Abstract
Herpes simplex encephalitis (HSE) caused by HSV-1 is the most common
non-epidemic viral encephalitis, and the neuropathogenesis of HSE
remains elusive. This work describes a 3D human neurovascular unit
(NVU) model that allows to explore the neuropathogenesis of HSE in
vitro. This model is established by co-culturing human microvascular
endothelial cells, astrocytes, microglia and neurons on a
multi-compartment chip. Upon HSV-1 infection, this NVU model exhibited
HSE-associated pathological changes, including cytopathic effects,
blood-brain barrier dysfunction and pro-inflammatory cytokines release.
Besides, significant innate immune responses were observed with the
infiltration of peripheral immune cells and microglial activation.
Transcriptomic analysis revealed broadly inflammatory and chemotactic
responses in host cells. Mechanistically, we found HSV-1 could induce
severe suppression of autophagic flux in glial cells, especially in
microglia. Autophagy activators could effectively inhibit HSV-1
replication and rescue neurovascular injuries, indicating the utility
of this unique platform for studying neurological diseases and new
therapeutics.
Subject terms: Cerebrovascular disorders, Viral pathogenesis,
Blood-brain barrier
__________________________________________________________________
Herpes simplex encephalitis caused by HSV-1 is a common viral
encephalitis with severe symptoms and poor prognosis. Here, the authors
developed a new 3D human neurovascular unit (NVU) model in vitro that
allows to probe the neuropathogenesis and therapeutics of
HSV-1-associated encephalitis.
Introduction
Herpes simplex encephalitis (HSE) is the most common sporadic
encephalitis with high mortality^[50]1, and over 90% of HSE cases in
adults are caused by HSV-1^[51]2–[52]4. Although the administration of
antiviral acyclovir has greatly reduced the mortality, most survivors
still suffer from permanent neurological sequelae^[53]4,[54]5. Besides,
increasing evidence implicates that HSV-1 infection is a potential risk
factor for Alzheimer’s disease^[55]1,[56]6,[57]7. So far, an effective
vaccine against HSV-1 has not yet been developed, and the only
antiviral drug (acyclovir) is facing the increasing problem of viral
drug resistance.
Clinically, HSE is characterized by severe cerebral damage, which is
considered to be attributed to a combination of viral replication and
excessive inflammatory responses^[58]1,[59]3,[60]8. In vivo, the brain
is the most complex organ, with multiple cell types and complicated
interactions between distinctive cell types in a 3D brain
microenvironment. In the last two decades, the concept of neurovascular
unit (NVU) was formalized, which refers specifically to nerve tissue
and its neighboring blood vessels, consisting of neurons, glial cells,
basement membrane and vascular cells (such as brain microvascular
endothelial cells and pericytes)^[61]9,[62]10. NVU emphasized the close
connections between brain cells and their neighboring microvasculature,
thus facilitating to study the various brain disease as a whole and to
find targeted therapy^[63]11–[64]15. It has been reported that NVU
injuries occur in a variety of neurological disorders, including
ischemic stroke, Alzheimer’s disease, and multiple
sclerosis^[65]11–[66]14. In particular, some pathological studies
revealed neurovascular injuries and dysfunctional blood-brain barrier
(BBB) occurred in the affected brain region of HSE
patients^[67]16–[68]19. However, the basis underlying HSV-1-induced
neurovascular injuries remains poorly defined.
At present, HSV-1 research mainly relies on mouse models. Due to the
inter-species differences, mouse models often showed inconsistent
findings with HSV-1-infected human cells or human clinical
trials^[69]20,[70]21. In recent years, brain organoids, as a new type
of 3D organ or tissue models, have been utilized for the
neuropathological studies of HSV-1 infection, such as the studying
neurodevelopmental disorders associated with neonatal HSV-1
infection^[71]22,[72]23, modeling HSV-1 encephalitis and drug
discovery^[73]24, and testing AD pathogenesis association with HSV-1
infection^[74]25. However, these brain organoids more resemble the
embryonic human brain and lack the vasculatures and immune cells that
are vital for constituting brain microenvironment. These features make
them not suitable for investigating the complex neurovascular injuries
and host immune responses underlying HSV-1 encephalitis.
Organ-on-a-chip, also called microphysiological system, is a form of
engineered microfluidic 3D culture system, that allows to recapitulate
the basic structure and key function of various human organs or tissues
in vitro^[75]26–[76]28. Based on this technology, some types of NVU or
BBB chips have been developed for the study of brain infection, such as
fungal brain infection^[77]29, SARS-CoV-2-induced
neuroinflammation^[78]30, Venezuelan equine encephalitis^[79]31 and
bacterial meningitis^[80]32.
In this work, we microengineered a new in vitro 3D NVU model that
allows to reconstitute the BBB interface and NVU structure by
co-culturing brain endothelial cells, microglia, astrocytes and neurons
in a biomimetic brain microenvironment. This model can not only emulate
the features of NVU in normal condition but also enables to monitor the
multi-stage intercellular interactions in pathological condition
dynamically, such as the responses of peripheral immune cells and
microglia. Upon HSV-1 infection, significant HSE-associated
pathological changes were detected on the NVU model. Mechanistically,
we found HSV-1 infection induced severe disturbance of autophagic flux
in glial cells. While autophagy activators could effectively inhibit
virus proliferation and rescue neurovascular injuries, revealing the
therapeutic potential of autophagy activators against HSV-1
encephalitis.
Results
Construction of a microengineered human 3D NVU model
The human brain is an organ of vast complexity in terms of the cell
types it comprises. In order to study brain infection associated with
HSV-1 (Fig. [81]1a), we initially designed and engineered a biomimetic
neurovascular unit device (also called NVU chip) with multi-compartment
chambers and different functional regions (Fig. [82]1b). The NVU chip
consisted of a top layer (named as blood side) and a bottom layer
(named as brain side), separated by a transparent porous membrane (8 µm
pore diameter, 7 µm thick) (Fig. [83]1c; Supplementary Fig. [84]1). The
top blood side included an open reservoir for seeding the brain
microvascular endothelial cells (HBMECs). The bottom brain side
consisted of five parallel channels for co-culturing of human glial
cells and neurons to mimic the in vivo brain microenvironment, as shown
in Supplementary Fig. [85]2. The astrocytes and neurons (astrocytes:
neurons = 10:1) mixed in 3D Matrigel were seeded in the central
compartment, and microglia were seeded into the two lateral channels.
The central and lateral culture compartments were connected via 12
trapezoidal channels (6 channels on each side) to enable the
intercellular communication of astrocytes/neurons and microglia. In
this way, the migration trajectories of microglia in the culture region
can be tracked visually under a microscope. Using this device, the four
types of cells (HBMECs, astrocytes, neurons, and microglia) were
co-cultured in an orderly manner on the chip to mirror the in vivo-like
3D NVU structure. According to this design, responses of distinctive
cell types to pathological stimuli can be monitored and recorded
conveniently.
Fig. 1. Establishment of a 3D human NVU model for HSE study.
