Eomes⁺T-bet^low CD8⁺ T Cells Are Functionally Impaired and Are Associated with Poor Clinical Outcome in Patients with Acute Myeloid Leukemia

Despite considerable effort, the prognosis of acute myeloid leukemia (AML) remains poor with the 5-year survival rate at only 25%. There is a clear need for improvement of leukemia therapeutics. Immunotherapy is a promising strategy for cancer treatment. Recent studies using reagents blocking negative immune regulatory pathways have achieved great success (1–5). These approaches target T-cell exhaustion, a state of T-cell dysfunction that develops in response to persistent antigen stimulation (6). Antagonist antibodies against programmed cell death protein 1 (PD-1), a major regulator of T-cell exhaustion, have been FDA approved for treating multiple solid tumors. Several studies, including ours, have demonstrated an involvement of PD-1 and other T-cell inhibitory pathways such as T-cell immunoglobulin and mucin domain 3 (TIM-3) and T-cell immunoglobulin and ITIM domain (TIGIT) in AML progression (7–13). These data provide a strong premise to develop effective AML treatment by suppressing inhibitory pathways to unleash patients' own antileukemia immune response. Clinical trials investigating the safety and efficacy of reagents inhibiting the PD-1 pathway in AML are under way (14, 15). To further optimize and successfully implement this approach, it is crucial to understand the molecular mechanisms that control the inhibitory pathway and antileukemia T-cell response in AML, which is currently unknown.

Eomesodermin (Eomes) and T-bet are both T-box transcription factors that play key roles in CD8 T-cell activation, differentiation, and memory development (16–20). Elegant studies in mouse models of infection have demonstrated a context-specific regulation of antiviral CD8 T-cell responses by these transcription factors. In acute viral infection, both T-bet and Eomes are upregulated and stimulate potent antiviral activity. Although T-bet primarily functions in terminal differentiated effector CD8 T cells, Eomes mainly controls the memory CD8 T-cell repertoire (16, 21, 22). In chronic viral infections, where T-cell exhaustion occurs, T-bet is involved in maintaining a subset of nonterminal progenitor cells among the exhausted CD8 T cells. In contrast, Eomes regulates the terminal exhausted CD8 T cells that are not able to be reinvigorated by PD-1 blockade (23). The phenomenon that T-bet and Eomes associate with different phenotype and function of exhausted CD8 T cells has been further confirmed in patients with HCV and HIV. In these patients, T-bet^hi T cells express an intermediate level of PD-1 and associate with resolution of infection. By contrast, Eomes^hi T cells display high expression of inhibitory receptors and severe functional defects, leading to the persistent stage of chronic infection (23, 24). Knowledge on the regulatory effect of Eomes and T-bet in cancer is limited. Studies in mouse models of melanoma suggest that both T-bet and Eomes are required for an effective antitumor T-cell response (25). However, a study of a mouse model of leukemia demonstrates that T-bet, not Eomes, is essential for leukemia-reactive CD8 T-cell function during immunotherapy (26). Importantly, elegant work from Galon and colleagues identified TBX21 and several T-bet target genes as part of an immune signature in human colorectal cancer that associates positively with patient survival (27). However, the clinical impact of these transcription factors in AML patients is unknown.

In this study, we determined the transcriptional mechanisms that govern CD8 T-cell response in AML and the impact on clinical outcome. We examined clinical samples from a large cohort of patients with AML (n = 59). Here, we report that the frequency of Eomes⁺T-bet^low CD8 T cells is increased in newly diagnosed AML. This cell subset is functionally impaired and knockdown of Eomes is able to restore T-cell function. In addition, Eomes binds to the promoter of TIGIT and positively regulates expression of this inhibitory receptor. Importantly, a high frequency of Eomes⁺T-bet^low CD8 T cells in AML patients at initial diagnosis associates with poor response to induction chemotherapy and shorter overall survival, suggesting a predictive and prognostic value in AML.

