key: cord-0839633-swgoy1ln
authors: Pellegrini, Laura; Albecka, Anna; Mallery, Donna L.; Kellner, Max J.; Paul, David; Carter, Andrew P.; James, Leo C.; Lancaster, Madeline A.
title: SARS-CoV-2 infects the brain choroid plexus and disrupts the blood-CSF-barrier in human brain organoids
date: 2020-10-13
journal: Cell Stem Cell
DOI: 10.1016/j.stem.2020.10.001
sha: 346a3059dca994fadc823e3d2e6b8356b88b4466
doc_id: 839633
cord_uid: swgoy1ln

Coronavirus disease-19 (COVID-19), caused by the SARS-CoV-2 virus, leads to respiratory symptoms that can be fatal. However, neurological symptoms have also been observed in some patients. The cause of these complications is currently unknown. Here, we use human pluripotent stem cell-derived brain organoids to examine SARS-CoV-2 neurotropism. We find expression of viral receptor ACE2 in mature choroid plexus cells expressing abundant lipoproteins, but not in neurons or other cell types. We challenge organoids with SARS-CoV-2 spike pseudovirus and live virus to demonstrate viral tropism for choroid plexus epithelial cells, but little to no infection of neurons or glia. We find that infected cells are apolipoprotein and ACE2 expressing cells of the choroid plexus epithelial barrier. Finally, we show that infection with SARS-CoV-2 damages the choroid plexus epithelium, leading to leakage across this important barrier that normally prevents entry of pathogens, immune cells and cytokines into cerebrospinal fluid and the brain.

The COVID-19 global pandemic caused by the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has infected more than 32 million people and caused around 990,000 deaths as of late September 2020 (worldometers.info/coronavirus) and it still poses a constant threat to public 40 health systems worldwide. Respiratory symptoms, predominantly associated with the infection, include fever, chest tightness and persistent cough (Wu et al., 2020) .

SARS-CoV-2 shares around 80% sequence similarity with SARS-CoV (Wang et al., 45 2020) and both use their spike (S) glycoprotein, composed of two subunits S1

and S2, to mediate host cell entry Wu et al., 2020) . S1 facilitates viral attachment to a cell surface receptor, angiotensin-converting enzyme 2 (ACE2), and S2 is essential for membrane fusion (Hoffmann et al., 2020b (Hoffmann et al., , 2020a . Co-entry factors such as TMPRSS2 (Hoffmann et al., 2020a) , 50

TMPRSS4 (Zang et al., 2020) and neuropilin1 (Cantuti-Castelvetri et al., 2020; Daly et al., 2020) have also been reported to potentiate infectivity.

Even though respiratory symptoms seem to be the most prominent, increasing evidence from clinical reports indicates a rise in both acute and chronic 55 neurological symptoms, including headache, seizures, confusion, and even psychosis (Varatharaj et al., 2020) , as well as long-term complications such as persistent fatigue (Montalvan et al., 2020) and meningitis/encephalitis (Moriguchi et al., 2020) . These symptoms may indicate viral tropism for brain cells or they may be due to more indirect effects as a result of systemic 60

inflammation. Interestingly, one case reported SARS-CoV-2 in the cerebrospinal fluid (CSF) (Moriguchi et al., 2020) , and additional autopsy findings have demonstrated viral presence within the brains of some patients (Song et al., 2020) . However, the prevalence of central nervous system (CNS) infection is not yet known, and viral presence in the brain and CSF is not a widely reported 65

finding (Helms et al., 2020; Neumann et al., 2020; Schuler et al., 2020) .

The CNS is protected from the rest of the body by two major barriers: the bloodbrain-barrier (BBB) and the blood-CSF-barrier (B-CSF-B). Both barriers prevent entry of blood-borne toxins and pathogens, including viruses. Thus, these 70 barriers represent a major obstacle to SARS-CoV-2 infection of the brain, and it is not yet clear how the virus may enter the brain. Because tropism for nasal epithelial cells has been demonstrated (Sungnak et al., 2020) , viral entry from the nasal cavity through the cribriform plate into the olfactory bulb has been suggested as a potential route of entry to the CNS (Montalvan et al., 2020) . 75 However, post-mortem autopsies suggest this route is unlikely in humans (Schuler et al., 2020) .

The BBB, which separates the systemic blood from the brain parenchyma, is a complex barrier constituted by multiple cells types and mainly formed by the 80 tight junctions between endothelial cells along with pericytes and glial end feet. It therefore represents a complex and highly insulated barrier. The B-CSF-B instead is much simpler, being formed by a single layer of epithelial cells of the choroid plexus (ChP) that separate the fenestrated, leaky capillaries of the stroma from the CSF (Ghersi-Egea et al., 2018; Lehtinen et al., 2011; Lun et al., 85 2015; Strazielle and Ghersi-Egea, 2013) . The stroma is a rich environment that also provides a site of immune surveillance, as well as acting as a gateway for immune cells (Schwerk et al., 2015) . This close interaction with the blood and immune cells makes the ChP epithelium particularly exposed, and previous studies have suggested its invasion may be responsible for the encephalitis 90 caused by lentivirus and the virus Coxsackievirus B3 (CVB3) (Schwerk et al., 2015) . In addition, the ChP itself may contribute to the immune response of the host by secreting proinflammatory cytokines such as IL6 and IL8 into the CSF (Schwerk et al., 2015) . 95

Because human brain tissue is difficult to access, particularly from patients with a contagious pathogen due to safety concerns (Hanley et al., 2020) , 3D in vitro models called cerebral organoids can provide a viable and safe alternative. These tissues can faithfully recapitulate various aspects of human neuronal organization and function (Giandomenico et al., 2019; Lancaster et al., 2017) . 100 Indeed, several published and preliminary reports have used neural organoids to demonstrate some degree of neurotropism (Ramani et al., 2020a; Song et al., 2020) . However, the physiological relevance is still unclear, in particular, the degree of infection relative to more susceptible cell types, as well as the route of entry in to the brain. 105

We recently developed an organoid model to study the ChP (Pellegrini et al., 2020) , which recapitulates the epithelial polarisation of ChP cells and the formation of a tight barrier which separates the surrounding media from the CSF-like fluid secreted by the ChP. To test viral tropism of SARS-CoV-2 in various 110 cells of the CNS, we examined the expression patterns of viral entry factors in cerebral and ChP organoids, and tested for infection with both pseudovirions carrying SARS-CoV-2 spike, and live SARS-CoV-2. We found that particular lipoprotein producing cells of the ChP expressed SARS-CoV-2 entry factors. Comparison with in vivo data supported these findings and suggested these cells 115 represent highly mature ChP epithelial cells. We then tested infection with SARS-CoV-2 spike pseudovirions and live virus, which could productively infect ChP epithelial cells. In contrast, neurons and other CNS cell types were not generally susceptible, except under infection with very large viral quantities. Finally, we observed that the primary effect of the virus was on ChP cells, which disrupted 120 integrity of this key CNS barrier and caused it to become leakier.

