Invariant phenotype and molecular association of biallelic TET2 mutant myeloid neoplasia

Increasingly, genomic data are being used to classify myeloid neoplasia (MN). Examples include BCR/ABL in chronic myeloid leukemia (CML)¹ ; t(8;21), inv(16), t(15;17), or MLL in acute myeloid leukemia (AML)^2-4 ; PDGFRA/B translocations in chronic myelomonocytic leukemia (CMML), and, in hereditary cases, germline mutations in CEBPA,⁵ RUNX1,^6,7 ETV6,^8,9 DDX41,¹⁰ GATA2,¹¹ TP53,¹²  etc. Such genetic alterations are beginning to supersede the use of morphologies in diagnoses, particularly when the pathomorphologies are less pronounced. For instance, ring sideroblasts are linked to SF3B1 mutations^13,14  or refractory anemia with ring sideroblasts and thrombocytosis with a combination of SF3B1 with either CALR, JAK2, or cMPL mutations.^13,15,16 

Some mutations are common and thus unlikely to be molecular markers of specific morphologic subentities. Examples include mutations in TET2, ASXL1, and DNMT3A.^17,18 TET2 is located on the long arm of chromosome 4 (4q24), a region susceptible to microdeletions, copy-neutral losses of heterozygosity, and rare translocations that also result in protein loss of function, producing TET2. TET2 is an Fe²⁺-dependent dioxygenase that converts 5-methylcytosine to 5-hydroxymethylcytosine (5hmC) to derepress silenced genes. TET2 mutations (TET2^MT) impair TET2’s ability to carry out this reaction. This decreases 5hmC DNA levels, which in turn skews differentiation toward monocytic progenitors. Indeed, engineered tet2^−/− mice have earlier onsets of myeloproliferative neoplasm (MPN) disease than tet2^+/− mice.^19,20  The role and function of TET2 has been studied in normal and malignant hematopoiesis.^21-24  The contribution of TET2^MT to clinicohematological features has, however, been controversial, possibly due to small-scale studies and combinatorial diversity of cooccurring lesions. Large studies accounting for clonal architecture and association between molecular and clinical features will be helpful to clarify the consequences of TET2^MT on disease phenotypes.

Here, we report the clinical course of patients with biallelic TET2 inactivation (biTET2ⁱ) in the context of MN. Compared with corresponding monoallelic mutations, these events might be associated with gene-dose–dependent greater intensity phenotypes and clinical outcomes and perhaps exaggerated morphologic features associated with increased risk to leukemia progression. We have comprehensively dissected the clonal nature of TET2^MT in 4930 patients with MN, of whom 40% harbored biTET2^MT.²⁵  We thus investigated possible associations between such abnormalities and clinical features and outcomes. We provide evidence supportive of the notion that biTET2ⁱ cases belong to a qualitatively distinct morphologic subentity of MN.

Peripheral blood and bone marrow (BM) samples from patients with MN were collected after receiving written informed consent according to the protocols approved by the Institutional Review Board of Cleveland Clinic in accordance with the Declaration of Helsinki. A total of 1045 patients were initially screened and enrolled in this study. Clinical parameters (age, sex, peripheral blood, BM counts, diagnosis, and overall survival) were obtained from medical records. Diagnosis was assigned based on the 2008 World Health Organization (WHO) classification criteria.²⁶  Genomic and germline DNA obtained from CD3⁺ lymphocytes was subjected to molecular screening for the coding regions of TET2 and other gene mutations. Samples that yielded low sequencing quality due to low depth were excluded from the study. Cases in which no TET2^MT was found by gene sequencing were investigated for possible deletions and microdeletions at chromosome 4q/24 by reviewing metaphase cytogenetics and results of single-nucleotide polymorphism (SNP) array, respectively. The ones with 4q deletion/microdeletion involving TET2 locus were considered monoallelic TET2^MT (monoTET2^MT), while those with absent aberrations were considered wild-type (TET2^WT) (supplemental Figure 1A). All TET2^MT were also screened for uniparental disomy (UPD) at chromosome 4q/24 by the SNP-array (SNP-A) method.

