The recent success of immune checkpoint-inhibition in the therapy of malignant melanoma and various other cancer entities emphasizes the significance of tumor-reactive cells, especially effector T cells (28, 46). However, tumors escape immune surveillance acquiring different accesses, including reduction of immune recognition and immune activation, developing resistance to immune effector mechanisms and establishing an immunosuppressive tumor microenvironment (47). As far as the latter is concerned MDSC were shown to be key mediators of immunosuppression in the tumor microenvironment, facilitating tumor outgrowth, metastasis, and negatively influencing the efficacy of immunotherapy of cancer (48–54).
In our experiments using a murine melanoma model we applied a combined immunotherapeutic approach, LRAST, consisting of lymphodepletion with the alkylating agent cyclophosphamide, followed by i.v. reconstitution with naive congenic spleen cells and active-specific tumor vaccination using GM-CSF-secreting whole tumor cells. In the poorly immunogenic D5-melanoma model we intended to improve T cell immunization at different levels simultaneously (55). Induction of a lymphopenic environment should empower T cells with a homeostatic drive to proliferate and, by simultaneous exposure to tumor-antigens via whole cell vaccine, ensure that preferentially tumor-directed T cells colonize empty lymphatic niches (56). GM-CSF is a hematopoietic cytokine and is often used as an adjuvant in immunotherapeutic regimes, especially vaccination strategies (57–60). It acts to promote the local recruitment of antigen-presenting cells and improves their maturation, thus enhancing antigen-presentation to T cells in TVDLN (61).
In line with previous reports (7, 56), we observed a remarkable increase in frequencies of CD11b+ Ly6Chigh Ly6G− (phenotype of M-MDSC) and CD11b+ Ly6C+ Ly6G+ (phenotype of PMN-MDSC) cells (together also attributable as CD11b+ Gr-1+ cells) following LRAST. Within 24 days after lymphodepletion with 4,0 mg i.p. cylophosphamide per animal, PMN- as well as M-MDSC increased, peaking at day 7 in peripheral blood and day 10 in spleen. Similar cell kinetics were reported by Salem et al. in blood, spleen and bone marrow of C57BL/6 mice after treatment with the same amount of i.p. cyclophosphamide (62). Although we observed CD45.1+ cells of the myeloid lineage (CD45.1+ CD11b+) in the blood, spleen and tumor, reconstituted cells, in contrast to their host counterparts, did not give rise to relevant amounts of progeny with the phenotype of PMN- or M-MDSC or displayed a similar behavior in frequency kinetics over time. Thus, we conclude that CD45.1+ MDSC do not contribute considerably to an immunosuppressive tumor micromilieu in mice with established D5 melanoma.
To enhance priming of tumor-specific T cells and anti-tumor effects of cytotoxic T cells we aimed to deplete CD11b+ Gr-1+ cells. Anti-Gr-1 mAb (clone: RB6-8C5) has already successfully been used to eliminate MDSC in tumor, spleen, peripheral blood and bone marrow of tumor-bearing and control mice (21, 23, 40–42). Thus, proof of principle for the efficacy of the anti-Gr-1 antibody and its therapeutic effect slowing down the growth of malignant tumors has already been brought forward. Srivastava et al., for example, used 200 µg anti-Gr-1 mAb (RB6-8C5) every other day for a total amount of 4 weeks in a model of 3LL-lung carcinoma, starting one week after tumor inoculation. They observed a reduction of CD11b+ Gr-1+ in blood, spleen, bone marrow and tumor and also a significant reduction in tumor volume and weight (23). Using a similar approach in mice carrying 3LL-tumors, Zhang et al. were able to significantly reduce tumor-infiltrating MDSC, slowing down tumor growth and improving survival of the animals, using repeated i.p. administrations of 250 µg RB6-8C5, every 3 days starting 2 weeks after tumor inoculation (41).
