The choroid plexus synergizes with immune cells during neuroinflammation

SUMMARY

The choroid plexus (ChP) is a vital brain barrier and source of cerebrospinal fluid (CSF). Here, we use longitudinal two-photon imaging in awake mice and single-cell transcriptomics to elucidate mechanisms of ChP regulation of brain inflammation. We used intracerebroventricular injections of lipopolysaccharides (LPS) to model meningitis in mice and observed that neutrophils and monocytes accumulated in ChP stroma and surged across the epithelial barrier into the CSF. Bi-directional recruitment of monocytes from the periphery and, unexpectedly, macrophages from the CSF to the ChP helped eliminate neutrophils and repair the barrier. Transcriptomic analyses detailed the molecular steps accompanying this process and revealed that ChP epithelial cells transiently specialized to nurture immune cells, coordinating their recruitment, survival and differentiation, and regulation of the tight junctions that control the permeability of the ChP brain barrier. Collectively, we provide a mechanistic understanding and comprehensive roadmap of neuroinflammation at the ChP brain barrier.

Graphical Abstract

In-brief:

The choroid plexus (ChP) is a hub of immune activity, in which a specialized population of epithelial cells coordinates stepwise immune cell recruitment, infiltration, differentiation, and adhesion to the ChP upon acute brain inflammation. Such coordination contributes to the timely breaking-down and repair of the ChP brain barrier.

INTRODUCTION

Brain barriers including the blood-brain barrier (BBB) and the blood-cerebrospinal fluid (CSF) barrier protect the central nervous system (CNS) from systemic insults^1–3. During acute inflammation, blood-borne pathogens including viruses (e.g., HIV, Zika, SARS-Cov-2)^4–8, bacteria (e.g., gram-positive and -negative bacteria)^9–11, and parasites (e.g., Trypanosoma brucei)^11–13 can directly gain access to the CNS via the brain borders^14,15. Inflammation following infection is associated with a growing number of lifelong neurologic conditions ranging from microcephaly due to congenital viral infections^16–18 and hydrocephalus due to meningitis^19,20, to enduring cognitive impairments associated with long-COVID²¹. Persistent CNS inflammation is also common in chronic, non-infectious conditions such as amyotrophic lateral sclerosis²² and Alzheimer’s disease²³. Elevated immune cell counts in the CSF are well-documented for many neurologic disorders^24–28. As such, there is a pressing need to increase our understanding of how the brain responds to and resolves inflammation.

As the principal blood-CSF barrier and source of CSF, the choroid plexus (ChP) critically regulates the physical and biological environment of the CNS^14,29. Located within each of the brain’s ventricles, the ChP comprises sheets of epithelial cells enclosing a vascularized stroma of mesenchymal and resident immune cells³⁰. ChP barrier permeability and secretion modulate CSF composition across development, circadian cycles, and nutritional states^31–33. The ChP has been long recognized to participate in neuroimmune functions by serving as a gateway for peripheral immune cell entry into the brain^1,2,28,34,35. The ChP and the brain’s ventricles are patrolled by resident macrophages (epiplexus / Kolmer cells) located on the apical, CSF-contacting surface of the ChP¹⁵. Recent transcriptome and fate-mapping analyses uncovered a transient expansion of resident and recruited macrophages at the brain borders including the ChP following Trypanosoma brucei exposure¹². Curiously, inflammation of the ChP itself is also emerging as a common feature of diverse and seemingly unrelated neurologic conditions including hydrocephalus³⁶, schizophrenia³⁷, amyotrophic lateral sclerosis³⁸, and Alzheimer’s disease³⁹. Pro-inflammatory transcriptional signatures identified in ChP samples from COVID-19 patients have been linked to chronic CNS inflammatory conditions⁴⁰. Despite the plethora of correlative evidence, a clear picture of the role of the ChP in acute or chronic inflammation has yet to emerge. Progress elucidating the underlying cellular and molecular mechanisms has been hampered by a lack of tools to visualize and manipulate the ChP barrier in vivo.

Here, we leverage recent breakthroughs in longitudinal two-photon imaging of ChP in awake mice⁴¹ and single-cell transcriptomics⁴² to track the cellular and molecular events at the ChP in real time in a murine model of acute brain inflammation by intracerebroventricular (ICV) delivery of lipopolysaccharides (LPS)⁴³, a bacterial endotoxin found in essentially all Gram-negative bacteria. Effects of ICV LPS resemble clinically diagnosed bacterial (e.g., N. meningitidis, S. pneumoniae, S. agalacie) meningitis characterized by high leukocyte / white blood cell count in the CSF⁴⁴. Following ICV injections of LPS, we observed a bi-directional recruitment and proliferation pattern of immune cells to the ChP, including macrophages to the ChP from the CSF, a paradigm not seen in most barrier tissues such as the intestine or airway, in addition to monocytes and neutrophils from the periphery. This recruitment pattern bolstered the available resident macrophages, especially those at the apical surface (the epiplexus cells) and promoted the resolution of inflammation and the healing of epithelial barrier. Informed by the imaging results, we performed single cell transcriptome analyses of CSF and ChP at specific times to reveal a molecular map and stepwise mechanisms for immune cell recruitment, survival, differentiation, and barrier repair. A key finding was the identification of a specialized and transient population of early responding epithelial cells (inf-Epi) that coordinated immune cell recruitment (by upregulating key chemokines), survival, and differentiation into macrophages (transient upregulation and secretion of colony-stimulating-factor CSF1/M-CSF), and regulating fundamental ChP barrier properties (matrix metalloprotease expression, upregulation of adhesion molecules for securing tissue-repair macrophages). Collectively, our data demonstrate a collaborative and synergetic relationship between immune cells and ChP epithelial cells. The ChP functions as an immune organ that actively addresses and repairs the blood-CSF barrier during inflammation, akin to other internal body organs. Our work provides a mechanistic understanding of neuroinflammation at the ChP brain barrier.

RESULTS

The choroid plexus is a central hub of brain immune cell activity in bacterial meningitis.

Examination of ChP samples from patients diagnosed with bacterial meningitis (18-year-old female, 10-day-old male, and 2-week-old male) revealed a striking number of leukocytes within the ChP and CSF compared to controls (Figure 1A). Indeed, the ChP was swollen and contained numerous peripheral leukocytes of high CD45-positivity that appeared to extravasate into the brain ventricles (Figure 1B, Figure S1A). The inflamed ChP tissues also harbored increased numbers of amoeboid, activated CD163+ macrophages (Figure 1C, Figure S1B), altogether indicating robust immune activation at the ChP-CSF barrier during meningitis.

Figure 1. The ChP is a key site of inflammation in meningitis.

(A-C) Pathology specimens from patients with meningitis and controls analyzed by H&E (A, CSF is pseudo colored in blue), CD45 (B, infiltrated immune cells), and CD163 (C, macrophages). Scale = 100 μm. (D) UMAP embedding of 11,765 immune cells collected from mouse CSF 24 hours or 48 hours following intracerebroventricular injection (ICV) LPS challenge, analyzed via scRNA-seq. Neu = Neutrophil; MΦ = Macrophage; IFN-stim = Interferon-stimulated. (E) UMAP colored by hash call assignment to each treatment group with pooled samples from 5 adult mice, including those without definitive hash identities (”Unassigned”). (F) Violin plot of representative genes used to assign cell type and state identity to clusters. (G) Representative histology images showing accumulation of S100A9+ leukocytes in wholemount mouse ChP explants. Scale = 500 μm (200 μm in G1-G3). (H) Representative images of wholemount mouse explants, showing increasing amounts of CX3CR1+ macrophages with amoeboid morphology following LPS. Scale = 500 μm (200 μm in H1-H3).

To elucidate the path that these cells take to enter the brain and related regulatory processes in the ChP, we modeled acute brain infection-associated inflammation in mice by a single intracerebroventricular (ICV) injection of lipopolysaccharides (LPS, E. coli), a major component of the cell walls of gram-negative bacteria. Using single-cell RNA-sequencing (scRNA-seq) we confirmed robust cellular infiltration into the CSF 24 hours following LPS delivery (Figure 1D–F, Figure S1C–F, Dataset S1). Leukocytes in mouse CSF 24 hours following LPS were predominantly neutrophils (Csf3r, Retnlg). These cells formed six distinct clusters including three Ngp+, Ly6g+ clusters, one Saa3+ cluster, and one IFN-stimulated cluster, consistent with reported diverse neutrophil populations following infection^45,46. By 48 hours, myeloid cells including monocytes and macrophages (Ccr2, Ly6c2, Cd74, H2-Aa) became the predominant CSF cell type. We confirmed the accumulation of these major cell classes at the ChP by immunostaining (Figure 1G–H; S100A9 pixel/μm² ChP: PBS 24 hrs = 0.29 ± 0.38 vs. LPS 24 hrs = 21.23 ± 8.77 vs. LPS 72 hrs = 4.73 ± 1.12, ** p = 0.0055, one-way ANOVA. CX3CR1 pixel/μm² ChP: PBS 24 hrs = 7.31 ± 1.59 vs. LPS 24 hrs = 12.09 ± 4.20 vs. LPS 72 hrs = 21.62 ± 3.52, ** p = 0.0049, one-way ANOVA) and flow cytometry, which also showed a consistent presence of these cell classes in the CSF following LPS and limited number of leukocytes in the CSF in control, baseline conditions (Figure S1G–H). These data demonstrate that ICV LPS in mice elicits the key diagnostic feature of bacterial meningitis in humans, an influx of leukocytes in the CSF⁴⁷. For comparison we ICV injected heat-killed and live Streptococcus agalactiae (Group B Streptococcus, GBS)^48,49, a Gram-positive bacterial pathogen that causes meningitis⁵⁰. In these mice we also observed increased ChP S100A9+ leukocytes, CX3CR1+ macrophages, and CD45-high leukocytes (Figure S1I–J; heat-killed GBS: 24 hrs S100A9 pixel/μm² ChP = 24.17 ± 10.85; 72 hrs CX3CR1 pixel/μm² ChP = 31.35 ± 7.44; live GBS: 48 hrs CD45 pixel/μm² ChP = 20.64 ± 8.02) suggesting that these responses were not limited to LPS but rather a general response to common bacterial infections.