[86]Fig. 1
[87]Open in a new tab
a Schematic diagram of the human neurovascular unit. b Image of a real
chip. c 3D schematic diagram showing cell culture on the NVU chip. The
yellow area indicates the HBMECs culture compartment of the blood side
(1). The dark gray area indicates the porous PCTE membrane between the
blood side and brain side (2). The red area indicates the
astrocytes/neurons culture compartment of the brain side (3). The blue
area indicates the microglia culture compartment of the brain side (4).
d 3 days after seeding cells on the chip, confocal micrographs showing
the HBMECs on the chip, immunostained for VE-cadherin, ZO-1 and
PECAM-1. e Confocal micrographs showing the astrocytes on the chip,
immunostained for S100β and GFAP. f Confocal micrographs showing the
neurons on the chip, stained for Tuj-1 and MAP2. g A confocal
micrograph showing the co-culture of astrocytes (GFAP) and neurons
(Tuj-1) in Matrigel. h Confocal micrographs showing the microglia on
the chip, immunostained for CD11b and IBA1. i A 3D configuration image
showing BBB interface formed by HBMECs (VE-cadherin) and astrocytes
(S100β) after co-culture for 3 days. d–i Representative images were
from 3 biological replicates. All experiments were repeated at least 3
times. j FITC-dextran permeability assays for BBB formed by HBMECs
alone, or HBMECs and astrocytes, after culture for 3 days (n = 6 for
biological replicates; n = 3 for technical replicates). Data are
presented as the mean ± SD and were analyzed using an unpaired
two-sided Student’s t-test.
Prior to the study of HSV-1 infection on the NVU model, we initially
characterized the structure and function of NVU chip by
immunofluorescent imaging after co-culture for 3 days. In the blood
side, an intact endothelium was formed with strong expression of
adherens junction protein VE-cadherin, tight junction protein ZO-1 and
endothelial marker PECAM-1 on the cell periphery (Fig. [88]1d). In the
brain side, astrocytes (S100β, GFAP) and neurons (Tuj-1, MAP2) were
co-cultured in 3D Matrigel exhibiting specific markers of the
individual cell type in the middle channels (Fig. [89]1e–g).
Quantification of GFAP+ astrocytes and Tuj-1+ neurons showed the ratio
of astrocytes to neurons is ~2.33 (Fig. [90]1g, Supplementary
Fig. [91]3). While, microglia exhibited polygonal shape with expressing
microglial markers (CD11b and IBA1) in the lateral channels
(Fig. [92]1h). As we know, in human NVU, HBMECs, astrocytes, pericytes
and basement membrane work in concert to form BBB in vivo, which
strictly controls the metabolites transport between blood circulation
and brain parenchyma, and prevents invasion of toxic substances and
pathogens. Here, 3D confocal configuration images revealed an intact
BBB interface formed by co-culturing HBMECs (VE-cadherin) and
astrocytes (S100β) on the porous membrane (Fig. [93]1i). Aquaporin-4
(AQP4) is a vital water channel protein which showed concentrated
expression at the end-feet of astrocytes near vasculature in human
brain^[94]33. Confocal micrographs showed many fluorescent puncta of
AQP4 were detected facing toward the brain endothelium (Supplementary
Fig. [95]4a), and some of the AQP4 puncta were located at the end of
end process of astrocytes (Supplementary Fig. [96]4b), indicating the
partially polarized localization of AQP4 on the NVU model. Also, the
barrier function of BBB was evaluated by permeability assessment using
dextran-FITC with various molecular weights (4 kDa and 40 kDa). It
appeared the co-culture of astrocytes and HBMECs significantly enhanced
the barrier function of BBB with integrated interface (Fig. [97]1j).
Collectively, these data indicated the formed human 3D NVU model
enables to recapitulate the major features of NVU in vitro at both
structural and functional levels.
HSV-1 induced severe BBB dysfunction and neurovascular injuries on the NVU
model
To probe the neuropathogenesis with HSV-1, we then tested HSV-1
infection on the NVU model. The virus (MOI = 0.5) was inoculated in the
bottom brain side (Fig. [98]2a), and viral infection was examined by
immunostaining for viral gG protein (HSV-1 gG). Immunofluorescence
analysis revealed obvious HSV-1 infection in the lower brain side
(13.68 ± 3.98% at 3 days post-infection (dpi), 20.73 ± 11.47% at 5 dpi
and 31.05 ± 8.14% at 7 dpi for astrocytes/neurons; 22.25 ± 5.72% at 3
dpi, 44.78 ± 8.46% at 5 dpi and 61.50 ± 8.33% 7 dpi for microglia)
(Fig. [99]2b, c). Meanwhile, significant cytopathies were detected in
the glial cells and neurons, such as syncytia, cell shrinkage, and cell
debris (Fig. [100]2b). Although no obvious viral infection was found,
marked downregulation of VE-cadherin was observed in the brain
endothelium (Fig. [101]2b). Permeability assay (4 kDa FITC-dextran)
revealed increased permeability on the infected-NVU model at 7 dpi,
indicating the loss of BBB integrity following HSV-1 infection
(Fig. [102]2d).
Fig. 2. HSV-1 infection on the NVU model.
[103]Fig. 2
[104]Open in a new tab
a Schematic diagram of HSV-1 inoculation on the NVU model. b Confocal
micrographs showing the HBMECs, astrocytes, microglia and neurons
following HSV-1 infection at 7 dpi on the chip by detecting for viral
gG protein (red) (n = 3 for biological replicates; n = 3 for technical
replicates). c Quantification of the infected cells for HBMECs,
astrocytes/neurons and microglia and at 3, 5 and 7 dpi (n = 3 for
biological replicates; n = 3 for technical replicates). Three fields
were quantified for each chip. Data are presented as the mean ± SD. d
BBB permeability assay using 4 kDa FITC-dextran for Mock or
HSV-1-infected NVU chips at 7 dpi (n = 6 for biological replicates;
n = 3 for technical replicates). Data are presented as the mean ± SD
and were analyzed using an unpaired Student’s t-test. e LDH assay for
Mock or HSV-1-infected NVU chips at 3, 5, and 7 dpi, respectively
(n = 3 for biological replicates; n = 3 for technical replicates). Data
are presented as the mean ± SD and were analyzed using an unpaired
two-sided Student’s t-test. f Confocal micrographs of HBMECs,
astrocytes/neurons, and microglia, stained with Annexin V-FITC/PI dyes
following HSV-1 infection at 7 dpi (n = 4 for biological replicates;
n = 3 for technical replicates). g Quantification of Annexin V positive
or PI-positive cells for HBMECs, astrocytes/neurons, and microglia
following HSV-1 infection, based on (f) (n = 4 for biological
replicates; n = 3 for technical replicates). Data are presented as the
mean ± SD and were analyzed using an unpaired two-sided Student’s
t-test. h ELISA results of medium from brain side for Mock or
HSV-1-infected chips at 3 dpi (n = 3 for biological replicates; n = 3
for technical replicates). Data are presented as the mean ± SD and were
analyzed using an unpaired two-sided Student’s t-test.
Furthermore, cell viability was examined by LDH (lactate dehydrogenase)
activity assay for the medium, and it was observed LDH activity was
significantly increased following viral infection in both blood side
and brain side at 3, 5, and 7 dpi (Fig. [105]2e). Cell death assay
(Annexin V/PI kit staining) showed, at 7 dpi, no obvious cell death was
detected in the brain endothelial cells, whereas marked cell death was
found in the glial cells and neurons (Fig. [106]2f, g). As increasing
evidences implicate that HSV-1 infection is a potential risk factor for
Alzheimer’s disease^[107]1,[108]6,[109]7, and several studies reported
HSV-1 infection can induce Aβ accumulation or deposition in
vitro^[110]25,[111]34,[112]35. We then determined the amyloid
pathological responses following HSV-1 infection on the NVU model. By
immunostaining for amyloid-β (Aβ) in astrocytes/neurons compartment, we
found HSV-1 infection had no obvious effect on Aβ40 accumulation at 4
dpi, while led to a significant increase in Aβ42 level (Supplementary
Fig. [113]5), indicating this NVU chip can recapitulate the amyloid
pathology caused by HSV-1 infection in vitro. Moreover, cytokine levels
in the medium of brain side were examined by ELISA experiment, and
results showed IL-1β, IL-6, TNF-α, CXCL1, and IFN-β were significantly
released following HSV-1 infection at 3 dpi (Fig. [114]2h).
