MR spectroscopy in HIV associated neurocognitive disorder in the era of cART: a review

Neuroimaging has been a critical tool for understanding the neuropathological underpinnings observed in HIV. The pathophysiology of HAND is chiefly driven by neuroinflammation. Despite adhering to cART, low levels of viraemia probably persist in the brain in some patients leading to chronic immune activation with resultant neuroinflammation and consequent neuronal injury. MR spectroscopy has been widely used as a biomarker for the presence and severity of HAND in several studies. By studying the MRS signatures, it is possible to characterise the presence of neuroinflammation and neural injury. Furthermore, metabolite concentrations measured by MRS could be used as a quantitative indicator of HIV cerebral involvement, thereby affording the opportunity to assess the efficacy of cART in HAND. However, currently there are three significant limitations in the MRS HIV research literature: the relative paucity of prospective studies, the small number of regions of interrogation due to current methodology (single voxel MRS), and the evolving understanding of the impact of co-morbidities (e.g. ageing, mood disorders, alcoholism etc.) on MRS measurements. This review critically addresses the current literature of MRS studies in people living with HIV (PWH) with HAND to determine its value, especially in the context of the current cART era. In addition, we discuss technical considerations related to the disease and the future direction in HAND using MRS.

1. MRS is a reproducible surrogate marker to measure the brain injury in HIV.
2. In the cART era, the neurocognitive impairment is characterised by increased Ch and Mi in frontal white matter and basal ganglia early on and in the later stages there is compromise in the neuronal integrity signalled by reduced NAA.
3. Creatine concentrations change with severity of dementia and therefore metabolite concentrations should be expressed either as absolute values or relative to internal water. 4. MRS changes can take as long as 6-12 months to correlate with the neuropsychological improvement (although paradoxical, this is a significant observation in two respects: that the cellular level changes take longer time to both clinically manifest as well as takes longer time to revert back in spite of apparent clinical improvement). 5. Glx compounds appear to be more sensitive to identify the early neurocognitive impairment and early indications are that the excitotoxic pathway can help to identify the treatment response as well. 6. MRS complements the other advanced imaging measures in the assessment of the influence of the comorbidities on cognition.
7. With the emergence of the newer evidence in the development of HAND, the metabolites Glx and GABA

Introduction
The diagnosis of HAND in the era of cART can be challenging. Currently HAND is best diagnosed by NC testing which requires trained staff. However, quantitative imaging techniques are emerging as a potential alternative. MRS is a robust technique which measures the chemical environment in the brain and offers insight into the neuronal integrity, cell membrane synthesis and turnover, inflammation status, and levels of microglial activation and astrogliosis within the sampled CNS tissue. The concentration of these neurochemicals can then be measured from the area under the curve of that particular spectral peak.
Given the very significant effect of cART on HIV disease, the literature will be reviewed in two sections: pre-cART period (prior to 1997) and the era of cART (after 1997), after a brief synopsis of the technical aspects of MRS.

