TP53-mutant variant allele frequency and cytogenetics determine prognostic groups in MDS/AML for transplantation Open Access

Myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) are myeloid malignancies that have variable prognoses.^1-3 A specific subtype of MDS/AML carrying TP53 mutations is reported to have poor outcomes,¹ and it often does not respond to currently available treatments.¹ Hematopoietic stem cell transplantation (HSCT) can be curative for myeloid malignancies.⁴ However, patients with TP53 mutations have high relapse rates with short survival expectancy even with HSCT.⁵ Although it is often recommended for eligible patients with TP53 mutations to undergo HSCT, it is unclear which patients can achieve a long-term remission through HSCT.⁶ 

There is well-defined heterogeneity among TP53 mutations including the TP53 allelic state, co-occurring somatic mutations, and position within the clonal hierarchy.¹ Specifically, multihit TP53 mutations in MDS have been found to have associations with genome instability, treatment resistance, disease progression, and dismal outcomes.⁷ Patients with monoallelic TP53-mutated and TP53 wild-type diseases have similar outcomes with regard to response to therapy, overall survival (OS), and progression to AML.⁷ In AML, data on the prognostic impact of the allelic state are conflicting.^8,9 In addition, variant allele frequency (VAF) of TP53, number of mutations, and type of mutations have been reported as important for prognosis in several cohorts of MDS and AML.^9-13 

Given the plethora of heterogenicity in TP53-mutated myeloid diseases and uncertainty on the benefit of HSCT in these patients, we decided to investigate the outcomes of patients who underwent HSCT at the MD Anderson Cancer Center.

We conducted a retrospective study with an aim to define prognostic subgroups of patients with TP53-mutated AML and MDS receiving HSCT. We screened adult patients, aged ≥18 years, with at least 1 TP53 mutation who underwent HSCT between December 2012 and August 2022. All donors and conditioning regimens were included.

We categorized regimens as reduced intensity conditioning (1) IV busulfan in a dose calculated to target an area under the curve (AUC) of 4000 μmol/min ± 10% with fludarabine or (2) melphalan 100 to 140 mg/m² as a single dose with fludarabine. Myeloablative conditioning regimens consisted of one of the following IV busulfan-containing regimens: (1) nonfractionated busulfan with pharmacokinetic dose adjustment to achieve a target of an average daily systemic exposure, represented by the AUC of 5000 to 6000 μmol/min ± 10%, in combination with fludarabine or (2) fractionated busulfan with a fixed dose of 80 mg/m² delivered daily on days –13 and –12, followed by busulfan on days –6 to –3, and dosed based on pharmacokinetic studies to achieve a target AUC of 16 000 to 20 000 μmol/min for the course with fludarabine.

The study was conducted at a single academic center (The University of Texas MD Anderson Cancer Center) approved by the institutional review board of The University of Texas MD Anderson Cancer Center, and in accordance with the Declaration of Helsinki.

Mutation analysis was performed on bone marrow specimens using a 28-, 53-, or 81-gene targeted next-generation sequencing (NGS) panel as previously described.¹⁴ Across all panels, the limit of detection of the assay was 2% VAF. Although a minimum bidirectional coverage of 250× was required for variant calling, the median coverage achieved across TP53 amplicons was 1500×. For cases with multiple mutations, the highest VAF of the TP53 mutations detected before the HSCT was used for analysis. These mutations were assessed usually in multiple timepoints, including diagnosis, follow-up, and pretransplant. Evolutionary action (EA) score of TP53 was calculated using the algorithm available at http://eaction.lichtargelab.org/Eap53, and it was categorized as high or low based on the cutoffs derived from previous work from our center.¹⁴ Multihit status of TP53 alterations was determined using the following criteria: presence of >1 TP53 mutation, TP53 mutations with VAF ≥ 50%, or 1 TP53 mutation plus 17p deletion based on cytogenetics. With the in-house targeted NGS panel, which uses bulk DNA, we applied the criteria outlined by Bernard et al⁷ to distinguish between monoallelic and biallelic mutations based on the number of mutations, VAF, and deletion of the 17p/TP53 locus. We did not exclude TP53 mutations with VAF < 10%. After exploring outcomes with classification and regression tree analysis (CART), we decided that the best VAF cutoff for analysis was 50%, but we also included 10% as a variable because it had been found before to hold prognostic significance.¹⁰ 

Measurable residual disease status by multicolor flow cytometry and/or NGS in a pre-HSCT marrow was available for a fraction of patients.

