Study participants were recruited at the WIR – Walk in Ruhr, Department of Dermatology, Ruhr-University Bochum. We included age-matched HIV-negative participants (controls; n=20) and PLWH (n=71). The written informed consent was obtained from all study participants. The clinical information of the study participants is presented in Supplementary Table 1 . Samples from PLWH were collected immediately before the first vaccination (T0), at the day of second vaccination (T1) within 4-6 weeks (T2) following the second vaccination, at the day of third vaccination (T3) and within 4-6 weeks (T4) following the third vaccination against SARS-CoV-2. Similarly, samples from HIV-negative controls were collected at T2, T3 and T4. No control samples were available for the T0 and T1 time points.

Blood collection for peripheral blood mononuclear cells (PBMC) isolation was conducted using KABEVETTE^® G EDTA tubes. 2.5 ml of blood was centrifuged at 1500 g for 10 min., the plasma was obtained and stored at -80°C until further use. The remaining blood was diluted 1:1 with PBS and slowly placed on top of Pancoll human (PAN-Biotech). Density gradient centrifugation was performed at 800 g for 30 min. without break. The interface containing the PBMCs was collected from the gradient and washed twice with PBS at 500 g for 10 minutes. The cell pellet was resuspended in PBS and the cell number was determined in a Sysmex KX-21N (Sysmex Europe GmbH). 1.8x10⁷ cells were cryopreserved in FCS (PAN-Biotech) containing 10% DMSO. The cryotubes were cooled down overnight in a Mr. Frosty (Sigma) at -80°C and stored in liquid nitrogen until further use.

Plasma samples were analyzed for the spike (receptor-binding domain [RBD]; sequence derived from the original wild type SARS-CoV-2 strain) specific immunoglobulin G antibodies by enzyme-linked immunosorbent assay (ELISA) as described previously (42). Briefly, samples were diluted from 1:100 to 1:12,500 and results were expressed as the dilution, which still gave the same signal as an internal calibrator of the ELISA, indicating a positive result. The values obtained for samples below the detection limit were interpreted as negative and set to 1. The assay was calibrated according to the World Health Organization international standards and values were expressed as binding antibody units (BAU).

SARS-CoV-2 pseudoviruses were prepared as described previously (43). Briefly, sera were incubated for 30 min at 56°C in order to inactivate complement factors. Single cycle VSV∗ΔG(FLuc) pseudoviruses bearing the SARS-CoV-2 spike (D614G) protein (44), SARS-CoV-2 B.1.617.2 (Delta) (EPI_ISL_1921353) or SARS-CoV-2 B.1.1.529 (Omicron) (EPI_ISL_6640919) spike in the envelope were incubated with quadruplicates of two-fold serial dilutions from 1:20 to 1:2560 of immune sera in 96-well plates prior to infection of Vero E6 cells (1x10⁴ cells/well) in DMEM with 10% FBS (Life Technologies). 18 hours post infection, firefly luciferase (FLuc) reporter activity was determined as previously described (45) using a CentroXS LB960 (Berthold). The reciprocal antibody dilution causing 50% inhibition of the luciferase reporter was calculated as pseudovirus neutralization dose 50% (PVND50). Detection range is defined to be between 1:20 and 1:2560.

