Background
Acute myeloid leukemias (AML) encompass a heterogeneous group of oligoclonal hematological malignancies characterized by rapid, uncontrolled proliferation of leukemic blasts in the bone marrow and peripheral blood [
1]. Initial treatment for AML typically includes intensive cytotoxic chemotherapy intended to achieve a morphological complete remission (CR) in the bone marrow, but current standard of care regimens have poor clinical outcomes, as over half of AML patients will relapse and only a quarter will survive past 5 years [
2].
Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a consolidative immunotherapy strategy to prevent relapse but is itself associated with significant morbidity and mortality [
3]. Nevertheless, given the poor clinical outcomes with chemotherapy alone, the addition of other immunotherapeutic modalities to prevent relapse after completion of cytotoxic chemotherapy is an appealing prospect [
4‐
6]. Cancer vaccination against tumor-specific antigens is a logical immunotherapeutic strategy to prevent relapse by further priming an immune response, and induction of leukemia-associated antigen-specific CD8+ T-cells have been observed in response to peptide vaccination, though responses are often short-lived [
7‐
10]. Vaccination to retain remission in AML patients has produced conflicting results, with some groups reporting long-term efficacy and sustained immune response, while others report deletion of high-avidity CD8+ T-cell clones and unsustainable responses [
11‐
14].
Though immunotherapeutic strategies to maintain remission after patients receive intensive chemotherapy are logical and being tested in clinical trials [
6], there is an incomplete description of the state of the adaptive immune system in AML patients who have completed chemotherapy. Previous work on peripheral blood lymphocyte recovery in AML after induction demonstrated a skewing of the T-cell compartment towards peripherally expanded oligoclonal activated T-regulatory cells (T-regs) in the time immediately following chemotherapy [
15]. Yet, the functional capacity of the immune system in AML patients after the completion of intensive chemotherapy is largely unknown, and this has important implications for the success of any subsequent immunotherapy intended to prevent relapse, especially cancer vaccination and immune checkpoint blockade meant to augment endogenous cellular-mediated anti-leukemic responses.
The intent of the present study was to quantify phenotypic and functional immune abnormalities in a cohort of AML patients that would be considered ideal candidates for immunotherapies intended to prevent relapse. To eliminate the potential confounding effects of testing novel leukemia vaccines of unknown antigenic potency, we instead studied the well-characterized seasonal influenza vaccine with well-known efficacy profiles in healthy subjects. Recent studies have illuminated the presence of influenza vaccine-induced cell signatures that are predictive of robust immune responses across a spectrum of healthy individuals [
16,
17]. Influenza is a significant cause of morbidity and mortality in patients with hematological malignancies [
18,
19]. Because perturbations in the adaptive immune system after chemotherapy have severe implications for the success of any immunotherapy, especially vaccine-based modalities, we sought to interrogate the recovery and presence or absence of immune cell populations potentially predictive of response to vaccination in AML patients. We performed comprehensive phenotypic, genetic, and functional immune profiling immediately before and 30 days after the seasonal influenza vaccination in ten AML patients in a first remission after consolidation chemotherapy.
Discussion
While immune system reconstitution after chemotherapy is important from an infectious disease standpoint, understanding the kinetics and biology behind this phenomenon is also important in the context of anti-cancer immunotherapy to prevent relapse and maintain remission in patients. Our goal with this study was to perform a deep assessment of the state of the adaptive immune system in AML patients in a first clinical remission after consolidation chemotherapy, not only to further the body of evidence for the timing of vaccinations against infectious disease in these patients, but also to characterize patients who may be considered candidates for post-chemotherapy immunotherapy. To do this, we sought to explore patterns of immune reconstitution following consolidation chemotherapy in AML patients in remission and to look for indications of a functional response, or lack thereof, to vaccination. Our approach, when compared to previous studies, was unique in that we established a comprehensive picture of the state of the adaptive “immunome” by simultaneously examining the genetic, phenotypic, and functional consequences of chemotherapy in AML patients. Importantly, we show a dramatic impact in the B-cell compartment, which appears slower to recover than the T-cell compartment after AML chemotherapy. We further show that the inability of AML patients to produce protective antibody titers in response to influenza vaccination is likely due to multiple B-cell abnormalities in this cohort of patients, including increased frequencies of transitional B-cells, a lack of affinity-matured, class-switched B-cells, and an antigen-inexperienced B-cell repertoire. Of particular interest are the generally recovered T-cell frequencies and the cytokine-secretion functionality of CD8+ T-cells in response to influenza vaccination, suggesting that the two arms of the adaptive immune system are not equal in how they are affected by, or recover from, AML chemotherapy.
