Blood samples were collected from 144 participants recruited in the study “EndoAfrica” (27). Patients were recruited upon presentation at HIV clinics or community health centres in Cape Town, South Africa. On the day of recruitment, participants underwent procedures, including informed consent, completion of a health questionnaire, measurement of body mass index, waist circumference, and waist–hip ratio, assessment of blood pressure and heart rate, provision of a urine sample, collection of fasting blood, and examination of flow-mediated dilatation (FMD) and carotid intima-media thickness (cIMT); most of the clinical characteristics were also assessed on the day of recruitment. The Table 1 is divided into groups: control volunteers (n = 50), PLWH on ART (PLWH ART+) (n = 50), and untreated PLWH (PLWH ART−) (n = 44). Patients were matched between groups for age and sex. The number of patients taking statins was not significant, as only one patient was under statins (simvastatin) at the time of blood collection. This patient belongs to the PLWH ART– group. Informed written consent was obtained from all the participants. The study adhered to the ethical guidelines of the Declaration of Helsinki and was reviewed and cleared by the Ethical Committee of the University of Stellenbosh (South Africa).

PLWH included in this study were over 18 years old, not pregnant, and over 3 months post-partum. PLWH on ART were treated with efavirenz 600 mg + tenofovir DF 300 mg + emtricitabine 200 mg (n = 46), Nevirapine 200 mg (NVP)/lamivitudine 150mg + zidovudine 300 mg (Lamzid) (1), Lamivudine/Efavirenz/Abacavir (1), and Ritonavir 50 mg + lopinavir 200 mg (Aluvia)/Tenofovir/Lamivudine (1). The median duration of treatment was 176 weeks, with an interquartile range of 108–300 weeks.

PLWH-ART naïve showed a median time between HIV diagnosis and sampling of 65 days, with an interquartile range of 23.5–406 days (n = 32).

A qualified research nurse conducted anthropometric measurements, urine collection, and phlebotomy. Anthropometric measurements, including height (cm) measured with a stadiometer, weight (kg) measured using an electronic scale, and waist and hip circumferences (cm) measured with a measuring tape, were performed per standardised protocols. Body mass index (BMI, kg/m²) and waist-to-hip ratio were calculated. Systolic blood pressure (SBP, mmHg) and diastolic blood pressure (DBP, mmHg) were measured thrice at 5-min intervals using an Omron M6 automatic digital blood pressure monitor (Omron Healthcare, Kyoto, Japan) around the left brachium. Subsequently, the average value was calculated. Study participants fasted from 10:00 pm the night before clinical assessments. Participants with unknown HIV status were tested for HIV using a rapid HIV test (SD Bioline HIV 1/2 3.0 immunochromatographic kit; Standard Diagnostics, Republic of Korea) to determine their HIV status. Urine and fasting blood samples were collected and sent to the National Health Laboratory Service (NHLS) for biochemical analyses using standard laboratory techniques. Plasma lipid profiles [total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and triglyceride (TG) levels, mmol/L] were determined using a chemiluminescence methodology (cobas® 301/501 analyser, Roche/Hitachi cobas® c systems, Basel, Switzerland). Levels of liver enzymes [gamma-glutamyl transferase (GGT, U/L) and alanine aminotransferase (ALT, U/L)] were determined via an enzymatic chemiluminescence methodology using a cobas® 311/501 analyser (Roche/Hitachi cobas® c systems, Basel, Switzerland). Levels of high-sensitivity C-reactive protein (hsCRP, mg/L) were obtained via an IMMAGE® Immunochemistry Systems and Calibrator 5 Plus assay kit (Beckman Coulter, Inc., CA, USA). Fasting glucose levels (mmol/L) and glycated haemoglobin (HbA1C, Hb%) were determined using a chemiluminescence methodology (haemolysate application on the cobas® 311/501 analyser, Roche/Hitachi cobas® c systems, Basel, Switzerland). Haemoglobin (Hb, g/dL) levels were determined using a chemiluminescence method (whole blood application on the cobas® 311/501 analyser, Roche/Hitachi cobas® c systems, Basel, Switzerland). Urine samples were analysed to determine microalbuminuria (mg/L) and creatinine (mmol/L) using an enzymatic chemiluminescence method by cobas® 501/502 and cobas® 311/501 analysers (Roche/Hitachi cobas® c systems, Basel, Switzerland), respectively, and the albumin-to-creatinine ratio (ACR) (mg/mmol) was computed. In HIV+ participants, the levels of cluster of differentiation four (CD4)+ T-cell count and viral load (VL) were determined by flow cytometry (FC 500 MPL) with MXP software (Beckman Coulter, Brea, CA, USA) and the COBAS® AmpliPrep/COBAS® TaqMan® HIV-1 Test, v2.0 (Roche Diagnostics Ltd., Burgess Hill, UK), respectively (28).