Conventional karyotypic analysis by G-banding was performed on a metaphase spread prepared from a bone marrow aspirate using standard techniques.¹⁵ Fluorescence in situ hybridization for TP53 deletion was performed on 96 cases usually at diagnosis.¹⁶ The results were reported using the International System for Human Cytogenetic Nomenclature.

The primary outcome of this analysis was progression-free survival (PFS). PFS was defined as time from HSCT to disease progression or death from any cause and was estimated using the Kaplan-Meier method. Univariate analysis evaluating prognostic factors for PFS was performed using a Cox proportional hazards model. The cumulative incidence method accounting for competing risks was used to estimate the incidence of nonrelapse mortality (NRM), disease progression, and graft-versus-host disease (GVHD). Competing risks included disease progression or relapse death for NRM, death of any cause for progression, and disease progression or death before GVHD for acute and chronic GVHD. CART was used in multivariate analysis to identify the optimal prognostic algorithm incorporating the most significant predictors. This method was selected for multivariate analysis given the high correlation among some covariates of interest in this study. CART analysis is based on a recursive partitioning process by which the predictive value of all covariates included in the analysis is repeatedly tested. Partitions occur each time a statistically significant covariate is identified; the prognostic value of all remaining covariates is then tested within the subsets of that covariate. The recursive process stops when no additional significant predictors can be identified. This recursive partitioning process generates a hierarchical ranking (or algorithm) of the significant predictors that is represented through the “tree” and its “branches.” Statistical significance is determined based on martingale residuals of a Cox model which are used to calculate χ² values for all possible cutpoints of covariates. A sample size of 20 was set as the minimum number of patients at each “node.” All covariates identified to be significant in univariate analysis were included in CART analysis. All statistical analyses were performed using Stata 16 (College Station, TX). Statistical significance was defined at the 5% level.

We identified 240 patients with AML (n = 126) and MDS (n = 114) whose neoplasm had TP53 mutations eligible for outcome analyses. Baseline characteristics of the study population are listed in Table 1. The median age at HSCT was 62 years (range, 18-75). There were 171 patients (71%) who had complex cytogenetics at diagnosis and 18 patients (7%) who had deletion 5q or deletion 7q. The median TP53 mutation VAF was 28.21% (range, 1.02-98.5). The median EA score was 84.11 (range, 3.4-100). Furthermore, 71 patients (30%) had >1 TP53 mutation detected (2 mutations, n = 57; >2 mutations, n = 14). Multihit TP53 alterations were observed in 166 patients (69%), and truncating TP53 mutations were observed in 39 patients (16%). A total of 94 patients (39%) who received HSCT are in complete remission (CR). Of those patients, 68 had pre-HSCT TP53 evaluated and only 22 (32.4%) had a negative result. All but 4 patients had treatment directed for the disease pre-HSCT; 195 received a hypomethylating agent alone or in combination with other agents as summarized in Table 1.

In Table 2, we present the characteristics of the patients who had TP53 mutation VAF < 50% or VAF ≥ 50%. Patients with VAF ≥ 50% had complex cytogenetics (89.4% vs 65%; P = .001) more frequently and were always multihit (100% vs 74%; P < .001) but had a lower number of TP53 mutations (80% with 1 mutation vs 65%; P = .05).

The median time from AML or MDS diagnosis to HSCT was 6.5 months. The donor sources were matched sibling donors in 60 patients (25%), matched unrelated in 126 (52.5%), haploidentical in 42 (17.5%), and mismatched unrelated in 11. Furthermore, 141 patients (59%) received myeloablative conditioning regimens and 67% received posttransplant cyclophosphamide for GVHD prophylaxis. According to our institutional guidelines, eligible patients were offered posttransplant maintenance. In this cohort, 60 of 240 patients received at least 1 cycle of posttransplant maintenance treatment after SCT.

The median follow-up in 87 surviving patients was 19 months (range, 3.26-93). The 2-year PFS for the whole cohort was 24% (Table 3). The 2-year OS was 34% (n = 82). The 2-year incidence of NRM was 22% and the 2-year progression was 53%. The incidence of grades 2 to 4 and grades 3 to 4 acute GVHD at 6 months was 42% and 7%, respectively, whereas the incidence of chronic GVHD at 2 years was 12%.

