Potential Mechanisms of Neurologic Disease in HIV Infection

Melvyn P. Heyes

Infection of humans with the human immunodeficiency virus (HIV) results in the eventual development of the acquired immune deficiency syndrome (AIDS). In the first month after infection, there is a phase of high virus replication in peripheral blood and entry of virus into lymphoid tissue and the brain (15, 21, 39). During seroconversion, flu-like symptoms and aseptic meningitis are experienced, and there is a transient phase of immunosuppression with a decrease in CD+4 cells. In subsequent years, viral load and replication rates in tissues where HIV is localized may be low, due to the effects of the hosts immune response (21). There is (a) chronic activation of the immune system, with increased levels of immune activation markers such as neopterin and b2-microglobulin, and (b) gradual development of immune dysfunction, including loss of antibody production, increased production of cytokines that accelerate HIV replication, destruction of the microenvironment of lymphoid organs, and increased apoptosis of CD+4 and CD+8 cells (21). Patients usually develop a persistent lymphadenopathy and a gradual decline in the number of CD+4 cells, and some may also develop fever, weight loss, night sweats, and oral candidiasis. There is a marked secretion of cytokines during the course of disease. It has been hypothesized that increased activity of immunoprotective cytokines in the early phases of disease is beneficial [TH-1 dominant phase; interleukin 2 (IL-2) and interferon g (IFN-g)], but is followed in the later phases of disease by secretion of proinflammatory cytokines (TH-2 dominant phase; IL-4, IL-5, IL-6, and IL-10) (14, 21, 53). Secretion of tumor necrosis factor a (TNF-a) from B cells and macrophages has been linked to the wasting syndrome associated with HIV infection (21). The secretion of cytokines is also important in the regulation of HIV expression; TNF-a, for example, promotes HIV replication (21, 49). Eventually, immunosuppression and immune abnormalities increase in severity, and CD4+ T lymphocytes become depleted. About 6–8 years after initial infection, patients suffer one or more opportunistic conditions of systemic tissues and brain (AIDS), including Pneumocystis carinii pneumonia, lymphoma, Cryptococcus neoformans meningitis, cytomegalovirus with encephalitis, and human papovavirus infection of brain, which causes progressive multifocal leukoencephalitis. Peripheral neuropathies and myopathies also occur. Death usually occurs within about 4 years of the onset of AIDS (21) (see Multi-Infract Dementia).

HIV infection is also associated with the development of a broad spectrum of central nervous system (CNS) and neuropsychologic deficits, originally termed the "AIDS dementia complex" (55). More recently, the terms "HIV-associated dementia" and "HIV-associated motor cognitive and motor complex" have evolved (9). Meningitis may recur during throughout the course of disease (40). Motor slowing may begin early in HIV infection when patients are otherwise asymptomatic, and resemble those noted in basal-ganglia-related disorders, such as Huntington's disease (2, 46). This has led to the notion that key components of HIV-related neurologic disease are a basal ganglia disorder or subcortical dementia (65). Nevertheless, it is clear that other brain regions are involved, particularly in the later stages of disease (52). Neuropsychologic symptoms that begin early in disease may be of constant severity. Once AIDS has developed, symptoms may rapidly increase in severity and can become an incapacitating dementia. These deficits are viewed as unrelated to opportunistic CNS conditions, although such conditions can cause additional neurologic problems. There is evidence that CNS disease is associated with macrophage-tropic variants of virus rather than T-cell-tropic variants (18). About 30% of vertically transmitted HIV-infected children also show bilateral and symmetrical mineralization of the basal ganglia and periventricular frontal white matter (17). Postmortem studies have shown the presence of multinucleated giant cells, microglial nodules, perivascular monocyte infiltrates, astrogliosis, white-matter pallor, and neurodegeneration or neuronal injury with loss of dendritic spines (9, 41, 48, 51, 65, 74, 75, 81, 82). The presence of multiple disseminated microglial nodules with macrophages and multinucleated giant cells, or the presence of HIV-infected cells within the CNS, is defined as HIV encephalitis (9). Vacuolar myelopathy occurs in about 20% of patients, particularly those with dementia (40).

