To investigate the impact of early ART initiation on residual inflammation in PLWH, we studied soluble mediators in longitudinal plasma samples from 39 very early treated HIV-positive (HIV+) participants selected from a Belgian multicentric AHI cohort and 21 HIV-negative participants (HIV-). In the HIV+ group the median age was 38.77, the large majority (92.31%) was male and the main risk group was MSM ( Table 1 ). The median age in the HIV- participants was similar to the HIV+ group (36, p=0.74). However, the HIV- group was more gender-balanced (61.90% vs. 92.31% males, p=0.011) and consisted of a smaller proportion of gay/lesbian people compared to the HIV+ group (4.76% vs. 74.36%, p<0.001). HIV+ participants were stratified based on the Fiebig stage at diagnosis of HIV infection; 8 (20.51%) participants were diagnosed in Fiebig stage II, 7 (17.95%) in Fiebig stage III, 2 (5.13%) in Fiebig stage IV, 20 (51.28%) in Fiebig stage V, and 2 (5.13%) in early chronic infection, Fiebig stage VI.

Acute retroviral syndrome (ARS), defined as the presence of fever or ≥2 ARS symptoms at the time of diagnosis, was present in 32 (82.05%) of HIV+ participants. The most commonly reported signs and symptoms were lymphadenopathy, fever, and rash ( Figure 1 ). On average, participants with ARS presented with 3 [IQR 2-5.25] signs or symptoms. In two participants fever was the only presenting symptom. There was a higher frequency of ARS in participants diagnosed in Fiebig stages II-III (14/15, 93.30%) versus Fiebig stages IV-VI (18/24, 75%), however this difference was not statistically significant (p=0.24). All HIV+ participants were placed on ART within 1 week after their first visit to the department.

The HIV+ group was followed longitudinally and plasma soluble mediators were measured on four timepoints: T0, during acute HIV infection; DVL at decreasing viral load on ART (median 157 copies/mL [IQR 74-322]) about one month (median 35 days [IQR 28-43]) after ART initiation; UD, after suppression of plasma viral load (median 278 days [IQR 176-365] after ART initiation); and UD+1, approximately one year later (median 726 days [IQR 623-817] after ART initiation).

Acute HIV infection was hallmarked by a drop in absolute CD4 count, decreased CD4/CD8 ratio, decreased absolute B cell counts and high plasma viral load ( Table 2 ). Participants who presented in Fiebig stages II-III had lower absolute CD4 count (p=0.0022) and higher plasma HIV viral load (p<0.001), compared to participants who presented in Fiebig stages IV-VI ( Supplementary Figures 1A, B ). On the other hand, participants in the later Fiebig stages had higher absolute CD8 counts (p<0.001) and lower CD4/CD8 ratios (p=0.036) ( Supplementary Figures 1C, D ). In contrast to previous studies (26, 29), HIV viral loads did not significantly differ between participants with and without ARS presentation (p=0.67) ( Supplementary Figure 1E ). However, there was a non-significant trend for lower CD4 T cell counts (p=0.11) and significantly higher levels of aspartate aminotransferase (AST) levels (p=0.028) in participants with ARS, the latter reflecting the systemic nature of ARS ( Supplementary Figures 1F, G ).

Two decades ago, the beneficial impact of initiating ART very early on immune recovery was well-established (30–32). In our early treated population, the median CD4/CD8 ratio normalized to >1, and the median CD4 count increased to 712/µL as soon as 1 month after treatment initiation (DVL). Additionally, normalization of the B cell subset was demonstrated at the DVL timepoint ( Table 2 ). The median time to viral suppression was 97 days [IQR 90-131]. Neither Fiebig stage (Fiebig II-III: median 97 [IQR 92-152] days, Fiebig IV-VI: median 99 [IQR 89.5-130] days, p=0.98), nor clinical presentation (ARS: median 99.5 [IQR 94-154] days, no ARS: median 88 [IQR 71.5-99.5] days, p =0.096) at diagnosis had a significant impact on the time to viral suppression.

