Caveolin-1 mediates blood-brain barrier permeability, neuroinflammation, and cognitive impairment in SARS-CoV-2 infection

Blood-brain barrier (BBB) permeability can cause neuroinflammation and cognitive impairment. Caveolin-1 (Cav-1) critically regulates BBB permeability, but its influence on the BBB and consequent neurological outcomes in respiratory viral infections is unknown. We used Cav-1-deficient mice with genetically encoded fluorescent endothelial tight junctions to determine how Cav-1 influences BBB permeability, neuroinflammation, and cognitive impairment following respiratory infection with mouse adapted (MA10) SARS-CoV-2 as a model for COVID-19. We found that SARS-CoV-2 infection increased brain endothelial Cav-1 and increased transcellular BBB permeability to albumin, decreased paracellular BBB Claudin-5 tight junctions, and caused T lymphocyte infiltration in the hippocampus, a region important for learning and memory. Concordantly, we observed learning and memory deficits in SARS-CoV-2 infected mice. Importantly, genetic deficiency in Cav-1 attenuated transcellular BBB permeability and paracellular BBB tight junction losses, T lymphocyte infiltration, and gliosis induced by SARS-CoV-2 infection. Moreover, Cav-1 KO mice were protected from the learning and memory deficits caused by SARS-CoV-2 infection. These results establish the contribution of Cav-1 to BBB permeability and behavioral dysfunction induced by SARS-CoV-2 neuroinflammation.


Introduction
Brain endothelial cells regulate the permeability of the cere-brovasculature, referred to as the blood brain barrier (BBB).BBB destabilization causes extravasation of blood proteins and immune cells into the CNS, leading to neuroinflammation, pruning of synapses and neurons, and behavioral changes (Merlini et al., 2019).BBB permeability in the hippocampus particularly contributes to cognitive impairment, consistent with the important role of the hippocampus in cognition (Garber et al., 2019).Disorders of cognition are frequent in COVID-19 (Spudich and Nath, 2022), and BBB permeability and neuroinflammation are pronounced in COVID-19 decedents and animal models (Fernández-Castañeda et al., 2022;Krasemann et al., 2022;Lee et al., 2021;Lee et al., 2022;Matschke et al., 2020;Soung et al., 2022;Thakur et al., 2021;Zhang et al., 2021).Indeed, the hippocampus is a target of inflammation and neurodegeneration in COVID-19 patients (Bayat et al., 2022;Nouraeinejad, 2023;Soung et al., 2022); hippocampal atrophy correlates to cognitive decline after COVID-19 (Douaud et al., 2022).Emerging evidence links infiltration of blood proteins and T cells into the CNS to neuroinflammatory processes in COVID-19 (Heming et al., 2021;Lee et al., 2021;Schwabenland et al., 2021;Soung et al., 2022;Vanderheiden and Klein, 2022).However, the contribution of BBB permeability to neuroinflammation and cognitive impairment in COVID-19 is incompletely understood.We designed the present study to address how the effect of SARS-CoV-2 infection on specific aspects of the BBB would influence cognition.
Thus, the goal of the present study was to evaluate the extent to which Cav-1 contributes to cognitive impairment by promoting BBB permeability in a COVID-19 mouse model.For this, we deployed a mouse adapted strain of SARS-CoV-2 (MA10) that has been generated by engineering the spike protein to bind to the murine homolog of the viral entry receptor, ACE2, and sequential passages of the mouse-adapted strain through mice (Leist et al., 2020).Intranasal inoculation with SARS-CoV-2 MA10 infects ACE2-expressing alveolar epithelium and alveolar type II (AT2) cells within the lung, yielding acute respiratory infection (Leist et al., 2020).Recently, MA10 has emerged as a promising tool to study mechanisms of SARS-CoV-2 neuropathogenesis because it recapitulates features of COVID-19 neuroinflammation, with greater severity in the aged (Amruta et al., 2022).We found that respiratory infection with SARS-CoV-2 MA10 upregulated Cav-1 and decreased Claudin-5 in brain endothelial cells.This was accompanied by heightened permeability to intravenously injected albumin, T cell infiltration, neuroinflammation, and learning/memory deficits in infected mice.Importantly, genetic deficiency in Cav-1 offered protection from SARS-CoV-2 induced BBB permeability, neuroinflammation and memory deficits.These data indicate that Cav-1-mediated BBB permeability is increased during acute SARS-CoV-2 infection and may contribute to neuropathology and cognitive impairment in COVID-19.