Peripheral blood samples collected from 59 patients with AML diagnosed per WHO criteria were used in this study. All patients were diagnosed at the Penn State Cancer Institute of Penn State University College of Medicine, Hershey, PA, between January 2014 and January 2017. The study was approved by the Institutional Review Board of Penn State University College of Medicine. Full written informed consent was obtained from all patients. Patients' clinical characteristics are summarized in Table 1. The median age was 64, gender distribution was largely even. Consistent with heterogeneity in AML, WBC counts and blast percentage had wide variability. Risk stratification based on cytogenetics was per Estey (28). The majority of enrolled patients carry intermediate or adverse cytogenetic features. Survival data were available for 57 of 59 patients (2 patients were lost to follow-up). Five patients opted to receive best supportive care (no induction chemotherapy). Among the 52 patients who received induction chemotherapy, 5 patients died before assessment of remission; therefore, response to induction chemotherapy was able to be analyzed on 47 patients.

Cells were stained with mAbs for 30 minutes at 4°C. These mAbs were anti-human CD3-BV786, CD4-BV711, CD8-APC-H7, PD-1-PE-Cy7, TIM-3-FITC (BD Biosciences), and TIGIT-APC (eBioscience). Samples were acquired using an LSRFortessa flow cytometer (BD Biosciences) in the Penn State College of Medicine Flow Cytometry Core Facility and data analyzed using FlowJo software (TreeStar). Thawed PBMCs were stimulated with anti-CD3/CD28 (eBioscience), plus GolgiPlug (BD Biosciences) for 5 hours. After surface staining, fixation, and permeabilization, intracellular staining was performed using anti-IFNγ-APC and TNFα-FITC (BD Biosciences). Staining of Granzyme B-AF700, Perforin-APC, Ki67-FITC, Eomes-eF610, and T-bet-PE (BD Biosciences) was performed without anti-CD3/CD28 stimulation.

Purified CD8 T cells were cocultured with T2 cells that were pulsed with 10 μmol/L WT1_126-134 or SV40 LT_281-289 peptide, in the presence of 50 IU IL2 (R&D Systems) for 6 days. IL2 was re-added on day 3. On day 6, CD8 T cells were restimulated with the peptides, plus GolgiPlug, for 5 hours and followed by flow-cytometric analysis.

Accell SMARTpool human EOMES and control siRNA were obtained from GE Dharmacon RNA Technologies. Control and specific siRNAs were applied at a final concentration of 1 μmol/L per well (96-well plate) for 72 to 96 hours. Cells were further stimulated with anti-CD3/CD28 antibodies for 5 hours for functional analysis, followed by flow cytometry.

293T cells were transfected with TIGIT promoter (−2,228/+70 bp), EOMES expressing plasmid (ORF of EOMES in pEnter, with C terminal Flag and His tag, Vigene) and pRL-TK. After 24 hours, luciferase assays were performed using a dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions.

Purified T cells were stimulated in vitro with anti-CD3 for 24 hours, followed by cross-linking, sonication, and chromatin immunoprecipitation (ChIP) with antibodies to Eomes or normal goat IgG (Abcam). The precipitated DNA was quantified by real-time PCR with SYBR green. Primer sequences: TIGIT-Promoter-F, CCCAGAGAGGGGAGAAAACTG; TIGIT-Promoter-R, GAAAATCTGCATCTGCACCCA.

GraphPad Prism was used for some statistical calculations. For data distributed normally, the comparison of variables was performed using unpaired or paired Student t test. For data not distributed normally, the comparison of variables was performed with a Mann–Whitney U test or a Wilcoxon signed-rank test for unpaired and paired data, respectively. Comparisons of categorical patient characteristics were analyzed using Fisher exact test. Pearson correlation coefficients were used to evaluate correlation. Kaplan–Meier curves for overall survival (OS) by the CD8 subgroup are provided to evaluate the survival difference with the P value obtained from log-rank tests. Also, a multivariable Cox proportional hazard (PH) model is conducted to further assess the association of CD8 subpopulations, with OS adjusted for potential confounding factors, where the PH assumption is checked based on graphical methods or statistical tests on Schoenfeld residuals. All tests are two-tailed, with P values less than 0.05 considered statistically significant.