The mechanisms responsible for neuropsychologic deficits has focused on the neuroanatomical lesions in the brain, i.e. HIV encephalitis. What is striking is that HIV is localized in macrophage infiltrates and microglia, rather than the "functional" elements of the brain, namely neurons, astrocytes and oligodendrocytes. These studies have led to the notion that neurologic deficits in HIV infection are mediated indirectly by either virus- or host-coded agents that produce neurologic symptoms by killing neurons, damaging neurons, disrupting neuronal electrical activity, or impairing neurotransmission. Neurologic dysfunction may also result from abnormalities in the functions of astrocytes or oligodendrocytes (20, 22, 57). Particular interest has focused on the macrophage, because of the distinct association between this cell type and the presence of HIV. Furthermore, although a direct cause-and-effect link between the presence of HIV encephalitis with neurologic deficits during life has not been firmly established, a large proportion of encephalitic patients are demented, which suggests that neurologic symptoms are linked to neuropathology, the presence of HIV, and inflammatory cell infiltrates. Nevertheless, there can be a disassociation between of HIV encephalitis and dementia (73).

Experimental investigations of these hypotheses have largely relied on studies of cells in vitro where neurons and other CNS cell types from various species and brain areas are incubated in the presence or absence of other cell types (macrophages, microglia, astrocytes, lymphocytes), or are treated from supernatants obtained from such cells. In some instances, the cell types other than neurons that are present in the cultures are not stated. Cells are then infected with HIV or stimulated by other HIV- or non-HIV-related agents, and the death of neurons is documented. A few studies have determined whether either systemic or intracerebral injections of putative neurotoxins produce functional or anatomical lesions in intact animals, usually rats.

Collectively, these studies have added significantly to the understanding of immune-mediated neurologic disease. Importantly, they have evolved new approaches to therapy. On the other hand, it has to be stated that the results and interpretations of some of these studies have been viewed as inconsistent, speculative, and perhaps difficult to accept. Some findings are unique or have not been reproduced between different laboratories, and consequently are difficult to evaluate (see also Exciatory Amino Acid Neurotransmission, Arachidonic Acid, and Nitric Oxide and Related Substance as Neural Messengers , for related topics).

Glycoprotein-160 (gp160) is a 160-kD glycoprotein on the surface of HIV, and it includes the gp120 viral envelope and the gp41 fusion protein. The gp120 envelope protein has a high affinity for binding to CD4, a membrane receptor on the surface of several types of immune system cells. gp120 can be released from HIV-infected cells, a critical means by which HIV targets CD+4 cells (27). Brenneman et al. (7) were the first to report that gp120 was neurotoxic to mouse hippocampal neurons at concentrations of 100 aM to 10 pM (10-16 M to 10-10 M). Toxicity associated with gp120 was attenuated by monoclonal antibodies to CD4 receptors. Both peptide T and vasoactive intestinal polypeptide, which share certain analogous sequences with gp120 from some HIV strains, appeared to block gp120 binding and reduce neurotoxicity (6, 7, 10). The neurotoxic effects of 20 to 200 pM (2 {ewc MVIMG, MVIMAGE,!times.bmp} 10-11 M to 2 {ewc MVIMG, MVIMAGE,!times.bmp} 10-10 M) gp120 on rat retinal ganglion cells have also been shown to be attenuated by either N-methyl-D-aspartate (NMDA) receptor antagonists or nimodipine, a calcium channel blocker (19, 44). In this latter respect, gp120 resembles the classic excitotoxic response exemplified by NMDA itself, glutamate, kainate, ibotenate, and quinolinate. The receptors that mediate the neurotoxic effect of gp120, as well as its binding characteristics and specificity, are unclear, because rodent cells do not express CD4, although receptors other than CD4 have been implicated in HIV infection of neural cell lines (28).