We next tested if Cav-1 is altered at the BBB during SARS-CoV-2 infection.At 4 DPI, mice were euthanized and Cav-1 expression in brain microvessels was assessed by immunostaining (Fig. 1A-C).We found that SARS-CoV-2 infection significantly increased Cav-1 positive area in the hippocampus, with a vascular distribution (Fig. 1C).Similarly, Cav-1 expression was high in the olfactory bulb, cortex, and brainstem in SARS-CoV-2 infected mice (Supplementary Fig. 3).We confirmed these results by quantifying Cav-1 by flow cytometry on endothelial cells acutely isolated from whole mouse brain.The percentage of brain endothelial cells with high Cav-1 expression was increased by SARS-CoV-2 infection (Fig. 1D-G).These data indicate that infection with SARS-CoV-2 upregulates Cav-1 in brain endothelial cells.
A consequence of BBB permeability to macromolecules and immune cells can include inflammatory changes in glial cells.We next investigated expression of GFAP in SARS-CoV-2 infection.We observed that GFAP immunoreactive area was significantly upregulated in Cav-1 +/+ mice infected with SARS-CoV-2 but not in Cav-1 −/− mice infected with SARS-CoV-2 (Fig. 5A-E).This suggests that in SARS-CoV-2 infection, upregulation of cerebrovascular Cav-1 and resulting BBB permeability contribute to reactive changes in astrocytes.
Previous studies indicate that brain endothelial cells undergo changes in SARS-CoV-2 infection including cell death (Soung et al., 2022).We investigated endothelial cell apoptosis by immunostaining for Caspase 3. We observed that SARS-CoV-2 induced brain endothelial cell apoptosis to a similar extent in Cav-1 +/+ and in Cav-1 −/− mice (Supplementary Fig. 5A-E).

Cav-1 promotes neurological signs of disease in SARS-CoV-2 infection
Neurological impairment can accompany acute SARS-CoV-2 infection (Spudich and Nath, 2022).The extent to which BBB inflammation influences neurological deficits in COVD-19 is unclear.We therefore tested whether Cav-1 deficiency offers protection from SARS-CoV-2 induced cognitive impairment.We utilized a novel object recognition task to measure learning and memory related to hippocampal function.Cav-1 −/− mice have some age-dependent neurobehavioral abnormalities, including spatial learning deficit and center avoidance (Gioiosa et al., 2008;Trushina et al., 2006).Nonetheless, in our study, mock infected Cav-1 −/− mice had similar novel object recognition as did mock infected Cav-1 +/+ mice (Fig. 6A).In Cav-1 +/+ mice, SARS-CoV-2 respiratory infection significantly impaired the ability to discriminate between known and unknown objects (Fig. 6A).Importantly, Cav-1 deficiency mitigated this effect of SARS-CoV-2 on learning and memory (Fig. 6A).As expected, no object preferences were observed during the training task in either genotype (Fig. 6B).Overall, our data suggest that Cav-1 upregulation on BECs promotes neuroinflammation and neurological deficits in COVID-19 by modifying BBB function.
We observed that brain endothelial cell Cav-1 is upregulated by SARS-CoV-2 respiratory infection.Our observation is consistent with recent reports of increased Cav-1 in the forebrain of COVID-19 decedents (Green et al., 2022;Premkumar and Sajitha, 2023).In our data, the upregulation of Cav-1 in the cerebrovasculature induced by SARS-CoV-2 was linked to worse BBB permeability, T-lymphocyte infiltration, gliosis, and cognitive impairment, whereas these signs of disease were attenuated in SARS-CoV-2 infected mice genetically deficient in Cav-1.These data suggest a potential role for Cav-1 in BBB permeability in COVID-19 neuroinflammation.
