The current studies were designed to investigate the glutamatergic mechanisms underlying HIV-1 Tat-mediated toxicity in a rat model of HAND. Prior studies have clearly demonstrated that Tat drives neurotoxicity in cell culture and cognitive impairment in animal models of HAND. Importantly, the physical interaction of Tat with NMDARs is critical for Tat’s induction of neuronal death. To examine Tat-induced glutamate toxicity in vivo and its impact on cognitive function, we administered a high or low dose of Tat to mPFC and examined performance on learning and memory tasks, and changes in gene expression. The results demonstrate that acute administration of HIV-1 Tat86 to mPFC at both a high and low dose corresponds with deficits in spatial learning and memory. Additionally, low-dose Tat86 and Tat101 caused changes in NMDAR subunit expression in the mPFC.

The rats injected with high dose of Tat86 (10 μg) specifically exhibit a lower mean ER in the SOR task when compared to saline-injected rats, as does the NMDA group, indicating a lower proportion of time exploring the moved object (Fig. 1A). Analysis of performance in NOR and TO shows no statistically significant differences observed between the Tat and saline groups. Though the high dose (10 μg) shows a trend toward impairment on TO, the more relevant low dose (40 ng) indicates no impairment on the TO task. Additionally, the NMDA-injected group did not exhibit significantly different mean ERs when compared to the saline-administered rats suggesting that overt glutamatergic toxicity did not affect NOR or TO behavior. The lack of NMDA-mediated toxicity may be impacted by the unmatched size of the NMDA group compared to the saline and Tat groups. Further, the relatively low group sizes may limit detection of differences in performance on the other behavioral tests. In particular, the high-dose Tat86 cohort suggests some effect on TO task that may be limited by small group size. Nonetheless, NOR and TO tasks are mediated by circuitry distinct from that underlying spatial learning and memory, and thus, the results suggest a circuit- or subregion-specific susceptibility to Tat-mediated glutamate toxicity. Performance in the SOR task relies on mPFC and the hippocampus, with some contribution from perirhinal cortex [43]. Temporal order/recency memory is also dependent on hippocampus [43,44,45]; additionally, TO and NOR performance are more dependent on perirhinal cortex [46, 47], whereas mPFC-hippocampus circuitry is most critical for spatial object recognition [48, 49]. The reduced performance of Tat86-injected rats in SOR taken together with no significant differences observed in NOR or TO implies that specific correlates of spatial learning and memory are affected by Tat, sparing mPFC-perirhinal circuitry involved in NOR and TO. Additionally, the impaired SOR performance observed in the NMDA-injected group suggests that Tat86 impacts circuitry underlying spatial learning and memory via NMDAR signaling. Thus, based on the findings of the current study, mPFC projections to CA1 and/or CA3 of the hippocampus may be particularly altered by Tat protein in mPFC.

Prior studies consistently demonstrate that exposure to Tat results in spatial learning deficits via glutamatergic mechanisms. Tat is known to impair spatial learning and memory including radial arm maze [50], likely via NMDAR signaling [51]. Tat expression in transgenic animals also impairs performance in the Morris Water Maze [52]. Spatial learning and memory deficits are also observed long-term following neonatal Tat exposure [53, 54]. These findings are in accordance with the abundant evidence of Tat potentiating NMDAR signaling. NMDAR potentiation in mPFC-hippocampal circuitry is suggested to contribute to impairment in mPFC-dependent tasks [55], and NMDAR signaling in the hippocampus is required for spatial learning and memory [56]. These studies additionally suggest that Tat reduces long-term potentiation [50, 51], offering another mechanism to examine in future studies. However, Tat impairment of spatial learning and memory is largely shown by injection to the hippocampus. It therefore remains critical to further investigate whether mPFC exposure to Tat leads to compensatory mechanisms or adaptation in specific subregions to affect connectivity to hippocampus.

To follow up on the consistent impairment of spatial learning and memory by both high and low doses of Tat86, and similar effects between high-dose Tat86 and NMDA treatment, we hypothesized that glutamatergic signaling in the mPFC may be altered in Tat-injected rats. It is also well established that cellular exposure to Tat stimulates pro-inflammatory signaling—particularly from activated microglia. Thus, we assessed transcript abundance for cytokines and chemokines as well as Grin1, Grin2a, and Grin2b which encode NMDA receptor subunits NR1, NR2A, and NR2B. While the NMDA receptor may also utilize NR2C or NR2D subunits, these are unlikely to be expressed in mPFC as they are restricted to early developmental time periods and/or specific cell populations [57]; NR2A is ubiquitously expressed, and NR2B is abundant in the frontal cortex [58].

Considering the importance of NMDARs in LTP and subsequent relation to spatial learning and memory, our findings regarding the upregulation of Grin1 and Grin2a transcripts in mPFC (Figs. 3 and 5a) are of particular interest. Moreover, Grin1 and Grin2a transcripts are also enriched in the contralateral hemisphere of Tat86-injected rats, which was not directly lesioned in the Tat86 cohort. These results suggest an effect beyond the site of injection, which may be due to spread of the injection or cross-talk within mPFC.

