Cytokine secretion from brain macrophages infected with human immunodeficiency virus in vitro and treated with raltegravir

Chronic human immunodeficiency virus (HIV) infection has detrimental effects on central nervous system (CNS) neuronal health. The principal targets of productive HIV infection in the brain are microglia and monocyte derived macrophages [1–4]. In the periphery, HIV infection in CD4 T cells is cytopathic, resulting in fast CD4 T cell depletion [5], whereas in resting memory T cells, tissue monocytes, macrophages, and brain microglia with integrated HIV survive longer and serve as an HIV reservoir that persists indefinitely [6]. In adult humans, parenchymal microglia are infrequently exchanged with peripheral blood monocytes [7]. Perivascular cells in the CNS with macrophage markers have a higher turnover and are more frequently replaced by blood monocytes. Therefore, perivascular brain macrophages may serve as a source of HIV in the brain, and immunohistological and DNA hybridization studies have confirmed the presence of HIV proteins and DNA in these cells [8], which has been shown to be associated with depleted pre- and post- synaptic markers, synaptophysin and microtubule associated protein-2 (MAP2) [9].

HIV effects on the CNS have behavioral consequences. Neurocognitive impairment in HIV infected individuals is associated with elevated viral load and microglial activation [10, 11]. There are several prevailing hypotheses for the mechanisms of neural damage in HIV: direct neurotoxic effect from viral proteins and neural damage via intracellular signaling and secondary neural damage from cytokines and chemokines from infected microglia and brain macrophages [12]. Drugs used to treat HIV have variable neurotoxic effects, as examined by MAP2 staining in neuronal cultures, ranging from 17% to 52% reduction in MAP2 area. Of six combination regimens which included protease inhibitor, non-nucleoside reverse transcriptase inhibitor, and reverse transcriptase inhibitor, four caused significant reduction in neuronal MAP2 area [13]. As new antiretroviral drugs become available, it is important to continue assessing neurotoxicity.

Integrase inhibitors are a relatively new class of antiretroviral compounds, one of which is raltegravir (RAL), which has potent and durable antiviral activity similar to that of efavirenz, but which achieved HIV suppression to below detectable limits at a faster rate in a 24 week study [14]. HIV integration into the host genome is a multistep process involving: a) 3′ endonucleolytic cleavage of the 3′ end of DNA in the HIV pre-integration complex to a conserved CA dinucleotide, b) Strand-transfer involving concerted cleavage of host DNA and ligation of viral 3′ DNA to 5′ staggered site on host DNA, and c) DNA repair of gaps and DNA synthesis. RAL functions by inhibiting the strand-transfer step [15], thus preventing the induction of DNA repair machinery which is associated with innate immune activation and cell death [16]. We were therefore interested in determining the CNS efficacy of RAL by assessing: 1) neurotoxicity using in vitro neuronal cultures following methods comparable to a previous study [13], 2) inhibition of HIV infection in cultured microglia, and 3) cytokine production in RAL-treated, HIV infected microglia.

In primary human neuroglial cultures, we found that RAL is not neurotoxic and we present evidence for reduced astrogliosis. In primary human HLA-DR-positive brain macrophages and microglia, we measured secretion of seven cytokines across 9 d of in vitro HIV infection, and found reduced rate of TNF-α, IL-10, and IL-8 production in those treated with 20 nM RAL.

In this study, we measured cytokine secretion from human microglia under four conditions, 1) No-treatment control, 2) Infected with HIV, 3) Infected with HIV but treated with 20 nM RAL, and 4) 20 nM RAL alone. The concentration of 20 nM RAL was chosen on the basis of evaluating several pharmacokinetics studies. The mean trough plasma concentration from RAL once-daily parmacokinetics studies was 40 nM [22], while an independent study calculated a CSF:Blood Plasma ratio of 0.058 with a median (over 24 hr) plasma concentration 540.7 nM (31.36 nM in CSF) [23]. One additional study evaluated intracellular RAL concentration and calculated a 24 hr area under the curve (AUC) 1,884 nMxh [24], which is averages to roughly 78.5 nM. Therefore 20 nM RAL seemed to be a reasonably relevant concentration to assess in vitro effects on microglia.

