Myeloid-derived suppressor cell activation by combined LPS and IFN-γ treatment impairs DC development

Authors


Abstract

Myeloid-derived suppressor cells (MDSC) and DC are major controllers of immune responses against tumors or infections. However, it remains unclear how DC development and MDSC suppressor activity both generated from myeloid precursor cells are regulated. Here, we show that the combined treatment of BM-derived MDSC with LPS plus IFN-γ inhibited the DC development but enhanced MDSC functions, such as NO release and T-cell suppression. This was not observed by the single treatments in vitro. In the spleens of healthy mice, we identified two Gr-1lowCD11bhighLy-6ChighSSClowMo-MDSC and Gr-1highCD11blowPMN-MDSC populations with suppressive potential, whereas Gr-1highCD11bhigh neutrophils and Gr-1lowCD11bhighSSClow eosinophils were not suppressive. Injections of LPS plus IFN-γ expanded these populations within the spleen but not LN leading to the block of the proliferation of CD8+ T cells. At the same time, their capacity to develop into DC was impaired. Together, our data suggest that spleens of healthy mice contain two subsets of MDSC with suppressive potential. A two-signal-program through combined LPS and IFN-γ treatment expands and fully activates MDSC in vitro and in vivo.

Introduction

Under steady state conditions myeloid/monocytic (Mo) precursor cells expressing the Gr-1 and CD11b markers have the capacity to differentiate into neutrophils, macrophages or DC in the presence of GM-CSF or other myeloid growth factors 1. This can be also mimicked in vitro2. The resulting immature or semi-mature DC present self-antigens in a tolerogenic fashion, which results in protection from autoimmunity 3. After microbial or inflammatory activation, DC can induce T-cell immune responses by presenting the foreign antigens in a different context, i.e. by upregulating costimulatory molecules and releasing proinflammatory cytokines 4, 5. The termination or prevention of these immune responses is controlled by various mechanisms including the modulation of mature DC functions, such as by Foxp3+ or IL-10-producing regulatory T cells 3, 6. Also myeloid-derived suppressor cells (MDSC) have been described to control immune responses and their major role in modulating CD8+ T-cell responses against tumors is well established 7–9. However, little is known about the mechanisms that regulate MDSC activation versus DC development from the same myeloid precursors.

MDSC function is mostly associated with the splenic expansion of Gr-1+ CD11b+ cells and is suggested to result from a specific activation of such myeloid precursor cells 1. Data from tumor-bearing mice and infection models indicate that MDSC require activation to exert their suppressor function 8–10. Proinflammatory cytokines, such as IL-1β, IL-6 and Prostaglandin E2 have been described to mediate accumulation of MDSC at tumor sites, despite the fact that it still occurred in their absence 11–16. More recently, the inflammatory mediators S100A8 and S100A9 proteins have been identified as the major MDSC attractants at tumor sites 17, 18. Furthermore, the combination of S100A8/S100A9 is also effective in blocking myeloid differentiation into macrophages and DC 18. The expansion and activation of MDSC has also been described after surgical trauma 19 and in polymicrobial sepsis, where LPS injection alone could expand Gr-1+ CD11b+ cells in the spleen 20. IFN-γ has been demonstrated to promote MDSC activity by inducing NO production and to interfere with DC development in vitro2, 21. We have shown before that LPS is able to interfere with DC in vitro where it blocks not only DC maturation, but also DC development, when added throughout the DC cultures 22. How bacterial LPS or inflammatory IFN-γ either alone or in combination can influence MDSC expansion versus suppressive activity has not been investigated.

More recent data indicate that two subsets of MDSC can be distinguished in tumor-bearing mice 23, 24. While both subsets suppress T-cell proliferation, one subset resembled monocytes and was thus termed Mo-MDSC, while the other had more similarities with PMN granulocytes and was therefore called PMN-MDSC 23. Recently, we identified two similar populations in the BM of healthy normal mice (ER and MBL, unpublished observations); however, whether both subsets of MDSC can be isolated and activated from spleens of healthy mice is not known.

Here, we report that combined LPS/IFN-γ treatment further enhances the suppressive function of in vitro generated MDSC. We identify MO-MDSC and PMN-MDSC in the spleens of healthy mice and show that, after isolation, they bear the capacity to suppress T cells in vitro. Injection of LPS/IFN-γ into healthy mice led to the activation of MO-MDSC and PMN-MDSC suppressor activity in the spleen blocking CD8+ T cell proliferation. We also show that injection of LPS/IFN-γ leads to the expansion of splenic myeloid precursors, blocking their development into DC. Together, we provide evidence for a dual activation program for MDSC in vitro and in vivo, which can be elicited by a microbial LPS signal in conjunction with IFN-γ.

Results

Treatment of day 3 BM-MDSC with LPS/IFN-γ induces NO production and impairs DC development

The development of DC from BM precursor cells in vitro is impaired in the presence of LPS 22. Since MDSC are also early products within BM-DC cultures and their suppressive function through NO release depends on IFN-γ 2, we wondered how LPS and IFN-γ in combination affects MDSC function and DC development.