Whole-exome sequencing libraries were prepared according to the Nextera Rapid Capture Exome protocol (Illumina, San Diego, CA) and subjected to massive parallel sequencing using HiSeq 2000. Average coverage of samples subjected to whole-exome sequencing and targeted deep sequencing was 115× and 250×, respectively. Variants with a variant allele frequency (VAF) >5% were included. Multiamplicon targeted deep sequencing included a panel of 36 genes commonly mutated in myelodysplastic syndrome (MDS)^18,27,28  and other myeloid malignancies (supplemental Table 1). Paired-end libraries were subjected to deep sequencing on MiSeq sequencers according to Illumina protocols. Variants were extracted using the GATK3.3 pipeline. TET2^MT were called somatic when absent or at very low frequencies in germline CD3⁺ lymphocytes. Alterations found in both the myeloid and lymphoid cells with an equal VAF were considered germline and excluded from the study. Previously, usage of T cells as germline^29,30  resulted in similar frequencies of TET2^MT compared with skin or buccal swab specimens.^18,31  For original data, please contact H.A. (awadah@ccf.org).

SNP-A karyotyping for confirming metaphase cytogenetics and detecting copy-number–neutral loss of heterozygosity was performed as previously described.^32,33  Briefly, Affymetrix 250K and 6.0 SNP-As were used to evaluate cop-number alterations and copy-number natural loss of heterozygosity. Using our internal database and a publicly available database (http://dgv.tcag.ca/dgv/app/home), the screening algorithm validated each lesion as somatic. Nonsomatic lesions were excluded from further analysis. Affected genomic positions in each lesion were visualized and extracted using CNAG (v3.0) or Genotyping Console (Affymetrix, Santa Clara, CA).^34,35  Metaphase cytogenetic requires cellular proliferation, and its sensitivity and resolution depend on the proportion of clonal cells in the sample and size of the lesion, respectively. SNP-A does not depend on the presence of dividing cells and is able to detect copy-number variations and UPD with a high resolution. For this purpose, we included this method to further investigate for 4q/24 cryptic chromosomal lesions not identified by metaphase cytogenetic in our cohort.³⁶ 

Metaphase cytogenetics was performed on BM aspirates. The median number of metaphases analyzed was 20. Chromosomal preparation was performed on G-banded metaphase cells using standard techniques, and karyotypes were described in 86% (862/1001) of patients according to the International System for Human Cytogenetic Nomenclature.³⁷ 

VAFs were used to categorize first-hit TET2^MT into ancestral (dominant or codominant) mutations vs subclonal secondary mutations. A mutation with the highest VAF that is at least 5% more than the second highest VAF in each sample was defined as an ancestral/dominant mutation; those with <5% difference from the highest VAF were defined as ancestral/codominant, while VAFs with a >5% difference from the highest VAF were considered subclonal/secondary mutations (supplemental Figure 2A-F). Mutations in selected genes (commonly mutated in myeloid neoplasms) were assessed for differences in biTET2ⁱ vs biTET2^- cases.

Fisher’s exact test was used to compare proportions. All P values were 2 sided; those <.05 were considered statistically significant. Kaplan-Meier methods were used to plot survival probabilities, and log-rank tests were used to compare such curves. Univariate and multivariate Cox model analyses were also performed. All statistical computations were performed using R 3.5.1 (www.r-project.org).

We analyzed configurations of TET2^MT using VAFs (see Materials and methods) in 1045 patients with MN. Patients with TET2^MT (n = 200) were classified into heterozygous biallelic TET2^MT (biTET2^MT; ≥2 TET2^MT with VAF sum of >55%; n = 66), biclonal TET2^MT (≥2 TET2^MT with VAF sum of <45%; n = 11), undetermined (either biallelic or biclonal TET2^MT, as their VAF sum lays between 45% and 55%; n = 33), hemizygous TET2^MT (single TET2^MT with VAF >55% in the presence TET2 locus alteration on chromosome 4q24; n = 13), homozygous TET2^MT (homozygous mutation at 4q24 with UPD detected by SNP analysis; n = 3), and monoTET2^MT (single TET2^MT with VAF <45% and normal cytogenetics; n = 74; Figure 1A). Biclonal TET2^MT and undetermined cases were ambivalent and thus filtered out of the study (n = 44), resulting in a cohort of 1001 MN patients. BiTET2^MT, hemizygous TET2^MT, and homozygous TET2^MT have all inactivation (impairment) of both parental copies of TET2 and therefore were grouped in as biTET2ⁱ cases (n = 82) (Figure 1D). The remainder of the population, with either monoTET2^MT (n = 96; 74 cases with single TET2^MT and normal metaphase cytogenetic/SNP-A screening; 22 cases with 4q/TET2 locus deletion in absence of TET2^MT) or wild-type configuration (TET2^WT, n = 823; normal metaphase cytogenetic/SNP-A screening and absent TET2^MT), were considered negative for biTET2ⁱ (biTET2⁻; n = 919). Patients were also divided into those with CMML and without CMML (−) according to the presence of WHO-defined CMML hallmark clinical features.¹  The CMML (−) cohort (n = 885) was then subgrouped according to the presence of monocytosis. Monocytosis was present in 56% (n = 497) of these cases. TET2^MT configuration and respective number of patients is summarized in (supplemental Figure 1A).