Technical considerations
The principle behind the MRS technique is suppression of the water signal and extraction of metabolites which have differing resonance frequencies; these can be separated from each other using Fourier transformation into different spectral peaks (1-5 ppm range). The concentration of these neurochemicals can then be measured.
The measured metabolites include NAA, Cho, Cr, Mi, Glx and GABA. The resonance of the metabolites is not uniformly seen on all MR acquisition sequences and is dependent upon the TR and TE. The short TE sequences capture the resonance of the metabolites with short relaxation times (eg. Mi and Glx, GABA as well as free lipids). The NAA, Cr and Cho resonance can be captured on both long and short TE sequences.
NAA is found predominantly in neurons and is known to be present at concentrations of 9-12 mmol in neurons, and as such is regarded as a marker of neuronal density. In addition, NAA is believed to play several roles: as a reserve for energy metabolism, a supply of acetyl CoA, a source of glutamate and an important role in osmoregulation. NAA is synthesized in the mitochondria of neurons and therefore mitochondrial dysfunction is known to cause reduced NAA. Neuronal cell death leads to irreversible loss of [1] NAA while mitochondrial dysfunction in neurons leads to reversible loss of NAA [1].
Cho is a membrane marker whose elevation reflects membrane inflammation and myelin breakdown. Cho mainly consists of glycerophosphocholine and phosphocholine, compounds involved in phospholipid metabolism in brain tissue. Cho is present in the cell membrane of all cells; however, it is more abundant in glial cells. The resonance is attributed to trimethyl ammonium residues of free choline, phosphorylcholine, glycerophosphorylcholine and other metabolites such as carnitine. Cho is typically higher in the white matter than in gray matter [2].
Mi is present almost exclusively in glial cells and as such is a putative marker for glial cells [3]. It acts as an osmoregulator and is increased in any process that causes glial activation. The brain and cerebrospinal fluid (CSF) are relatively enriched in Mi compared to plasma, with estimated typical concentrations of 6 mM in brain, 0.2 mM in CSF, and 0.03 mM in plasma [4]. Within the brain Mi is an intracellular molecule. It is taken up into cells via two sodium-Mi cotransporters, SMIT1 and SMIT2, and a hydrogen-myo-inositol symporter, HMIT. SMIT1 and SMIT2 are expressed by both neurons and glia, although SMIT1 is predominantly astrocytic and SMIT2 predominantly neuronal [5]. Mi concentrations appear generally to be higher in glia than neurons, and glial uptake of Mi exceeds that of neurons, possibly due to expression of the HMIT transporter.
Cr is a complex of amino acids and is present in most of the cell types in the brain. It is usually regarded as relatively constant in the brain parenchyma though there are some disease states where it may be reduced [6,7]. CrT (total creatine) is the concentration of Cr and PCr (phosphocreatine) in the brain tissue. It is utilised as an energy reservoir in cells with high energy demands as it is a part of creatine kinase energy metabolism buffer system used to maintain ATP levels. Glial cells have a four times greater concentration compared to neuronal cells [1].
Glx is a unitary term for two excitatory amino acids (glutamate and glutamine) involved in normal neuronal communication; raised levels are associated with excitotoxicity [8]. GABA is a neuroinhibitory amino acid potentially involved in the disrupted default neuronal network associated with impaired memory as a compensatory response to excitotoxicity [8] (Fig. 1).
Metabolites' measurements are not only dependent on the acquisition methods but also on editing techniques. Newer sequences such as 2D L-COSY (two dimensional (2D) localised chemical shift correlated spectroscopy) is an acquisition technique that uses several additional frequency applications to detect the molecular resonance that otherwise is undetectable in 1D-1H MRS. MEGA-PRESS uses a post processing editing technique to unmask the spectra which otherwise cannot be detected due to a stronger resonating molecule at similar resonating frequency (eg. GABA and total creatine) utilising their unique spectral characteristic of J-coupling (GABA) [9].
The metabolite values derived from MRS are often expressed as a ratio with reference to creatine as an internal standard. However, Cr has been shown to vary with age [6,7] as well as the severity of HAND [10], potentially leading to inaccurate results. Therefore, internal water has been used as a standard given that it is more robust and reproducible (Fig. 2).
Initial MRS reports were either long TE single voxel PRESS studies [10][11][12] or long TE Multivoxel techniques in HIV Dementia patients [13][14][15]. The long TE techniques attenuate the signal from many metabolites considered important in the measurement of HIV associated brain injury (the compounds such as Mi, Glx and GABA cannot be measured on long TE sequences). As a result, long TE PRESS (point RESolved spectroscopy) studies were replaced by short TE PRESS methods. While short TE PRESS can show metabolites that reflect brain injury associated with cognitive impairment, this technique also has limitations, chiefly the small area of interrogation. Furthermore, the metabolites have overlapping spectra and require specialised post processing techniques for separation such as GABA and creatine and Glt and Gln.