Intracellular expression of Eomes and T-bet in CD8 T cells of peripheral blood from newly diagnosed AML patients was examined by flow cytometry. Based on the expression of Eomes versus T-bet, three subpopulations were defined among activated CD8 T cells: Eomes⁻ T-bet^hi (fraction I), Eomes⁺ T-bet^hi (fraction II), and Eomes⁺T-bet^low (fraction III) as shown in Fig. 1A. We observed a significant increase in the percentage of Eomes⁺T-bet^low CD8 T cells from AML patients (n = 59) compared with that of healthy controls (n = 22; P = 0.0297; Fig. 1B). In contrast, there was no significant difference in fractions I and II between AML and healthy controls. We also assessed blood samples from 11 AML patients at initial diagnosis compared with when the patient was in complete remission (CR). We found that the frequency of Eomes⁺T-bet^low CD8 T cells was significantly lower in CR (P = 0.0483; Fig. 1C). This finding suggests an association of Eomes⁺T-bet^low CD8 T cells with AML progression.

We then evaluated the clinical correlation of each CD8 subpopulation in AML. We first analyzed their impact in the response to induction chemotherapy. Based on the frequency of Eomes⁺T-bet^low (fraction III) in CD8 T cells, we defined high III% (Eomes⁺T-bet^low ≥ 20.1%) versus low III% (Eomes⁺T-bet^low < 20.1%) subgroups among the AML patients in this study. The median value of III% was used as the cutoff. Among the 47 AML patients of whom the response status was evaluable, we observed a significantly higher rate of primary refractory disease (failure to achieve CR) in the high III% subgroup compared with that of the low III% subgroup (7/22 vs. 1/25, P = 0.018; Fig. 2A). In contrast, when the same analysis was applied to Eomes⁻ T-bet^hi (fraction I) of CD8 T cells, we found an opposite trend (although not statistically significant) that primary refractory disease tended to occur more in the low I% subgroup. No difference was detected between high II% and low II% in the rate of primary refractory disease (Fig. 2A). We next examined the association of each CD8 subpopulation to OS. Strikingly, we observed a strong negative association of the Eomes⁺T-bet^low CD8 T cells to OS. With a median follow-up of 1,006 days, patients with high percentage of Eomes⁺T-bet^low CD8 T cells displayed significantly lower OS compared with that of low-percentage subgroup (median: 457 vs. 191 days; P = 0.021; Fig. 2B). No survival association was detected when fractions I and II were analyzed. In order to evaluate whether there are confounding factors contributing to the OS, we compared high versus low III% subgroup of patients for crucial clinical factors including age, gender, WBC count, blast count, cytogenetic features, and treatment modalities. No statistically significant difference was observed in any of the factors (Supplementary Table S1). A multivariable Cox PH regression analysis confirmed the significant survival difference between high versus low III % subgroup (P = 0.011; Supplementary Table S2) upon adjustment for potential confounding factors. Taken together, these data demonstrate that Eomes⁺T-bet^low CD8 T cells associate with poor clinical outcome in AML.

