Background
It has been established that HIV-1 crosses the blood-brain barrier (BBB) and enters the brain in the early stages of infection, which confers to the virus protection from the immune system and certain antiretroviral drugs [
1]. While the existing combined antiretroviral therapy (cART) has considerably improved the patients’ quality of life and prevented the progression of the disease and the occurrence of the most severe forms of HIV-1-associated neurocognitive disorders (HAND), the incidence of the mildest forms of neuropathologic abnormalities has actually increased. Moreover, cART fails to completely eliminate HIV-1 in reservoirs and viremia rebounds upon treatment interruption due to the reactivation of latent HIV-1.
The so-called "shock and kill" approach was proposed to eliminate latently infected cells persisting despite long-term effective cART [
2]. In this strategy, latent HIV-1 would be reactivated by latency-reversing agents (LRA) (shock) and the infected cells would then be eliminated by the immune system (kill) while cART would protect from new rounds of virus infection. However, none of the current LRA target exclusively latently infected cells, and although their effects have been extensively studied in vitro in the blood compartment [
3], the consequences of these treatments in the central nervous system (CNS) are still unclear. Importantly, the outcome of the “shock and kill” approach could be adversely affected by some unique characteristics of the CNS such as (i) constrained LRA penetration which may limit the “shock” [
4], (ii) reduced immune surveillance which may compromise the “kill” [
5], (iii) extensive virus compartmentalization leading to genetically distinct variants responding differently to LRA [
6,
7], and (iv) altered cART bioavailability which may allow sustained viral replication [
8]. Our group has previously shown that the LRA bryostatin-1 and prostratin, two protein kinase C (PKC) agonists, caused inflammation and disruption of the BBB, as well as alteration of leukocyte adherence and transmigration [
9]. Thus, a careful assessment of the potential impact of LRA on long-lived infected cells of the CNS such as astrocytes is essential to evaluate if the “shock and kill” approach would be appropriate or even viable for the brain reservoir.
The aim of this manuscript is to evaluate the overall safety of different LRA in a tightly regulated microenvironment such as the CNS. For this purpose, we treated human astrocytes with various LRA that can cross the BBB and analyzed their overall effects on some specific cellular functions. Our results demonstrate that bryostatin-1 induces astrogliosis and disturbs the astrocytic glutamate uptake/release balance, which can lead to excitotoxicity. Moreover, bryostatin-1 could induce neuroinflammation since we report that it drives secretion of certain chemotactic factors and proinflammatory cytokines such as CCL2 (also known as monocyte chemoattractant protein-1, MCP-1), interleukin-6 (IL-6), IL-8, and granulocyte-macrophage colony-stimulating factor (GM-CSF) by astrocytes. Using an in vitro BBB model, we show that transmigration of neutrophils is increased in response to bryostatin-1 as well as neutrophil extracellular trap formation. Taken together, these results suggest that bryostatin-1 could induce an inflammatory syndrome in the brain that could eventually lead to neurological disorders.
Methods
HIV-1 LRA and positive controls used in this study
BIX-01294 (2-(hexahydro-4-methyl-1H-1,4-diazepin-1-yl)-6,7-dimethoxy-N-[1-(phenylmethyl)-4-piperidinyl]-4-quinazolinamine trihydrochloride hydrate), SAHA (suberoylanilide hydroxamic acid), HMBA (N,N′-hexamethylene bis(acetamide)), disulfiram (tetraethylthiuram disulfide), bryostatin-1, and JQ1 were all purchased from Sigma-Aldrich. The positive controls Pam3Csk4 and poly(I:C) were purchased from InvivoGen.
Virus production
Virions were produced by calcium-phosphate transient transfection in 293T cells as previously described [
10]. As astrocytes do not express the primary cell surface receptor for HIV-1 (i.e., CD4), we used VSV-G-pseudotyped HIV-1-based reporter viruses for our infection experiments by co-transfecting 293T with pHCMV-G and NL4-3 eGFP-IRES-Nef
env [
11]. Infectivity of our virus stocks was assessed using the genetically modified HeLa-derived indicator cell line TZM-bl [
10].