To assess whether SARS-CoV-2 entry factors are present in various cell types in brain organoids, we looked at the expression of the receptor ACE2 and the coentry factor TMPRSS2 in different clusters of cells from previously published scRNA-seq data from ChP and telencephalic organoids (Pellegrini et al., 2020) 130 ( Fig. 1A) . Expression of ACE2 and TMPRSS2 was detected predominantly in ChP clusters, but not in the neural progenitor or neuron clusters (Fig. 1A) . To examine whether these results were in agreement with the expression in vivo, we analysed data from the Allen Brain Atlas (Hawrylycz et al., 2012; Miller et al., 2017) reporting expression levels of ACE2 in different human brain regions (Fig.  135  1B) . Among all the different brain regions compared, we found highest levels of ACE2 in the ChP (Fig. 1B) , validating our findings in vitro.

To better characterise the distribution of SARS-CoV-2 entry factors in the different ChP populations, we performed sub-clustering of all the ChP cell 140 populations and we identified four prominent clusters of immature ChP, maturing ChP, hem and ChP stroma, as well as two smaller clusters, one that appeared to be neural crest cells, and one that was enriched in expression of lipoprotein genes, such as APOA1 and APOA2 (Besler et al., 2012; J o u r n a l P r e -p r o o f 2019), suggesting a function in lipid-production and transfer ( Fig. 1C-D, Fig.  145 S1A). We then looked at the main SARS-CoV-2 entry genes in these cell populations and found the highest expression of ACE2 in lipoprotein-expressing ChP cells ( Fig. 1D-E, Fig. S1B ) while some cells of other ChP clusters expressed ACE2 to a lower extent (Fig. 1E, Fig. S1B) . Similarly, the co-entry factors TMPRSS2 and TMPRSS4 were most abundant in ChP cells with high levels of 150 lipoprotein expression (Fig. 1D, Fig. S1B) .

The presence of a cluster of ChP cells expressing abundant lipoproteins was intriguing. In order to examine whether such lipoprotein producing cells were similarly present in vivo, we examined available scRNA-seq datasets from 155 embryonic and adult mouse (Cao et al., 2019; Zeisel et al., 2018) . We extracted those cells specifically expressing ChP markers and performed cluster analysis. This revealed clustering mainly by age, with embryonic and adult cells confined to separate clusters (Fig. 1F) . Examination of apolipoproteins revealed their enrichment in the major adult ChP cluster, which exhibited much higher levels 160 than cells of the embryonic ChP (Fig. 1F, Fig. S1C ). This suggests that the lipoprotein producing cells of ChP organoids may represent a more mature ChP stage.

To explore this timing further, and to examine whether protein levels displayed a 165 similar pattern, we performed immunoblotting for both ACE2 and APOJ, an abundant CSF lipoprotein present in various sized lipoprotein particles (Wang and Eckel, 2014) . We could detect both APOJ and ACE2 in ChP organoid lysates with increasing abundance over time (Fig. 1G) . Furthermore, we examined a previously published proteomic dataset of CSF-like fluid from ChP organoids 170 spanning day 32 to day 146 (Pellegrini et al., 2020) , which revealed increasing secretion of lipoproteins with age (Fig. 1H) . These findings further point to lipoprotein production as a hallmark of maturity, and suggest that ACE2 expression is upregulated in these more mature cells. 175

Next, to validate the observed differences in expression of ACE2 in neuronal and ChP cells, we performed immunostaining of organoids containing cortical ( Fig.  S1D ) or ChP identities (Fig. 1I) . Consistent with scRNA-seq data, the ChP epithelial tissue, marked by the serotonin receptor HTR2C, showed sparse but strong positive signal for ACE2 compared to cortex. Together, these findings 180

indicate that ACE2 is expressed in cells of the ChP, but no specific expression of ACE2 is present in neuronal progenitors or neurons. Expression of TMPRSS2 was also detected by immunostaining in the membrane of ChP epithelial cells from organoids ( Fig. S1E) .

To further explore the expression profile of these identified ACE2 expressing ChP cells, we investigated which genes showed correlated expression pattern with ACE2. Gene Ontology Analysis of genes co-expressed with ACE2 revealed enrichment in GO:MF categories "exogenous protein binding", "viral receptor activity", "toxic substance binding" and "aminopeptidase activity"; GO:BP 190 category "protein-lipid complex subunit organization" and GO:CC category "brush border" (Fig. 1J) . We found enriched expression of DPP4 (dipeptidyl peptidase 4) and ANPEP (alanyl aminopeptidase), which both encode for known receptors of human CoVs (Qi et al., 2020) , and were similarly enriched in the lipoprotein producing subcluster (Fig. S1F) . In particular, DPP4 is the receptor 195

for MERS-CoV, whereas ANPEP is a receptor for human coronavirus 229E, among other human CoVs (Qi et al., 2020) . Interestingly, ChP cells of this subcluster also express lipoproteins that are also required for assembly of hepatitis C virus (HCV), as well as some HCV entry factors such as SCARB1 and CD81 (Aizawa et al., 2015; Grassi et al., 2016; Wrensch et al., 2018) (Fig. S1G) .

