After Treatment Decrease of Bone Marrow Tregs and Outcome in Younger Patients with Newly Diagnosed Acute Myeloid Leukemia

An emerging body of evidence demonstrates that defects in antileukemic effector cells in patients with acute myeloid leukemia (AML) can contribute to the development and/or persistence of the disease. In particular, immune suppressive regulatory T cells (Tregs) may contribute to this defective antileukemic immune response, being recruited by bone marrow leukemic cells to evade immune surveillance. We evaluated Tregs (CD4+/CD45RA-/CD25high/CD127low), performing multiparametric flow cytometry on freshly collected bone marrow aspirate (BMA), in addition to the usual molecular and cytogenetic work-up in newly diagnosed AML patients to look for any correlation between Tregs and the overall response rate (ORR). We studied 39 AML younger patients (<65 years), all treated with standard induction chemotherapy. ORR (complete remission (CR)+CR with incomplete hematologic recovery (CRi)) was documented in 21 out of 39 patients (54%); two partial responder patients were also recorded. Apart from the expected impact of the molecular-cytogenetic group (p = 0.03) and the NPM mutation (p = 0.05), diagnostic BMA Tregs did not show any correlation with ORR. However, although BMA Tregs did not differ in the study population after treatment, their counts significantly decreased in responder patients (p = 0.039), while no difference was documented in nonresponder ones. This suggested that the removal of Treg cells is able to evoke and enhance anti-AML immune response. However, the role of BMA Tregs in mediating immune system-AML interactions in the diagnostic and posttreatment phase should be confirmed in a greater number of patients.


Introduction
In recent years, considerable progress has been made in deciphering the molecular and genetic heterogeneity of acute myeloid leukemia (AML) [1] and in defining new diagnostic and prognostic biomarkers [2]. In particular, having an impact on the clinical outcome, the therapeutic strategy largely depends on the European LeukemiaNet genetic risk stratification [3].
Moreover, a growing number of recurring genetic changes have been recognized in the new World Health Organization (WHO) classification of AML [4]. It appears that the prognostic relevance of integrated genetic profiling [5] is crucial in the diagnostic phase and seems to explain the clinical complexity of AMLs which inevitably differ from each other. Consequently, novel therapies, targeting some of the known genetic lesions, are now amply appealing [6]. However, in spite of these new targeted drugs, the outcome of patients affected by AML remains dismal paving the way for emergent players.
In AML patients, the bone marrow (BM) aspirate, studied through a genomic and epigenomic-based approach [7], is part of the altered BM-AML-immune environment [8,9]. The AML microenvironment may be analyzed to explore its potential ability to contribute to the disease onset and outcome since it is immunosuppressive and antiapoptotic by itself thus favoring the immune escape and survival of malignant hematopoietic cells [10].
Given the complex interactions between AML cells and the many components of their BM-environment [11][12][13], we should expect additional diagnostic and/or prognostic roles for bone marrow aspirate T regulatory cells (BMA Tregs), which previous studies have emphasized [10,14,15]. However, in solid tumors, the role of Tregs seems to be firmly associated with tumor escape from immunosurveillance and consequently a worse outcome [16,17], whereas in the AML setting, their action is still matter of debate [18]. In fact, apart from selected lymphomas in which tumor infiltrating Treg cell density seems to be associated with a better outcome [19], there are conflicting results in terms of a worse [20][21][22] or better prognosis [23] in AML patients. Moreover, the possible role of Tregs was indistinctly studied analyzing BM [23], peripheral blood (PB) [20,21,24,25] and BMA plus PB [22,26] at different time points (diagnosis [20][21][22][23][24]26], and recovery phase posttreatment [23][24][25]) which made it difficult to reach firm conclusions.
Therefore, the aim of our study was to update our previous report [27] in an attempt to clarify the possible prognostic impact on the overall response rate (ORR) and outcome of the BMA Tregs, evaluated at the diagnostic phase and after treatment, in a prospective series for younger patients affected by AML.