While Tat is known to potentiate NMDAR signaling, alteration of NMDAR expression is not well established. One observational clinical study suggests that overall NMDAR expression is reduced in the frontal cortex of individuals with AIDS and HAD, whereas we observe increased Grin1 and Grin2a transcript expression specifically in rat mPFC. In vitro and in vivo models provide varying results regarding the effects of Tat on NMDAR expression. In an ethanol withdrawal model, Tat administration to CA1 of hippocampus does not alter NMDAR density, as assessed by MK801 binding in tissue lysates [51]. Primary rat hippocampal neurons treated with an LTP blocker following Tat-induced synaptic loss exhibited NMDAR reactivity in new postsynaptic densities, suggesting that synaptic loss caused by Tat may specifically affect NMDARs [59].

Other studies indicate that Tat exerts subunit-specific effects on NMDARs in vitro and in vivo. Selective antagonism of GluN2A-containing NMDARs can prevent Tat-induced synapse loss, whereas GluN2B antagonism limits cell death [60]. Further, GluN2B-specific antagonism after Tat-induced synaptic loss stimulates synapse replacement in vitro while rescuing Tat-induced dendritic spine loss and restoring fear conditioning in vivo [60, 61]. This study suggests that GluN2A stimulation enables neuronal pro-survival signaling, whereas antagonism of GluN2B containing NMDARs in the presence of Tat may limit pro-death signaling to enable the recovery of NMDAR expressing postsynaptic densities. The subunit-specific findings likely contribute to the observed behavioral outcomes; differential recruitment of GluN2A and GluN2B affects long-term potentiation in the hippocampus; thus, the observed upregulation of Grin1 and Grin2a may particularly alter the effects of mPFC afferents on hippocampal LTP to promote impairment in SOR [62]. In addition, knockout studies suggest that the NR1 subunit in CA1 pyramidal cells of the hippocampus is required for intact object recognition [63], while other studies suggest that object recognition tasks increase NR1 and NR2A expression [64]. These findings suggest that increased expression of NR1 and NR2A in the hippocampus supports memory consolidation, and various studies suggest that expression of NR1 and NR2A subunits is increased following induction of long-term potentiation [65]. These studies may support our observation of increased expression of NR1 transcripts in the Tat group in combination with intact NOR performance; however, prior studies are specific to hippocampal expression. Specific findings regarding NMDAR subunit involvement in learning and memory in mPFC are contradictory. Whereas hippocampus exhibits increased NR1 and NR2A expression reflective of LTP following novel environment habituation, subunit expression in mPFC is not altered [66]. However, in the context of spatial tasks, NR1 and NR2A are increased in mPFC synapses following radial maze training, followed by further NR1 increase and NR2B increases after performance in additional tests [67].

Glutamatergic signaling in the mPFC is suggested to support amyloid beta-induced impairment in object recognition, as post-training NMDAR antagonist administration rescues performance in the object recognition test [68]. In contrast, systemic administration of NMDAR antagonist MK801 induced an animal model of schizophrenia involving spatial memory deficits [69]. Similarly, chronic MK801 induction of schizophrenia is suggested to limit LTP in the mPFC-hippocampus circuitry [70].

It is important to highlight that a number of studies suggest that behavioral testing alters gene expression. Memory consolidation following behavioral tasks may alter NMDAR subunit expression specifically [64, 65]. It is therefore possible that behavioral testing in our Tat86 cohort is contributing to the observed increases in Grin1 and Grin2a expression. Notably, our RNA-seq study was carried out in animals that did not perform behavioral testing providing additional context. While Grin2a upregulation is confirmed in the RNA-seq data, suggesting that this is truly an effect of Tat exposure in mPFC, Grin1 upregulation was not upheld. Thus, future studies should consider the potential effects of behavior training on Grin1 and other NMDA subunit expression to understand if exposure to Tat protein alone influences gene upregulation.

