Subjects with chronic HIV-1 infection and undetectable viral loads, more than 2 years of ART, were enrolled at Tianjin Second People’s Hospital from December 2020 to May 2021. The participants were stratified either as INRs (CD4 <350 cells/μL) or as IRs (CD4 >500 cells/μL). Additionally, age-matched healthy controls (HCs) were recruited. Blood samples were collected; then, PBMCs were separated from the whole blood, and were subsequently stored at -80°C. The Tianjin Second People’s Hospital Ethics Committee authorized the study (2020-12). Each participant provided written informed consent prior to enrolment, which is in line with the Helsinki Declaration.

Cryopreserved PBMCs were thawed in RPMI 1640 media (Invitrogen, Carlsbad, CA, USA) with 10% fetal bovine serum (Gibco, Invitrogen, NY, USA). To determine the frequency and phenotypes of MDSCs, PBMCs were stained with the following antibodies: CD45-APC-H7, CD33-PE-Cy7, CD11b-BV605, HLA-DR-BV510, CD14-FITC, CD15-PerCP-Cy5.5 (BD Biosciences). To detect PD-L1 expression on MDSCs, PD-L1-PE (BD Biosciences) was added to the previously listed antibodies. T-cell phenotypes were stained with CD3-PerCP, CD4-PE-Cy7, CD8-APC-Cy7, and PD-1-BV605 (BD Biosciences) for 20 min at room temperature. Samples were then acquired and analyzed on FACS Canto Plus and LSRFortessa with Diva software (BD Biosciences). Fluorescence minus one controls or relative isotype controls were prepared to facilitate gating. Data were analyzed using the Flowjo 10 software (Tree Star Inc., Ashland, OR, USA).

MDSC subpopulation phenotypes were defined as follows: PMN-MDSC: CD45+HLA-DR–CD33+CD11b+CD14–CD15+, M-MDSC: CD45+HLA-DR–CD33+CD11b+CD14+CD15– ( Supplementary Figure 1 ).

PBMCs were isolated from fresh blood from 10 HIV-1-infected individuals by using Ficoll-Paque PLUS (GE Healthcare), and then PMN-MDSCs were sorted using CD15 MicroBeads and magnetic assisted cell sorting MS separation columns (Miltenyi Biotec, Bergisch Gladbach, Germany).

PBMCs, PMN-MDSCs depleted PBMCs (DEPL), and DEPL plus autologous sorted PMN-MDSCs were labeled with a cell tracker dye (1:4 ratio), carboxyfluorescein diacetate succinimidyl ester, (CFSE, Invitrogen, Carlsbad, CA, USA) at a final concentration of 5μM as recommended by the manufacturer. Then, cells in each group (PBMCs, DEPL, DEPL+MDSCs) were stimulated with anti-CD3/CD28 coated beads (1 μg/mL; BD Biosciences) in an incubator at 37°C and 5% CO₂ for 4 days in R10. Finally, the cells were washed, stained with CD4-PE-Cy7 at day 5, and subjected to flow cytometry to assess T-cells proliferation.

For intracellular cytokine detection, these cells were incubated in R10 for 6 hours and stimulated with leukocyte activation cocktail (BD Biosciences), which contained phorbol 12-myristate-13-acetate (PMA), ionomycin, and brefeldin A. Cells were surface stained with CD3-BV650 (BD Biosciences), CD4-AF700 (Biolegend), and CD8-Percp-Cy5.5 (BD Biosciences), then further permeabilized with a IntraSure™ kit (BD Biosciences). Series of intracellular staining were performed with IFN-γ-BV605 (Biolegend), IL-2-APC (Biolegend), and TNF-α-PE-Cy7 (Biolegend). Samples were acquired and analyzed as described above.