A recent study of post-vaccination responses to seasonal and pandemic influenza vaccination identified several immune subsets that were predictive of a robust and specific antibody response [
17]. Among these predictive subsets, several CD38+ B-cell populations (ID91, ID96, ID103, ID108) were very highly correlated with a strong antibody response to influenza vaccination. However, in our data set, we saw no differences in the frequencies of these CD38+ populations in AML patients compared to HD, nor did we see any difference in their frequencies between the AML-NR and AML-R. The lack of a relationship between these populations and response to vaccination in our AML cohort and HD indicate that applying a signature to predict post-vaccination responses is difficult in patients who receive chemotherapy, likely due to fluctuating immune parameters as immune reconstitution occurs. T-cell responses may be better surrogates of response to vaccination, especially in elderly individuals, but again, it has not been determined if this holds true in AML patients after they have received chemotherapy [
26].
Previous studies in solid tumors and in lymphoid hematological malignancies have reported a delayed B-cell recovery and function during and in the year following allo-HSCT, cytotoxic chemotherapy, or B-cell depleting agents [
27‐
29]. In a small cohort of pediatric AML patients either still receiving or having just completed treatment, it was shown that chemotherapy appeared especially deleterious for naïve and memory B-cells [
28]. However, it remained unknown whether these effects would be the same in adult AML patients specifically in a complete remission post-consolidation. Furthermore, while both cellular and humoral immune reconstitution has been characterized in the allo-HSCT setting [
29‐
35], it remained unclear what immune deficiencies could be ascribed to the consolidation chemotherapy patients received to achieve a CR before they receive conditioning therapy before transplant.
In our study, the total frequencies of CD19+ B-cells were similar between AML patients before vaccination and HD. The increased frequency of circulating immature, transitional B-cells suggest higher output by the bone marrow of these cells to repopulate the humoral immune niche [
29], and the near absence of memory B-cell subsets and class-switched B-cells suggests either an inherently increased susceptibility of both memory and effector B-cells to chemotherapy, an insufficient amount of time to fully recover these cell populations, or both. CD27+ memory B-cells, and especially resting memory B-cells, are responsible for generating antibody responses, and circulating influenza-specific memory B-cells can last for decades [
36‐
39]. The consistent finding of a reduction of memory B-cells in all the AML patients implicates this loss in the poor responses observed to influenza vaccination and suggests that humoral immunity reconstitution is a very long process.
The large number of differentially up-regulated genes in the 8 AML patients who did not respond to influenza vaccination compared to HD points towards a very active cell machinery across many cell pathways in various immune cell subsets. Furthermore, gene set enrichment analyses with very stringent criteria produced a very large number of up-regulated gene sets, underscoring the magnitude of perturbation in the intracellular state of PBMCs from the AML patients. The over-expression of genes related to cell death, metabolism, and genes more generally related to the immune response are strongly suggestive that the immune system is subject to lymphopoiesis and reconstitution in AML patients. Of particular interest is the upregulation of galectin-9 (LGALS9), which is the ligand for TIM-3 [
40]. LGALS9 can exert pleiotropic effects through TIM-3, and it has been shown to promote leukemic stem cell renewal, suggesting that this gene that appears to be up-regulated by the process of immune reconstitution after chemotherapy may also have a role in AML relapse [
41,
42].
Since the loss of memory B-cells is a central finding of our work, we sought to further interrogate the status of these cells through B-cell receptor IGH chain repertoire sequencing. The sequencing data shows that AML patients after chemotherapy have the same success rate at forming productive sequences as HD and can eventually form BCRs against specific antigen. Interestingly, the percentage of rearranged IGHV regions with point mutations indicative of somatic hypermutation in each patient tracked with the total frequency of memory B-cells, which were highest in frequency in patients furthest from last chemotherapy. Although we did not specifically look at B-cells involved in germinal center reactions as we did not sample peripheral lymphoid tissue in this study [
43], our findings show that the humoral immune system in AML patients after chemotherapy is still capable of generating B-cell memory.
