Rhesus Macaque Myeloid-Derived Suppressor Cells Demonstrate T Cell Inhibitory Functions and Are Transiently Increased after Vaccination

A number of conditions associated with inflammation, autoimmune disease, and cancer lead to expansion of myeloid-derived suppressor cells (MDSCs) that play a key role in regulating T cell responses. MDSCs represent a heterogeneous population of innate immune cells with three main features: myeloid origin, immature state, and suppressive effect of T cell responses in particular (1). MDSCs interfere with the functions of T cells and NK cells either by direct receptor-mediated cell–cell contact, via the release of suppressive mediators, or disrupting the contact between other innate cell subsets, for example, dendritic cells with T cells or NK cells (2).

MDSCs have been intensively investigated in humans and mice in recent years (3). In contrast, MDSCs in nonhuman primates (NHPs) have been found, but knowledge of the cells is very limited (4, 5). NHPs serve as a critical animal model for a wide range of late-stage biomedical investigations because of their genetic and physiological similarities to humans. Characterizing MDSCs in this model is therefore warranted. The first aim of this study was to identify phenotypic markers suitable for MDSCs in rhesus macaques and perform a side-by-side comparison with human MDSCs. This was accompanied by functional confirmation of rhesus MDSCs for their suppressive effect on T cell response.

In human blood, monocytic (M)-MDSCs can be identified as HLA-DR^−/lowLin⁻CD33⁺CD11b⁺CD14⁺CD15⁻ and polymorphonuclear (PMN)-MDSCs as HLA⁻DR⁻Lin⁻CD33⁺CD15⁺CD14⁻ (3). Notably, human PMN-MDSCs are thought to be a unique subset of neutrophils within the heterogeneous low-density neutrophils (LDNs) that cosegregate with PBMCs after Ficoll centrifugation. This is in contrast with normal density neutrophils (NDNs), which sediment together with erythrocytes and consist of a homogenous population with immune-stimulatory function (6). It is currently difficult to discriminate PMN-MDSCs from other LDNs because of limited availability of specific markers. Human PMN-MDSCs are therefore usually defined as the LDN population as a whole. Immature HLA-DR⁻Lin⁻CD33⁺CD15⁻CD14⁻ progenitors named early-stage MDSCs were recently described in humans as a subset that may differentiate into M-MDSCs or PMN-MDSCs (3, 7).

MDSCs were first identified in cancer to support tumor progression via the dysregulation of immune responses in the tumor microenvironment (8). Recent studies have shown that MDSCs also appear in several other conditions including infection, transplantation, autoimmunity, and hypertension (1, 9). Therefore, the role of MDSCs in immune regulation appears to be widespread, although several details of their functions remain elusive. Their suppressive function that contributes to immune deficiency has been proposed to have a protective effect in other contexts, evidenced by preventing tissue damage, alleviating inflammation, or even facilitating clearance of pathogens (10). To this end, it is currently unclear whether the numbers of MDSCs increase and influence T cell responses in temporary inflammation, such as after vaccine administration. Because rhesus macaques are important for testing new vaccines prior to clinical trials (11), we monitored the fluctuation of circulating and tissue MDSCs after successful vaccination.

All animal experiments were performed in accordance with the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care and the Swedish Animal Welfare Agency. The study was approved by the Local Ethical Committee. Rhesus macaques were housed at Astrid Fagræus Laboratory in Karolinska Institutet, Sweden. Animals were i.m. immunized with an influenza vaccine consisting of modified mRNA encoding the hemagglutinin of H10N8 influenza A virus (A/Jiangxi-Donghu/346/2013) formulated in a lipid nanoparticle as previously described (12). The vaccine did not contain additional adjuvant. The animals received prime and boost immunization at weeks 0 and 4. Peripheral venous blood was collected at different time points before and after vaccination and processed within 1 h. For the monitoring of MDSCs in peripheral sites at early time point, rhesus macaques received vaccine injection at several distal sites (left calf and left quadriceps) and PBS injection on contralateral sites (left deltoid) as control. Biopsies of muscle tissues from injection sites and injection-site draining lymph nodes (dLNs; axillary lymph node [LN], popliteal LN, and inguinal LN) were collected after 24 h and processed immediately to obtain single-cell suspensions (13) or stored in RNAlater solution (Ambion) for further microarray analysis.