Taken together, these studies reflected this NVU model could
recapitulate the neuropathological features associated with HSV-1
infection in vitro, including BBB dysfunction, viral infection, cell
injuries, Aβ42 accumulation and pro-inflammatory cytokines release,
which indicated it can be further utilized for studies of disease
pathogenesis and development of new therapeutics.
HSV-1 infection induced microglial response on the NVU model
In the central nervous system, microglia are the resident immune cells
and function as the first line of defense against pathogen
invasion^[115]36,[116]37. When pathogens invade the brain tissue,
microglia respond quickly and participate in a range of immune
responses, such as migration to the infected area, recognition and
phagocytosis of pathogens and dying cells, antigen presentation, or
secretion of a variety of bioactive substances^[117]38,[118]39. To
track the microglial responses on the NVU model following HSV-1
infection, microglia were labeled with RFP by lentivirus-RFP vector and
were visualized in real time. HSV-1 (MOI = 0.1) was inoculated in the
right channel of astrocytes/neurons compartment, and 2 h later,
microglia labeled with RFP were seeded in the left microglial channel.
At 0, 1, 3 and 5 dpi, microglial migration was recorded under a
fluorescent microscope. The results showed more microglia migrated to
the astrocytes/neurons compartment on the HSV-1-infected model
(Fig. [119]3a, b; Supplementary Fig. [120]6), which is reminiscent of
the pathological changes of microglia in HSE patient
samples^[121]18,[122]19. It was reported that pro-inflammatory
cytokines, such as IL-1β and TNF-α, can stimulate microglia
proliferation^[123]40,[124]41, which is a key component of the
responses to brain injuries^[125]42. The ELISA experiments showed,
HSV-1 infection triggered a release of IL-1β and TNF-α on the NVU model
(Fig. [126]2h). We then examined the proliferation state of microglia
by immunostaining for Ki67, and the result showed more proliferating
microglia were detected on the HSV-1-infected NVU model (Fig. [127]3c,
d). Based on these findings, we speculated that enhanced proliferation
and increased migration might jointly lead to more appearance of
microglia in the astrocytes/neurons compartment following HSV-1
infection.
Fig. 3. Changes of microglia following HSV-1 infection on the NVU model.
[128]Fig. 3
[129]Open in a new tab
a Fluorescent images showing microglial migration (red) following HSV-1
infection on the chip at 0, 1, 3, and 5 dpi (n = 3 for biological
replicates; n = 3 for technical replicates). The dotted white line
indicates the interface between astrocytes-neurons compartment and
microglia compartment. b Quantification of the migrated microglia at 0,
1, 3 and 5 dpi (n = 3 for biological replicates; n = 3 for technical
replicates). Data are presented as the mean ± SD and were analyzed
using an unpaired two-sided Student’s t-test. c Confocal micrographs of
microglia (red) immunostained for Ki67 at 4 dpi (n = 3 for biological
replicates; n = 3 for technical replicates). The dotted white line
indicates the interface between astrocytes-neurons compartment and
microglia compartment. d Quantification of Ki67+ microglia based on c.
Three fields were quantified for each chip. Data are presented as the
mean ± SD and were analyzed using an unpaired two-sided Student’s
t-test. e Confocal micrographs of astrocytes/neurons immunostained for
viral gG protein (green) and MAP2 (red) at 4 dpi, with or without
microglia (n = 4 for biological replicates; n = 3 for technical
replicates). f Quantification of the infected neurons based on (c)
(n = 4 for biological replicates; n = 3 for technical replicates).
Three fields were quantified for each chip. Data are presented as the
mean ± SD and were analyzed using an unpaired two-sided Student’s
t-test.
Microglia are known to phagocytose damaged cells or cell debris under a
number of pathological conditions, including viral
infections^[130]43,[131]44. We then examined microglial phagocytosis on
the NVU model 5 days after HSV-1 infection (MOI = 0.1). As shown in
Supplementary Fig. [132]6, HSV-1 infection resulted in more cell debris
in the astrocytes/neurons compartment compared with Mock-infected NVU
model. Intriguingly, some cell debris was engulfed by the microglia
(Supplementary Fig. [133]7), indicative of microglial phagocytosis
induced by HSV-1 infection. To further determine the antiviral effect
of microglia, we conducted a microglia ablation study. To exclude the
possibility that less cells affect the infection efficiency, microglia
were replaced with the same quantity of astrocytes-neurons on the
microglia-deprived NVU model (w/o microglia). After 4 days infection
(MOI = 0.5), more infected neurons were detected on the
microglia-deprived NVU model, compared with NVU model with microglia (w
microglia) (Fig. [134]3e, f), indicating the protective roles of
microglia against acute HSV-1 infection.
HSV-1 infection elicited infiltration of peripheral immune cells on the NVU
model
To gain a deeper understanding of the molecular signatures underlying
HSV-1-associated encephalitis, we performed RNA-sequencing (RNA-seq)
analysis of cells on the NVU model at 3 dpi. Genes differentially
expressed with fold changes of >2.0 and P < 0.05 were identified as
differentially expressed genes (DEGs). Transcriptomic analysis revealed
that HSV-1 infection induced global perturbations in all three types of
cells (1049 DEGs in HBMECs, 5093 DEGs in astrocytes, and 5610 DEGs in
microglia) (Fig. [135]4a; Supplementary Fig. [136]8). Consistent with
ELISA results in Fig. [137]2h, RNA-seq data revealed many cytokine
genes (GO: 0005125) elevated in HBMECs following HSV-1 infection
(Fig. [138]4b), some of which are enriched in “TNF signaling pathway”
according to KEGG (Kyoto Encyclopedia of Genes and Genomes) enrichment
analysis (Supplementary Fig. [139]9). Gene set enrichment analysis
(GSEA) indicated “TNF signaling pathway” was significantly activated in
HBMECs following HSV-1 infection (Fig. [140]4c), and upregulated genes
included chemokines (CXCL1, CXCL2, CCL2, CXCL3, CX3CL1 and CCL20),
pro-inflammatory cytokine (IL-6) and cellular adhesion molecule (ICAM1)
(Supplementary Fig. [141]10). These findings implicated HSV-1 infection
induced an inflammatory state in HBMECs, which might trigger
recruitment of peripheral immune cells to further participate in
inflammation.
Fig. 4. HSV-1 infection triggered chemotactic response and infiltration of
peripheral immune cells on the NVU model.
[142]Fig. 4
[143]Open in a new tab
a Scatter plots showing the differentially expressed genes in HBMECs
following HSV-1 infection at 3 dpi. Genes differentially expressed with
fold changes of >2.0 and P < 0.05 are marked in color. P-values were
calculated using a two-sided, unpaired Student’s t-test with equal
variance assumed (n = 3). b Heatmap indicating the expression levels of
upregulated genes annotated to cytokine activity (GO: 0005125) in
HBMECs (n = 3). c GSEA analysis reveals the correlation between HSV-1
infection and genes involved in the TNF signaling pathway in HBMECs.
The analyses were all one-sided and adjustments were made for multiple
comparisons. NES normalized enrichment score, FDR false discovery rate.