PBMCs were thawed in a 37°C water bath and diluted in 10 ml RPMI 1640 with Glutamine (Capricorn) with 5% human AB serum (PAN-Biotech) and 5 U/ml Benzonase (Merck/Sigma) and centrifuged at 400 g for 5 min. The pellet was washed with 10 ml thawing medium, resuspended in 5 ml culture medium (RPMI 1640 with Glutamine (Capricorn) plus 5% human AB serum (PAN-Biotech)) and cells were rested for at least 2 hours. Cell number was determined using 7-aminoactinomycin D in a Cytoflex LX (Beckman Coulter, RRID: SCR_019627). 100 µl per sample were seeded in duplicates in a flat bottom 96 well plate (Sarstedt) at a concentration of 1x10⁷ cells/ml. Cells were stimulated with PepTivator pools (Miltenyi Biotec) for 16 hours according to manufacturer’s instructions. The peptide mixes consisted of a pool of mainly 15-mer sequences with 11 amino acids overlap, covering the complete protein coding sequence (aa 5–1273) of the surface or spike glycoprotein (“S”, 130-127-953, Miltenyi Biotec) of SARS Coronavirus 2 (GenBank MN908947.3, Protein QHD43416.1), as well as a mixture of peptides covering the membrane glycoprotein (“M”, 30-126-702, Miltenyi Biotec) and the nucleocapsid phosphoprotein (“N”, 130-126-698, Miltenyi Biotec) protein of SARS-CoV-2. Additional wells included a positive control (CytoStim™, Miltenyi Biotec) and an unstimulated control, in which sterile water instead of the PepTivator pool was added. Except the positive control, all conditions were done in duplicates.

After peptide stimulation for 16 hours, PBMC were stained with the reagents contained in the SARS-CoV-2 T cell analysis kit (PBMC), human plus anti-CD137-PE-Vio 615 (Miltenyi Biotec; see Supplementary Table 2 ) according to manufacturer’s recommendations and analyzed by flow cytometry.

For analysis of different T cell subsets, 2.5 x 10⁶ PBMCs were stained with viability stain LIVE/DEAD™ Fixable Blue Dead Cell Stain Kit, for UV excitation (L23105, Thermo Fisher) for 30 min. at 4°C. Afterwards, Fc receptors were blocked by incubating the cells with Human TruStain FcX™ (422302, Biolegend, RRID: AB_2818986) for 15 min. at 4°C. Subsequently, surface markers were stained for 15 min. at 4°C. Fixation and permeabilization were performed with Foxp3 staining buffer set (130-093-142, Miltenyi Biotec). Next, intracellular proteins Ki-67 and Foxp3 were stained for 30 min. at 4°C. Antibody dilutions are presented in Supplementary Table 3 . Flow cytometry measurement was performed using a Cytoflex LX (Beckman Coulter, RRID: SCR_019627). Flow cytometry data were analyzed with FlowJo™ (Becton Dickinson & Company, version 10.8.0, RRID: SCR_008520) using either smoothing or large dot plot function for representation purpose. Barnes-Hut algorithm was implemented to generate t-Distributed Stochastic Neighbor Embedding (t-SNE) plots with FlowJo™ for high-dimensional data visualization.

Mann-Whitney or Kruskal–Wallis one-way ANOVA tests were performed to calculate statistical significance using Prism (GraphPad Software v9.2.0, San Diego, CA. RRID: SCR_002798). Antibody titers were compared by using two-tailed Welch´s test. The data were imported in R program (v3.6.3). Correlation analysis was performed using “rcorr” function of Hmisc package (v4.4-0) and visualized by “ggscatter” function of ggplot2-based ggpubr package (v0.2.5). Spearman method was used to compute the correlation coefficient (R) and p-value ≤ 0.05 was considered as significant correlation.

We recruited a cohort of 71 people living with HIV (PLWH) receiving anti-retroviral therapy (ART) (mean age = 46.1 ± 10.9; 62 male and 9 female) and 20 HIV-negative donors (controls; mean age = 39.4 ± 11.9; 8 male and 12 female) to analyze humoral and cellular immune responses after two doses of vaccination. All participants received the BNT162b2 vaccine (Pfizer-BioNTech) apart from two participants of the control group, who got mRNA-1273 (Moderna) on the third vaccination time point. Except three participants, all of the PLWH cohort had a CD4 count higher than 200 cells/μl before and 3 months after vaccination (mean CD4 count: 756.4 cells/mm³; range: 79 – 1562 cells/mm³). None in the PLWH cohort reported a confirmed SARS-CoV-2 infection during the sampling period. One participant of the control group had a confirmed SARS-CoV-2 infection and got a BNT162b2 vaccination six months after the infection as recommended by the German Robert Koch Institute (RKI, Berlin). Since the responses of this donor were in the range of other HIV-negative donors, we did not exclude these values. HIV viral load was >30 copies/ml (range: 58-1573 copies/ml) in 6 participants at time of first vaccination. All but one (191 copies/ml) declined during follow up below detection level. The clinical data are summarized in Supplementary Table 1 . Blood samples were drawn immediately before the first (T0) and second (T1) vaccination, four to six weeks after the second vaccination (T2), immediately before the third (T3) and four to six weeks after the third vaccination (T4) ( Figure 1A ).