Cell culture
The hCMEC/D3 cell line was kindly provided by Dr. Pierre Couraud (Institut Cochin) under the license from the Institut National de la Santé et de la Recherche Médicale (INSERM). This immortalized human BMVEC line possesses the morphological and functional characteristics of cerebral endothelial cells [
12] and was maintained in culture as described previously [
9]. Briefly, hCMEC/D3 cells were grown in endothelial basal medium-2 (EBM-2; Lonza Group Ltd.) supplemented with 5% fetal bovine serum (Corning Life Sciences), 1% penicillin-streptomycin solution (GIBCO Life Technologies, Invitrogen), 1.4 μM hydrocortisone, 5 μg/mL ascorbic acid, 10 mM HEPES (all three from Sigma-Aldrich), 1% chemically defined lipid concentrate (GIBCO Life Technologies), and 1 ng/mL of basic fibroblast growth factor (ProSpec-Tany Technogene Ltd.). Passage numbers 28-34 were used throughout our experiments.
Neutrophils were purified from human blood samples as described previously [
13]. Briefly, blood samples were centrifuged to concentrate the blood and remove platelets before erythrocyte sedimentation in 2% dextran T-500 followed by centrifugation on Ficoll-Paque cushion. Contaminating erythrocytes were removed by hypotonic lysis for 10 s. Neutrophils were resuspended in a medium containing 50% complete EBM-2 and 50% DMEM supplemented with 10% fetal bovine serum.
Isolation and purification of astrocytes
Human fetal astrocytes were isolated from fetal brain samples (15 to 24 gestational weeks) as previously described [
9,
14]. Briefly, blood vessels and meninges were removed from the fetal brain tissues. Thereafter, the tissues were minced and treated with 0.2 mg/mL DNase I (Roche) and 0.25% trypsin (GIBCO Life Technologies) for 30 min before being passed through a 70-μm cell strainer (Corning). The flow through was plated in T75 tissue culture flasks for adherent cells (Sarstedt) at a final concentration of 6–8 × 10
7 cells/flask in MEM supplemented with 10% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, 0.3 mg/mL
l-glutamine, 1 mM sodium pyruvate, 1× MEM nonessential amino acids, 0.5 μg/mL amphotericin B (all from GIBCO Life Technologies), and 0.1% dextrose (Sigma-Aldrich). Astrocytes were grown at 37 °C under a 5% CO
2 atmosphere and left untouched for 2 weeks before being passaged once a week. To ensure cell purity, all experiments were conducted on the third or fourth passage.
Infection with HIV-1 (VSV-G) pseudotypes and treatment with LRA
Astrocytes were seeded in 24-well plates at 2 × 10
5 cells/well in X-VIVO™ 20 hematopoietic serum-free culture medium (Lonza) and cultured for 24 h before virus infection. Cells were then incubated with VSV-G-pseudotyped HIV-1 (MOI 0.1) for 24 h before being washed thoroughly to remove unbound viral particles and left untouched for six more days to ensure there is no more residual viral activity from the virus inoculum and initial cell responses to acute infection are completed [
15]. Cells were then treated with various LRA for 24 h at subcytotoxic doses, which allow HIV-1 reactivation based on previous in vitro and clinical studies.
MTS assay
The metabolic activity of astrocytes was assessed using CellTiter 96™ AQueous Nonradioactive Cell Proliferation Assay following the manufacturer’s instructions (Promega). Absorbance at 490 nm was measured using an ELX808 microplate reader (Biotek instruments).