There is other evidence that the neurotoxicity of gp120 may be indirect. Lipton et al. (44) have noted that the increase in intracellular calcium (19) and NMDA-receptor-mediated neurotoxicity of gp120 in rat retinal ganglion cells was attenuated by depletion of cell cultures of glutamate. Because gp120 alone had no discernible effect on ionic currents of ganglion cells, as determined by patch-clamp techniques, nor did gp120 enhance glutamate/NMDA-activated currents, a "synergistic" effect of gp120 on glutamate toxicity was postulated. Similarly, an increase in intracellular calcium levels and subsequent translocation of protein kinase C from the cytosol to the cell membrane of rat cortical neurons in response to gp120 was also blocked by removal of glutamate from the incubation medium (72). The mechanisms by which gp120 interacts with glutamate and NMDA receptors remains unclear. However, Sweetnam et al. (67) have reported that gp120 at a concentration of 10 nM to 100 nM (10-8 M to 10-7 M) inhibited the specific binding of NMDA receptor agonists, and blocked inward ion currents in Xenopus oocytes in response to NMDA and glycine (67). They also found that gp120 actually protected against neurotoxicity of rat cerebellar neurons in response to NMDA and glycine, and blocked calcium currents in response to NMDA (67). Peptide T had no effect on NMDA-induced calcium currents. Pulliam et al. (57) have noted no binding of biotinylated gp120 to human fetal neuronal cultures.

Piani et al. (54) showed that murine brain macrophages released glutamate, particularly after stimulation with either endotoxin or IFN-g, and that such supernatants killed cerebellar granule cells via an NMDA-receptor-mediated mechanism. The effects of gp120 were not studied. In contrast, glutamate was eliminated as a neurotoxin involved in HIV-macrophage-mediated toxicity in other studies (24, 56). Further evidence that gp120 can affect NMDA receptor function has been obtained in 7-day-old rats, where an injection of 100 ng of gp120 (but not 1 or 50 ng) into the hippocampus has been reported to produce focal pyramidal cell loss and to exacerbate the neurotoxic effects of NMDA (3). An intracerebroventricular injection of 5 mg of gp120, however, protected mice against an intraperitoneal injection of NMDA (67).

gp120 has also been shown to stimulate nitric oxide production from a human astrocyte cell line by 70% (50). Nitric oxide is an arginine-derived neuroactive free radical that has been implicated in NMDA-receptor-mediated neurotoxicity. Killing of cortical neurons in response to picomolar quantities of gp120 has also been linked to nitric oxide production (16). However, other studies of gp120 report neurotoxicity that is independent of nitric oxide (22, 24, 25, 56).

While some studies suggest a direct effect of gp120 on neurons (64!popup(ch147ref64)), it should be noted that other studies have reported no direct neurotoxic effects of gp120 on neurons obtained from chicken ciliary ganglion cells [100 pM to 10 nM (25)], fetal rat cerebral cortex [4 nM (4)], or rat cerebellar granule cells [10 nM to 100 nM (67)] or from cultures of neurons, astrocytes, and oligodendrocytes obtained from human fetus [1 pM to 1 nM (56, 57)]. Giulian et al. (25) did, however, report that gp120-stimulated macrophages (but not lymphocytes or H9 cells) released neurotoxin(s) via binding of gp120 to monocytoid CD4 receptors. These toxin(s) co-purified with toxin(s) released from HIV-infected THP-1 cells, and could be blocked by antagonists to NMDA receptors (24, 25). In contrast, Pulliam et al. (57) have reported that supernatant obtained from human macrophages that were stimulated with gp120 (1 pM to 1 nM) was not neurotoxic to human fetal brain cultures.