Frequencies of various T-cell subpopulations were grossly normal in our cohort of AML patients. The relatively recovered and intact T-cell immune system in these patients and the ability of CD8+ T-cells to secrete IFNγ in an antigen-specific manner suggest that T-cell mediated immunotherapies are perhaps better suited at preventing and treating relapsed and refractory AML. We noted increased PD-1 expression on effector and memory T-cells shortly after the completion of chemotherapy in some of our AML patients. We believe there may be a window for immunotherapeutic intervention using immune checkpoint blockade to target residual disease in patients in remission. We do not fully address if the seemingly intact T-cell compartment in the blood of our patients were thymically derived or if surviving cells underwent peripheral expansion after chemotherapy. Recovery of CD4+ T-cells is thought to be generally thymus-independent, and one piece of corroborating evidence in our cohort lies in the higher frequency of memory T-regs and the reduced frequency of naïve T-regs in the AML patients, suggesting a proliferation of memory T-regs in the periphery as has been reported in the month immediately following chemotherapy [
15,
44]. However, the inverse has been reported in other hematological malignancies [
45,
46], which has been ascribed to a reduced susceptibility of naïve T-regs to oxidative stress [
47]. While T-regs are important in the maintenance of B-cell homeostasis and tolerance [
48] in mice models receiving influenza vaccination, CD4+ FOXP3+ T-regs had no effect on B-cell responses [
49]. However, this same study reported increased frequencies of T-regs in response to influenza vaccination, but we did not observe any appreciable increase in the frequencies of naïve or memory T-regs in our AML cohort after vaccination [
49]. Thus, our observations on the relative frequencies of T-reg subsets in AML patients after chemotherapy differ somewhat with what has been reported in the literature, and more work into the underlying mechanisms of T-reg reconstitution and responses in the period immediately following chemotherapy and its effects on the success of non-transplant immunotherapeutic strategies is warranted.
A limitation of our work is the exclusion of CD161 and ICOS in our T-cell phenotyping panels, as recent reports identified populations of CD4+ CD161+ T-cells and ICOS+ CD38+ T-follicular helper cells that clonally expanded and persisted years after influenza vaccination in AML patients receiving chemotherapy and in healthy individuals receiving successive influenza vaccinations, respectively [
50,
51]. Additionally, the flow cytometric panels used in this study did not include natural killer (NK) cells. As a previous report identified recovered NK cell frequencies in an AML patient in remission [
52], future planned studies will examine the frequencies of NK cells in AML patients in remission, as this cell population is an active area of exploration as an anti-leukemic immunotherapy [
53]. Additionally, we will compare T-cells in the marrow compartment with those sampled from the blood to examine whether there are phenotypic and functional differences between cells from these two related yet distinct sites of disease. Furthermore, one consideration we have not addressed is the possibility of AML minimal residual disease (MRD) under the threshold of detection by morphologic examination contributing to local immunosuppression in the tumor microenvironment and its effect on lymphocyte recovery. It has been shown that AML blasts can induce IL-10 and IL-17 production by Th17 cells [
54], and IL-10 can have differential effects on B-cells depending on their differentiation and activation status [
55]. The effect of AML MRD on B-cell recovery in particular remains to be determined, and studying Th17 cells as well in the context of AML remission is also warranted.
While our study was small and our cohort of AML patients was diverse in terms of disease etiology, clinical outcomes, and duration into their first remission, we strikingly found similar patterns of immune dysfunction across all the patients. Notably, when we ranked the patients based on time elapsed since chemotherapy and used cell frequency trends as surrogates for immune reconstitution over time, the degree of dysfunction was less in those patients that were further away from chemotherapy. Our better understanding of the changes in adaptive immune cell subsets after chemotherapy in AML patients will be useful in designing immunotherapies that can work with existing immune capacity to eliminate residual disease and maintain a remission in patients with AML.
Authors’ contributions
MG performed experiments, analyzed data, wrote the manuscript, and made the figures. GP enrolled and cared for patients on the protocol and provided comments on the manuscript. AB, SM, LK, BS, JC, RS, HZ, HG, JM, and LK performed experiments and provided comments on the manuscript. FC and YK analyzed data and provided comments on the manuscript. KN collected and collated all the patient samples and provided comments on the manuscript. BDS and IMB supervised the clinical protocol and sample acquisition, contributed to study design, and edited the manuscript. CSH designed and supervised the research, wrote the clinical protocol, established the collaborations, and edited the manuscript. All authors read and approved the final manuscript.