During SARS-CoV-2 infection, most of the antibody response is directed against the receptor-binding domain (RBD) of the SARS-CoV-2 spike protein and serum levels of anti-RBD antibodies correlate well with the humoral immune response upon vaccination as well as protection against SARS-CoV-2 (6, 46). Therefore, we quantified anti-RBD antibodies in the plasma of study participants by ELISA. Notably, two donors in the PLWH group had detectable antibody titers already at T0 suggesting that they had a previous asymptomatic SARS-CoV-2 infection ( Figure 1B and Supplementary Table 4 ). After the first vaccination, antibody titers did not rise significantly in PLWH ( Figure 1B ). After two doses of vaccine administration, antibody titers of PLWH were reduced (GMT: 838.42 BAU/ml; 95% confidence interval: 430.79-1246.05) compared to the HIV-negative control group (GMT: 1841.94 BAU/ml; 95% confidence interval: 850.35-2833.53) ( Figure 1B ). Although we did not have access to T1 samples in the control group, we could observe a 5-6 times increase in antibody titers in PLWH between the first and the second vaccination, while other studies reported 10-20 times titer increase in HIV-negative individuals (17, 47). Nevertheless, vaccination induced anti-RBD antibody titers in PLWH and additional doses of vaccination further increased these titers ( Figure 1B ). During the time between the second and third vaccination (time points T2 and T3), antibody titers dropped significantly in PLWH, and were boosted again by the third vaccination (T3 vs. T4) ( Figure 1B ). Of note, the PLWH group (anti-RBD titer average: 2394 BAU/ml) had lower antibody titers against the RBD after the third vaccination compared to the control group (anti-RBD titer average: 3733 BAU/ml), although this was not statistically significant ( Figure 1B ).

Next, we analyzed the capacity of sera to neutralize SARS-CoV-2 infection. To this end, we employed an established pseudovirus assay (43) and determined the neutralization dose against the viral strain first reported in Wuhan, China (referred to hereafter as WT) as well as Delta and Omicron SARS-CoV-2 variants of concern. The amounts of neutralizing antibodies were reduced in PLWH compared to controls for all, WT, Delta and Omicron SARS-CoV-2 strains after the second and the third vaccination ( Figure 2 ). We conclude that although SARS-CoV-2 vaccination induces antibody responses in PLWH the extent of the humoral immune response is reduced compared to HIV-negative controls.

Next, we analyzed cellular immune responses towards SARS-CoV-2 vaccination in PLWH and control groups. Here, multi-parameter flow cytometry was employed and the gating strategy is shown in Supplementary Figure 1 . As expected, we detected no differences in the frequencies of total CD3+ T cells between PLWH and the control group ( Supplementary Figure 2A ), reduced frequencies of CD4+ T cells in PLWH ( Supplementary Figure 2B ), and increased frequencies of CD8+ T cells ( Supplementary Figure 2C ). Accordingly, the ratio of CD4 to CD8 T cells was significantly lower in PLWH compared to the control group ( Supplementary Figure 2D ).

