A long-standing observation in patients infected with HIV is that different parts of the brain appear to be variably affected by the virus. Evidence for variable response to the virus by different brain regions has come from clinical, pathological and imaging studies. To help gain a better understanding of this phenomenon, we used in vivo
1H MR spectroscopy to probe how different regions of the macaque brain respond metabolically during the first month of SIV infection, a time point not normally accessible by clinical studies. At the time of peak viremia, a transient decline in the NAA resonance, reflecting temporary neuronal injury, was only detected in the frontal cortex. On the other hand, changes in Cho and MI were similar across brain regions. The Cr resonance appeared stable in all regions and times with the exception of white matter where a transient increase was observed at peak viremia.
Evidence of neuronal injury is detected only in the frontal cortex
A decrease in NAA is only observed in the frontal cortex at 2 wpi, the time of peak viremia. NAA is almost exclusively located in neurons, and a decrease in NAA reflects neuronal loss [22, 23] or reversible neuronal injury [24, 25]. Lentz et al. showed that encephalitis severity correlates with decreased NAA/Cr levels in the frontal cortex in the macaque model of neuroAIDS [26, 27]. With the reduction of SIV plasma viral load by 1–2 orders of magnitude at 4 wpi, NAA levels are also no longer abnormal, suggesting that partial immunologic control of the virus is sufficient to result in neuronal recovery. The selective neuronal vulnerability of frontal cortex to SIV infection is further supported by the observation of significant decreases in synaptophysin (SYN) in the frontal cortex but not in the basal ganglia (unpublished results). SYN is an integral membrane glycoprotein in presynaptic vesicles. Its decline indicates a disruption in synaptic transmission and it is therefore a sensitive marker of neuronal dysfunction. We have previously shown that SYN is highly correlated to in vivo NAA/Cr measurements in this same SIV/macaque model of neuroAIDS .
Choline changes in the same manner in the brain regions
The Cho levels appear to change in a similar fashion in all brain regions after SIV infection. This implies that the changes are global, and not specific to any particular brain region. MI increased significantly in the FC and WM, and MI changes correlated between frontal cortex and basal ganglia (P = 0.043, Rs = 0.37), but not between white matter and the other two regions. It must be noted that when the correlations are subjected to Bonferroni correction, the MI correlations lose significance, however, the choline correlations between the regions remain significant after Bonferroni correction. At the time of peak viremia, Cho levels are elevated, and they decrease to baseline levels or lower with a reduction of plasma viral load at 4 wpi. An increase in Cho may reflect altered membrane metabolism , and may be associated with cerebral inflammation. Its elevation is believed to reflect cellular responses, including cell proliferation, recruitment, microgliosis, and/or astrocytosis in neuroAIDS [29, 30]. Evidence of the association with astrocytosis is given by our observations of elevated GFAP at the same time point in this model .
In a previous report  we reported that Cho levels in the frontal cortex decreased to levels significantly below baseline by 4 wpi. In extending our analyses to other regions we find that at 4 wpi Cho levels are below baseline in the white matter and basal ganglia, but these are not significant. However, significant declines in Cho were observed in all 3 regions between 2 and 4 wpi, a period when viral loads are substantially reduced. The best interpretation of this observation is that it is a manifestation of brain recovery and healing.
Cerebral MI increased almost linearly after SIV infection. MI is located primarily in glia; thus, elevation of MI is considered to reflect increased glial cell activity . An elevation of MI has been commonly reported in studies of chronically infected HIV patients [9, 16]. The increases in MI were not as highly correlated between the brain regions, but they appeared to increase in all 3 regions almost linearly up to 4 wpi. We have previously reported that there is a complex relationship between astrocytosis and MI, as well as between astrocytosis and Cho . The data reported here on Cho and MI continue to support a complex relationship among these measurements that occur when the animal is infected. Further studies are necessary to elucidate the relationships.
Creatine increases during acute SIV infection in white matter
At the time of peak viremia, Cr concentrations were significantly increased in the white matter of the centrum semiovale. Changes in total Cr (phosphocreatine and creatine) are associated with altered energy metabolism. Elevated Cr levels have been reported in the frontal white matter of chronic HIV patients . It is believed that the virus enters the brain through infected monocytes that later differentiate into macrophages [34, 35]. During this process of mononuclear cell infiltration, glial-cell activation and proliferation may manifest high tissue metabolism which would explain an increase in Cr. Energy change as a function of enhanced glial activation is supported by our finding that the greatest increases in MI and Cho are in the same regions as the highest increase in Cr (Figure 2). Our finding that creatine changes are not uniform across the brain regions is consistent with the study by Chang et al.  who reported creatine increases in the white matter but decreases in the basal ganglia in later stages of AIDS dementia complex.
Comparison of HIV and SIV
Most HIV MRS studies are performed during the chronic stage of infection, often with subjects who are on antiretroviral therapy. Our macaque study focuses on acute SIV infection, a period that has not yet been studied by MRS in HIV infection. Acute SIV infection induces the most profound NAA changes in the frontal cortex, and more subtle and difficult to detect changes in the basal ganglia and white matter. We hypothesize that HIV in humans might produce similar responses during the primary infection as those described here. During primary infection, HIV and SIV proliferate to very high levels for a short time. As the immune system responds to the infection, plasma virus drops to lower levels. There may be recurrence of high viremia with the progressive weakening of the immune system by the virus. Anti-retroviral therapy will suppress viral production at this stage. It is only at the last stage, with the collapse of the immune system, that very high viremia is once again observed. There are very few reported MRS studies of people with early  or late high viremia [36, 37]. It is possible that the difference in viral loads in patients is the most important factor responsible for differences MRS patterns that have been reported.