Somatic TET2^MT was found in 156 out of 1001 of cases (16%), of which 53% were biTET2ⁱ. A total of 83% of biTET2ⁱ cases were truncating (frameshift deletion/insertion and nonsense), while 27% of somatic alterations were missense (Figure 1B). A comparison of the VAF ratio of the first and second TET2^MT per type of mutation and succession in biTET2^MT is presented in supplemental Figure 1D. The majority of monoTET2^MT patients (77%) harbored a TET2^MT with a VAF <45%; the minority had 4q/24 aberrations (locus deletion) (23%) (supplemental Figure 1B). MonoTET2^MT cases included 65% truncating and 35% missense mutations (supplemental Figure 1C). Truncating mutations were significantly more common in biTET2ⁱ cases than monoTET2^MT cases (odds ratio [OR], 2.7; P = .02; Figure 1B).

Clinical analysis revealed that biTET2ⁱ, in comparison with biTET2⁻, was associated with older age (91% ≥60 years vs 74%, P = .0004; Table 1). Among biTET2ⁱ cases, 27% were classified as MDS, 50% as MDS/MPN, and 23% as AML (10% pAML and 13% sAML). BiTET2ⁱ was enriched in patients with CMML1/2 (44%; P < .0001), predominantly in lower-risk cases (62% vs 47% in biTET2⁻; P = .003) and more commonly had normal metaphase cytogenetics (65%; P = .0007; Figure 1C). We also assessed phenotype/genotype association of biTET2ⁱ (Table 2). In biTET2⁻ cases, leukopenia (81%; P < .0001), neutropenia (52%; P = .008), pancytopenia (27%; P = .008), and increased marrow blast percentages (blasts ≥5% in 33%; P = .01) were more prevalent than in biTET2ⁱ cases, which in return cosegregated with monocytosis (84%; P < .0001; Table 2; Figure 1E), marrow hypercellularity (cellularity ≥70% in 67%; P < .0001), and more pronounced myeloid dysplasia (68%; P = .0003).

We next compared biTET2ⁱ cases to those with monoTET2^MT to evaluate the consequence of a second TET2 inactivation on disease features (supplemental Tables 2 and 3). Biallelic inactivation of TET2 was more likely to occur with MDS/MPN (P = .001), particularly the CMML1/2 subtype (P < .0001; Figure 1C,E), and with normal cytogenetics (P = .003). In addition, it was correlated with a lower odds of leukopenia (P = .002) and ring sideroblasts (≥15%; P = .02) and a higher likelihood of monocytosis (P = .003) and marrow hypercellularity (P = .02) (Figure 1E).

Because we observed a highly significant (P < .0001) relationship between biTET2ⁱ and CMML diagnosis and/or monocytosis, we focused on patients without obvious CMML (monocytosis; absence of BCR/ABL1, PDGFRA/B, or 11q23; presence of <20% blasts; and myeloid dysplasia) and compared biTET2ⁱ and biTET2⁻ for the association with monocytosis and myeloid dysplasia (supplemental Figure 1E). Increased monocyte counts among CMML (−) was also significantly overrepresented in biTET2ⁱ cases (72%; P = .03) compared with biTET2⁻ cases (55%), as was myeloid dysplasia (72% vs 46%; P = .0001).