Within the CD8+ compartment, the naïve CD8+ T cells (CD27+ CD45RA+ CCR7+) were reduced in PLWH compared to controls ( Supplementary Figure 3A ). Moreover, we detected increased frequencies of effector (E; CD27- CD45RA+ CCR7-; Supplementary Figure 3B ) and effector memory CD8+ T cells (EM; CD27- CD45RA- CCR7-; Supplementary Figure 3C ) as well as decreased frequencies of transitional memory CD8+ T cells (TM; CD27+ CD45RA- CCR7-; Supplementary Figure 3D ) in PLWH. The observation that CD8+ T cells are less naïve and more in an activated and effector-like state in PLWH than in controls was also supported by unsupervised clustering using t-SNE for two representative donors ( Supplementary Figure 3E ). Thus, the phenotypic profiling of the CD8 T cell compartment suggests a high constitutive activation of these cells. In line with this notion, we detected increased frequencies of activated, i.e. CD137 expressing, CD8+ cells that co-expressed the inflammatory cytokines IFNγ and TNFα even in the absence of any antigenic peptide stimulation making it impossible to measure antigen-specific responses in the CD8 compartment ( Supplementary Figure 4 and Supplementary Figures 5A, B ). Moreover, increased frequencies of CD8+ CD137+ IFNγ+ TNFα+ T cells were detected in PLWH upon stimulation with peptide pools derived from spike (S) protein as well as membrane plus nucleocapsid proteins (M+N) ( Supplementary Figures 5C, D ) suggesting that this response is unlikely to be related to the vaccination. Thus, in contrast to what has been reported for SARS-CoV-2 infection in PLWH (41), we were not able to measure antigen-specific CD8 T cell responses in our cohort. However, our data are in line with the activated phenotype described for CD8+ T cells in other studies (48).

Furthermore, we analyzed cellular responses of CD4+ T helper cells upon vaccination. Surprisingly, we detected decreasing frequencies of circulating follicular T helper (cTFH) cells, which are characterized by the expression of CXCR5 and PD-1, over the course of vaccination in PLWH ( Supplementary Figure 6A ). However, we did not observe any correlation between cTFH cells and antibody responses (neutralizing or anti-RBD; data not shown). Consistent with the situation observed for CD8+ T cells in PLWH, we found reduced frequencies of naïve Tcon cells and increased frequencies of terminally differentiated Tcon cells in PLWH after vaccination with two doses ( Supplementary Figures 6B, C ). Despite a similar tendency in these populations, the differences between the groups were not significant in T3 and T4 time points ( Supplementary Figures 6B, C ). Interestingly, we observed significant differences in effector memory T cells for all time points ( Supplementary Figure 6D ). Again, unsupervised clustering using t-SNE of two representative donors supported our findings that CD4+ T cells of PLWH are rather in an activated and differentiated state and that the naïve phenotype is reduced ( Supplementary Figure 6E ).

Finally, we analyzed antigen-specific responses in CD4+ T helper cells. We detected only minimal cytokine responses in cultures without any peptide stimulation or in cultures stimulated with peptide pools derived from membrane (M) and nucleocapsid (N) proteins ( Figure 3A and Supplementary Figure 7A ). In contrast, polyclonal stimulation (positive control (Pos) with CytoStim™) resulted in strong responses at all time points ( Figure 3A ). Stimulation with spike (S) protein peptide pools resulted in upregulation of the activation markers CD137 and CD154 on CD4+ T cells of PLWH after the first and the second vaccination ( Figure 3B ). After a certain reduction of the two activation markers at time point T3, the third vaccination increased CD137 and CD154 expression again at time point T4 ( Figure 3B ). Of note, we did not observe statistically significant differences in the frequencies of activated CD154+ CD137+ CD4+ T cells upon S peptide re-stimulation between the two groups ( Figure 3B ). Surprisingly, we found a significant anti-correlation between S-peptide antigen-specific CD4 T cells expressing activation markers CD154+ CD137+ and the ratio of CD4 to CD8 in the PLWH cohort (R=-0.39; p-value=0.0033; Figure 3C lower scatter plot), which was insignificant in the HIV-negative control group (R=-0.31; p-value=0.18; Figure 3C upper scatter plot). Thus, low CD4+ relative to CD8+ T cell counts are surprisingly associated with better activation of CD4+ T cells.