In our experiments, we administered 230 µg anti-Gr-1 mAb (RB6-8C5) or isotope control via intraperitoneal injection every other day, starting with the day of cyclophosphamide administration and continuing until the animals were euthanized, thus ensuring the period of RB6-8C5 administration would cover the days with maximum MDSC frequencies. To maintain comparability, dose and time intervals between the single doses of antibody were set in accordance to previous reports (22, 41). However, assessing the depletion status of MDSC is not trivial since fluorochrome labelled Ly6G antibodies (clone 1A8), which we and others used to distinguish between PMN- and M-MDSC, do not bind due to the same binding site, when anti-Gr-1 antibodies were administered previously (21, 22, 24). Thus, we performed co-staining and quantification of RB6-8C5-bound cells using a secondary antibody-approach with fluorochrome-labeled antibodies directed against goat-IgG heavy chains of the anti-Gr-1 antibodies (21, 22, 24, 45). Complete absence of MDSC after conventional and secondary antibody staining indicated that MDSC were reliably depleted. Absence of MDSC regarding the conventional staining, but detection of secondary Ab positive cells in the same gate or neighboring areas, indicated the persistence of MDSC. Noteworthy, in some cases, we observed, (a) secondary Ab bound cells negatively stained for Ly6G, (b) a cell cloud with a positive and negative portion regarding Ly6G-staining, or (c) secondary Ab bound cells within the PMN-gate. The fact, that cells of all three cases retained the same Ly6C intensity suggests that in every case PMN-MDSC with different degrees of Gr-1 epitope saturation with RB6-8C5 were visualized. Hence, the absence of PMN-MDSC in their designated gate was not necessarily accompanied with complete absence or depletion of PMN-MDSC.
This in mind, we observed a successful depletion of PMN-MDSC in the peripheral blood, 2 and 4 days after treatment initiation, as there were no secondary antibody-bound cells in the PMN-gate or neighboring areas. In contrast, the frequencies of M-MDSC appeared to be reduced, but not completely depleted. In the time-span, when maximum frequencies of MDSC were to be expected following LRAST treatment (day 7 and day 10), frequencies of PMN-MDSC appeared to be significantly reduced, but, at the same time, a major portion of cells in the PMN-gate showed secondary Ab binding. Since anti-Gr-1 antibodies (RB6-8C5) were shown to persist on the cell surface of MDSC for up to 4 days and might retain suppressive activity (21, 22), the persisting portion of RB6-8C5-bound cells might represent an obstacle to therapy due to preserved immunosuppressive properties. Therefore, we aimed at investigating whether the observed reduction of PMN-MDSC, despite the occurrence of RB6-8C5-bound cells, would be sufficient to improve a tumor-specific T cell response and display a measurable therapeutic effect.
IFN-γ secretion is often used as a marker for the cytotoxic properties of T cells, including anti-tumor reactivity (63–66). Van den Engel et al. have already demonstrated an increase in IFN-γ secretion from TVDLN after LRAST (7). Here, we hypothesized that tumor-specific T cells in TVDLN from mice in the group with LRAST + RB6-8C5 treatment would exert improved IFN-γ producing capability due to better T cell-priming after MDSC-depletion. TVDLN from mice treated with LRAST alone already presented an increased tumor specific IFN-γ production and a delay in tumor outgrow compared to untreated mice. We could show that RB6-8C5 administration in addition to LRAST could further increase the IFN-γ-secretion from TVDLN, though significant results could not be obtained due to the variability of results per mouse. In accordance to that, tumor growth in mice treated with LRAST + RB6-8C5 appeared significantly reduced in the initial treatment phase up to 13 days after tumor inoculation, including the point in time with maximum IFN-γ secretion and covering the timespan of successful MDSC-depletion in our experiments - at least regarding PMN-MDSC. Since