Gene sets were considered significant when P < 0.05 and FDR < 0.25. d
Schematic diagram of adding PBMCs following HSV-1 inoculation on the
NVU model. e Fluorescent micrographs of PBMCs (red) attached to the
HBMECs (green) following HSV-1 infection at 3 dpi (n = 4 for biological
replicates; n = 3 for technical replicates). f Quantification of the
attached PBMDs per filed for the Mock or HSV-1-infected chips (n = 4
for biological replicates; n = 3 for technical replicates). Two fields
were quantified for each chip. Data are presented as the mean ± SD and
were analyzed using an unpaired two-sided Student’s t-test. g
Fluorescent micrographs of migrating PBMCs through the brain
endothelium on the HSV-1-infected chip (n = 4 for biological
replicates; n = 3 for technical replicates). h Heatmap indicating the
expression levels of upregulated genes annotated to chemokine activity
(GO: 0008009) in astrocytes (n = 3). i Heatmap indicating the
expression levels of upregulated genes annotated to chemokine activity
(GO: 0008009) in microglia (n = 3). j Fluorescent micrographs of
infiltrated PBMCs in the astrocytes culture compartment 3 h after PBMCs
addition (n = 3 for biological replicates; n = 3 for technical
replicates). k Quantification of the infiltrated PBMCs per filed for
the Mock or HSV-1-infected chips (n = 3 for biological replicates;
n = 3 for technical replicates). Two fields were quantified for each
chip. Data are presented as the mean ± SD and were analyzed using an
unpaired two-sided Student’s t-test. l Fluorescent micrographs showing
infiltrated T cells (CD45^+CD3^+) in the astrocytes culture compartment
24 h after PBMCs addition (n = 4 for biological replicates; n = 3 for
technical replicates). m Fluorescent micrographs showing infiltrated B
cells (CD45^+CD19^+) in the astrocytes culture compartment 24 h after
PBMCs addition (n = 4 for biological replicates; n = 3 for technical
replicates). n Fluorescent micrographs showing infiltrated macrophages
(CD45^+CD68^+) in the astrocytes culture compartment 24 h after PBMCs
addition (n = 4 for biological replicates; n = 3 for technical
replicates). o Quantification of the infiltrated T cells, B cells and
macrophages per field for the Mock or HSV-1-infected chips (n = 4 for
biological replicates; n = 3 for technical replicates). Two fields were
quantified for each chip. Data are presented as the mean ± SD and were
analyzed using an unpaired two-sided Student’s t-test.
To probe the responses of peripheral immune cells to HSV-1 infection,
peripheral blood mononuclear cells (PBMCs) were added to the HBMECs
compartment to monitor the recruitment of immune cells (Fig. [144]4d).
Two hours later, more PBMCs were found to attach to the brain
endothelium of the HSV-1-infected group (Fig. [145]4e, f). Moreover,
transmigration of PBMCs through the brain endothelium was further
detected on the HSV-1-infected group (Fig. [146]4g). Next, we analyzed
the transcriptome data of glia cells and found more chemokine genes
were upregulated in both astrocytes and microglia in the infected group
(Fig. [147]4h, i). In line with the RNA-seq analysis, obvious
infiltration of peripheral immune cells was observed in the
astrocytes/neurons compartment of the HSV-1-infected models, while
almost no peripheral immune cells were detected in the Mock-infected
models (Fig. [148]4j, k). Twenty-four hours after addition of PBMCs,
the types of infiltrated immune cells were determined by immunostaining
for specific markers of immune cells, including T cells (CD45^+CD3^+),
B cells (CD45^+CD19^+) and macrophages (CD45^+CD68^+). The results
showed the main infiltrated immune cells are T cells and macrophages,
while no obvious infiltration of B cells was detected following HSV-1
infection (Fig. [149]4l–o), which was consistent with the components of
parenchymal inflammatory infiltrates identified in the HSE pathological
samples^[150]8,[151]18,[152]19. Intriguingly, the infiltrated immune
cells exhibited in vivo-like morphology, such as spherical T cells
(enlarged images in Fig. [153]4l) and irregularly shaped macrophages
with many pseudopodia (enlarged images in Fig. [154]4n).
Collectively, it appeared this NVU model can reflect the host immune
responses in brain to HSV-1 infection via brain-resident microglia and
peripheral immune cells, indicating it can be further utilized to
explore the roles of immune cells in the pathogenesis of HSV-1
encephalitis.
Blockage of autophagic flux in glial cells following HSV-1 infection
As above, HSV-1 infection induced obvious dysfunctions of NVU, we then
tried to identify the molecular basis of the host responses to the
viral infection by KEGG enrichment analysis (Fig. [155]5a, b).
Intriguingly, we noticed some vital biological processes were commonly
modulated in both astrocytes and microglia, especially the autophagic
process (e.g., autophagy or mitophagy). Next, we examined the NVU
models by transmission electron microscope (TEM), and discovered more
autophagic vacuoles in the microglia and astrocytes of HSV-1 infected
models at 3 dpi (Fig. [156]5c; Supplementary Fig. [157]11), especially
in the microglia. Further, the microglia were examined by
immunofluorescence for autophagic markers (LC3B and p62), and the p62
and LC3B positive dots were significantly increased in both rapamycin
(a potent autophagy activator) treated and HSV-1 infected NVU models at
3 dpi (Fig. [158]5d).
Fig. 5. Autophagy dysregulation in glial cells induced by HSV-1 infection on
the NVU model.
[159]Fig. 5
[160]Open in a new tab
a KEGG pathway enrichment analysis of upregulated genes and
downregulated genes in astrocytes following HSV-1 infection at 3 dpi. b
KEGG pathway enrichment analysis of upregulated genes and downregulated
genes in microglia following HSV-1 infection at 3 dpi. a, b
Autophagy-related terms were marked with red boxes. The analyses were
all one-sided and adjustments were made for multiple comparisons. c TEM
images showing microglia on Mock or HSV-1-infected NVU models at 3 dpi
(n = 3 for biological replicates; n = 3 for technical replicates). The
autophagic vacuoles were indicated by red asterisks. d Confocal
micrographs showing the microglia immunostained for p62 (green) and
LC3B (red) following HSV-1 infection at 3 dpi on the chip (n = 3 for
biological replicates; n = 3 for technical replicates). The chip
treated with rapamycin was used as a positive control group. e Confocal
micrographs showing autophagic flux in microglia following HSV-1
infection at 3 dpi (n = 3 for biological replicates; n = 3 for
technical replicates). The red dots indicate autolysosome, and the
yellow dots indicate autophagosome. The chip treated with rapamycin was
used as a positive control. f Quantification of autolysosome and
autophagosome numbers in microglia following HSV-1 infection based on
(e) (n = 3 for biological replicates; n = 3 for technical replicates).
Ten cells were quantified for each chip. Data are presented as the
mean ± SD and were analyzed using an unpaired two-sided Student’s
t-test.
As autophagic flux is a dynamic process, it is difficult to distinguish
whether the increased autophagosomes are due to increased autophagosome
formation or suppressed autophagosome clearance. We then used GSEA as a
targeted method to analyze the potential correlation between HSV-1
infection and the autophagic process in all three types of cells. The
results showed that the pathway related to autophagy exhibited
significant differences between the Mock group and HSV-1 group in both
astrocytes and microglia, while not in HBMECs (Supplementary
Fig. [161]12). Especially, we found this difference was more
significant in the microglia (P = 0.299 in HBMECs; P = 0.048 in
astrocytes; P = 0.020 in microglia), in which HSV-1 infection was
negatively correlated with the autophagic process (Supplementary
Fig. [162]12). To verify this GSEA analysis, autophagic flux was
examined in microglia using an adenovirus-mRFP-GFP-LC3 tool, which can
directly reflect the level of autophagic flux by the fluorescent
dots^[163]45. The red dots indicate the autolysosomes and the yellow
dots (overlay of GFP and RFP) indicate the autophagosomes. The NVU
model treated with rapamycin was used as a positive control (increased
RFP dots) (Fig. [164]5e). The results showed the number of
autophagosomes was significantly increased following HSV-1 infection at
3 dpi, while the number of autolysosomes was decreased at the meanwhile
(Fig. [165]5e, f). These findings reflected that HSV-1 infection caused
excess accumulation of autophagosomes in the microglia, by blocking the
fusion of autophagosomes with lysosomes.