ChP displays temporally and spatially distinct immune activation programs

To profile ChP immune activation temporally and spatially, we used longitudinal two-photon in vivo imaging to visualize in real-time leukocyte accumulation at the ChP and subsequent leukocyte entry into the brain (Figure 2A). To visualize peripheral leukocytes, mice harboring the ZsGreen reporter (JAX Ai6) were crossed with mice expressing Cre recombinase under the control of the lysozyme 2 (Lyz2) promoter⁵¹ to generate Lyz2^ZsGreen progenies. While resident ChP macrophages express Lyz2, our whole-mount explant histology confirmed that 24 hours following LPS the vast majority of Lyz2-ZsGreen cells in the ChP co-expressed S100A9 (59.28% ± 13.65%, N=3; Figure S2A), a marker for infiltrating neutrophils and monocytes. Longitudinal imaging of the same ChP subregion before and after LPS exposure captured the full spectrum of leukocyte infiltration from a low level at 8–12 hours following LPS, when occasional cells appeared in the ChP, to a “flood” of infiltrated leukocytes by 24–28 hours, in the ChP and CSF (Figure 2B–D, Videos S1 and S2). We also identified discrete “hot spots” of the ChP-CSF interface where Lyz2+ leukocytes extravasated into the ChP, then burst across the epithelial barrier, streaming into the CSF (Video S2). These data suggest that immune infiltration may have spatial preferences.

Figure 2. The ChP allows rapid leukocyte infiltration into the CSF across its epithelial barrier.

(A) Schematic of the cranial window, head post, and injection cannula that allow live two-photon recording of the ChP in awake mice following ICV injections of LPS. (B) Representative still images extracted from 3 serial in vivo imaging sessions of the same mouse, showing the gradual infiltration of Lyz2+ leukocytes into the ChP over the course of 24 hours following LPS ICV delivery. Dotted lines denote major vasculature. See also Video S1. Scale = 100 μm. (C) Representative still images extracted from in vivo imaging showing high numbers of Lyz2+ infiltrating leukocytes both in the ChP and in the CSF, 24 hours following LPS delivery. See also Video S2. Scale = 100 μm. (D) the amplitude of Lyz2+ signal fluorescence 24 hours following LPS was significantly increased from baseline (** p = 0.0078, N=3). (E) Representative image demonstrates spatial correspondence between the in vivo view and the wholemount explant. Scale = 500 μm. (F-G) Representative images of wholemount explants showing selective expression of E-selectin 24 hours following LPS. Scale = 500 μm; Yellow square indicates typical in vivo field of view. (H) Schematic and 6 representative images demonstrating infiltration of Lyz2-ZsGreen S100A9+ immune cells, injected IC 16 hours following LPS ICV, from peripheral blood into the ChP (examples 1–4) and CSF (examples 5–6). Scale = 50 μm. (I) Schematic depicting strategy to breed mice harboring CCR2^RFP and tamoxifen-inducible Cx3cr1^Zsgreen, which enables simultaneous visualization of infiltrated monocytes and monocytes-derived macrophages (RFP+, ZsGreen-) and resident macrophages (ZsGreen+), in both stromal and epiplexus spaces of the ChP. (J-K) Representative images and counts of fluorescent cells showing significant increases in epiplexus (**** p < 0.0001) and stromal (** p = 0.0016) RFP+ cells, as well as a small, but significant, increase in numbers of of epiplexus ZsGreen+ cells (**** p < 0.0001). Scale = 100 μm. (L-N) Representative images and cell counts showing significant increases in EdU+ CCR2+ monocytes (** p = 0.0028) and EdU+ ZsGreen+ resident macrophages (total: **** p < 0.0001; epiplexus: ** p = 0.0079). Scale = 100 μm. All quantitative data are presented as mean ± SD.

Following in vivo imaging, we prepared whole mounts of the ChP and used its characteristic vasculature to guide post hoc analyses of the imaging locations. The two-photon imaging field of view contained the caudal aspect of the anterior domain of the lateral ventricle ChP (Figure 2E). We observed robust endothelial expression of E-selectin 12–24 hours following LPS in the ChP, including regions within the two-photon imaging field of view (Figure 2F–G; E-selectin pixel/μm² ChP: PBS 24 hrs = non-detectable vs. LPS 24 hrs = 6.28 ± 0.28). E-selectin expression was consistent with a post-capillary venule localization, likely indicating sites of leukocyte extravasation from the blood into the ChP prior to bursting through the epithelial barrier.

Further supporting the concept that the ChP is a crucial hub of leukocyte activity following LPS, we found that peripherally labeled leukocytes entered the ChP and were closely positioned on the basal, stromal, and apical, CSF sides of the epithelium. First, we collected peripheral blood leukocytes from LPS-stimulated Lyz2^ZsGreen mice. Then we transfused these cells into the blood of LPS-stimulated WT mice by intracardiac injection, and subsequently detected Lyz2-ZsGreen cells within ChP stroma (Figure 2H, examples 1–2), on the ChP surface (Figure 2H, examples 3–4), and in the CSF (Figure 2H, examples 5–6). Second, we labeled blood-sourced neutrophils by intracardiac delivery of an anti-Ly6G antibody conjugated to Alexa-594, which appeared within the ChP shortly thereafter, as shown by two-photon in vivo imaging (Figure S2B, Video S3). Third, we captured the positioning of neutrophils along the blood vessels, in the ChP stroma, and then on the surface of epithelial cells with electron microscopy (Figure S2C). Fourth, we applied imaging flow cytometry (ImageStream) to the ChP following LPS ICV to visualize Ly6G/C+ neutrophils/monocytes and epithelial cells (Kiravia dye, “Kirva”) at single-cell resolution following tissue dissociation. Unlike most dissociated cells, many Ly6G/C+ cells remained attached to epithelial cells following tissue digestion (Figure S2D–E; 24 hrs: 21.6 ± 16.7% of total live cells, N =3, and 48 hrs: 5.6 ± 2.1%, N=4), indicating tight physical interactions. Fifth, we performed histological analyses to visualize peripheral leukocyte infiltration through the epithelial bilayers. Ttr^mNeon mice with fluorescently labeled ChP epithelial cells³¹ revealed S100A9+ cells in the ChP stroma. More cells were positioned at the base of the ChP compared to the upper region, and many cells appeared to breach the epithelial layer 24 hours following LPS delivery (Figure S2E). Accumulation of neutrophils in the stroma and on the apical, ChP-CSF surface was accompanied by breaks in tight junctions between epithelial cells (Figure S2F). Similarly, more Ly6G+ cells were observed at the base of the ChP connecting with the brain compared to the upper free margin region of the ChP (Figure S2G). Notably, whole brain sections revealed many Ly6G+ neutrophils in the ChP, with only small numbers of neutrophils observed within the subventricular zone and the septum (Figure S2G, arrows; cortex/ChP Ly6G+ cell ratio = 2.56% ± 1.08%), strongly suggesting that in this model, the ChP is the primary site of immune cell infiltration into the ventricles.

We next examined the infiltration and proliferation of monocytes. Using mice co-expressing Ccr2^RFP and tamoxifen-inducible CX3CR1^ZsGreen, which labels only resident macrophages 5 weeks after tamoxifen treatment (Figure 2I), we found that infiltrated CCR2+ monocytes accumulated in both stromal and epiplexus locations following LPS (Figure 2J–K). ScRNA-seq of CSF cells captured Mki67+ myeloid cells (Figure 1F), and EdU labeling confirmed that many infiltrated CCR2+ monocytes and some resident macrophages were mitotically active (Figure 2L–N). Additionally, Cx3cr1^GFP;Ccr2^RFP dual labeled mice further showed that 74.7% ± 6.0% of EdU+/CCR2+ monocytes were also CX3CR1+, suggesting they were differentiated to macrophages. These data cannot fully distinguish if monocytes divided in the periphery or locally following infiltration into the ventricles. However, collectively, these data demonstrate that in response to ventricular inflammation, peripheral leukocytes can infiltrate the brain’s ventricles via the ChP epithelial barrier.

Following the initial wave of infiltrated leukocytes, we tracked real-time macrophage responses at the ChP 48–72 hours following LPS (Figure 3A–B, Video S4), the time frame when we observed the most robust macrophage changes by histology (Figure 1H). During this time frame, the number of visibly motile ChP macrophages were increased drastically, encasing the entire tissue (Video S4). Cx3cr1^GFP;Ccr2^RFP dual labeled mice revealed increased double-labeled CX3CR1-GFP and CCR2-RFP cells at 72 hours following LPS ICV (Figure 3C–D), suggesting the infiltrated monocytes differentiated into macrophages and that monocyte-derived macrophages accounted for the majority of newly gained ChP macrophages. Consistent with increased CCR2+ monocytes both in the stromal space and on the epiplexus surface (Figure 2K–L) and increased CCR2+/CX3CR1+ macrophages (Figure 3D), we also found increased Iba1+ macrophages throughout the ChP (Figure 3E–G).

Figure 3. The ChP recruits macrophages from peripheral monocytes and CSF macrophages.

(A) In vivo imaging schematic (same as Figure 2A) and representative still in vivo images showing increasing numbers of ChP macrophages over the course of 48 hours following LPS delivery. Also see Video S4. Scale = 100 μm. (B) Plot showing significant increase of CX3CR1+ fluorescence signal over the course of 48 hours following LPS ICV delivery (* p = 0.0385, N=3). (C) Representative images from wholemount explant histology, showing increasing numbers of CCR2+ monocytes and CCR2+/CX3CR1+ macrophages following LPS delivery. Scale = 500 μm (200 μm in C1-C3). (D) Quantitative analysis showing increasing number of cells that are CCR2+/CX3CR1+ in ChP wholemount explants. * p = 0.0231, one-way ANOVA. (E-G) Representative images and quantifications showing increases in epiplexus (F, * p = 0.0218) and total (G, ** p = 0.0084) Iba1+ macrophages in the ChP following LPS ICV delivery. Scale = 100 μm. (H) Representative images compressed from one-hour long in vivo video recordings, showing increased numbers of Cx3cr1+ macrophages traveling through CSF (green arrows) following LPS delivery. Also see Video S4. Scale = 100 μm. (I) Representative images showing CX3CR1+ macrophages traveling from CSF and landing on the ChP. Also see Video S7. Scale = 50 μm. All quantitative data are presented as mean ± SD.

Two-photon imaging revealed that in addition to blood-sourced monocyte-derived macrophages, epiplexus macrophages arrived at the ChP via the CSF. We recorded substantial numbers of CX3CR1+ cells traveling through CSF, lasting approximately 48 to 72 hours following LPS delivery (Figure 3H, Video S4) and consistent with results from CSF scRNA-seq (Figure 1D–F). CSF-macrophages expressed neither microglial signature genes (e.g., Sall1, P2ry12)⁵² nor barrier-associated macrophage genes (e.g., Pf4)⁵³, strongly suggesting that the CSF cells originated either peripherally or arrived from another location in the brain parenchyma. In support of long-range travel of macrophages, we observed dynamic translocation of macrophages from the ventricle walls into the CSF, from the CSF towards the ventricle walls (Video S5), and from the CSF to the ChP (Video S6). Some CSF macrophages landed on the ChP where they appeared to remain as epiplexus macrophages (Figure 3I, Video S7).