The macaque model has similar MR spectroscopic changes (diminished NAA, elevated Cho and MI) as in human neuroAIDS, suggesting that neurochemical and cellular responses to SIV and HIV are similar. However, the specific pattern of MRS changes in various brain regions of the acutely SIV-infected macaque appears different from that seen in humans. Both, single voxel MRS and magnetic resonance spectroscopic imaging (MRSI) studies performed before the advent of anti-retroviral therapy demonstrated profound declines in the levels of NAA in advanced cognitively impaired patients [29, 38, 39]. In the early stages of cognitive dysfunction, Cho/Cr and MI/Cr were shown to be elevated prior to changes in NAA .
More recently in the anti-retroviral era, in vivo
1H MRS studies of HIV patients have consistently found metabolic abnormalities in different brain regions [5, 6]. Some groups have reported that the most profound MRS changes occur in the basal ganglia and white matter [7, 8]. Others have reported significant changes in the frontal cortex of HIV patients, even in patients free of neurological symptoms . Additionally, recent high spatial resolution neuroimaging investigations using sophisticated analytic methods have found diminution in the thickness of the cortical mantle .
In this study we found region specific responses with respect to NAA but not choline. MRSI studies investigating the regional metabolic responses in HIV infected patients with HIV associated dementia were found to be quite variable. Some report evidence of uniform metabolic abnormalities (decreased NAA/Cr  and increased Cho/Cr) in various brain regions and conclude that abnormalities of cerebral metabolites in HIV-infected patients may be part of a diffuse process . However, some MRSI studies observe NAA/Cr decreases  and Cho/Cr increases  in gray matter regions and not in white matter in patients with AIDS dementia complex. Additionally, Barker et al. reported that the largest reductions in NAA and increases in Cho were in the frontal lobe white matter and there were no changes in the cortical gray matter . This great variability could be the result of many factors including: disease stage; the presence, type and duration of antiretroviral treatment; co-morbidities including drug abuse; and differences in imaging parameters such as pulse sequence, TE and TR.
We also investigated potential correlations between plasma viral load and metabolite concentration changes. We previously reported  a significant relationship between changes in Cho/Cr and plasma viral loads (Rs = 0.79, P < 0.01) by Spearman Rank correlation analyses. Here, we found a correlation between plasma viral loads and Cho/Cr changes in the basal ganglia and Cho changes in BG. There was no relationship between WM Cho changes and plasma viral loads or between any of the other metabolite concentrations/ratios and viral loads. These preliminary findings are different from what has been reported for patients with HIV, however, our study focuses on the acute phase of infection, unlike human studies which involve chronically infected individuals. Among HIV+ patients higher plasma viral loads are typically associated with lower NAA  and higher MI . Another study reported that CSF but not plasma viral load correlated with Cho/Cr in the white matter . Again, there are many potential reasons for differences in correlations that have been reported between viral loads and MRS metabolites.
Possible mechanisms of regional brain metabolic abnormalities
The leading theory of neuroAIDS pathogenesis is that infected/activated monocytes traffic into the CNS, where they differentiate into macrophages, and set off a cascade of events that ultimately result in neuronal injury . One possibility that may explain regional variations in brain injury is that there are differences in the rate of monocyte trafficking into various brain regions. However, the similarity in the changes in Cho and MI observed in all regions suggests that the trafficking is globally similar. This is supported by the finding of similar astrocytic activation in frontal cortex and basal ganglia (unpublished results). However, while the evidence suggests that viral invasion is globally similar, neuronal injury is selective to the frontal cortex. Cortical neurons may be more susceptible to injury or neuronal protective/recovery mechanisms may be more effective in other regions.
The relationship of brain metabolic changes with respect to the level of virus in the blood is important. The greatest changes in NAA, Cho and Cr occur at the time of peak viremia followed by near normalization within 2 weeks. This sensitivity of the brain to events in the blood indicates a highly dynamic pathogenic process with the rapid transit of virally-infected monocytes into the brain and an equally rapid response by the brain to these events. The highly dynamic nature is further supported by the rapid normalization as plasma viral load decreases. Notably, this normalization occurs despite persistent virus in the blood, although not at peak levels. This observation suggests there may be a competition between pathogenic and protective mechanisms, and that a threshold level of viral load and trafficking, virally-infected monocytes must be exceeded for brain metabolic response to be manifested.
Our results imply that the virus has a global effect on microgliosis and astrogliosis. However, only cortical regions, here the frontal cortex, are affected by neuronal injury, suggesting that some regions of the brain are more vulnerable to invasion by SIV or HIV infected macrophages. These results may support the use of a combination of antiretroviral therapy early to decrease influx of infected/activated monocytes/macrophages that lead to inflammation and neuroprotective drugs which pass the BBB in order to prevent neuronal injury in region of the brain in which the innate recovery mechanisms fail.
MR spectroscopy should be considered for use in clinical HIV patient care to monitor changes in metabolism indicative of neuronal dysfunction possibly predictive of cognitive dysfunction. Single voxel MRS [29, 33, 47] as well as magnetic resonance spectroscopic imaging (MRSI) [16, 30, 36, 38, 39, 42, 43, 48, 49] have been performed on patients with HIV dementia. Single voxel MRS has been commonly used since it is a shorter scan (for one voxel location) and requires less data processing. However, MRSI is a much more efficient technique to record MR spectroscopy data since it examines multiple regions in one data acquisition. Both techniques have shown significant changes in metabolite concentration in individuals with neuroAIDS and offer complementary information and should be used in conjunction as suggested by Sacktor  if time allows.