Pre-cART era
Whilst the pre-cART era is a matter for history for most populations its characteristics are still relevant    a backdrop for the efficacy of cART and can serve as a  guide to the MRS findings in cART naïve patients as well  as those who have multiresistant HIV. However, studies in this period were often limited by the data acquisition and analysis methods. Long TE sequences were dominantly used hindering identification of Mi and did not specifically target areas known to be vulnerable in HAD [13,15,16]. Jarvik et al. moved the field forward by using steam sequence (STimulated Echo Acquisition Mode-STEAM) (Short TE) to study the proton MRS (Hydrogen proton MR spectroscopy) role in HIV induced neural injury in the areas of abnormal appearing white matter and compared these to normal areas/centrum semiovale [17]. They showed that there are significant differences: reduced NAA/Cr, increased Cho/Cr, increased Mi/Cr ratios in HAD patients and an increase in a marker peak/Cr (this peak identified between 2.2 and 2.4 and likely Glx). Initial studies in the pre-cART period showed the Cho/Cr ratios were abnormal in HIV infected patients regardless of neurocognitive impairment and were seen in most parts of the brain (FWM, BG, Centrum Semiovale and FC [10,11,17,18]. The Cho elevation was postulated to be due to either increased cellularity (increased macrophages and microglia) or increased cell membrane break down or both. The putative glial marker Mi was abnormal as a ratio to Cr in various stages of HAND [19,20] and shown to be proportional to the dementia severity in the FWM [19]. The NAA/Cr levels were reduced in cognitively impaired while asymptomatic seropositive patients were shown to have normal NAA/Cr ratios [11,17,19,20]  However, later studies were not uniform in this observation of normal NAA/Cr in cognitively unimpaired subjects as compared to the earlier studies [10,21,22]. These differing observations are most likely due to the use of Cr as internal standard. Cr is known to be in higher concentrations in astrocytes than in neurons. Astrogliosis, a known histopathological sequel in chronic HIV patients would lead to increased Cr on MRS. (Cr was shown to increase in the frontal white matter while reduced in the basal ganglia in HAD group). Studies in animal models have shown that the levels of Cr do not remain stable in HIV associated brain injury [8,23]. As Cr concentrations could change with dementia severity, metabolite concentrations rather than metabolite ratios should be measured [23].
In summary in the pre-cART era the characteristic findings on MRS are reduced NAA/Cr and NAA/Ch and elevation in Choline and Myoinositol in cognitively impaired patients while those who were asymptomatic showed proportional elevation in Choline and Myoinositol only.
Global changes in the pre cART era

MRS in cART era
"The overarching findings in the era of cART in PWH, are reduced neuroinflammation and reduced neuronal loss, which are reflected in normalised or reduced elevation of Cho and Mi and normalised or low normal NAA when compared to the normal controls [22]". These findings also correlated well with improvement in CD-4 counts and neuropsychology performance [18,24]. However, these observations were not uniform with several later studies showing improvement but not necessarily normalisation with cART [21,22,25]. These inconsistencies have subsequently been addressed by using longitudinal studies.

