Prognostic impact of pre-transplant chromosomal aberrations in peripheral blood of patients undergoing unrelated donor hematopoietic cell transplant for acute myeloid leukemia

To improve risk stratification and treatment decisions for patients with acute myeloid leukemia (AML) undergoing hematopoietic cell transplantation (HCT). We used SNP-array data from the DISCOVeRY-BMT study to detect chromosomal aberrations in pre-HCT peripheral blood (collected 2–4 weeks before the administration of conditioning regimen) from 1974 AML patients who received HCT between 2000 and 2011. All aberrations detected in ≥ 10 patients were tested for their association with overall survival (OS), separately by remission status, using the Kaplan–Meier estimator. Cox regression models were used for multivariable analyses. Follow-up was through January 2019. We identified 701 unique chromosomal aberrations in 285 patients (7% of 1438 in complete remission (CR) and 36% of 536 not in CR). Copy-neutral loss-of-heterozygosity (CNLOH) in chr17p in CR patients (3-year OS = 20% vs. 50%, with and without chr17p CNLOH, p = 0.0002), and chr13q in patients not in CR (3-year OS = 4% vs. 26%, with and without chr13q CNLOH, p < 0.0001) are risk factors for poor survival. Models adjusted for clinical factors showed approximately three-fold excess risk of post-HCT mortality with chr17p CNLOH in CR patients (hazard ratio, HR = 3.39, 95% confidence interval CI 1.74–6.60, p = 0.0003), or chr13q CNLOH in patients not in CR (HR = 2.68, 95% CI 1.75–4.09, p < 0.0001). The observed mortality was mostly driven by post-HCT relapse (HR = 2.47, 95% CI 1.01–6.02, p = 0.047 for chr17p CNLOH in CR patients, and HR = 2.58, 95% CI 1.63–4.08, p < 0.0001 for chr13q CNLOH in patients not in CR. Pre-transplant CNLOH in chr13q or chr17p predicts risk of poor outcomes after unrelated donor HCT in AML patients. A large prospective study is warranted to validate the results and evaluate novel strategies to improve survival in those patients.


Pre-transplant chromosomal aberrations associated with overall survival. Univariate analyses
in patients with advanced disease showed poor post-HCT overall survival with the following aberrations: copy losses in chr5q (N = 24, log-rank p = 0.046), 17p (N = 16, p = 0.01), 12p (N = 12, p = 0.04), 15q (N = 9, p = 0.02), and CNLOH in 13q (N = 25, p < 0.0001). Only CNLOH chr13q remained statistically significantly associated with patient survival in a multivariable model with other aberrations and important clinical factors (Supplemental Table 2). The 3-year overall survival (OS) probabilities in patients with advanced disease with and without chr13q CNLOH were 4% vs. 26%, log-rank p < 0.0001, with a median survival of 2.7 vs. 7.4 months, and 96% vs. 73% of the death happened in the first year after HCT, respectively ( Fig. 2A and Supplemental Table 3). Multivariable analysis adjusted for clinical factors, showed that patients with advanced disease and chr13q CNLOH were at three-fold higher risk of post-HCT mortality compared to those with advanced disease but not with chr13q CNLOH (HR = 2.68, 95% CI 1.75-4.09, p < 0.0001) ( Table 2).