PMN-MDSC are known to mainly target T cell priming accounting for tumor-specific T cell tolerance (20), the improvement in tumor-specific INF-γ secretion from TVDLN and the delay of tumor-growth which is chronologically fitting to the time-span of successful reduction of PMN-MDSC, implies that the observed effects in fact are attributable to the administration of MDSC depleting antibodies. In contrast to Srivastava et al. and Zhang et al. (23, 41), who gained good long-term results due to a presumably successful long-term depletion of MDSC in their tumor models, our results clearly demonstrate the recurrence of MDSC, especially the PMN-subset after repetitive administration of anti-Gr-1 antibodies. Interestingly, the presence of RB6-8C5-bound cells, as indicated by secondary Ab binding, initially increased over time. After approximately 4 weeks of repetitive RB6-8C5-administration binding of depleting antibodies could not be observed anymore and MDSC depletion appeared insufficient. It is noteworthy though, that the comparability of our data to pre-existing literature on MDSC-depletion using RB6-8C5 is impaired due to a pervasive pre-treatment, which even bears the risk to be a MDSC driving stimulus itself. We assume that most likely side effects of the components of LRAST work in synergy and oppose MDSC eradication. Hence, despite its positive immunomodulating features, low-dose cyclophosphamide is known to increase levels of various cytokines (GM-CSF, IL-1b, IL-5, IL-10, IFN-γ, TNF-α) which can contribute to the expansion and activation of MDSC (67). GM-CSF, a major component within LRAST, and in turn a driving force in MDSC recruitment. It is not only found manifold as an adjuvant in immunotherapeutic regimes, but also produced by many human and murine tumor cell lines (68). GM-CSF has been shown to recruit MDSC into secondary lymphoid tissues with a consecutively impaired function of tumor-specific CD8+ T cells (38). In experiments with irradiated GM-CSF producing B78H1-GM cells, a cell line derived from the B16 melanoma, Serafini et al. proposed a cut-off concentration at 1500 ng / 106 cells / 24 hours, which – if exceeded – was associated with a suppression of antigen-specific T cell response (69). Irradiated D5G6 cells used in this work showed an in vitro GM-CSF production of 154 ng / 106 cells / 24 hours with an ascending tendency over 6 days (data not shown). Although, the GM-CSF concentrations observed in our experiments are below the proposed cut-off. A positive contribution to MDSC recruitment by GM-CSF cannot be ruled out, especially considering the potentially synergistic effects when combined with cyclophosphamide. Furthermore, even the anti-Gr-1 depleting antibodies themselves may act as a driver for MDSC expansion and thereby stand in the way of their own therapeutic purpose. Single application of RB6-8C5 was shown to be accompanied by enlarged spleens with increased cell numbers 9 days after antibody injection (40); observations which we also could obtain in our experiments. An increase in numbers of PMN- and M-MDSC was attributed to a pronounced proliferative stimulus on early myeloid precursors due to depletion (40). More to the point, following repetitive administration of RB6-8C5 every other day, Ribechini et al. found an induction of STAT1, STAT3 and STAT5. STAT3 in particular acts as an inducer of myeloid cell lineages and thereby may promote MDSC differentiation and activation (22).
Overall, the increasing frequencies of PMN-MDSC despite administration of anti-Gr-1 mAb on the one hand, and an increasing number of cells with cell-bound RB6-8C5 on the other, indicate that antibody administration becomes insufficient in keeping up with the reproduction / regeneration of MDSC, especially the PMN-subset. The abrogated depleting efficacy of long-term use of RB6-8C5 together with the fact, that, in the long run, no RB6-8C5 bound cells were detectable in the gates of both MDSC-subsets, implies the existence of a neutralizing mechanism (e.g. neutralizing self-antibodies) directed against RB6-8C5 antibodies.