Day 3 BM cultures, known to suppress T-cell proliferation in vitro, were treated with various cytokines or LPS alone or in combination. After 24 h, neither TNF, IL-6 nor IL-1β induced NO release. However, NO production was induced by either LPS or IFN-γ, but LPS and IFN-γ in combination elicited the highest release (Fig. 1A). In addition, the same combination dramatically affected the subsequent DC development. The cultures were treated on day 3 with IFN-γ, LPS or LPS/IFN-γ, further cultured with GM-CSF until day 8, and then analyzed for the surface expression of CD11c, MHC class II and CD86 molecules. Whereas IFN-γ did not influence the development of immature MHC class IIlowCD86neg DC or spontaneously matured MHC class IIhighCD86pos DC in the cultures, LPS showed the expected inhibition and the LPS/IFN-γ combination completely blocked MHC class II and CD86 expression, despite the surprising fact that the proportion of CD11c+ cells was the same under all conditions (Fig. 1B). Taken together, our data show that combined LPS/IFN-γ treatment of day 3 BM-MDSC cultures greatly induces NO release and blocks DC development.

Figure 1.

Combined treatment of BM cells with LPS/IFN-γ induces high NO production and blocks DC development. Day 3 MSDC generated from C57BL/6 mice were cultured in 24 well plates using 1×106 cells/well and stimulated over night with LPS, TNF, IL-1β or IL-6 either alone or in combination with IFN-γ. (A) On day 4 the cell supernatants were tested for NO production by Griess reaction. (B) Replicate wells were further cultured to develop into DC. On day 8 cells were analyzed for MHC class II, CD86 and CD11c by flow cytometry to determine the percentage of DC using mAb against MHC class II, CD86 and CD11c. Data are representative of three independent experiments with similar results.

Combined LPS/IFN-γ treatment enhances the suppressive capacity of in vitro generated MDSC

Next, we wanted to examine the effect of LPS/IFN-γ on the suppressive capacity of day 3 BM-GM-CSF cultures. For this purpose, we stimulated day 3 BM-GM-CSF cultures with IFN-γ, LPS or the LPS/IFN-γ combination overnight or left them untreated. Then, the cells were titrated into an allogeneic MLR (allo-MLR) to test their suppressive potential. The untreated cells served as controls and suppressed the T-cell proliferation as described before 2, while day 4 cells had a stronger suppressive capacity (Fig. 2A). Cells stimulated with IFN-γ alone partially, and cells treated with LPS alone completely lost their suppressive function. In contrast, cells stimulated with LPS/IFN-γ dramatically increased their suppressive potential as compared with the control day 3 or day 4 MDSC (Fig. 2A). To test whether this activation of suppressor function by LPS/IFN-γ would be maintained, we cultured the LPS/IFN-γ treated cells for another 5 days with GM-CSF and then used them as suppressors in an allo-MLR. The results indicated that cultures treated with LPS/IFN-γ completely lost their suppressive potential (Fig. 2B). Thus, LPS/IFN-γ treatment of MDSC activates their suppressive potential transiently. Further, culture in GM-CSF led to impaired DC development and does not maintain the suppressive function.

Figure 2.

Only the LPS/IFN-γ combined treatment enhances the suppressive capacity of in vitro generated day 3 MDSC. Day 3 BM cells of C57BL/6 mice were stimulated overnight with IFN-γ, LPS or combined LPS/IFN-γ. As positive suppressive controls the MDSC of days 3 or 4 were left untreated. (A) Capacity of stimulated and control cells to suppress allo-MLR. Proliferation was tested after 3 days by [3H]-Thymidine incorporation. (B) The LPS/IFN-γ-treated cells were washed at day 4 and further cultured until day 8. Suppressive capacity was tested by titrating the cells into an allo-MLR. Proliferation was tested after 3 days by [3H]-Thymidine incorporation. Data are representative of four independent experiments with similar results.

Characterization of six spleen cell subsets differentially expressing Gr-1 and CD11b