Chromosomal aberration profile in patients with CNLOHs in chr13q or chr17p. A total of 28 patients had chr13q CNLOH, with the majority having advanced disease (N = 25). On the other hand, chr17p CNLOH was detected in 17 patients with advanced disease and 10 patients in CR. Figure 3A,B show aberration profile of patients with CNLOH in chr13q or chr17p by CR status. None of the patients had both aberrations regardless of their remission status. Thirty-nine percent (N = 11) of the patients with chr13q CNLOH carried aberrations in other chromosomes vs. 56% (N = 15) of those with chr17q CNLOH. Chromosomal regions affected by chr13q or 17p CNLOH in this study include FLT3 in 13q12.2 and TP53 in 17p13.1 (Supplemental Fig. 1A Table 3. In a multivariable model, the presence of CNLOHs in chr13q in patients with advanced disease or in chr17p in CR patients was associated with more than twofold increased risk of leukemia progression or relapse, respectively (HR = 2.58, 95% CI 1.63-4.08, p < 0.0001; and HR = 2.47, 95% CI 1.01-6.02, p = 0.047) ( Table 2). CNLOH in chr13q or chr17p in patients with normal cytogenetics at AML diagnosis and their association with overall survival after HCT. In analysis restricted to patients with normal cytogenetics at AML diagnosis (N = 572), approximately one-quarter (N = 157, 27.4%) received HCT while not in CR, with CNLOHs in chr13q (N = 13, 8.3%) the most frequently detected aberration in this group. Among CR patients with normal cytogenetics at diagnosis, 5% had chromosomal aberrations at HCT but none had chr17p CNLOH. Supplemental Fig. 3A,B summarize the chromosomal aberration profile in patients with normal cytogenetics at AML diagnosis by CR status. Similar to results from the full cohort, the presence of chr13q CNLOH in patients with advanced disease in this subgroup analysis was associated with excess risk of post-HCT mortality (HR = 2.78, 95% CI 1.48-5.21, p = 0.002) (Supplemental Table 4).  Table 2. Association between copy-neutral loss of heterozygosity in chr13q or chr17p and post-HCT outcomes in patients with AML. a Model was adjusted for recipient race, Karnofsky Performance Status scores, study cohort, GvHD prophylaxis, donor-recipient CMV serostatus matching, and graft type. b Model was adjusted for Karnofsky Performance Status scores, and graft type. c Model was adjusted for recipient age, donor age, GvHD prophylaxis, Karnofsky Performance Status scores, donor-recipient CMV serostatus matching, and stratified on recipient sex, conditioning intensity, graft type, and year of transplant. d Model was adjusted for conditioning intensity, donor sex, study cohort, and stratified on Karnofsky Performance Status scores. Chr17p CNLOH was associated with a threefold increase risk of mortality and more than a twofold increased  CNLOH in chr13q or chr17p and their association with outcomes after HCT in patients with de novo AML. Supplemental Fig. 5A-D show chromosomal aberration profile in patients with de novo and treatment-related AML stratified by remission status. In de novo AML (N = 1611), the most frequent aberration in patients in advanced disease (N = 455) was chr13q CNLOH (N = 23, 5.1%), followed by chr5q loss (N = 18, 4%), and in those in CR (N = 1156) was CNLOH in chr9p (N = 11, 1%), followed by chr17p CNLOH (N = 7, 0.6%) and chr5q loss (N = 7, 0.6%). In treatment-related AML (N = 145), the most common aberration was chr5q loss in advanced disease (N = 4, 15.4%). In an analysis restricted to patients with de novo AML (N = 1611), we observed similar results to that of the full cohort. In patients with advanced disease, the presence of chr13q CNLOH was associated with approximately 3-folds increased risks of death or leukemia progression (OS: HR = 2.82, 95% CI 1.81-4.40, p < 0.0001; progression: HR = 2.86, 95% CI 1.78-4.62, p < 0.0001) (Supplemental Table 5). In CR patients, the presence of chr17p CNLOH was associated with threefold increased risk of death (HR = 3.14, 95% CI 1.38-7.15, p = 0.01) and twofold excess risk of relapse, although not statistically significant (HR = 2.33, p = 0.15). Small number of patients with treatment-related AML (N = 145) did not allow for robust outcome analyses.