Treatment initiation during acute HIV infection curtails reservoir seeding and thereby limits the size of the reservoir (33–36). Total and intact viral reservoir was measured at T0, UD, and UD+1. In our cohort, total HIV DNA was significantly higher in Fiebig II-III compared to Fiebig stages IV-VI at T0 (p=0.0036, median 45231 and 8834 copies/10⁶ CD4 T cells respectively) ( Figure 2A ). However, no difference was found during suppressed viremia (UD: p=0.97, median 299 and 313 copies/10⁶ CD4 T cells and UD+1: p=0.74, median 114 and 133 copies/10⁶ CD4 T cells respectively). Even though total HIV DNA is a robust and subtype-independent measurement of the total cell-associated HIV load (37), it cannot distinguish between defective and intact proviruses. To distinguish the total from the intact reservoir, we performed a digital polymerase chain reaction (PCR)-based assay targeting the packaging signal (psi) and envelope (env) region of the HIV genome, based on previously reported techniques (38, 39). Simultaneous detection of psi and env was only possible for 19/39 (48.72%) of HIV+ participants. For intact HIV DNA no significant differences were found between participants treated in Fiebig II-III and Fiebig IV-VI (T0: p=0.15, median 10549 and 3594 copies/10⁶ CD4 T cells; UD: p=0.43, median 136 and 288 copies/10⁶ CD4 T cells; UD+1: p=0.11, median 62.60 and 34.40 copies/10⁶ CD4 T cells respectively) ( Figure 2B ). As expected, intact HIV DNA was significantly (p<0.001) lower than total HIV DNA at all timepoints ( Table 2 , Figure 2C ). After ART initiation, a significant decline in total (p<0.001) and intact (p=0.032) HIV DNA was observed between T0 and UD. During viral suppression a further decline of the HIV reservoir was observed between UD and UD+1 both in total as in intact HIV DNA (p=0.0012 total HIV DNA, p=0.032 intact HIV DNA).

We observed a large interparticipant variation in terms of reservoir composition within our cohort ( Figure 2D ). The ratio of intact to total HIV DNA was significantly higher at UD than at UD+1 (p=0.0081, 76% [IQR 47-81.8] and 42.4% [IQR 29.3-60.9] respectively). It has been established that during the first year after ART initiation, there is a rapid decay of the viral reservoir, after which the decay rate slows down (40). Paired longitudinal analysis of the HIV reservoir in our cohort revealed a more pronounced fold decrease of the intact reservoir versus the total reservoir in the period between UD and UD+1 (p=0.013), which spanned a median of 379 days [IQR 359-462] ( Figures 2E, F ).

In order to establish in-depth inflammatory profiles during AHI and longitudinally after introduction of ART, we longitudinally measured plasma concentrations of 40 soluble mediators consisting of adaptive immune system cytokines, chemoattractants, growth factors, inflammasome-related mediators, interferons, pyrogens and TNF superfamily cytokines. Representation of all measured plasma samples in a heatmap illustrates the main trends across the four timepoints, but also highlights the large inter-individual variability ( Figure 3 ). During acute HIV infection, a pleiotropic increase of plasma cytokines and chemokines was observed. Compared to the HIV- group, a significantly higher expression of type II interferon (IFN-γ), and type II IFN stimulated chemokines (CXCL10, CXCL11), inflammasome-associated cytokines (IL-18, IL-1RA), type I IFN (IFN-α2a), tumor necrosis factor (TNF) family cytokines (TNF-α, TNF-β, TRAIL), monocyte chemoattractants (CCL1, CCL2, CCL3, CCL4), monocyte secreted chemokines (CXCL8), Th1 polarizing cytokines (IL-12/23p40) and lymphocyte growth factors (IL-2, IL-7) was found on the T0 timepoint ( Figure 3 , Table 3 ). Furthermore, our data confirmed the temporal and transient induction of soluble mediators during AHI as described previously (1, 41), with participants in the earlier phase of AHI (Fiebig stage II-III) showing higher levels of IFN-α2a (p<0.001), IL-15 (p<0.001), IL-6 (p=0.0014), CXCL8 (p=0.0018), IFN-γ (p=0.003), CCL2 (p=0.02), IL-27 (p=0.031)), M-CSF (p=0.045), and CCL4 (p=0.047), as compared to those in a later phase of AHI (Fiebig stage IV-VI). ( Figure 3 , Supplementary Figure 2A ). Plasma concentrations of IL-22 showed a similar trend, but did not reach significance (p=0.09). Some markers, such as IL-27, IL-6, and IL-22 normalized spontaneously by the later Fiebig stages. Notably, during the later phase of AHI, the concentrations of IL-15 appeared lower compared to the levels observed in the HIV- group, although this trend was not statistically significant (p=0.08).