Previous reports indicated that the MDSC activity of Gr-1+ CD11b+ cells could only be measured with cells isolated from tumor-bearing mice or animals with infections. In the following, we wanted to elucidate whether Gr-1+ CD11b+ cells from healthy mice already have MDSC potential. From our previous experiments, we knew that BM-derived MDSC represent Gr-1low, CD11b+, CD11cneg and F4/80+ myeloid cells with ring-shaped nuclei 2. More recent results from others obtained using the spleens of tumor-bearing mice 23, 24, and our own results from the BM (ER and MBL, unpublished observations), indicate that there are at least two MDSC subsets with differential suppressive capacity present in mice. To identify the putative suppressive cell populations in the spleens of healthy mice, we stained fresh splenocytes with mAb against Gr-1 and CD11b and isolated different fractions by magnetic bead technology or cell sorting. For analysis forward scatter (FSC) and side scatter (SSC), and the morphological analysis of cytospin preparations were also used. Spleens of healthy mice consisted of at least six different subpopulations (Fig. 3). Gr-1high cells with an SSCint profile can be divided into Gr-1highCD11bint cells with ring-shaped nuclei (Fig. 3, population P1), with similarities to the BM-derived MDSC, and Gr-1highCD11bhigh cells with PMN shape (P2), indicative for neutrophils. Gr-1lowCD11bint CD115+ splenocytes comprise two different subpopulations with respect to their granularity. Within these, we found SSChigh (P3) and SSClow (P4) cells, which consisted of eosinophils as indicated by their red granular staining by eosin in the cytoplasm (Fig. 3, P3, arrows labelled E), cells with myelomonocytic morphology (Fig. 3, P3, arrows labelled M) and small non-granular cells with a lymphocyte-like morphology (Fig. 3, P4, arrows labelled L). Gr-1neg CD11blowCD115+ SSClow cells represent the fifth splenic subpopulation that also showed a myelomonocytic morphology (P5) and Gr-1pos CD11bneg CD115neg SSClow cells the sixth (P6) with a lymphocyte-like morphology.

Figure 3.

Splenocytes of C57BL/6 mice can be divided into six different subpopulations with regard to their surface markers CD11b and Gr-1, granularity and morphology. Freshly isolated spleen cells of C57BL/6 mice were stained with mAb against CD11b and Gr-1 and analyzed by flow cytometry. The dot plot shows the pattern of distribution of spleen cells regarding Gr-1 versus CD11b. Gated subsets are further shown as FSC versus SSC profiles or were sorted and cytospin preparations were stained with H&E dye. The resulting six different populations are termed P1-P6. In the cytospin preparation of P3/P4 cells with various morphologies appear as eosinophils (E), myelomonocytic cells (M) or lymphoid-like cells (L). Data are representative of three independent experiments with similar results.

The six different subpopulations assigned P1–P6 in Fig. 3 were further analyzed for their expression of a selected panel markers (Fig. 4). As expected the P1 and P2 expression patterns were consistent with early myeloid cells (Ly-6C+ F4/80+), while the P3 population expressed CCR3, which is characteristic for eosinophils. Within the P4 gate fractions of CD11c+ DC and Ly-6Chigh monocytes could be detected. In the subpopulation P5, a few cells expressed not only CD11c, but also MHC class II, NK 1.1 and DX5 (data not shown, a remarkable profile that could represent the NK cell subset with DC features that was originally named NK-DC or IK-DC 25. The Gr-1+ CD11bneg splenocytes within gate P6 expressed the B-cell marker CD45-R/B220 (data not shown) and a small subset the PDCA-1 marker, and therefore constituted a mixture of B cells and plasmacytoid DC 26. Having addressed the morphology and surface marker profile of these subpopulations, we wanted to assess the suppressive capacities of these subsets.

Figure 4.

Surface marker expression of the individual splenic Gr-1+ CD1b+ subsets P1-P6. Spleen cells were triple stained on their cell surface for Gr-1, CD11b and the indicated markers or the respective isotype controls. Analysis gates were set according to their differential Gr-1 and CD11b expression as indicated in Fig. 3 and designated P1–P6. The mAb-stained (solid line) or the respective isotype control (filled graph) histograms are shown for the indicated populations. Data are representative of three independent experiments with similar results.

Two subsets of suppressive spleen cells in healthy mice

Since suppression of T-cell proliferation is the best functional evidence for MDSC, we separated the spleen subpopulations by magnetic beads or cell sorting and tested their capacity to suppress T-cell proliferation in an allo-MLR. In vitro generated day 3 BM-MSDC served as control suppressor cells. In a first step, we sorted fresh spleen cells into Gr-1pos or CD11bpos splenocytes by magnetic bead separation (MACS) and utilized them in titrated numbers as potential suppressor cells. While the Gr-1pos fraction could not suppress T-cell proliferation substantially, the CD11bpos cells were strongly suppressive (Fig. 5A). In the next step, we isolated different subpopulations with a cell sorter to examine their suppressive potential. First, we compared Gr-1highCD11bint (P1, “ring” cells) with Gr-1highCD11bhigh cells (P2, granulocytes). As expected, the Gr-1highCD11bint subpopulation P1 was able to suppress T-cell proliferation, while splenic granulocytes (P2) did not influence T-cell proliferation (Fig. 5B). The latter was expected but has not been appreciated so far because usually all Gr-1+ CD11b+ spleen cells have been considered to have MDSC function. This, however, does not exclude that neutrophils may exert suppressive functions in other assays or experimental settings in vivo.

Figure 5.

Gr-1highCD11bint and Gr-1lowCD11bint splenocytes are able to suppress T-cell proliferation in vitro. Freshly isolated spleen cells of C57BL/6 mice were stained with mAb against CD11b and Gr-1 and sorted (A) by magnetic beads using Gr-1 or CD11b mAb or (B-D) by FACS for the indicated antibodies and for their FSC/SSC profile for the populations P3 and P4 to isolate the different splenic subpopulations. As suppressive positive control cells, day 3 BM-MDSC were used. Cells were then added as potential suppressor cells in triplicates at graded concentrations into an allo-MLR. Proliferation was tested after 3 days by [3H]-Thymidine incorporation. Data show mean±SEM (n=3) and are representative of 8 independent experiments.