Univariate analysis for PFS identified several predictors as found in Table 4. Cytogenetic abnormalities without complex karyotype and without 5q/7q (hazard ratio [HR], 0.3; 95% confidence interval [CI], 0.2-0.6), CR as the disease status at HSCT (HR, 0.7; 95% CI, 0.5-0.9), and having a matched sibling donor for HSCT (HR, 0.7; 95% CI, 0.5-0.9) were associated with favorable prognosis. However, TP53 mutation VAF ≥ 50% (HR, 3.6; 95% CI, 2.2-5.7), TP53 mutation multihit status (HR, 2.1; 95% CI, 1.4-3.1), and detectable measurable residual disease by multicolor flow cytometry while in CR (HR, 1.8; 95% CI, 1.1-3.2) were all associated with inferior outcomes. Interestingly, having MDS vs AML did not affect the outcome (HR, 0.9; 95% CI, 0.7-1.2). Conditioning intensity and GVHD prophylaxis chosen had no impact on the primary outcome of PFS. EA score and TP53 mutation detection by NGS pre-HSCT did not reach statistical significance either.

Multivariate analysis was performed using CART analysis as described previously. For this analysis, complex cytogenetics and 5q or 7q deletion were combined because of similar outcomes. Among all covariates included, CART identified VAF as the most significant prognostic variable and cytogenetics as the second most prognostic for PFS. These were also the most significant variables in the Cox univariate analysis. Of note, the corresponding Cox regression model, including VAF level and cytogenetics, yields a C-statistic of 60.5%. Considering the largest subset (VAF < 50% without complex cytogenetics and without del5q/del7q) as a reference, the HR was 2.03 (95% CI, 1.4-2.9; P < .001) for VAF ≥ 50% and 0.4 (95% CI, 0.2-0.7; P = .001) for VAF < 50% without complex cytogenetics and without del5q/7q deletions. These 2 variables formed an algorithm with 3 prognostic subgroups as found in Figure 1. (1) There were 49 patients (20.4%) in the worst prognostic group with 2-year PFS of 3% (TP53 mutation VAF ≥ 50%); (2) 131 patients (54.6%) were in the intermediate prognostic group with 2-year PFS of 22% (TP53 mutation VAF < 50% but having complex cytogenetics or 5q or 7q abnormalities); and (3) 39 patients (16.3%) were in the best prognostic group with 2-year PFS of 60% (TP53 mutation VAF < 50% and not having complex/5q/7q cytogenetics). Of note, CART suggested an extra subcategory within the VAF < 50% and complex cytogenetics group, depending on the VAF being less or more than 10%. However, owing to the small difference of the 2 resulting subgroups (∼5% in PFS), we decided not to include it in the final model. The reasons of PFS failure from each subgroup are detailed in Table 5. As expected, most failures for the 2 worse subgroups came from disease progression (72.2% in the VAF < 50% and complex/5q/7q group and 67.4% in the VAF ≥ 50% group).

In this study, we demonstrated that integrating VAF and cytogenetic status improves risk stratification and enhances long-term outcome predictions for patients with TP53-mutated AML or MDS undergoing HSCT. In this cohort of 240 patients with AML/MDS, outcome after HSCT was extremely poor with PFS < 5% if VAF ≥ 50 and that group represented approximately 20% of our HSCT cohort. Patients with VAF < 50% had much improved outcome expectation especially if they did not have cytogenetic abnormalities consistent with complex karyotype or 5q/7q abnormalities. The favorable prognosis group with 2-year PFS of 60% only represented 16.3% of our cohort, and most patients (54.6% of the study cohort) fell in between for prognosis with 2-year PFS of 22%. Our data clearly confirm the heterogeneity of patients with TP53-mutated AML/MDS even in the post-HSCT setting for outcome expectation.

Recently, a City of Hope study reported a 50% 5-year OS for 18 patients with MDS with TP53 mutation, which was similar to the other therapy-related MDS within their cohort.¹⁷ This was dissimilar to our results, which is likely a result of a smaller sample. A large Center for International Blood and Marrow Transplantation retrospective study that sequenced specimens revealed a detriment in survival for MDS with TP53 mutations. Specifically, TP53-mutated cases had 20% 3-year OS vs ∼50% for the rest.^10,18 The number of TP53 mutations, karyotype, and VAF was not associated with survival, but patients with truncating TP53 mutations had a worse prognosis.¹⁰ Another study of MDS/Myeloproliferative neoplasm (MPN) in Japan revealed that the combination of TP53 mutation and complex karyotype resulted in <20% 2-year OS.¹⁹ VAF affected OS but less so than complex karyotype. More recently, a study from Memorial Sloan Kettering revealed that biallelic inactivation of TP53 worsened the prognosis in the post-HSCT setting but not if monoallelic.⁷ Recently, subgroup analysis of the BMT-CTN 1102 study analyzing 87 patients with TP53-mutated MDS revealed that reduced intensity conditioning HSCT might improve long-term survival in patients with mutated TP53, independent of other risk factors with 3-year OS of 23% vs 11% without HSCT.²⁰ In that analyses, allelic state and VAF were not prognostic.

