Tumor necrosis factor-related apoptosis-inducing ligand induces neuronal death in a murine model of HIV central nervous system infection

Reconstitution and HIV-1 Infection of hu-PBMC-NOD-SCID Mice.

NOD (NOD/Shi) scid/scid (NOD-SCID) mice were used at 6–8 weeks of age. The experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of the participating institutions. Reconstitution with human PBMCs and infection with HIV-1 were performed as described (9). Briefly, human PBMCs were isolated from a healthy HIV-1-seronegative donor and injected i.p. Five days later, mice were inoculated i.p. with 1,000 ID₅₀ of JRFL HIV-1 isolate (11), 100,000 ID₅₀ of NL4-3 (12), or recombinant HIV-1 expressing green fluorescent protein (GFP). The T-tropic HIV-1 expressing GFP (NL-EGFP) was constructed from NL4-3 by inserting an enhanced GFP (EGFP) gene and an internal ribosome entry site (IRES) sequence between gp41 and the nef sequence by PCR-based subcloning. The ATG codon of EGFP was placed 2 bp downstream of gp41 termination codon and nef expression was rescued from insertion of the IRES sequence. The M-tropic HIV-1 expressing GFP (NL-CSFV3-EGFP) was constructed by replacing the V3 sequence in the NL-EGFP with the V3 sequence from JRCSF (11). One hundred micrograms of LPS (Sigma) was injected i.p. 7 days after HIV-1 inoculation. Three or 21 days later, the brains were fixed in 4% periodate-lysine-paraformaldehyde (PLP) fixative. The brains were cut into four serial coronal sections at the level of corpus callosum, hippocampus, and cerebellum. Then they were subjected to immunohistological analyses.

Administration of mAbs.

One milligram of anti-human TRAIL mAb (clone RIK-2, IgG; ref. 13), anti-human Fas ligand (FasL) mAb (clone NOK-1, IgG; ref. 14), anti-human TNF-α mAb (clone 28401.111, Genzyme/Techne, Minneapolis), or a control mouse IgG (Inter-Cell Technologies, Hopewell, NJ) was administered i.p. at the time of the LPS injection. The mice were killed 3 days later, and the brains were subjected to histological analyses.

Immunohistological Staining.

Abs against the following molecules were used for PLP fixed section as primary Abs: human CD3 (rabbit polyclonal IgG, Dako), human CD68 (clone PGM1, mouse monoclonal IgG, Dako), glial fibrillary acidic protein (GFAP) (rabbit polyclonal IgG, Dako), HIV-1 gag p24 (clone Kal-1, mouse monoclonal IgG, Dako), and mouse F4/80 (clone F4/80, rat monoclonal IgG, Serotec). Detection of the immunohistological staining was performed as described (10). Nonimmunized rabbit, mouse, and rat Ig (Dako) were used as negative controls. Terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick-end labeling (TUNEL) staining of frozen sections was carried out using an indirect method as described (10). Absence of either digoxigenin-labeled dUTP or TdT served as a negative control. DNase-treated sections served as positive controls.

Frozen sections were incubated with Abs against human CD68 (clone PGM1), human TRAIL (clone RIK-2), IL-1β (rabbit polyclonal IgG, Endogen, Woburn, MA), human TNF-α (rabbit polyclonal IgG, Endogen), human FasL (rabbit polyclonal IgG, Santa Cruz Biotechnology), HIV-1 gag p24 (goat polyclonal IgG, ViroStat, Portland, ME), or the active form of caspase-3 (rabbit polyclonal IgG, R & D Systems). The sections were then incubated with FITC-conjugated secondary Abs as described (10). Dual-color staining, including TUNEL, was performed FluoroNissl Green (FNG, FMP) (15), NeuroTrace blue fluorescent Nissl stain (FNB, FMP), or anti-microtubule-associated protein (MAP)-2 mAb (clone HM-2, mouse monoclonal IgG, Sigma). HIV gag p24 (Kal-1, Dako) and human CD68 (PGM1, Dako) were also used. Secondary Abs were used as described (10). Triple-color analysis for GFP, CD68, and TRAIL, TNF-α, or FasL was also performed as described (10). An absence of primary Ab was used as a negative control. Sections were covered with Vector Shield mounting medium (Vector Laboratories) and observed under a Zeiss LSM 310 confocal laser-scanning microscope.

Serial coronal sections (at least three slices) were scored for the immunostained cells by a single investigator in a blinded manner. All over the sections were evaluated in each sample. Apoptotic cells were also identified using active form of caspase-3 staining and hematoxylin/eosin (H&E) staining. Cells undergoing apoptosis were identified by characteristic morphology including nuclear fragmentation and cell shrinkage with condensed nuclei on H&E-stained sections. In dual-color detection of GFP and human CD3 or CD68, the percentage of double-positive cells was calculated. In dual-color staining of CD68 or CD3 and TNF-α, IL-1β, TRAIL, or FasL, the percentage of double-positive cells in CD68⁺ or CD3⁺ cells was calculated.

Statistics.

Data are presented as the mean ± SD of slices. Welch's t test was used to compare the mean values between two groups. P values <0.05 were regarded as statistically significant.

