Myeloid Derived Suppressor Cells Expansion Persists After Early ART and May Affect CD4 T Cell Recovery

Myeloid-derived suppressor cells (MDSC) are expanded during HIV-1 infection and correlated with disease progression. MDSC expand in the early phase of primary infection depending on TRAIL level. In this study we evaluated the effect of ART on the frequency of MDSC in patients with primary HIV infection (PHI), and their impact on CD4 T cell reconstitution. MDSC frequency was evaluated by flow-cytometry in 60 PHI patients at 12, 24 and 48 weeks after ART initiation. Cytokine plasma levels were evaluated by Luminex technology at the same time points. The capacity of MDSC to modulate hematopoietic early progenitor cells' expansion was evaluated using the OP9/Dl1 in vitro system. As previously described, polymorphonuclear-MDSC (PMN-MDSC) frequency was higher in PHI compared to healthy donors. Interestingly, 48 weeks of successful ART failed to normalize the PMN-MDSC frequency. Moreover, PMN-MDSC frequency was not correlated with residual viral load, suggesting that the persistence of PMN-MDSC was not due to residual viral replication. Interestingly, patients with low PMN-MDSC frequency (<6%) at T0 had a higher HIV DNA at the same time point than individuals with high PMN-MDSC frequency (>6%). We also found an inverse correlation between PMN-MDSC frequency and CD4-T cell count at 48 weeks post-ART, which was confirmed by multivariate analysis adjusting for age and CD4 T cell number at baseline. These data suggest that the persistence of PMN-MDSC may impact CD4 T cell recovery. Indeed, in vitro PMN-MDSC impaired the expansion of CD34+CD38- hematopoietic early progenitors. Further, a balance between TRAIL and GM-CSF may be necessary to maintain a low MDSC level. In conclusion, early ART initiation was not able to normalize PMN-MDSC frequency that might impact the CD4 T cell recovery. These data open new questions regarding the clinical impact of MDSC persistence in HIV+ patients, in particular on non-AIDS related diseases.


INTRODUCTION
Several conditions associated with various types of chronic inflammation result in aberrant and sustained myelopoiesis, characterized by the accumulation of myeloid cells with regulatory functions and ultimately defined as myeloid-derived suppressor cells (MDSC) (1)(2)(3). The first observation of myeloid cells with suppressive functions was reported 20 years ago in cancer-bearing mice (4). However, the importance of this cell population has only recently been pointed out, due to accumulating evidence on its contribution to the negative regulation of immune responses during cancer and other diseases. It has been shown that MDSC are a heterogeneous cell population at different stages of differentiation, comprising myeloid cell progenitors and mature cells (5). In humans, two major MDSC subsets have been identified, the monocytic-MDSC (Mo-MDSC) and polymorphonuclear-MDSC (PMN-MDSC) subsets, expressing low level of HLA-DR, myeloid markers as CD11b and CD33, and CD14 or CD15 respectively (6). A third immature subset of MDSC (immature-MDSC, i-MDSC) comprises cells with immaturity characteristics and is defined essentially as lineage negative cells (7). Suppressive functions are mediated by several major molecular pathways, including inducible nitric oxide (NO) synthetase (iNOS), arginase-1 (Arg-1), NADPH oxidase (NOX2), and TGFβ (6). MDSC are able to inhibit a broad range of immune cell functions such as T cell proliferation and activation (8,9), cytokine production by macrophages (10), and NK cell function (11). They are also involved in the induction of regulatory T cells (12), thus playing a pivotal role in regulating immune response. It has been reported that MDSC increase in several acute and chronic infections [reviewed in (13)]. We and others showed the expansion of Mo-MDSC and PMN-MDSC during chronic HIV infection (14)(15)(16), which correlated with CD4 T cell number and viral load. In HIV-infected patients, MDSC have a potent inhibitory capacity toward HIV-specific CD8 T cell response, thus participating in the maintenance of HIV persistence. More recently, we observed the spread of suppressive PMN-MDSC in the acute phase of HIV infection (17), in particular in the first Fiebig stages. Furthermore, PMN-MDSC expansion during primary HIV infection may be, at least in part, regulated by tumor necrosis factor-related apoptosisinducing ligand (TRAIL), which is able to limit PMN-MDSC survival in the absence of pro-survival factors as GM-CSF (17,18). However, different issues on MDSC during HIV infection remain still unclear, in particular whether the inhibition of HIV replication by ART directly induces MDSC constriction. A reduction in MDSC frequency has been observed after 3 months of therapy in a small group of chronically infected patients (14), but no data are available for longer treatment. The aim of the present work was to better understand the dynamic of MDSC during ART started in the early phase of HIV infection and the possible impact of their persistence on immune reconstitution.