By comparing the inflammatory profile of HIV+ participants with and without ARS, we identified the soluble mediators contributing most to systemic clinical manifestation. ARS was associated with higher levels of the endogenous pyrogen IL-6 (p=0.021), the inflammasome regulator IL-1RA (p=0.038), and the IFN-γ induced chemokine CXCL10 (p=0.032), compared to participants without ARS. IFN-γ was higher in participants with ARS, compared to the HIV- group (p<0.001), while no statistically significant difference was observed between the control group and the participants without ARS ( Figure 3 , Supplementary Figure 2B ). In contrast, participants without ARS had higher plasma levels of the neutrophil chemoattractant CXCL1, compared to those with ARS (p=0.033), as well as compared to the HIV- group (p=0.0054). The hematopoietic growth factor IL-7 showed a similar trend.

We then assessed the longitudinal course of the inflammatory profile after ART initiation ( Figure 3 , Table 3 ). One month after ART initiation (DVL), most soluble mediators showed a decreasing trend. Despite the detectable plasma viral load, some cytokines and chemokines had completely normalized by this timepoint. Examples are the rapidly induced IL-27, IL-6, IL-22, M-CSF, and IFN-α2a which already showed a normalization or declining pattern between early and late Fiebig stages. In contrast, VEGF showed a delayed peak, with increasing levels one month after AHI diagnosis, and a spontaneous normalization afterwards ( Supplementary Figure 3A ). Interestingly, some soluble mediators did not normalize with suppressive ART, even at the UD+1 timepoint, on average 2 years after treatment initiation: type II IFN induced chemokines CXCL11 (p=0.002) and CXCL10 (p=0.028); inflammasome-related cytokines IL-18 (p<0.001) and IL-1β (p=0.015); chemoattractants CCL2 (p=0.0081), CCL4 (p=0.015), CXCL8 (p=0.035); and IL-21 (p=0.028) ( Supplementary Figure 3B ). On the other hand, CX3CL1, a key regulator of cytotoxic T-cells, was significantly lower at the UD+1 timepoint, compared to the HIV- group (p<0.001), despite an initial upregulation during acute HIV infection. Plasma levels of IFN-γ, the inducer of CXCL11 and CXCL10 showed a gradual decline over time and were still significantly higher than the HIV- group at the UD timepoint (p=0.044), but no longer at the UD+1 timepoint (p=0.13) ( Supplementary Figure 3C ).

To understand the contributing factors to the persistent inflammatory profile during viral suppression, we examined differences in pretreatment characteristics. Interestingly, clinical presentation with ARS during acute HIV infection, was associated with higher plasma concentrations of CXCL11 (p=0.038) at the UD+1 timepoint. CXCL10, another type II IFN induced chemokine, showed a similar trend, but did not reach statistical significance (p=0.14) ( Figure 4A ). The comparison to the HIV- group suggests that the persistently elevated levels of CXCL11 and CXCL10 at UD+1 were mainly driven by the participants presenting with ARS during AHI. Plasma concentrations of IL-4 at UD+1 were marginally lower in participants without ARS during AHI, compared to those with ARS (p=0.042) and the HIV- group (p=0.018), but were generally very low. Furthermore, we looked whether earlier ART initiation, in Fiebig stages II-III versus Fiebig stages IV-VI had an impact on the levels of soluble mediators under stable, suppressive ART. ART initiation during the earlier Fiebig stages was associated with higher levels of IFN-α2a (p=0.013) at UD+1, compared to ART initiation in later Fiebig stages ( Figure 4B ). Levels of CXCL10 at UD+1 showed a similar, but non-significant trend (p=0.088). Only the participants diagnosed in Fiebig stages II-III showed a significant difference in plasma CXCL10 with the HIV- group (p=0.0012), not the participants diagnosed in the later Fiebig stages (p=0.19). Collectively, these findings suggest that events early in the course of HIV infection have a lasting impact on low-grade inflammation.

When specifically examining the correlations between the identified group of persistently elevated cytokines and chemokines at UD+1 and the pretreatment immunological parameters at T0, we found that the CD4 nadir during acute HIV infection correlated inversely with plasma concentrations of CXCL11 and CXCL10 at UD+1 ( Figure 4C ). Generally, however, the most significant positive correlations between the persistently elevated levels of CXCL10, IL-18, IL-1β, CCL2 and CCL4 (UD+1) were found with their pretreatment levels during AHI (T0) ( Figure 4D ). During acute HIV-infection, these mediators clustered together and showed a positive correlation to plasma viral load and levels of aspartate aminotransferase, suggesting that they might be driven by viral replication. ( Figure 4E ).

