Impact of HIV and recreational drugs on cognitive functions in young men having sex with men

Participants

From December 2017 to June 2020, 25 western European, and native French-speaking HIV+ men (mean age: 38.8 years, age range: 18-60 years, >9 years of education) not declaring any cognitive impairment, were prospectively included. The exact duration of HIV infection was available for all of them (either the diagnosis was done during the acute infection or the presumed duration was based on negative HIV screening close to HIV diagnosis). All HIV+ subjects were taking cART with undetectable viral load for at least 6 months before inclusion and had a CD4+ cell count greater than 500/mm3. Those with a coinfection with hepatitis C, untreated syphilis, psychiatric disease, efavirenz based regimen (cART with brain side effects), or any history of brain disorder were excluded.

A total of 51 men matched for age with HIV+ subjects were also included. Those participants satisfied to the same criteria in terms of age, education and ethnicity as the HIV+ group. 26 of them were selected from the HIV-PrEP outpatient clinic. Fourteen HIV-PrEP users were taking TDF/FTC daily (30 pills/month), while 11 were taking it on demand before having a risky sexual intercourse (mean number of 15 pills/month) (11). The other 25 men were selected from the general population and constituted the healthy controls group based on their clinical background, lifestyle, and normal cognitive evaluation.

Participants contributed to the study approved by the institutional Ethics Committee after written informed consent (References: P2017/541, CCB: B4062017344197).

Clinical and neuropsychological assessment

The clinical evaluation included an exhaustive anamnesis in all participants to collect their medical and psychiatric background. HIV+ subjects and HIV-PrEP users also underwent a comprehensive neuropsychological assessment investigating the main cognitive functions such as working memory, long-term episodic memory (on both verbal and visual material), gestural and visuo-constructive praxis, language (verbal fluencies and picture naming), attentional and executive functions. Questionnaires were used in all participants to assess depression, anxiety, previous night of sleep and cognitive functioning in daily living. Healthy controls’ broad cognitive functions were assessed with the Montreal cognitive assessment (MoCA) (12). RDU was evaluated in all participants through a semi-directed interview inspired by “the interview for research on addictive behavior” (13) that evaluated the frequency, the chronicity and the active/past character of the consumption of a series of drugs commonly used by MSM in the context of chemical sex (i.e., cocaine, synthetic and crystallized methamphetamine, ketamine, gamma-hydroxybutyrate/gamma-butyrolactoneheroin (GHB/GBL), poppers, cannabis) (14). An operational composite score (RDU score) was computed and incorporated the current and past exposition to recreative drugs. It was computed by adding the number of RDU on a monthly basis at the time of the study as well as those used on a monthly basis in the past (RDU= number of drugs used in the past + number of drugs currently used). This quantification was however a coarse-grained appreciation of the actual and past exposure as the majority of participants were unable to precisely evaluate the frequency and dosage of each current and past recreational drug due to the high variability in their use. Furthermore, to assess the potential toxicity of TDF/FTC (i.e., the cART shared by HIV+ subjects and HIV-PrEP users) on brain structure and function, an index was computed based on the dosage since the start of use (dosage=number of months of use). For HIV-PrEP users taking on demand regimen, the number of months of use was divided by two based on a mean number of fifteen pills/month (17).

Analyses of neuropsychological data

To classify our participants according to Frascati’s criteria, individual performance of HIV+ subjects and HIV-PrEP users in the different tasks was then expressed as z-scores calculated by comparison with normative data. Here, we calculated z-scores for each task by subtracting the mean score of the normative data from a participant raw score, and then dividing the difference by the standard deviation of the normative data. For between-group comparison, we computed composite and index scores to quantify cognitive processes and reduce the number of data and variables. Scores of HIV+ subjects and HIV-PrEP users were compared using Student’s t-tests. Multiple comparisons were controlled using the Bonferroni correction. Relationships between cognitive scores and indexes, and TDF/FTC index or RDU score were assessed using Pearson’s correlations.