Study Population
Patients with primary HIV infection (PHI, n = 60) were enrolled at the National Institute for Infectious Diseases (INMI) "Lazzaro Spallanzani" (Rome, Italy). At baseline the median CD4 T cell number was 570.5 (IQR 427. 8-731.5), and median viral load (VL) was 5.3 Log copies/ml (IQR 4.31-6.26). Healthy individuals (HD, n = 23) were included as controls. Baseline patients' characteristics are summarized in Table 1. PHIs were recruited before therapy initiation (T0), 12, 24, and 48 weeks (T12, T24, T48) after ART onset. After 48 weeks of therapy, median CD4 T cell number was 806 (IQR 623-997), and median VL < 40 copies/ml (range: not detected-<40 copies/ml). The study was approved by the Institutional Review Board of the INMI "Lazzaro Spallanzani" (ALPHA and SIREA studies) and signed written informed consent was obtained from all patients in accordance with the Declaration of Helsinki.

PMN-MDSC Purification
PMN-MDSC were purified from PBMC by immunomagnetic sorting by using CD15 microbeads (Miltenyi Biotec) following the manufacturer's procedure. The purity of PMN-MDSC was >90% as verified by flow cytometry (data not shown). Purified cells were rested for 18 h in R10.

Residual Viral Load and HIV DNA Quantification
Residual viral load (RVL) was quantified with the ultrasensitive protocol of the Abbott RealTime HIV-1 assay (LLOD: 5 copies/ml), obtained applying the following modifications to the standard procedure (19): (1) higher sample volume (3. Total HIV-DNA was quantified in PBMCs collected by density gradient centrifugation as previously reported. Automated DNA extraction from PBMC was performed on the QIAsymphony platform using an inhouse modification of the DSP Virus/Pathogen Midi Kit protocol (Qiagen, Milan Italy). A Real-Time PCR targeting HIV-LTR was used to quantify total HIV-1 DNA as previously described (20), whereas another Real-Time PCR targeting the housekeeping gene hTERT was performed to refer HIV-DNA copies to a million PBMCs (21) on LightCycler 2.0 platform (Roche, Monza Italia).

Plasma Cytokine Levels
Plasma samples were obtained after speed centrifugation for 10 min at 2,000 rpm and immediately stored at −80 • C. Plasma samples were assayed by using the multiplex bead-based assays

Statistical Analysis
GraphPad Prism version 4.00 for Windows (GraphPad Software) and STATA 15.1 were used to perform statistical analyses. The non-parametric Kruskal-Wallis/Friedman with Dunn's post test were used to compare continuous variables. Correlations were evaluated with the non-parametric Spearman test. Logistic regression analysis was used to estimate OR and 95% CI of factors associated with high/low CD4 cell count at baseline and at 48 w. A p-value < 0.05 was considered statistically significant.

ART During PHI Failed to Decrease PMN-MDSC
We previously demonstrated that PMN-MDSC are expanded during PHI (17). We wondered whether successful ART was able to decrease PMN-MDSC frequency. In agreement with previous research, before ART the frequency of PMN-MDSC was higher in PHI than HD. In contrast, Mo-MDSC were not detected in both PHI and HD. After ART initiation, the frequency of PMN-MDSC persisted higher than HD during all time points (Figures 1A,B). Because of the high variability of PMN-MDSC frequency in PHI, we grouped patients based on PMN-MDSC frequency at T0: low PMN-MDSC group (<6%) and high PMN-MDSC group (>6%); we decided to use the 6% cut off since it represented the highest PMN-MDSC frequency found in the HD group. We observed that in the high PMN-MDSC group the frequency of MDSC decreased at T12 compared to T0, but at T24 it went back to a level similar as T0 (Figure 1C Pt1, and Figure 1D left panel). Moreover, at all time points except T12, PMN-MDSC were higher than HD ( Figure 1C). Differently, in the low PMN-MDSC group, no modulation of PMN-MDSC frequency was observed, remaining stably higher compared to HD at all time points (Figure 1C Pt2, and Figure 1D right panel).
In order to understand whether PMN-MDSC maintain a suppressive capability after 48 weeks from ART initiation, we tested the percentage of HIV-specific CD8 T cells producing IFN-γ in a patient subgroup. We found a lower percentage of cells from patients with high PMN-MDSC frequency than those with low PMN-MDSC (Figure 2A). Accordingly, an inverse correlation was observed between PMN-MDSC frequency and the percentage of CD8 T cells producing IFN-γ after HIVderived peptide stimulation ( Figure 2B). These data suggest that PMN-MDSC retained their suppressive capacity after 48 weeks of therapy.