Together these data suggest that: 1. More mature lipoprotein-expressing ChP cells express entry factors required for SARS-CoV-2 infection, and 2. Neuronal progenitors and neurons do not specifically express SARS-CoV-2 entry factors.

To examine SARS-CoV-2 neurotropism, we incubated brain organoids with mixed identities including ChP and cortical tissue with SARS-CoV-2 Spike pseudovirions carrying a 19 aminoacid deletion (c19) from the C-terminus, 210

which allows for better expression and integration of the spike into the lentivirus (Giroglou et al., 2004; Ou et al., 2020) . SARS-CoV-2 Spike pseudovirions allow investigation of viral entry without other effects of live CoV such as cell death or viral replication, and because they encoded for GFP we could visualise infected cells only (Giroglou et al., 2004) . As a negative control, pseudovirions lacking 215 viral envelope glycoprotein (Δenv) were used for the infection. VSV-G pseudotyped lentivirus (VSV) with broad viral tropism was used as a positive control ( Fig. S2A-C) . A biochemical assay of the viral reverse transcriptase (RT) was used to assess viral particle production and subsequent titration in ACE2 overexpressing 293T cells ( Fig. S2A-C) . We initially examined infection by direct 220 observation of GFP fluorescence (Fig. S2D) , which clearly showed positive cells, but also some putative autofluorescence in the green channel. Indeed, we found that dead or dying cells displayed quite bright fluorescence in the same channel as GFP (Fig. S2E) . Therefore, to be sure we were observing true GFP signal from infected cells, we added a step of immunostaining with a GFP antibody, which 225 allowed for accurate assignment of infected cells.

We performed three independent infections with SARS-CoV-2 Spike pseudovirions of organoids aged more than 55 days to ensure a mature identity, which revealed viral tropism for ChP epithelial tissue, as indicated by the GFP 230 antibody-positive signal ( Fig. 2A) . Positive control for the infection with VSV lentivirus confirmed a broader tropism (Fig. 2B, Fig. S2F ), whereas no positive signal was detectable in organoids infected with Δenv lentivirus (Fig. 2C, Fig.  S2F ). Quantification of the ratio of GFP-positive cells to total cells showed that SARS-CoV-2 infected around 13% of cells in the ChP epithelium (Fig. 2D) . 235

Interestingly, we found that neuronal regions of organoids with mixed identity did not seem to get infected with the virus (Fig. S2G) , and the only signal we could detect in the green channel was autofluorescence (Fig. 2E, Fig. S2G ). To further investigate the ability of SARS-CoV-2 Spike pseudovirus to infect cortical 240 tissue and neurons, we infected Air-Liquid Interphase Cerebral Organoids (ALI-COs) (Giandomenico et al., 2019) , which are long-term cultures of organoids that lead to improved neuronal maturity and function. ALI-COs infected with SARS-CoV-2 and immunostained for SMI312 to visualise neurons showed no specific GFP-positive signal (Fig. 2F) , when compared to the signal detected in ALI-COs 245 infected with the positive control ( Fig. 2G) . This indicated the lack of signal in upon SARS-CoV-2 Spike infection was not due to defective GFP expression, but rather a lack of viral entry.

These findings support the hypothesis that also in the brain susceptibility to 250 infection by SARS-CoV-2 is governed by expression of ACE2, which is only present on ChP epithelial cells of the organoids. Further supporting this claim, confocal imaging of infected cells immunostained for ACE2 showed positive signal on the membrane of SARS-CoV-2 infected ChP cells (Fig. 2H) .

Recent reports have similarly explored neurotropism of SARS-CoV-2 using brain organoids and neurospheres, but with contradictory findings to those we observed here (Ramani et al., 2020b; Song et al., 2020) . One possible explanation is the use of pseudovirus versus live SARS-CoV-2, with other reports making use of live viral isolates. Therefore, to explore this possibility we turned to infections 260 with a live SARS-CoV-2 clinical isolate (Papa et al., 2020) . We first infected mature ChP organoids with an equivalent amount of virus as we had performed using pseudovirions, and observed highly comparable infection (Fig. 3A, B) . We next tested whether this same amount of virus was capable of infecting neurons and other cortical cells of telencephalic organoids with a mixed identity 265 containing also ChP. We still found infection exclusively of the ChP in these conditions ( Fig. 3C) , as also confirmed by costaining for the ChP marker TTR (Fig. S3A) , suggesting that also live SARS-CoV-2 specifically infects the ChP.

We next sought to test whether there could be any conditions in which neurons 270 or other CNS cell types may be infected with SARS-CoV-2. We did not observe any viral spreading from ChP to nearby cortical regions, even when in very close proximity and left for as long as 4 days (Fig. 3D ). However, one possibility is that the virus preferentially infects ChP resulting in lower amounts of viral particles available to infect the surrounding neurons. We therefore tested 275 infection with live virus on "pure" cortical organoids containing no ChP, but still saw no specific staining for viral protein in neurons or neural progenitors ( Fig.  3E, Fig. S3B ). Finally, we tested whether a combination of longer exposure to sliced organoids with internal neural regions exposed, and much higher viral quantities would lead to infection. We performed infection of ALI-COs with 10 280

times the viral MOI used for ChP infection, and were able to observe very sparse, but specific, neuronal and glial infection after 2 days post-infection ( Fig. 3F, Fig.  S3C ). This suggests that SARS-CoV-2 has a much lower infectivity of human neural cell types compared with ChP epithelium. 285 We next examined expression of ACE2 and markers of lipoprotein production to test whether these may explain the susceptibility of the ChP epithelium. Again, this infection matched ACE2 expression, with infected cells co-staining for the receptor (Fig. 4A ) in the ChP epithelium, whereas stromal cells of the ChP were uninfected (Fig. S4A) . We also observed expression of APOA1 and abundant lipid 290 vesicles in cells of the ChP epithelium infected with SARS-CoV-2 ( Fig. 4B, C) , whereas these cells were absent from cortical neuronal regions (Fig. S4B) .