Collectively, these data establish a time course of ChP immune infiltration and activation following brain LPS exposure that involves both infiltrated leukocytes and dynamic macrophages. We show that peripheral neutrophils and monocytes infiltrated as first responders across spatially distinct regions of the ChP, which allowed the expansion of the ChP macrophage pool from two distinct sources: (1) peripheral monocyte differentiation and (2) CNS macrophage recruitment from the CSF.

Epithelial cells govern a stepwise inflammatory response at the ChP

To determine the cell types and mechanisms involved in regulating ChP immune infiltration revealed by in vivo two-photon imaging, we performed scRNA-seq of the ChP from LPS-treated mice (Figure 4, Figure S3A–C, Dataset S2). We identified two clusters of neutrophils in the ChP, Retnlg+ and Rpl-high, at 24 hours and 72 hours post-LPS respectively with representing genes along the spectrum of neutrophil maturity⁴⁵. Retnlg+ neutrophils exhibited higher levels of granule genes (Chil3, Lcn2, Wfdc21, Ngp, Mmp9), inflammatory chemokines (Ccl3, Ccl4, Cxcl2), cytokines (Il1b), and mediators (Nfkbia, S100a8, S100a9) (Figure 4A–C), whereas Rpl-high neutrophils had slightly elevated expression of ribosomal genes (Rpl18a, Rps3, Rpl8) and a signature defined by expression of Siglecf and Ptma, which have been previously described as markers of late-stage neutrophils⁵⁴. The number of neutrophils present at 24 hours exceeded those observed at 72 hours (Figure S3D), corroborating findings from CSF scRNA-seq (Figure 1D–F) and imaging studies. Additionally, we observed a population of Ccr2+ Plac8+ monocytes infiltrating into the ChP at 24 hours and Ccr2+ Cx3cr1+ infiltrating macrophages and monocyte-derived dendritic cell (moDC) populations at 72 hours, including cells that were proliferating (Mki67+) (Figure 4A–C, Figure S3D), consistent with EdU labeling of the ChP (Figure 2L–N). In agreement with our previous findings³⁰, cDC1s were present in the ChP at baseline, suggesting that these cells naturally patrol the tissue and may cross-present antigens to promote CD8+ T-cell responses. Accordingly, we observed infiltration of CD8+ T cells 72 hours following ICV LPS (Figure 4A–C, Figure S3D).

Figure 4. scRNA-seq of the inflamed ChP reveals complex and dynamic immune signatures to support immune infiltration.

(A) UMAP embedding of 16,291 cells collected from ChP 24 hours or 72 hours following LPS ICV or 24 hours following PBS ICV for scRNA-seq. Neu = Neutrophil; MΦ = Macrophage; Epi = epithelial cells. (B) UMAP as in (A) colored by hash call assignment, including those without definitive hash identities (”Unassigned”). (C) Violin plot of literature-curated marker genes used to assign cell type and state identity to clusters. (D) CellChat analysis showing cell-to-cell interactions mediated by secreted ligand (outgoing) and receptor (incoming) pairs at its peak strength 24 hours after LPS ICV. (E) CellChat analysis showing chemokine signaling from all cell types (outgoing arrows) to immune cells (incoming arrows).

We observed heightened cell-cell interactions through secretory signaling at 24 hours following LPS (Figure 4D), including extensive chemokine signaling from multiple sources within the ChP (Figure 4E). We first identified a population of resident macrophages (Cx3cr1+ Pf4+ Ccr2-) that displayed an intensely inflammatory transcriptional response including high levels of chemokine expression at 24 hours (Nfkbia, Ccl2, Ccl5, Ccl7, Figure 4A–C, E, Figure S3D) and reverted to baseline at 72 hours, indicating their roles in initiating leukocyte recruitment. In addition, we identified one Chil1+ Icam1+ epithelial state out of 6 clusters of Ttr+ epithelial cells that was induced at 24 hours and produced high levels of chemokines (Ccl2, Cxcl5, Cxcl16, Cxcl1, Cxcl10, Figure 4E, Dataset S2). One out of two clusters of fibroblasts (Pdgfra, Coch, Alpl, Col1a1) was enriched at 24 hours and marked by relatively elevated expression levels of chemokines (Cxcl1, Cxcl16, Ccl2, Ccl7, Figure 4A–C, E, Dataset S2) and acute phase proteins (Saa1, Saa3) (Dataset S2). We further identified TNF signaling from inflamed macrophages and infiltrated leukocytes (Tnf) to epithelial cells (Tnfrsf1a) following LPS ICV (Dataset S2). TNF signaling is a well-characterized mechanism of inflammatory cascade initiation and has been shown to drive the upregulation Icam1^55,56. Consistent with these data, we detected high CSF levels of cytokines and chemokines, including CCL2 and TNFα, 24 hours following LPS exposure, which decreased at 72 hours (non-detectable in mice receiving PBS) (Figure S3E). The ChP also displayed NF-κB signaling in Chil1+ Icam1+ epithelial cells and IFN signaling in Chil1+ Icam1+ epithelial cells (Cxcl10, Ifit3, Isg15, Dataset S2) and endothelial cells associated with LPS at 24 hours (Ifitm3, Irf7, and Isg15, Figure S3F, Dataset S2). These cellular changes are consistent with the hypothesis that ChP epithelial cells and resident macrophages respond directly to the presence of CSF-sourced pathogenic triggers including LPS and in turn, engage ChP endothelial cells and fibroblasts. Indeed, removing resident macrophages with PLX5622 (CSF1/M-CSFR inhibitor) prevented LPS-mediated immune infiltration into the ChP (Figure S3G). Injections of an equal dose of LPS directly into peripheral blood also failed to induce substantial leukocyte infiltration compared to ICV injection (Figure S3H), suggesting that the observed ChP responses were unlikely to be secondary to peripheral effects of CSF LPS entering the blood.

A previously unknown cluster of inflammatory epithelial cells identified by gene expression patterns (Chil1+ Icam1+ epithelial cells, from here on referred to as inf-Epi) appeared 24 hours following LPS ICV. Inf-Epi cells not only secreted chemokines but also exhibited a multitude of interactions with immune cells consistent with their coordinating the stepwise progression of ChP inflammation: GO analysis revealed that, in comparison to Pcp4+ Cntn1+ epithelial cells that were primarily associated with baseline condition, the inf-Epi cells displayed a shifted functional priority from ion transport and transmembrane transport, which are the most common functions of ChP epithelial cells, towards immune recruitment and support (Figure S4A–B). First, inf-Epi, as well as inflammatory fibroblasts, expressed a high level of extracellular matrix remodeling factors (Mmp3, Timp1, Adamts4) complementing the ones expressed by immune cells (Mmp8, Mmp9, Mmp14) (Figure 5A–B). Concurrently, epithelial tight junctions were weakened 24 hours following LPS ICV, as reflected by a loss of Occludin-positive junctions and disrupted organization of Claudin-2 and ZO-1 (Figure 5C), enabling infiltration of leukocytes from ChP to CSF. Using an alternative set of antibodies, we confirmed reduced Occludin/ZO-1 co-localization and an increased linearity index of the cell-cell border as marked by Occludin (defined as the ratio between the total length of the shared cell border and the shortest distance connecting the two ends), both indicators of disrupted tight junctions and barrier integrity (Figure S4C–G). Exposure to a broad MMP inhibitor (GM6001) 6 hours following LPS improved Occludin/ZO-1 colocalization and the linearity index (Figure S4C–G), confirming that MMPs contributed to barrier disruption.

Figure 5. Epithelial cells emerge as key coordinators of choroid plexus immune responses.

(A) Violin plot showing expression of extracellular matrix remodeling proteins by epithelial cells, fibroblasts, and other non-immune cells, as well as neutrophils and macrophages of the ChP. (B) Representative images showing increased MMP3 expression in ChP epithelial cells following LPS ICV. Scale = 50 μm. (C) Representative images of wholemount ChP explants showing disrupted patterns of Occludin, Claudin 2, and ZO-1 staining by LPS. Scale = 50 μm. (D) Cellchat analysis showing robust CSF1/M-CSF signaling from ChP epithelial cells, fibroblasts, endothelial cells towards resident and infiltrated macrophages at 24 hours following LPS ICV. (E) Representative images showing increased CSF1/M-CSF protein expression in the ChP 24 hours following LPS ICV. The expression levels were reduced at 72 hours and undetectable in PBS ChP; Scale = 200 μm. (F) Immunoblot of CSF showing increased level of CSF1/M-CSF in CSF from mice 24 hours following LPS ICV delivery. 9 μl of CSF was loaded for each lane. (G-H) Representative images and quantification showing reduced epiplexus macrophages (Iba1+) in mice treated with anti-CSF1/M-CSF antibody ICV following LPS ICV. Scale = 100 μm. (I) Schematics demonstrating the experimental procedure to collect ZsGreen+ brain resident macrophages within LPS-stimulated donor CSF and transplant to LPS-stimulated recipient mice by ICV. (J-K) Representative images and quantifications showing reduced numbers of both ZsGreen+ donor CSF macrophages (* p = 0.0189) and total Iba1+ epiplexus macrophages (**** p < 0.0001) in mice treated ICV with anti-VCAM1/ICAM1 neutralizing antibodies prior to cell transplant. Scale = 100 μm. All quantitative data are presented as mean ± SD.