Evaluation of the Association Between RVL or HIV Reservoir and PMN-MDSC Frequency
After 48 weeks of treatments, all enrolled patients had a viral load under the limit of detection ( Table 1), indicating that persistent viral replication was not involved in the maintenance of PMN-MDSC. To confirm that viral load was not the direct driving force of PMN-MDSC expansion, we evaluated the RVL after 48 weeks from therapy initiation. We did not find difference in PMN-MDSC frequency between patients with and without detectable RVL ( Figure 3A); moreover, no correlation was observed between RVL and PMN-MDSC frequency ( Figure 3B). We also evaluated RVL in the low and high PMN-MDSC frequency groups, and no difference was observed.
Since PMN-MDSC increased very early after infection and can inhibit different immune cell functions, we wondered whether they could affect the HIV reservoir establishment. To this aim, we evaluated the correlation between MDSC frequency and HIV DNA before and after 48 weeks of ART. We did not find any correlation between the two variables (Figures 3C,D). However, grouping patients in low and high PMN-MDSC, we found a higher HIV DNA level in the low PMN-MDSC group compared to the high group (Figure 3E), suggesting that PMN-MDSC might interfere with HIV reservoir establishment. After 48 weeks from therapy initiation a decrease of HIV DNA was observed in both groups, and no difference was observed between the two groups at T48 (Figure 3E).

Effect of PMN-MDSC on Early Progenitor's Expansion During in vitro T Cell Differentiation
As previously demonstrated (17), in PHI patients before starting ART, no correlation between PMN-MDSC frequency and CD4 T cell number was observed ( Figure 4A); differently, after 48 weeks of therapy, an inverse correlation was found (Figure 4B). After adjustment for age (< or >=40 years) and CD4 T cell count at baseline, PMN-MDSC frequency was still associated with CD4 T count >500 cells/mmc at T48 ( Table 2). These data suggest that PMN-MDSC persistence may reflect a worse immune reconstitution.
We wondered whether PMN-MDSC may affect the commitment of CD34+ hematopoietic progenitor cells toward T cells differentiation. To this aim, we cultured purified CD34 stem cells from HD on OP9-DL1 stromal cells in the presence of IL-7, FLT3, and SCF to support the expansion of lymphoid progenitors. We found a reduction of early progenitor frequency when PMN-MDSC were added at a ratio of 1:3 (Figures 4C,D), suggesting that PMN-MDSC may interfere with T cell commitment of early hematopoietic progenitor cells. Moreover, to exclude that the decrease of CD34+Lin-CD38cells was not due to PMN-MDSC-induced CD34 cell death, we evaluated the cell recovery after culture, and found no difference among CD34 alone and CD34 cultured with PMN-MDSC at 1:1 and 1:3 ratios (mean ± SD: 397500 ± 142916, 523750 ± 330842, 515000 ± 312250, respectively). This finding was confirmed by evaluating the induction of 7AAD on CD34+ cells when PMN-MDSC were added, corroborating that PMN-MDSC did not induce CD34+ cell death ( Figure 4C).

Plasmatic TRAIL and GM-CSF Level After ART
We recently reported that, in treatment-naïve PHI patients, PMN-MDSC frequency inversely correlated with plasmatic TRAIL level, while no correlation was found with the other inflammatory cytokines (17). We also showed that TRAIL was able to induce PMN-MDSC apoptosis, and GM-CSF inhibited the pro-apoptotic function of TRAIL. Herein, we evaluated the TRAIL plasmatic level during ART, and found that 48 weeks after ART initiation TRAIL significantly decreased compared to T0 (Figure 5A). Interestingly, an increase of GM-CSF was observed at T48 ( Figure 5B). However, no correlation was found between PMN-MDSC frequency and TRAIL or GM-CSF levels (data not shown).
To further study the regulatory mechanisms that operate in patients with low (<6%) and high (>6%) PMN-MDSC frequency after 48 weeks of therapy, correlations were performed between TRAIL and GM-CSF at T48 in the two groups. We found a direct correlation between TRAIL and GM-CSF levels in patients with low PMN-MDSC frequency at T48 (Figure 5C). In contrast, such a correlation was not present in the group with high PMN-MDSC frequency ( Figure 5D). These data suggest the loss of co-regulated expression of TRAIL and GM-CSF may induce uncontrolled PMN-MDSC expansion. We did not find any correlations between PMN-MDSC frequency and other proinflammatory cytokines (data not shown).