We next examined whether ChP cells represent a permissive cell type for viral replication. We performed qRT-PCR for viral N1 and observed a significant 295

increase in viral genome copies in organoid supernatant over the course of 4 days (Fig. 4D, S4C ) with the largest increase between 1 and 2 days of infection. This matched staining where we similarly observed larger numbers of cells infected in the ChP at 2 and 4 days post-infection ( Fig. 3C-D) . 300

The ChP epithelium is highly polarized with its basal side facing the vascularized stroma. Thus, potential viral entry in vivo would have to occur from the basal surface. However, we observed strongest ACE2 expression on the apical side, whereas there was much less present on basolateral surfaces (Fig. 4A) . To test whether this low level of ACE2 would nonetheless be sufficient to allow infection 305

of ChP epithelium, we performed infections of intact tissues with clear, fluidfilled cysts, which we have previously demonstrated exhibit the same polarity as in vivo (Pellegrini et al., 2020) . This experiment revealed abundant infection from the basal side ( Fig. 4E) , suggesting that blood-borne virus would indeed be able to infect from the vascular compartment of the ChP. 310

We next examined the impact of SARS-CoV-2 infection on the function of the ChP epithelium, specifically its role as an integral part of the blood-CNS barrier. We observed a striking effect on tight junctions, labelled by Claudin-5, as early as 2 days post-infection (Fig. 4F) , with increasing disruption by 4 days post-infection 315 (Fig. S4D) , suggesting a potential breakdown of barrier integrity. To further test this, we observed ChP organoid morphology over time, which revealed striking damage to organoid integrity in 3 of 5 organoids at 4 days post-infection (and no change in morphology in all 5 mock infected organoids) (Fig. S4E) . A breakdown of barrier integrity would be expected to lead to leakage of the internal CSF-like 320

fluid from within, which should be measurable as a decrease in volume of fluid within the organoid. Indeed, we observed a dramatic decrease in internal fluid in those organoids that also exhibited disrupted morphology (Fig. 4G) , indicating a complete breakdown of the barrier. Furthermore, because CSF has a much lower protein content that serum or media, we measured the protein concentration in 325 the surrounding media and observed a reduction in the media of SARS-CoV-2 infected organoids, further indicating leakage and therefore dilution of media proteins (Fig 4H) . Finally, we performed immunoblot of samples of the surrounding media for the CSF component APOJ, and similarly detected leakage in those organoids with damaged morphology (Fig. 4I) . Taken together, these 330 data suggest that SARS-CoV-2 is capable of infecting ChP epithelial cells leading to a breakdown of barrier integrity.

Given the increasing reports of neurological symptoms associated with COVID-19 (Montalvan et al., 2020; Moriguchi et al., 2020) , understanding viral tropism is of significant interest for the development of better treatments and for prevention of long-term adverse effects. Our findings indicate that SARS-CoV-2 does not readily infect neuronal cells, but rather infects ChP epithelial cells of the 340 brain. This finding is consistent with the high levels of expression of SARS-CoV-2 entry factors such as ACE2 and TMPRSS2 in the ChP in vivo and in vitro, compared to other brain regions , as well as recent data from neural organoids showing minimal neuronal susceptibility to SARS-CoV2 and rather efficient ChP infection leading to transcriptional deregulation and cell 345 death (Jacob et al., 2020) . Furthermore, our findings of susceptibility of lipoprotein-producing cells matches closely the susceptibility seen in other organs such as the intestine (Lamers et al., 2020) .

A likely interpretation of these results is that the neurological symptoms 350 reported in COVID-19 patients are mainly due to an indirect, secondary consequence of viral infection of support cells in the brain, rather than neurons themselves. We show that infection of the ChP epithelium by SARS-CoV-2 leads to disruption of the B-CSF-B, which is in close agreement with recent clinical data demonstrating leakage of blood proteins into CSF in more than 40% of patients 355

tested (Neumann et al., 2020) . While this could then allow entry and spread of the virus into the brain, our results suggest that neural cells are minimally susceptible, even when exposed to high quantities of virus. Furthermore, substantial SARS-CoV-2 within the brain and CSF does not seem to be a widely reported finding (Neumann et al., 2020; Schaller et al., 2020) . Nonetheless, 360

barrier breakdown could allow abnormal entry of immune cells and cytokines leading to harmful neuroinflammation.

A prerequisite for infection of the ChP epithelium, or indeed other parts of the brain, would be presence of the virus in the circulating bloodstream. Given the 365 fact that detection of SARS-CoV-2 in serum is uncommon (Puelles et al., 2020) , and the primarily respiratory symptoms of COVID-19, this may also explain the relatively rare occurrence of severe neurological symptoms in patients. However, once present in the blood, the ChP would represent an easily accessible site within the brain due to its fenestrated and leaky capillaries. This exposure is 370 likely why it is also a site of immune surveillance by macrophages, dendritic cells and monocytes, and infection of ChP epithelial cells can lead to pro-inflammatory cytokine production and recruitment of T-cells that can initiate neural tissue injury (Ransohoff and Engelhardt, 2012) . Furthermore, the B-CSF-B might be a particularly vulnerable barrier to invasion of pathogens and immune cells 375

because the tight junctions between cells have lower electrical resistance compared to the blood-brain-barrier (BBB) (Redzic, 2011; Schwerk et al., 2015) . Thus, the niche surrounding the B-CSF-B makes it a likely target for both viral infection and entry of inflammatory cytokines and immune cells. 380

It is interesting to note that among the neurological disturbances seen in this and previous CoV outbreaks, chronic fatigue and nonrestorative sleep disturbances have been noted (Moldofsky and Patcai, 2011) . Indeed, multiple viruses, such as HCV, HIV and EBV, have been linked to chronic fatigue syndrome (CFS), also known as myalgic encephalomyelitis (ME) (Payne et al., 2013; Rasa et al., 2018; 385 Williams et al., 2019) . While the exact cause of CFS/ME is still obscure, abnormal levels of a number of cytokines, such as IL-10 have been observed in CSF from patients with CFS/ME (Natelson et al., 2005) , suggesting an imbalance in J o u r n a l P r e -p r o o f neuroimmune modulation. In HCV in particular, the most common extrahepatic symptom in patients is chronic fatigue (Poynard et al., 2002) , and key mediators 390 of HCV assembly are apolipoproteins with regulation of lipoprotein metabolism being important for the HCV replication cycle (Aizawa et al., 2015; Grassi et al., 2016) . Lipoproteins are mainly synthesized by hepatocytes and intestinal cells (Aizawa et al., 2015) ; however, we also describe the expression of viral entry factors for HCV and for SARS-CoV-2 in mature lipoprotein-producing cells of the 395

ChP.