In the next step, we wanted to know whether there were any differences regarding suppressive potential between the MDSC (P1) and the Gr-1lowCD11bint fraction (P3+P4) comprising cells of the myeloid lineage in diverse differentiation stages (Fig. 5C). Both subpopulations were able to suppress T-cell proliferation as efficiently as control day 3 BM-MDSC. Additional separation of the P3 and P4 subsets by their FSC/SSC profiles as shown in Fig. 3 indicates that the P4 subset is responsible for the suppressive effect within the Gr-1lowCD11bint fraction (Fig. 5D). When we used Gr-1negCD11blow splenocytes (P5; presumed NKDC) as a potential suppressor population (Fig. 5E), no reduction of T-cell proliferation was obtained in this setting. In addition, Gr-1pos CD11bneg cells (P6) also did not show suppressive activity in this type of assay (data not shown).

To summarize, we identified two suppressive subpopulations in the spleens of healthy mice by their differential expression of CD11b and Gr-1, FSC, SSC and morphology. The suppressive populations found here correlate with previous descriptions of such cells in the spleens of tumor-bearing mice 23, 24 and BM of healthy mice (ER and MBL, unpublished observations). The ring-shaped Gr-1highCD11bint MDSC correlate with the described PMN-MSDC and the morphologically more heterogeneous Gr-1lowCD11bint myeloid cells resemble the MO-MDSC 23.

Accumulation and activation of MDSC after injection of LPS/IFN-γ

Since LPS/IFN-γ enhanced the MDSC potential in vitro, the question remained how it would affect the suppressive activity of the MDSC subsets in vivo. To investigate this, mice were left untreated or injected at days 0, 2 and 4 with LPS/IFN-γ. On day 7 splenic Gr-1+ CD11b+ cells were sorted from spleens of these mice and tested for their potential to inhibit an allo-MLR in vitro. To our surprise, the suppressive capacity of all isolated subsets in the inhibition of an allo-MLR in vitro remained comparable and was not enhanced by LPS/IFN-γ treatment (data not shown) as we had observed using the BM-derived MDSC.

To analyze the suppressive capacity after LPS/IFN-γ injection in vivo, mice were injected with LPS/IFN-γ and then immunized s.c. with OVA-loaded and CpG-matured DC (exogenous DC) or OVA and CpG in incomplete Freund's adjuvant (IFA; endogenous DC). Control mice were left untreated or received only LPS/IFN-γ or the immunizations alone. To monitor specific T-cell suppression, all mice received CFSE-labeled OVA-specific, TCR-transgenic CD8+ OT-I cells. Injection of LPS/IFN-γ led to higher frequencies of both Gr-1+ CD11b+ cells in the spleens but not LN (Fig. 6A). Both Gr-1highCD11bint (P1) and Gr-1lowCD11bint (P3/4) MDSC splenic subsets, and also Gr-1highCD11bhigh granulocytes (P2) were increased by this treatment, while immunization by exogenous or endogenous DC alone led to only minor increases of the MDSC subpopulations (Fig. 6A). Further subgating of the Gr-1lowCD11bint cells by their FSC/SSC profiles into P3 and P4 populations after LPS/IFN-γ injection indicated a drop in P3 cell numbers from 29.2±6.7% (n=2) in untreated animals to 9.0±3.0% (n=3) in LPS/IFN-γ treated animals (n=3) and thereby an increase of P4 cells from 56.6±15.2% (n=2) to 81.3±1.5% (n=3). This may indicate a specific expansion of myelomonocytic P4 cells but not the eosinophils within P3. Analysis of the injected OT-I T cells revealed that vigorous proliferation was induced after both types of immunizations, but pre-injections with LPS/IFN-γ inhibited the proliferation of OT-I cells in the spleens (Fig. 6B). Correlated with the absence of MDSC in the LN, the proliferation of OT-I cells in these organs was largely unaffected (Fig. 6B). These data indicate that MDSC accumulation in lymphoid organs after LPS/IFN-γ injection was associated with impaired CD8+ T-cell proliferation in the same organ. To further investigate whether the depletion of Gr-1+ cells would reconstitute CD8+ T-cell proliferation, we injected Gr-1 antibodies into the immunized control mice and mice injected with LPS/IFN-γ. Unfortunately, the adoptively transferred OT-I cells were depleted after Gr-1 injection (data not shown), similar as described for CD8+ memory T cells 27.

Figure 6.