Autophagy activators ameliorated HSV-1-induced neurovascular injury by
inhibiting viral replication
As HSV-1 infection caused severe blockage of the autophagic process in
host cells, we then tried to test whether autophagy could be a therapy
target for HSV-1 encephalitis. To verify the hypothesis, we performed a
drug screening by testing autophagy activators or inhibitors on the
HSV-1 encephalitis model (Fig. [166]6a). In this drug screening test,
acyclovir (Acyc) which is a common antiviral drug to treat HSV-1
infection, was used a positive control drug. Autophagy activators
include rapamycin (Rapa) and two natural compounds (Euphpepluone K
(EupK) and Munronin V (MunV)). Euphpepluone K was isolated from
Euphorbia peplus Linn^[167]46, and Munronin V was isolated from
Munronia henryi Harms^[168]47. Previous studies reported that both of
the two compounds significantly activated autophagic flux in
cells^[169]46,[170]47. As a comparison, 3-Methyladenine (3-MA) was
chosen as an autophagy inhibitor in this test. Cell viability assay
showed, similar to acyclovir, all 3 autophagy activators (rapamycin,
Euphpepluone K and Munronin V) could effectively rescue the cell death
caused by HSV-1 infection, while autophagy inhibitors (3-MA)
significantly exacerbated the cell death (Fig. [171]6b). Besides, BBB
permeability assay and LDH assay were performed, and the results showed
acyclovir and all 3 autophagy activators could restore the BBB injury
and LDH release caused by HSV-1 infection to different extend
(Fig. [172]6c, d).
Fig. 6. Evaluation of HSV-1-induced neurovascular injury in response to
autophagy activators treatment on the NVU model.
[173]Fig. 6
[174]Open in a new tab
a Experimental flow diagram showing the drug testing on the HSV-1
encephalitis model. b Cell viability assay showing the therapeutic
effects of different drugs on HBMECs, Astrocytes/Neurons, and microglia
following HSV-1 infection on the model at 7 dpi (n = 3 for biological
replicates; n = 3 for technical replicates). Acyclovir (Acyc) was used
as a positive control drug. Rapa rapamycin, 3-MA
3-methyladenine, Euphpepluone K EupK, Munronin V MunV. Cells were
stained with LIVE/DEAD® Viability/Cytotoxicity Kit, and red color
indicates the dead cell. c 4 kDa FITC-dextran assay showing the effects
of different drugs on the BBB permeability following HSV-1 infection at
7 dpi (n = 3 for biological replicates; n = 3 for technical
replicates). Data are presented as the mean ± SD and were analyzed
using a one-way analysis of variance (ANOVA) followed by the Bonferroni
post hoc test. d LDH assay showing the therapeutic effects of different
drugs on the model following HSV-1 infection at 7 dpi (n = 3 for
biological replicates; n = 3 for technical replicates). Data are
presented as the mean ± SD and were analyzed using a one-way ANOVA
followed by the Bonferroni post hoc test.
Next, we continued to investigate whether the autophagy activators
could rescue HSV-1-caused neurovascular injury by inhibiting HSV-1
replication. Immunofluorescent analysis showed the infected cells were
dramatically decreased following treatment with acyclovir or 3
autophagy activators at 7 dpi, while were significantly increased
following treatment with autophagy inhibitor (3-MA) (Fig. [175]7a).
TCID50 assay was performed to determine the viral load in medium of
brain side at 7 dpi, and the results showed the viral load was
decreased following treatment with autophagy activators, while was
increased following treatment with autophagy inhibitor (Fig. [176]7b),
indicating that the autophagy activators rescued neurovascular injury
by inhibiting HSV-1 replication. Furthermore, autophagic flux assay was
also performed, and the results showed that autophagy activators
effectively decreased the excess accumulation of autophagosomes and
restored the blockade of autophagic process on the HSV-1-infected NVU
model (Fig. [177]7c, d).
Fig. 7. Assessment of HSV-1 infection and autophagic flux in microglia
following autophagy activators treatment on the NVU model.
[178]Fig. 7
[179]Open in a new tab
a At 7 dpi, confocal micrographs showing the HSV-1 infection (red,
immunostained for viral gG protein) in astrocytes/neurons and microglia
after adding the indicated drugs (n = 3 for biological replicates;
n = 3 for technical replicates). b TCID50 assay showing the antiviral
effects of different drugs on the brain side (n = 3 for biological
replicates; n = 3 for technical replicates). c Confocal micrographs
showing autophagic flux in microglia with the addition of indicated
drugs following HSV-1 infection at 3 dpi n = 3 for biological
replicates; n = 3 for technical replicates. The red dots indicate
autolysosome, and the yellow dots indicate autophagosome. d
Quantification of autolysosome and autophagosome numbers in microglia
of the HSV-1-infected chips after adding indicated drugs (n = 3 for
biological replicates; n = 3 for technical replicates). Ten cells were
quantified for each chip. Data are presented as the mean ± SD and were
analyzed using a one-way ANOVA followed by the Bonferroni post hoc
test.
Discussion
This work described a 3D human NVU model containing a functional BBB
interface and neural components, which allows to simulate the brain
microenvironment in normal and pathophysiology of HSE associated with
HSV-1 infection. Upon HSV-1 infection, this model revealed obvious
pathological changes, such as cytopathic effects, BBB dysfunction,
pro-inflammatory cytokines release, microglial response and
infiltration of peripheral immune cells. These results are consistent
with the findings of pathological studies in SAE
patients^[180]16–[181]19. Intriguingly, it was found that HSV-1 induced
severe suppression of autophagic flux in glial cells, and autophagy
activators could effectively rescue the neurovascular injuries induced
by HSV-1.
In vivo, the brain has the largest number of cell types and the most
complex physiological microenvironment. The pathogenesis of HSE still
remains elusive, partially due to the lack of models that accurately
mimic the complex physiological structure of the brain and reflect the
human-relevant responses to HSV-1 infection. Although mouse models are
widely used in HSV-1 studies, mice are not natural hosts for this
virus, which leads to host responses different from those in humans.
For example, HSV-1 infection leads to inhibition of necroptosis in
human cells, but activation of necroptosis in mouse cells^[182]20. Many
anti-HSV-1 candidate drugs that showed promise in mouse models, all
failed in human clinical trials, such as vaccines. In addition,
organoids have recently been utilized in the neuropathological studies
of HSV-1 infection^[183]22–[184]25. Although the brain organoid models
show great potential in the study of brain HSV-1 infection, they still
face some challenges at this stage, such as a lack of vasculatures or
immune cells, which makes them not suitable for investigating
neurovascular injuries and the immune cells-mediated inflammatory
responses underlying HSV-1 encephalitis. Till now, the model that can
accurately reflect the pathogenesis of HSV-1-associated encephalitis is
still lacking.
In this work, to probe the neuropathogenesis with HSV-1 in a
human-relevant manner, we created a human NVU-on-a-chip model,
including brain endothelial cells, neurons, astrocytes and microglia in
3D microenvironment. A multi-compartment co-culture chip was designed
for the evaluation of host responses to HSV-1 infection and enabled
real-time tracking of distinctive types of cells. Following viral
infection in the brain side, we detected prominent viral replication in
the brain cells, including glial cells and neurons, while no visible
HSV-1 replication was observed in HBMECs. As we know, brain endothelium
is the core component of BBB, which line the brain capillaries.
Although it is not the direct target for HSV-1 infection, brain
endothelium still suffered from severe injuries, such as disruption of
cell junctions, endothelial cell death, increased BBB permeability and
more LDH release. This indicates the injuries of HBMECs are caused in
an indirect manner, and this might be mediated by the cross-talk
between infected brain cells and endothelial cells.