Blood samples were collected in Becton Dickinson vials to obtain serum or EDTA plasma. After collection, the serum was allowed to coagulate at room temperature for 30 min. Serum and plasma samples were centrifuged at 2,300g for 5 min at room temperature, aliquoted, and stored at −80°C until further analyses. Apolipoproteins A1 and B were analysed by a Cobas e501 automated system using electrochemiluminescence technology from Roche Diagnostics (Roche, Rotkreuz, Switzerland) at the University Hospital of Geneva, Geneva, Switzerland.

Cardiovascular risk was assessed using the 10 years Framingham risk score (FRS) calculation, the extent of carotid atherosclerotic vascular disease was evaluated using the carotid intima-media thickness (cIMT) measurement, and the endothelial function was assessed using the flow-mediated dilation (FMD) measurement.

The FRS was calculated using the calculator prepared by R.B. D'Agostino and M.J. Pencina based on a publication by D'Agostino et al. (29). FRS calculation is based on gender, age, systolic blood pressure, treatment for hypertension, smoking, presence of diabetes, total cholesterol, and HDL cholesterol.

FMD assessment was performed using a MyLab™ Five mobile ultrasound system (Esaote, Italy). The FMD protocol followed previously published recommendations (30, 31). Participants were positioned supine on an examination bed, with their right arm abducted and supinated. A blood pressure cuff (deflated) was placed around the proximal part of the forearm. Subsequently, the ultrasound probe was positioned proximal to the cubital fossa (mid to distal humerus), just below the biceps brachii muscle belly, until the brachial artery was visually located on the ultrasound image. The ultrasound probe was then secured in this position using a probe holder. Next, cross-sectional still images were captured with the ultrasound probe at three different locations along the designated section of the artery to obtain baseline brachial artery diameter measurements. Following this, the blood pressure cuff was inflated to 200 mmHg (or 50 mmHg supra-systolic in the case of individuals with systolic blood pressure greater than 150 mmHg), and blood flow to the forearm was occluded for 5 min. After the blood pressure cuff was deflated, additional cross-sectional ultrasound stills were taken along the same section of the artery for a duration of 2 min. For still analysis, the built-in manual measurement tool of the MyLab™ Five mobile ultrasound system was used to measure the brachial artery diameter in millimetres, consistently at the end of diastole, in all the stills. The three baseline measurements were used to calculate a mean baseline brachial artery diameter, and the maximum post-occlusion measurement, usually at approximately 60 s after blood pressure cuff release, was used to calculate the FMD percentage according to the following formula:FMD%=maximumpost-occlusion(diameter(mm)−meanbaselinediameter(mm))meanbaselinediameter(mm)×100To ensure reliable data collection, the ultrasound operators were subjected to stringent training by experts and multiple practice sessions with student volunteers and colleagues.

To keep inter-operator variability to a minimum, only three trained and experienced operators were employed in this study, and image analysis and data acquisition were performed independently by a single person (32).

Of note, for statistical correlation tests, the FMD values used for correlation analysis included negative values. To overcome negativity, four was added to the FMD value to shift them from negative value and then the values were log2-transformed.