We found that RAL administered at 20 nM was effective at suppressing HIV infection in microglia (Figure 1) for at least 9 d. Only IL-8, IL-10, and TNF-α were above the detection limit. We calculated the rate of cytokine production for all three cytokines across the different treatment groups (Table 1). The mean IL-8, IL-10, and TNF-α concentrations and linear regression, separated by conditions are shown in Figure 2. For IL-10, IL-8 and TNF-α, there was a significant effect of RAL on the cytokine production in the context of HIV infection. Alone, the RAL-treated microglia had the highest concentration of IL-8, IL-10, and TNF-α. However, in the context of HIV infection, RAL-treated microglia had lowest production of TNF-α and IL-8. This makes sense because TNF-α autocrine signaling leads to IL-8 production via NF-kB. However, it is important to note that the highest production of TNF-α was in the presence of RAL alone (Table 1). IL-8 and IL-10 production were lowest in RAL treated microglia in the context of HIV, but not significantly different among the other groups (Table 1).

It is remarkable that HIV and RAL alone induced cytokine expression, but in combination, was below control levels. This implies that RAL may be pro-inflammatory alone, and in the presence of HIV replication complex it is anti-inflammatory. One possibility is off-target effects on DNA-binding proteins, mimicking NF-kB - like activation, however in the presence of HIV-replication complex, it is bound and not interacting with off-target proteins.

We assessed cytotoxicity through measuring LDH-release in HIV-infected cells exposed to RAL and found background levels cytotoxicity due to HIV-infection, which was neither enhanced or diminished by RAL. The mean ± standard deviation cytotoxicity in HIV-alone was 43 ± 7.9%; and for HIV + RAL was 45 ± 3.2%. This may have been due to the protocol, which was acute infection via exposure to high levels of HIV for 4 hr, followed by media change including the drug. RAL may have suppressed further integration and infection of more cells with HIV, but it did not prevent lysis/death of those already infected.

Activation of the IL-8 gene is enhanced by signaling from NF-kB, and based on reduced IL-8 secretion in the HIV + RAL treated cultures, we hypothesized lower NF-kB activation and quantitated the nuclear translocation of NF-kB after 9 d infection, we calculated the percent-nuclear NF-kB compared to total NF-kB. The proportion of NF-kB localized to the nucleus was significantly lower in the HIV + RAL than the HIV + alone. Nuclear localization of phosopho-NF-kB was quantitated as percent-nuclear, then arcsine transformed for comparison by Dunnett’s test vs. Control. RAL alone (21%) and HIV + RAL (22%) were not significantly different from Control (20%), while HIV + alone was significantly higher from Control (28%, P = 0.02).

There are several cellular markers of neurons and astrocytes in cell culture, one of which is class-III β-tubulin, a structural microtubule protein specifically expressed in the cell body, axon, and dendrites of neuronal cells. To determine the neurotoxicity, we measured the area of coverage of β-III tubulin, similar to Robertson et al. [13], because retraction or loss of neurites would be detected by this method. In order to measure the relative amount of astrocytes in the culture, we similarly measured the area covered by the astrocyte-specific protein, GFAP. Diffuse astrocytosis was observed during HIV associated dementia and HIV-encephalitis [25], and in the CNS of transgenic mice expressing HIV proteins [26] and is a marker of astrogliosis or astrocytosis. In the absence of RAL, neuronal cultures had 110 ± 41 μm² β-III tubulin and 244 ± 49 μm² GFAP. While with 20 and 100 nM RAL, there was no significant effect on β-III tubulin, with 142 ± 48 and 89.4 ± 34 μm², respectively, with P = 0.17 and P = 0.43 as assessed by Dunnett’s test vs. Control. However, with 20 and 100 nM RAL, there was a significant effect on GFAP, with 107 ± 26 and 115 ± 163 μm², respectively, with P = 0.0094 and P = 0.01 as assessed by Dunnett’s test vs. Control. Thus, RAL had no significant effect on β-III tubulin area at 20 or 100 nM and an inhibitory effect on GFAP area. The possibility that RAL is toxic to astrocytes should be noted, considering the important role that they play in maintaining the blood brain barrier and neuronal maintenance. Figure 3 shows the quantitation of β-III tubulin and GFAP area as well as representative photomicrographs.