Longitudinal studies
"Several of the longitudinal studies have shown that there is spectroscopic improvement with institution of cART " [18,24,26,27]. Some of subsequent studies have shown that these changes persist in chronically infected patients despite restoration of immunological status and effective viral suppression in response to antiretroviral therapies [28][29][30]. These varied results in metabolite ratios in multiple studies are likely multifactorial and driven at least partially by expression of the metabolite concentrations as ratios relative to Cr [23]. In addition, there are likely several pathological processes driving neural injury as patients are living longer and developing various comorbidities as well as differing degrees of response to cART [31].
Some longitudinal studies have revealed improvement in the metabolite ratio of NAA/NAA + Cr + Cho compared to the baseline in the first few months of treatment (↑ NAA) [26,27]. Chang et al. demonstrated significant improvement in the metabolites (Choline compounds elevated at 3 months in FWM, FGM, BG and continued to be elevated even at 6 months but showed reversal after 9 months) of the HIV + cohort after 6-9 months, while improvements in CD4 cell counts and viral loads (measured in the plasma and CSF) occurred after 3 months of treatment [18,23]. Tarasow et al., also demonstrated similar improvements in neuronal integrity (increased NAA/Cho ratios) approximately after 6 months in FWM and BG [25]. Both these studies also demonstrated that the spectral improvement correlated with clinical improvement in multiple domains (Improvement in neurocognition: Chang et al., CD-4 counts: Wilkinson et al., HIV encephalopathy) [19,27]. The early reversal of metabolites in some of the initial studies [24,26] was likely to be due to the lower creatine ratios at the base line due to treatment effect at the base line cohort [21].
However, the improved metabolites concentrations were later found to be inconsistent in some of subsequent studies. Salvan et al. [10] have shown that the improvement was only observed in those patients who had decreased NAA/CR at baseline while those who had elevation in the Cho/Cr did not improve much [10,13]. Some studies, also showed persistent elevation of Mi, but not Cho, in the FWM, suggesting ongoing neuroinflammation with glial activation despite viral suppression. Nevertheless, the degree of elevation was less than that observed in cART-naïve HIV patients [23,32].
More recently, two studies have explored the longitudinal effects of patients with HIV infection and response to cART. Young et al. [33] recruited fifty-three participants at a median of 3.7 months post HIV transmission and followed for a median of 6.0 months. They observed increases of Cho/Cr and MI/Cr in the frontal white matter and parietal gray matter as well as Glu a marker of excitotoxicity, in the basal ganglia in treatment-naive patients which reduced after initiation of cART [33]. Sailasuta et al. have used single voxel MRS in four brain regions (FWM, FGM, BG, PC) in cART-naïve subjects before (n = 59, 27 with HAND) and after 12 months of cART. The salient observation was Persistent elevation of Cho was noted in individuals who remain impaired after 12 months of cART in the posterior cingulate (PC) [34]. This cohort of cognitively impaired group also exhibited decreased glutamate in both FWM (P = 0.027) and BG (P = 0.013) as compared with those without HAND. This study has raised an interesting association of Choline but not myoinositol in patients who were on stable therapy with NCI similar to earlier study by Chaganti et al. [35].
Three recent studies have explored the interactions of cART and neurometabolites in virally suppressed people living with HIV in a longitudinal design. Cysique et al.; in their longitudinal design (23 month ± 5) (areas of interrogation: FWM, BG, PCC using water as an internal standard) in an aviremic cohort have observed progressive reduction in creatine in the FWM and PCC in stable HAND and a sharper decline in progressive HAND [36]. Other metabolites (however, this study showed reduced baseline NAA in CA and PC and increased Mi in PCC) did not show any significant change over time in this cohort with stable neurocognition [33]. Boban et al. used a longitudinal design to study an aviremic cohort using 2D CSI with short TE in a 5 year follow up and interrogated prefrontal cortices, anterior and PCC, intraparietal sulci, and frontal centrum semiovale white matter; the ratios of creatine to NAA, Cho and Mi were assessed. "Although this study demonstrated significant increase in the NAA/Cr ratios in those who showed NC improvement, it has used creatine as the internal standard which may confound the outcomes" [37]. In this study, none of the other metabolites revealed any consistent changes across the brain regions interrogated.
The improvement of NAA metabolite concentrations is interesting and is most likely due to one of the mechanisms detailed below. The role of NAA in myelin lipid synthesis, particularly in early development, is well established. The acetic acid from NAA becomes incorporated into CNS myelin [38]. Under metabolic stress, a shortage of acetyl-CoA could result in reduced NAA synthesis and increased hydrolysis of NAA to provide acetate for myelin repair [38] It is therefore likely that there are several mechanisms in play in the increase in NAA after treatment. 1: The creatine reduction in patients on cART when measured as a ratio to creatine. 2: Correction of transient depression of synthesis due to reduced availability of acetyl-CoA and 3: Reduced NAA hydrolysis and neuroplasticity could play a significant role in improvement of NAA.
Another longitudinal study of interest is from Gongvatana et al. They explored two groups; HIV on stable cART and NAS and followed them over two years. This study explored the midfrontal cortex (MFC), FWM, and BG and found that the HIV-infected subjects showed significant annual decreases in brain metabolite levels in all regions examined, including NAA (2.95%) and Cho (2.61%) in the FWM; NAA (1.89%), Cr (1.84%), Cho (2.19%), and Glx (6.05%) in the MFC; and Glx (2.80%) in the BG. They further identified that the neurocognitive decline was associated with longitudinal decreases in Glx in the FWM and the BG, and in NAA in the BG. "This study concluded that there are widespread progressive changes in the brain in chronically HIV-infected persons despite stable antiretroviral treatment and virologic suppression and can lead to neurocognitive declines" [39].
"Putting these studies together, the outcome measures based on metabolite improvement could take longer and may not entirely dependent upon the antiretroviral therapy alone. Other factors such as comorbidities and the nadir CD4 counts, presence of NCI, and limited efficacy of cART or potential cART neurotoxicity [40] may influence the final outcomes. In patients who are aviremic, although there are no conclusive observations, there are early indications to suggest that there is a trajectory of metabolites with normal NAA and persistently increased choline and variable myoinositol".