Together, these observations suggest the intriguing possibility that the ChP epithelium is particularly susceptible to infection by viruses, and that this susceptibility may lead to breakdown of the B-CSF-B and development of 400 neurological complications, such as CFS/ME. In the case of SARS-CoV-2, this susceptibility seems to be dictated by the expression of ACE2, which is primarily expressed in mature lipoprotein producing ChP epithelium. Thus, our findings suggest that the neurological symptoms are not due to a direct effect on neurons, but rather a consequence of the damage to this important barrier resulting in 405 leakage and proinflammatory changes in the CSF.

One limitation of this study is the lack of examination of post-mortem ChP from 410 COVID-19 patients. Obtaining these samples is notoriously difficult and the collection of brain tissue from COVID-19 patients is complicated by safety concerns with regard to accessing the brain of a patient with a contagious pathogen (Hanley et al., 2020) . A second limitation comes from further safety concerns when working in BSL3 conditions, in particular the inability to use 415

sharps. This limits our ability to selectively extract fluid from within ChP organoids without media contamination, making it impossible to specifically test for virus or proteins within the CSF compartment in a safe manner. Third, the lack of vasculature in the organoid might represent an additional limitation, but since the capillaries surrounding the ChP epithelium are fenestrated and the 420 stromal tissue is cell-sparse, the media compartment of these studies should model the environment surrounding ChP epithelium in vivo. Thus, if virus is present in the blood it would be likely to reach the epithelial cells in a similar fashion to the experiments carried out in this study. Finally, while ChP organoids seem to reach quite an advanced stage of maturity, neurons within telencephalic 425 or cortical organoids are not thought to reach such an advanced stage. Thus, while we show that these fetal-stage neuronal tissues do not appear to be susceptible to SARS-CoV-2 (matching the fact that no congenital brain malformations have so far been reported), it is possible that more mature, adult staged neurons may be susceptible. This would require further investigation, and 430 additional post-mortem analyses would greatly improve our understanding.

The authors would like to thank members of the Lancaster lab for helpful feedback and discussions. We also thank G. Papa for cell lines used in these studies. We also thank the Light Microscopy and BSL3 facilities of the MRC Laboratory of Molecular Biology. The Lancaster lab is supported by the Medical 440

Research Council (MC_UP_1201/9), and the James lab is supported by the Medical Research Council (U105181010). supervised the project, analysed data, and wrote the manuscript.

L.P. and M.A.L. declare that they have filed a patent based on the ChP organoid 455

protocol.

Single cell sequencing data were previously published and accessible on NCBI 460 GEO, accession number GSE150903, and through the UCSC cell browser. Raw data is available upon request. H9 (WA09) cells are available through MTA with WiCell. 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Madeline A. Lancaster (madeline.lancaster@mrc-lmb.cam.ac.uk) . 585

All data are available in the main text or the supplementary materials. H9 cells are available from WiCell under a material transfer agreement with WiCell. Pseudovirions carrying SARS-CoV-2 spike, live SARS-CoV-2 isolate, and ACE2 overexpressing HEK293 cells have been previously published (Papa et al., 2020) . 590

These tools are available upon request. This study did not generate new unique reagents.

The single cell data from (Pellegrini et al., 2020) are available at chporg.cells.ucsc.edu and have been deposited on NCBI GEO (GSE150903). The 595 data that support these findings are available from the Lead Contact, Madeline A. Lancaster (madeline.lancaster@mrc-lmb.cam.ac.uk) upon request.

HEK 293T CRL-3216 cells were purchased from ATCC and authenticated by the 600 supplier. All cells were regularly tested and were mycoplasma free. HEK 293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% FBS, 2 mM Lglutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (Gibco) at 37°C with 5% CO2. Vero cells were purchased from ATCC and maintained in DMEM supplemented with 10% FBS, 100 U/ml penicillin and 100mg/ml 605 streptomycin. Cells tested negative for mycoplasma before virus production and infection experiments. Vero-hACE2 were generated by transducing Vero cells with lentiviral particles expressing hACE2 ORF and cultured in DMEM 10% FCS with 5 μg/ml blasticidin. Vero-hACE2-TMPRSS2 were generated by transducing Vero-hACE2 with lentiviral particles expressing TMPRSS2 ORF and maintained 610

in DMEM 10% FCS with addition of 5 μg/ml blasticidin and 1 μg/ml puromycin. H9 ES cells were obtained from WiCell (WA09) and have been approved for these studies by the UKSCB Steering Committee. Human ES cells were maintained in culture with StemFlex culture media (Gibco A3349401) on growth factor reduced Matrigel coated dishes (Corning). 615

Vectors for viral production were: pCRV-1 for HIV-1 Gag/Pol (Naldini et al., 1996) , and CSGW for GFP expression . Lentiviral packaging plasmid pMDG2, which encodes VSV-G envelope, was used to pseudotype infectious virions (Addgene plasmid # 12259). pCAGGS-Spike ∆c19 was generated from pCAGGS-Spike by digesting the vector backbone with EcoR1 and 620

NheI and subsequently gel purified. Q5 polymerase (NEB) was used to amplify the spike protein fragment using the 3' primer to make a C-terminal 19 amino acid deletion. Primers with a 15bp overlap with the vector backbone were used. After PCR, the fragments were treated with Dpn1 for 1h and subsequently gel purified. The fragment was then inserted into the vector using In-Fusion 625 assembly (Takara Inc.). The plasmid was checked by sequencing.

