Previous studies found that in patients with cognitive impairment and HIVE, neurodegeneration might be associated with a reduction in the expression of neurotrophic factors such as FGF1 [69]. In contrast, in a subset of patients with HIV with preserved cognitive performance, FGF1 levels are up-regulated. Interestingly, neuronal populations prone to degenerate in patients with HIV are dependent on FGF1 and FGF2 and express FGF receptors and CXCR4 [6,63, 70]. The FGFs belong to a family 13 trophic factors that are involved in neurogenesis [71] and angiogenesis [72]. Since FGF1 and 2 are expressed in the CNS in neurons vulnerable to the neurotoxic effects of HIV proteins [69], this indicates that the molecular pathways modulating FGF1 effects might be important mediators of the signaling involved in neurodegeneration and neuroprotection in patients with HIV [73].

In vitro studies in primary neurons and neuronal cell lines have shown that the neuroprotective effects of FGF1 and FGF2 are mediated by activation of PI3K-Akt that in turn inactivate GSK3β via phosphorylation at the Ser 9 residue [63,64]. In addition to FGF1 and FGF2, other growth factors that exert their effects via receptor tyrosine kinases also lead to inactivation of GSK3β through phosphorylation. These include growth factors such as insulin growth factor-1 (IGF-I), epidermal growth factor (EGF) and platelet-derived growth factor (PDGF) [74,75]. To further investigate the neuroprotective effects of GSK3β regulation by FGF1 in vivo, we generated lines of tg mice expressing the human FGF1 under a neuronal promoter (PDGFβ).

Human FGF1 cDNA was obtained by reverse transcriptase polymerase chain reaction (RT-PCR) from human brains and cloned into PCRII vector (TA Cloning from Invitrogen, CA) and 100% fidelity of nucleotide sequence was confirmed by dideoxy sequencing. Subsequently the FGF1 cDNA fragment was subcloned into the PDGFβ transgene cassette. The PDGFβ promoter was a gift of Dr. Tucker Collins at Harvard Medical School. The final construct contains the PDGFβ promoter, SV40 intron, hFGFl cDNAs, and SV40 polyA (Figure 2A). Constructs were microinjected and 5 lines of founder mice were obtained. Of them, based on the levels of mRNA expression two transgenic lines (line 15 low expresser; line 12 moderate expresser) were selected. RPA and Western blot analysis showed that both lines expressed human (h)FGFl at levels comparable to the levels in the human brain (Figure 2B–F). Immunocytochemical analysis confirmed that hFGFl was primarily expressed by neurons in the neocortex, hippocampus and basal ganglia, regions selectively susceptible to the neurotoxic effects of HIV products. Both lines of hFGFl tg mice were viable, bred well and the nervous system developed normally. To determine the effects of FGF1 expression on the GSK3β signaling pathway, immunoblot analysis was performed with an antibody against phosphorylated GSK3β. This showed that in the mouse line expressing moderate levels of hFGFl (line 12), levels of phosphorylated GSK3β (inactive form) were increased, while levels of pGSK3β in the low expresser line (15) were similar to nontg controls (Figure 2C).

In order to test the hypothesis that hFGFl protects against the neurotoxic effects of HIV products, tg mice (3 mo old, 5 per group) from lines 12 and 15 received intracerebral gp120 injections (lmM, total 2μl) in the neocortex and hippocampus. In nontg mice (3 mo old, 5 per group), gp120 promoted significant neuronal damage and astrogliosis compared to nontg saline-treated mice (Figure 3). In hFGFl tg mice from line 12 (moderate expresser) neurons were protected against the toxic effects of gp120, while line 15 mice (low expresser) were susceptible, supporting the contention that hFGFl was bioactive and protected neurons in a dose-dependent manner, probably via inactivation of GSK3β. Therefore, treatments directed at increasing the expression of FGF1 or targeting the signaling pathways affected, such as GSK3β, might represent a possible neuroprotective strategy.

Similar to FGF1-mediated neuroprotection against HIV toxicity, recent studies have shown that at low concentrations lithium—a GSK3β inhibitor—protects the CNS from toxins [76,77] via GSK3β inactivation. Our previous studies have shown lithium pre-treatment is neuroprotective in mice that received intrahippocampal injections of recombinant gp120 protein [76]. To expand on these studies, and to investigate the protective effects of lithium-mediated regulation of GSK3β against HIV-gp120-mediated toxicity in a tg model, mice expressing high levels of gp120 in the brain were treated with lithium and analyzed by immunocytochemistry and immunoblot. For this purpose, 4-month old gp120 tg mice (n=6) received intraperitoneal injections of lithium chloride (2 mg/kg/day) (Sigma, St. Louis, MO) for 14 days, and 6 control mice received saline injections (100μl/day for 14 days).