DISCUSSION
The expansion of different subsets of MDSC has been clearly demonstrated during HIV infection (14)(15)(16). However, contrasting data have been reported on the effect of ART on the modulation of MDSC frequency. We previously reported an increased frequency of PMN-MDSC in chronically HIV-infected patients regardless of ART (16). In line with these results, herein we demonstrated that PMN-MDSC frequency of PHI patients remained stably higher than HD during therapy. In contrast, Vollbrecht and colleagues showed a decrease of PMN-MDSC in 7 patients after 6 weeks of therapy (14). This paper is only partially in contrast with our data, in fact we also observed a transient decrease in PMN-MDSC frequency at 12 weeks of ART in patients with a high frequency of PMN-MDSC (>6%) before ART initiation. These data suggest that during the first weeks of treatment, PMN-MDSC may decrease, but afterwards they increased to a level comparable to T0. Accordingly, it has been recently reported that ART intensification in PHI and CHI has no effects on PMN-MDSC frequency, remaining higher than HD in both three-and five-drug ART regiments (23). Importantly, PMN-MDSC maintain their suppressive activity after 48 weeks of ART, thus contributing to the persistence of immune dysfunction. Altogether, the data demonstrate that the anti-HIV treatment, even if efficient in clearing the virus, is not able to induce a normalization of PMN-MDSC frequency after 48 weeks. We excluded that PMN-MDSC persistence could be associated with the residual viral load, confirming that after ART, HIV particles are not the main driver of PMN-MDSC maintenance.
Immune activation and inflammation are among the factors that may contribute to the HIV reservoir size (24) by favoring homeostatic or antigen-driven CD4 T CM and T TM cell proliferation latently infected with HIV (25). Given the high suppressive activity of MDSC on T cell proliferation and activation, our data suggest that PMN-MDSC might have an impact on viral reservoir size before starting therapy; however, after therapy a similar viral reservoir size was achieved, suggesting that the early ART initiation is able to reduce viral reservoir regardless of PMN-MDSC level. MDSC frequency has been reported to be inversely correlated with CD4 T cell count in chronic HIV+ patients (14,15). In contrast, we did not observe that correlation in treatment-naïve PHI patients (17), but after 48 weeks an inverse correlation was found, opening the question of whether and how PMN-MDSC may affect CD4 T cell recovery after ART. Our data show that in vitro PMN-MDSC may interfere with the expansion of the early hematopoietic progenitors, suggesting that PMN-MDSC have the capacity to modulate CD34+ hematopoietic stem cell commitment. The concept of MDSC functions other than direct immune suppression is not new, and include degradation of extracellular matrix, promotion of tumor cell invasion, angiogenesis, and formation of pre-metastatic niche (26,27). Moreover, Chen and colleagues demonstrated that MDSC from myelodysplastic patients can inhibit CD34+ hematopoietic stem cell differentiation toward myeloid and erythroid lineages (28), confirming that MDSC may impact the stem cell compartment.
Long-term production and accumulation of inflammatory factors lead to local and systemic immunosuppression associated with MDSC accumulation [reviewed in (29)]. In particular, GM-CSF is an important growth factor that drives the accumulation and suppressive function of MDSC in both mice (30) and patients (31) with cancer. We recently reported that TRAIL is able to control PMN-MDSC frequency in PHI, while GM-CSF correlates with PMN-MDSC accumulation during chronic HIV infection (17). In the present report, we found that GM-CSF increased overtime in PHI under ART, and the balance between TRAIL and GM-CSF may regulate circulating PMN-MDSC frequency. TRAIL is a member of the TNF family able to induce apoptosis in target cells, including MDSC (32). We previously demonstrated that GM-CSF is able to abrogate the TRAIL-induced apoptosis of PMN-MDSC (17). Thus, we can speculate that, to better control PMN-MDSC accumulation, the increase of GM-CSF needs to be balanced by TRAIL. The reason why the GM-CSF level is raised during efficient therapy remains to be investigated.
In conclusion, we found that early ART initiation was not able to normalize PMN-MDSC frequency, and this persistence may have an impact on the HIV-specific T cell response, and on the hematopoietic stem cell compartment, too. Our data open new question on the effect of the long-term persistence of PMN-MDSC during HIV infection.

DATA AVAILABILITY
All datasets generated for this study are included in the manuscript and/or the supplementary files.