LPS/IFN-γ injections accumulate Gr-1+CD11b+ MDSC subsets in the spleen but not LN leading to suppression of CD8+ T-cell proliferation only in the spleen. Mice were injected with a combination of LPS/IFN-γ i.p. at days 0, 2 and 4 or remained untreated. At day 10, they were immunized s.c. into the footpads with IFA/CpG/OVA peptide or mature OVA peptide-loaded DC. CFSE+ OT-I cells were transferred i.v. at the same day 10. The animals were sacrificed 4 days after the immunization/OT-I-cell transfer. LN and spleen were removed and analyzed for the subsets P1 and P3/P4 of Gr-1+ CD11b+ cells (A) and Vβ5+ OT-I-cell proliferation (B). The data are representative of two independent experiments.

Impaired capacity of Gr-1+ CD11b+ cells to develop into DC after LPS/IFN-γ injections

To evaluate whether LPS/IFN-γ activation would affect the potential of Gr-1+ CD11b+ cells to develop into DC, we investigated the splenic CD11c+ DC frequency 6, 9 or 12 days after LPS/IFN-γ treatment. The total number of spleen cells increased after this treatment until day 9 (Fig. 7A), similar to the absolute numbers of splenic DC, which resulted in a stable percentage of DC during the whole period investigated (Fig. 7B). This would indicate that the splenic DC populations are not affected by LPS/IFN-γ injections. However, since the turnover of spleen DC is very high with half-lifes of 1.5–2.9 days 28, and the splenic DC pool may not be recruited from BM precursors but a splenic precursor pool 29, potential effects of the LPS/IFN-γ treatment may be rapidly corrected for in the splenic DC pool. In order to test the capacity of the Gr-1+ CD11b+ cells regarding their potential to differentiate into DC, these cells were sorted from mice injected with LPS, IFN-γ, LPS/IFN-γ or untreated animals. After in vitro culture with GM-CSF for 3 days the animals injected with LPS or LPS/IFN-γ generated far fewer CD11c+ CD86neg immature DC as compared with the control or IFN-γ injection alone (Fig. 7C), similar to that which we observed in vitro. Thus, activation by LPS/IFN-γ blocks the capacity of splenic MDSC to develop into DC.

Figure 7.

In vivo activation of Gr-1/CD11b cells by LPS or LPS/IFN-γ impairs their capacity to develop into DC. Mice were injected i.p. with LPS/IFN-γ or with PBS at days 0, 2 and 4. At days 6, 9 and 12 the spleen cell numbers were counted (A), stained for CD11c and analyzed by flow cytometry (B). In another experimental setting the mice were injected i.p. with LPS, IFN-γ, the combination, or with PBS at days 0, 2, and 4 and at day 7 the splenic Gr-1+CD11b+ cells from individual mice of each group were isolated by cell sorting and cultured in the presence of GM-CSF. At day 10 FACS analyses were performed with the indicated markers (C) to analyze their precursor/granulocyte by Gr-1/CD11b or immature/mature DC phenotype by CD86/CD11c. The data are representative dot plots of duplicate mice per experiment. Three experiments with similar results for all individual mice have been performed.

GM-CSF culture modulates the suppressive capacity of LPS/IFN-γ treated Gr-1+ CD11b+ cells

The question remained whether LPS/IFN-γ activated Gr-1+ CD11b+ cells would lose their suppressive capacity after further culture in GM-CSF. To test this, the same experimental setting as shown above (Fig. 7C) was used to obtain Gr-1+ CD11b+ cells from spleens of mice that were injected with LPS, IFN-γ, LPS/IFN-γ or PBS as a control. These cells were then cultured for 3 days in GM-CSF and then tested for their capacity to suppress an allo-MLR (Fig. 8A). While cells from control mice and IFN-γ treated mice remained suppressive, the Gr-1+ CD11b+ cells from the other two groups seemed to be less suppressive. Given that, for this experiment, the high turnover of cells in the spleen may influence the populations, we directly tested the effect of GM-CSF on the splenic Gr-1+ CD11b+ cells ex vivo. CD11b+ cells were sorted with magnetic beads from spleens of untreated mice and cultured for 24 h with GM-CSF only or in addition with LPS, IFN-γ, LPS/IFN-γ before they were titrated into an allo-MLR to test their suppressive potential. Only the control cells retained suppressive capacity but the LPS, IFN-γ, LPS/IFN-γ plus GM-CSF treated cells completely lost this function (Fig. 8B). These data indicate that pretreatment with GM-CSF and more strongly simultaneous treatment of splenic Gr-1+ CD11b+ cells conteracts their suppressive potential.

Figure 8.

GM-CSF counteracts the suppressive potential of LPS/IFN-γ activated MDSC. (A) In analogy to the experimental setting shown in Fig. 7C, mice were injected i.p. with LPS, IFN-γ, the combination, or as a control with PBS at days 0, 2 and 4. At day 7 the splenic Gr-1+ CD11b+ cells from individual mice of each group were isolated by cell sorting and cultured in the presence of GM-CSF. At day 10 cells were harvested, counted and their suppressive function was tested by titrating them into an allo-MLR. (B) Freshly isolated and magnetic bead-sorted CD11b+ cells from spleens of untreated mice were simultaneously treated with GM-CSF plus either LPS, IFN-γ or the combination or remained untreated for 24 h. Capacity to suppress allo-MLR was tested by adding graded doses of the cells. Data show mean±SD (n=3) and are representative of three independent experiments.