Next, we tested the role of colony-stimulating-factor (CSF1/M-CSF) signaling in ChP inflammatory responses. CSF1/M-CSF supports macrophage differentiation and survival^57–59. We found robust Csf1 expression in inf-Epi and some inflamed fibroblasts and endothelial cells at 24 hours, which was reduced by 72 hours (Figure 5D). This induction of CSF1/M-CSF in the ChP corresponded with the infiltration of CCR2+ monocytes (24 hours) and preceded their differentiation to CX3CR1+ CCR2+ macrophages and DCs (72 hours), suggesting an active system at the inflamed ChP that supports the survival of infiltrated myeloid cells and guides their differentiation towards macrophages. We confirmed this finding with histology showing increased CSF1/M-CSF expression in ChP epithelial cells (Figure 5E; pixel/μm² ChP: PBS 24 hrs = 0.12 ± 0.06 vs. LPS 24 hrs = 4.31 ± 0.88, ** p = 0.0024, Welch’s unpaired t-test). We further detected increased CSF1/M-CSF levels in the CSF 24 hours following LPS ICV by immunoblotting, consistent with the secretion of CSF1/M-CSF by ChP epithelial cells into the CSF (Figure 5F; band intensity: PBS 24 hrs = 358.41 ± 480.27 vs. LPS 24 hrs = 24235.08 ± 12121.08, * p = 0.0116, Welch’s unpaired t-test). Neutralizing CSF1/M-CSF by ICV delivery of antibodies 24 hours following LPS reduced epiplexus CX3CR1+/Iba1+ macrophages at 72 hours, consistent with CSF1/M-CSF’s role in supporting macrophage differentiation and survival (Figure 5G–H). The number of CX3CR1+/Iba1+ macrophages that had positive TUNEL staining also showed a moderate increase at 72 hours in the presence of CSF1/M-CSF neutralizing antibody (TUNEL+ macrophages/μm² ChP: IgG = 32.78 ± 20.78 vs. anti-CSF1/M-CSF = 78.35 ± 18.44, * p = 0.0299), but the numbers may not fully reflect the extent of cell death if dead cells were removed via CSF flow. In addition, we found that inf-Epi contributed to ChP macrophage gain (as shown in Figure 3) by retaining epiplexus macrophages with increased expression of adhesion molecules, including ICAM1 and VCAM1, as shown both by scRNA-seq (Figure 4C) and histology (Figure S4H–I), consistent with prior reports^12,60. We performed CSF macrophage transplant studies with neutralizing antibodies against VCAM1 and ICAM1 to confirm their functions (Figure 5I). CSF cells were isolated from tamoxifen-induced CX3CR1^ZsGreen mice 48 hours following LPS ICV and delivered ICV into wild-type mice pretreated with LPS ICV for 24 hours. Before cell transfer, the WT mice received anti-VCAM1 and anti-ICAM1 antibodies or dose-matched rat IgG isotype control by ICV. Brains from WT mice were harvested 48 hours after cell transfer to look for epiplexus ZsGreen+ cells on the ChP. Indeed, we found that mice treated with VCAM1/ICAM1 neutralizing antibodies had ~50% fewer ZsGreen+ cells on their ChP apical surface, as well as ~50% fewer total Iba1+ epiplexus macrophages (Figure 5J–K). Intraventricular injection of antibodies targeting VLA-4α (CD49d) and LFA-1α (CD11a), which are leukocyte-expressed integrin receptors for VCAM1 and ICAM1 (Figure S4J), reduced numbers of epiplexus Iba1+ macrophages to a comparable extent (Figure S4K–L). Collectively, our data demonstrate that ChP epithelial cells, and in particular the inf-Epi population, carry out multiple layers of support to coordinate immune actions and transport at the ChP.

Choroid plexus epiplexus macrophages contributed to barrier healing

How the ChP epithelium returns to its baseline physiological state following the course of inflammation remains unknown. ScRNA-seq indicated that the expression of several epithelial genes (e.g., Htr2c, Kcnj13) and key inflammatory markers (e.g., Icam1, Cxcl1, Mmp3, Csf1) showed trends of recovery towards baseline levels from 24 hrs to 72 hrs following LPS ICV (Figure S5A), suggesting the epithelial cells restore a baseline state following resolution of inflammation.

Because macrophages are well recognized for their roles in wound healing^61,62, we tested and found that epiplexus macrophages contributed to epithelial healing following LPS-induced inflammation. We first investigated their behaviors using in vivo two-photon imaging. We found a range of altered cellular dynamics consistent with phagocytic and pro-healing status. A significant portion of CX3CR1+ macrophages became amoeboid in shape and formed large vacuoles inside their cytoplasm 24 hours after LPS ICV (Figure 6A–B). The macrophages also became highly mobile and formed large aggregates throughout the ChP from 24 hours to 72 hours (Figure 6A–B, Video S8), similar to their reported responses to tissue injury. Further, we specifically labeled resident epiplexus macrophages by crossing Ttr^mNeon mice and mice expressing tamoxifen-inducible Cx3cr1^TdTomato. We found that, following LPS ICV, tdTomato+ epiplexus resident macrophages also became highly mobile and motile with large vacuoles in either cell bodies or at the end of processes (Figure 6C, Video S9), but did not form any large clusters. The differences in observed cellular dynamics suggest both shared and distinguished functions between resident and differentiated macrophages. The number of epiplexus resident macrophages increased over the course of 48 hours (Figure 6D), consistent with lineage tracing data (Figure 2I–K). To determine the possible content of the vacuoles, we used ImageStream and identified a population of macrophages (F4/80+) that contained Ly6G/C+ neutrophils or monocytes (24 hours: 5.3 ± 2.3% of all live cells, N=3, and 48 hours: 0.6 ± 0.4%, N=4, few Ly6G/C+ cells were detected in PBS condition. Figure S5B), which was further confirmed by immunohistology on brain sections (Figure 6F). Consistent with enhanced phagocytotic activity, scRNA-seq revealed high expression of lysosomal marker CD68 in all enriched macrophage and monocyte populations following LPS ICV (Dataset S2). We further identified significant increases in genes associated with myeloid cell phagocytosis following LPS ICV (Figure S5C). Macrophages upregulated calreticulin, which is known to mark aging neutrophils for clearance in the absence of programmed cell death⁶³. This aligns with the lack of a programmed cell death signature in the neutrophils (Dataset S2). These findings suggest that some of the macrophages contributed to clearance of neutrophils.

Figure 6. Macrophages at the ChP aided inflammation resolution and barrier repair.

(A) Schematics depicting morphological changes of ChP macrophages following LPS ICV. (B) Representative images extracted from in vivo recordings showing ChP macrophages at baseline and with large vacuoles and forming clusters following LPS ICV. Also see Video S8. Scale = 50 μm. (C) Breeding scheme for Ttr^mNeon mice to label epithelial cells while expressing Tamoxifen-inducible Cx3cr1^TdTomato to label resident macrophages. The image on the right is a representative image extracted from in vivo imaging, demonstrating morphological changes in epiplexus resident ChP macrophages. Also see Video S11. Scale = 50 μm. (D) Quantification showing increased CX3CR1+ resident epiplexus macrophages from baseline to 72 hours following LPS ICV delivery (** p = 0.0034, N=3). (E) Representative images from ImageStream showing Ly6G+ neutrophils internalized by F4/80+ macrophages. Scale = 10 μm. (F) Representative images showing co-localization of Iba1 (macrophage) and S100A9 (neutrophil) by immunohistochemistry. Scale = 50 μm. (G) Representative 3D image and orthogonal views showing epiplexus and stromal CX3CR1+ macrophages that contain Occludin in ChP explants from mice 72 hours following LPS vs. PBS ICV delivery. (H-J) Representative images and line profile quantification showing progressive repair of ChP epithelial tight junctions. Mice treated with anti-VCAM1/ICAM1 neutralizing antibodies (N=4) had delayed Occludin repair compared to controls (N=3). Statistical significance was calculated by the maximum height of peaks. * p = 0.0320. Scale = 50 μm. All quantitative data are presented as mean ± SD.

Next, we questioned the roles and contributions of newly differentiated epiplexus macrophages in restoring epithelial cell tight junctions following LPS-induced breakdown. Occludin was detected within CX3CR1+ immune cells in both LPS-treated mice and PBS controls, but LPS-treated mice had many Occludin+ epiplexus immune cells (14 ± 2.65 cells in one 421.2 μm x 421.2 μm FOV) while PBS control mice only had a few Occludin+ stromal macrophages and no epiplexus ones were found (Figure 6G, more examples in Figure S5D). This finding suggests that, while stromal macrophages may be involved in barrier maintenance at baseline, epiplexus macrophages might be involved in epithelial barrier repair following LPS ICV, possibly by cleaning up epithelial debris such as the disrupted tight junctions, an established function of macrophages during wound healing^64,65. In additional control experiments, we found that male mice treated with VCAM1/ICAM1 neutralizing antibodies 24 hours following LPS to reduce epiplexus macrophages had delayed recovery of Occludin over 5 days compared to their IgG-treated controls, as reflected by weakened signal peaks between two adjacent cells indicating incomplete barriers (Figure 6H–J). In female mice, the recovery rate varied across all conditions (tested with N=8 in two independent experiments, quantified in Figure S5E). Taken together, our data demonstrate that newly recruited ChP epiplexus macrophages require epithelial cell expression of apical VCAM1/ICAM1 for repairing tight junctions and implicate sexually dimorphic control of this process.

DISCUSSION

The data presented here suggest that the ChP is an immune organ for the brain. Key to this function is a specialized population of ChP epithelial cells, termed inf-Epi, that responded to LPS with patterns of gene expression that support recruitment, adhesion, survival, and differentiation of immune cells. The inf-Epi cells also produced extracellular matrix remodeling enzymes that enabled leukocyte infiltration into the brain and adhesion molecules that facilitated CSF-to-ChP macrophage recruitment for barrier repair following inflammation.

Many features of ChP mediated responses to inflammatory stimuli resemble those of other tissues⁶⁶. Infection or sterile injury in the skin, gut, lung, and heart activates resident macrophages, which together with other cells, stimulates the endothelium to upregulate adhesion molecules that in turn induce the stepwise recruitment of additional cell types. Other similarities include neutrophils being first responders and monocytes being recruited later; the latter differentiate into macrophages that promote tissue remodeling and repair^67–70. A similar series of steps takes place in the brain parenchyma following acute traumatic injury where leukocytes infiltrate through broken blood vessels and specialized microglia rebuild the vessels and repair the blood-brain barrier⁷¹. Finally, the recruitment of macrophages from the CSF to the ventricle-facing surface of ChP resembles the mobilization of macrophages from organ-cavities and appears to serve an analogous function in tissue repair ⁶⁹.

Our data also revealed distinct aspects of the ChP response to inflammation. We highlight the key coordinating role played by a specialized population of inflamed epithelial cells (inf-Epi) through multiple immune cell regulatory mechanisms, including recruiting macrophages from the CSF. Another specialized aspect of the ChP-CSF axis is the use of CSF-distributed factors to coordinate a unified response to infection and injury across brain barriers and compartments¹. Finally, the dual role of the ChP as both a self-regulating brain barrier and a governor of immune cell activity warrants its classification as an immune organ with a distinct and vital function.