Regional specificity of metabolites and relationship to the neurocognitive function in PWH
Several studies in the cART era in PWH have shown that there is persistence of metabolite abnormalities despite restoration of immunological status and effective viral suppression in response to antiretroviral therapies [28][29][30].Common areas of interrogation are include FWM, FGM, BG and in some instances the parietal white matter, parietal gray matter and cingulate.
The most common spectral abnormalities in MND are those of marginally elevated Ch/Cr, MI/Cr in the frontal white matter and frontal cortex [39] whereas in those with dementia (HAD), these metabolite ratios were elevated in both in the basal ganglia and frontal white matter with reduced NAA [19,40]. However, single voxel MRS studies suffer from the limited areas of interrogation and a global understanding of the extent of this injury is therefore limited. To overcome this limitation multivoxel MRSI was employed [22,27,41]. In their study with short TE, Lopez-Villages observed elevated MI/CR in the frontal white matter in NAS and only decreased NAA/ CR and normal MI/CR in gray matter of patients with HAND [41].
A recent study by Boban et al. using long TE MRSI demonstrated that the impact of the HAND is more widespread low NAA/Cr ratios in the cingulate gyrus and subcortical frontal and parietal and deep frontal white matter in the chronic aviremic group. In the other two groups chronic HIV untreated and normal controls, Mi increased in the FWM as well as AC in the untreated group. However, it's important to note that this study only evaluated the blood but not CSF to characterise the aviremia and used the creatine as reference standard to measure the metabolites [37].

MRS and neuropsychology measures MRS showed a positive correlation with neurocognitive tests and hence can provide a positive complimentary information
The International HIV Dementia Scale (IHDS) was developed to screen for HIV-associated dementia (HAD), but it has been used more generally for HIV-associated neurocognitive disorder (HAND). These batteries of tests comprise of Trail Making Test A, WAIS-III Digit Symbol (DS) and HVLT-R Total Recall and DS, BVMT-R Total Recall and Grooved Pegboard Test-Dominant Hand. The objective is to study the functional domains that are most likely to get impacted in HIV [42].
HIV patients may show significant neurocognitive improvements in psychomotor speed as early as 6 months after cART [43]. However, the limited by availability of trained staff as well as practice effects makes the routine use of neuropsychology tests difficult to implement. Neuropsychological improvement has been shown to correlate with MRS changes thereby overcoming the latter mentioned difficulties with neuropsychological evaluation.
There have been several studies that have described the metabolic substrates underlying more detailed neuropsychological performance and these studies broadly indicate that the neuropsychological impairment is associated with markers of neuronal damage and increased markers of gliosis in the basal ganglia and frontal white matter.
Studies revealed a strong positive correlation between measures of gross and fine motor function and NAA/Cr ratios in the FWM (frontal white matter) and negatively with MI/Cr in the basal ganglia [41,44]. Similarly, cognitive processing speed was negatively correlated with MI/Cr in the basal ganglia [23,45,46] and frontal lobe dysfunction (as defined by neuropsychology) with Mi [23,41]. The ACTG 301/700 trial assessed changes in the NPZ-8 score (a summary score of a brief battery of neuropsychological tests) and MRS markers in the context of a trial of the efficacy of memantine in HAND. Improvement in NAA/Cr but not NPZ-8 scores was observed at 16 weeks of therapy [47].
"It's likely that MRS can play a role in the early detection of HAND, in determining its progression, and in assessing response to therapeutic interventions". [31].

Early HIV infection and MRS
Interesting research in early HIV infection by Lentz et al. (early was defined as HIV seroconversion and imaged within 60 days of an evolving Western blot, while still having detectable plasma virus) demonstrated neuronal dysfunction soon after infection: "reduced NAA and Glx in the cortical grey matter". They correlated the MRS markers with T-Cell phenotypes and found reduction in NAA and Glutamates in the frontal cortex [48]. These findings correlate with another similar study using volumetric and DTI analysis in the first 100 days of HIV infection wherein they demonstrated that there are volumetric (loss of parenchymal volume) and diffusion changes. This study contributes to the postulation that the initial early inflammation secondary to unchecked viremia may lead to significant neural injury. Furthermore, the focality of these findings suggests intrinsic brain inflammation but the effect of systemic inflammation is still possible.

MRS and other imaging correlates
Volumetric imaging, DWI techniques and Single Photon Emission Computerized Tomography (SPECT) studies have been found to correlate with MRS derived metabolites and neurocognitive scores. However, volumetric measures and SPECT perfusion techniques appear to be less sensitive than MRS, while DWI and fMRI appears to show equal sensitivity overall.