For blocking experiments, T-cell suppression assays were performed using sorted MDSCs cocultured with autologous CD4+ T-cells in the presence or absence of inhibitors. Magnetic cell sorting CD4+ T-cell isolation kit (Miltenyi Biotec) was used to isolate autologous T-cells (CD4+ T-cells) from freshly isolated PBMCs. Flow cytometry revealed cell purity of >95% after all separations (data not shown). Prior to coculture, purified CD4+ T cells were labeled with CFSE, then mixed with sorted PMN-MDSCs (4:1 ratio) in the presence of anti-CD3/CD28 beads for 4 days. T-cells alone were used as controls. At the commencement of the experiments, either anti-PD-L1 antibody (10μg/mL), pure anti-human TGF-β neutralizing antibody (20 μg/mL), or anti-PD-L1 isotype control were administered to the co-cultured system. T-cell proliferation and intracellular cytokine detection were analyzed as described above.

TRIzol reagent was used to extract total RNA from PMN-MDSCs and DEPL. Then, RNA was reverse-transcribed to obtain cDNA using Superscript™ III First-Strand Synthesis system (Invitrogen, Carlsbad, CA, USA). cDNA product was set in a 20 μl amplification reaction, which contained 10 μl 2×SuperMix (Platinum SYBR Green qPCR kit; Invitrogen), 4 μl cDNA, 0.4 μl of each primer (10 μM), and 5.2 μl DEPC water. The reaction condition was as follows: 50 °C for 2 min and 95 °C for 5 min, then 50 cycles of 95 °C for 15s and 60 °C for 30s. ARG1, iNOS, IL-10, TGF-β, and IDO mRNA expressions were analyzed with real-time PCR using the following primers:

ARG1 Forward: 5′-CGCCAAGTCCAGAACCATAG-3′

Reverse: 5′-TCCCCATAATCCTTCACATCAC-3′;

iNOS Forward: 5′-AGATAAGTGACATAAGTGACCTG-3′

Reverse: 5′-CATTCTGCTGCTTGCTGAG-3′;

IL-10 Forward: 5′- GCCAAGCCTTGTCTGAGATG-3′

Reverse: 5′-AAGAAATCGATGACAGCGCC-3′;

TGF-β Forward: 5′-GACATCAACGGGTTCACTAC-3′

Reverse: 5′-GTGGAGCTGAAGCAATAGTT-3′;

IDO Forward: 5′-AGTTCTGGGATGCATCACCA-3′

Reverse: 5′-ACTGCAGTCTCCATCACGAA-3′.

The relative level of target mRNA expression was normalized to GAPDH (Forward: 5′-CCAGAACATCATCCCTGCCT-3′; Reverse: 5′-CCTGCTTCACCACCTTCTTG-3′) using the equation 2^−ΔΔCt.

TGF-β levels in plasma samples were assessed using an ELISA kit (Invitrogen, CA, USA). Experiments were performed in accordance with the manufacturer’s instructions. The thresholds for detection were 0.098 ng/mL.

CD4+ T-cell counts and plasma HIV-1 RNA were measured following an established procedure, which has been published previously (Xia et al., 2018).

Graphical presentation and statistical analyses were performed using GraphPad Prism version 8.0 (GraphPad Software, San Diego, CA, USA). Mann–Whitney U test or Student’s t-test (between two groups) and one-way ANOVA (for multiple groups) are used to compare continuous variables. Correlation between variables is estimated with Spearman’s nonparametric test. All tests are two-tailed, and P values < 0.05 is considered statistically significant.

Blood samples from 62 HIV-1-infected males, displaying undetectable HIV-1 viral loads as receiving ART for at least two years (31 INRs and 31 IRs), and 30 healthy controls were examined ( Table 1 ). At the time of inclusion, the median CD4+ T-cell counts for INR and IR were 254 cells/μL and 736 cells/μL, respectively. INR had a lower median nadir CD4 counts than IR (35 vs. 309, P < 0.0001). Age, ART duration, and pre-ART viral loads did not significantly differ among the different HIV-1-infected groups.