The rank of TET2^MT within the clonal hierarchy can be determined according to VAF methodology (see Materials and methods). We first defined each TET2^MT as ancestral vs secondary and then identified other hits in relation to TET2 status. Due to resolution limits of the VAF approach and the difficulty of distinguishing subclonal from ancestral hits, we applied an arbitrary cutoff of 5% between VAFs to discriminate ancestral first hits (dominant and codominant) from subsequent “secondary” hits (supplemental Figure 2A-F). A summary of clonal hierarchy of somatic mutations, cytogenetic findings, and diagnoses in all biTET2ⁱ cases is presented in Figure 2A. Seventy-two percent of first TET2 hits in biTET2ⁱ were founder (dominant/codominant) lesions (P < .0001), while only 28% were secondary to another antecedent somatic mutations (Figure 2B). In monoTET2^MT cases, only 38% TET2 hits were dominant. In biTET2ⁱ, when the first TET2^MT was subclonal, the preceding founder clone was most likely characterized by BCOR/BCORL1^MT, PRC2-family^MT, ZRSR2^MT, ASXL1^MT, and others (Figure 2C). When the first TET2 hit was ancestral, the most common secondary mutation affected TET2, followed in frequency by SRSF2^MT, ASXL1^MT, RAS^MT, and DNMT3A^MT (Figure 2D-E).

When we investigated associations between concurrent mutations and TET2^MT configuration, ASXL1^MT (25%), TP53^MT (16%), and CBL^MT (7%) were more frequent in monoTET2^MT cases; SRSF2^MT (33%), KRAS/NRAS^MT (16%), RUNX1^MT (16%), and ZRSR2^MT (6%) were more frequent in biTET2ⁱ cases; and DNMT3A^MT (13%) and U2AF1^MT (9%) were more common in the TET2^WT population (Figure 2F). When compared with patients with and without biTET2ⁱ (Figure 3A), a significant co-occurrence with SRSF2^MT (P < .0001) and KRAS/NRAS^MT (P = .03) in biTET2ⁱ and TP53^MT (P = .03) in biTET2⁻ was noted. SRSF2^MT was also found to be significantly associated with biTET2ⁱ when compared with monoTET2^MT (P = .02) (supplemental Figure 3). In contrast, in biTET2ⁱ without monocytosis (16%; n = 13), SRSF2^MT was absent and KRAS/NRAS^MT was only detected in 8% (n = 1) of the cases.

In CMML, among biTET2ⁱ cases, SRSF2^MT was the most commonly found co-occurring lesion (53%; P = .005), followed by KRAS/NRAS^MT (28%), ASXL1^MT (28%), and RUNX1^MT (22%). Investigation for the incidence of a secondary/subclonal ASXL1^MT among CMML with preexisting biTET2ⁱ (20%) vs CMML with monoTET2^MT (72%) tended to be significant (P = .05). In biTET2⁻ cases, ASXL1^MT (31%), SRSF2^MT (25%), KRAS/NRAS^MT (23%), and RUNX1^MT (18%) were seen in high frequencies. In MDS/MPN (excluding CMML), SRSF2^MT was present in 40% and TP53^MT in 20% of biTET2ⁱ cases, respectively, while DNMT3A^MT occurred in 14% and ASXL1^MT in 11% of biTET2⁻ cases. Similar to what we observed in CMML, patients with MDS/MPN carrying biTET2ⁱ showed a striking concordance with SRSF2^MT (P = .05). In MDS/sAML, ASXL1^MT (24%) and SRSF2^MT (12%) were most commonly found together with biTET2ⁱ, while ASXL1^MT (13%), DNMT3A^MT (12%), and TP53^MT (12%) clustered with biTET2⁻ cases. Finally, in pAML, DNMT3A^MT (63%), SRSF2^MT (25%), and RUNX1^MT (25%) were the most frequent molecular events in biTET2ⁱ cases, while DNMT3A^MT (22%) and KRAS/NRAS^MT (18%) were seen in biTET2⁻ cases (Figure 3B).

We then grouped biTET2ⁱ cases into those with and without diagnosis of CMML (supplemental Figure 4A) and found that biTET2ⁱ cases were strongly associated with SRSF2^MT (P = .0009) and KRAS/NRAS^MT (P = .01) in CMML. Among biTET2ⁱ cases with SRSF2^MT and those with KRAS/NRAS^MT, CMML was diagnosed in 70% (P = .001) and 77% (P = .01), respectively (supplemental Figure 4B), which is higher than what was seen in the biTET2ⁱ population (44%; Figure 1E). We then investigated the overall impact of biTET2ⁱ by comparing the effect of biTET2ⁱ vs biTET2⁻ in relation to CMML diagnosis, the presence and absence of monocytosis, and the concomitant presence of SRSF2^MT or KRAS/NRAS^MT. No differences in survival outcomes could be attributed to the presence of biTET2ⁱ as a sole factor or in combination with others (supplemental Figure 5A-J).