Consistent with our previous study in lithium-treated mice that received intrahippocampal injections of gp120 [76], lithium treatment in gp120 tg mice protected the hippocampus of mice from gp120-mediated toxicity (Figure 4) [76]. Lithium ameliorated the dendritic damage in gp120 tg mice (Figure 4A-D). Similarly, pre-exposure of neuronal cultures to lithium significantly reduced gp120-associated neurotoxicity [76] (Figure 4E). The protective effects of lithium in vitro were partially blocked by LY294002, an inhibitor of the PI3K/Akt/GSK3β pathway (Figure 4E). Immunoblot analysis showed that lithium activates Akt and blocks GSK3β activity via phosphorylation at Ser 9 (Figure 4F). This is consistent with previous studies showing that lithium and other inhibitors of GSK3β activation, such as valproic acid, protect the brain against HIV-associated neurotoxicity [78–82]. Therefore, the neuroprotective effects of FGF1 and lithium or valproic acid involve targeting analogous pathways, namely the Akt/GSK3β signaling cascade.

In view of this evidence and previous studies showing that GSK3β activation is involved in the mechanisms of HIV neurotoxicity [60], we concluded that targeting the GSK3β pathway with lithium might be of value in the management of the cognitive impairment [76,81–83]. Remarkably, in a pilot clinical trial we found that treatment with lithium for 12 weeks significantly ameliorated the cognitive deficits in HIV patients [83]. Similarly, another pilot clinical trial demonstrated that valproic acid is well tolerated in patients and can ameliorate HACI [78,80]. Taken together, these studies support a role for the GSK3β signaling pathway as a potential therapeutic target in the treatment of neuroAIDS.

Currently, we are in the process of extending the lithium studies as well as investigating the potential therapeutic value of analogous molecules and signaling targets. In this regard, it is becoming clear that the neuroprotective effects of FGF1 and lithium might involve regulation of other signaling cascades [84]. Immunoblot analysis of brain homogenates from gp120 tg mice treated with saline or lithium showed that in addition to activating Akt and blocking GSK3β, lithium treatment also resulted in reduced levels of CDK5 expression (Figure 4F), but no effects were observed on p38 and SAPK. These findings support the contention that both the PI3K/Akt/GSK3β and the CDK5 signaling pathways might be involved in mediating neurotoxicity in HIVE, and that regulation of CDK5 in pathological states might be an important target for neuroprotection.

Recently, and as part of a gene array study, we found that several components of the CDK5 signaling pathway are altered in patients with HIVE [85]. Moreover, other studies have shown that activation of CDK5 by calpains contributes to HIV-induced neurotoxicity [58]. In the mature CNS CDK5 is the predominant CDK, and is expressed at high levels in neurons [86]. While in peripheral dividing tissues other CDKs function as regulators of the cell cycle, CDK5 is unique among the CDKs in that its primary role is in regulating the phosphorylation of a multitude of downstream targets involved in cytoskeletal and synaptic function, among other activities [87]. Through its diverse targets, CDK5 plays an important role in critical processes in both the mature and developing brain, including neurogenesis [50,88], neuronal migration [89], and synapse formation and plasticity [87,90,91] (Figure 5).

CDK5 is a Ser-Thr protein kinase with postmitotic activity that phosphorylates KSP motifs on cytoskeletal proteins (MAPlb, tau, NF, nestin, DCX), synaptic proteins (PSD95, synapsin, cadherin) and transcription factors (MEF2) [92–94]. While in dividing neurons CDKs are activated by cyclins, in the nervous system CDK5 is activated by forming a complex with p35 or p39 [93,95] (Figure 5). In contrast to other CDKs, which are inhibited by the cell cycle regulator proteins p21 and p27 (Kip-1) [96], CDK5 activity is regulated primarily by the metabolism of the activating proteins p35 and p39 [97].