Depending on the tissue localization the ratio of MDSC subpopulations and their suppressive activity vary, as does their susceptibility to RB6-8C5 (13). Overall, we observed a differing behavior of the M-MDSC subset compared to their PMN counterparts in response to MDSC depletion. For a period of 4 weeks, repetitive RB6-8C5 administration did not – apart from an initial reduction in peripheral blood - have any significant depleting effect on M-MDSC in the peripheral blood of tumor bearing mice. In previous reports particularly tumor-localized M-MDSC were shown to be resistant to depletion with anti-Gr-1 mAb and the frequency of Ly6Chigh cells was not altered 48 hours after a single administration of 250 µg RB6-8C5 (22, 70). Although we did not evaluate the efficacy of RB6-8C5 regarding M-MDSC depletion within the tumor tissue, we expect a similar resistance. Still, given the initial occurrence of RB6-8C5 bound cells in the M-MDSC gate and a short-term reduction in numbers of M-MDSC after treatment initiation with LRAST + RB6-8C5, we cannot assume a general resistance of M-MDSC to anti-Gr-1 mAbs. However, M-MDSC frequency kinetics appeared to be independent of anti-Gr-1 Ab administration. This has to be evaluated in future experiments.
Apart from the MDSC-depleting properties of anti-Gr-1 antibodies, a secondary focus was to elucidate the effect of RB6-8C5 on CD8+ cells and CD8+ memory T cell subsets (Tcm and Tem) in mice after LRAST treatment. Both, the anti-Gr-1 antibodies as well as lymphopenia induced by cyclophosphamide provably have an impact on CD8+ T cells and their memory subsets (11, 44, 45), but - as to our knowledge - the combined effect has not been investigated so far. In line with previous work by Matsuzaki et al., we confirmed the expression of Gr-1 on memory CD8+, but not CD4+ T cells with FACS analysis (data not shown) and observed a depletion of CD8+ memory T cell subsets following RB6-8C5 administration (45). In accordance to results from Salem et al. the relative cell frequency kinetics of CD8+ cells were obviously influenced by lymphopenia / lymphopenia driven homeostatic proliferation (62), but no significant impact of MDSC depletion on the population of CD8+ cells could be observed. However, looking at the CD8+ memory subsets (Tcm and Tem), reduced staining capability with fluorochrome-conjugated mAb directed towards Gr-1 and reduced overall frequencies indicated that not only targeting of memory T cells by anti-Gr-1 mAb occurred, but in fact a portion of both memory T cell subsets was depleted following RB6-8C5 administration, independent from pretreatment with LRAST. Nevertheless, despite of continuous injections with anti-Gr-1 mAb, we observed that frequencies of Tem, more than Tcm, in the peripheral blood strongly increased following LRAST, reflecting in a Tem:Tcm ratio of 3.5:1 by day 7, compared to a ratio of 1.2:1 in the control group. Tem in mice with RB6-8C5 monotherapy and LR + RB6-8C5 treatment (no vaccination) remained reduced in the same period of time. These results go in line with observations from Ma and coworkers. They found that the exposure to a tumor vaccine during homeostatic recovery after induction of lymphopenia resulted in strong expansion of CD4+ and CD8+ CD44high CD62Llow effector memory T cells, accompanied by a pronounced tumor-specific IFN-γ production and better tumor reactivity (11). In our experiments the proliferative drive seems to exceed the depleting properties of the anti-Gr-1 antibodies exerted upon the Gr-1 expressing CD8+ memory T cells which are capable of producing significant amounts of IFN-γ in a tumor-specific manner (11, 45). Nevertheless, it remains to be evaluated whether the observed increase in IFN-γ secretion from TVDLN is attributable to the increased occurrence of Tem, since these cells lack the ability to home lymphatic tissue (33). Although, compared to CD8+ Tem, CD8+ Tcm are referred to as the cell population with a more pronounced role in antitumor-immunity. Downregulation of CD62L enables Tem to quickly migrate to peripheral tissues and exert effector functions upon antigen encounter (35). Thus Tem might successfully invade tumor tissue and kill transformed cells. We therefore hypothesize that the increased frequencies of Tem cells might be reflected in the initial delay of tumor growth of mice treated with LRAST + RB6-8C5. Nevertheless, the influence of RB6-8C5 on the CD8+ memory T cell compartment might limit the anti-tumor efficacy of LRAST + RB6-8C5 treatment, to provide suitable data that can be translated into improved patient care.