A total of 1.5 million fresh rhesus PBMCs were seeded per well into 96-well U-bottom plates in the presence of overlapping peptides of H10 protein (2 μg) and brefeldin A (5 μg) overnight. Intracellular staining of IFN-γ, IL-2, and TNF was performed using Fixation/Permeabilization Solution kit (BD Biosciences) as previously described (14). Frequency of cytokine-producing CD4⁺ T cells was evaluated by FACS analysis. Vaccine-specific B cell responses were evaluated by measuring hemagglutination inhibition (HAI) titers (12). HAI assay was conducted with 0.5% turkey erythrocytes (Rockland Abs and Assays) diluted in PBS (pH 7.4). Serial diluted (1:2) serum samples starting from 1:10 dilution were incubated with recombinant HA of H10N8 influenza A virus (4 U) in 96-well plates for 30 min at room temperature. The reciprocal of the last serum dilution that resulted in nonagglutinated RBCs represented the HAI titer.

Peripheral venous blood was collected in EDTA vacuum tubes and processed within 1 h. PBMCs were isolated using Ficoll (GE Healthcare) as earlier described (14). After centrifugation, cells from low-density fraction (mainly PBMCs and LDNs) were collected, washed, and suspended in medium. CD33⁺ LDNs were isolated from low-density cells using anti-CD33 microbeads (Miltenyi Biotec). CD33⁻ LDNs were purified by depleting CD33⁺ cells from low-density cells using anti-CD33 microbeads followed by positive selection using anti-CD66abce microbeads. M-MDSCs (HLA-DR⁻CD14⁺) were purified by depleting HLA-DR⁺ cells first using anti–HLA-DR microbeads and then positively selected using anti-CD14 microbeads. All purifications were performed by MACS using LD or LS columns and MACS separator. The purity of cells is >90% with viability >95%. NDNs were isolated using dextran sedimentation assay as described with modifications (15). In brief, the NDN-erythrocyte pellet was suspended in PBS and mixed with equal volume of 3% dextran, and the tube was standing in an upright position to allow sedimentation for 30 min at room temperature. The NDN-rich upper layer was collected, and contaminating RBCs were eliminated by hypotonic lysis (0.2% NaCl for 30 s followed by addition of an equal volume of 1.6% NaCl). Cells were cultured in complete RPMI 1640 medium (Sigma-Aldrich) including 10% FBS (Life Technologies), 100 U/ml penicillin, 100 μg/ml streptomycin, and 292 μg/ml l-glutamine (Hyclone) at 37°C in a 5% CO₂ atmosphere.

Fresh PBMCs from healthy NHPs were cultured in T-75 flasks at a concentration of 1 × 10⁶ cells/ml in complete RPMI 1640 medium supplemented with different cytokines for 7 d. PBMCs cultured in medium alone served as control. Medium and cytokines were changed every 2 or 3 d. The following recombinant human cytokines were used in this study: GM-CSF (50 ng/ml), IL-6 (10 ng/ml), IL-1β (10 ng/ml; R&D Systems), PGE₂ (1 μg/ml; Sigma-Aldrich), and IL-11 (10 ng/ml; PeproTech).

PBMCs or PBMCs depleted of CD33⁺ LDNs were labeled with CFSE (Molecular Probes) as previously described (14). Purified autologous CD33⁻ LDNs, CD33⁺ LDNs, NDNs, or M-MDSCs were cocultured with CFSE-labeled PBMCs at a ratio of 1:5 in the presence of staphylococcal enterotoxin B (SEB). In some experiments, CD33⁺ LDNs were cocultured with CFSE-labeled PBMCs at different ratio (1:20, 1:10, 1:5, and 1:2) in the presence of SEB. Unless otherwise stated, 0.05 μg of SEB was used. After 3 d of culture, the cells were washed and stained with anti-CD3, anti-CD4, and anti-CD8 Abs. T cell proliferation was measured by flow cytometry calculated by the percentage of CFSE^low T cells.

Cells were collected and stained with LIVE/DEAD Fixable Blue Dead Cells Stain (Thermo Fisher Scientific) before incubation with human FcR blocking reagent (Miltenyi Biotec) and a mixture of Abs for 20 min at room temperature. For all of the experiments, Abs were titrated to allow for the obtainment of brightest signal, and fluorescence minus one control was set to avoid nonspecific binding or spectral overlap. After incubation, samples were washed with PBS and subsequently acquired on LSRFortessa cell analyzer (BD Biosciences). Data were analyzed using FlowJo V.10.1 (Tree Star). A list of Abs is available in Supplemental Table II.