The carotid intima-media thickness (cIMT) was measured using an Esaote MyLab Five portable ultrasound device (Genoa, Italy) equipped with an Esaote Doppler probe (LA523, 12 MHz) and QIMT software, which automatically and accurately measures all the parameters needed for carotid IMT measurements. For the IMT protocol, participants were asked to lie in a supine position with their head resting comfortably and their neck slightly hyperextended and tilted 45° to the opposite side of the carotid artery being assessed. The operator (author of this dissertation) used the index and middle fingers to locate the pulsating carotid artery to guide the position of the probe on the participant's neck. The position of the probe was further adjusted until a clear and stable image was obtained on the ultrasound. The lateral angle of the image was assessed as far as possible. In cases where a lateral view was not optimal (clear), an anterior or posterior angle was used. Once the image was clear and stable, the region of interest was manually placed 5 mm proximal to the dilatation of the carotid bulb. Thereafter, QIMT software automatically detected and analysed the vascular boundaries in radio frequency (RF) mode. The software further calculated the diameter and the thickness of the intima-media layer, as well as the median and standard deviation IMT values, using high spatial resolution. At this stage, the IMT measurements were also calculated and expressed in micrometres (μm), with the standard deviation changing. Once the standard deviation fell below 25 μm (the value also turns green), the IMT image was frozen and saved. These automatic and accurate measurements are largely independent of the investigator and the device settings. For this study, both the left and right common carotid arteries were assessed, and values from both sides were averaged and used in data analysis.

All inflammatory biomarker analyses were performed on serum samples at the Geneva University Hospitals. Inter-cellular adhesion molecule-1 (ICAM-1), vascular cell adhesion molecule-1 (VCAM-1), interleukin (IL)-6, -8, -10, interferon (INF)-gamma, tumour necrosis factor (TNF)-alpha, macrophage inflammatory protein 1-alpha and -beta (MIP1 alpha and MIP1 beta), serum amyloid A (SAA) and monocyte chemoattractant protein-1 (MCP-1), were measured using the Meso Scale Discovery (MSD) platform (Rockville, MD, USA). Analyte concentrations were determined by Discovery Workbench® software 4.0 using a four-parameter logistic fit model. The lower limits of detection in pg/mL were as follows: IL-6 and IL-8, 0.04; IL-10, 0.03; INFγ, MIP1 alpha, MIP1 beta, SAA, TNF-α, 0.04; and MCP-1, 0.09. Intra-run CVs were below 7%, and inter-run CVs were below 15%.

Anti-apoA-1 IgG levels were measured as previously described (17, 23, 33). Briefly, MaxiSorp plates (NuncTM, Denmark) were coated with purified, human-derived delipidated apolipoprotein A-1 (20 μg/mL; 50 μL/well) for 1 h at 37°C. After washing, all wells were blocked for 1 h with 2% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) at 37°C. Patient samples were also added to a non-coated well to assess individual non-specific binding. After six washing cycles, 50 μL/well of signal antibody (alkaline phosphatase-conjugated anti-human IgG; Sigma-Aldrich, St. Louis, MO, USA), diluted 1:1,000 in a PBS/BSA 2% solution, was added and incubated for 1 h at 37°C. After six more washing cycles, phosphatase substrate p-nitrophanylphosphate disodium (Sigma-Aldrich) dissolved in diethanolamine buffer (pH 9.8) was added and incubated for 30 min at 37°C (Molecular Devices TM Filtermax 3). The optical density (OD) was determined at 405 nm, and each sample was tested in duplicate. Corresponding non-specific binding was subtracted from the mean OD for each sample. The specificity of detection was assessed using conventional saturation tests by western blot analysis. As previously described, elevated levels of anti-apoA-1 IgG (seropositivity) were defined by an OD cut-off of OD ≥0.64 and a ratio between the sample net absorbance and the positive control net absorbance × 100 above 37, corresponding to the 97.5th percentile of a reference population.

HAECs were purchased from Ruwag (Bettlach, Switzerland). Briefly, HAECs were grown in a complete EBM-2 medium supplemented with 5% FBS (foetal bovine serum). HAECs were stimulated in EBM-2 medium supplemented with 0% FBS for 24 h without or with polyclonal goat anti-apoA1 IgG (Academy Bio-Medical Co) (40 µg/mL) or polyclonal goat IgG (Meridian Life Science) (40 µg/mL). The protocol was chosen based on previous evaluations of the optimal time course for inducing a pro-inflammatory response, which determined that 24 h of stimulation with anti-apoA1 IgG at a concentration of 40 µg/mL yielded better results (21–24, 26, 33, 34). Afterwards, cells were collected, and RNA was isolated and analysed using a LightCycler 480 Real-Time PCR System (Roche) in 96-well plates. The amplification curves were analysed using Roche LightCycler software to determine Cp (by the second derivative method). Primers for human ICAM-1 (Hs 00164932_m1), VCAM-1 (Hs 01003372_m1), and GAPDH (Hs 99999905_m1) as housekeeping genes were used and purchased from Thermo Fisher (Basel, Switzerland). The stimulating medium was collected, and ICAM-1 and VCAM-1 were quantified using the Meso Scale Discovery platform, as described above. In the same supernatant, the levels of kynurenine, tryptophan and kynurenic acid were measured using liquid chromatography-multiple reaction monitoring/mass spectrometry (LC-MRM/MS).