Our findings confirm and extend previous work. Cultivating a niche for immune cells is a natural role extension for secretory ChP epithelial cells which have already been shown to do the same for neural stem cells^72,73. It has been long hypothesized that the ChP plays an active role in guiding peripheral immune cell entry though its barrier^10,28,35,74. Ventricle-facing ChP macrophages (epiplexus / Kolmer cells) are poorly understood despite being first described over 100 years ago^10,15. Our data reveal that at least some of these cells are recruited from the CSF. We and others^12,60, have previously shown that adhesion molecules are upregulated on the CSF-facing surface of ChP epithelial cells, but the significance of this unconventional expression pattern was unknown. We now show that these molecules are required for epiplexus cell binding to the ChP. Taken together with data showing tight junction recovery with the aid of epiplexus macrophages, these findings strongly suggest a functional role for epiplexus cells in the last step of the ChP inflammatory response - barrier repair. It is remarkable that the same ChP macrophages that respond rapidly to acute tissue injury also participate in brain barrier repair.

Many questions remain to be answered. First, while we expect that tight junctions regulate barrier integrity in the ChP as they do in the gut, this has yet to be directly tested. Second, the molecular mechanisms underpinning our finding that preventing macrophage attachment to the ChP hindered the recovery of ChP epithelial tight junctions, should be studied further. The antibodies we delivered ICV to neutralize membrane ICAM1/VCAM1 may also have interfered with the function of adhesion molecules expressed by ependymal cells, and/or soluble ICAM1/VCAM1 in the CSF. The physiological significance of soluble adhesion molecules is unknown but reports of increased levels of soluble ICAM1 and VCAM1 in CSF samples from patients diagnosed with schizophrenia⁷⁵, multiple sclerosis⁷⁶, and subarachnoid hemorrhage⁷⁷ warrant further investigation. Third, the origin of the CSF macrophages that move to the ChP following inflammation is uncertain. These cells lacked signature genes of either microglia or border-associated macrophages (BAMs) (e.g., Sall1 and Pf4). Transcriptional changes in microglia and macrophages are highly dynamic, thus CSF macrophages may represent mobilized microglia and BAMs, newly differentiated macrophages from infiltrated monocytes, or a combination of cells recruited from multiple sources. Fourth, our data point to sexually dimorphic regulation of barrier repair. Sexual dimorphism in immune responses including infection rates for viral, bacterial, and parasitic pathogens, as well as long COVID, is well-documented^78–80 and will be important to study in future examinations of the ChP-CSF system. Finally, although we predict that key features of ChP responses to LPS will be generalizable to other insults, the exact cellular and molecular mechanisms may depend on several variables. We found that responses to LPS and GBS were similar, but further study will be needed to assess the full range of inflammatory pathogen types including bacteria, viruses, parasites, and other inflammatory agents in chronic neurologic conditions, including but not limited to schizophrenia, multiple sclerosis, amyotrophic lateral sclerosis, and Alzheimer’s disease. Further, it will be important to understand ChP responses to peripheral as wells as intraventricular stimuli. Peripheral stimuli access the ChP through fenestrated capillaries, traverse stromal cells, and first encounter the basal surface of ChP epithelial cells. The significance of this is difficult to predict as polarization and differential apical/basal receptor expression by ChP epithelial cells is unknown. Nonetheless, our findings suggest that ChP macrophage responses to brain inflammation differ from those to peripheral inflammation, including other models involving LPS^41,81–84. For example, peripheral LPS delivery triggers elongation of stromal ChP macrophages along blood vessels⁴¹, appearing to bolster the physical barrier against systemic challenge. A similar process was recently reported for microglia following blood vessel injury in the brain⁷¹.

Technical Limitations of study

First, our imaging preparation has anatomical limitations. We can reliably capture approximately the top one-third of the ChP which extends from its brain attachment site ventral to the hippocampus and fornix up into the lateral ventricles. Although histological analyses revealed leukocytes accumulating in the stroma at the “base” of the ChP where it connects to the brain parenchyma (Figures S2E and S2G), and macrophages can attach to the ventricular surface with the help of ependymal cell adhesion molecule expression (e.g., VCAM1, Figure S4I) and subsequently may enter the brain parenchyma, we were not able to obtain live imaging data of leukocyte movements beyond the top of the ChP and surrounding ventricles. We noted proliferating myeloid cells in the ventricles, but further analyses will be necessary to track in real-time where cell division occurs. Moving forward, it will be important to incorporate new technologies (e.g., three-photon imaging⁸⁵) to improve access throughout the ventricles and to track the path of cells along ventricle walls and their potential entry into the brain. Ultimately, a toolbox consisting of mouse models, fluorescent reporter lines, live imaging, lineage tracing, cell labeling, and omics approaches will help unravel the dynamics of all cell types at the ChP (including fibroblasts, T cells, and B cells).

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Maria K. Lehtinen, maria.lehtinen@childrens.harvard.edu.

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability Data Availability

1. Sequencing data (Figure 1, Figure 4, Figure S1, and Figure S3) are available in GEO upon publication (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE239318, accession number GSE239318). All other original data are available from the Lead Contact upon request. All biological materials were either directly commercially available or are available upon request.

2. Custom MatLab code generated for this study is provided in supplemental files. All other code is published⁴¹ and available at https://github.com/LehtinenLab/Shipley2020.

3. Any additional information required to reanalyze the data reported in this work paper is available from the Lead Contact upon request

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animal studies

The Boston Children’s Hospital IACUC approved all experiments involving mice in this study. The following genetic strains were obtained from JAX laboratory: Cx3cr1^GFP (JAX 005582), Ccr2^RFP (017586), Cx3cr1^CreER (020940), Ai6 (007906, ZsGreen reporter), Ai14 (007914, TdTomato reporter), Lyz2^Cre (004781). C57BL6 mice were obtained from Charles River Laboratories. C57BL6-Ttr^mNeonGreen was created with Boston Children’s gene manipulation core (referred to as Ttr^mNeon in the manuscript)³¹. Both male and female mice (3–6 months adults) were equally included in the study except for single cell sequencing and CSF cell transplanting assays, where only male mice were used. Animals were housed in a temperature-controlled room on a 12-hr light/12-hr dark cycle and had free access to food and water. All animals were healthy without any previous procedures prior to experiments reported in this study.

Human samples

All human samples were obtained using an IRB-approved protocol at Boston Children’s Hospital. Three specimens from Boston Children’s Hospital archive were included in the study: 18-year-old female, diagnosed with diffusely hyperemic meninges with purulent exudate, consistent with bacterial meningitis; 10-day-old male, diagnosed with acute meningoencephalitis, with gram-negative rods present. Gram negative sepsis (E. coli positive blood and urine cultures); and 2-week-old male, diagnosed with acute bacterial meningitis with focal superficial extension into cerebral cortex (Staphyloccus aureus culture positive). Two non-infectious specimens were used as control cases. All samples were analyzed by immunohistochemistry. All specimens were collected postmortem. Details regarding the ancestry and ethnicity of the patients were not available. Images were acquired using an Olympus brightfield microscope with DP71 camera and cellSens Entry software.

METHOD DETAILS

Intracerebroventricular injection (ICV)

Adult mice were anesthetized with isoflurane from a vaporizer. Following a midline scalp incision to expose the skull, the lateral ventricle was located by stereotaxic coordinates (bregma −0.4 mm, midline +1.0 mm, dura −2.2 mm). A single delivery path was prepared above the lateral ventricle. 5 μl of lipopolysaccharide solution (0.5 mg/ml) was delivered using a Hamilton syringe into the lateral ventricle over the course of 5 min.

The following reagents were delivered by ICV: Lipopolysaccharides (Sigma L3024, 0.5 mg/ml x 5 μl). Dosages of LPS used to induce brain inflammation in mice range from 10 ng to over 20 μg^86–89. We chose 2.5 μg per mouse to induce acute brain inflammation without severe systemic responses. Rat Ultra-LEAF^™ purified anti-mouse CD106 antibody, clone 429 (MVCAM.A) (BioLegend 105728, 1 μg/mouse), rat Ultra-LEAF^™ purified anti-mouse CD54 antibody, clone YN1/1.7.4 (for ICAM1, BioLegend 116133, 1 μg/mouse). Rat purified anti-mouse CD11a antibody (For LFA-1α, BioLegend 101101, 1 μg/mouse), rat purified anti-mouse CD49d antibody (For VLA-4α, BioLegend 103701, 1μg/mouse), rat IgG2b κ control antibody (ThermoFisher Scientific, 17-4031-82, matching dose to other antibodies).

Live and heat-killed S. agalactiae (GBS) were also delivered by ICV. All procedures related to culture and work with pathogenic bacteria were approved by the Committee on Microbiological Safety of Harvard Medical School and conducted under Biosafety Level 2 guidelines. GBS clinical isolate COH1 (serotype III) was grown in Todd-Hewitt broth (THB)(Sigma) at 37° C and growth was monitored by measuring optical density at 600 nm (OD600). To prepare bacteria for intracerebral ventricular injection, GBS was grown to mid-log phase, pelleted by centrifugation, and resuspended in PBS (Sigma). Heat-killed bacteria were prepared by boiling the bacterial suspension for 5 min. THB was inoculated with a portion of the heat-killed suspension and incubated at 37° C for 2 days to confirm lack of bacterial growth.

CSF collection and analysis

CSF was collected by inserting a glass capillary into cisterna magna, and collected CSF was centrifuged at 1000 x g for 10min. at 4° C to remove any tissue debris. The supernatants were used to measure cytokines and chemokines using LEGENDplex^™ Mouse Inflammation Panel (13-plex) (BioLegend 740446). In the case of CSF cell transplant, the samples were centrifuged in the same way, and the pellet was resuspended in sterile PBS for subsequent analysis.

Tissue collection and processing: explants and brain blocks

Animals were anesthetized by ketamine, perfused with ice cold PBS, and then followed by cold 4% paraformaldehyde (PFA). For cryosectioning, brains were dissected, further fixed in 4% PFA at 4° C overnight, and then incubated in 30% sucrose for at least 48 hours, followed by OCT (1 hour on ice) prior to freezing by dry ice and 2-met-butane bath. For wholemount ChP explant, the LV ChPs were dissected and incubated in 4% PFA at room temperature for 5 min., and then immunostained.

Immunostaining

Cryosections and explants were blocked and permeabilized (0.3% Triton-X-100 in PBS; 5% serum), incubated in primary antibodies overnight and secondary antibodies for 2 hours. Sections and explants were counterstained with Hoechst 33342 (Invitrogen H3570, 1:10,000) and mounted using Fluoromount-G (SouthernBiotech).