In intact rats, intracerebroventricular injections of 12 ng, but not 1.2 ng, of gp120 resulted in decreased spatial learning in rats (26). This effect of gp120 could be blocked by co-administration of vasoactive intestinal polypeptide, but was not replicated by injections of gp160. In subsequent studies, daily subcutaneous injections of 5 ng of gp120 for 28 days into newborn rats resulted in morphological damage of pyramidal neurons in the cerebral cortex consisting of reduced dendritic branching and length (38). A number of neurologic milestones were also delayed by gp120, although some were unaffected. Both peptide T and vasoactive intestinal polypeptide reduced the severity of neuroanatomical damage and behavioral abnormalities in response to gp120. To investigate whether a fragment of gp120 could be involved in neurotoxicity, 3.3 ng of 125I-gp120 was injected subcutaneously into newborn rats, and gel-filtration chromatography was performed on brain extracts to isolate 125I-labeled gp120 fragments (38). Several fragments were demonstrated and were shown to kill mouse spinal cord neurons. One fraction (#54) was diluted 1:400 (to an estimated concentration of 6 {ewc MVIMG, MVIMAGE,!times.bmp} 10-15 M) and was shown to kill neurons. Peptide T was able to block the toxic effects of fraction #54 (38). The metabolism of gp120 to these neuroactive fragments was not investigated. It is not known whether the production of fragments is essential for the neurotoxic effect of gp120, although filtration of gp120 to remove fragments of gp120 of <30 kD was associated with a 20% decrease in inward ion currents in Xenopus oocytes in response to NMDA and glycine (67). The composition of these molecules were not investigated, and it remains to be established whether they are derived from gp120 or not. Their presence in other preparations of gp120 has not been determined.

In a new approach, Toggas et al. (69) have created the expression of gp120 in the brains of mice by inserting the env sequence that codes for gp120, and placing it under the control of a modified glial fibrillary acidic protein gene promoter that is localized predominantly in astrocytes. Extensive vacuolization of dendrites, reductions in synaptodendritic complexity, astrogliosis, and increased numbers of F4/80-positive cells (microglia) were noted in these mice. Similarly, increased expression of IL-6, also under the regulation of the glial fibrillary acidic protein gene promoter in transgenic mice, was associated with tremors, ataxia, and seizures, as well as neurodegeneration, astrogliosis, angiogenesis, and induction of acute-phase protein production (11).

Tat, nef, rev, env, and gag are genes that code for structural, regulatory, and enzyme proteins (27). When HIV enters the hosts cells, viral RNA is converted to DNA by the action of the enzyme reverse transcriptase and is incorporated into the host cells' chromosomes (provirus) (27). Mistakes in the transcription of HIV nucleotide sequences occur during the reverse transcription step and contribute the high degree of virus mutation. Certain proteins from the host cell initiate transcription, and RNA molecules leave the nucleus to produce proteins via the hosts own translation system. The actual HIV genes code these different proteins. The long-term repeat is a binding site for host transcription factors (27). Tat is one of the first genes to be transcribed, and it binds to the enhancer element of the long-term repeat to enhance viral replication. Nef may suppress transcription or facilitate the manufacture of virus. Rev facilitates the export of intron containing HIV mRNA from the nucleus to the cytoplasm. Tat, nef, and rev are nonstructural regulatory proteins and are expressed early in HIV replication. Env codes for viral coat proteins that mediate CD4 binding and membrane fusion. Core proteins are coded for by gag. Other genes are also involved, including pol, which codes for reverse transcriptase, integrase, and ribonuclease (27). Release of these genes from HIV-infected cells have been reported or implicated.

Sabatier et al. (60) have studied the effects of tat protein and certain fragments of tat. 125I-labeled tat38–68 was shown to bind to rat brain synaptosomes in a manner that could be blocked by very high concentrations of unlabeled tat derivatives with a 50% binding coefficient (K0) of 3 mM. At 5 mM, tat2–86 caused a rapid depolarization of frog muscle fibers and cockroach giant interneuron synapses, via increases in non-ion-selective ion permeability. Murine neuroblastoma and rat neuroglioma were also damaged by tat2–86 at concentrations of 130 nM to 13 mM. Several tat derivatives and certain rev peptides were then shown to kill mice following intracerebroventricular injection in 10- to 180-mg quantities. The authors concluded that this was a "neurotoxic" effect, because muscular tremors, convulsions, wasting, and spastic paralysis were noted (45, 60). No "neurotoxicity" was observed following systemic injection.