Purified CD33⁻ LDNs, CD33⁺ LDNs, NDNs, and CD14⁺ cells adhered to slides by cytospin followed by staining with May/Grünwald-Giemsa solution (Sigma-Aldrich). In brief, cells were fixed with methanol at room temperature for 5 min and then placed in May/Grünwald solution for 12 min. Slides were gently rinsed with distilled water before staining in Giemsa solution for 15 min. Pictures were taken using Nikon Eclipse E800 at 100× magnification.

The concentration of IFN-γ, IL-17A, IL-6, and IL-1β in supernatants from indicated cell cultures were measured using a customized MILLIPLEX MAP Multiplex Kit (Merck Millipore). Experiments were performed according to the manufacturer’s protocol and analyzed using MAGPIX system (Merck Millipore).

Biopsies of muscle and LNs were collected, and total RNA was extracted using TRIzol (Invitrogen) plus TissueLyser (Qiagen) according to the manufacturer’s instruction. Two micrograms of total RNA was used for probe synthesis of cyanine-3–labeled cRNA using Quick Amp Labeling Kit (Agilent) before purification with RNeasy column (Qiagen). Cyanine-3 cRNA was then hybridized to Agilent Rhesus Macaque Gene Expression Microarray (G2519F; Design ID: V2-026806) and processed according to manuals. The background corrected data from output were further normalized using the Bioconductor package to achieve consistency between samples. A set of genes relevant to MDSCs was selected, and gene expression in vaccine/PBS injection muscle sites or injection-site dLN was measured and compared. Agilent gene data are deposited in Gene Expression Omnibus (accession number GSE98211, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE98211).

Statistical analysis was conducted by Prism Version 6.0 software. All values are presented as mean ± SEM for at least four independent experiments. Differences between groups were analyzed by unpaired or paired Student t test. A p value <0.05 is considered statistically significant.

Because the identification and function of MDSCs in NHPs is to a large extent unexplored, we first tested multiple Abs for markers used for human cells to identify MDSCs in rhesus macaques. We used the NHP Reagent Resource Web site (http://www.nhpreagents.org) and confirmed the cross-reactivity of the human Ab clones to rhesus macaques (Supplemental Table I). Because CD15 is not ideal for identification of granulocytes in rhesus macaques, we used CD66abce, which identifies the same cell population as we reported previously (14). CD8 was used instead of CD56 to discriminate NK cells because CD56 is expressed on subsets of rhesus monocytes and most rhesus NK cells express CD8 (16). We found distinct cell populations in the low-density cell fraction (regularly also referred to as PBMC fraction after Ficoll centrifugation) (Fig. 1A). This fraction contained CD14⁺ classical monocytes (HLA-DR⁺Lin⁻CD14⁺), M-MDSCs (HLA-DR⁻Lin⁻CD14⁺), as well as LDNs (HLA-DR⁻CD66abce⁺). The higher granularity of LDNs versus monocytes and M-MDSCs was confirmed by the difference in side scatter (SSC) parameter (Fig. 1B). The LDNs were composed of two populations based on CD33 expression: CD33⁻ LDNs representing around two thirds of the cells and CD33⁺ LDNs. The fraction of sedimented cells consisted of NDNs, which were found to be a uniform population of CD33⁺ cells (Fig. 1C). The frequencies of the different populations of cells were estimated to be 0.43 ± 0.16% CD33⁻ LDNs, 0.20 ± 0.07% CD33⁺ LDNs, and 0.10 ± 0.03% M-MDSCs out of PBMCs (n = 13). In contrast with rhesus LDNs and NDNs, the human counterparts all showed CD33 expression (Fig. 1D), although human LDNs are known to consist of a heterogeneous population of immature/mature and inactivated/activated neutrophils (17). LDNs and NDNs showed a similar level of granularity (SSC) in both humans and rhesus macaques (Fig. 1C, 1D).