Cav-1 contributes to transcellular and paracellular BBB permeability.Cav-1 functions in endocytosis and transcytosis of proteins and cells from the luminal to the abluminal endothelial cell surface (Jones and Minshall, 2020;Ohi and Kenworthy, 2022;Parton, 2018;Pol et al., 2020;Zimnicka et al., 2016).In development, transcellular BBB permeability is suppressed by pericyte and astrocyte regulation of Cav-1 (Andreone et al., 2017;Armulik et al., 2010;Ayloo et al., 2022;Chow and Gu, 2017;Daneman et al., 2010;Guérit et al., 2020), whereas caveolar transcytosis can also be induced in the mature BBB by endothelial autonomous mechanisms (Chang et al., 2022;Liebner et al., 2018;Pandit et al., 2020;Villaseñor et al., 2017).This transcellular activity is heightened by inflammation and advanced age (Carman and Springer, 2004;Hu et al., 2008;Marchiando et al., 2010;Millan et al., 2006;Yang et al., 2020).In our data, SARS-CoV-2 infection induced transcellular BBB permeability to albumin by upregulated brain endothelial Cav-1.Furthermore, Cav-1 contributes to the membrane removal, recycling, and degradation of tight junction molecules such as Claudin-5, especially in the context of inflammatory cytokines (Li et al., 2015;Liu et al., 2012;Marchiando et al., 2010;Stamatovic et al., 2009).Because tight junctions are a critical component of paracellular BBB function, heightened Cav-1 activity indirectly promotes paracellular BBB permeability.In our data, SARS-CoV-2 infection decreased cerebrovascular tight junction coverage in a Cav-1 dependent way, which we posit is due to caveolar endocytosis and degradation of BBB tight junctions.Future studies could intravitally probe rates of caveolar endocytosis of BBB junctional proteins and functionally probe paracellular BBB permeability using small and large molecular weight fluorescent tracers in infected mice and further define the downstream mechanisms by which Cav-1 contributes to loss of tight junctions in this model.Our findings implicate Cav-1 as a regulator of multiple aspects of BBB leakage in SARS-CoV-2 infection.
No animal model can fully recapitulate complex human diseases.Advantages of the SARS-CoV-2 MA10 infectious model include that because the virus binds to mouse ACE2, it results in infection of those cells that have endogenous ACE2 expression.In mice, like in humans, expression of ACE2 is robust in the alveolar epithelium and alveolar type II (AT2) cells of the respiratory tract.Consequently, AT2 cells are the predominant population of infected cells in the respiratory tract in mice infected with MA10 and in humans infected with the SARS-CoV-2 ancestral (WA-1) strain.However, the cellular patterns of ACE2 expression and MA10 infection in mice may not completely match infection by SARS-CoV-2 in humans; indeed the cellular targets of infection by SARS-CoV-2 in humans may differ based upon viral strain and variant.As such, caution is warranted in generalizing our findings to the human condition.
Because our goal in this study was to define the contribution of the BBB protein Cav-1 to neurobehavior, we focused on the hippocampus as a relevant neuroanatomic structure with readily measurable behavioral outcomes.Nonetheless, we also observed SARS-CoV-2 induced upregulation of vascular Cav-1 in other neuroanatomic regions notable for inflammation in COVID-19, including the olfactory bulb, cortex, and brainstem, and corroborated this finding in endothelial cells isolated from whole brain.Thus, SARS-CoV-2 induced upregulation of Cav-1 in multiple brain regions.Our data raise the key question of whether Cav-1 mediated changes to the BBB in other brain regions impacted by SARS-CoV-2 infection also influence neurobehavioral outcomes of disease.Larger animal models may be better suited for testing the neuroanatomic and mechanistic basis of complex neurobehavioral deficits reflecting neural processes superseding the hippocampus and difficult to recapitulate in mice, such as brain fog, attention deficit, and affective change.