For the generation of ACE2 over-expressing HEK 293T cells, the human ACE2 ORF was PCR amplified from Addgene plasmid 1786 and C-terminally fused with the porcine teschovirus-1-derived P2A cleavage sequence (ATNFSLLKQAGDVEENPGP) followed by the blasticidin resistance gene. This 630 continuous, single ORF expression cassette was transferred into pLenti6-Dest_Puro by gateway recombination. Lenti-viral particles were generated by cotransfection of HEK 293T cells with pLenti6-Dest_Puro_ACE2-2A-Bla, pCMVR8.74 (Addgene plasmid 22036) and pMD2.G (Addgene plasmid 12259) using PEI. Supernatant containing virus particles was harvested after 48 635 h, 0.45 um filtered, and used to infect naive HEK 293T cells. Transduced cells stably expressing ACE2 were selected with 5 ug/ml blasticidin.

Telencephalic cerebral organoids were prepared from single cell suspension of human ES cells as previously described (Lancaster and Knoblich, 2014 ) and 640 cultured with Stem Cell Technologies Cerebral Organoid kit (catalogue n. 08570, 08571). Briefly, EBs were generated by seeding 2000 cells in a 96-well U-bottom low attachment plate with EB media and 50µM Y-27632 ROCK inhibitor for 3 days. EB media was replaced on day 5 with NI media in the same 96 well plate. At day 7, EBs were embedded in 30µl Matrigel (Corning) using sheets of dimpled 645 parafilm and incubated for 20min at 37°C as previously detailed (Lancaster and Knoblich, 2014) . EBs were then transferred to a 6-well plate in 3ml of Expansion media per well. From day 30, dissolved Matrigel (1:50) was added to the Maturation media. ChP organoids were generated by adding a pulse treatment of BMP4 (20ng/ml) and CHIR (3µM) from day 10 to day 14 of the cerebral organoid 650 protocol, as previously published (Pellegrini et al., 2020) .

Air-Liquid Interface Cerebral Organoid (ALI-CO) culture was performed as previously described (Giandomenico et al., 2019) . Briefly, mature organoids (day 55) were collected, washed in HBSS without Ca2+ and Mg2+ and embedded in 3% low-gelling-temperature agarose (Sigma, A9414) at 40°C in embedding 655 molds (Sigma, E6032). Agarose blocks were then incubated for 10-15min on ice and sectioned using a vibratome (Leica VT100S vibrating microtome) in cold HBSS. Sections (300µm-thick) were collected onto Millicell CM culture inserts (Millipore, PICM0RG50) in 6-well plates and left at 37°C for 1h to equilibrate with serum-supplemented slice culture media (SSSCM). SSSCM was then 660

replaced with serum-free slice culture media (SFSCM). ALI-CO cultures were fed with SFSCM daily.

Replication deficient VSV-G or SARS Cov2 pseudotyped HIV-1 virions were produced in HEK293T cells by transfection with pMDG2 or pCAGGS-Spike ∆c19, pCRV GagPol and CSGW as described previously (Price et al., 2014) . Viral supernatants were filtered through a 0.45μm membrane at 48 hours posttransfection and pelleted through a 20% sucrose cushion for 2hrs at 28K. 670

Pelleted virions were drained and then resuspended in DMEM. Viral stocks were quantified by qRT-PCR as described previously (Vermeire and Verhasselt, 2013) with modifications. In brief, 5 µl of virus stock was mixed with 5 µl lysis buffer (0.25% Triton X-100, 50 mM KCl, 100 mM Tris-HCl (pH 7.4), 40% glycerol, 0.1 µl RNase Inhibitor). After incubation at room temp for 10 mins, 90 μl nuclease-free 675 water was added. 2 µl of lysate was mixed with 5 µl TaqMan Fast Universal PCR Mix, 0.1 µl MS2 RNA, 0.05 µl RNase Inhibitor and 0.5µl MS2 primer/probe mix (7.5µM primers Fwd: AACATGCTCGAGGGCCTT, Rev: GCCTTAGCAGTGCCCTGTCT, 3.7µM probe, TGGGATGCTCCTACATG), in a final volume of 10µl.

For titer determination of pseudotyped viruses, 293T ACE2 cells were plated 680 into 96 well plates at a density of 7.5x10 3 cells per well and allowed to attach overnight. Viral stocks were titrated in triplicate by addition of virus onto cells. Infection was measured through GFP expression measured by visualisation on an Incucyte Live cell imaging system (Sartorius). Infection was enumerated as GFP positive cell area. 685 SARS-CoV-2 virus, named "SARS-CoV-2/human/Liverpool/REMRQ0001/2020", used in this study was isolated by Lance Turtle (University of Liverpool), David Matthews and Andrew Davidson (University of Bristol). SARS-CoV-2 stock was prepared in Vero hACE2-TMPRSS2 cells by infecting monolayer of cells with 0.01 MOI of virus. Virus stock was harvested after 3 days by three freeze-thaw cycles 690

and 5 min 300xg spin to remove cell debris. Virus titers were assessed by plaque assays in Vero ACE2/TMPRSS2 cells. For plaque assays, Vero hAce2-TMPRSS2 cells were seeded on 12-well dishes day prior infection. Next day serial dilutions of the supernatant (-1 to -6) were prepared and used to infect cells for 1h and then overlayed with 0.05% agarose in 2% FBS DMEM. After 3 days monolayers 695

were fixed with 4% formaldehyde and the plaques were revealed with 0.1% toluidine blue staining.

scRNA-seq data was previously published (Pellegrini et al., 2020) and was analysed using the Seurat v3 R package. The already normalized and scaled 700 matrix was visualized by UMAP dimensionality reduction. Subclustering of ChP cell types was performed by first extracting and combining cells of the ChP mature epithelial, immature/hem, and stromal clusters, followed by FindNeighbors, FindClusters and UMAP dimensionality reduction on PCs 1-12 (as determined by ElbowPlot). For expression correlation analysis, pearson 705 correlation was performed between ACE2 expression in all ChP cells and all other genes. The top 20 correlated genes were used for GO term enrichment analysis using gProfiler (Reimand et al., 2011) .

In vivo human data were obtained from the Allen Human Brain Atlas (http://human.brain-map.org/). Z-scaled data were downloaded for the two 710 ACE2 probes available for all samples and brain regions. Mean Z-score was calculated for all samples within each brain region for each probe.