The following primary antibodies were used: chicken anti-GFP (Abcam ab13970; 1:1000), rabbit anti-RFP (Rockland 600-401-379, 1:500), rat anti-PECAM (BD Pharmingen 550274, 1:100), rat anti-CD45 (Fisher Scientific BDB550539, 1:50), goat anti-S100A9 (R&D Systems AF2065, 1:200), rabbit anti-Iba1 (Wako 019-19741, 1:200), goat Anti-Type IV Collagen-Alexa Fluor^® 488 (Southern Biotech 1340-30, 1:200), PE rat anti-mouse CD62E (for E-selectin, BD Biosciences 553751, 10 μg per mouse), Alexa594 anti-mouse Ly-6G/Ly-6C (Gr-1) antibody (BioLegend 108448, 10 ug per mouse), rat Ultra-LEAF^™ purified anti-mouse CD106 antibody, clone 429 (MVCAM.A) (BioLegend 105728, 1:50), rat Ultra-LEAF^™ purified anti-mouse CD54 antibody, clone YN1/1.7.4 (for ICAM1, BioLegend 116133, 1:50), rat purified anti-mouse CD11a antibody (For LFA-1α, BioLegend 101101, 1:50), rat purified anti-mouse CD49d antibody (For VLA-4α, BioLegend 103701, 1:50), rabbit anti-Occludin (ThermoFisher 71-1500, 1:50), rabbit anti-Occludin-488 (Proteintech CL488-27260, 1:100), rabbit anti-claudin2 (Thermo-Fisher, 51-6100, 1:100), rabbit anti-ZO1 (Thermo-Fisher, 61-7300, 1:100), rabbit anti-ZO1-647 (Cell Signaling Technology, D6L1E, 1:100), goat anti-mouse MCSF antibody (for CSF1/M-CSF, R&D AF416, 1:100), rabbit anti-MMP3 (Abcam, ab52915, 1:100). Secondary antibodies were selected from the Alexa series (Invitrogen, 1:500). Images were acquired using Zeiss LSM880 confocal microscope.

Tissue dissociation for flow cytometry and single cell sequencing

The lateral ventricle (LV) ChP was dissected in cold RPMI media with 10mM HEPES and minced with a pair of fine scissors in a 1.5 ml tube (approximately 200 times until only small pieces remained). LV ChP from 3 adult male mice were pooled into each sample. Tissue pellets were digested in an enzyme cocktail containing collagenase P (0.5 mg/ml, Sigma cat. 11213865001), dispase (0.8 mg/ml, Worthington cat. LS02104), and DNAse1 (250 U/ml, Worthington cat. LK003172) for 30 min. at 37° C with slow head-to-toe rotation. Post-digestion tissues were washed, triturated, and filtered with 70 μm mesh. All pipette tips and tubes were coated by incubating in HBSS containing 2% BSA. For scRNA-seq preparation, to avoid transcriptional changes during the dissociation process, dissection and digestion solutions contained actinomycin (Sigma, cat. A1410, 5 μg/ml), triptolide (Sigma, cat. T3652, 10 μM), and Anisomycin (Sigma, cat. A9789, 27.1 μg/ml)⁹⁰. For CSF analysis, CSF from 5 adult male mice were collected and pooled into each sample. The CSF was treated with ACK lysing buffer (Thermo Fisher A1049201) to remove red blood cells. Cells from CSF were pelleted at 1000 g x 10 min. (4° C) and washed.

Flow cytometry

Single cell suspensions were stained with the following antibodies and analyzed by BD LSRFortessa^™ Cell Analyzer: CD45-APC, Ly6G-PE, CD11b-BV421, F4/80-PECy7, and Ly6C-FITC. Results were analyzed and plotted by FlowJo v10.

Confocal microscopy and image analysis

All confocal images were acquired using Zeiss LSM 880 with Airyscan FAST confocal microscope. Explant images were taken with 2.5 μm z-stacks. All images were processed with FIJI ⁹¹. CX3CR1-CCR2 colocalization quantification (Figure 3D) was performed with MATLAB (2020b). The code is included as a supplemental file (ChP_red_green_Colocalization.m).

Line profile analysis of Occludin

Line profile analysis from FIJI was used to quantify the strength and distribution of Occludin between adjacent cells (approach was modified from⁹²). 3 images from each explant were used for analysis. A total of 20 adjacent cell pairs were selected based on nucleus position. A line of 8 μm was drawn between the two nuclei and registered as ROI. The same line was then placed on the Occludin image. Move the line, if needed, to have the midpoint at the cell-cell border so all line profiles have their peak at 4μm, unless the Occludin signal was absent and cell-cell borders not visible, in which case the line was placed evenly in between the 2 nuclei. The line profile was analyzed to determine the peak signal at the cell border (4 μm distance). The average height of all 20 peaks were used to represent each mouse. Peak height was used for statistical analysis between conditions.

Co-localization of Occludin and ZO-1

Individual cell borders were isolated in FIJI and threshold adjusted for both Occludin and ZO-1 channels. Multiply the two images to yield pixels where Occludin and ZO-1 colocalize. Normalize the total co-localization pixels by the length of the cell border.

Linearity index

Individual cell borders were isolated in FIJI. The total length of the border was measured by manual tracing. The shortest distance was measured by drawing a straight line connecting the two ends of the border. Linearity index was calculated as Total length / straight line length.

Intracardiac injection (IC)

Adult mice were anesthetized using isoflurane vaporizer. The left chest was shaved, and 200 μl antibody solution was delivered by intracardiac (IC) injection. Alexa594 anti-mouse Ly-6G/Ly-6C (Gr-1) antibody (BioLegend 108448, 10 μg per mouse) and PE rat anti-mouse CD62E (for E-selectin, BD Biosciences 553751, 10 μg per mouse) were introduced using this approach for in vivo labeling. Anti-VCAM1 and anti-ICAM1 antibodies were delivered IC to block endothelial expression. LPS (2.5 μg) was delivered IC to compare with ICV.

Headpost, cranial window, and ICV canula placement

Mice used for in vivo two-photon imaging (3–6 months) were outfitted with a headpost, a 3 mm cranial window over the left lateral ventricle, and a contralateral trans-occipital cannula for ICV injections as previously described⁹³. Briefly, mice were outfitted with a metal/titanium head post, and then had a 3 mm diameter craniotomy (enough to snugly fit the cannula) drilled at the stereotactically correct location through the frontal/parietal bone (top of skull). Brain tissue was carefully irrigated and suctioned away to reveal the lateral ventricle, at which point the cranial imaging cannula, a 3 mm diameter cannula (with 2.5mm height) was placed in the resulting hole. One injection cannula was inserted on the contralateral side through a drill hole placed in the superior- and lateral-most border of the suboccipital bone and advanced to a lateral ventricle target. All mice were given at least 2 weeks to recover from the surgery before imaging takes place. A guide cannula was inserted into the injection cannula and held in place with a screwcap to prevent infection and tissue growth into the injection cannula. This guide cannula was removed prior to ICV injections.

Two-photon imaging and image processing

Two-photon microscopy (Olympus FVMPE-RS two-photon microscope; 512 × 512 pixels / frame) was used to record immune cells activities in ChP in vivo from the following mice: Cx3cr1GFP (JAX 005582), Lyz2Cre (004781) crossed with Ai6, Ttr^mNeon crossed with Cx3cr1CreER (020940) and Ai14 reporter line. 200 μm-400 μm Z-stack was acquired with 5 μm stepping size at each field of view over the course of approximately 1 hour. A 25X magnification, 8 mm working distance objective was used.

LPS solution (0.5 mg/ml) was delivered through the injection cannula into the lateral ventricle prior imaging at a rate of 1 μl/min. over 5 min.

Images were registered and processed following the established algorithm (available on GitHub)⁴¹. MATLAB code used to match two-photon in vivo views with post hoc explant images is included in the supplemental files (PostHocMatching.m).

Tamoxifen induction of gene expression

Tamoxifen was dissolved in canola oil at 20 mg/ml. The solution was incubated at 37° C overnight with shaking and stored at 4° C with light protection. Mice were injected intraperitoneally (i.p.) daily for 4 days, 100 μl per day. All mice treated with tamoxifen were analyzed at least 5 weeks after the final treatment.

Blood cell transfusion

Lyz2^ZsGreen mice and WT mice both received LPS ICV. 18 hours later, blood was collected from Lyz2^ZsGreen mice, treated with citric acid to prevent clotting, and incubated with 10x volume of ACK lysis buffer to remove red blood cells. The remaining cells were pelleted, washed, and resuspended in sterile PBS. The cell suspension was injected to WT mice by IC. WT mice were harvested 6 hours after cell transfusion for brain histology.

ImageStream

Freshly dissected LV and 4V ChP were placed in digestion solution (HBSS + collagenase/dispase, Sigma-Aldrich #10269638001) for 30 mins at 37° C with shaking at 600 rpm. Then, DNase1 was added (4 μl for 100 μl digestion solution) and incubated for 15 min. 1 ml DMEM was added to stop enzyme activity, and cells were collected by centrifuging (10 min. at 400g, 4° C). The cells were first stained for outer markers with KIRAVIA 1:100 (BioLegend 405172) + F4/80-PE 1:100 (BioLegend, #157340) in 0.5%BSA/HBSS for 25 min. on ice, and then fixed with Cytofix/Cytoperm (BD bioscience, cat # 554714, 50 μl for sample for 20 min. at room temperature). Next, cells were stained for inner marker with Ly6G-AF594 (BioLegend 108448) 1:100 for 20 min. at room temperature and incubated with Hoechst 1:2000 (Sigma # 23491-45-4) prior to analysis with ImageStream (Flow Cytometry PCMM core facility at BCH and Harvard Medical School).

Electron microscopy

All tissue processing, sectioning, and imaging was carried out at the Harvard Medical School Electron Microscopy Core Facility. The ChP were fixed in 2.5% Glutaraldehyde/2% Paraformaldehyde in 0.4% CaCl2 and 0.1 M sodium cacodylate buffer (pH 7.4). They were then washed in 0.1M cacodylate buffer and postfixed with 1% Osmium tetroxide (OsO4)/1.5% Potassium ferrocyanide (KFeCN6) for one hour, washed in water three times and incubated in 1% aqueous uranyl acetate for one hour. This was followed by two washes in water and subsequent dehydration in grades of alcohol (10 min. each; 50%, 70%, 90%, 2 × 10 min. 100%). Samples were then incubated in propylene oxide for one hour and infiltrated overnight in a 1:1 mixture of propylene oxide and TAAB Epon (Marivac Canada Inc. St. Laurent, Canada). The following day, the samples were embedded in TAAB Epon and polymerized at 60° C for 48 hours. Ultrathin sections (about 80nm) were cut on a Reichert Ultracut-S microtome and placed on copper grids stained with lead citrate. Sections were examined in a JEOL 1200EX Transmission electron microscope or a TecnaiG² Spirit BioTWIN. Images were recorded with an AMT 2k CCD camera.