Regulated activation of the CDK5 pathway via p35 [98] plays an important role in neuronal development [99] and synaptic plasticity [86,91] (Figure 5). The p35/p39-CDK5 complex has been shown to be constitutively active, however CDK5 phosphorylation by casein kinase 1 (CK1) [100,101] and CDK7 [102] at Ser159 also contribute to CDK5 activation. CDK5 can be further activated by phosphorylation at Tyr15 by Src-related tyrosine kinase [100]. However, it is unclear to what extent phosphorylation plays a role in the regulation of CDK5 activity, and most studies suggest that the primary mechanism controlling CDK5 activation is expression and metabolism of p35 [97].

Synthesis of p35 is stimulated by NGF and BDNF [103] as well as by ECM molecules like laminin [104] (Figure 5). Interestingly, HIV-1 Tat inhibits NGF-induced Egr-1 transcriptional activity and consequent p35 expression in neural cells [57]. The phosphorylation of p35, which is catalyzed by CDK5, produces an upward shift observed by gel electrophoresis and can be utilized as an indicator of CDK5 activity [100]. Although regulated activation of CDK5 by p35 and p39 under physiological conditions promotes neuronal development [99], axonal transport and synaptic activity [86,87,91], recent evidence suggests that abnormal activation of the CDK5 pathway might be also involved in cell death and neurodegeneration [105–108] (Figure 6). In this regard, recent studies indicate that calpain I mediated cleavage of p35 into the N-truncated fragmented p25 results in hyperactivation of CDK5 which in turn results in abnormal phosphorylation of toxic substrates that have been associated with cell death [106] (Figure 6).

For example, in AD, it has been proposed that amyloid-β (Aβ) protein triggers calpain-mediated cleavage of p35 into p10 and p25, which in turn results in uncontrolled activity of CDK5 and tau hyperphosphorylation [106]. Similarly, in the nervous systems of patients with AIDS, it is possible that HIV products such as gp120 might trigger calcium influx via chemokine or glutamate receptors. This, in turn, may stimulate calpain-mediated proteolysis of p35 into p25 and subsequent hyperactivation of CDK5 with abnormal substrate phosphorylation (Figure 6). Alternatively, HIV proteins might regulate CDK5 activity by sequestering p35, or interfering with factors that modulate levels of p35 expression or stability. Since CDK5 plays a role both in synaptic function and neuronal development, then abnormal activation of this molecule by HIV proteins might not only impair the functioning of mature neurons but also contribute to alterations in neurogenesis.

Although HAART is an indispensable tool for the management of the neurological complications associated with HIV neuro-invasion, this approach does not completely eradicate HIV from the CNS, nor does it provide protection from the neurotoxicity of HIV proteins. Therefore discovery and testing of neuroprotective agents continues to be an important goal as an adjuvant therapy for the cognitive and neurological complications of HIV. Efforts in this direction include the testing of NMDA-R blockers such as memantine [109–111]. In addition, we have recently shown in a pilot clinical trial that modulation of signaling pathways involved in neurodegeneration with lithium might have a potential role as a neuroprotective treatment in HIV patients [83]. Lithium’s effects involve regulation of GSK3β and probably CDK5 among other potential targets [84]. Interestingly, and as indicated before, both of these signaling pathways are altered in patients with HIVE. Although lithium [76,83] and valproic acid [78,80,81] might hold some promise as therapeutic regulators of the GSK3β and CDK5 signaling pathways, it is necessary to test and develop new and more specific compounds that target CDK5.

In recent years considerable efforts have been devoted at developing CDK inhibitors for the treatment of neoplastic, viral and neurodegenerative disorders [112]. Among them, roscovitine has been identified as a CDK5 inhibitor with therapeutic potential. This compound belongs to the family of purine-derived CDK inhibitors that includes olomoucine and purvalanols [113]. Roscovitine crosses the blood brain barrier [114] and has been shown to be protective in experimental models of neurotoxicity [115]. This compound has been shown to be safe in humans and in clinical trials for glial cell tumors has been shown to have a promising effect. Roscovitine also blocks to a lesser extent cdc42, CDK2 and ERK [113]. Remarkably, CDK inhibitors are also under development as anti-retroviral agents, because HIV recruits these kinases for viral replication in mitotic cells [112]. Of the CDK inhibitors, flavopiridol, which targets CDK1, 2, 4, 7 and 9, in combination with roscovitine, appears to be effective [112]. Thus CDK inhibitors could target both the viral infection as well as the pathogenic mechanism.