In addition to the direct injury of brain cells by viral infection,
inflammatory response was considered another cause for severe cerebral
injury in HSE. We then evaluated the host immune responses to HSV-1
infection on the NVU model. By real-time recording of microglial
responses following HSV-1 infection, it appeared more microglia
migrated to the infected area and phagocytosed cell debris. Ablation of
microglia significantly exacerbated HSV-1 infection in the
astrocytes/neurons compartment, and this result indicated the
protective role of brain-resident immune cells against viral infection
in HSE. Furthermore, we also probed the responses of peripheral immune
cells to HSV-1 infection. This NVU model contains a continuous layer of
brain endothelial cells that, together with extracellular matrix and
astrocytes, form a functional BBB interface. It is thus convenient to
record the adhesion and infiltration of peripheral immune cells from
the blood side into the brain side. The results indicated HSV-1
infection triggered the secretion of chemokines from HBMEs and glial
cells, which might cause peripheral immune cells (mainly T cells and
macrophages) to adhere to the brain endothelial cells, transmigrating
through the endothelium, followed by infiltration into brain side.
These findings demonstrated the unique advantages of this model to
emulate the complex inflammatory responses mediated by resident immune
cells (microglia) or peripheral immune cells in HSE^[185]18,[186]19.
To better understand the molecular basis underlying HSE, we performed
RNA-seq analysis on the NVU at 3 dpi. Consistent with pathological
studies, transcriptomic analysis revealed the upregulation of many
immune-related genes, including pro-inflammatory cytokines and
chemokines. Further GO enrichment analysis of upregulated genes
indicated several vital inflammation-associated pathways were activated
following HSV-1 infection, such as the TNF signaling pathway, IL-17
signaling pathway, NF-kappa B signaling pathway, etc. However, type I
interferon (INF-1) signaling pathway, which plays pivotal roles in
defense against viral infection^[187]48,[188]49, is not significantly
activated by HSV-1 infection on the NVU model, according to the RNA-seq
analysis. This might be due to the immune evasion mechanism exploited
by HSV-1. Previous studies reported HSV-1 has evolved several
strategies to interfere with IFN responses at multiple levels^[189]50.
For example, HSV-1 ICP0 protein was reported to degrade the host
deubiquitinase BRCC36 and further downregulate IFN-I receptor
IFNAR1^[190]51, and HSV-1 ICP27 protein was found to interfere with
STAT-1 activation and hamper the downstream transcription of ISGs
(interferon-stimulated genes)^[191]52. Moreover, another study reported
the levels of IFN antiviral responses to HSV-1 are different among
distinctive tissues, and IFN induction is more robust in skin than in
brain^[192]53. Taken together, we suggested that the immune evasion
mechanism of HSV-1 might be responsible for the failure to trigger a
significant IFN response on the NVU model, which in turn increased the
degree of brain damage.
Intriguingly, we found HSV-1 infection induced severe suppression of
autophagic flux in glial cells. Autophagy has been shown to act as a
vital cellular defense mechanism against pathogens by eliminating
intracellular microorganisms^[193]54,[194]55. Several studies reported
HSV-1 infection causes autophagy inhibition, by which the virus evades
the innate immune immunity to promote its propagation^[195]56–[196]58.
In support of this view, deeper studies revealed HSV-1 neurovirulence
factor ICP34.5 reverses PKR-mediated eIF2a phosphorylation and
translational shutoff in host cells^[197]59, and it could also bind to
BECN1 and inhibits BECN1-mediated autophagy^[198]57,[199]60. Other
studies, however, reported HSV-1 infection induces autophagy in some
types of cells, such as macrophages, fibroblast, neurons and glial
cells, which contributes to antigen presentation or promoting cell
survival^[200]61–[201]63. With regard to the controversial findings,
further studies indicated that HSV-1 might modulate autophagy in a
cell-type-specific manner^[202]55,[203]64. For example, autophagy plays
a critical role in limiting HSV-1 replication in terminally
differentiated neurons, while it is not required for HSV-1 control in
mucosal epithelial cells and other mitotic cells^[204]65. In this work,
we found HSV-1 infection caused excess accumulation of autophagosomes
and decreased autolysosomes in glial cells, which indicated HSV-1
infection interfered with the fusion of autophagosomes with lysosomes
and blocked the autophagic flux. In addition, the addition of autophagy
inhibitor promoted HSV-1 replication and exacerbated neurovascular
injuries, while treatment of three types of autophagy activators
obviously rescued the injuries caused by HSV-1 infection. Intriguingly,
we found the degrees of autophagic blockage caused by HSV-1 infection
were different in three cell types (HBMECs, astrocytes and microglia),
and this effect was particularly pronounced in the microglia. These
data supported that HSV-1 can suppress autophagic flux in host cells
and promote viral replication, which indicates the autophagy pathway
might be a valuable therapeutic target for HSV-1 encephalitis.
The brain is a complex organ with various cell types, and each cell
type has different functions. Therefore, it is necessary to consider
the proportion of different cells, especially the ratio of astrocytes
to neurons. Previous studies reported, the glia/neurons ratio in the
whole human brain is close 1^[205]66–[206]68. However, the glia/neuron
ratio varies dramatically among distinctive brain regions, and the
ratio is ~3.7 in cerebral cortex, ~0.2 in cerebellum, and ~11.3 in the
rest of brain^[207]66. Clinical studies reported HSV-1 mainly attacks
cortical areas of the brain in HSE patients, including temporal lobe,
frontal lobe and limbic system^[208]69,[209]70. Besides, considering
that astrocytes are the most abundant cells among glial
cells^[210]71,[211]72, we set the cell seeding ratio of
astrocyte/neuron to 10 in the NVU chip. As SH-SY5Y cells have a higher
proliferation rate than HA cells, the final ratio of astrocytes to
neurons reaches ~2.33 after co-culture for 3 days.
Despite the advantage of this NVU model for the study of HSV-1
encephalitis, there are still some limitations that need to be
improved. First of all, immortalized cell lines were used to represent
neurons and microglia in this NVU model. Especially for neurons,
SH-SY5Y cells lack some vital neuronal features, such as mature
synapses (Supplementary Fig. [212]13) and electrophysiological
characteristics. Neurons and microglia differentiated from iPSCs could
be a better choice for the construction of a more biomimetic NVU model.
Second, a vascular channel design with an open reservoir was applied in
our model, without fluidic flow acting on the brain endothelium. It is
known that dynamic fluid flow and shear stress significantly contribute
to the normal structure and function of blood vessels. Third, pericyte
is also a vital cell type contributing to a functional BBB, which
should be incorporated in the future to improve the overall functions
of the NVU model. Finally, although Matrigel is widely used for the
establishment of BBB or NVU models^[213]33,[214]73,[215]74, it is
limited in the ill-defined composition and high laminin (~60%)
content^[216]75, which is different from the real composition of
extracellular matrix in the human brain^[217]76. Thus, it is suggested
to reconstitute the human NVU model using an extracellular matrix with
brain-rich and chemical defined substances in further studies, such as
collagen IV or fibrin/fibrinogen.
Taken together, we established a new 3D NVU model that allows to
recapitulate the features of NVU in normal and pathological condition
of HSV-1 infection in vitro. The novelty of this work lies in its
ability to mimic the physiologically relevant microenvironment of NVU
and the host immune responses to viral infection in a human-organ
context. In particular, it enables the visualization of interactions
between the diverse brain cells (astrocytes, neurons) and immune cells
(peripheral immune cells, microglia) dynamically, which is not easily
achieved with animal models. This work provides a unique platform for
the study of pathophysiology in HSV-1 encephalitis and the development
of new therapeutics for this devastating disease.