To assess the functional status of Eomes⁺T-bet^low CD8 T cells from AML patients, we performed functional assays to test T-cell cytokine release upon in vitro stimulation. Eomes⁺T-bet^low (fraction III) CD8 T cells had significantly lower production of intracellular IFNγ and TNFα compared with the fraction I and II CD8 T cells (Fig. 3A and B). We also evaluated the cytotoxic potential by examining the level of granzyme B and perforin expression. Consistently, we observed significantly lower expression of both granzyme B and perforin in Eomes⁺T-bet^low CD8 T cells (Fig. 3C and D). To further dissect the function of leukemia-reactive Eomes⁺T-bet^low CD8 T cells, we tested CD8 T cells for cytokine release in response to a WT-1 peptide. WT-1 is a well-known tumor-associated antigen, in which HLA-A*0201–restricted WT-1_126-134 is the most studied epitope in AML (29, 30). Purified CD8 T cells derived from peripheral blood of newly diagnosed HLA-A*0201 AML patients were cocultured in vitro with T2 cells (used as antigen-presenting cells) that were pulsed with HLA-A*0201–binding WT-1_126-134 or SV40 LT_281-289 (as negative control) peptide. Intracellular cytokine production by CD8 T cells was detected (1.2% for IFNγ and 0.38% for TNFα) after 6 days of stimulation with WT-1 peptide, while essentially no CD8 T cells responded to control peptide (Fig. 3E). Using this experimental system, we compared the leukemia-reactive response of Eomes⁺T-bet^low CD8 T cells with that of Eomes⁻T-bet^hi CD8 T cells. We found significantly lower production of both IFNγ and TNFα by Eomes⁺T-bet^low CD8 T cells in response to WT-1 stimulation (Fig. 3F). Collectively, these results demonstrate that both bulk and leukemia-reactive Eomes⁺T-bet^low CD8 T cells from AML patients exhibit decreased function.

To evaluate whether the expression of Eomes contributes to the functional defect of CD8 T cells in AML patients, we used a specific siRNA to knockdown Eomes expression in CD8 T cells from AML patients (Fig. 4A). Intracellular cytokine staining assays were performed on CD8 T cells that had Eomes knocked down. We observed a significant increase of IFNγ and TNFα production after Eomes knockdown (Fig. 4B and C). Of note, cytokine production was assessed here 4 days post in vitro stimulation in serum-free culture condition as this is required in the knockdown experiment. With this culture condition, low level of IFNγ and TNFα was detected in response to anti-CD3/CD28 (control siRNA). However, significantly higher amount of cytokines was detected when Eomes was inhibited (Eomes siRNA). Consistently, granzyme B expression was increased in CD8 T cells upon Eomes knockdown (Fig. 4D), indicating an improved cytotoxic capacity. These data demonstrate a pivotal role for Eomes in suppressing T-cell function in AML patients.

Eomes is a crucial transcription factor facilitating CD8 T-cell exhaustion in chronic viral infection (23, 24). We hypothesize that the suppressive effect of Eomes on CD8 T cells in AML is through its transcriptional regulation of exhaustion-related inhibitory receptors. To examine whether these inhibitory receptors associate with the expression of Eomes in AML, multicolor flow cytometry analysis was performed to detect the expression of inhibitory receptors on each CD8 T-cell subpopulation (defined by different pattern of intracellular expression of Eomes and T-bet). As shown in Fig. 5A, although PD-1 expression was higher in fraction III (Eomes⁺T-bet^low) than that in fraction II (Eomes⁺T-bet^hi) and I (Eomes⁻T-bet^hi), there was no significant correlation between the expression of PD-1 and frequency of each CD8 T-cell subpopulation. TIM-3 followed the same pattern with no correlation detected (Fig. 5B). Strikingly, the expression of TIGIT was significantly higher in fraction II and III CD8 T cells compared with that in fraction I. Importantly, we observed a significant positive correlation between the frequency of TIGIT and that of fraction II or fraction III. There was no correlation between TIGIT and fraction I (Fig. 5C). Both fractions II and III are Eomes⁺ cells; therefore, these results demonstrated that expression of Eomes is associated with the upregulation of TIGIT on CD8 T cells from AML.

