Diffuse large B-cell lymphoma (DLBCL) is the most common non-Hodgkin’s lymphoma, representing about 30%–40% of new cases.1 Approximately, one-third of DLBCL patients develops relapsed/refractory (R/R) disease after standard therapies, that include rituximab.2 The standard second-line treatment in R/R DLBCL is salvage chemotherapy (CT), followed by autologous stem cell transplantation (ASCT). In most cases, however, patients relapse after ASCT or are not eligible for transplantation due to ineffective salvage treatment.3

At this point, rituximab, gemcitabine, dexamethasone, and cisplatin (R2-GDP) schedule, which includes lenalidomide combined with R-GDP, was used in R/R DLBCL patients in the study ‘phase II clinical trial to evaluate the combination of lenalidomide with R-GDP in patients diagnosed with R/R DLBCL not candidates for high-dose CT and hematopoietic progenitor cell transplantation’ (EudraCT Number: 2014-001620-29), performed by the Spanish Lymphoma Oncology Group (GOTEL). Lenalidomide is an immunomodulatory drug derived from thalidomide and its functions have been extensively described, mainly increasing T cell immune responses against tumors4 and in combination with rituximab, overcoming its resistance in patients with indolent B-cell and mantle cell lymphomas who were previously rituximab resistant.5

Some studies have evaluated alternative schedules to ASCT or first-line treatments since the outcome of both R/R patients is poor. The largest retrospective research to characterize overall response rates (ORR) and overall survival (OS) in the refractory DLBCL subpopulation is the SCHOLAR-1 study that communicated an ORR of 26% (7% of complete responses) and OS of 6.3 months.6 Specifically, rituximab has been combined with other schedules,7 8 showing more than 60% of ORR and a good OS in elderly patients after three or more years. Lenalidomide combined with rituximab (R2) has already been tested in R/R DLBCL patients. Phase II clinical trials concluded that R2-CHOP could improve the response and outcome of patients after six cycles of treatment.9 10 R2 has also been tested alone,11 showing interesting activity. In addition, a phase 1b study from GELTAMO group showed an ORR of 78.9% and an OS of 63% after 2 years combining lenalidomide with R-ESHAP.12

Considering the heterogeneity of the disease, the dismal prognosis of elderly R/R DLBCL patients and the different clinical outcomes reached,13 the search for new biomarkers may be considered as a research priority. Some prognostic markers have been widely described and tested in DLBCL, such as the International Prognostic Index (IPI) or the most recent revised version in the rituximab era (R-IPI), the immunohistochemical or molecular findings as those defining double and triple hit DLBCL, among others. Furthermore, the immune system seems to play an important role in this disease.1 In this context, expansion of myeloid-derived suppressor cells (MDSCs) has been found in the tumor microenvironment (TME) and in peripheral blood (PB). MDSCs orchestrate potent immune effects favoring tumor escape and have been correlated with poorer clinical outcomes in many neoplastic diseases.14 15

Tumor cell-mediated suppression causes immune disruptions, including altered hematopoiesis. This results in a considerable expansion and accumulation of immature myeloid cells (IMCs) in the TME, in secondary lymphoid organs and tissues, driven by tumor-derived growth factors and others via STAT1, STAT3, STAT6, COX2 or the NF-κB pathway.15 As a consequence, tumor prevents IMCs to follow their differentiation to normal cells, such as neutrophils, monocytes, and dendritic cells, and induce their activation, which finally confer them immunosuppressive properties. For this reason, activated IMCs were proposed to be called as MDSCs by Gabrilovich et al.16 MDSCs consist of monocytic (M-MDSCs) and granulocytic (G-MDSCs) precursors that mainly inhibit T cell immune responses using different mechanisms such as reactive oxygen species or nitrogen oxide production. MDSCs also promote angiogenesis via VEGF and suppress innate immune responses by inhibiting maturation and functions of natural killer cells and inducing macrophages into their M2 phenotype by interleukin 10 (IL-10) and transforming growth factor (TGF)-β secretion.14 15 Moreover, MDSCs also contribute to tumorous dissemination by inducing regulatory T cells (Tregs).17 Tregs inhibit immune responses through different mechanisms such as IL-10, TGF-β and IL-35 production or IL-2 recruitment.17 18 Nevertheless, the prognostic role of Tregs in DLBCL remains controversial.19 20

In the last few years, immune activation and inhibitory checkpoint molecules have emerged as promising biomarkers in cancer. In this sense, OX40 promotes effector T cell expansion and functions after its binding to OX40L, induced on antigen-presenting cells (APCs) such as dendritic cells, B cells or macrophages.21 By contrast, Programmed Death-1 (PD-1) binds to PD-L1 or PD-L2, expressed on APCs and tumor cells, to generate inhibitory signals that downregulate T cell-mediated immune responses.22 In fact, the interaction between PD-1 and its ligands is being promising to develop checkpoint blockade immunotherapies in B-cell lymphomas.23

In this work, we aimed to analyze M-MDSCs, G-MDSCs and total MDSCs in PB, Tregs and activated OX40 +PD-1- and inhibited PD-1 +OX40 T cells, in R/R DLBCL patients treated with the R2-GDP schedule, paying special attention, and comparing the immune profile evolution in two groups, choosing the landmark of 24 months of OS as clinical outcome of interest (OS <24 months vs OS >24 months groups). We also compared this immune profile before starting treatment with healthy donors (HD). The results obtained here support the hypothesis that consider these immune cells as promising prognostic biomarkers in DLBCL and, even more, these data point to MDSCs and Tregs as potential therapeutic targets to overcome immunosuppression in R/R DLBCL.