The CNS has been proposed as a compartment for a latent HIV reservoir which has long-term neurological effects with downstream neurocognitive consequences, attributed to cytokine production by infected perivascular CNS macrophages and parenchymal microglia [10]. Therefore, understanding whether antiretroviral compounds themselves are neurotoxic becomes important when infected individuals are on therapy for decades. The purpose of the present study was to assess whether RAL, an inhibitor of strand-transfer, has neurotoxic properties and whether RAL slows the rate of cytokine secretion by HIV infected microglia. We found that RAL is not neurotoxic at 20 or 100 nM and that it significantly inhibits cytokine secretion in HIV infected microglia in vitro.

Robertson et al.[13] found that exposure of primary human neurons caused statistically significant reduction of MAP2, a dendritic marker of mature neurons, when exposed to some antiretroviral drugs. Individually, abacavir, 2′,3′-dideoxyinosine, and nevirapine are predicted to have relatively high risk of neurotoxicity, with at least a 10% drop in MAP2 intensity at observed typical cerebrospinal fluid (CSF) concentrations of the drugs [13]. Median RAL concentration in the CSF of treated HIV patients was observed to be 14.5 ng/mL (30.05 nM) [23], which is within range of the concentrations we tested for neurotoxicity. This suggests that RAL has relatively low neurotoxic risk.

We measured cytokine production by withdrawal of supernatant from infected microglia cultures and quantification using a multiplex assay and found production of IL-8, IL-10, and TNF-α, which increased linearly. RAL on its own caused significant increase in the rate of production of these three cytokines, mainly IL-10, an anti-inflammatory cytokine; but resulted in a significant decrease when administered with HIV. The most abundant was IL-8, a molecule with potent chemotaxic properties for neutrophils [27] and monocytes [28]. IL-8 gene transcription is induced by NF-kB activation and nuclear translocation, and is dependent on TNF-α. One preliminary positron emission tomography (PET) study found slightly increased retention of the PET ligand ¹¹C-PK11195, which binds to activated microglia, in neurologically asymptomatic and ARV-treated HIV-infected individuals [11]. One human study assessed IL-8 concentration in CSF of patients with HIV associated dementia compared to HIV-infected patients with no neurocognitive impairment and found higher IL-8 in CSF of those with dementia [29]. In our experiments, phospho-NF-kB nuclear translocation was significantly higher only in the HIV-infected cultures, not in Control, RAL-alone, or HIV + RAL.

We also observed production of IL-10, an anti-inflammatory cytokine which inhibits T cell proliferation and is putatively produced early in HIV infection via NF-kB signaling [30], which would be an antiviral action, allowing the incipient reservoir to avoid detection.

In comparison with nucleoside reverse transcriptase inhibitors, tenofovir and zidovudine, RAL has relatively modest effects on monocyte cytokine production. Zidovudine dose-dependently increased secretion of IL-8 and tenofovir decreased it. Likewise, the anti-inflammatory IL-10 was reduced in the presence of tenofovir while zidovudine did not affect it [31].

One possible explanation for reduced TNF-α, IL-8, and IL-10 production in our cultures is simply due to fewer infected cells. Another possible explanation is that inhibition of the strand transfer step of HIV integration prevents the DNA damage and repair process from being initiated, which may lead to downstream microglial activation. One weakness to the present study is the lack of comparison with another antiretroviral compound or lack of testing for combinations. Additionally, with measuring the cytokines at one-day intervals, it is difficult to determine the order of events except by interpolating from what is already known. An important future experiment would be to assess the effect of conditioned media from brain macrophage cultures on neuronal and glia cultures.

Results from this study suggest a low probability for neurotoxicity of RAL and likely neuroprotective effect in HIV-infection by suppressing the production of chemotaxic inflammatory cytokine, IL-8.