Discussion

Here, we showed that two specific subsets of splenic Gr-1+ CD11b+ cells of healthy naive mice bear the potential to become MDSC and suppress T-cell proliferation. The combined LPS/IFN-γ signaling but none of the single components led to the activation of myeloid precursors into functionally suppressive MDSC, which impaired their developmental potential into DC. Additional exposure to GM-CSF counteracted the LPS/IFN-γ effects.

Our earlier findings indicated that LPS signals on myeloid precursor cells interfered with DC development in vitro. This LPS blocking generated CD11c+ cells that expressed only little surface MHC class II and no costimulatory molecules. Consequently, these immature DC were functionally tolerogenic and induced T-cell anergy in vitro22. Here, we extended our findings in vitro by showing that combined LPS/IFN-γ treatment enforced this effect, while IFN-γ alone showed no changes. After injection of LPS, Gr-1+ CD11b+ cells accumulated in the spleen as reported before 20 and, similar to our in vitro findings, these cells showed reduced capacity to develop into CD11c+ DC. These data indicated that LPS could expand Gr-1+ CD11b+ cells and block DC development, but not whether LPS alone is sufficient to activate their suppressor function.

NO has been shown to be one of the suppressive tools secreted by MDSC 1. Here, LPS transiently induced the NO production in our BM-MDSC cultures. However, when the NO was washed off from the cells they lost their suppressive potential after continued culture in GM-CSF. These findings indicate that LPS treatment of myeloid precursors leads to a partial activation as indicated by NO release that allows a transient suppressive activity. This also may indicate that the suppressive mechanism that could be observed in the allo-MLR by using LPS plus IFN-γ for MDSC stimulation could be mediated through a mechanism other than NO production or involve a second wave of NO production that may, however, depend on IFN-γ.

IFN-γ is the major cytokine released by CD4+ Th1 cells and CD8+ CTL and has been also shown to control MDSC activity 1, 30. Our results with in vitro generated MDSC indicated that their suppressive activity could be partially abrogated by blocking IFN-γ 2, similar to that which others observed in vivo30. Treatment of BM-MDSC cultures with IFN-γ led to low NO production, but after washing, the blocking activity in an allo-MLR was reduced when compared with untreated cells. Injections of IFN-γ did not induce T-cell suppression in vivo and did not increase the number of myeloid cells in the spleen. Thus, these data indicate that neither myeloid cell accumulation in the spleen nor MDSC activity could be induced by IFN-γ alone.

The combined in vitro treatment or in vivo injection of LPS/IFN-γ led to an accumulation of myeloid cells in culture or in the spleen but with an impaired capacity to develop into DC. When LPS/IFN-γ-treated day 3 BM cell cultures were further propagated with GM-CSF they not only showed an impaired capacity to develop into DC, but also lost their suppressive potential. This GM-CSF effect was also observed with splenic MDSC ex vivo. At this point, it is speculative whether activated MDSC may further undergo a special myeloid development toward tolerogenic DC, since they acquire the CD11c marker without leading to fully functional immunogenic DC, similar to that observed with LPS treatment alone 22.

Only the simultaneous presence of LPS plus IFN-γ activated MDSC suppression in vitro. During the course of a bacterial infection, LPS will be detectable to immune cells at the earliest time points, initiating an anti-microbial immune response. Subsequently, as a result of the adaptive immune response IFN-γ producing T cells will be generated, which produce IFN-γ only locally when they are restimulated by antigens at the infection site. The simultaneous treatment of mice with LPS/IFN-γ may reflect a chronic infection 31–33 or sepsis 20, where pathogens carrying LPS or other immunostimulatory molecules are systemically present, together with high levels of IFN-γ. In this situation, MDSC activation may occur as a beneficial mechanism to control immunopathology in the host. We can only speculate why the single treatments with either LPS or IFN-γ have opposing effects as compared with the combination of both. However, the single detection of either LPS or IFN-γ by MDSC also results in useful mechanisms. Recognition of pathogen alone requires an immune response and should not be associated with increased suppression. Similarly, detection of IFN-γ by MDSC is just part of a normal ongoing immune response where suppression would also be counter-productive.

The suppression of T cells in the spleen but not in the LN would further support this model, as only the systemic T-cell activation is blocked in the spleen but not in the local tissue-draining LN where T-cell activation may continue. Although increased frequencies of MDSC in LN have been observed by others using models for sepsis, or tumors, the percentages remained low around 2–4% as compared with about 20% in the spleens 20, 34. Under the conditions tested here, there was clearly no induction of Gr-1+ CD11b+ cell in the LN. The slightly lower rate of CD8+ T-cell proliferation detected in the LN of our mice must therefore reflect the reduced recirculation of T cells from the spleen. Together, this may indicate that the major site of suppression in mice is the spleen rather than the LN.