In the last decades, the heterogeneity of DLBCL and its poor outcome, especially in the R/R setting, have led to seek different prognosis factors that may predict the clinical outcome of patients, such as IPI or R-IPI, immunohistochemical or molecular tools, micro-RNAs and even vitamin D determinations. Nevertheless, the composition of the TME also seems to be a promising prognostic factor. In this regard, the stromal signature composed by monocyte/macrophage infiltrates has been correlated with a good prognostic in DLBCL, while stromal signature composed by factors that promote angiogenesis and blood vessel density has been correlated with a worse prognosis.24 Although monocyte infiltrate is associated with a better outcome, monocytes with a low HLA-DR expression have been correlated with more aggressive DLBCLs and progression,25 like immunosuppressive CD14 +HLA DR low/- monocytes mediated by the arginine metabolism in PB of B-cell non-Hodgkin’s lymphoma patients, considered as a phenotypically different population of monocytes from MDSCs,26 that also express CD33 and CD11b, among others.15

Due to the high variety of monocytes, which give them plasticity, M-MDSCs show a close relationship with other monocyte subsets. Canè et al reviewed that common monocyte progenitors (CMPs) are differentiated into some types of cells, such as inflammatory monocytes, M-MDSCs or other immature monocytes. Particularly, CMPs can be differentiated into M-MDSCs via VEGF by upregulating ARG1 and iNOS through NF-kB and elements related to hypoxia, avoiding their maturation to normal monocytes, macrophages, or dendritic cells (although M-MDSCs could also become APCs by inflammation-induced activation of p53).27 In addition, M-MDSCs have a close connection with tumor-associated macrophages (TAMs), which are present in a high (M2 phenotype) and a low (M1 phenotype) proportion in the TME, while M1 macrophages are predominant in healthy individuals. In this sense, M-MDSCs can also prevent their differentiation to mature cells maintaining the STAT3 activity; however, because of the effects of hypoxia, STAT3 activity can be highly reduced in the TME, leading the differentiation from M-MDSCs to TAMs.15

In our work, R/R DLBCL patients from the R2-GDP-GOTEL trial showed increased baseline levels of M-MDSCs, G-MDSCs and, consequently, total MDSCs in PB compared with HD. Therefore, tumor escape and immunosuppression mediated by MDSCs seem evident before receiving treatment, as described in other studies.26 28–30 Azzaoui et al were the first authors to study both M-MDSCs and G-MDSCs in PB of DLBCL patients at diagnosis, and they also found higher levels of these subsets in comparison with HD.29 Their conclusions included an interest in monitoring these cell populations during the course of the treatment. Thus, we decided to monitor MDSCs in all patients and put the focus in the two main populations (M-MDSCs and G-MDSCs) with better (>24 months) and worse (<24 months) OS. To our knowledge, this is the first time that circulating MDSCs are monitored during the follow-up of patients in DLBCL. Patients with OS >24 months showed a significant reduction in all MDSC populations, whereas the group with the worse OS experienced an increase in MDSC levels. At the end of the induction, there were also significant differences between the two OS groups in all MDSCs levels. These findings confirm that a decrease of MDSC levels is correlated with a better OS, improving the clinical outcome.

As previously mentioned, MDSCs induce T cells to their regulatory phenotype. Nevertheless, the role of Tregs in cancer seems to be uncertain since they have been associated with both good and poor prognosis, possibly due to environmental factors, tumor-derived products, and cytokine production that may affect Treg generation, recruitment and survival.31 Specifically, in DLBCL, Treg levels can also vary due to the location (TME and PB) or the phenotype (germinal center B-cell like (GCB) and ABC/non-GCB).20 Furthermore, Tregs can be gated in different ways, but commonly CD4+CD25+ and CD127- or FOXP3 +must be included. Chang et al explained that Tregs gated as CD4+CD25+ in PB of DLBCL patients lead the tumor escape in the disease.20

In the R2-GDP-GOTEL study, we have found that immunosuppression may not be mediated only by MDSCs, but also by Tregs, as this T cell subset was also higher in PB of patients before starting therapy in comparison with healthy subjects. Baraka and Salem gated Tregs as CD4+CD25+FOXP3+ and obtained similar results between 45 B-cell non-Hodgkin’s lymphoma patients (including 15 with DLBCL) and 20 healthy subjects.32 Yin et al quantified CD4+CD25+CD127- Tregs from 56 patients with DLBCL and 26 healthy controls and obtained similar results.33 Chang et al measured CD3+FOXP3+Tregs and also found higher levels of cells in 77 patients with the disease compared with 30 HD.20 By contrast, low concentrations of CD4+CD25^highFOXP3+ and CD4+CD25^highCD127^lowFOXP3+ Tregs in PB have also been observed in patients compared with healthy subjects,19 34–36 which has generated some controversy that still needs to be clarified.