PET-MR data acquisition

Cerebral FDG-PET and structural MRI data were obtained simultaneously using a 3T hybrid PET-MR scanner (SIGNATM, GE Healthcare). Structural MRI sequences for parenchymal evaluation (whole-brain axial 3D T1-weighted imaging (WI), axial T2WI, sagittal 3D T2 Fluid-Attenuated Inversion Recovery, axial 3D susceptibility-WI, and axial diffusion-WI) were acquired as previously described (15). Intracranial arteries were evaluated in HIV+ subjects to exclude active primary angiitis of CNS related to HIV status (16) by means of an axial 3D time-of-flight sequence, along with a pair of 3D T1-weighted “Black Blood” volumes, acquired before and after gadolinium-based contrast medium injection. For PET data acquisitions conducted as previously described (21), participants fasted for at least 4 hours, were awake in an eye-closed rest, and received an intravenous bolus injection of 3-5 mCi (111-185 MBq) of FDG. Forty minutes post-injection, a 20-min data acquisition was performed.

MRI data preprocessing and analyses Qualitative analyses

MRI data were reviewed by one experienced neuroradiologist (T.C.) following a systematic and comprehensive visual assessment procedure searching for chronic cerebral small vessel disease and other parenchymal abnormalities, including extra-axial masses.

Quantitative analyses

The Freesurfer 6 image analysis suite (http://surfer.nmr.mgh.harvard.edu/) was first used to obtain cortical volume and thickness using conventional surface-based approach (17-19) as well as to extract brain regions of interest (ROIs) based on anatomical brain parcellation (see supplemental table 1 (20, 21).

Student t-tests were used to search for differences in (i) maps of whole brain cortical volume and thickness and (ii) ROIs averaged volume (for cortical and subcortical ROIs) and thickness (for cortical ROIs only). These differences were tested between HIV+ subjects (taken individually or as a group) compared to HIV-PrEP users, and in HIV+ subjects and HIV-PrEP users (taken as separate or one group(s)) with that of healthy controls. For whole brain surface-based analyses, statistical maps were corrected for multiple comparisons using the False Discovery Rate (FDR) as implemented in Freesurfer 6, with the significance level set to p<0.05. For ROIs analyses, significance was considered at p<0.05 Bonferroni-corrected for the number of ROIs.

PET data preprocessing and analyses

FDG-PET data were preprocessed and analyzed using the voxel-based Statistical Parametric Mapping software (SPM8, http://www.fil.ion.ucl.ac.uk/spm/, Wellcome Trust Centre for Neuroimaging, London, UK) based on conventional preprocessing, (individual and group level) subtractive and correlation analyses previously described (22-25).

Subtractive analyses identified brain areas where glucose metabolism was significantly lower or higher in HIV+ subjects (taken individually or as a group) compared with HIV-PrEP users. Additional group-level analyses compared the regional cerebral glucose metabolism of HIV+ subjects and HIV-PrEP users (taken as separate or one group(s)) with that of healthy controls. To search for a pathophysiological link between significant hypo- and hypermetabolic brain areas found in HIV+ subjects and HIV-PrEP users compared with healthy controls, we also performed pathophysiological interaction (PPI) analyses as previously described (25, 27). We finally performed correlations between the regional cerebral glucose metabolism of HIV+ subjects and HIV-PrEP users taken as one group and RDU score, TDF/FTC index, and neuropsychological data. Results were considered significant at p < .05 corrected for multiple comparisons over the entire brain volume (Family Wise Error (FWE)) or at a more liberal threshold (p < .001 uncorrected, height threshold: 0.001, cluster size k ≥ 50 voxels; or small-volume-corrected P<0.05 using a 10-mm radius spherical volume of interest).

Clinical characteristics

Table 1 describes the clinical characteristics of each group of participants.

No significant difference was found between HIV+ subjects and HIV-PrEP users in age, levels of anxiety/depression/quality of sleep, RDU score, and mean number of abnormal neuropsychological fields. Based on Frascati’s criteria, only 2 HIV+ subjects and 2 HIV-PrEP users were classified as normal, while 21 HIV+ subjects and 23 HIV-PrEP users were classified as having asymptomatic neurocognitive impairment (ANI) as they did not report any cognitive complain but were impaired in 2 or more cognitive domains (>1.0 standard deviation (SD) below the mean in a demographically appropriate normative mean). At the group level, performance was significantly lower in HIV+ subjects compared with HIV-PrEP users in executive functions (inhibition, p_corr=0.038; combined EF, p_corr=0.011; planification, p_corr=0.031), attentional functions (vigilance, p_corr=0.01; divided attention, pFWE=0.019) and working memory (WM updating, p_corr=0.009), see Fig. 1. No correlation was found between RDU score and neuropsychological performance. Significant correlations, were found between TDF/FTC index and executive functions (combined EF, p_corr=0.005, r=0,40; planification, p_corr=0.039, r=0,3) in the form of a negative impact of TDF/FTC on executive functions.