Our AML cohort with TP53 mutation had poor outcomes similar to previous reports. A 2014 multicenter study in patients with abnl(17p) who underwent HSCT revealed a 3-year OS of 15%.²¹ A similar report with TP53-mutated AML revealed only 10% 3-year OS.²² A 2017 European Society for Blood and Marrow Transplantation (EBMT) study revealed a 2-year OS of 28% for AML with abnl(17p), with del5q or del7q worsening the prognosis even more.²³ A recent meta-analysis of all TP53-mutated AML cases published revealed a pooled 2-year OS of 29.7% and a relapse rate of 61.4%.⁵ A Japanese study revealed that TP53-mutated cases without complex cytogenetics had a better prognosis post-HSCT which was later confirmed by an EBMT study published in 2022.¹⁹ The EBMT study revealed that patients with TP53 mutation who did not have complex karyotype or abnl(17p) had a much-improved outcome (2-year OS 64% vs 24.6%).²⁴ This finding is in agreement with our results. More recent studies from a multi-institutional cohort revealed that the outcome in TP53-mutated AML has not improved in the era of novel therapies, but HSCT seems to still offer a survival advantage.^25,26 

To summarize, the HSCT decision in patients with TP53 mutation remains to be challenging considering the lack of established benefit to improve survival, the high risk of failure due to relapse, older age of patients with AML/MDS with higher comorbidity profile that may also increase NRM, and unmeasured impact of HSCT on quality of life. In this population, our results do provide an algorithm on how to approach patients with TP53 mutations by identifying prognostic subgroups. It would be reasonable not to consider HSCT in patients with 2-year PFS < 10% unless a clinical trial to improve outcomes is available. It would also be reasonable to consider proceeding with HSCT with any donor available as soon as possible in patients with 2-year PFS estimate ≥ 50%. Our results in 240 patients with lack of impact of conditioning intensity may demonstrate that rather than increasing intensity, innovative approaches, such as integrating immune-mediated treatment approach to the backbone of HSCT, could be explored.

Our study has several strengths. The TP53-specific HSCT cohort is the largest single-center cohort ever published which allowed for robust statistical analysis. The cohort had patients treated with a variety of regimens and donors allowing us to dissect the impact of such variables on prognosis. All genetic information including number of mutations, VAF, biallelic status, and EA score was available to us. Finally, given that we had access to all the patient charts, we were able to verify all information that we needed centrally.

Inherent in retrospective design, our research had several limitations. There was a selection bias of TP53-mutated and transplanted patients that they were medically eligible for a high-risk procedure such as HSCT. This study also represents the experience of a single center, reflecting its guidelines, treatment preferences, and unique patient population. Our patients received intensive regimens with either busulfan backbone or melphalan with no regimen truly nonablative. The current study is based on TP53 mutation determined by bulk DNA analysis, rather than a single-cell mutation analysis. Therefore, it is not possible to determine with absolute certainty whether 2 different TP53 mutation are present in the same clone or in 2 separate clones. To address this, we have adopted the criteria from Bernard et al which include the following: presence of 2 different TP53 mutations, a single TP53 mutation at VAF > 49%, or a single TP53 mutation (with any VAF) plus a deletion. Finally, in some of our patients, TP53-specific fluorescence in situ hybridization was not performed, and as such we might have misclassified the biallelic status by relying purely on cytogenetics.

In conclusion, we believe that patients with TP53-mutated AML/MDS represent a unique and heterogeneous high-risk group that should not be treated as 1 single entity. Our results when interpreted with previously published reports clearly indicate that there are different subgroups, and they should be the target of different treatment approaches to improve their outcomes. Our good-risk group in this population had PFS estimates of 60% at 2 years, suggesting that the presence of a TP53 mutation should not be the sole criterion for excluding patients from proceeding with HSCT. Clinical trials through multicenter collaborations are urgently needed to provide the best care with improved outcomes to this unique population.

Contribution: K.L., R.M.S., R.K.-S., and B.O. designed the research, performed the research, contributed vital new reagents or analytical tools, analyzed the data, and wrote the manuscript; and J.R., G.O., T.K., N.J.S., H.K., D.M., P.K., U.P., R.C., and E.S. performed the research, analyzed the data, and wrote the manuscript.

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

Correspondence: Betül Oran, Department of Stem Cell Transplantation and Cellular Therapy, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Unit 423, Houston, TX 77030; email: boran@mdanderson.org.