In contrast, CXCL1 at T0 showed an inverse correlation with the pro-inflammatory signature at UD+1, with negative correlations with CXCL11, CXCL10, CXCL8 and CCL2 ( Figure 4D ). Remarkably, among the cellular laboratory parameters, no clear immunological signature during viral suppression was found to correlate with the identified group of persistently elevated cytokines. Only absolute NK cell counts showed a positive correlation with CCL2 and CD4/CD8 ratio correlated with CXCL10 ( Supplementary Figure 4 ).

CD4/CD8 ratio is often used as a proxy for immune activation and residual inflammation in PLWH and correlates with clinical outcomes (42, 43). Indeed, we found inverse correlations between the CD4/CD8 ratio during viral suppression and plasma levels of CXCL10 both at T0 (R=-0.47, p=0.017) and UD+1 (R=-0.39, p=0.043) ( Figure 5A ). Time to viral suppression is another parameter which is often taken into account in assessing immunovirological outcomes. Because of the potent integrase strand transfer inhibitor (INSTI) regimens used in our cohort, rapid viral suppression was obtained in most HIV+ participants after approximately 100 days. However, in some participants time to viral suppression was notably longer, raising the question whether the inflammatory profile during AHI or during viral suppression could be associated with this phenomenon. Interesting, albeit modest correlations were found between time to viral suppression and T0 plasma levels of CXCL1 (R=-0.33, p= 0.039) and UD+1 levels of CX3CL1 (R=-0.38, P=0.04) ( Figure 5B ).

Next, we assessed the relationship between the plasma inflammatory profile and HIV reservoir. Immunological signaling molecules might not only be produced in response to cells harboring HIV, their presence might also impact homeostatic maintenance or clearance of the viral reservoir. We first examined correlations between the HIV reservoir and contemporaneous soluble mediator levels at UD+1. In general, more correlations were found with the intact reservoir, than the defective or total reservoir ( Supplementary Figure 5 ). CX3CL1 showed an inverse correlation with total HIV DNA (R=-0.54, p-p=0.0026) and intact HIV DNA (R=-0.58, p=0.033) ( Figures 6A, B ). Interestingly, intact HIV DNA exhibited a positive correlation with levels of CXCL11 (R=0.65, p=0.014), one of the chemokines which remained persistently elevated at UD+1 ( Figure 6B ). In addition, levels of IL-27, a pleiotropic cytokine with both pro-inflammatory and immunosuppressive capabilities, were also correlated with higher intact HIV DNA (R=0.59, p=0.03) ( Figure 6B ). To assess whether the inflammatory profile during acute HIV infection had a lasting impact on the viral reservoir during viral suppression, we correlated both total and intact HIV DNA with pretreatment levels of soluble mediators. Total HIV DNA was inversely correlated with levels of GM-CSF at T0 (R=-0.53, p=0.0036) ( Figure 6C ). Intact HIV DNA was inversely correlated with IL-2 (R=-0.57, p=0.035) and TNF-β (R=-0.53, p=0.0036) and positively associated with IL-27 (R=0.6, p=0.043) ( Figure 6D ). The fold decrease in total HIV DNA was positively associated with levels of CX3CL1 (R=0.43, p=0.02) during viral suppression (UD+1) ( Figure 6E ). Plasma levels of the pro-inflammatory chemokine CXCL8 at UD+1 showed a strong inverse correlation with both the fold decrease in total (R=-0.61, p<0.001) and intact reservoir (R=-0.79, p=0.0012) ( Figures 6E, F ).

All HIV-positive samples and data were obtained from the ongoing ACS-study, “Accurate staging of immuno-virological dynamics during acute HIV-1 infection”, a multicentric prospective AHI cohort in Belgium coordinated at the HIV Cure Research Center at Ghent University (ClinicalTrials.gov identification NCT03449706). Inclusion criteria were adult age (18-65 years) and documented acute HIV infection defined as a new diagnosis of HIV with clinical symptoms of acute seroconversion, a negative screening test within the past 6 months or the presence of a risk contact within the last 3 months, and negative or incomplete line immunoassay. Exclusion criteria were recent use of immunomodulatory drugs and active hepatitis C infection. Participants were included at Ghent University Hospital, AZ Sint-Jan Bruges, CHU Saint-Pierre, and UZ Brussels. Participants were included during AHI and started on ART within one week. Blood draws were performed before ART initiation (T0) and longitudinally during follow-up. Leukapheresis was performed after suppression of plasma viral load (UD) and one year later (UD+1).