To investigate the role of MDSCs in immunological recovery, we compared the proportions of MDSCs in HIV-1-infected individuals’ peripheral blood to that of HCs. As shown in Figure 1A , compared to INRs and IRs, lowest proportions of PMN-MDSCs were noted in HCs (all P < 0.0001). Interestingly, PMN-MDSCs were more abundant in INRs than in IRs (P < 0.0001, Figure 1A ). On the other hand, we noted that the proportions of M-MDSCs were analogous across all groups (all P > 0.05, Figure 1B ).

Then, we assessed the relationship between the proportions of MDSCs and HIV-1 disease progression (via IRs and INRs). In INRs, the proportions of PMN-MDSCs were shown to be inversely correlated to CD4+ T-cell counts (r_s = -0.6750, P < 0.0001, Figure 1C ) and nadir CD4 counts (r_s = -0.7242, P < 0.0001, Figure 1D ), but not to pretreatment HIV-1 viral loads (r_s = 0.0765, P = 0.6825) ( Figure 1E ). Conversely, in IRs, the proportions of PMN-MDSCs were not correlated to CD4+ T-cell counts (r_s = 0.2577, P = 0.1616, Figure 1C ), nadir CD4 counts (r_s = –0.2799, P = 0.1272, Figure 1D ), and pretreatment HIV-1 viral loads (r_s = 0.1663, P = 0.3711, Figure 1E ). However, no correlations were found between M-MDSCs and disease progression markers such as nadir CD4 counts, CD4+ T-cell counts, or HIV-1 viral loads when data from both INRs and IRs were combined (data not shown). These findings indicated that PMN-MDSCs are potentially important in the immunological recovery process.

To assess the suppressive activity of PMN-MDSCs, we examined their capacity to inhibit the proliferation of autologous CD4+ T-cells using CFSE dilution analysis. With the use of CD15 magnetic beads, pure PMN-MDSCs were isolated from freshly collected PBMCs. We examined T-cell proliferation after stimulating PBMCs, DEPL, and DEPL+MDSCs with anti-CD3/CD28 beads (ratio 1:4) for 4 days. The flow-chart for the experiment is shown in Figure 2A .

We found that PMN-MDSCs removal significantly enhanced CD4+ T-cells proliferation in INRs only. When PMN-MDSCs were re-added at a ratio of 1:4, CD4+ T-cells proliferation dropped down to a level comparable to that of the freshly collected PBMCs ( Figure 2C ). Compared to HC and INRs, MDSC-depleted INR PBMCs supplemented with autologous MDSCs have a statistically significant reduction in proliferation, (intermediate level, Figure 2C ). Notably, in IRs, CD4+ T-cells proliferation remained relatively stable in PBMCs, MDSC-depleted PBMCs, and MDSC-depleted PBMCs supplemented with MDSCs.

Then, we assessed the inhibitory effects of MDSCs on cytokines release from autologous T-cells. Flow cytometry was used to detect intracellular IL-2, TNF-α, and IFN-γ after stimulation of CD4+ T-cells with PMA ( Figure 2D ). The frequency of IFN-γ–secreting CD4+ T-cells increased significantly after PMN-MDSCs depletion and declined after PMN-MDSCs addition to DEPL ( Figure 2G ). However, we did not find any increase in CD4+ T-cells generating IL-2 and TNF-α ( Figures 2E, F ). In HCs and IRs, no comparable outcomes were seen. In other words, the results revealed that PMN-MDSCs are not suppressive in HCs and doesn’t seem to have any effect on IRs.