Methods
Cell culture
Human microvascular endothelial cells (HBMECs) were purchased from
ScienCell Corporation (no. 1000) and cultured in an endothelial cell
medium (ECM; ScienCell; no. 1001) supplemented with 5% FBS. Human
astrocytes (HA cells) were purchased from ScienCell Corporation (no.
1800) and cultured in AM medium (ScienCell; no. 1801). Human
neuroblastoma cell line SH-SY5Y was purchased from Procell Corporation
(no. CL-0208) and maintained in DMEM/F-12 medium (Gibco; no.
C11330500BT) supplemented with 10% FBS and 1% P/S. HMC3 cells were
purchased from Procell Corporation (no. CL-0620) and maintained in MEM
medium containing NEAA (Procell; no. [218]MP150410) supplemented with
10% FBS and 1% P/S. HMC3-RFP cells were transfected with a
lentivirus-GFP vector for 48 h, and RFP-positive cells were selected
using 10 μg/ml puromycin dihydrochloride (Beyotime; no. ST551). PBMCs
were isolated from fresh human peripheral blood using Ficoll (GE
Healthcare) density gradient centrifugation and cultured in RPMI 1640
medium containing 10% FBS, 1% P/S, and 50 IU/ml IL-2. All cells were
cultured at 37 °C in a humidified atmosphere of 5% CO[2].
Chip fabrication
The microfluidic NVU model device was composed of two
polydimethylsiloxane (PDMS) layers and a PCTE membrane (Whatman; no.
10417512; 1 × 10^5 pores/cm^2). First, SU-8 molds with different
heights were prepared by standard photolithography technology as
previously described^[219]77. The base and curing agent (Sylgard 184,
Dow Corning company) were mixed at a 10:1 ratio (wt/wt) and poured on
the molds, and cured for 45 min at 80 °C. Peeled off the cured PDMS
replica and punched a rounded rectangle (10 mm long × 5 mm wide) in the
center of the upper layer. Punched round holes (1.6 mm diameter for 6
middle pores; 3.5 mm diameter for 4 lateral pores) at both ends of PDMS
channels for the inlets and outlets of cell-containing compartments.
Using uncured PDMS as glue, a porous polycarbonate track-etched (PCTE)
membrane (7 μm thick, 8 μm pore size) was glued onto the central hole
in the upper layer. Then, two PDMS layers were bonded using oxygen
plasma treatment (250 W, 30 s, TS-PL30MA, Plasma Systems). The
assembled devices were sterilized by ultraviolet irradiation overnight
and then were pre-coated with 100 μg/ml collagen I (1:100, Corning) at
37 °C for 12 h to promote cell adhesion.
Cell culture on the NVU chip
Before seeding cells on the chips, SH-SY5Y cells were treated with
retinoic acid (10 µM) for 5 days and differentiated to a neuron-like
phenotype. Astrocytes and SH-SY5Y cells were mixed at a ratio of 10:1,
then were mixed with half volume of Matrigel on ice to get a final
density of 5 × 10^5 cells/ml. 60 μl cell-Matrigel suspension was
injected into the astrocytes/neurons culture compartment of the lower
brain side, and the chips were flipped immediately. The height of the
middle three channels was designed to be lower than that of two lateral
channels. Height difference between two channels could generate surface
tension in the junction while cell loading. The mixture flowed into the
middle three channels, stayed within the junction of the two
compartments and formed a well-defined interface after curing. Thirty
minutes later, the Matrigel was solidified and the chips were flipped
back. Then, 5 × 10^4 HBMECs (in 100 μl endothelial medium) were seeded
in the central compartment of the upper blood side. Next, 2.5 × 10^4
microglia were seeded in each lateral channel of the lower brain side.
Two hours later, 100 μl mixed medium (AM: DMEM/F-12: MEM = 1:1:1) was
infused into each lateral channel of the lower layer. Medium was
changed every day, and the chips were maintained in an incubator for
three days with 5% CO2 at 37 °C.
Virus
HSV-1 virus was provided by Professor Jumin Zhou’s lab (Kunming
Institute of Zoology, CAS). The HSV-1 infection experiments were
performed in a biosafety level-2 (BSL-2) laboratory.
Direct viral infection
For HSV-1 infection on the 3D NVU model (except for microglial
migration assay), 20 μl medium containing indicated multiplicity of
virus (MOI = 0.5) was infused in the lateral channels of the lower
brain side. Two hours after virus inoculation, cells on the NVU model
were washed twice with PBS and kept in a fresh medium for continued
culture.
Immunofluorescent imaging
Cells on the chips were washed twice with PBS and fixed with 4% PFA at
4 °C overnight. The fixed cells were permeabilized and blocked with
0.2% Triton X-100 in PBS (PBST) buffer containing 5% normal goat serum
for 30 min at room temperature. Antibodies were diluted with PBST
buffer and added into the upper and lower channels, respectively. Cells
were stained with the primary antibodies at 4 °C overnight and
corresponding secondary antibodies at room temperature for 2 h. The
chips were disassembled after counterstaining with DAPI, and the porous
membrane and bottom layer were mounted on slides. Images were acquired
using an Olympus FV3000 confocal fluorescent microscope. Image
processing was done using ImageJ software (NIH).
PBMC adhesion and infiltration assay
PBMC adhesion and infiltration assay on NVU model was performed at 2
dpi following HSV-1 infection (MOI = 0.5). Briefly, PBMCs were
collected from fresh peripheral blood and stained with CellTracker red
dye (Invitrogen, Cat no. [220]C34552) in RPMI 1640 medium at the
concentration of 5 µM at 37 °C for 30 min. Meanwhile, cells in NVU
model were stained with CellTracker green dye (Invitrogen, Cat no.
C7025). After washing with medium twice, the PBMCs were resuspended in
ECM medium to 3 × 10^6 cells/ml. 100 µl cell suspension was added in
the central compartment of the upper blood side and incubated at 37 °C
for 3 h. Then, the central compartment of the upper layer was washed
with PBS three times to wash out unattached PBMCs, and the whole chip
was fixed with 4% PFA. Finally, the attached PBMCs in the upper blood
side and infiltrated PBMCs in the bottom astrocytes/neurons compartment
were imaged using a fluorescent microscope, respectively.
BBB permeability assay
Permeability was assessed by detecting the diffusion of FITC-dextran
from the upper blood channel to the bottom brain channel. Briefly,
medium containing 50 μg/ml FITC-dextran was infused into the upper
blood channel of the chip. Two hours later, the medium in the upper
blood channel and the bottom brain channel were collected,
respectively. The FITC-dextran concentration was determined by FITC
fluorescence intensity using a microplate reader system at 488 nm
(excitation) and 525 nm (emission). The apparent permeability
coefficient (P[app]) was calculated according to the equation
below^[221]78,[222]79: P[app] = (V[b]*C[b])/(A*C[v]*t)
[MATH:
Papp=Vb
Cb
ACv<
/mi>t :MATH]
where V[b] (ml) and C[b] (ng/ml) indicate the volume (200 μl) and
concentration of FITC-dextran in the bottom brain channel,
respectively; A (cm^2) indicates the contact area between upper layer
and bottom layer (0.27 cm^2); C[v] (ng/ml) is the FITC-dextran
concentration in the upper channel; and t (s) is the diffusion time of
FITC-dextran (7200 s).
ELISA assay
Culture supernatant was harvested from the lower brain side at 3 dpi.
Cytokines in the culture supernatants, including IL-1β, IL-6, TNF-α,
CXCL-1 and IFN-β were analyzed using ELISA kits (Beyotime
Biotechnology, China) according to the manufacturer’s instructions.