Other workers have reported that tat peptides derived from the maedi-visna virus or HIV produced excitotoxic-like lesions 7 days following direct injections into rat striata, consisting of neuronal loss, astrocytosis, microglial reactions, and macrophage infiltrates (29). Nanomoles of the toxin were injected, although the small volume of the injectate rendered the resultant solution in the milimolar range. These neurotoxic effects were attenuated by co-administration of NG-nitro-L-arginine methyl ester, an inhibitor of the synthesis of nitric oxide, or systemic administration of MK-801, an antagonist of NMDA receptors. The authors also speculated that the large numbers of arginine residues in tat provided substrate for the synthesis of nitric oxide.

Nef protein has been reported to share sequence and structural similarities to neuroactive scorpion peptides and can reversibly increase total potassium currents in chick dorsal root ganglion cells (77). A recombinant fusion peptide, env-gag, has been reported to potentiate NMDA neurotoxicity in newborn rat hippocampus (3).

Giulian et al. (25) reported no neurotoxic effects of gag protein or nef protein in chicken cilliary ganglion cells.

Cytokines are well-established mediators of immune function. Within the CNS, cytokines have many effects including toxicity to neurons and oligodendrocytes, stimuli to astrocyte proliferation, and mediators of fever (11, 49). Although selective increases cytokine expression in brain have been shown to occur in HIV-1-infected patients, there is minimal correlation to the severity of encephalitis (1, 70, 78, 80). The most consistent response has been an elevation in TNF-a, which is highest in patients with dementia (1, 70, 78, 80). Cytokines are also important in protecting cells against intracellular parasites—including Toxoplasma gondii, an opportunistic infection in AIDS (13).

Quinolinic acid (QUIN) is a neurotoxic tryptophan–kynurenine pathway metabolite (see ref. 66) which increases neuronal activity and the entry of calcium and which can cause neurodegeneration and promote lipid peroxidation (59). QUIN is an agonist of NMDA receptors, and some brain regions and neuronal types are particularly sensitive to the neurotoxic effects of QUIN, including striatum, hippocampus, and spinal cord (66). Neurotoxicity has been reported in the nanomolar to micromolar range (23, 24, 42, 79).

Immune activation in experimental animals was reported to result in an increase in QUIN levels in brain and blood (33). Elevated levels of QUIN were noted in patients with AIDS, thus implicating NMDA receptors in HIV-associated neurologic disease (34). Substantial increases in the concentrations of QUIN have also been observed in the brain and other tissues following immune activation unrelated to HIV infection (5, 35, 83). In the case of patients infected with HIV, the accumulations of QUIN in cerebrospinal fluid (CSF) begin soon after seroconversion (30, 32), in association with the development of motor deficits (46, 47)). CSF QUIN levels are highest in the later stages of disease, particularly in those patients with neurologic deficits, inflammatory lesions in the brain, aseptic meningitis, or opportunistic CNS conditions (1, 30, 31, 34, 35, 80). In brain tissue, QUIN levels were highest in the basal ganglia (1). Importantly, the concentrations of QUIN in CSF are also correlated with quantitative measures of neuropsychologic deficits in adults (30, 47)), children (8), and SIV-infected macaques (58). Treatment of HIV-infected patients with azidothymidine and antimicrobial therapies were associated with reductions in CSF QUIN levels (8, 30). CSF QUIN concentrations are also correlated with measures of immune activation (neopterin, b2-microglobulin), but not correlated with measures of blood–brain barrier abnormalities (31, 35). In this context, QUIN may be a "marker" of immune activation and neurologic status. The mechanisms involved in increasing QUIN production involve increased activity of, and regulation by, indoleamine-2,3-dioxygenase and kynurenine-3-hydroxylase (35, 36, 63). Macrophages rather than astrocytes appear to be an important source of QUIN within the CNS (36, 37, 61). Cytokines released from activated lymphocytes, macrophages, and astrocytes are an important link between immune activation and QUIN production (36, 61, 62).