Quantification of mRNA and protein levels
Total RNA was extracted from astrocytes using the RNeasy Kit (Qiagen) according to the manufacturer’s instructions and reverse transcribed to cDNA with M-MLV RT Polymerase (Promega). Cytokine expression was assessed by SYBR-Green Quantitative RT-PCR using QuantStudioTM 6 Flex Real-Time PCR System (Thermo Fisher Scientific) following the manufacturer’s instructions. Primer (used at 0.4 μM) sequences were forward 5′-CCCCAGTCACCTGCTGTTAT-3′ and reverse 5′-TGGAATCCTGAACCCACTTC-3′ for CCL2, forward 5′-TAGCAAAATTGAGGCCAAGG-3′ and reverse 5′-AAACCAAGGCACAGTGGAAC-3′ for IL-8, forward 5′-CCTTCCAAAGATGGCTGAAA-3′ and reverse 5′-CAGGGGTGGTTATTGCATCTC-3′ for IL-6, forward 5′-AAAAGGGGCGCAACAAGTTC-3′ and reverse 5′-GATGCCTTCCGGGTTCTCAA-3′ for C3, and forward 5′-TAGAGGGACAAGTGGCGTTC-3′ and reverse 5′-CGCTGAGCCAGTCAGTGT-3′ for 18S. Amplification of target genes was normalized to the geometric mean of 18S ribosomal cDNA. A standard curve was drawn for each gene of interest using serial dilutions of pooled cDNA from all samples. Levels of CCL2, IL-6, IL-8, GM-CSF, CCL5, and TNF produced by astrocytes were determined by ELISA MAX™ Deluxe assays (Biolegend), and IL-1β concentrations were assessed by ELISA Ready-Set-Go! (eBioscience) following the manufacturer’s instructions. IFNα/β levels were determined using HEK-Blue™ IFN-α/β indicator cells per manufacturer’s instructions (InvivoGen).
Quantification of TGFβ1
Levels of transforming growth factor beta 1 (TGFβ1) were assessed using the PAI-1/luciferase assay as previously described [
16]. Briefly, transfected mink lung epithelial cells (MLEC) kindly provided by Dr. Daniel Rifkin (New York University Medical Center) were plated in a 96-well plate at 1.6 × 10
4 cells per well in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. Thereafter, 100 μL of cell supernatant from astrocytes was added directly to cells. Recombinant TGFβ was used as standards (15.625 to 500 pg/mL; 100 μL/well). Samples and standards were incubated overnight at 37 °C under a 5% CO
2 atmosphere. Cells were then lysed, and luciferase activity was measured using a Varioskan flash multi-mode reader (Thermo Fisher Scientific).
Astrocyte glutamate uptake
Glutamate uptake by astrocytes was monitored using the Glutamate assay kit (Abcam). Briefly, mock- and HIV-1-infected astrocytes were either left untreated or treated for 24 h with JQ1, bryostatin-1, or both JQ1 and bryostatin-1 in X-VIVO culture medium. Glutamate levels in the medium were quantified after 2, 4, 8, and 24 h of treatment following the manufacturer’s instructions. A standard curve was drawn using serial dilutions of glutamate.
In vitro human BBB model system
We used a co-cultivation model including hCMEC/D3 cells and astrocytes seeded on each side of a porous insert allowing cell-cell contacts as previously described [
9,
17]. Briefly, cell culture inserts for 24-well plates with 3.0-μm pore translucent PET membrane (Corning Life Sciences) were coated with 150 μg/mL collagen-1 (Sigma-Aldrich). Astrocytes were then seeded on the basal side of the membrane and incubated for 4 h before hCMEC/D3 cells were seeded on the upper side of the membrane. Cells were allowed to grow in 150 μL (upper chamber) and 750 μL (collector) of BBB medium, composed of 50% complete EBM-2 and 50% DMEM supplemented with 10% FBS, during 6 days to reach confluence. BBB integrity was assessed by measuring permeability to dextran-rhodamine. The culture medium in the upper chamber was replaced with BBB medium supplemented with 1 mg/mL 70-kDa dextran-rhodamine (Life Technologies). After 6 h, the fluorescence intensity in collectors was measured using a Varioskan flash multi-mode reader (Thermo Fisher Scientific). Samples displaying dextran-rhodamine permeability superior to 20% of the empty control were discarded.