To further define the involvement of Eomes in the transcriptional regulation of TIGIT, we determined whether there is direct binding of Eomes to the promoter of TIGIT. We analyzed the promoter sequence and found one binding site for Eomes (Fig. 6A). In a ChIP assay using T cells purified from PBMCs of a healthy volunteer, we observed a clear interaction, in both CD4 and CD8 T cells, between Eomes and its binding site on the TIGIT promoter (Fig. 6B). We then performed a luciferase assay using TIGIT luciferase reporter. Eomes had significant transcriptional activity for the expression of TIGIT (Fig. 6C). To determine whether Eomes regulates CD8 T-cell expression of TIGIT in AML, we assessed the effect of Eomes knockdown on TIGIT expression using CD8 T cells derived from AML patients. We observed a significant decrease of TIGIT on CD8 T cells upon Eomes knockdown (Fig. 6D). These data demonstrate that the transcription factor Eomes can bind to the promoters of TIGIT and positively regulate the expression of this inhibitory receptor.

Similar to chronic infection, persistent antigens in the tumor environment can promote T-cell exhaustion. Reagents blocking inhibitory mechanisms involved in T-cell exhaustion, such as the PD-1 pathway, have proven to be safe and effective in treating multiple types of tumors. However, little is known about the transcriptional regulation of T-cell exhaustion in cancer. In the present study, we demonstrate that in newly diagnosed AML patients, Eomes⁺ T-bet^low CD8 T cells are elevated and display impaired function. Importantly, Eomes knockdown by siRNA reverses these functional defects. Results of our study suggest a reciprocal effect of Eomes and T-bet in patients with newly diagnosed AML. In the context of this disease, low T-bet and high Eomes expression associates with poor immune response in both bulk and leukemia-reactive CD8 T cells, which may contribute to the progression of leukemia. Our data are consistent with the finding in colon cancer patients that low T-bet expression in tumor-infiltrating CD8 T cells correlates with increased relapse rate and worse clinical outcome (31). However, different conclusions were made from studies using mouse tumor models. In a mouse model of melanoma (B16), it has been demonstrated that both T-bet and Eomes promote antitumor CD8 T-cell responses in that T-bet/Eomes-deficient tumor-infiltrating CD8 T cells produce more IFNγ (25). In addition, a recent report showed that constitutive expression of Eomes in tumor-specific CD8 T cells improved tumor rejection in a mouse lymphoma model system (EG7; ref. 32). These findings support a positive immune regulatory role of Eomes in cancer, which is in contrast to our observation that Eomes suppresses CD8 T-cell response in AML patients. Both B16 and EG7 are fast-growing mouse tumor cell lines, so that the observed antitumor CD8 T-cell responses in these mouse model systems likely represent an acute phase immune response, where both T-bet and Eomes are involved in promoting an effective CD8 T-cell response. In contrast, cancer patients may have undergone a relatively indolent disease course before presentation at diagnosis, a setting that more closely mimics chronic infection, where T-cell exhaustion occurs and T-bet and Eomes reciprocally regulate the immune response. In line with this idea, we found that knockdown of Eomes in CD8 T cells collected from AML patients increases cytokine production and cytotoxic capacity, which is opposed to a previous report that Eomes knockdown in healthy donor-derived CD8 T cells decreased IFNγ and perforin expression (33). The different functional status of CD8 T cells in AML patients versus healthy donors likely explains this discrepancy. It is worth noting that recent studies have shown that PD-1⁺CXCR5⁺TCF-1⁺ CD8 T cells are more functional than PD-1⁺CXCR5⁻TCF-1⁻ terminally exhausted T cells (34). We assessed CXCR5 and TCF-1 expression among each subpopulation of T cells based on the expression of Eomes and T-bet in AML. We observed that the frequency of cells expressing CXCR5 and TCF1, single or double positive, is significantly higher in Eomes⁺T-bet^low (fraction III) CD8 T cells compared with other subpopulations (Supplementary Fig. S1). This finding is consistent with the study of Im and colleagues that CXCR5⁺ cells expressed more Eomes (34). However, it is somewhat paradoxical as CXCR5⁺TCF-1⁺ CD8 T cells are more functional in the studies of Im and colleagues, whereas we observed less functional (more exhausted) status in Eomes⁺ cells. The discrepancy is likely attributed to the different model systems and disease status (mouse model of chronic viral infection vs. highly heterogeneous human AML) in the studies. These observations highlight the context-specific character for the transcriptional control of CD8 T cells.