Patients.
We prospectively analyzed newly diagnosed AML patients (≤65 years) treated at our institution from March 2016 to May 2020. Patients gave written informed consent to the collection of personal data in accordance with the Declaration of Helsinki and Italian law.
All patients underwent induction chemotherapy (i.e., "3 + 7") with cytarabine 100 mg/m 2 , intravenously, on days 1 to 7 and an anthracycline (daunorubicin 60 mg/m 2 on days 1 to 3 or mitoxantrone 10 mg/m 2 on days 1 to 3) and midostaurin 50 mg twice daily, in case of FLT3 mutation (ITD or D835) on days 8-21 (since January 2018). Thereafter, responders underwent consolidation chemotherapy with cytarabine at higher doses (up to 2 cycles) and midostaurin 50 mg twice daily in case of FLT3 mutation (ITD or D835) on days 8-21 (since January 2018), and nonresponders underwent salvage chemotherapy (FLAG-Ida for all patients) as a bridge to allotransplantation. The AML response was evaluated according to the ELN-2017 [3] limiting the evaluation of the quality of response to the morphologic complete response (CR). At the time of the evaluation of response, BMA-T cell population (CD3+,CD4+; CD8+; Tregs) together with B (CD19+) and natural killer (NK) cells count was repeated for comparison with diagnostic samples. A subanalysis of BMA-Treg count reduction in midostaurin-treated FLT3-mutated AML and low-risk patients was also performed.

Statistical Analysis.
We were interested in studying Tregs in responder patients. Accordingly, assuming that 70 ± 15% of AML patients were responder, along with a confidence coefficient ð1 − αÞ = 0:95 [confidence level =95%] and a confidence interval = 0:7 ± 0:15, respectively, we needed 36 AML patients to treat. The Mann-Whitney rank sum test was used to compare absolute cell counts, while chi-square or Fisher's exact test (2tailed) was performed to compare proportions. The comparison of the diagnostic BMA (dBMA) population with the posttreatment one was performed with paired t-test or Willcoxon signed rank test, as appropriate. The variables analyzed for a correlation with ORR were age, white blood cells (WBC), integrated molecular-cytogenetic risk, the NPM mutation, the FLT3 ITD or D835 mutation, NPM mut FLT3 wt-Normal Karyotype, BMA Tregs, and de novo vs secondary AML. Covariates in the multivariate logistic regression models were chosen by stepwise-with-backward elimination variable selection procedures. The discriminatory power of the dBMA Treg value to predict response was assessed by estimating the area under the ROC curve (AUC). The optimal cutoff was determined by maximizing both sensitivity and specificity, computed at the optimal cutoff, as reported along with the 95% confidence intervals. The variables analyzed for a correlation with the dBMA Treg median value were age, WBC, integrated molecular-cytogenetic risk, the NPM mutation, the FLT3 ITD or D835 mutation, and de novo vs secondary AML. The overall survival curves were plotted with the Kaplan-Meier method and compared by the log-rank test, not censoring patients at allotransplantation. Significance was defined as a p value of ≤0.05. Table 1. The study included 39 AML patients (18 males and 21 females, median age 56 years, range 19-65). According to cytogenetic-molecular risk stratification [31], 3 (8%), 26 (66%), and 10 (26%) patients were assigned to favorable, intermediate, and adverse prognosis groups, respectively. Molecular evaluation (i.e., NPM, FLT3, and CEBPA) was performed in all cases. NPM1 (A or B mutation) and FLT3 mutations (ITD or D835) were positive in 8 (20%) and 9 (23%) patients, respectively. There were no CEBPA positive cases. Median values of white blood cells (WBC) and of dBMA Tregs were 17000/μL and14/μL, respectively. 3.3. dBMA Cell Count. dBMA CD4 lymphocytes showed a correlation with dBMA Tregs (r = 0:7, p < 0:001, Figure 1).

Overall Survival (OS).
The whole group median OS was 18 months. The responder patient median OS was better than the nonresponder patient one (not reached vs 9 months, respectively, p = 0:002, log-rank test, Figure 7).

Discussion
Bearing in mind the complexity of the BM-AML microenvironment in which Tregs act, we have investigated their possible impact on response, adding a match-paired analysis of their absolute value differences before and after the antineoplastic treatment the study patients underwent.
It is well known that Tregs increase not only in the peripheral blood [20][21][22]26] but also in the BM, where they seem to be higher and also more immunosuppressive [22,26]. Confirming the dBMA Treg higher frequencies (10% of CD4 + T cells), our study population seemed suitable for studying Treg correlation with posttreatment outcome.
Interestingly, at the diagnostic phase, with regards to the BMA population other than the leukemic one, Williams et al. [32] did not report significant differences in the composition of the BMA Tregs by age, cytogenetic subgroup, or somatic myeloid-associated mutations. Accordingly, in our study population, dBMA Tregs neither seemed to correlate with clinical and molecular-cytogenetic risk category (Table 2) nor showed any ROC-threshold level predicting response ( Figure 6).