Quantification of neutrophil transmigration
BBB-containing inserts were placed into a 24-well plate containing astrocyte-conditioned medium (ACM) before addition of 2 × 106 fleshly isolated neutrophils in the upper chamber. After 4 h of cell transmigration, the plate was centrifuged for 2 min at 100×g in order to collect neutrophils bound to astrocytes on the basal side of the insert. Next, inserts were discarded and the absolute number of neutrophils in the collector was counted by flow cytometry using 123count eBeads (eBioscience). In some experiments, neutrophils were pretreated with the noncompetitive allosteric inhibitor of IL-8 reparixin (1 μM) (Cayman Chemical) for 45 min before performing the transmigration assay.
Microscopy-based detection of NETs
Transmigration assays were performed as described above but using as collector a μ-Plate 24-well ibiTreat (Ibidi) allowing fluorescent microscopy. In parallel, 1 × 106 neutrophils were cultured in the same μ-Plate and remained untreated or treated for 4 h with bryostatin-1 (25 nM) or PMA (100 nM). Thereafter, DNA was labeled with 1× GreenGlo™ Safe DNA Dye (Denville Scientific Inc.) for 30 min at room temperature. Next, neutrophils and NETs were visualized with a fully automated inverted Leica DMI6000 B microscope (Leica Microsystems). Images were acquired using both FITC and UV filter cubes. In some experiments, DNA was labeled with NucBlue® and imaged using a UV filter cube. Image acquisition was done using Volocity Software Version 5.4.0 (PerkinElmer).
Statistical analysis
Means of raw data were compared using either two-tailed paired Student’s t test or one-way ANOVA with the appropriate post-test or the non-parametric Friedman test. P values < 0.05 were deemed statistically significant. Calculations were performed with the GraphPad PRISM 7 software for Windows (GraphPad Software).
Discussion
The use of LRA to reactivate HIV-1 from latently infected cells has been proposed as part of the “shock and kill” therapeutic strategy, which is aimed at achieving a sterilizing cure [
2]. Although many reports have described the effectiveness of various LRA in reactivating the latent virus in CD4
+ T cells [
3], the possible eradication of brain reservoirs with this strategy is still undefined. To our knowledge, only one study has assessed the impact of LRA on some functions of CNS cell types such as BMVEC [
9] and the present work is extending these investigations to astrocytes, the most abundant cell type in the brain.
Our initial findings demonstrate that certain basic functions of astrocytes are not affected by LRA. Indeed, LRA do not disturb astrocyte metabolic activity and do not trigger ROS production. However, we report that bryostatin-1, when used either alone or in combination with other LRA, triggers secretion by astrocytes of several chemokines and proinflammatory cytokines such as CCL2, IL-6, IL-8, and GM-CSF. The LRA-mediated effect on production of such soluble factors is seen in both uninfected and HIV-1-infected astrocytes and is exerting its effect at the mRNA level. This observation is of high clinical significance given that a very low percentage of HIV-1-infected astrocytes is detected under in vivo conditions. Bryostatin-1 is a PKC modulator that has been recently tested in a clinical trial for HIV-1 reactivation [
28]. Plasma concentrations reported by Gutierrez and colleagues after a 20 μg/m
2 single-dose administration were 100-fold lower than bryostatin-1 concentrations used in previously described in vitro studies including the current one [
9,
29‐
31]. However, Gutierrez and co-workers did not detect any HIV-1 reactivation at this very low plasma concentration. Thus, higher doses of bryostatin-1 will be necessary to reactivate the virus in vivo and the adverse effects described in our work might be reached under such conditions. Indeed, as previously mentioned, astrocytes release CCL2, IL-6, IL-8, and GM-CSF in response to a treatment with bryostatin-1. All these soluble factors can exhibit neuroprotective effects in the CNS during some trauma or diseases. CCL2, by reducing glutamate levels increased by
N-methyl-
d-aspartate (NMDA) and Tat protein, and IL-6, by preventing ROS and Ca
2+ excitotoxicity in Parkinson’s and Huntington’s diseases, exert neuroprotective properties [
32,