Single cell sequencing

Single cell suspension from digested ChP or CSF was evaluated under microscope with trypan blue staining for cell viability and cell number. In ChP study and CSF study respectively, 0.5–1 million live cells from each condition were stained with TotalSeqTM B0301, B0302, and B0303 (BioLegend, cat. 155831, 155833, 155835), and pooled at equal ratios. Cells were counted again after hashing and adjusted to 1500 live cells per μl. Approximately 20,000 cells per sample were loaded onto the Chromium controller. Libraries were prepared according to manufacturer protocols (Chromium Next GEM Single Cell 3’ Reagent Kits v3.1 (Dual Index)). Sequencing was performed on a NovaSeq SP at a depth of ~20,000 paired reads/cell.

Single cell data Preprocessing and quality control

Gene expression and feature barcoding data were processed using CellRanger (v3.1.0) with alignment to the mm10 reference genome. To assign ChP cells to different conditions, feature barcodes (TotalSeq) were demultiplexed using the HTODemux() function in Seurat v4 with a quantile threshold of 0.99, yielding 7,987 singlets, 1,876 doublets, and 10,412 unassigned cells. Doublets and low-quality cells with greater than 30% of reads mapping to mitochondrial genes were removed. To assign CSF cells to conditions, feature barcodes were demultiplexed similarly, yielding 11,770 singlets, 1,270 doublet, and 27 unassigned cells. Doublets and cells with greater than 10% of reads mapping to mitochondrial genes were removed.

Cell Clustering and Annotation

Dimensional reduction, cell clustering, and differential gene expression analysis were conducted using Seurat v4.2.1⁹⁴. We filtered cells that were identified as likely doublets through hash demultiplexing then gene expression was scaled and normalized using SCTransform^95,96 with method set to “glmGamPoi”⁹⁷ and dimensionality reduction was conducted using principal component analysis (PCA). For CSF cells, we used 31 PCs for dimensional reduction and clustering at resolution of 0.5. For ChP cells, we selected PCs visually using an elbow plot then used these PCs for cell clustering at high resolution to further identify likely doublets and contaminating erythrocytes based on high levels of co-expression of marker genes from disparate classes of cell types. We iteratively removed these cells and re-clustered until our analysis yielded a dataset of 16,294 cells. ChP cells were assigned cell type identities based on expression of known canonical marker genes, including epithelial (Kcnj13, Wfdc2, and Ecrg4), fibroblast (Dcn, Alpl, and Igfbp4), myofibroblast (Acta2, Myl9, and Tagln), pericyte (Pdgfrb, Cox4i2, and Kcnj8), endothelial (Ly6a, Kdr, and Flt1), myeloid (Ms4a6c, Cx3cr1, and Trem2), neutrophil (Csf3r, Il1b, and S100a8), and lymphocyte (Ptprcap, Trbc2, and Nkg7). For detailed annotation of cell subtypes and states, we subclustered each cell type individually and annotated cells based on their gene expression profiles and known markers. Stacked violin plots were generated using Seurat::VlnPlot() using default parameters. For volcano plots, differential gene expression was performed between the indicated groups using the Wilcoxon rank sum test in Seurat:FindMarkers and data were visualized using EnhancedVolcano v1.11.3⁹⁸. For heatmaps, gene expression was row normalized and visualized using ComplexHeatmap (v2.13.4⁹⁹).

Cell-Cell Communication Analysis using CellChat

We used CellChat (v1.5)¹⁰⁰ to infer cell-cell communication based on the expression of ligand-receptor pairs and downstream signaling mediators from our scRNA-seq data. We loaded in normalized counts and applied standard parameters to assess signaling between cell types. For comparison of signaling between different timepoints, we collapsed cell states into higher-level clusters (e.g. Resident Macrophages and Inflammatory Resident Macrophages were combined into “Resident Macrophages”) and ran CellChat with multiple comparisons. Circlize (v0.4.16)¹⁰¹, ggplot2 (v3.4), and scCustomize (v0.7.0)¹⁰² were used for visualization.

PLX5622 treatment

PLX5622 chow was manufactured by ScottPharma at 1200 ppm, and standard chow provided by Boston Children’s Hospital was used as control. PLX5622 compound was provided by Cayman Chemical (#28927). Mice were placed on PLX5622 diet or control diet for 10 days before LPS ICV delivery and maintained on the same diet for duration of experiment.

CSF cell transplant

CX3CR1^ZsGreen mice were injected ICV with LPS as described. 48 hours following LPS injection, CSF was collected as described. CSF cells were collected by centrifuging the CSF at 1000 x g for 10 min at 4° C and washed once with cold sterile PBS. The cells were resuspended in PBS for transplant at 10 μl per donor mouse (i.e. when 5 mice were used as donors, the CSF cells were resuspended to 50 μl). 5 μl of cell suspension was injected ICV to WT recipient mice that received LPS ICV 24 hours before. The recipient mice were treated by ICV injection of 1μg each VCAM1/ICAM1 antibody cocktail or control rat IgG 10 min. before ICV injection of cell suspension. Recipient mice were harvested 48 hours following cell transplant.

QUANTIFICATION AND STATISTICAL ANALYSIS

Biological replicates (N) were defined as samples from distinct individuals except where it was necessary to pooled samples across multiple individuals to obtain sufficient material (e.g., CSF) to make a measurement. Sample sizes were informed by estimated mean values from preliminary data and previous studies. Data analyses were performed in a blinded manner whenever possible. Statistical analyses were performed using Welch’s unpaired t-test except for scRNA-seq analysis, which is detailed in STAR METHOD above and in figure legends. Data are presented as mean ± standard deviation (SD). Each datapoint represents one biological replicate and was plotted separately in figures. All datapoints were included (no outlier removed). See Results or figure legends for statistical tests and figure panels for sample sizes for each experiment. p values < 0.05 were considered significant (*p <0.05, ** p <0.01, *** p <0.001, **** p <0.0001).

Supplementary Material

1

Data S1. Zip file containing two MATLAB code: (1) PostHocMatching. MATLAB code to match two-photon in vivo field of view and regions of the ChP wholemount explant, related to Figure 2; and (2) ChP_red_green_Colocalization. MATLAB code to quantify ChP myeloid cells that co-express CX3CR1-GFP and CCR2-RFP, related to Figure 3.

2. Figure S1. ChP histology images and CSF scRNAseq QC, related to Figure 1.

(A-B) Additional histological images of human ChP from meningitis patients and controls. Scale = 100 μm. (C-E) QC plots from CSF scRNAseq showing comparable gene detection, cell counts, and cell viability across samples. (F) Cell frequencies and relative proportions by assigned identity. (G) Representative gating strategy to isolate neutrophils (CD45+/CD11b+/Ly6G+), monocytes (CD45+/CD11b+/Ly6G-/Ly6C+/F4/80-), monocytes derived macrophages (CD45+/CD11b+/Ly6G-/Ly6C+/F4/80+), and resident macrophages (CD45+/CD11b+/Ly6G-/Ly6C-/F4/80+) by flow cytometry. (H) Quantification of ChP and CSF immune cells by flow cytometry following LPS ICV. (I-J) Representative images of wholemount ChP explant showing accumulation of neutrophils following heat-killed Group B Strep (GBS) (I) and live GBS ICV (J) ICV delivery. Scale = 500 μm (J) and 100 μm (J’, J”).

3. Figure S2. Peripheral neutrophils cross the ChP epithelial barrier following LPS ICV delivery, related to Figure 2.

(A) Lyz2^ZsGreen mice had many S100A9+ neutrophils in ChP, as revealed by explant staining 24 hours following LPS ICV delivery. Scale = 100 μm. (B) Schematics and images extracted from two-photon in vivo video recordings showing peripherally labeled Ly6G+ neutrophils present and moving in the ChP. Also see Video S3. Scale = 50 μm. (C) Representative pseudo colored electron photomicrographs showing neutrophils attached to a blood vessel (top), within the ChP stroma (middle) and on the apical sides of ChP epithelial cells (bottom), 24 hours following LPS ICV. Scale = 2μm. (D) Examples of neutrophils in close contact with epithelial cells observed by ImageStream. (E) Representative image showing S100A9+ peripheral immune cells accumulating within the ChP stroma (yellow arrows) and penetrating epithelial layers (white arrows). Epithelial cells labeled by endogenous Ttr^mNeon. Scale = 50 μm. (F) Representative H&E images from mice 24 hours following LPS ICV delivery showing presence of neutrophils in both the ChP stromal space and on the CSF-facing apical surface (green arrows), and breakage of epithelial bi-layers (black arrows). Scale = 50 μm. (G) Representative image of brain coronal section showing that neutrophils, marked by Ly6G, predominantly accumulated in the 3V and LV ChP, with small numbers in the septum pellucidum and paraventricular zone 24 hrs following LPS ICV delivery. Scale = 1mm.

4. Figure S3. ChP scRNAseq analysis and validation, related to Figure 4.

(A-C) QC plots from ChP scRNAseq showing comparable gene detection, cell counts, and cell viability across samples. (D) Cell frequencies and relative proportions by assigned identity. (E) CSF cytokine and chemokine levels following LPS ICV delivery by ELISA. (F) Volcano plots showing endothelial cell genes that were differentially expressed 24 hours following LPS vs. PBS. (G) Representative images showing that mice fed PLX5622 chow diet failed to recruit neutrophils and monocytes following LPS ICV delivery. Scale = 200 μm. (H) Representative images showing minimal S100A9+ leukocytes infiltration (purple arrows) into the ChP at 24 hours when LPS was delivered to the blood at the same dose as ICV. Scale = 500 μm.