In vivo mouse single cell data was obtained from cells.ucsc.edu. For adult mouse ChP, the expression matrix from the Mouse Nervous System single cell dataset that was previously published was loaded into R and all cells 715

with Ttr values greater than 2 were extracted and taken forward for analysis. For embryonic mouse ChP, cells with Ttr greater than or equal to 1 and no Alb were extracted and taken forward for analysis. These two datasets were then merged and analysed using the Seurat v3 R package. Data were then normalized for read depth across cells, scaled, and underwent variable feature finding using 720

SCTransform. Unbiased clustering was performed by principle component analysis (PCA) using ElbowPlot to guide selection of the number of dimensions (4 principal components), followed by FindNeighbors, FindClusters, and UMAP dimensionality reduction visualization.

Organoid infections were performed by addition of virus to the culture medium at an expected MOI of 0.5 for each virus based on viral titre determined in 293T ACE2 cells (in the case of pseudotyped virus) or plaque assay (in the case of live SARS-CoV-2). To calculate an MOI, counts from previous single cell dissociations (Pellegrini et al., 2020) were used to estimate cell numbers in organoids. Thus, 730 7x10 3 PFU virus was added to ChP tissues with an estimated number of 17,000 cells to achieve MOI 0.5. For mixed identity or pure cortical organoids, with an increased cell number estimated at approximately 35,000 cells, we used 1.4x10 4 PFU virus, or 1.4x10 5 in experiments using 10-fold increased viral quantity. In the case of organoid ChP tissue, ChP epithelium was physically separated from 735 the organoid and broken into three large pieces before infection, and SARS-CoV-2 virus or pseudovirions, negative control, or positive control was added to each of the three pieces. This provided an added internal control so that the experiment and control conditions were tested on tissues obtained from the exact same organoid. 740

RNA extractions from 140µl of inoculated culture medium were performed using the Qiagen QIAmp Viral RNA mini kit and eluted in 80µl of AVE elution buffer. 5µl of extract were used per 20µl RT-qPCR reaction. Reaction mixes were set up according to the manufacturer's instructions with the addition of a CDC-N1 745 SARS-CoV-2 probe/primer mix (IDT 2019-nCoV RUO kit) with a final concentration of 500nM for each primer and 125nM of the hydrolysis probe. One-step RT-qPCR reactions were run on a Viia7s real-time qPCR instrument with the following parameters: 1. 55°C for 10 minutes for reverse transcription 2. 95°C for 1 minute for enzyme activation. 3. 45 cycles of 10 seconds 750 denaturation at 95°C and 30 seconds annealing/extension at 55°C. Fluorescence was recorded at the end of the extension step. To estimate viral copies per reaction, a dilution series of nCoV-2019 N-gene positive control (IDT) was performed in technical duplicates. Sample Cq values were then converted to viral copies per reaction using the standard curve's linear regression model. Values in 755 dot plots represent log10 transformed copies per reaction of individual biological samples over time. For samples where no virus was detected, a pseudo value of 0 was assigned to represent the data point on the graph. Raw data is available upon request.

Monolayers of Vero ACE2 TMPRSS2 cells were infected with ten-fold serial dilutions of supernatant from infected cells as previously described (Papa et al., 2020) to determine viral titers for organoid infections. After 1 hour of infection, cells were overlayed with 0.05% agarose in culture media and incubated for 3 days. Cells were then fixed and stained with 0.1% toludine blue to visualise the 765 plaques.

Organoids were fixed in 4% PFA overnight at 4°C and then moved to 30% sucrose buffer at 4°C for at least 24h. Organoids were then embedded in gelatin and sectioned as previously described (Lancaster and Knoblich, 2014) . After 770 blocking and permeabilisation with 0.25% Triton and 1% donkey serum buffer, sections were incubated overnight with primary antibodies according to their optimized instructions. To mark the nuclei, Dapi (1:1000) was used added with the secondary antibody incubation. Secondary antibodies labelled with Alexa 488, 568 and 647 were applied 1h at room temperature. Incubation with HCS 775

LipidTox (1:1000 in PBS, ThermoFisher, H34477) was carrier for 20min after secondary antibody to visualize lipid droplets. Slides were then washed in PBS and mounted using Prolong Diamond mounting media. All the staining steps for ALI-COs were carried out in permeabilisation buffer and their duration was extended as previously described (Giandomenico et al., 2019) . 780

Images were acquired using a Zeiss LSM 780 confocal microscope (Carl Zeiss) and prepared using Fiji (NIH). The Fiji Huttner plugin was used for the quantification of GFP-positive cells over total cells: 100 Dapi-stained nuclei were counted in each experiment and number of GFP-positive cells over total cells counted was then recorded. 785

For immunoblotting, organoids were snap-frozen in liquid nitrogen and homogenized in RIPA buffer with protease inhibitors (Roche) to produce total protein lysate. Protein samples were prepared using NuPAGE LDS Sample Buffer 4x and DTT 1M and heated at 95°C for 15min. Protein samples and ladder were 790 loaded into a polyacrylamide gel and run for 2h at 100mV. Wet transfer in PDVF membrane (Immobilion) was carried out for 3h or overnight at 4°C. Membranes were blocked in 5% milk in PBS-T and incubated overnight at 4°C with primary antibodies (rabbit anti-ACE2, rabbit anti-Clusterin) and mouse anti-β-actin). After 3 washes in PBS-T, secondary antibodies Alexafluor conjugated were added 795 for 1h at room temperature and membranes were images using a Li-COR Odyssey CLx Infrared Imaging System.

For measurements of internal and external volumes, organoids were carefully handled throughout the experiment to avoid accidental barrier breakage. On day 800 4 post-infection, media was carefully removed and its volume measured. Organoids were then lysed in RIPA buffer and total lysate volume was measured. Excess volume in the media and in the lysates beyond the volumes of media and RIPA added were then calculated, and excess internal volume was reported as a ratio of the total excess volume. 805

For measurements of leakage into the media, Bradford assay was performed using Bio-Rad Protein Assay Kit (Cat#5000001) to measure total protein content of the media. Western blot for APOJ on equal volumes of media samples was performed as described above.