• Download figure
• Open in new tab

Figure 1

Subtractive analyses comparing FDG-PET data of HIV+ subjects and HIV-PrEP users taken as a group with those of healthy controls. Significant relative hypometabolic areas were observed in dorso-lateral prefrontal cortex bilaterally, the right dorso-medial prefrontal cortex, as well as in the left prefrontal and insular cortices.

Structural MRI analyses

Regional cerebral glucose abnormalities

Figures 1-2 and Supplementary Tables 2-3 describe the group-level regional cerebral glucose abnormalities.

• Download figure
• Open in new tab

Figure 2

Subtractive analyses comparing FDG-PET data of HIV+ subjects and HIV-PrEP users taken as a group with those of healthy controls. Significant relative hypermetabolic areas were observed in the precuneus and cuneus bilaterally, the right primary somatosensory and posterior cingulate cortices, as well as the left superior temporal gyrus.

No group-level difference in regional cerebral glucose metabolism was found between HIV+ subjects and HIV-PrEP users (even at p_uncorrected<0.001). When HIV+ subjects and HIV-PrEP users were pooled in one single group and compared with healthy controls, significant (p_FWE<0.05) hypometabolism was found in the lateral and mesial (right > left) prefrontal cortex as well as significant (p_FWE<0.05) hypermetabolism in posterior midline cortices bilaterally (precuneus and posterior cingulate cortex (PCC)). Similar prefrontral hypometabolism was found when comparing the brain metabolism of HIV+ subjects or HIV-PrEP users with that of healthy controls. When considering the peak voxel value in the hypometabolic right dorso-lateral prefrontal (DLPFC, MNI coordinates: [60, 14, 28]) and dorso-mesial prefrontal (DMPFC, [16, 64, 4]) cortices found in the pooled group of HIV+ subjects and HIV-PrEP users, no significant (event at p_uncorrected<0.001) correlation was found between the level of metabolism in those brain regions and the rest of the brain, suggesting that prefrontal hypometabolism and posterior midline hypermetabolism were unrelated.

Correlation analyses were then performed between regional cerebral glucose metabolism and RDU score or TDF/FTC index in HIV+ subjects/HIV-PrEP users taken as one group. These analyses revealed a significant (p_FWE <0.05) negative correlation between RDU score and the level of metabolism in a right prefrontal cluster (right DLPFC, [60, 14, 28], (p_FWE<0.01), r=-0.51; right DMPFC, [16, 64, 4], (p_FWE<0.01), r=-0.52), see figure 3). No significant correlations were found between TDF/FTC index and prefrontal cortex metabolism. Finally, we did not find any significant correlation between regional cerebral glucose metabolism and the score of neuropsychological tests that significantly differed between HIV+ subjects or HIV-PrEP users. Figure 4 illustrates the significant hypometabolism observed at the individual level in 6 HIV+ subjects compared with HIV-PrEP users in the DLPFC or DMPFC (2 at p_FWE<0.05, voxel level; 2 p_FWE<0.05, cluster level; 2 at p_uncorrected<0.001).

• Download figure
• Open in new tab

Figure 3

Results of the correlations analyses performed between the PET data of the HIV-PrEP and HIV+ and the cumulated drugs scores. Regression plots of cumulated drugs scores and adjusted metabolic responses were obtained by considering the peak voxel in (A) the right dorso-lateral prefrontal cortex [60, 14, 28] (Pearson’s correlation: r = -0.48 p < 0.001) and (B) the right dorso-medial prefrontal cortex [16, 64, 4] (Pearson’s correlation: r = -0.54, p < 0.001). These plots showed significant positive correlation between cumulated drugs scores and cerebral glucose metabolism in the considered voxels.

• Download figure
• Open in new tab

Figure 4

Subtractive analyses comparing FDG-PET data of HIV+ subjects taken individually with those of the group of HIV-PrEP users. Significant relative hypometabolic areas were observed in lateral or mesial prefrontal cortices of 25% of the HIV+. The figure shows for illustrative purposes two of these subjects exhibiting the aforementioned metabolic abnormalities.