It is a matter of debate whether MDSC exist in the steady state. Our ex vivo and in vivo data demonstrate that two subsets of Gr-1+ CD11b+ cells with suppressive potential exist within the spleen, indicating preformed MDSC in the steady state. However, our previous findings indicated that blocking IFN-γ within the allo-MLR that is undergoing suppression by BM-derived MDSC limited NO production and partially reconstituted T-cell proliferation 2. Similar data have been acquired in vivo27. This clearly indicates that during the allo-MLR, factors such as IFN-γ plus presumably others are instantly produced to activate MDSC function. We could observe this especially at high antigen doses during the in vitro stimulation of TCR-transgenic T cells 2. Enhancing the doses of IFN-γ in vivo by injection in combination with the presence of tumor-derived or microbial factors such as LPS then may further unfold the full potential of MDSC and increase their numbers in the spleen. Thus, MDSC suppressor activity requires activation, which may have more than one activation level.

However, Gr-1+ CD11b+ cells isolated from mice that were pre-injected with LPS/IFN-γ and tested ex vivo for suppression did not show an elevated suppressive potential as compared with untreated mice (data not shown). The reasons for this discrepancy are unclear but could also be related to the high cellular turnover within the spleen. Particularly, the LPS/IFN-γ injections impaired the OT-I-cell proliferation, indicating that MDSC activity increased. Taken together, we observed the increased OT-I T-cell suppression in the spleen after LPS/IFN-γ injections without a higher intrinsic suppressive potential of the splenic MDSC, but increased frequencies of the two splenic MDSC subsets. These data indicate that MDSC activity in vivo may be predominantly regulated by their number in the spleen.

The elimination of MDSC would be a tool to investigate their role in vivo. However, injecting Gr-1 antibody to deplete MDSC in the steady state was not successful in our hands since the antibody depleted activated CD8+ OT-I T cells, similar to that which has been described before for memory CD8+ T cells by others 27. Genetic models to specifically ablate MDSC or their functions will clarify this point in the future.

Finally, it is of note that here we identified the Gr-1highCD11bhigh expressing neutrophils and the Gr-1+ CD11blowSSChigh eosinophils as non-suppressive populations in the spleen of healthy mice. This fact is worth mentioning since many reports generalize all Gr-1+ CD11b+ cells in the spleen as candidates for MDSC activity. However, differentiated neutrophilic granulocytes and eosinophils do not seem to have suppressive potential in this type of assay.

In conclusion, our data indicate that healthy mice contain two splenic subsets of Gr-1+ CD11b+ cells with suppressive capacity. Their full activation requires two simultaneous signals such as LPS/IFN-γ. This activation by the simultaneous presence of these factors resembles chronic infections or sepsis and is triggered to inhibit immunopathology.

Materials and methods

Mice

C57BL/6 and BALB/c mice between 4–12 wk of age were used for generating single cell suspensions of BM, spleen and LN as well as for the in vivo experiments (C57BL/6 mice). Animals were purchased from Charles River, Germany, or obtained from our internal breeding facilities in Erlangen or Würzburg, Germany. All experiments were performed according to the animal protection laws and under control and with permission of the local authorities (Regierung von Mittelfranken AZ: 621.2531.32-06/02, TS-99/14; Regierung von Unterfranken AZ: 54-2531.01-08/07).

Media and reagents

For cell culture R10 medium was used consisting of RPMI 1640 (Lonza) supplemented with 100 U/mL penicillin (Sigma), 100 μg/mL streptomycin (Sigma), 2 mM L-glutamine (Sigma), 50 μM β-mercaptoethanol (Sigma) and 10% heat-inactivated FBS (PAA, Cölbe, Germany).

Isolation and preparation of cells and treatments in vitro

To generate single cell suspensions of spleens, LN or BM organs were removed under sterile conditions and the popliteal, inguinal, axillary, cervical and mesenteric LN were disrupted with glass slides and resuspended in PBS (Lonza) as described in detail before 35. Stimulation of MDSC were performed over night with 0.1–1 μg/mL LPS (Sigma), 500 U/mL TNF, 200 U/mL IL-1β or 200 U/mL IL-6 either alone or in combination with 100 U/mL IFN-γ (all from Preprotech).

BM-derived MDSC and DC cultures

The culture of BM cells from C57BL/6 mice to generate d3 MSDC or DC was performed as described previously 2. Shortly, fresh BM cells were cultured in 10 mL R10 medium with 10% culture supernatant of a murine GM-CSF transfected cell line (equivalent to >200 U/mL). After 3 or 4 days of culture MSDC could be harvested as non-adherent cells.

Flow cytometry and cell sorting for MDSC

Cells were stained with PerCP-conjugated mAb against Gr-1 and either PE- or FITC-CD11b or FITC- or PE-conjugated mAb against CD11c, F4/80, CD45R/B220, DX5, NK1.1, MHC class II (I-A; M5/114), or CCR3-AlexaFluor647 (all BD Pharmingen, Hamburg, Germany) or for Ly-6C (ER-MP20, AbD Serotec), or CD115-APC, PDCA-1-AlexaFluor647 (eBiosciences) or the appropriate fluorochome-conjugated mAb or supernatants as isotype controls at 2–5 μg/mL in PBS containing 0.1% sodium azide and 5% FBS for 30 min on ice in the dark. Samples were washed once in staining buffer, measured and analyzed with a FACScan (Becton Dickinson, Heidelberg, Germany).