Metabolites were extracted using a methanol–ethanol solvent mixture in a 1:1 ratio. After protein precipitation, the supernatant was evaporated to dryness and finally resuspended with 100 µL of 10% methanol (MeOH) in H₂O. The samples were analysed by LC-MRM/MS on a hybrid triple quadrupole-linear ion trap (QqQ_LIT) system (Qtrap 5500, Sciex) coupled to an LC Dionex Ultimate 3,000 (Dionex, Thermo Scientific). Analysis was performed in positive and negative electrospray ionisation modes using a TurboV ion source. The MRM/MS method included 299 and 284 transitions in positive and negative modes, respectively, corresponding to 583 endogenous metabolites. The chromatographic separation was performed on a Kinetex C18 column (100 × 2.1 mm, 2.6 µm). For the positive mode, mobile phase A consisted of 0.1% FA in H₂O and mobile phase B consisted of 0.1% FA in acetonitrile (ACN). For the negative mode, mobile phase A consisted of 0.5 mM ammonium fluoride in H₂O and mobile phase B consisted of 0.5 mM ammonium fluoride in ACN.

The linear gradient program was as follows: 0–1.5 min, 2% B; 1.5–15 min, up to 98% B; 15–17 min, held at 98% B; 17.5 min, down to 2% B, with the flow rate maintained at 250 µL/min.

The MS instrument was controlled by Analyst software v.1.6.2 (AB Sciex). Peak integration was performed with MultiQuant software v.3.0 (AB Sciex). The integration algorithm was MQ4 with a Gaussian smoothing of a half-width equal to 1.5 points.

Data obtained after MSTUS (MS Total useful signal) normalisation were used (35, 36). For better visualisation, a factor of 10⁶ was applied after normalisation.

Non-parametric Kruskal–Wallis and Mann–Whitney tests were used for assessing differences between groups, and Spearman tests were used for correlation analysis. In in vitro experiments with HAECs, paired Wilcoxon tests were used for analysing differences between groups. For metabolite analyses, the data were expressed as log-normalised. The correlation between anti-apoA1 IgG, cIMT, FMD, and metabolites was analysed using Spearman’s rank test. Differences between groups were tested using the Mann–Whitney test. P < 0.05 was considered as significant. Statistical analyses were performed using Prism 9 (GraphPad Prism, CA, USA).

Participants were divided into three groups: HIV-free control volunteers (n = 50), HIV-positive patients on ART (n = 50), and HIV-positive ART naïve patients (n = 44) (Figure 1). Cardiovascular risk profile, subclinical atherosclerosis status, and endothelial dysfunction were assessed using FRS, cIMT, and FMD, respectively. The baseline demographic and clinico-biological characteristics of the 144 patients are summarised in Table 1. Across the three groups (healthy volunteers, PLWH ART-treated, and PLWH ART-naïve), we observed a decreasing trend for HDL-C and apoA1 levels, with the lowest levels observed in PLWH ART-naïve (Table 1), while increasing trends were observed for most markers of systemic inflammation measured in negative controls vs. the PLWH group, with a few exceptions. MIP-1 beta, MCP-1, and SAA levels were similar between the three groups. IL-8 and hsCRP were significantly increased in PLWH ART-treated compared to healthy volunteers, while we observed no differences between healthy volunteers and PLWH ART-naïve for these parameters (Table 1). Regarding the tryptophan pathway metabolites, similar trends were observed for most tryptophan metabolites across these three groups, with the kynurenine/tryptophan ratio being the highest in PLWH ART-naïve. Conversely, a decreasing trend was observed for xanthurenic acid across these groups (Table 1).

Despite the trends observed in the biochemical variables and biomarkers, which were overall suggestive of a pro-atherogenic profile, there were no differences in FRS, cIMT, and FMD between the three groups.