Because neutrophils at different stages of differentiation display differential patterns of cellular elements, we performed hematological staining of the purified subsets to further characterize their heterogeneity and maturation status. We found that CD33⁻ LDNs were mainly composed of immature neutrophil precursors (promyelocytes and myelocytes) with round or oval nucleus and bluish pink cytoplasm containing granules (Fig. 1E). Eosinophils were also present within the CD33⁻ LDNs. In contrast, CD33⁺ LDNs consisted of neutrophils of later stages of differentiation, where most of the cells were hypersegmented and larger compared with normal segmented neutrophils. CD33⁺ LDNs were also more heterogeneous, indicated by the presence of other neutrophils, that is, banded neutrophils. Distinct from LDNs, NDNs were found to be uniform mature neutrophils with segmented nucleus. Purified CD14⁺ monocytes were stained as controls and exhibited round nuclei and blue-stained cytoplasm as expected.

To further define the phenotype of CD33⁺ and CD33⁻ rhesus LDNs, we measured the expression of CD10, CD49d, CD11c, and CD45RA associated with neutrophil development (18) and typical neutrophil activation markers CD11b and CD62L. We found that CD10, uniquely expressed on segmented mature neutrophils, was highly expressed on CD33⁺ LDNs, but was low or undetectable on CD33⁻ LDNs (Fig. 1F, 1G). Vice versa, CD49d, present on immature neutrophils and absent on segmented neutrophils, was expressed on CD33⁻ LDNs, but not on CD33⁺ LDNs. Both subsets expressed CD11c, but not CD45RA. Downregulation of CD62L was observed only on the CD33⁻ LDNs, whereas CD11b was expressed on both subsets. Collectively, this suggests that rhesus CD33⁺ LDNs display a mature phenotype and that CD33⁻ LDNs represent immature neutrophils with activated state.

As expected, we found that CD33 was expressed on human M-MDSCs, as well as CD14⁺ classical monocytes and CD11c⁺ myeloid dendritic cells (19) (Fig. 1H). In contrast, the rhesus counterparts did not show surface expression of CD33 (Fig. 1H). This was verified using two different anti-CD33 Ab clones (HIM3-4 and AC104.3E3), of which the latter bound CD33 on rhesus LDNs. Expression of CD33 is therefore not a strategy to identify M-MDSCs in rhesus macaques. Human M-MDSCs and monocytes have been reported to express both CD11b and CCR2 (3, 20). Similarly, rhesus M-MDSCs and monocytes coexpressed CD11b and CCR2 (Fig. 1I). Based on these data, we defined a strategy for identifying the different populations of MDSCs in rhesus macaques.

To functionally assess whether any of the rhesus neutrophil subsets represented PMN-MDSCs with suppressive effects on T cells, we cocultured purified CD33⁻ LDNs, CD33⁺ LDNs, or NDNs with CFSE-labeled autologous PBMCs in the presence of SEB. As expected, SEB induced strong proliferation of T cells. In contrast, the addition of CD33⁺ LDNs induced a clear reduction in T cell proliferation (Fig. 2A, 2B). However, CD33⁻ LDNs and NDNs were both unable to reduce T cell proliferation. Furthermore, the suppressive effect exerted by CD33⁺ LDNs was cell ratio dependent (Fig. 2C, 2D). In addition to the inhibition on cell proliferation, CD33⁺ LDNs also showed evidence of suppressing the production of IFN-γ and IL-17A by T cells (Fig. 2E). To further validate this effect, we exposed PBMCs depleted of CD33⁺ LDNs to SEB, which resulted in an increased T cell proliferation (Fig. 2F, 2G). This demonstrates that rhesus PMN-MDSCs are represented by CD33⁺ LDNs.

Several mechanisms have been proposed as the suppressive function of MDSCs. Arginase-1, which is stored in granules under steady-state, is a primary candidate (21). Exocytosis of arginase-1 from MDSCs or activated neutrophils inhibits T cell responses via metabolizing l-arginine needed for T cell survival. Consequently, high mRNA levels but low intracellular protein levels of arginase-1 have been reported in MDSCs (22). We found that NDNs and CD33⁻ LDNs constitutively express intracellular levels of arginase-1, whereas CD33⁺ LDNs showed much lower or undetectable levels indicating release of arginase-1 (Fig. 2H). In addition, cocultures of PBMCs and CD33⁺ LDNs supplemented with l-arginine partly recovered the suppressed T cell proliferation (Fig. 2I, 2J). Addition of l-arginine to PBMC cultures with or without SEB stimulation did not increase T cell proliferation (data not shown). This suggests that release of arginase-1 from rhesus CD33⁺ LDNs is one mechanism causing T cell inhibition.