Indeed, although our data supports a central role for Cav-1 in neuroinflammation and cognitive impairment, it is important to note caveats.Importantly, no differences were noted in viral RNA in the lung, systemic hypoxia, or weight loss between Cav-1 +/+ and Cav-1 −/− mice with SARS-CoV-2 infection.Nonetheless, variability was observed within groups in the amount of viral RNA in the lung; variability in the severity of the pulmonary infection, as well as natural variability found in aged mice, may have contributed variability to the experimental results.
In this study, we did not identify the proximal cause of cerebrovascular Cav-1 upregulation in SARS-CoV-2 infection.There are a number potential mechanisms.Brain endothelial cells are not the primary cellular targets of SARS-CoV-2 infection in the respiratory inoculation model.Nonetheless, direct infection of brain EC could have led to dysregulation of Cav-1.However, data are limited to support that idea, especially as Cav-1 is not reported to be upregulated in brain endothelial cells infected with SARS-CoV-2 in vitro (Motta et al., 2023;Yang et al., 2022).Nonetheless, SARS-CoV-2 infection is closely associated with production of numerous circulating proinflammatory mediators including viral proteins, cytokines, and activated leukocytes, all of which are anticipated to directly and indirectly impinge upon brain EC homeostasis (Vanderheiden and Klein, 2022).Additional studies will be required to elucidate the molecular mediators between respiratory infection and brain EC dysfunction.

Likewise, we have not yet fully mapped the mediators between elevated cerebrovascular
Cav-1 and impaired neurologic function.For example, we have not defined the contribution of Cav-1 to neuronophagia, synaptophagia, or impaired neurogenesis in SARS-CoV-2 infection.Nonetheless, correlations between features of BBB permeability, neuroinflammation, and cognitive impairment are consistent with a model in which Cav-1 mediated BBB permeability leads to neuronal dysfunction.
Because SARS-CoV-2 is a BSL3 pathogen, we were limited in the kinds of neurobehavioral tests we could conduct to assess neuronal function, because of the practical limitations to conducting behavior tests within the confines of the biosafety cabinet in the BSL3 suite.
This study exclusively examined acute infection, so we cannot draw conclusions regarding neurological post-acute sequelae of .Interestingly, however, neuroPASC correlates to serum markers of BBB leakage (Bonetto et al., 2022;Hanson et al., 2022).These observations suggest the value of future mechanistic studies into BBB Cav-1 dysregulation in neuroPASC.

Mice
The study design is depicted in Supplementary Fig. 1.All animal studies were approved by the UIC Animal Care and Use Committee .Wild-type C57Bl/6 and Cav-1 KO (Jackson Laboratory 000664, 004585, respectively) mice were purchased from the Jackson Laboratory and backcrossed 9 generations.eGFP:Cldn5 transgenic mice express Claudin-5 (Cldn5) labeled with enhanced green fluorescent protein to facilitate visualization of BBB tight junctions (Knowland et al., 2014;Lutz et al., 2017).eGFP:Cldn5 Tg/− Cav1 −/− mice were generated by breeding these lines.All the mice used in this study were male, born in house, and maintained in a specific pathogen free vivarium suite.Mice were transferred to Animal Biosafety Level 3 facilities at least 2 days prior to inoculation.Mice were maintained on standard light-dark cycles with ad libitum food and water in micro-isolation cages.Mice were assigned numeric codes which were used to track the samples in a blinded fashion during in vivo and post-mortem processing, imaging, and analysis.