Reductions in CSF and blood L-tryptophan levels are associated with induction of indoleamine-2,3-dioxygenase. (31, 76). L-Tryptophan is a precursor of indoleamines, such as serotonin. Evidence for reductions in serotonergic activity have been reported in the CSF of HIV-1-infected patients (43), which raises the possibility that such disturbances may have neuropsychologic consequences. Paradoxically, despite the long history of investigating neurotransmitter abnormalities in human neurodegenerative and psychiatric disorders, minimal studies of the neurochemistry of HIV-1-related CNS disease have been published. Interestingly, induction of indoleamine-2,3-dioxygenase and depletion of L-tryptophan has been implicated in the mechanisms by which interferon-g exerts antiproliferative and antimicrobial effects (12). Depletion of L-tryptophan, an essential amino acid, may have a role in the wasting syndrome of AIDS and other inflammatory conditions.

While the micromolar concentrations of QUIN that are achieved in CSF and tissue QUIN of HIV-infected patients and simian immunodeficiency virus (SIV)-infected macaques (1, 30, 32, 80) are neurotoxic and neuroactive to certain types of CNS neurons (23, 24, 25, 42, 66, 79), there is currently no proof that QUIN is involved in the pathogenesis of HIV-associated neurologic disease.

A number of independent studies with quite different paradigms have suggested the presence of additional factors related to HIV toxicity. Many of these factors have not been identified, and their presence in patients remains to be confirmed.

Giulian et al. (24) reported that chronic infection of either THP-1 cells or U-937 cells in vitro was associated with the release of a neurotoxin(s), as demonstrated by death of either rat spinal cord neurons or chick ciliary ganglion cells following application of monocyte supernatants. In this study, neurons were kept separate from the HIV-infected cells. A number of potential toxins were eliminated as candidates for this toxic activity, including gp120, free radicals, nitric oxide, QUIN, glutamate, aspartate, cysteine, cytokines, and other peptides or molecules larger than 2000 daltons. A subsequent study suggested that toxin(s) are also released from gp120-stimulated monocytes (25). Pulliam et al. (56) have also reported neurotoxic activity of supernatants derived from HIV-infected macrophages, although this toxic activity was associated with high-molecular-weight molecules that were heat-labile. Toxicity was proportional to the levels of p24 expression, but unrelated to gp120 production (56).

Other studies have reported no neurotoxic activity in supernatants obtained from relatively acutely HIV-infected THP-1 cells when the fluids were applied to cultures of human neurons (68), SK-N-MC human neuroblastoma cells, or rat neurons (22). However, neurotoxicity associated with THP-1 infection was observed if PBMC or THP-1 cells had actually adhered to the neuron/astrocyte cultures ("cell-to-cell adhesion") (22, 68). Free radicals were eliminated as mediating toxins, and neurotoxicity was not replicated by TNF-a, IL-1, or IFN-g or by using fibroblasts instead of macrophages (68). Similarly, Genis et al. (22) noted no neurotoxic activity in supernatants obtained from HIV-infected monocytes when applied to cultures of SK-N-MC human neuroblastoma cells or rat brains. Neither TNF-a nor IL-1b was neurotoxic in their system.

Initially, Bernton et al. (4) reported that the neurotoxic effects of culture media obtained from HIV-infected monocyte cultures on rat cortical neurons could be explained by contamination of the HIV stock used with either Mycoplasma arginini or M. hominis. Both Pulliam et al. (56) and Giulian et al. (24), however, had already stated that such contamination was not present in their experimental systems. Bernton et al. (4) also showed that neurotoxicity could be replicated by exposure of neurons to mycoplasma, endotoxin, or TNF-a.