Volumetry and MRS
For volume loss to occur, there should be significant neuronal and axonal loss. It is therefore not surprising that MRS is more sensitive than volumetry to identify cognitive dysfunction (Paul et al. 2008: Caudate volumes and cognitive function) [49]. Further analyses have shown that NAA reduction correlates with total cortical volume loss while increased choline correlated with increased brain volume (possibly related to inflammation and oedema). Glx, osmoregulator and excitotoxic ion, is notable for its close relation to subcortical structural volumes [29].

SPECT and MRS
Ernst et al. studied the comparative sensitivity of the SPECT derived blood flow and the metabolite concentrations in patients with HAND. While the SPECT studies showed some trend in decrease in the rCBF in the temporoparietal white matter, MRS was abnormal in multiple areas (reduced Cr concentration in the BG and increased Mi concentrations in the BG and the temporoparietal white matter) pointing to the superiority of MRS [50].

DWI and MRS
Diffusion imaging explores the function of Brownian motion of the hydrogen proton in the CNS environment which allows assessment of the micro-architectural detail of white matter tracts and its integrity. DWI MR metrics showed strong positive correlations with glial metabolites. The increased diffusion (mean diffusivity) is thought to be associated with increased glial activation and inflammation [51]. The relationship of the metabolites and the treatment responses were also correlated with DTI (Diffusion Tensor Imaging) metrics FA (Fractional Anisotropy) and MD (Mean Diffusivity). Further, one study of Lithium as a treatment showed improving neuropsychology scores corresponded with an increase in the FA and reduction in MD along with reducing levels of the neuronal metabolite complex Glx [52].

Gliosis associated with neuroinflammation plays a crucial role in influencing the patterns of fMRI. The striatofrontal cortical and subcortical involvement in HIV influences working memory impairment and is reflected in fMRI.
The Blood oxygen dependent neurovascular coupling response that is associated with any motor task leads to local tissue level perfusion changes which is exploited in fMRI. fMRI has been extensively investigated in HAND to assess different neurocognitive domains with specific tasks. The working memory network which comprises the posterior parietal cortex and lateral prefrontal cortex were correlated with neurometabolites. The glial metabolites Mi, Ch and Cr-T in the basal ganglia and frontal white matter were positively correlated with increasing loads on the working memory derived signal strength [53]. This study also observed that the working memory network was not correlated with the concentration of NAA, which are markers of neuronal viability. This is in keeping with the mechanism of HIV associated injury which results in increased glial activation usually without significant neuronal abnormality until the disease is more advanced.

Biochemical measures and MRS
Chemokines are multifunctional, immunomodulatory proteins that influence HIV neuropathogenesis by multiple mechanisms. In raised concentrations they are considered to be potentially neurotoxic. Several studies have shown the correlation between neurometabolites and chemokines (CXC chaemokine IP-10 and CC chaemokine MCP-1 (monocyte chemoattractant protein).
Metabolite improvement has been correlated with chemokines especially CSF MCP-monocyte chemoattractant protein (MCP-1) and IP-10. Higher MCP-1 levels are inversely associated with neuronal dysfunction (NAA and Glutamates) in untreated patients. After cART, MCP-1 is correlated with high glial response (increased Cho and Mi) rather than NAA. After 3 months of cART, the decreased systemic factors (viral burden, systemically derived MCP-1) were no longer associated with neuronal dysfunction, but subjects with the strongest glial response in the brain continue to produce the highest levels of MCP-1 [46]. Letendre et al. also identified higher levels of IP 10 correlating with reduced neuronal scores and higher basal ganglionic and inflammatory pattern scores by using advanced statistical paradigms [54].
"These studies showed that the metabolite factor analysis method has a direct correlation with neuroinflammation and neuronal dysfunction as measured by chemokine pathway of IP 10 and MCP"