Journal of Immunology Research
On the other hand, the factors which have an impact on ORR are amply defined in an AML setting and have been confirmed in our analysis (Table 3). In particular, as expected, molecular and cytogenetic findings remain the major impacting variable both in univariate and in multivariate analysis, while no impact was demonstrated for dBMA Tregs. Therefore, our data might appear contradictory with the observations showing dBMA Tregs to be correlated with poor prognosis [22,26], apart from their impact if studied in peripheral blood [20-23, 25, 26]. However, in our study population, we have analyzed the BMA Treg absolute values, instead of BM Treg frequencies, with the aim of studying the cytoreductive effect of antineoplastic treatment on BMA-T, -NK, and -B populations and the possible correlation with outcome.
It has already been shown [24] that treatment-induced lymphopenia is not a random process, and susceptibility to intensive chemotherapy differs between T cell subsets (i.e., CD4, CD8, and Tregs) and/or NK and B cells. Therefore, we were interested in deciphering the differential effects of treatment on the BM microenvironment with regard to Tregs and cells other than Tregs. While the only population that decreases after treatment is the B one (Figure 3), Tregs did not seem to reduce their absolute values in the whole group. Thus, our results seem to suggest that CD3+ T cells are less sensitive to intensive chemotherapy than B lymphocytes. Moreover, the preferential negative impact on B lymphocytes by chemotherapy in AMLtreated patients obtaining CR has been already shown in studies that have investigated response to seasonal influenza Tre re e n nTre Tre T nTre T T Tre re nTre nTr nTr T nTre nTre T T Tre Tre nTre nT nT T T T nT nT n nT nTr nT T Tr re re e n nT T Tre Tr r r re r re e e e e e e n nT T T T T T Tr r r r re e e nTre Tre nT nT T T T Tr r r r r re e re nTre nTre Tre n nTr n n n nT nT T T nTr r r re e e e e n n nT nTr nT T nTre T T Tr r re e n n n nT T T T Tr T Tr r re r r re e T T T Tr r r re e e Tre e n n n nT nT T T T nT T T Tr r r r r r re nTre e e e n n nT Tr r re e e e e n n n n nT Tr re e eg g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g Effecto

CD4
Naive ing no g n g n no g no g n no g no g n n n no g g no n no n g no n n n n no nTre e e e n nTre T Tr T Tr re e nTre T Tr re T T T Tr T n n n r n n r re e e e e n re re e e e e eg g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g g Effecto  Journal of Immunology Research vaccination as a surrogate for the robustness of the immune system [33]. In the same studies, frequencies of their T cell populations were similar to those seen in healthy controls [33]. Interestingly, BMA Tregs significantly decreased after treatment only in responder patients (Figure 4), thus miming, in the human setting, the in vitro evidence that Tregs and other T lymphocyte removal from the microenvironment lead to augmented immune responses to AML [15] and suggesting how dBMA Treg decrease matters irrespective of the higher pretherapeutic diagnostic values. Also noteworthy is that this treatment-correlated Treg reduction has been recently reported [34] in the context of the midostaurin-treated [35] AML patients with FLT3   Journal of Immunology Research mutation [36]. Although performed on peripheral blood, the study reported how midostaurin treatment significantly reduced the regulatory T cell population, suggesting an "off-target effect of this multikinase inhibitor on T cell signaling pathways" [34] as our subanalysis, though limited by a small number of patients and performed on BM, seemed to suggest ( Figure 5). Moreover, it is known that AML cells secrete factors which inhibit T cell activation and proliferation [10,15] and may directly drive Treg expansion [37].

Conclusions
It is widely recognized [18,37,38] that BMA-Treg values are higher in, and act at, the neoplastic AML site (i.e., bone marrow) primarily favoring leukemia growth. Additionally, in our study, unlike cytogenetics and molecular AML findings, dBMA Tregs do not correlate with response after antineoplastic treatment, confirming their role in sustaining the disease without being a trigger. Nonetheless, it is well known that early lymphoid reconstitution after chemotherapy is associated with decreased risk of leukemia relapse [39], and these observations suggest that immunological events early after chemotherapy are clinically important too. In this context, the observed after-treatment BMA-Treg reduction in responder patients seems to suggest a new scenario. However, how the immunological status after treatment might be conditioned by preexisting diseaseinduced abnormalities or chemotherapy-induced defects needs to be further investigated.

Data Availability
The data used to support the findings of this study are included within the article.

Conflicts of Interest
There are no conflicts of interest to report.

Authors' Contributions
M. Delia wrote the paper and compiled and analyzed the data. A. Mestice performed flow cytometry. P. Carluccio, R. Frappampina, and F. Albano collected the data. M. Delia and G. Specchia designed the study. P. Musto revised and approved the paper.