Pulliam et al. (57) have reported that gp120 in human fetal brain cultures or isolated human fetal brain astrocytes was toxic to astrocytes, resulting in decreased GFAP staining. Astrocytes have been reported to attenuate the neurotoxic effects of glutamate released from stimulated murine brain macrophages (54), and zymosanactivated astrocytes have been reported to release protein(s) that promote neuronal survival (23). In contrast, no astrocyte toxicity was found by Genis et al. (22) following exposure of rat cultures of supernatants from HIV-infected monocytes. However, cocultures of astrocytes and HIV-infected monocytes were reported to produce neurotoxic factor(s) (22). In association with toxicity, increased expression of TNF-a and IL-1b were demonstrated, although these cytokines were not toxic in and of themselves. Production of arachidonic acid metabolites were implicated in the production of TNF-a and IL-1b. Platelet-activating factor, leukotrienes B4 and D4, and lipoxin A4 were also shown to be produced. Neurotoxic activity in HIV-infected glial cultures was stated to be blocked by dexamethasone. Platelet-activating factor added to cocultures of uninfected monocytes, and astroglia produced neurotoxic activity (20). These results suggest that arachidonate metabolites, cytokine production, and neurodegenerative responses are linked in this particular system, and that "interactions" between HIV-infected monocytes and astrocytes were required for neurotoxicity.

Buzy et al. (10) have shown that CSF obtained from HIV-infected patients was not toxic when applied undiluted to mouse hippocampal neurons. However, neurotoxic activity appeared if the CSF was first diluted between 1:1000 to 1:1000000. Neurotoxicity was attenuated by peptide T and monoclonal antibodies to CD4. Neurotoxic activity was not found in CSF from non-HIV-related neurologic diseases.

Certain features of HIV encephalitis can be replicated in severe combined immunodeficiency mice by intracerebral injections with human peripheral blood mononuclear cells and HIV (71). Other models include macaques infected with SIV, mice infected with murine retroviruses, and cats infected with the feline immunodeficiency virus.

Neurotoxicity associated with HIV-1 infection has been described in a variety of experimental paradigms. Neurotoxic activity may be related to the presence of HIV and agents derived from the virus. Other toxins released from the host in response to the presence of HIV and/or immune activation are also candidates. The clinical manifestations of HIV-associated neurologic disease may result from the combined and separate effects of several of factors that operate to different degrees at different times in disease. NMDA receptors may be common mediators of toxicity in several experimental conditions.

The conflicting findings in the different experimental systems used indicate that caution should be exercised in the interpretation of results to generalized phenomenon in patients. It is clear that no one toxin can account for all neurotoxic activity described in vitro, and several remain to be identified or shown to be present in HIV-infected patients in significant and neuroactive amounts. Most studies that evaluate the "neurotoxin" hypothesis of HIV-associated neurologic disease use cell death as a measure of neurotoxic activity. More studies are required that evaluate subtle or noncytolytic neurologic dysfunction, particularly because there is no proof that HIV-associated neurologic disease in AIDS is dependent on nerve cell death per se. Alternative systems, such as transgenic mice, xenographs, and cells implanted into severe combined immunodeficiency mice, may prove useful in modeling chronic mechanisms, rather than the acute systems currently available. The effects of anti-HIV drugs on neurotoxic activity also need to be studied. Certain mechanisms suggest generalized therapeutic strategies, including attenuation of HIV replication, reductions in the synthesis, or effects of key cytokines or antagonists to NMDA receptors. Other agents suggest highly specific approaches, such as drugs that attenuate the synthesis of quinolinate or attenuate the effects of gp120.

I appreciate the useful comments made by Drs. A. Lackner, S. P. Markey, D. Rausch, K. Saito, and C. A. Wiley regarding this manuscript.

published 2000