We then measured IFNγ and TNFα expression in the CD154 and CD137 co-expressing CD4+ T cells and found a robust increase in single (IFNγ or TNFα) or double (IFNγ and TNFα) cytokine-producing cells in PLWH after the first, second and third vaccination ( Figures 3D-F ). Importantly, the expression of cytokines exhibited no significant differences between PLWH and the control group. Furthermore, comparable findings were obtained when activated CD4+ T cells that only expressed the activation marker CD154 were analyzed for S-peptide stimulation ( Supplementary Figures 7B, D ) Therefore, despite the known reduced frequency of CD4+ T helper cells in PLWH, the antigen-specific responses in the CD4 compartment are comparable between PLWH and control groups.

After performing correlation analysis of the parameters obtained after the second vaccination (T2) of the study, we found a statistically significant correlation between anti-RBD antibody response and neutralization capacity against WT SARS-CoV-2 strain in HIV-negative participants (R=0.82, p=3.8e-06) as well as PLWH cohorts (R=0.83, p<2.2e-16) ( Figure 4A ). However, the correlation analyses for anti-RBD antibody and neutralization against SARS-CoV-2 Delta strain were remarkably decreased in PLWH patients (R=0.69, p=1.7e-10) and controls (R=0.84, p=4.5e-06) ( Figure 4B ). No significant correlation was found for anti-RBD antibody and neutralization against the Omicron variant in the control group ( Figure 4C ). The neutralization capacity against WT, Delta and Omicron strains of SARS-CoV-2 correlated better in PLWH patients compared to HIV-negative individuals ( Figures 4D–F ), which might be due to the higher number of participants in the PLWH cohort.

Regarding antigen-specific T cell responses, the correlation analyses at the T2 time point in the control participants showed a significant positive correlation between S-peptide antigen-specific CD4+ CD154+ CD137+ T cells producing IFNγ and TNFα cytokines and neutralizing capacity against SARS-CoV-2 WT strain (R=0.54; p-value=0.015; Figure 4G left graph), Delta variant (R=0.64, p-value=0.0022; Figure 4H left graph), Omicron variant (R=0.48, p-value=0.03; Figure 4I left graph) and with anti-RBD titer (R=0.66; p-value=0.0016; Figure 4J left graph). Interestingly, the correlation value R of these comparisons in PLWH cohort exhibited a significant weaker association ( Figures 4G–I right graphs) and a non-significant correlation for anti-RBD (R=0.18; p-value=0.18, Figure 4J right graph). Moreover, we found comparable correlations between TNFα- and IFNγ-producing CD4+ CD154+ cells with neutralization against Delta and Omicron variants ( Supplementary Figures 7E, F ).

We then performed correlation analysis of the parameters obtained at time point T4 (after the third vaccination). Similar to time point T2, we found comparable correlations between anti-RBD antibody response and neutralization capacity against WT SARS-CoV-2 strain in HIV-negative participants and in PLWH after the third vaccination ( Figure 4K ). In the same line, PLWH presented weaker correlation for anti-RBD antibody and neutralization against SARS-CoV-2 Delta than the control group ( Figure 4K ). Similar to T2, the control group had a weak correlation for anti-RBD antibody and neutralization against the Omicron variant ( Figure 4K ). Except for the case between the neutralization capacity against the Delta strain, which showed comparable results, the correlations for the neutralization capacity against WT and Omicron were stronger in PLWH compared to control group after the third vaccination ( Figure 4K ). Of note, there was practically no correlation for anti-RBD and S-peptide antigen-specific CD4+ CD154+ CD137+ T cells producing IFNγ and TNFα cytokines in both groups ( Figure 4K ). Although weak in general, the latter parameter showed stronger positive correlation with neutralization capacity against WT and Delta strains in PLWH ( Figure 4K ). On the contrary, the correlation between S-peptide antigen-specific CD4+ CD154+ CD137+ T cells producing IFNγ and TNFα cytokines and neutralization capacity against Omicron strain was stronger in the control group than in PLWH ( Figure 4K ).

In conclusion, while the cellular immune response by CD4+ T cells is surprisingly comparable to the control group, the humoral response mediated by antibodies is reduced suggesting that PLWH might be less protected from COVID-19 by vaccination.