HIV-negative control participants were included at Ghent University Hospital in the VIM-study, “Virological, immunological and microbiome monitoring platform in HIV-infected populations”. HIV-testing was negative in all participants.

Clinical and laboratory parameters were assessed at every study visit. Plasma viral loads were measured using the Cobas® HIV-1 test (Roche Diagnostics, Basel, Switzerland). Classification of AHI was performed as described by Fiebig et al (73), using results of plasma HIV RNA determination, fourth-generation immunoassay (VIDAS DUO), and INNO-LIA HIV I/II Score (Fujirebio).

Acute retroviral syndrome was defined based on the presence of fever or ≥2 of the following signs and symptoms around the time of the positive HIV diagnostic test: adenopathy, night sweating headache, fatigue, myalgia, arthralgia, pharyngitis, oral ulcer, genital ulcer, weight loss, anorexia, nausea/vomiting, diarrhea, odynophagia, skin rash, oral candidiasis, vaginal candidiasis, or other neurologic symptoms.

Study data were collected and managed using REDCap electronic data capture tools hosted at Gent University Hospital (74, 75).

Whole blood was collected from study participants by blood draw or by leukapheresis. Plasma was collected after peripheral blood mononuclear cell (PBMC) separation, with Lymphoprep™ (# 07851, Stemcell technologies) in Leucosep™ tubes (# 227290, Greiner Bio-One), aliquoted and stored at -80°C. PBMCs were collected and cryopreserved in FCS/DMSO 10% in liquid nitrogen.

Plasma samples were thawed on ice and quantified using four Mesoscale Discovery U-Plex electrochemiluminescent immunoassay 10-plex panels (#K15067L-1, MSD) according to the manufacturer’s instructions in two batches. The following 40 analytes were included: CX3CL1, GM-CSF, CXCL1, CCL1, IFN-α2a, IFN-β, IFN-γ, IL-12p40/IL-23p40, IL-12p70, IL-13, IL-15, IL-16, IL-17A, IL-18, IL-1RA, IL-1α, IL-1β, IL-2, IL-21, IL-22, IL-27, IL-29, IL-3, IL-4, IL-5, IL-6, IL-7, CXCL8, IL-9, CXCL10, CXCL11, CCL-2, M-CSF, CCL3, CCL4, CXCL12, TNF-α, TNF-β, TRAIL, VEGF. Analyte concentrations were inferred from standard curves. Values below the fit curve were replaced by the square root of the minimum value per analyte per run. A linear mixed model with participant ID as random factor and batch and timepoint as fixed factors was used to correct for batch effects between the two batches.

CD4+ T cells were isolated from fresh or cryopreserved PBMCs using EasySep™ Human CD4+ T Cell Isolation Kit (#17952, StemCell technologies), after which genomic DNA was extracted using QIAamp DNA Mini Kits (#51304, Qiagen). Total and intact HIV DNA were measured simultaneously with the Rainbow proviral DNA digital PCR assay (38), using the QIAcuity digital PCR platform. This multiplex assay includes RU5 primers and probes for total HIV DNA detection (76) and the psi and env primers and probes of the IPDA assay (39). RPP30 was used as a reference gene (38). Intactness was quantified based on the detection of env and psi and corrected for shearing using a DNA Shearing Index (DSI) correction, as previously described (39). Defective HIV DNA was calculated by subtracting intact from total HIV DNA.

Categorical variables are presented as frequency counts and percentages, continuous variables are presented as medians and interquartile ranges. Categorical data between groups were compared using Chi square tests. Unpaired continuous data between groups were compared using Mann Whitney U tests (2 groups). Paired continuous data were compared using Wilcoxon signed-rank tests (2 groups) or Friedman tests with post-hoc Dunn tests (>2 groups). Correlations were calculated by two-sided Spearman’s rank tests for pairwise complete observations. P values <0.05 were considered statistically significant. Due to sample size limitations and the exploratory nature of this study, no correction for multiple testing was applied. All statistics were performed using R-4.2.1 software.

The ACS-study (early treated HIV-positive participants) and VIM-study (HIV-negative participants) were approved by the Ethics Committee of Ghent University Hospital, Belgium (BC-00812 and BC-01373, respectively). All participants provided written informed consent.