The function of M-MDSCs was evaluated using a similar strategy as described earlier. Because M-MDSCs could not be isolated based on CD33 expression, we purified HLA-DR⁻ CD14⁺ cells, representing M-MDSCs. The cells were compared with total HLA-DR⁺ cells for their ability to interfere with T cell proliferation. M-MDSCs were able to suppress T cell proliferation and again slightly reduce the production of IFN-γ, although not significantly (Fig. 3A–C). In contrast, HLA-DR⁺ cells showed no influence on T cell responses.

Human M-MDSCs can be differentiated in vitro from monocytes either using sorted cells or bulk PBMCs by exposure to cytokines (23, 24). It was reported that generation of rhesus M-MDSCs from monocytes using GM-CSF, IL-4 plus PGE₂ was not successful, although this mixture was able to generate human M-MDSCs (25). We therefore tested several different cytokine combinations for generating M-MDSC-like cells from rhesus PBMCs. In cultures supplemented with GM-CSF either alone or in combination with IL-6, IL-11, IL-1β, or PGE₂ for 7 d, a population of CD14⁺CD11b⁺ cells was differentiated (Fig. 3D, 3E). This population appeared to contain both monocytes and M-MDSCs based on the differential expression of HLA-DR (Fig. 3F, 3G). The combination of GM-CSF and IL-6 was particularly effective in differentiating HLA-DR^low/− cells. The generated M-MDSC-like cells were assessed for their T cell inhibitory function. Due to the small numbers of HLA-DR⁻CD14⁺ cells that could be retrieved from the cultures, it was only technically feasible to sort the total CD14⁺ cells and coculture them with PBMCs plus SEB. Nevertheless, we found that the T cell proliferation was strongly reduced in the presence of CD14⁺ cells isolated from GM-CSF/IL-6 cultures compared with CD14⁺ cells from unstimulated cultures (Fig. 3H, 3I). A population of cells reminiscent of M-MDSCs with suppressive effects on T cells can therefore be generated in vitro from rhesus PBMCs.

Numerous studies have been performed to understand the relationship between MDSCs and different pathological conditions, whereas the role of MDSCs during temporary inflammatory conditions like vaccination is largely unexplored. We analyzed how the levels of MDSCs were affected by vaccine administration by monitoring rhesus macaques receiving a novel influenza vaccine based on mRNA encoding for hemagglutinin formulated in lipid nanoparticles (12). This vaccine formulation induced hemagglutinin-specific CD4⁺ T cell response detected by IFN-γ production, as well as neutralizing hemagglutinin Ab titers above the protective levels reported for seasonal influenza vaccination (Fig. 4A, 4B). We found that the frequencies of M-MDSCs in the blood rapidly increased at day 1 after vaccine administration (Fig. 4C). Classical monocytes also increased after administration. CD33⁺ LDNs and CD33⁻ LDNs also showed a trend toward increased frequencies (Fig. 4D). The levels of all subsets had returned to prevaccination levels after 7 d (data not shown).

It is well described that MDSCs regulate immune responses locally in inflamed tissues or tumors. We have earlier found that administration of different formulations of vaccines leads to a local inflammation at the site of injection and dLNs as a result of infiltration of immune cells and cell activation (14, 26). We therefore analyzed the frequency of MDSCs in biopsies from the site of injection (muscle), as well as dLNs collected at 1 d after vaccination. We found that similar to blood, the frequencies of both M-MDSCs and classical monocytes were significantly elevated at the site of vaccine injection compared with the donor-matched PBS injection site (Fig. 4E). Monocytes also accumulated in the vaccine dLNs, whereas this was not found for M-MDSCs (Fig. 4F). Infiltration of granulocytes was also found both at the vaccine injection site and dLNs (Supplemental Fig. 1). However, because of the limited knowledge about the phenotype of PMN-MDSCs in tissues, we were unable to conclude whether the infiltrating granulocyte population contained suppressive PMN-MDSCs. We recently found that maturation of infiltrating innate cells and type I IFN production were dependent on mRNA, but that the lipid nanoparticles alone were able to cause infiltration of cells to the site of injection (27). In this study, we found that administration of the lipid nanoparticles alone without the mRNA cargo was sufficient to induce infiltration of M-MDSCs to the site of injection (Supplemental Fig. 2), as we reported previously for monocytes, DCs, and granulocytes (27). This indicates that the lipid nanoparticles induce inflammation and likely provide an adjuvant effect for the vaccine.