33]. GM-CSF participates in the neuronal repair after a traumatic injury to the CNS [
34] and IL-8 may increase neuronal survival after a traumatic brain injury by promoting the production of nerve growth factor (NGF) by astrocytes [
35]. However, it has also been shown that a CCL2 exposure of murine astrocytes and BMVEC induced a structural change of the actin cytoskeleton and a redistribution of tight junction proteins leading to a more porous BBB and a facilitation of leukocyte migration into the CNS [
36]. Moreover, Ehrhart and colleagues showed significant increases in levels of IL-8 in the blood from amyotrophic lateral sclerosis (ALS) patients compared to controls, these higher levels of IL-8 being correlated with disease progression [
37]. It has also been demonstrated that IL-6 levels are elevated in the CSF of patients with chronic schizophrenia. Moreover, IL-6 is able to activate the kynurenine pathway leading to an increase of the kynurenic acid production, a compound that has been consistently reported at elevated levels in the CSF or in the post-mortem brain of patients with schizophrenia [
38]. Finally, McQualter and co-workers have detected the formation and expansion of inflammatory lesions within the CNS during experimental autoimmune encephalomyelitis, a mouse model for human multiple sclerosis (MS) that is highly dependent on GM-CSF [
39]. All these studies and many others have established or suggested detrimental roles in various brain disorders and their progression (MS, Alzheimer disease (AD), cerebral stroke, ALS, and more) and BBB permeability for these four cytokines that are all upregulated by bryostatin-1 in astrocytes [
32,
33,
40,
41].
Astrogliosis is an astrocyte response to neuronal insults due to injury or disease. Although constitutive astrogliosis exerts beneficial functions for the brain such as restriction of CNS inflammation, neuronal protection, BBB repair, and wound closure, this process can also lead to harmful effects such as exacerbation of inflammation or interference with synapse sprouting or axonal growth [
42]. Human astrogliosis is commonly assessed in vivo by an increase of GFAP expression, which was never shown in vitro. Thus, even if we did not observe any increase in GFAP expression after LRA treatments, bryostatin-1 might induce astrogliosis. Indeed, it has been suggested that (i) bryostatin-1 activates latent HIV-1 through a PKC- and NF-κB-dependent mechanism [
30] and (ii) reactive astrocytes might be induced by NF-κB signaling [
24]. It is thus possible that bryostatin-1 induces the so-called A1 neuroinflammatory reactive astrocytes. Our results showing a bryostatin-1-mediated increase in C3 expression confirm this postulate as C3 is specifically upregulated in inflammatory reactive astrocytes and has therefore been proposed as a marker of astrogliosis [
24].
It is well known that a tight regulation of brain glutamate concentrations is vital, as elevated levels of extracellular glutamate may lead to excitotoxicity [
43]. We demonstrate here that the glutamate concentrations are significantly higher in supernatants from bryostatin-1-treated astrocytes compared to untreated cells. Considering that astrocytes maintain glutamate homeostasis by a delicate balance between uptake and release processes, it can be proposed that bryostatin-1 may disrupt this equilibrium by either decreasing glutamate uptake and/or increasing its release by astrocytes. Excitatory amino acid transporters 1 and 2 (EAAT1/GLAST and EAAT2/GLT-1, respectively) expressed at the astrocyte plasma membrane are responsible for extracellular glutamate uptake by astrocytes. Their cellular localization is controlled by PKC, and it has been shown that PKC activation drives internalization of EAAT1/2 to endosomal compartments [
44]. As bryostatin-1 is a PKC activator, the observed increase in extracellular glutamate concentration could be caused by internalization of the transporter. Compared to glutamate uptake, astrocytic glutamate release is far more complex and can occur by numerous mechanisms [
45] in which the precise bryostatin-1- or PKC-dependent modulatory effects are still unknown. If an occasional use of bryostatin-1 should not lead to a chronic excitotoxicity, it could induce an acute excitotoxicity, which is mainly mediated by an increase of extracellular glutamate levels [
43].