Data were reported as the mean ± standard deviation, except where indicated in the figure legends, using a significant level of p<0.05. The number of replicates is indicated in figure legends and "n" denotes the number of independent experiments. For statistical comparisons, data were analysed by Student's t-test, comparing mock and infected. Cells infected for SARS-CoV-2 were manually 815 counted over total DAPI+ cells using the Cell Counter of ImageJ (NIH). No statistical methods were used to pre-determine sample size.

• SARS-CoV-2 entry factors are expressed in choroid plexus (ChP) cells • More mature lipid-producing ChP cells could be more susceptible to SARS-CoV-2 infection • SARS-CoV-2 productively infects ChP, but not neurons in organoids • SARS-CoV-2 infection damages the ChP epithelium causing leakage of this brain barrier eTOC Blurb

Increasing reports show neurological symptoms in COVID19 patients, but it is still unclear whether SARS-CoV-2 can directly damage neurons. Lancaster and colleagues investigate viral neurotropism using brain organoids and discover that SARS-CoV-2 does not significantly infect neurons but productively infects choroid plexus, leading to damage of this brain barrier. 

Chronic hepatitis C virus infection and lipoprotein metabolism

Molecular mechanisms of 825 vascular effects of High-density lipoprotein: Alterations in cardiovascular disease

Neuropilin-1 facilitates SARS-CoV-2 cell entry and provides a possible pathway into the central nervous 830 system

The single-cell transcriptional landscape of mammalian organogenesis

The spatial and celltype distribution of SARS-CoV-2 receptor ACE2 in human and mouse brain

Neuropilin-840 1 is a host factor for SARS-CoV-2 infection

Molecular anatomy and functions of the choroidal bloodcerebrospinal fluid barrier in health and disease

Cerebral organoids at the air-liquid interface generate diverse nerve tracts with functional output

Retroviral Vectors Pseudotyped with Severe Acute Respiratory Syndrome Coronavirus S Protein

Hepatitis C virus relies on lipoproteins for its life cycle

Autopsy in suspected COVID-19 cases

An anatomically comprehensive atlas of the adult human brain transcriptome

Neurologic Features in Severe SARS-CoV-2 Infection

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

Multibasic Cleavage Site in the Spike Protein of SARS-CoV-2 Is Essential for Infection of Human Lung 870 Cells

Human Pluripotent Stem Cell-Derived Neural Cells and Brain Organoids Reveal SARS-CoV-2 Neurotropism Predominates in Choroid Plexus Epithelium

SARS-CoV-2 productively infects human gut enterocytes

Generation of cerebral organoids from human pluripotent stem cells

Guided self-organization and cortical plate formation in human brain organoids

The Cerebrospinal Fluid Provides a

Proliferative Niche for Neural Progenitor Cells

Development and functions of the choroid plexus-cerebrospinal fluid system

Neuropathological 890 and transcriptomic characteristics of the aged brain

Chronic widespread musculoskeletal pain, fatigue, depression and disordered sleep in chronic post-SARS syndrome; a casecontrolled study

Neurological 895 manifestations of COVID-19 and other coronavirus infections: A systematic review

A first case of meningitis/encephalitis associated with SARS-Coronavirus-2

In Vivo Gene Delivery and Stable Transduction of Nondividing Cells by a Lentiviral Vector

Spinal Fluid Abnormalities in Patients with Chronic Fatigue Syndrome

Cerebrospinal fluid findings in COVID-19 patients with neurological symptoms

Characterization of spike glycoprotein of SARS-CoV-2 on virus entry and its immune cross-reactivity with SARS-CoV

Furin cleavage of SARS-CoV-2 Spike promotes but is not essential for infection and cell-cell fusion

HIV-associated fatigue in the era of highly 920 active antiretroviral therapy: Novel biological mechanisms

Human CNS barrier-forming organoids with cerebrospinal fluid production

Fatigue in patients with chronic hepatitis C

Pharmacologic Ligands Share an Essential Interface in HIV-1 Capsid That Is Lost upon Disassembly

Multiorgan and Renal Tropism of SARS-CoV-2

Single cell RNA sequencing of 13 human tissues identify cell types and receptors of human coronaviruses

SARS-CoV-2 targets neurons of 3D human brain organoids

SARS-CoV-2 targets cortical neurons of 3D human brain organoids and shows 945 neurodegeneration-like effects

The anatomical and cellular basis of immune surveillance in the central nervous system

Chronic viral infections 950 in myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS)

Molecular biology of the blood-brain and the bloodcerebrospinal fluid barriers: Similarities and differences

G:Profiler -A web server for functional 955 interpretation of gene lists

Postmortem Examination of Patients with COVID-19

Age-determined expression of priming protease TMPRSS2 and localization of SARS-CoV-2 infection in the lung epithelium

The choroid plexus -a multi-role player during infectious diseases of the CNS

Neuroinvasive potential of SARS-CoV-2 revealed in a human brain organoid model

Physiology of blood-brain interfaces in 970 relation to brain disposition of small compounds and macromolecules

Comprehensive Integration of Single-Cell Data

SARS-CoV-2 entry factors are highly expressed in nasal epithelial cells together with innate immune genes

Neurological and neuropsychiatric complications of COVID-19 in 153 patients: a UK-wide surveillance study

Quantification of Retro-and Lentiviral 985 Reverse Transcriptase Activity by Real-time PCR

What are lipoproteins doing in the brainα

The genetic sequence, origin, and diagnosis of SARS-CoV-2

Apolipoprotein A-IV involves in glucose and lipid metabolism of rat

Epstein-Barr virus dUTPase induces neuroinflammatory mediators: Implications for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome

Hepatitis C virus (HCV) -Apolipoprotein interactions and immune evasion and their impact on HCV vaccine design

A new coronavirus associated with human respiratory disease in China

TMPRSS2 and TMPRSS4 promote SARS-CoV-2 infection of human small intestinal enterocytes

Molecular Architecture of the Mouse Nervous System

APOBEC3G Incorporation into Human Immunodeficiency Virus Type 1 Particles

APOBEC3G Incorporation into Human Immunodeficiency Virus Type 1 Particles