Development and Characterization of an HIV-CNS Infection Model.

It has been proposed that HIV infection in the brain is initiated by the migration of M-tropic HIV-infected macrophages (16). Thus, we first examined the migration of HIV-infected human PBMCs into the brain of hu-PBMC-NOD-SCID mice after infection with a M-tropic JRFL or a T-tropic NL4-3 virus. Low levels of human CD3⁺ T cell migration [24.4 ± 14.4 cells per slice (cps) in JRFL-infected brains and 30.0 ± 19.7 cps in NL4-3-infected brains] and, less frequently, HIV gag p24⁺ cell migration (10.2 ± 7.69 cps in JRFL-infected brains and 7.25 ± 4.57 cps in NL4-3-infected brains), but not human CD68⁺ macrophages, were found in the meninx and around the vessel of the brain 10 days after infection (Table 1). Because we observed similar levels of CD3⁺ T cell migration in mice 28 days after infection (data not shown) and in brains from uninfected mice (21.6 ± 12.3 cps, Table 1), these data appear to represent the baseline distribution of T cells and macrophages in hu-PBMC-NOD-SCID mice. We observed no apparent neuropathology excepting sparse astrocytosis in the brains of infected mice (data not shown).

To facilitate the migration of HIV-infected macrophages into the brain, we administered a sublethal dose of LPS (100 μg) i.p. to the JRFL- or NL4-3-infected hu-PBMC-NOD-SCID mice 7 days after infection, because LPS has been reported to enhance the transendothelial migration of macrophages and T cells into the brain through its activation (17). A marked increase in human CD68⁺ macrophages (21.3 ± 5.3 cps in JRFL-infected brains and 13.0 ± 4.16 cps in NL4-3-infected brains) and HIV gag p24⁺ cells (58.4 ± 21.5 cps in JRFL-infected brains and 26.8 ± 7.46 cps in NL4-3-infected brains), as well as human CD3⁺ T cells (91.3 ± 34.4 cps in JRFL-infected brains and 36.0 ± 12.7 cps in NL4-3-infected brains), was found in the brain parenchyma of closed regions to vessels (Fig. 1, arrows) 3 days after LPS treatment (10 days after infection; Fig. 1, Table 1) and 21 days after LPS treatment (data not shown), as compared with the LPS-untreated mice (P < 0.05). In contrast, the number of human cells [human leukocyte antigen (HLA)-ABC⁺ (HLA⁺)] recovered from mouse peritoneal lavages decreased after LPS treatment (213.2 ± 159.7 × 10⁴ cells without LPS treatment versus 47.1 ± 37.2 × 10⁴ cells with LPS treatment in JRFL-infected mice; Table 1, P < 0.05). These infiltrated cells were mainly localized in the perivascular region of cortex, basal ganglia, hippocampus, and cerebellum. We also found an increased number of GFAP⁺ astrocytes (astrocytosis) and clustering of F4/80⁺ microglia (microglial nodule) 3 days after LPS treatment in both JRFL- and NL4-3-infected hu-PBMC-NOD-SCID mice, as well as HIV-uninfected hu-PBMC-NOD-SCID and untransplanted NOD-SCID mice, compared with mice that were not treated with LPS (Fig. 1, Table 1), indicating that these features were independent of HIV-1 infection.

Neuronal apoptosis is a hallmark of HIV encephalopathy leading to neuronal dysfunction (6). We examined whether our mouse model had such a feature of the disease. After LPS treatment, histological examination clearly demonstrated apoptosis in the brain. We initially performed TUNEL staining for apoptotic cells. The LPS treatment markedly increased the number of TUNEL⁺ apoptotic cells (78.8 ± 21.5 cps in JRFL-infected brains and 33.3 ± 9.36 cps in NL4-3-infected brains) in cortex, basal ganglia, hippocampus, and cerebellum of HIV-1 infected hu-PBMC-NOD-SCID mice (Table 1) 3 days after LPS treatment (10 days after infection), compared with the LPS-untreated mice (11.4 ± 5.37 cps in JRFL-infected brains and 10.0 ± 4.24 cps in NL4-3-infected brains; Fig. 2a, Table 1). A similar level of increase was also observed 21 days after LPS treatment (28 days after infection; Fig. 2a). To identify apoptotic neurons, we next performed FNG staining to label Nissl bodies in the cytoplasm as a neuron-specific marker, along with TUNEL staining (Fig. 2b). The FNG stains nucleus of all cells and only cytoplasmic Nissl bodies of neurons. Notably, 10–20% of the TUNEL⁺ cells were cytoplasmic FNG⁺ (cFNG⁺) neurons (Fig. 2 b and c), and we observed these only after LPS treatment and only in the JRFL-infected mice (13.1 ± 4.47 cps in JRFL-infected brains, Table 1), not NL4-3-infected mice (Table 1). We further verified the identity of the apoptotic neurons by dual-color staining for TUNEL and another neuron-specific marker, anti-microtubule-associated protein (MAP)-2 (Fig. 2d). We also confirmed the apoptosis event in the brain of HIV-1-infected mice by immunostaining using an Ab specific for active form of caspase-3 (Fig. 2e). In addition, we found a modest increase in apoptotic cells (25.0 ± 10.8 cps) in the brain of HIV-uninfected hu-PBMC-NOD-SCID mice at 3 days after LPS treatment (Fig. 2a, Table 1), but these cells did not include cFNG⁺ neurons (Table 1) and disappeared at 21 days after LPS treatment (data not shown).