Spearman correlations indicated positive and significant correlations between cIMT and FRS across all three subgroups (r ranging between 0.41 and 0.51), but no correlation was observed between FMD and FRS (Figure 2).

Except for IL-6, which modestly correlated with the FRS across the three groups, no significant associations between the other cytokines and the FRS were observed (Figure 2A). Similarly, no associations were found between FRS and tryptophan pathway metabolites (Figure 2A) or between cIMT and cytokines or tryptophan pathway metabolites (Figure 2B).

As shown in Table 1, participants with HIV displayed higher median anti-apoA1 IgG levels and seropositivity rates than healthy volunteers, while PLWH ART-naïve displayed the highest values and seropositivity rates. The association of anti-apoA1 IgG with the different parameters was analysed in the three groups. Anti-apoA1 IgG was not associated with FRS, cIMT, or FMD, despite a possible trend with FRS and cIMT in the PLWH ART-treated group (p = 0.061 and p = 0.083, respectively) (Table 2).

When dichotomising the three study groups according to anti-apoA1 IgG seropositivity status, we observed that HIV-negative anti-apoA1 IgG seropositive participants had higher circulating median hsCRP, ICAM-1, SAA, kynurenine, indole-3-acetate, and 5-hydxoxyindolacetate levels compared to seronegative participants (Table 2). Among PLWH, anti-apoA1 IgG seropositive individuals displayed higher median values of INFγ, TNFα, MIP1alpha, ICAM-1, VCAM-1, kynurenine, kynurenic acid, 5-hydxoxyindolacetate, and kynurenine/tryptophan ratio than seronegative ones. Anti-apoA1 IgG seropositive individuals displayed lower CD4+ cell counts and higher viral load than seronegative individuals. Despite limited statistical power, we further explored the unique study group consisting of PLWH ART-naïve to better understand the possible contribution of ART to inflammation, the kynurenine pathway, and cardiovascular phenotype. In the PLWH ART-treated group, anti-apoA1 IgG was positively correlated with IL-8, hsCRP, ICAM-1, VCAM-1, and SAA, while in the PLWH ART-naïve group, anti-apoA1 IgG was positively correlated with INFγ and TNFα and inversely correlated with SAA (Figure 2C). Regarding the tryptophan/kynurenine pathway metabolites, the kynurenine/tryptophan ratio and xanthurenic acid were increased and decreased, respectively, in PLWH anti-apoA1 IgG seropositive participants (Table 2). The tryptophan pathway metabolites were associated with HIV parameters. While CD4+ cell counts were positively correlated with xanthurenic acid and negatively correlated with kynurenine, 5-hydroxyindoleacetate, 3-hydroxykynurenine, and kynurenine/tryptophan ratio, the viral load was positively correlated with kynurenine, 5-hydroxyindoleacetate, and kynurenic acid and negatively correlated with xanthurenic acid (Figure 3). Of note, ART affected the levels of kynurenine, kynurenine/tryptophan ratio, 5-hydroxindoleacetate, kynurenic acid, and xanthurenic acid. Interestingly, xanthurenic acid was associated with anti-apoA1 IgG seropositivity in PLWH ART-naïve and HIV parameters (Table 2 and Figure 3). In addition, in the HIV ART-treated group, anti-apoA1 IgG seropositive patients displayed a significantly lower body mass index (Table 2).

To identify a possible causal link between higher levels of circulating adhesion molecules (ICAM-1 and VCAM-1) and some metabolites of the tryptophan pathway, we evaluated in vitro the impact of anti-apoA1 IgG stimulation on VCAM-1 and ICAM-1 production on HAECs, as well as the levels of tryptophan metabolites in the cell supernatant. HAECs were incubated for 24 h with polyclonal anti-apoA1 IgG or the isotype control antibody (40 µg/mL), followed by the measurement of the ICAM-1 and VCAM-1 intra-cellular mRNA expression levels as well as their presence in the cell supernatant. The results indicated that anti-apoA1 IgG treatment of HAECs induced the mRNA expression as well as the release into the medium of ICAM-1 and VCAM-1 in vitro (Figures 4A–D). In the same supernatant, the levels of kynurenine, tryptophan, and kynurenic acid were measured, and we did not observe a significant change in tryptophan, kynurenine, kynurenine/tryptophan ratio, or kynurenic acid upon anti-apoA1 IgG stimulation (Figures 4E–H).