Transcriptomics analyses to measure MDSC-associated genes in the biopsies were also performed. Fifty-five genes were selected and divided into four distinct functional properties. Gene expression of inflammatory molecules such as S100A8/A9, IL-1β, COX-2, and IL-6, which are proposed to be critical in the expansion of MDSCs, were markedly upregulated at the vaccine injection sites compared with PBS injection sites (Fig. 4G, far left panel). Several genes encoding for mediators involved in the suppressive mechanisms by MDSCs such as the typical soluble molecules arginase-1, IDO, NO synthase, and IL-10, as well as the surface molecule PD-L1 that mediates T cell suppression via cell–cell interaction, were also upregulated (Fig. 4G, middle, left panel). Elevated gene expression of chemokines such as CCL2, CCL4, and CCL5 known to attract MDSCs to tumor and infection sites (8) was observed in vaccine injection sites (Fig. 4G, middle, right panel). Mediators in inflammatory pathways such as MyD88 and NLRP3 were elevated at the vaccine injection sites (Fig. 4G, far right panel). The increase of genes was also found in vaccine dLNs but was much lower than in the vaccine injection sites. Genes such as arginase-1, IL-10, CCL2, and CCL5 that were elevated at the vaccine injection sites showed a smaller change in the dLNs (Fig. 4H, 4I). Although it is important to note that indicated genes are not exclusively linked to MDSCs but reflect a general induction of proinflammatory cytokines and inflammatory signaling pathways at the vaccine injection sites, this suggests that such inflammatory milieu formed locally after vaccine administration attracts multiple immune subsets including MDSCs. We therefore speculate that MDSCs are important for the prevention of excessive immune activation caused by vaccine exposure.

As one of the major immune regulators, MDSCs have been intensively studied, with most focus on their contribution to tumor development, but also in other inflammatory conditions. Although the significance of MDSCs in immune regulation has been supported by numerous studies, critical information about their heterogeneity and phenotypic identification is still lacking. Initial studies mainly defined MDSCs based on expression of myeloid markers and absence of HLA-DR plus lymphoid-associated Ags. However, this criterion is not sufficient due to the phenotypic similarity of MDSCs to other cell subsets. The main challenge is the identification of PMN-MDSCs, which share almost all surface markers with other neutrophil subsets. Increasing evidence shows that human PMN-MDSCs represent a unique subset of LDNs, and Ficoll centrifugation is therefore regarded as the only method to enrich PMN-MDSCs from blood (3). Additional functional assessment is therefore critical to accurately identify them.

We found that CD33⁺ LDNs represent PMN-MDSCs in rhesus macaques and possess suppressive effects on T cells, which is partly mediated by release of arginase-1. Other mechanisms are likely also involved and remain to be investigated. Interestingly, we found that suppressive rhesus CD33⁺ LDNs display a mature phenotype, which was unexpected because MDSCs are considered immature myeloid cells. However, a recent study also showed that suppressive human CD10⁺ LDNs display a mature phenotype (28). In addition, some studies suggest that suppressive LDNs consist of activated or degranulated mature neutrophils with lower SSC profile relative to NDNs (17, 28). Although in patients with lung cancer or septic shock, LDNs and NDNs showed a similar SSC (6, 29). In this study, we found that rhesus LDNs and NDNs showed no difference in SSC. The different results may depend on the health status of the individual or in this case may be a rhesus-specific observation. The relationship among PMN-MDSCs, LDNs, and NDNs continues to be a growing discussion, particularly with regard to cell origin, diversity, and nomenclature (30).

M-MDSCs have been investigated in more depth than PMN-MDSCs because of their longer life span and better cell viability after cryopreservation (31). M-MDSCs have been reported to be more potent than PMN-MDSCs at excreting a suppressive effect (32). Human M-MDSCs are few in healthy individuals but still show a suppressive effect on T cell proliferation (33). Consistent with this, we found that the few rhesus M-MDSCs purified from healthy animals had a suppressive effect on T cells. In vitro generation of human M-MDSCs using different combinations of cytokines has been proposed as a promising therapeutic strategy for treatment of autoimmune diseases or GVHD by adoptive transfer of in vitro–generated autologous M-MDSCs (34, 35). We found that rhesus M-MDSC-like cells can also be generated using similar cytokine combinations, especially IL-6 and GM-CSF. Although it is possible that these in vitro–generated cells are different from the MDSCs existing in vivo, it provides an in vitro model to enrich cells reminiscent of MDSCs. NHPs may therefore be valuable for preclinical investigations of MDSC-based immunotherapy and offer a high degree of clinical translatability.