Discussion
In this large study of 1974 AML patients who underwent unrelated donor HCT, we showed that CNLOH in chr17p or chr13q (detected in approximately 3% of the patients) can identify new high-risk groups. The 3-year OS for CR patients carrying chr17p CNLOH was similar to those with advanced disease in the absence of chr13q CNLOH (3 years OS = 20% vs. 26%). Yet, harboring chr13q CNLOH in patients with advanced disease resulted in a detrimental outcome with a 3-year OS of 4% (with almost all death happened in the first year after HCT). The observed poor survival associated with such aberrations were primarily driven by the excess risk of relapse or disease progression. Although relatively small numbers, poor survival associated with CNLOH in chr17p or chr13q can refine pre-HCT risk stratification and guide precision medicine strategy in HCT decisions for patients with AML. Responses to induction therapy dictated HCT prognosis in patients with AML for many years, with patients receiving HCT in first CR having the best prognosis and those with advanced disease with the worst outcome 4 . Recent studies showed a prognostic improvement when considering measurable residual disease in patients with morphological remission [12][13][14] . It is plausible that the observed inferior survival associated with CNLOH in chr17p or chr13q in our study is a reflection of residual leukemia or its aggressive phenotype. We showed that these aberrations affect genomic regions harboring TP53 and FLT3, respectively, both known to be associated with risk of leukemia relapse and poor prognosis 15,16 . The presence of mutated TP53 or high allelic ratio of FLT3-ITD at AML diagnosis are included in the adverse risk category of the European Leukemia Net 2017 (ELN-2017) 2 . We were unable to examine the role of ELN risk scheme in our analysis due to the lack of information but the presence of chr13q CNLOH in patients with FLT3-ITD was associated with a shorter survival (median OS = 2.3 vs.10.1 months, p = 0.009, in patients with and without chr13q CNLOH, respectively) in one study 17 . It is possible that the loss-of-heterozygosity in those genomic regions are subsequent events to mutations in TP53 or FLT3 that may modify patient outcome. If valid, pre-HCT testing for clonal CNLOH in chr17p or chr13q may guide the identification of patients at highest risk of inferior outcomes among patients harboring somatic mutations in TP53 and FLT3, a hypothesis that warrants investigation.
In line with previous literature in AML 17 , our finding showed that chr13q CNLOH was detected primarily in patients with advanced disease (25/28); this was also true for the subset of patients who had normal cytogenetics at AML diagnosis. The observed detrimental post-HCT outcome in patients with advanced disease and chr13q CNLOH (median survival = 2.7 months; 3-year OS = 4%) need to be validated in a larger study as it questions the role of HCT in those patients and calls for careful risk-benefit assessment for this vulnerable patient population. Also, the presence of CNLOH in chr13q at HCT further stratified patients who were originally classified as intermediate risk based with normal cytogenetics at AML diagnosis. All those patients did not achieve CR and had an approximately three-fold risk of post-HCT leukemia relapse when compared to advanced disease patients free of such aberration. This observation calls for a longitudinal evaluation of chromosomal loss-of heterozygosity in AML patients with normal cytogenetics to understand their dynamics through the disease treatment course.
Surprisingly, the effect of chr17p CNLOH on post-HCT survival was restricted to CR patients. It is possible that the higher frequency of reduced intensity regimens in CR patients with chr17p CNLOH (60% in CR vs. 18% in advanced disease) contributed to this observation. Alternatively, the use of fludarabine-based regimen (in 80% of CR patients with chr17p CNLOH) was less effective. In-vitro studies suggested that cells with chr17 abnormalities are resistant to fludarabine-induced apoptosis 18,19 . A recent CIBMTR study in myelodysplastic syndrome showed that the presence of TP53 mutation was associated with an approximately twofold increased risk in post-HCT mortality (HR = 1.71, p < 0.001) 20 ; this was not modified by conditioning regimen intensity. A prospective study is needed to determine the best conditioning regimen strategy for those patients.
In the current study, almost none of the 168 pediatric AML patients carried either of our identified prognostic aberrations; small sample size may be a limiting factor. Yet, a recent large study of young AML patients (age at diagnosis ranging between days and 29 years) showed different molecular landscape than that of adult AML with distinct FLT3 mutations but almost no TP53 mutations 21 .
In the current study, we found no associations between pre-HCT SNP array-detected copy loss in chr5q or chr7q and patient post-HCT survival in either advanced or CR patients. Monosomy 7 and monosomy 5 or 5q deletion are known adverse cytogenetic prognostic markers for AML 2  www.nature.com/scientificreports/ CNLOH (N = 7, 2 in CR and 5 not in CR) had pre-HCT SNP-array detected copy loss in either chr5 or chr7, while no patient with 13q CNLOH had such aberrations. This was also consistent in the subset of patients with available cytogenetic information at AML diagnosis, where approximately 63% of the CR patients with chr17p CNLOH had deletions in chr5 or chr7 (N = 5/8), but only one patient with chr13q CNLOH did. Taken together, this suggests that the observed association with chr13q CNLOH is mostly independent of other known risk cytogenetic findings.