Further studies were conducted on the mechanisms whereby MDSCs inhibit CD4+ T-cell response. Given that it is expressed on myeloid cells, PD-L1 is regarded as a crucial biomarker and a promising target in current immunotherapies (Yi et al., 2022). The interaction between PD-1 on T-cells and the inhibitory ligand PD-L1 expressed on myeloid cells was proven capable of inducing T-cell anergy (Bowers et al., 2014). In order to determine if the PD-L1/PD-1 axis is involved in the MDSCs-mediated inhibition of CD4+ T-cell function, flow cytometry was used to measure PD-L1 expression on PMN-MDSCs and PD-1 on CD4+ T-cells.

Figure 3A demonstrates that the expression of PD-L1 on PMN-MDSCs was considerably higher in INRs and IRs than in HCs (P < 0.0001 for all comparisons). In INRs, the proportions of PD-L1+ PMN-MDSCs were significantly higher than in IRs (P < 0.05). In the same line, we observed that PD-1 expression on CD4+ T-cells was greater in INRs and IRs than in HCs (HCs vs. INRs, P < 0.01; HCs vs. IRs, P < 0.01; Figure 3B ). However, between INRs and IRs, the proportions of CD4+ T-cells expressing PD-1 were statistically similar. Notably, we found a correlation between the expression of PD-L1 on PMN-MDSCs and PD-1 on CD4+ T-cells (INRs: r_s = 0.4331, P = 0.0189 and IRs: r_s = 0.5867, P = 0.0005; Figure 3C ). Therefore, we speculate that PD-L1/PD-1 pathway may be crucial for the restoration of CD4 counts.

In addition to surface PD-L1, freshly sorted PMN-MDSCs and DEPL from 10 INRs and 10 IRs were examined to assess mRNA expressions of immune regulatory molecules relevant to MDSC activities (ARG1, iNOS, IL-10, TGF-β, and IDO). In INRs and IRs, mRNA from ARG1, iNOS, IL-10, and IDO were expressed at relatively analogous levels in PMN-MDSCs and DEPL ( Figure 4A ) Within the INRs group, the results showed that PMN-MDSCs had much higher levels of TGF-β than DEPL (P < 0.001, Figure 4A ). Compared to IRs, TGF-β mRNA levels were significantly increased in PMN-MDSCs from INRs (P < 0.0001). Of note, none of the preceding molecules was upregulated in PMN-MDSCs from IRs. We subsequently examined the levels of TGF-β in plasma from INRs (n = 31) and discovered a clear association between PMN-MDSCs and TGF-β (r_s = 0.3883, P = 0.0309; Figure 4B ), demonstrating that TGF-β production and release are affected by PMN-MDSCs.

Further investigations were undertaken to analyze the effects of the blockade of PD-L1 and TGF-β signaling on MDSCs-induced suppression of CD4+ T-cell responses. In the presence of anti-CD3/anti-CD28 beads, purified CD4+ T-cells from INRs were cocultured with autologous PMN-MDSCs at different ratios (2:1, 4:1, 8:1). Autologous CD4+ T-cell proliferation was similarly inhibited by PMN-MDSCs in a dose-independent manner ( Supplementary Figure 2 ). Inhibitors/antibodies targeting PD-L1, TGF-β, or control antibodies were added. A schematic representation of the experiment procedure is shown in Figure 5A . As shown in Figures 5B and E , we discovered that PD-L1 neutralizing antibodies, TGF- antibodies, and dual PD-L1 and TGF-β blocking greatly increased CD4+ T-cell proliferation and IFN-γ secreting capacity. However, there was no increase in the frequency of CD4+ T-cells producing IL-2 and TNF-α ( Figures 5C, D ). Notably, a simultaneous blockade of both, PD-L1 and TGF-β had a greater impact on reviving CD4+ T-cells proliferation than either anti-PD-L1 or anti-TGF-β therapy alone ( Figure 5B ). Furthermore, inhibiting both PD-L1 and TGF-β were observed to relatively enhance IFN-γ production ( Figure 5E ). Our results suggest that to reverse the inhibitory activities of MDSCs during HIV-1 infection, an optimal strategy should simultaneously target both TGF-β and PD-L1 pathways.