NHPs are and have been critical in the development of several vaccines against infectious diseases. The use of NHPs for validating the potential of new powerful vaccine platforms such as modified mRNA vaccines before proceeding to clinical trials has been of great importance (12). We therefore took the opportunity to study MDSCs after administration of an mRNA vaccine encoding for influenza hemagglutinin that induces protective levels of Abs. We have earlier shown that administration of successful vaccine formulations induces a robust temporary inflammation and innate immune activation culminating in priming of vaccine-specific responses (26, 36). Whether MDSCs are induced as a result of the inflammation and have a role in regulating the generation of adaptive responses is not known. In this study, we found that circulating MDSCs were induced rapidly after vaccination, accompanied by the nonsuppressive counterparts of myeloid cells. MDSCs also infiltrated vaccine injection sites together with the counterpart monocytes. Vaccine adjuvants have been recently proposed to cause expansion or activation of MDSCs (37). Although the vaccine used in this study does not contain an adjuvant, both the lipid nanoparticle and the mRNA component can provide an adjuvant effect. We recently found that maturation of infiltrating innate cells and type I IFN production were dependent on mRNA, but that the lipid nanoparticles alone were able to cause infiltration of cells to the site of infection (27). In line with this, we found in this study that MDSC infiltration to the site of injection was observed with lipid nanoparticles alone. The increase of MDSCs likely depends on that there is an inflammatory response induced by the vaccine.

However, the expansion of MDSCs did not suppress the vaccine outcome, indicated by well-detectable vaccine-specific T and B cell responses in all animals. Elevated levels of MDSCs have earlier been proposed to suppress vaccine responses at later time points (4). Future studies of whether the high baseline levels of MDSCs in individuals with immune-compromised conditions (e.g., cancer) is a reason for poor responses to vaccination would be highly relevant. Considerable advances have been made in the development of cancer vaccines as a means of immunotherapy. MDSCs have been proposed to impair their efficacy (37, 38). This emphasizes the need for studies in different patient cohorts to understand whether MDSCs reduce the ability to sufficiently elicit responses to vaccination. In our study, because the accumulation of M-MDSCs was observed in vaccine injection sites and not in vaccine dLNs, MDSCs may not mobilize to the location where most of the T cell response occurs. A large number of MDSC-relevant genes were upregulated at vaccine injection sites, but to a lower degree in dLNs. However, although these genes are pivotal in regulating expansion, activation, and function of MDSCs, they are not unique to MDSCs and are likely operative in many other cell subsets. Nevertheless, innate immune activation at the vaccine injection site is necessary to induce sufficient adaptive vaccine responses, but it needs to resolve rapidly to avoid local or systemic side effects (39, 40). We speculate that infiltration of immune-suppressive cells along with immune-stimulatory cells is physiologically important to prevent overreactive responses. The rapid and transient expansion of circulating MDSCs and their differential distribution in peripheral sites suggest an early immune-balancing role of these cells. Because the majority of vaccines are administrated by i.m. injection, local innate immune events play an important role in shaping adaptive immunity and finally determine vaccine outcome (36). The inflammatory microenvironment formed at the local injection site recruits circulating MDSCs that in turn produce immune-suppressive mediators and prevent excessive inflammation. However, compared with monocytes and DCs, much fewer numbers of MDSCs appear to migrate to the vaccine dLNs, which could explain why the generation of adaptive immune responses is not suppressed.

Overall, this study provides original data for identifying MDSCs in NHPs both phenotypically and functionally. In addition, we demonstrate that MDSCs are induced by vaccination. Further investigations of the relationship between MDSCs and vaccine outcome are required to understand how innate immune activation dictates vaccine responses.

We thank Dr. Mats Spångberg, Dr. Bengt Eriksson, and animal caretakers at Astrid Fagræus Laboratory for providing technical assistance.

G.C. and K.H. are employees of Valera LLC, and H.S. is a former employee of Moderna Therapeutics. The other authors have no financial conflicts of interest.