When univariate analyses were conducted in the biTET2ⁱ vs TET2^WT population (Figure 4A), biTET2ⁱ was associated with older age, lower risk, MDS/MPN, CMML, normal cytogenetics, monocytosis, marrow hypercellularity, myeloid dysplasia, and the presence of SRSF2^MT and NRAS/KRAS^MT. In contrast, TET2^WT status correlated with MDS, leukopenia, neutropenia, pancytopenia, elevated BM blasts, and TP53^MT. We then conducted a multivariate Cox regression analysis (Table 3). For biTET2ⁱ vs TET2^WT, older age (≥60 years; OR, 4.2; P = .002), CMML (OR, 3.4; P = .03), monocytosis (OR, 2.1; P = .05), myeloid dysplasia (OR, 1.8; P = .04), marrow hypercellularity (OR, 2.4; P = .005), and SRSF2^MT (OR, 2.2; P = .02) were independent features associated with biTET2ⁱ. In biTET2ⁱ vs monoTET2^MT, CMML (OR, 6.7; P = .02), truncating TET2^MT (OR, 3.5; P = .02), and ancestral TET2^MT (OR, 5.5; P = .0002) were independently associated with biTET2ⁱ. When we compared the cohort of biTET2ⁱ vs biTET2⁻, CMML (OR, 6.9; P = .02), truncating TET2^MT (OR, 3.5; P = .02) and ancestral TET2^MT (OR, 5.5; P = .0002) were found to be independent prognostic factors in biTET2ⁱ, while elevated marrow blasts (OR, 0.2; P = .02) were more common in biTET2⁻. Finally, for monoTET2^MT vs TET2^WT, elderly (age ≥60 years; OR, 3.3; P = .002), TP53^MT (OR, 2.5; P = .01), and SRSF2^MT (OR, 2.1; P = .03) were found to be distinct for monoTET2^MT.

Objective molecular tools are complementing morphological methods in clinical practice. Morphology remains, however, the golden standard for identifying strong associations between phenotype and genotype in MN. We report here on phenotypical and morphological characteristics of cases harboring biallelic TET2 defects.

To date, only a few studies have investigated the clinical consequences of biTET2ⁱ in MN. We used a well-characterized cohort of patients with biTET2ⁱ. Our hypothesis was that biTET2ⁱ associates with a group of pathomorphological features that independently define a distinct MN subtype. To test our idea, we first identified biTET2ⁱ cases among a cohort of 1045 patients with MN and then studied correlations between mutational configuration and clinicohematological morphology in comparison with biTET2⁻ cases.

While our finding that most TET2^MT cases are truncating (frameshift or nonsense changes) agrees with previous studies,^38-40  we further show that a second TET2 hit in biTET2^MT cases significantly increases the chances of accumulating more truncating changes in those already harboring a TET2^MT. The prevalence of biTET2ⁱ among older patients demonstrated in this investigation is also in agreement with patterns observed in other studies.²⁵  This can be explained by the effect of aging on the accumulation of TET2^MT in clonal hematopoiesis of indeterminate potential and subsequent subclonal hits over time.^41,42 

TET2^MT is detected in a large fraction of myeloid disorders^38,43  but predominantly in CMML.⁴⁴  Here, we show that the MDS/MPN CMML subtype significantly correlated with biTET2ⁱ events (Figures 1C,E and 4). This relationship is modified by mutations in additional genes, such as SRSF2^MT in cases of CMML,^44,45  that are found to correlate strongly with biTET2ⁱ even when CMML is absent. This observation was further demonstrated by comparing biTET2ⁱ with monoTET2^MT, which showed a significant accumulation of MDS/MPN, CMML subtype, monocytosis, and SRSF2^MT solely resulting from the clonal succession following the second TET2 hit. When criteria for CMML diagnosis⁴⁶  were not clinically fulfilled, biTET2ⁱ remained invariably associated with monocytosis and myeloid dysplasia, both hallmarks of CMML proliferative and dysplastic features.^43,47 

Along with studies suggesting that TET2^MT has a neutral impact on the rate of progression to AML,^25,48  we also identified that biTET2ⁱ tended to occur in lower-risk patients with a lower likelihood of leukemic transformation, unless additional deleterious events co-occur. Given that most TET2 hits were ancestral to subsequent secondary mutations (second TET2^MT, SRSF2^MT, ASXL1^MT, RAS^MT, and DNMT3A^MT), we concluded that these founder lesions represent a leukemogenic predisposition (mutator phenotype) rather than driving leukemia. BiTET2ⁱ correlated with normal karyotype, as did the second TET2 hit in those harboring TET2^MT, implying a significantly lower likelihood of cytogenetic abnormalities and an association with lower-risk disease.