Analysis of Tat101-exposed mPFC tissue by RNA-seq confirmed the upregulation of Grin2a and reveals other potential mechanisms of Tat toxicity. One of the most prominent upregulated genes in Tat-injected rats is Slc24a2, which encodes the calcium-sodium exchanger NCKX2 [71]. NCKX2 is widely expressed in the brain and suggested to play a critical role in axonal Ca2+ ion regulation, synaptic plasticity, and learning and memory [72, 73]. Given the importance of calcium influx to Tat toxicity through NMDAR potentiation [10, 17, 74], the upregulation of Slc24a2 in Tat-injected mPFC may occur in response to Tat-mediated neuronal calcium influx. Other upregulated genes highlight the central role of calcium in Tat-induced toxicity in our study—Cacna1e, encoding a high voltage activated subunit Cav2.3 which plays a role in synaptic plasticity [75, 76]. Interestingly, two GABA signaling-related genes are also upregulated. While Igsf9b and Gabrb2 have not specifically been reported in prior studies of HAND, GABAergic changes have been noted in Tat models of HAND [25, 26]. Tat-transgenic mice have been reported to exhibit PFC-specific reduction of Syt2 and gephyrin, but increased GAD67, indicating specific molecular changes rather than overall reduction of GABAergic signaling [77]. The upregulation of Igsf9 and Gabrb2 revealed by our RNA-sequencing analysis emphasize this variable effect of Tat on different GABAergic molecules. The additional finding of Ankk1 and Ankfn1 upregulation supports an overall effect of synaptic remodeling. Upregulation of these ankyrin-repeat containing molecules along with receptor subunit changes (Grin2a, Gabrb2) and other synapse-related adhesion molecules (Igsf9b, Pcdh11x) suggests the potential for synaptic remodeling after Tat injection to mPFC. Additional assessment of the differentially expressed genes at the threshold of − 2 ≤ log2FC ≥ 2 against the Interferome database [78] suggests interferon regulation of Slc24a2, Mis18a, and Ankfn1. By lowering the threshold for − 1 ≤ log2FC ≥ 1, many additional genes from Tat-treated mPFC match the database of interferon-regulated genes, including Adamts9, Aif1, Cxcl10, Egfr, Fgf2, Gabrg3, Map3k1, Ppargc1b, Slc26a2, and Cldn1. Interestingly, comparison of our DE genes against a published rat blood–brain barrier transcriptome [79] revealed no hits, and searching our gene list for genes associated with gap junctions, tight junctions, and established modes of Tat-mediated barrier disruption [80] similarly revealed few matches. However, a number of the aforementioned genes may play roles in cellular adhesion or blood–brain barrier permeability (namely CXCL10, Cldn1). While these genes are not featured in the differentially expressed genes in our RNA-seq results (Fig. 5) due to strict threshold limits, they can be useful candidates for future investigation. Our current results may also be limited by the bulk RNAseq approach, whereas single-cell RNA-seq may provide a refined understanding of cell-type specific Tat toxicity in future studies.

Additional analysis of our differentially expressed genes by GO analysis for pathway enrichment suggests broader changes beyond glutamate and GABAergic signaling (Online Resource 4). The top 20 significantly upregulated pathways identified by GO analysis highlight axon development and transport, cytoskeleton and filament organization, and glutamatergic transmission. Notably, axonal and synaptic transport has previously been implicated in Tat toxicity [81, 82]. MAP2 is commonly indicated across multiple pathways, and the loss of MAP2-positive dendrites has been reported in HIV-1-exposed neurons [83, 84]. In pathways related to cytoskeleton organization, axon and projection guidance, Mef2c and Cntn1 (contactin-1) stand out for their existing relation to synaptic plasticity and memory [85, 86], with Mef2c particularly having suggested protective roles [87,88,89]. The upregulation of neurotransmitter and glutamatergic regulation highlight Syt1 and Slc1a2 which have previous associations with Tat-induced toxicity [23, 90], providing some validation to our model. Particularly, Slc1a2 (encoding the glutamate transporter GLT-1) within neurotransmitter regulation pathways recapitulates the relevance of glutamate dysregulation and excitotoxicity in Tat-exposed mPFC.

GO analysis did not indicate significantly downregulated pathways at the threshold of Padj < 0.05 and log2FC ≤  − 1. Below this threshold, pathways appear emphasizing ribosomal and translation processes, and importantly mitochondrial pathways including metabolic and oxidative stress responses. These pathways are notable for their involvement in neurodegenerative diseases and cognitive functions and may be important candidates for future studies with higher group sizes and statistical power, as mitochondrial dysfunction and mitophagy are important factors in HAND [91] and in Tat toxicity in neurons and glial cells [92,93,94].

Overall, the results of GO enrichment analysis support our findings in RT-PCR while also broadening the scope our findings to recapitulate some of the known hallmarks of neuronal and glial toxicity mounted in HAND and Tat-based models of the disease state.

Some transcriptomic analyses of HAND models recapitulate some of these expression changes. Animal models suggest that chronic exposure to Tat is correlated with upregulation of immune and inflammatory genes in the hippocampus [95, 96], while our RNA analyses did not highlight this, the result may be limited by the acute injection of Tat weeks before tissue collection. Notably, human studies utilizing RNA sequencing demonstrate that glutamate receptor subunits and transporters, ion channels, and immune signaling aspects are upregulated in brain tissue from PLWH [97]. The similarity between this clinical study and our differentially expressed genes suggests that the use of Tat101 at a clinically relevant isoform provides an accurate model of HAND-related transcriptional changes. However, distinct region-specific transcriptional changes, mechanisms of Tat-induced gene modulation, and the relationship between multiple neurotransmitter and ion signaling changes must be further investigated.