These results indicated that LPS treatment of M-tropic JRFL-infected hu-PBMC-NOD-SCID mice induced some neuropathological changes resembling HIV encephalopathy, especially including infiltration of HIV-infected macrophages and neuronal apoptosis in the brain. In contrast, LPS treatment of T-tropic NL4-3-infected hu-PBMC-NOD-SCID mice induced infiltration of only HIV-infected T cells and apoptosis in nonneuronal cells (Table 1), human T cells, and mouse cells including astrocytes, microglias, and endothelial cells (data not shown). These results indicated that the HIV-infected T cells in the brain were not competent to induce neuronal apoptosis, suggesting a critical role of HIV-infected macrophages. Although this model has limitations, it is possible to examine the mechanism of neuronal damage through invading HIV- 1-infected macrophages.

To confirm the requirement of HIV-1-infected macrophage in brain for inducing the neuronal apoptosis, we used recombinant M- (NL-CSFV3-EGFP) and T-tropic (NL-EGFP) HIV-1 expressing GFP as a marker for replicative infection. These recombinant viruses are identical except for the V3 sequence of env that determines the tropism. We infected hu-PBMC-NOD-SCID mice with these viruses and administered LPS. A significant level of systemic infection was established, as estimated by HIV gag p24 plasma levels (data not shown). Both 40–50% of CD68⁺ macrophages and 10–20% of CD68⁻ CD3⁺ T cells in the brain were infected (GFP⁺) by M-tropic NL-CSFV3-EGFP, whereas 10–20% of CD68⁻ CD3⁺ T cells were infected by NL-EGFP in the brain (Fig. 3). We could not find infected (GFP⁺) CD68⁺ macrophages in the brain of NL-EGFP-infected mice. Importantly, apoptotic neurons exhibiting a blue cytoplasm with a pink nucleus, observed by dual-color detection with NeuroTrace blue fluorescent Nissl stain (FNB, blue) and TUNEL (red), were frequently found (10.8 ± 4.97 cps) in the M-tropic virus-infected brain, but not in the T-tropic virus-infected brain (Fig. 3, Table 1). These results further substantiated that the HIV-infected macrophages in the brain were necessary for the induction of neuronal apoptosis.

Critical Contribution of TRAIL to Neuronal Apoptosis.

We next explored the molecular mechanism for the HIV-mediated neuronal apoptosis in our model. The expression of TNF-α, IL-1β, TRAIL, and FasL was examined, because these molecules were reportedly expressed in HIV-1-infected macrophages (6, 18, 19). Dual-color staining demonstrated that ≈40–60% of the infiltrating human CD68⁺ macrophages expressed TNF-α, IL-1β, and/or TRAIL, but not FasL (data not shown). In contrast, in the HIV-1-uninfected brain, ≈40–50% of the infiltrating human macrophages expressed TNF-α and/or IL-1β, but not TRAIL, after LPS treatment (data not shown). Thus, the expression of TRAIL on infiltrating human macrophages was found to be preferentially augmented in the infected brains after LPS treatment. Furthermore, triple-color analysis after infection with NL-CSFV3-EGFP indicated that TRAIL was expressed in ≈50% of HIV-infected CD68⁺ macrophages and 20% of HIV-infected T cells, whereas TNF-α was expressed in ≈80% of HIV-infected CD68⁺ macrophages and FasL was expressed ≈30% of the infected T cells but not in the infected CD68⁺ macrophages (Fig. 4a). Dual-color staining for apoptotic cells by TUNEL and for human CD68, HIV gag p24, or human TRAIL showed that some apoptotic cells were predominantly apposed to HIV-infected macrophages expressing TRAIL but not to HIV-1-infected T cells (Fig. 4b). These results suggested that the occurrence of neuronal apoptosis might be associated with the expression of TRAIL in HIV-infected macrophages.

Finally, we examined the contribution of TRAIL, TNF-α, and FasL to the neuronal apoptosis in the brain of JRFL-infected hu-PBMC-NOD-SCID mice by administering neutralizing mAbs against these molecules at the time of LPS treatment. The infiltration of human CD3⁺ T cells, human CD68⁺ macrophages, and HIV gag p24⁺ cells in the brain was not affected by the administration of these mAbs (data not shown). Neuronal apoptosis was significantly inhibited by the anti-human TRAIL mAb (RIK-2; P = 0.0005 by Welch's t test), but not by the anti-human TNF-α mAb (28401.111) or the anti-human FasL mAb (NOK-1), as compared with control mouse IgG (Table 2). These results indicated a critical contribution of human TRAIL to the neuronal apoptosis in the brain of M-tropic HIV-1-infected hu-PBMC-NOD-SCID mice.