Neuropsychological assessment

Based on Frascati’s criteria, the prevalence of ANI (i.e., >1.0 to <2.0 SD below mean of a normative population in two or more domains but no decrease in everyday functioning) in our group of HIV+ subjects appeared especially high compared to the 26-76% previously reported (1). Surprisingly, the incidence of ANI was similar in our HIV-PrEP users. Thus, Frascati’s criteria do not optimally distinguish HIV+ subjects with HAND from HIV-subjects with similar life and health habits. They result in a high frequency of false-positive diagnosis due to the use of 1.0 SD deviation below mean, leading to an overestimation of HAND prevalence (1). Studies relying on those criteria might therefore suffer from an inadequate selection of neurocognitively impaired individuals. As previously suggested (1), our data further ask for the revisit of those criteria. The use of 2.0 SD below means of a normative population, as classically used in clinical neuropsychology, would be more appropriate to identify HIV+ subjects with HAND. Still, group-level analyses disclosed that HIV+ subjects had significantly lower performances in executive, working memory and attentional functions compared with HIV-PrEP users, which confirmed the dysexecutive/inattentive profile of HIV+ subjects previously reported (2).

PET/MR results

Group-level analyses did not reveal any difference in regional brain structure or glucose metabolism between HIV+ subjects and HIV-PrEP users. This finding questions the pathophysiological role of HIV in the prefrontal hypometabolism found in HIV+ subjects when compared with healthy controls selected from the general population. Still, individual-level analyses showed that 25% of HIV+ subjects exhibited lateral or mesial prefrontal hypometabolism when compared with HIV-PrEP users. These changes at the individual-level were in line with the localization of the relative hypometabolism previously described in HIV+ subjects (6-9). The prefrontal hypometabolism found in some HIV+ subjects might potentially relate to the abnormalities noted in executive functions, although prefrontal metabolism did not significantly correlate with cognitive scores. Alternatively (but not exclusively), this hypometabolism, especially in the ACC, might also play a causative role in the adoption of high-risk behaviors in a drug and sexually-transmitted disease setting. Indeed, this brain area plays a pivotal role in decision making and risks/benefits balancing when emotional drive and objective reasoning come into conflict (26, 27). It may thus be causative rather than the consequence of the contracted disease (26, 27). This hypothesis warrants further investigations. The absence of group-level metabolic difference between HIV+ subjects and HIV-PrEP users suggests that their common brain hypometabolic abnormalities in comparison to healthy controls might be driven by another cause than HIV. Correlation analyses pointed to the well-known vulnerability of the prefrontal areas to the effects of RDU (9, 28). In the absence of noticeable structural brain changes, hypometabolism was probably related to neural dysfunction rather than neuronal loss per se. RDU-induced neural dysfunction might be related to dopamine dysregulation, neuroinflammation or neurodegeneration through the accumulation of intraneuronal hyperphosphorylated TAU protein (pTAU) (29-31). In HIV+ subjects, recreational drugs might exert their neural toxicity synergistically with HIV-related neuroinflammation, neurodegeneration or the reactivation of deep latency virus (30, 32). A complex interplay between RDU and HIV might thus be involved in the induction and development of the prefrontal hypometabolism and dysexecutive/inattentive profile found in our HIV+ subjects. Still, results obtained in HIV-PrEP users suggest that RDU might play a predominant pathophysiological role compared with HIV. The absence of correlation between regional brain metabolism and cognitive alterations observed in our subjects might be explained by the inability of FDG to explore neuroinflammation or cerebral pTAU deposition. Other radioligands targeting neuroinflammation-related or pTAU proteins could thus be used and correlate with cognitive dysfunctions (33).

Hypermetabolism in posterior midline cortices (precuneus and PCC) belonging to the default mode network (DMN, (34) were also found at the group level in HIV+ and HIV-PrEP compared to healthy controls. Correlation analyses failed to find any relationship between the level of prefrontal hypometabolism and posterior midline cortices hypermetabolism, suggesting that these metabolic abnormalities have no pathophysiological link. The posterior midline cortices hypermetabolism, which did not correlate with RDU score, could be related to attention/dysexecutive alterations (35).