In this study, we stereotactically injected a single dose of HIV-1 Tat into murine hippocampus and cortex to study the temporal sequence of events that leads to the neuroinflammation and abnormal synaptic architecture associated with HAND. Introduction of Tat into the CNS resulted in acute infiltration of leukocytes, adjacent and remote to the injection site, with engagement of microglia in cell-to-cell contact between synapses and invading leukocytes with inflammatory phenotypes. A pragmatic caveat to note is that we used an in vivo
dose of Tat (3 µg/µl) that is ≥1000× the dose commonly used in in vitro
experiments (1–100 nM). There is no reliable way to measure the actual concentration of Tat in the extracellular space in the CNS because no successful ELISA strategy exists and because native Tat is incredibly sensitive to oxidation.
Thus, there are no published reports of local Tat concentrations in the brains of patients with HAND, but ample evidence of Tat transcripts in HAND brains 
. When cell lines are transfected with Tat, supernatants have extraordinary neurotoxicity, with a theoretical LD50
in the femtomolar range, far too small to be detected on a gel (Avi Nath, personal communication). When Tat is produced for experimental use, it is made and stored under reducing conditions, which are thought to represent the extracellular environment in the CNS. However, Tat is also extraordinarily sticky, and between its marked sensitivity to oxidation and ability to bind even to silanized glass surfaces (as in the syringe barrel and needle used in stereotactic injections), we used Tat1–72
at a pharmacologic concentration (i.e. 3 µg/µl in our study) to abrogate these confounds.
While the majority of the population of microglia that we surveyed in hippocampus exposed to Tat remained in a surveillance phenotype without evidence of transformation into a rod-like, reactive morphology (, insets), there was widespread evidence of reciprocal engulfment () of microglia and peripheral myeloid cells, raising the question of whether microglia may attack invading inflammatory leukocytes.
Introduction of Tat into the CNS was also associated with disruption of normal synaptic architecture, as reported previously 
. Our data cannot unequivocally differentiate the relative contribution of microglia vs. peripheral myeloid cells to phagocytic engulfment of normal synaptic architecture. However, qualitatively, our data in aggregate suggest that microglia do not simply mimic the initial neurotoxic response of infiltrating leukocytes 24 hr after exposure to Tat. While infiltrating inflammatory leukocytes robustly destroy microglia and neuronal dendritic spines in their vicinity (, , green arrowhead), microglia remain largely ramified, albeit with hypertrophied processes (, Panels B, D), and appear to ensheath neurons (, , blue arrowheads).
The sequence of signaling events after Tat injection in our model is complex and can be assumed to include release of the chemokine monocyte chemoattractant protein type 1 (MCP-1/CCL2) well as up-regulation of VCAM-1 and ICAM-1, with the ultimate biologic effect of transmigration of mononuclear cells across the BBB into the CNS 
. Inflammatory leukocyte infiltration is likely to be further amplified by the ability of Tat to increase the production of platelet-activating factor, PAF, a chemotactic factor for granulocytes 
. We do not yet know whether the microglial response to Tat-induced chemotaxis of peripheral inflammatory leukocytes in vivo
is an attempt at host defense of normal synaptic architecture, but it is interesting to speculate that microglia may attack invading leukocytes, which enter the CNS in response to local neuroinflammation and chemokine release 
. Microglia are also well recognized for their contribution to the normal “pruning” and maintenance of dendritic architecture 
. However, in the setting of Tat-induced neuronal damage and neuroinflammation, it is quite possible that their ability to strip synapses may become amplified to a pathologic level - with the result that microglia may elicit enduring neurotoxic responses 
. Our morphometric data (Figure S2
, Panel D) demonstrate a 41% increase in leukocyte and microglial contact with neuronal structures, and when considered in the context of synaptic elements present as inclusions in microglia at 7 d (, Panels E, F) and enduring microgliosis at 28 d (, Panel D; Figure S1
, Panels B, D), this seems like a reasonable supposition to be investigated in future studies.
In addition to demonstrating the rapid infiltration of inflammatory leukocytes into the brain following intracerebral delivery of HIV-1 Tat, our data suggest that the gene dose of CX3CR1 exerts a profound regulatory effect on infiltration of inflammatory leukocytes. Indeed, mice engrafted with homozygous marrow (CX3CR1-GFP+/+
) had significantly increased numbers of infiltrating mononuclear cells in response to both saline and Tat injection into the brain (), when compared with mice engrafted with heterozygous marrow (CX3CR1-GFP+/−
). This effect was especially striking in mice that received saline alone. In contrast, head shielding had no effect on the magnitude of Tat-mediated leukocyte infiltration into the CNS (Figure S1
, Panel B) 
Data in this study as well as a previous study 
demonstrate that a single injection of Tat into the CNS induces persistent changes in normal neuronal and microglial cellular architecture. While in vivo
exposure to Tat alone as an experimental model for HIV-1 neuropathogenesis clearly has its limitations, other murine models of HIV-1 infection of the CNS 
are not compatible with the adoptive transfer methodologies required to dissect the relative contribution of peripheral leukocytes vs. microglia in HIV-1 induced neuroinflammation. Moreover, the profound and durable CNS changes following a single injection of Tat are striking, and speak to the ability of this simple, reductionist model to recapitulate some of the striking features of HAND – including the puzzling persistence of CNS damage in the absence of detectable virus loads.
Here we show that, following a single injection of Tat into the CNS, phagocytic elimination of synaptic structures by microglia persisted for up to 7 days () and microgliosis persisted for at least 28 days ( and S1
), while infiltration of peripheral inflammatory leukocytes was far more ephemeral. Although ephemeral, this transient leukocyte infiltration may be devastating to normal neuronal function. Data from experimental models of stroke and intracerebral hemorrhage suggest that the magnitude of neutrophil and activated monocyte/macrophage infiltration, as well as release of soluble inflammatory mediators are associated with greater neurologic dysfunction 
In conclusion, our data make a compelling case for a sequence of events that are initiated by infiltration of mononuclear and granulocyte subsets of leukocytes in response to the presence of Tat in the CNS. This is associated with the subsequent engulfment of microglial processes and dendritic spines by infiltrating leukocytes and by a microglial response that includes an apparent “attack” on the invading leukocytes. This suggests that peripheral leukocytes and microglia may make quite different contributions to the pathogenesis of brain injury associated with HIV-1 infection of the CNS. The role of microglia in disease may also change over time, since intercellular communication between invading leukocytes and neurons may trigger the release of ‘find me’ signaling molecules such as ATP or UTP 
. These signals have the potential to trigger a switch to an inflammatory phenotype for microglia – leading to enduring microgliosis and neuroinflammation, long after the initial viral insult has been cleared. Studies in progress are focused on the critical signaling cascades that may be involved in the biphasic microglial response to HIV-induced neuroinflammation.