Excitotoxic injury and glutamatergic pathway
HAND is likely underpinned at least in part by neurotoxicity due to glial activation. Several recent studies have elaborated on this by showing the importance of Glu mediated excitotoxicity in the extracellular compartment. Glu is an amino acid which performs an important role in neurotransmission though when concentrations are high excitotoxic brain injury may occur. The possible mechanisms for increased extracellular glutamates include attenuation of astrocytic reuptake due to HIV infection and excessive production of Glu by HIV infected macrophages eventually leading to loss of the glutamates in the intracellular space [55,56].
Several authors have demonstrated reduced intracellular Glx in the frontal white matter. Frontal grey matter and parietal grey matter and basal ganglia also showed abnormal Glx levels. These studies showed a gradient in the reduction in Glx in NCS from MND to Dementia with increasing levels of reduced Glx with increasing neurocognitive impairment [56,57].
As stated earlier, study by Lentz et al, reduced Glu in the frontal cortical gray matter is a sensitive marker for identification of the CNS involvement by HIV virus [48].
Glx also appears to be sensitive to the effects of nucleoside reverse transcriptase inhibitors on the CNS. Reduced parietal grey matter Glx appeared to correlate with toxicity associated with nucleoside reverse transcriptase inhibitors [reduced astrocytic reuptake of Glu, secondary excitotoxicity, and mitochondrial toxicity from antiretroviral treatments [58].
"The current evidence points to Glu may be not only useful as sensitive marker for very early neuronal injury, but also a good surrogate marker for disease severity and treatment effects".

Patterns of pathological substrate: correlation between metabolites and HAND staging
The consequences of chronic HIV are multidimensional and the clinicopathological spectrum appears to be influenced not only by inflammation and neuronal loss but also by area of involvement. Mohammed et al. used these three patterns to prognosticate HAND staging with a specific pattern using advanced statistical algorithms [59].
Mohamed et al. employing factor analysis has identified three metabolic patterns: 1: the inflammatory factor.
(which was associated with mainly MI/Cr elevations in all three regions (BG, FWM and Parietal cortex) plus Ch/Cr increases in the centrum semiovale and parietal cortex), 2: the BG factor (associated with mostly NAA/ Cr and Ch/Cr elevations in the BG), and 3: the neuronal factor (associated primarily with NAA/Cr reductions in the centrum semiovale and the parietal cortex). These factors were found to be useful in discriminating between the groups of cognitively impaired and unimpaired participants, with the neuronal pattern being strongly associated with HAND, that is ADC staging as it was known then [48,60].

MRS and comorbidities
Patients living with HIV have several comorbidities and characterisation of their contribution to NCI has been a challenge. Some of these comorbidities include aging, alcohol abuse, drug abuse and cART neurotoxicity.

Aging
The increased life expectancy of the HIV positive individuals on cART has given rise to new challenges in characterising the cognitive impairment in this subgroup. Premature brain aging can be defined as greater than normal age-related deficits, but with similar or no greater rate of decline across the age spectrum, while accelerated brain aging can be demonstrated by a steeper rate of decline in brain measurements with age.
MRS studies performed during the pre-cART era or on antiretroviral-naive subjects suggest an accelerated aging process, while those on cART subjects suggest premature brain atrophy [61]. In normal aging there are changes in the metabolite concentrations in various regions of the brain with a slow steady increase in Mi and possibly some increase in the Ch and Cr [7] similar to HIV associated neural injury. However, HIV subjects who were naïve to cART showed greater than age-related decline in NAA and tCr in the BG and showed greater than age-related increases in Mi (+ 12% instead of 3%/ decade) and Cho (+ 10% instead of 2%/decade) in the FWM, indicating that there is accelerated ageing in HIV disease [62]. Those who were on cART, showed premature ageing, with higher than normal Mi in the FWM across the age span and lower than normal tNAA/tCr in the FGM n the medial frontal cortex [59]. A multicentre study by Harezlak et al. [62] also reflected similar observations indicating that there is premature ageing. However, study by Cysique et al. have shown only a synergistic effect between the age and serostatus. Decreased NAA/ Chaganti and Brew AIDS Res Ther (2021) 18:65 unsuppressed water signal in frontal WM was associated with older age, particularly in the HIV + group. Ernst et al. observed decline in glutamate concentration with aging in parietal gray matter. In addition, this study also observed a trend for lower brain glutamate levels in the parietal and frontal cortex in HIV + individuals, more pronounced in individuals with cognitive impairment.
In conclusion, HIV related MRS abnormalities are variably worse with ageing. The MRS ageing changes in the cART era are more prominent with age, suggesting premature ageing. However, all these studies are cross sectional and may suffer from bias from intersubject variability. Longitudinal studies are needed.