We also demonstrate that ACM from bryostatin-1-treated astrocytes causes a significant transmigration of neutrophils through an in vitro BBB model. It is possible that neutrophil transmigration is facilitated by a bryostatin-1-mediated disruption of the BBB as we previously described [
9] and the elevated production of IL-8 by astrocytes in response to the LRA. Neutrophils are essential for protective immunity during infection and tissue repair. However, by their release of antimicrobial proteins, proteases, and oxidants during the inflammatory response, recruited neutrophils may also cause severe side effects in the brain. For example, when activated in the inflammation site, neutrophils release CCL20 that mediates recruitment of Th17 cells [
46]. In turn, Th17 cells, along with migrating neutrophils, produce Il-17, which is toxic to neurons, induces BBB breakdown by decreasing the expression of tight junction proteins, promotes monocyte infiltration into the CNS via an ICAM-1-dependent mechanism, and activates glial cells that could release mediators contributing to brain damage [
26,
47]. Furthermore, our data show that apart from inducing neutrophil transmigration across the BBB, bryostatin-1 induces NET formation in the recruited neutrophils. Since bryostatin-1 is a PKC activator similar to PMA, a well-known NETosis inducer, this observation is not surprising. NETosis is a defense mechanism by which neutrophils eliminate pathogens by releasing their chromosomal DNA, histones, and granule contents to the extracellular space [
27]. It has been shown that NETs act in the antiviral response against HIV-1 through virus capture and neutralization via production of α-defensin and myeloperoxidase, and would be therefore a potential mechanism to protect against HIV-1 [
48]. Although there is a paucity of data about the formation of NETs in humans, it has nonetheless been suggested that neutrophils and NETs may play a role in AD pathology as they have been observed close to the amyloid β plaques in the CNS of AD patients [
26]. Moreover, components released during NETosis contribute to the loss of BBB integrity [
49], and NET formation may play a role in systemic lupus erythematosus through the induction of type-I IFN production [
50].
Bryostatin-1 is actually in a phase II clinical trial (NCT02431468) to assess its potential for the treatment of moderately severe to severe stages of AD. However, we estimated that the 40-μg dose administrated in this trial is similar to the 20 μg/m2 dose used in the clinical trial conducted by Gutierrez and his group. Thus, this low concentration of bryostatin-1 may be safe and efficient enough for the treatment of AD, but as specified above, this dose would not be sufficient per se to reactivate latent HIV-1 and higher doses could lead to the adverse effects described in the present work.
Interestingly, JQ1 presents quite different outcomes than bryostatin-1. For instance, JQ1 inhibits secretion of multiple soluble factors (i.e., CCL2, IL-6, IL-8, and GM-CSF) by astrocytes and suppresses neutrophil recruitment across an in vitro BBB while it does not induce astrogliosis and has no effect whatsoever on glutamate uptake/release and NETosis. JQ1 is also the only LRA tested, when used alone or in combination with bryostatin-1, that induces production of IFNα/β by astrocytes (data not shown). JQ1 is a small-molecule inhibitor of BET bromodomain (BRD) protein binding that has been shown to abolish inflammation, endothelium activation, and leukocyte transmigration [
51,
52]. It has been reported that, under inflammatory stimuli, the transcription factor NF-κB recruits the member of BET family BRD4 [
52]. Both NF-κB and BRD4 direct the formation of dynamic super-enhancers that control transcription of genes driving the inflammatory response. By inhibiting BRD4, JQ1 prevents formation of those super-enhancers and acts as an anti-inflammatory molecule. Moreover, a pretreatment with JQ1 inhibits the phenotypic features of BMVEC proinflammatory activation by reducing leukocyte rolling and neutrophil transmigration both in vivo and in vitro using EC monolayers [
52]. In our hands, we noticed a decrease in the number of transmigrated neutrophils in response to ACM from JQ1-treated astrocytes in an in vitro BBB experimental model made of BMVEC and astrocytes (Fig.
5b, c) but not with a BBB containing only BMVEC (Fig.
5d, e). This suggests that the inhibitory activity of JQ1 on neutrophil transmigration through the BBB is mainly due to a modulatory effect on astrocyte-mediated secretion of chemokines in response to JQ1.