Microglial migration assay
For microglial migration assay, HBMECs, astrocytes, neurons and
microglia were seeded on the NVU model for co-culture without microglia
seeded on the right lateral channel of the lower layer. Three days
later, HSV-1 (MOI = 0.1) was inoculated in the left lateral channel of
the lower layer, and 2.5 × 10^4 microglia labeled with RFP
(microglia-RFP) were seeded in the right lateral channel of the lower
layer. From 0 to 5 dpi, migration of microglia-RFP was recorded by
fluorescent microscope every day.
TCID50 assay
The tissue culture infectious dose 50 (TCID50) was determined as
described previously with minor modifications^[223]80. Briefly, 10,000
HMC3 cells were seeded per well in 96-well plates in 100 μl medium and
incubated overnight. HSV-1 containing medium was serially diluted
10-fold with fresh medium, then the diluted medium was used to
inoculate HMC3 cells, resulting in final HSV-1 dilutions of 1:10^2 to
1:10^8 on the cells in sextuplicates. Seventy-two hours later, the
cytopathic effect (CPE) was recorded under a microscope, and the TCID50
was calculated using the Reed and Muench method.
Autophagic flux detection
Autophagic flux in microglia was evaluated using an HBAD-mRFP-GFP-LC3
adenovirus (Hanbio Co. Ltd., Shanghai, China). Briefly, microglia on
the chips were transfected with the BAD-mRFP-GFP-LC3 adenovirus
(MOI = 10) for 4 h, then half of the medium containing adenovirus was
renewed with fresh normal medium. Forty-eight hours later, HSV-1 virus
(MOI = 0.5) was inoculated in the lateral channels of the lower layer.
Three days later, cells in NVU model were fixed with 4% PFA and
counterstained with DAPI. Images were acquired using a confocal
fluorescent microscope (Olympus FV3000). The number of yellow dots
(overlay of RFP and GFP) and mRFP dots was determined by manual
counting of fluorescent puncta in ten cells from three chips.
For evaluation assay of autophagy activators, 2 h after HSV-1
inoculation, medium containing virus in lower brain side was replaced
with fresh medium containing indicated chemicals for another 3-day
culture. The subsequent detection procedure was consistent with the
above experiment.
Drug testing
To test the drugs on the 3D NVU model after viral infection, 20 μl
medium containing HSV-1 (MOI = 0.5) was infused in the lateral channels
of the lower brain side. Two hours after virus inoculation, cells on
the NVU model were washed twice with PBS and kept in fresh medium
containing indicated chemicals (50 µM Acyclovir, 2 μM Rapamycin, 50 µM
3-Methyladenine, 20 μM Euphpepluone K, 20 μM Munronin V) for continued
culture. At 7 dpi, BBB permeability was examined by 4 kDa FITC-dextran
assay. Medium supernatants of the lower brain side were analyzed for
LDH and TCID50 assays. Cell viability was examined using LIVE/DEAD®
Viability/Cytotoxicity Kit, and viral infection was detected by
immunostaining for HSV-1 gG protein.
RNA extraction, library preparation, and sequencing
For RNA-seq experiments, HBMECs, astrocytes and microglia were seeded
on vascular channel, central compartment and lateral channels,
respectively. Three days after HSV-1 infection (MOI = 0.5) in the brain
side, HBMECs, astrocytes and microglia were collected, respectively.
Total RNAs of three cell types were extracted using TRIzol (Invitrogen)
following the methods of Chomczynski et al.^[224]81. RNA samples were
treated with DNaseI for DNA digestion and then determined by examining
A260/A280 using a Nanodrop OneC spectrophotometer (Thermo Fisher
Scientific Inc). RNA integrity was confirmed by 1.5% agarose gel
electrophoresis. Finally, qualified RNAs were quantified by Qubit 3.0
with a Qubit RNA Broad Range Assay kit (Life Technologies).
Samples with 2 μg total RNA were used for stranded RNA-sequencing
library preparation using a KC-Digital Stranded mRNA Library Prep Kit
for Illumina (Catalog NO. DR08502, Wuhan Seqhealth Co., Ltd. China),
following the manufacturer’s instructions. The kit eliminates
duplication bias in PCR and sequencing steps by using a unique
molecular identifier (UMI) of 8 random bases to label the pre-amplified
cDNA molecules. The library products corresponding to 200-500 bps were
enriched, quantified, and finally sequenced on a Hiseq X 10 sequencer
(Illumina).
RNA-Seq data analysis
Raw sequencing data were first filtered by Trimmomatic (version 0.36).
Low-quality reads were discarded, and the reads contaminated with
adapter sequences were trimmed. Clean reads were further treated with
in-house scripts to eliminate duplication bias introduced in library
preparation and sequencing. Briefly, clean reads were first clustered
according to the UMI sequences, in which reads with the same UMI
sequence were grouped into the same cluster, resulting in 65,536
clusters. Reads in the same cluster were compared by pairwise
alignment, and then reads with sequence identities of over 95% were
extracted to a new sub-cluster. Finally, sub-clusters were generated,
and multiple sequence alignment was performed to obtain one consensus
sequence for each sub-cluster. These steps eliminated any errors or
biases introduced by PCR amplification or sequencing.
The de-duplicated consensus sequences were used for standard RNA-seq
analysis. They were mapped to the reference genome of Homo sapiens from
the Ensembl database
([225]ftp://ftp.ensembl.org/pub/release-87/fasta/homo_sapiens/dna/)
using STAR software (version 2.5.3a) with default parameters. Reads
mapped to the exon regions of each gene were counted by featureCounts
(Subread-1.5.1; Bioconductor), and then RPKMs were calculated. Genes
differentially expressed between groups were identified using the edgeR
package (version 3.12.1). An FDR-corrected P-value cutoff of 0.05 and a
fold-change cutoff of 2 were used to judge the statistical significance
of gene expression differences. KEGG enrichment analysis of
differentially expressed genes were both implemented using KOBAS
software (version: 2.1.1) with a corrected P-value cutoff of 0.05 to
judge statistically significant enrichment. The gene set enrichment
analysis (GSEA) was performed via clusterProfiler (version 3.14.3) to
determine the KEGG signaling pathways. Annotated gene sets
(c2.cp.v7.2.symbols.gmt) were selected for the GSEA analysis. The gene
list input for the GSEA was ranked by values of log2 fold-change. The R
package ggplot 2 (version 3.3.3) was used for data visualization.
Transmission electron microscopy
Chip samples were fixed in PBS buffer containing 4% PFA (Electron
Microscopy Sciences) and 2.5% glutaraldehyde (Electron Microscopy
Sciences) at 4 °C overnight. After being washed with PBS three times
and fixed in 1% OsO[4] buffer for 2 h, the samples were dehydrated with
graded ethanol solutions and then embedded in Epon812 resin (SPI).
Then, 70-nm ultrathin sections were stained with 2% uranyl acetate for
30 min, followed by lead citrate for 10 min. Images were acquired using
a JEM-1400PLUS electron microscope.
Ethics
This study does not involve experiments involving animals, human
participants, or clinical samples.
Statistical analyses
Data were recorded using Excel (Microsoft) software. GraphPad Prism 8
software was used for statistical analysis. Differences between the two
groups were analyzed using unpaired two-sided Student’s t-test.
Multiple comparisons were performed using a one-way analysis of
variance (ANOVA) followed by the Bonferroni post hoc test. Data are
presented as the mean ± SD, and P < 0.05 was considered significant.
P-values are presented in the figures.
Reporting summary
Further information on research design is available in the [226]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[227]Supplementary Information^ (1.3MB, pdf)
[228]Reporting Summary^ (84.2KB, pdf)
[229]Transparent Peer Review file^ (1.4MB, pdf)
Source data
[230]Source Data 1^ (22.5KB, xlsx)
[231]Source Data 2^ (9KB, xlsx)
Acknowledgements