The clinical usefulness of SNP microarray platform in hematological malignancies is widely supported particularly as technology advances and cost drops 22 . Its possible use to guide HCT decision-making is still under investigation. Using the same technology, we previously showed a possible value to its use for transplant decisions in patients with Fanconi anemia 23 , an inherited marrow failure syndrome caused by defects in DNA repair genes, and acute lymphoblastic leukemia 24 .
The strengths of this study include the large sample size, the use of the genome-wide sensitive array, and the availability of well-captured clinical and outcome data. Yet, our study was limited by the unavailability of serial samples, which restrict our ability to explain whether those detected alterations are markers of residual disease or clonal drivers. Our analyses focused on aberrations observed in ≥ 10 patients, a larger study is warranted to validate our findings and evaluate less frequent aberrations. The study also lacked information on pre-HCT measurable residual disease, ELN risk classification, and prognostic mutations. SNP array analysis was completed using peripheral blood DNA; the use of bone marrow samples to validate our findings is warranted.
In conclusion, we showed poor survival associated with CNLOHs in chr13q and 17p in patients with AML. Further, larger scale study with diagnostic and pre-HCT samples for both genomic sequencing and SNP array analyses are warranted to elucidate the possible interplay between somatic point mutations and chromosomal alterations in AML pathogenesis, prognosis, and treatment response. Detection of chromosomal aberrations. DNA was extracted from peripheral whole blood samples collected 2-4 weeks prior to HCT conditioning regimen administration. We used SNP-array genotype data generated by the HumanOmniExpress-12v1_A BeadChip to calculate the log 2 R ratio (LRR) and B allele frequency (BAF). LRR is computed as the log 2 ratio of observed to expected signal intensity for each SNP; LRR > 0 indicates copy number gain, whereas LRR < 0 indicates copy number loss 26 . BAF is calculated as the frequency of the B allele at a given biallelic SNP, where BAF = 0.5 indicates a heterozygous genotype. A deviation from heterozygosity in the BAF without LRR changes indicates a copy-neutral loss of heterozygosity (CNLOH) 27 . We used quantile normalization and GC and CpG waves correction to improve accuracy, as described previously 28 . We used a custom software pipeline that included BAF Segmentation software 29 . All potential events were plotted, and false positive calls were removed after manual review. We only included chromosomal aberrations ≥ 2 Mb to minimize false positive discovery 28 . The detection method we used have shown to yield higher sensitivity and specificity for detecting allelic imbalances when compared with other CNV detection methods 29 . Partek Genomics Suite software, version 7.0 30 (Partek Inc., St. Louis, MO, USA, https:// www. partek. com/ partek-genom ics-suite/) and R package "OmicCircos", version 1.28.0 31 (https:// bioco nduct or. org/ packa ges/ OmicC ircos/) were used for visual representation of chromosomal aberration data. Study end points. Two end points were tested for this study: (1) overall survival (OS) defined as time from date of HCT to death from any cause or last follow-up, and (2) Leukemia relapse or disease progression. Complete remission in this study is primarily based on hematologic remission. Statistical analysis. We compared differences in clinical and transplant-related factors between patients with and without aberrations using chi-square or Fisher's exact test, as appropriate. All outcome analyses were completed separately by patient CR status to account for the differences in clinical management, HCT outcomes, and frequencies and types of detected aberrations between the two groups (N in CR = 1438 and N not in CR, i.e. advanced disease = 536). To identify chromosomal aberrations of prognostic effect, we first used the Kaplan-Meier estimator to calculate OS with each chromosomal aberration that occurred in ≥ 10 patients in the full cohort. All aberrations with statistically significant associations in the univariate models were then included in a multivariable Cox model to evaluate the independent effect of each controlling for other aberrations and important clinical factors. Aberrations that remained statistically significant from the previous step were tested for their associations with OS and relapse or disease progression in multivariable Cox proportional hazards models. Clinical factors included in the models were selected using a stepwise selection strategy with a threshold of p < 0.25 for entry and p < 0.15 for stay. Proportional hazard assumptions were examined using Schoenfeld residuals and violation was accounted for through stratification for non-proportional hazards effect. We used the cause-specific hazard analysis for relapse/disease progression with transplant related mortality treated as competing risk. Follow-up started at date of HCT and ended at outcome under-study, last follow-up, or in