For some experiments fresh spleen cells were sorted either by MACS technology (Miltenyi, Bergisch-Gladbach, Germany) or with a Mo-Flo highspeed sorter (Cytomation, Freiburg, Germany) or a FACS-Vantage (BD). The purity of sorted cells was generally above 90%. Sorted cells were then cultured as indicated.

Allo-MLR

The different Gr-1+ CD11b+ splenocyte populations from C57BL/6 mice were sorted and cultured as triplicates in a 96 well flat-bottomed plate (Falcon) in R10 medium at titrated numbers. In vitro generated untreated day 3 or 4 BM-derived MDSC served as positive control. LN cells from BALB/c mice (4×105 well) were used as responder population. Mature day 9 DC from C57BL/6 mice were used as stimulator cell population (1×104 well). After 3 days of culture cells were pulsed with 1 μCi [3H]-thymidine (Amersham) in HL-1 medium for 16 h and harvested onto filtermats with an ICH-110 harvester (Inotech, Dottikon, Switzerland). Filters were counted in a 1450 microplate counter (Wallac, Turku, Finland).

Measurement of NO as nitrite production

NO was measured as nitrite production using the Griess reaction 36. Briefly, 50 μL of cell culture supernatant were put into a 96 well ELISA-plate (Corning) as duplicates with titrated NaNO2 (Sigma) in R10 medium serving as a standard. An aqueous solution of 0.1% naphtylethylendiamine dihydrochloride (Sigma) and 1% sulfanilamide (Sigma) in 5% conc. H3PO4 (Merck, Darmstadt, Germany) in water were mixed 1:1 and 50 μL of this solution were added to 50 μL of the samples. The evoked color reaction was measured after 10 min in the ELISA reader (Molecular Devices) at 492 nm and nitrite concentrations were calculated from the sodium nitrite standard curve.

Cytospins

Sorted splenocytes (2×105) were resuspended in 200 μL R10 medium and centrifuged onto a microscope slide using a Cytospin-3 (Shandon, Life Sciences International, Astmoor, UK). Then, slides were stained with hematoxylin/eosin dye according to standard protocols.

Ex vivo analysis of spleen cells after injection of LPS/IFN-γ

Female C57BL/6 mice received 1 μg IFN-γ and 10 μg LPS i.p. per mouse on day 0. IFN-γ- and, where indicated, LPS injections were repeated on day 2 and 4. At 6, 9 and 12 days after first injection spleens of the different test groups were removed and analyzed by FACS staining. Fractions of the sorted cells at day 6 after the LPS/IFN-γ injections were cultured for another 4 days in saturating doses GM-CSF. mAb against Gr-1 and CD11b were used to determine the amount of potential suppressor cells in the spleen; mAb toward CD11c should reveal the effect of IFN-γ/LPS on the percentage of splenic DC. In addition their suppressive capacities was tested by titrating the cells into an allo-MLR.

In vivo analysis of CD8 T-cell suppression

Mice were injected with a combination of 1 μg IFN-γ and 10 μg LPS i.p. at days 0, 2 and 4 or remained untreated. At day 10, they were immunized s.c. into the footpads with 50 μL of a 1:1 mixture of IFA (Sigma): PBS containing 6 μg CpG oligonucleotides (MWG biotech AG, Ebersberg, Germany) and 100 μg OVA peptide (SIINFEKL) or 4×106 matured and OVA peptide (10 μM)-loaded DC. Mature DC were generated after 8 days of culture in GM-CSF-supernatant as described 35 and by 1 μg/mL CpG and 10 μM OVA peptide for 4 h. For preparation of OT-I cells, a single cell suspension from spleens and LN was prepared, lysed for erythrocytes with 0.8% ammoniumchloride, washed and then labeled with CFSE (5 μM in PBS for 15 min, RT; Molecular Probes). OT-I cells (1.3×107 cells per mouse) were transferred i.v. at the same day 10. The animals were sacrificed at day 14 and the single cell suspensions of spleens and LN used for flow cytometry analysis with biotinylated mAb against Vβ5 detected by an streptavidin-PerCP conjugate (both BD Pharmingen).

Acknowledgements

We thank Gerold Schuler for his continuous support, Thomas Hünig for support and helpful discussions, Gudrun Schell for expert technical assistance and Christian Linden for cell sorting. This work was supported by the Interdisciplinary Centre for Clinical Research (IZKF) Erlangen for VG, DFG through LU851/4-1 for SR and MBL and the Collaborative Research Centre SFB479 for ER and MBL.

Conflict of interest: The authors declared no financial or commercial conflict of interest.

Ancillary