Alcoholism and HIV
"Alcohol together with HIV appears to heighten the risk of neuronal loss in the brain".
Growing numbers of HIV patients abuse alcohol which is well known to be associated with accelerated brain aging [63]. One study comparing HIV + Alcohol and HIV -Alcohol groups using long TE MRSI demonstrated almost a full standard deviation reduction in NAA and Cr independent of cART status in the parietal-occipital region for the HIV + alcohol group [63]. On proton MRS, the Cho and Mi in the striatum were higher both in alcoholics with HIV and those who were not alcoholics and did not have any statistically significant change [64]. Although neither HIV infection nor alcoholism alone resulted in such a deficit, each disease carried a liability that put dually affected individuals at a heightened risk of neuronal compromise. These findings support the hypothesis that the alcohol has cumulative effect. Meyerhoff in pre cART era using 31P MRS observed that individuals with both conditions had augmented metabolite deficits but not interactive or synergistic.

Methamphetamine effects of HIV (Meth)
"Limited studies show that there is likely an additive effect on metabolites in patients with meth-abuse. " The effects of Meth on the brain are similar to HIV and the metabolic abnormalities are mainly localised to the frontal cortex, FWM and BG areas which are also typically involved in HIV. There are only a few studies available in patients with HIV with meth abuse, and these largely showed concordant results. The characteristic observations on MRS include markedly reduced NAA in both the BG and FWM and reduced Cr elevated Cho in the basal ganglia [64][65][66].
Metabolite changes in chronic meth users with HIV appears to be additive. HIV-negative subjects with a history of chronic meth use showed lower concentrations of the NAA in the FWM BG and higher concentrations of Cho compounds and Mi in the frontal cortex, relative to subjects with no history of drug abuse [66,67].
However, Taylor et al. study concluded that there is likely no effect [68]. Although Taylor et al. argues that the increase in the volumes of the BG could have influenced the levels of the metabolites, it is difficult to explain how that may be so, given that the meth only group consistently demonstrated increased Cho in FWM and reduced NAA in the BG which are also observed in HIV [68].

Antiretrovirals and NCI
Whether antiretrovirals can cause brain neurotoxicity is still in debate. Reduction in NAA secondary to mitochondrial injury may occur with Zidovudine (AZT) and other nucleoside reverse transcriptase inhibitors [69,70]. Didanosine and Stavudine use has been associated with lower NAA when compared to the sero negative controls in the FWM [70]. However, a study using Glx as a marker did not show any differences in Glx between naïve and cART treated patients and concluded that the toxic effects of cART are therefore unlikely [71]. Nonetheless Ernst et al. identified reduced parietal gray matter Glu in HAND patients while normal in seronegative controls and opined that the resultant excitotoxic injury is secondary to NRTIs [58]. "The likely cause for these differing observations may be secondary to differing cART regimens as well as individual host susceptibility. The reduction in the Glu could be secondary to reduced astrocyte reuptake of Glu, secondary excitotoxicity, and mitochondrial toxicity from antiretroviral treatments".

Conclusion
MRS can be a useful biomarker to measure brain injury in HIV and is likely to guide understanding of neurocognitive impairment at the cellular level. In the setting of cognitive decline, the MRS findings are increased Ch and Mi both in the FWM and BG early on and in the later stages there is compromise of the neuronal integrity signalled by reduced NAA. Creatine concentrations can change depending on the severity of dementia and therefore concentrations either absolute or relative to internal water are more sensitive. Although well established as early as in 2002 by Chang et al., multiple studies thereafter have noted the metabolite changes based on creatine; this is perhaps an issue that should be incorporated into guidelines. MRS changes correlating with cART improvement can take as long as 6-12 months. However, recent evidence with measurement of the excitotoxic pathway using as Glx and neuroinhibitory pathway using GABA have shown that they are the earliest to be identified in MRS [42].
A recent metaanlysis by Chelala et al. on a pooled study of 61 spectroscopic studies concluded that there is consistently lower tNAA/tCr, higher tCh/tCr and higher mI/tCr ratios associated with chronic HIV infection. However, this study does acknowledge that the Creatine contribution needs to be assessed independently and interpretation is difficult because of varying treatments, duration of infection, and comorbidities [72].
In summary then MRS is a useful tool in HIV infection to assist in HAND identification, response to cART, and insight into neuropathogenesis. NRTIs suppress HIV replication, but with mitochondrial toxicity

T FWM and gray mater
Reductions in NAA in individuals taking didanosine and/or stavudine may be the result of depleted brain mitochondria and/ or alterations in cellular respiration. Measurement of brain metabolites sensitive to impairments in energy metabolism, including NAA, may aid in early detection of subclinical NRTI-mediated mitochondrial toxicity