Our finding that Eomes inhibits CD8 T-cell function in AML provides a strong rationale for targeting this transcription factor for effective leukemia therapeutics. Targeting transcription factors for drug development has been considered challenging as most transcription factors do not contain small-molecule-targetable pockets, making it difficult for developing small-molecule inhibitors. However, significant progress has been achieved to overcome these challenges (35). Strategies including inhibition of protein/protein interactions (36), manipulating the transcription factor DNA binding domain (37, 38), targeting chromatin remodeling or epigenetic reader proteins (39, 40), and blocking protein/DNA binding (41, 42) have led to the successful development of transcription factor-targeting drugs. Therefore, it is highly feasible to regain CD8 T-cell activity by directly suppressing Eomes in AML. Furthermore, our better understanding of the mechanisms involved in the regulation of Eomes makes it possible to develop novel efficient approaches targeting this transcription factor. It has been demonstrated that IL12 increases T-bet while suppressing Eomes in CD8 T cells during infections (43). Recent studies also identified microRNA-29b as a strong negative regulator for both T-bet and Eomes (44). These observations suggest a strong therapeutic potential of IL12 and microRNA-29b in boosting CD8 T-cell response by inhibiting Eomes expression in AML.

We made the important observation that AML patients with a high frequency of Eomes⁺ T-bet^low CD8 T cells at initial diagnosis respond poorly to induction chemotherapy and have a significantly lower OS. Of note, we compared the percentages of Eomes⁺ T-bet^low CD8 T cells versus OS (OS for death subjects, but observed time for subjects who are alive at the time of evaluation). We observed more subjects that survived longer in the low% group and a trend of negative correlation (Supplementary Fig. S2). In the Kaplan–Meier analysis, we appreciated significantly lower OS in patients with high frequency of Eomes⁺ T-bet^low CD8 T cells (Fig. 2B). Importantly, the statistical significance was confirmed upon adjustment of confounding factors by multivariable Cox PH regression analysis (Supplementary Table S2). This discovery is highly clinically relevant as it suggests a vital role for Eomes⁺ T-bet^low T cells as both a predictive and prognostic biomarker in the clinical course of AML. The majority of patients in our study received intensive induction with daunorubicin and cytarabine (Table 1), which has been the standard first-line treatment for the last 4 decades (45). However, this regimen carries a high risk of frequent complications, some of which are severe and life-threatening. The treatment-related mortality is around 10% to 15% (46). For elderly patients or patients with significant comorbidities, it is often difficult to make decisions between taking the high risk-intensive treatment versus approaches that are less aggressive albeit with lower response potential. The median age at diagnosis for AML is 68; thus, this is a frequently encountered dilemma clinically. Identification of predictive biomarkers would make it possible to select patients who are likely to respond to treatment and offer the poor responders alternative therapeutic modalities. Our study revealed a tight correlation of high frequency of Eomes⁺ T-bet^low CD8 T cells at initial diagnosis to poor clinical outcome, suggesting a predictive value of this cell subset in AML. Importantly, our data demonstrated that Eomes⁺ T-bet^low CD8 T cells are functionally impaired. We suggest that AML patients with high frequent Eomes⁺ T-bet^low CD8 T cells display an immune dysfunctional status, leading to the failure of antileukemia immunity and subsequently poor treatment response and OS. Therefore, adding immunotherapy to improve antileukemia immune response may benefit this patient population and improve clinical outcome.