Other notable features of biTET2ⁱ included, in addition to monocytosis and SRSF2^MT, significantly higher leukocyte and neutrophil counts and less pancytopenia. When reviewing BM morphology, in addition to myeloid dysplasia, we observed low percentages of marrow blasts despite prominent hypercellularity. Moreover, lower odds of leukopenia and ring sideroblasts and an increased marrow cellularity were strongly correlated with a second hit in TET2 in comparisons of biTET2ⁱ and monoTET2^MT.

Inactivating TET2^MT leads to low 5hmC levels. 5hmC is the first oxidative product of the TET demethylation pathway marking tissue and cell-specific genes. A decrease in 5hmC levels is associated with malignant phenotypes and poor survival outcomes.⁴⁹  5hmC is highly localized at binding sites of the p300/CREB binding protein, and in vitro studies have shown that upon TET2 acetylation, 5hmC levels increase significantly. As previously reported, TET2^MT displayed low levels of 5hmC showing hypomethylation compared with healthy subjects at the majority of differentially methylated CpG sites. The greater degree of deficiency by the impairment of both TET2 alleles, the more the hydroxylation function is affected.²²  Consequently, more methylation of CpG sites would be expected. As a result, the previously described expansion of the stem cell compartment and differentiation that skewed toward monocytic differentiation should be expected to be more pronounced. In our study, biTET2ⁱ did not lead to a strong deleterious phenotype, suggesting that the acetylation mechanism might have been stabilized TET2 protein.

Given a more pronounced phenotype and genotype, it is important to speculate why biTET2ⁱ does not result in more serious clinical consequences. It is possible that the impairment of TET2 via inactivating mutations might be compensated by other TET enzymes (eg, TET1) or by decreased posttranslational modification (eg, acetylation) leading to a higher fraction of catalytically active protein. TET2 activity is regulated by acetylation, which increases TET2 stability (by protecting TET2 from ubiquitination and proteasome degradation) and promotes the cooperation with DNA methyltransferase 1 (DNMT1).⁴⁴  Acetylation might represent a regulatory mechanism of TET2 protein. TET2 is acetylated by p300/CREB binding protein at key lysine residues (K110 and K111) located in the N terminus of the protein, and this acetylation can be switched by HDAC1/2 and SIRT1/2 deacetylases. The N terminus of TET2 seems to contain high levels of enzymatic activity and positive regulatory feedback, possibly because the acetylation of lysine residues is a natural mechanism increasing TET2 activity.⁵⁰ 

In conclusion, our collective observations demonstrate that biTET2ⁱ is a frequent event in MN. Furthermore, the presence of biTET2ⁱ contributes additional information to the genetic complexity of MN and thus might represent a putative assessment tool of certain morphologic MN subentities associated with invariant phenotypic and molecular characteristics. The increasing collection of omics data and their correlation with clinical factors will further shed the light on the nature of biTET2ⁱ in MN.

This work was supported by the National Institutes of Health, National Heart, Lung, and Blood Institute (grants R01HL118281, R01HL123904, R01HL132071, and R35HL135795) and the Edward P. Evans Foundation (J.P.M.).

Contribution: H.A. designed the study, collected, analyzed, and interpreted clinical and molecular data, and wrote the manuscript; Y.N. collected data and provided important feedback to the manuscript; A.G., M.F.A., B.P., and Y.G. collected and analyzed data; C.M.H. and B.P.P., performed and analyzed DNA sequencing data and edited the manuscript; T.K., M.A., V.A., W.S., and L.W. collected clinical data and edited the manuscript; A.N., M.E.A., M.A.S., T.H., and B.K.J. provided clinical specimens and important insights on the manuscript; T.R. provided statistical advice and important editing in the manuscript; V.V. provided important editing and wrote the manuscript; J.P.M. designed the study, conceptualized and sponsored the overall research, and wrote the manuscript; and all authors read and approved the final manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jaroslaw P. Maciejewski, Department of Translational Hematology and Oncology Research, Lerner Research Institute, 9620 Carnegie Ave, N Building, NE6-250, Cleveland, OH 44106; e-mail: maciejj@ccf.org.