To characterize the distribution of apoptotic neurons and their relationships with the stage of disease, a history of HIV-dementia, and the degree of productive HIV infection, microglial activation and axonal damage, we examined the brains of 40 patients. Samples of frontal and temporal cortex, basal ganglia and brain stem were taken post-mortem from 20 patients with AIDS (including three with HIV-dementia, and eight with cognitive disorders that did not fulfil the criteria for HIV-dementia), 10 HIV-positive asymptomatic cases and 10 seronegative controls. Neuronal apoptosis was demonstrated by in situ end labelling in 18 AIDS cases and two pre-AIDS cases; a single apoptotic neuron was present in the temporal cortex of a control. Semiquantitative evaluation showed that the severity of neuronal apoptosis in the cerebral cortex correlated with the presence of cerebral atrophy, but not with a history of HIV dementia. There was no global quantitative correlation between neuronal apoptosis and HIV encephalitis or microglial activation. However, there was some topographical correlation between these changes. In the basal ganglia, apoptotic neurons were much more abundant in the vicinity of multinucleated giant cells and/or p24 expressing cells. Microglial activation was constantly present in these areas. Axonal damage was identified using beta-amyloid-precursor protein (betaAPP) immunostaining in 17 AIDS and eight pre-AIDS brains. Although no global quantitative correlation could be established between axonal damage and neuronal apoptosis there was an obvious topographic correlation supporting the view that axonal damage, either secondary to local microglial activation or due to the intervention of systemic factors, may also contribute to neuronal apoptosis.

Primary cultures of human embryonic neurons and astrocytes have been used to test the interactions between neural cells and either human immunodeficiency virus type 1 (HIV-1) or HIV-1-infected monocytes. After direct infection with HIV-1, neither morphological alteration of neurons and astrocytes nor signs of viral replication were observed. Similarly, cultured human neurons and astrocytes were resistant to incubation with the supernatant of HIV-1-infected U937 cells, a human monoblastoid cell line. In contrast, HIV-1-infected U937 monocytic cells adhered to neural cells and induced large plaques of necrosis surrounding them. This cytopathic effect began at the time of viral replication (day 16 after infection). Its intensity depended on that of viral replication, and its range was identical to the region of diffusion of viral antigens, as judged by immunocytochemistry. The cytopathic effect was not dependent on the release of free radicals. It could not be induced by cytokines or cytokine-stimulated U937 cells. It is likely that this cytopathic effect depends on the release of viral antigens either within the site of adherence itself or within close range of the astrocyte membrane.

Neuropathological studies have shown that human immunodeficiency virus type 1-infected cells within the brain express several markers characteristic of macrophages and could either be microglial cells, or monocytes invading the CNS, or both. To better define the target cells of human immunodeficiency virus type 1 within the brain, we have studied human microglial cells, both in vivo and in vitro, and compared them to monocytes for their antigenic markers and their susceptibility to human immunodeficiency virus type 1 infection. Brain-derived macrophages were isolated from primary cortical and spinal cord cultures obtained from 8 to 12-week-old human embryos. The isolated cells presented esterase activity, phagocyted zymosan particles, expressed several (Fc receptors, and CD68/Ki-M7 and CD11b/CR3 receptors) of the macrophagic antigenic markers, and appeared to be resident microglial cells from human embryonic brain. Conversely, brain-derived macrophages did not express antigens CD4, CD14, or CD68/Ki-M6, which are easily detected on freshly isolated monocytes. Using these antigenic differences between isolated microglial cells and monocytes, we have observed that two populations of macrophages could be individualized. In the normal adult brain, microglial cells were numerous in both the gray and the white matter. The infrequent cells sharing antigens with monocytes were found almost exclusively around vessels. In 8 to 12-week-old human embryos, microglial cells were found in both the parenchyma and the germinative layer. Cells sharing antigens with monocytes were only found at the top of and inside the germinative layer. In brain tissue from patients with human immunodeficiency virus type 1 encephalitis, cells sharing antigens with monocytes are abundant not only around the vessels but also in the parenchyma. In double-labeling experiments, human immunodeficiency virus type 1-infected cells showed monocyte antigens. Finally, microglial cells also differ from monocytes in their in vitro susceptibility to human immunodeficiency virus type 1 infection; after stimulation by r-TNF alpha or GmCSF, monocytes but not microglial cells can replicate human immunodeficiency virus type 1. This in vitro difference in human immunodeficiency virus type 1 susceptibility between monocytes and microglial cells together with the presence of monocytic antigens within the brain tissue of human immunodeficiency virus type 1-infected patients suggest that human immunodeficiency virus type 1-infected cells within the brain are either monocytes that have crossed the blood-brain barrier and spread through the tissue or perivascular microglial cells that, after phagocyting infected blood lymphocytes, subsequently contain viral antigen and migrate to brain tissue.