in (E) are shown, with enriched terms on y axes, − log 10 of Benjamini-Hochberg method adjusted p-values on x axes, dots colors encoding Enrichr combined scores, dot sizes indicating gene ratios.(H-J) Volcano plots of DEGs in (H) OPC (I) imOLIGO and (J) myelinating, mOLIGO of LPS treated animals.

Fig. 6 .
Fig. 6.Characterization of cell culture purity and developmental stage.(A-E) Immunofluorescence labelling of oligodendrocyte lineage and microglia markers.Representative ICC microscopy images are presented at 40× magnification (scale bars = 50 μm).Corresponding quantifications are presented on the right side of the images as bar graphs (Means +/-SD), representing counts of cells positive for the respective marker, normalized to total cell numbers per 20× field of view (N = 4 per staining, from N = 2 independent experiments).P-values are derived from Student'st-tests (ns ≥ 0.05, ** ≤ 0.01, *** ≤ 0.001).(A) Primary rat OPCs cultures at DIV2 stained for NG2 (green).(B) Iba1 (green) staining of OPC cultures (B, left image) and microglia (MG) cultures (B, right image) at DIV2. (C) NG2 (green) and BCAS1 (orange) double labelling of OPC cultures at DIV2 (upper panel) and mOLIGO cultures at DIV6 (lower panel).Student'st-tests revealed a significant decrease in the number of NG2 positive cells (t = 4.96, df = 6, p = 0.0026) and a significant increase of cells positive for BCAS1 (t = 4604, df = 6, p = 0,0037).(D) CNPase (green) and NogoA (orange) double labelling of OPC cultures at DIV2 (D, upper panel) and mOLIGO cultures at DIV6 (D, lower panel).Student'st-tests revealed a significant increase in the number of NogoA positive cells (t = 5760, df = 6, p = 0.0012).(E), MBP (orange) staining of OPC cultures at DIV2 (E, left image) and mOLIGO cultures at DIV6 (E, right image).Student'st-tests revealed a significant increase in the number of MBP positive cells (t = 6258, df = 6, p = 0.0008).(A-E), Nuclei in all stainings visualized with DAPI.F, Representative phase contrast images of OPCs at DIV2 (F, left image) and mOLIGOs at DIV6 (F, right image) in culture (scale bars = 50 μm).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)