In contrast to the findings in chronic viral infection that high expression of Eomes is associated with elevated levels of PD-1 and TIM-3 (23, 24), we did not observe this correlation in our AML study. However, we observed a positive correlation of Eomes expression with that of TIGIT. PD-1, TIM-3, and TIGIT are all reported to be involved in immune suppression in AML (7–13). Although all these negative T-cell receptors are upregulated in patients during leukemia relapse (10, 11, 13), significant elevation at newly diagnosed AML was observed only for TIGIT (11, 12). These findings suggest a disease status–specific usage of inhibitory immune pathways by tumor for immune evasion. Therefore, the TIGIT pathway may be predominant in the immune inhibition at newly diagnosed AML, whereas multiple suppressive pathways contribute to leukemia relapse. Our novel finding of the distinct correlation between Eomes and TIGIT in patients with newly diagnosed AML confirmed the dominant role of the TIGIT pathway in this disease setting and suggested that Eomes might transcriptionally regulate TIGIT expression, a hypothesis that is supported by our results identifying direct binding of Eomes to the TIGIT and the transcriptional activity. Taken together, our studies demonstrate that Eomes is a transcription factor that positively regulates TIGIT. In line with the emerging data demonstrating the inhibitory role of TIGIT in tumor immunity (47–49), this crucial finding suggests that upregulation of TIGIT likely contributes to the mechanisms by which Eomes suppresses T-cell function in AML.

It is important to evaluate which CD8 T-cell subpopulations contain the majority of tumor reactive T cells. We have made considerable effort in the study of MHC tetramer staining to address this question. However, commercially available tetramers for WT-1 that restricted to HLA-A2 were not reliably specific in our experiments. These also bind to samples from non–HLA-A2 patients. Nevertheless, we were able to develop an in vitro assay and the tumor reactive T-cell response was largely addressed when lower IFNγ production by Eomes⁺T-bet^low CD8 T cells was observed upon stimulation with WT-1_126-134, a leukemia-associated antigen peptide (Fig. 3E and F). This important finding suggests less functional status of leukemia-reactive Eomes⁺T-bet^low CD8 T cells. In addition, in line with the recent report of CD39 expression as a marker for chronically stimulated tumor-specific T cells (50), we examined CD39 expression in the samples of our AML patients and observed that the frequency of CD39⁺ cells are significantly increased in Eomes⁺T-bet^low CD8 T cells (Supplementary Fig. S3), indicating that this T-cell subpopulation contains more tumor reactive T cells.

In conclusion, we demonstrate an inhibitory effect of Eomes on T-cell response in AML patients. Eomes may do so by transcriptionally upregulating the TIGIT inhibitory receptor. Importantly, high frequency of Eomes⁺T-bet^low CD8 T cells associates with poor response to induction chemotherapy and low OS. Our findings could have a significant clinical impact as they not only identify a predictive and prognostic biomarker for the clinical course of AML, but also provide an important target for the development of effective leukemic therapeutics.

W.B. Rybka is consultant/advisory board member for Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B. Jia, H. Zheng

Development of methodology: B. Jia, C. Zhao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Jia, C. Zhao, K.L. Rakszawski, D.F. Claxton, W.C. Ehmann, W.B. Rybka, S. Mineishi, M.G. Bayerl, R.J. Hohl, H. Zheng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Jia, C. Zhao, K.L. Rakszawski, S. Mineishi, M. Wang, M.G. Bayerl, H. Zheng

Writing, review, and/or revision of the manuscript: B. Jia, C. Zhao, K.L. Rakszawski, D.F. Claxton, W.B. Rybka, S. Mineishi, M. Wang, J.M. Sivik, T.D. Schell, J.J. Drabick, R.J. Hohl, H. Zheng

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Shike, M.G. Bayerl, J.M. Sivik

Study supervision: J.J. Drabick, R.J. Hohl, H. Zheng

We thank all our patients for their trust, understanding, and willingness to provide their blood samples for our research. This work was supported by the American Society of Hematology Scholar Award Grant (#191839 to H. Zheng), Penn State Cancer Institute Funds, and the Kiesendahl Endowment funding.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.