SARS-CoV-2 inoculation
Mouse adapted SARS-CoV-2 MA10 (Leist et al., 2020) was propagated and titered on Vero-E6 cells expressing ACE2 and TMPRSS2 (ATCC, CRL1586).MA10 was generated by engineering the spike protein to bind to the murine homolog of the viral entry receptor, ACE2, and 10 sequential passages through mice (Leist et al., 2020).Mouse adapted SARS-CoV-2 (MA10) has been previously characterized (Amruta et al., 2022;Leist et al., 2020).Intranasal inoculation with SARS-CoV-2 MA10 infects ACE2-expressing alveolar epithelium and alveolar type II (AT2) cells within the lung, yielding acute respiratory infection (Leist et al., 2020).SARS-CoV-2 MA10 was delivered by intranasal inoculation with 1 × 10 4 foci-forming units (FFU) MA10 or vehicle (saline) in a volume of 25 μl in animal Biosafety Level 3 facilities.Characteristics of infection are described in Supplementary Fig. 2. Body weight was assessed with a standard postal scale.Arterial oxygenation was assessed with a pulse oximeter fitted with a sensor adapted for the mouse paw (MouseSTAT Jr. Pulse Oximeter & Heart Rate Monitor).Pulse oximetry was conducted on the hind paws.We present oximetry % change (Supplementary Fig. 1D) and absolute values (Supplementary Fig. 1E), because we noted that the baseline oximetry readings were lower than expected.Previous reports indicate that dark skin pigmentation decreases the accuracy of pulse oximetry readings, especially at lower oxygen saturation (Feiner et al., 2007), which might have influenced the oximetry readings in our black-pigmented mice (Supplementary Fig. 2D-E).

Behavioral assays
Behavior tasks were conducted in a dim biosafety cabinet laminar flow hood in the BSL3 facility.Testing arenas were white plastic bins 13 in.× 19 in.(Ikea) with pebbled floor.For novel object recognition (NOR), we first tested a catalog of 10 objects for intrinsic preference.Objects were similar in size (1-2 in.wide, 3-4 in.tall), visually interesting, without smell, and made of easily cleaned non-porous materials, e.g. 25 ml suspension flasks filled with pebbles, 50 ml conical tubes filled with corncob bedding, or 3-D printed flagpoles.Neodynium magnets affixed to each object were used to ensure consistent object placement relative to magnets permanently affixed to the underside of the arena.For familiarization, we placed individual mice in an arena containing two suspension flasks and allowed 10 min exploration.Behavior was filmed with an overhead mounted wide-angle webcam (Logitech C920S HD Webcam).Intersession interval was 14 h.For the testing session, mice were reintroduced into the arena containing one suspension flask and one novel object and filmed for 10 min.Objects and field were cleaned with ethanol and dried in between mice.Testing was conducted between 7:00-10:00 AM.Videos were coded and independently scored by two blinded scientists for duration of exploration of each object.Preference was calculated as (sec investigating novel object)/(sec investigating any object)*100.50% indicates no preference.

Fig. 1 .
Fig. 1.Cav-1 is increased in brain endothelial cells in SARS-CoV-2 infection.A-B) Immunofluorescence detection of Cav-1 (green) in hippocampal sections from wild-type mice euthanized 4 days post inoculation (DPI) with SARS-CoV-2.Monochromatic images (A, B) and overlays between Caveolin-1 (green), brain endothelial cell protein Glut-1 (red), and DAPI (blue) (A', B′).C) Quantification of % area immunoreactive for Cav-1 in hippocampus sections.Three brain sections analyzed per mouse.**p < 0.01, unpaired student t-test.D) Flow cytometric plot of CD31 and viable exclusion dye demonstrates gating strategy for endothelial cells (box) in single cell suspension prepared from isolated brain microvessels of WT mice euthanized 4 days after respiratory inoculation with SARS-CoV-2.E-F) Histograms of Cav-1 fluorescence intensity in viable CD31+ endothelial cells isolated as in D. G) Quantification of Cav-1 fluorescence intensity in viable CD31+ endothelial cells from mice euthanized at 4 DPI with SARS-CoV-2, expressed as ratio to healthy WT.n = 6 mice/group from two independent experiments.p < 0.05, unpaired student t-test.