In terms of response and OS, the role of Tregs in DLBCL also remains unclear and different conclusions have been extracted. Despite no differences have been found before and after treatment in some cases,19 35 other studies have elucidated an increase of Tregs.33 34 36 In our case, R2-GDP therapy induced a decrease of circulating Tregs in patients with OS >24 months, following a similar trend to that observed in MDSCs, while Tregs were found constant in the other group of patients throughout treatment. At the end of the therapy, significant differences between OS groups were also found. Based on these results, the immunosuppression caused by accumulation of Tregs in PB was diminished in patients with the best clinical outcome, reaching Treg values of HD. Pollari et al gated Tregs as CD4+FOXP3+ and FOXP3 +T-Bet+cells and only the last population was associated with a poor outcome in testicular DLBCL.37 In combination with rituximab, lenalidomide has caused the depletion of Tregs in responding patients with indolent B-cell and mantle cell lymhpomas5 and an increase irrespective of the response to the treatment in R/R primary DLBCL of the central nervous system.38

In addition to their regulatory functions, T cells can stimulate or inhibit the immune responses by inducing the expression of surface markers, such as OX40 and PD-1, respectively.39 In the R/R DLBCL patients from the R2-GDP-GOTEL study, we could establish that immunosuppression was also mediated by significant high levels of PD-1 +OX40 T cells and low levels of OX40 +PD-1- T cells compared with healthy subjects. The clinical outcome improved in patients when CD4+, CD8+, and total inhibited T cells significantly decreased and all activated T cell populations increased in PB, while a reverse trend was found in the group of patients with inferior OS. At the end of the induction, statistically differences were found between the groups in both activated and inhibited T cell levels, what represent quite promising results according to the clinical outcome. Although PD-1 +T cells in tissue seem to be favorable prognostic factors in DLBCL,40 41 higher concentrations of CD4+PD-1+T cells in PB have been associated with a low treatment efficacy and poor event-free survival and OS.42 43 Lenalidomide has also demonstrated to remarkably reduce PD-1 expression on T cells in patients with chronic lymphocytic leukemia.44 On the other hand, OX40 has been found in the TME of DLBCL patients in a low proportion compared with PD-1 and has been positively correlated with a better prognosis.45 Our results further support those results and highlight the importance of the findings in PB not only to monitor the OS, but also the clinical response. In fact, we previously reported preliminary results in which circulating MDSCs were reduced in R/R DLBCL patients with clinical benefit (CB) using the R2-GDP schedule46 and now we are showing similar results according to the 2 year-OS. In the same line, our group also reported the depletion of MDSCs in CB patients in advanced breast cancer.47 48 Those results suggest the importance of the drop in the levels of MDSCs irrespective of the therapy used.

The results we are discussing here are also supported by the positive correlations among MDSCs, Tregs and PD-1 +OX40 T cells in the R/R DLBCL patients (correlation data shown in figure 3), making clear that these cell populations have immunosuppressive functions and shape a complex network that could be dismantled by an active immune response. In this sense, tumor-infiltrating lymphocytes play an important role since their functions can be enhanced by immunotherapies, such as R2-GDP in R/R DLBCL. Currently, modern immunotherapy has definitely proved to be beneficial to oncohematological patients, since its efficacy in terms of OS impact has been demonstrated in a wide variety of tumors with different biological origin and clinical behavior.49

In conclusion, the research of this study has been focused on potential cell biomarkers found in PB of patients with R/R DLBCL monitoring the OS up to 24 months using R2-GDP as an immunotherapeutic schedule. We have demonstrated a significant depletion of MDSCs and Tregs in patients with OS >24 months. Consequently, both cell populations could be potential therapeutic targets to overcome immunosuppression in R/R DLBCL, especially in the case of MDSCs, since their role in cancer seems to be clearer. In fact, the decrease in MDSCs may contribute to the increase in activated OX40 +PD-1-, and the decrease in inhibited PD-1 +OX40 T cells in DLBCL.

Even though a limitation of this work could be the sample size of the study, it should be considered the high statistically significance found in the concentration of the different cell populations depending on the OS group. Nevertheless, additional studies are required to confirm these results.

In any case, our data support the hypothesis that MDSCs are good predictive markers of OS in R/R DLBCL patients and the consistent data of positive correlations of MDSCs with Tregs and inhibited T lymphocytes points to a possible immunoregulatory role of MDSCs in R/R DLBCL patients, that could therefore be considered as a target for immunotherapy.