5. Figure S4. Inflamed epithelial cells coordinate choroid plexus immune responses, related to Figure 5.

(A) GO-terms of differentially expressed genes between PBS PcP4+ Cntn1+ and 24 hours LPS Icam1+ Chil1+ (Inf-Epi) epithelial cells, showing increased immune-related terms and decreased ion transport terms. (B) Volcano plot showing differentially expressed ion transport related genes. (C-G) Representative images and quantification showing changes in Occludin/ZO-1 co-localization and Occludin+ cell border linearity index following LPS ICV and MMP inhibitor GM6001 treatment. Scale = 4 μm (H-I) Representative images showing increased VCAM1 and ICAM1 expression on the ChP apical surface and ventricle wall 24 hours following LPS ICV delivery. Scale = 100 μm. (J) Schematic demonstrating how CSF immune cells may be attracted to ChP epithelial cells through VCAM1 - VLA-4α interactions (similar scenario expected with ICAM1 – LFA-1α interaction). Blocking either side of the interaction was anticipated to prevent CSF immune cell landing on ChP. (K-L) Representative images and quantifications showing reduced numbers of total Iba1+ epiplexus macrophages in mice treated with anti-VLA-4 α /LFA-1α antibodies following LPS ICV delivery (* p = 0.0288). Scale = 100 μm. Data presented as mean ± SD.

7

Video S1. Timecourse montage showing Lyz2^ZsGreen neutrophils (green) entering the ChP over the course of 24 hours, related to Figure 2. Red: dextran-Texas Red. Scale = 100 μm. Timestamps are in units of minutes:seconds.

8

Video S2. Representative video showing Lyz2^ZsGreen neutrophils (green) travel rapidly in CSF and across the ChP 24 hours following LPS ICV, related to Figure 2. Red: dextran-Texas Red. Scale = 100 μm. Timestamps are in units of minutes:seconds.

9

Video S3. Representative video showing neutrophils peripherally labeled by IC Ly6G-568 antibody in the ChP, related to Figure 2. Red: Ly6G-568 (neutrophils); Green: Dextran-FITC. Scale = 50 μm. Timestamps are in units of minutes:seconds.

10

Video S4. Representative montage showing increased number and mobility of CX3CR1^GFP myeloid cells (green) in the ChP, as well as CX3CR1^GFP myeloid cells (green) traveling in the CSF 48 hours following LPS ICV (right), in comparison to baseline (left), related to Figure 3. Red: dextran-Texas Red. Scale = 100 μm. Timestamps are in units of minutes:seconds.

11

Video S5. Montage showing two examples of CX3CR1^GFP myeloid cells (green) traveling from and towards the ventricle wall across multiple z-planes within one imaging stack of Supplemental Video S4, related to Figure 3. Red: dextran-Texas Red. Scale = 100 μm. z-plane numbers are marked on the upper left corner.

12

Video S6. Montage showing three examples of CX3CR1^GFP myeloid cells (green) traveling towards the ChP across multiple z-planes within one imaging stack of Supplemental Video S4, related to Figure 3. Red: dextran-Texas Red. Scale = 100 μm. z-plane numbers are marked on the upper left corner.

13

Video S7. Representative video showing one CX3CR1^GFP myeloid cells (green) traveling through CSF and landing on the ChP, related to Figure 3. Red: dextran-Texas Red. Scale = 100 μm. Timestamps are in units of minutes:seconds.

14

Video S8. Representative video showing CX3CR1^GFP myeloid cells (green) converging into one large cluster. Video was taken 72 hours following LPS ICV, related to Figure 6. Scale = 25 μm. Timestamps are in units of minutes:seconds.

15

Video S9. Representative video showing epiplexus CX3CR1^{CreER-TdTomato} myeloid cells (red) moving along the apical surface of the ChP, as marked by Ttr^mNeon-labeled epithelial cells. Video was taken 48 hours following LPS ICV, related to Figure 6. Scale = 100 μm. Timestamps are in units of minutes:seconds.

16

Table S1. CSF scRNAseq cell type and subtype markers, related to Figure 1.

17

Table S2. ChP scRNAseq cell type and subtype markers, related to Figure 4.

Figure S5. Figure S5. ChP macrophages participate in neutrophil removal and barrier repair, related to Figure 6.

(A) Epithelial gene expression profile showing return towards PBS baseline by 72 hrs following LPS ICV. (B) Example cell images showing distinct staining and identification of neutrophils (Ly6G/C), macrophages (F4/80), and neutrophils engulfed by macrophages in the ChP by ImageStream. (C) Heatmap showing upregulation of phagocytosis related genes in myeloid cells following LPS ICV. (D) Additional representative 3D images and orthogonal views showing epiplexus and stromal CX3CR1+ macrophages that contain Occludin in ChP explants from mice 72 hours following LPS ICV delivery. In PBS controls, Occludin+ CX3CR1+ cells are predominantly stromal. Scale = 20 μm. (E) Line profile quantification in female mice showing less robust repair of ChP epithelial tight junctions +/− neutralizing antibodies against ICAM1/VCAM1.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Alexa594 anti-mouse Ly-6G/Ly6C	BioLegend	Cat# 108448; AB_2563227
CD11b-BV421	BioLegend	Cat# 101235; AB_10897942
CD45-APC	BioLegend	Cat# 103111; AB_312976
Chicken anti-GFP	Abcam	Cat# ab13970; AB_300798
F4/80-PECy7	BioLegend	Cat# 123113; AB_893490
Goat anti-MCSF	R&D Systems	Cat# AF416; AB_355351
Goat anti-S100A9	R&D Systems	Cat# AF2065; AB_2184263
Goat anti-type IV Collagen-Alexa Fluor 488	Southern Biotech	Cat# 1340-30; RRID N/A
Ly6C-FITC	BioLegend	Cat# 128005; AB_1186134
Ly6G-PE	BioLegend	Cat# 127607; AB_1186104
PE rat anti-CD62E	BD Biosciences	Cat# 553751; AB_395031
Rabbit anti-Claudin2	Thermo Fisher	Cat# 51-6100; AB_2533911
Rabbit anti-MMP3	Abcam	Cat# Ab52915; AB_881243
Rabbit anti-Iba1	WAKO	Cat# 019-19741; AB_839504
Rabbit anti-Occludin	Thermo Fisher	Cat# 71-1500; AB_2533977
Rabbit anti-Occludin 488	Proteintech	Cat# CL488-27260; RRID N/A
Rabbit anti-RFP	Rockland Immunochemicals	Cat# 600-401-379; RRID N/A
Rabbit anti-ZO1	Thermo Fisher	Cat# 61-7300; AB_2533938
Rabbit anti-ZO-1 647	Cell Signaling Technology	Cat# D6L1E; RRID N/A
Rat anti-CD11a	BioLegend	Cat# 101101; AB_312774
Rat anti-CD31 / PECAM	BD Biosciences	Cat# 550274; AB_393571
Rat anti-CD45	BD Biosciences	Cat# 550539; AB_2174426
Rat anti-CD49d	BioLegend	Cat# 103701; AB_313042
Rat anti-ICAM-1	BioLegend	Cat# 116133; AB_2813982
Rat anti-VCAM-1	BioLegend	Cat# 105728; AB_2819810
Rat IgG2b k control	ThermoFisher	Cat# 17-4031-82; AB_470176
TotalSeq B0301	BioLegend	Cat# 155831; AB_2814067
TotalSeq B0302	BioLegend	Cat# 155833; AB_2814068
TotalSeq B0303	BioLegend	Cat# 155835; AB_2814069
Bacterial and virus strains
Group B Strep clinical isolate COH1 (serotype III)	Doran et al., 200348 Kuypers et al., 198949	N/A
Biological samples
Human ChP	Boston Children’s Hospital	N/A
Chemicals, peptides, and recombinant proteins
ACK RBC lysis buffer	Thermo Fisher	Cat# A1049201
Actinomycin	Millipore Sigma	Cat# A1410
Anisomycin	Millipore Sigma	Cat# A9789
Clarity™ Western ECL Substrate	BioRad	Cat# 170-5060
Collagenase P	Millipore Sigma	Cat# 11213865001
Collagenase/dispase	Millipore Sigma	Cat# 10269638001
Cytofix/Cytoperm	BD Bioscience	Cat# 554714
Dispase	Worthington	Cat# LS02104
Dimethyl sulfoxide (DMSO)	Thermo Fisher	Cat# D128-500
DNase1	Worthington Biochemical	Cat# LS002007
F4/80-PE	BioLegend	Cat# 157340
Fluoromount-G	Southern Biotech	Cat# 00-4958-02
GM6001 MMP Inhibitor	Tocris Bioscience	Cat# 2983
HEPES	Millipore Sigma	Cat# 83264-100ML-F
KIRAVIA Blue 520™ Streptavidin	BioLegend	Cat# 405172
LEGENDplex™ Mouse Inflammation Panel	BioLegend	Cat# 740446
Lipopolysaccharides	Millipore Sigma	Cat# L3024
Paraformaldehyde	Millipore Sigma	Cat# P6148
Tamoxifen	Thermo Fisher	Cat# 4306311
Triptolide	Millipore Sigma	Cat# T3652
Critical commercial assays
Chromium Next GEM Single Cell 3’ Reagent Kits, Dual Index, V3.1	10x Genomics	Cat# PN-1000269
Chromium Next GEM Chip G Single Cell Kit	10x Genomics	Cat# PN-1000127
3’ Feature Barcode Kit	10x Genomics	Cat# PN-1000262
Dual Index Kit TT Set A	10x Genomics	Cat# PN-1000215
Dual Index Kit NT Set A	10x Genomics	Cat# PN-1000242
Deposited data
ChP scRNAseq	GEO	GSE239318
CSF scRNAseq	GEO	GSE239318
Experimental models: Organisms/strains
Ai14, tdTomato reporter	JAX Laboratory	007914
Ai6, ZsGreen reporter	JAX Laboratory	007906
C57BL6	JAX Laboratory	000664
Ttr mNeonGreen	Fame et al., 2023	N/A
Ccr2 RFP	JAX Laboratory	017586
Cx3cr1 CreER	JAX Laboratory	020940
Cx3cr1 GFP	JAX Laboratory	005582
Ly2 Cre	JAX Laboratory	004781
Software and algorithms
Cell Ranger	10x Genomics	V3.1.0
FIJI/ ImageJ	Schindelin et al. 201291	http://imagej.net/Fiji;https://imagej.nih.gov
FlowJo	BD Biosciences	v10
MATLAB	MathWorks	R2020b
Prism	GraphPad	v7
Two-photon microscopy	Olympus	FVMPE-RS
ZEN Black/Blue	Zeiss	N/A
ImageStream	Cytek	N/A

Highlights:

Choroid plexus (ChP) is hub of immune activity following acute brain inflammation

Live imaging reveals immune cell recruitment to the ChP from the brain and periphery

Specialized ChP epithelial cells coordinate stepwise responses to inflammation

Macrophages participate in ChP barrier repair

INCLUSION AND DIVERSITY

One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in their field of research or within their geographical location. One or more of the authors of this paper self-identifies as a member of the LGBTQIA+ community. One or more of the authors of this paper received support from a program designed to increase minority representation in their field of research.