A Clinically Practicable Approach to Predict TP53 Allelic Configurations in Myeloid Neoplasia

The TP53 gene can be affected by hypomorphic/loss of function lesions occurring in diverse configurations. In addition to truncating hits, missense mutations in various hotspots may exert a dominant negative effect. Li-Fraumeni syndrome and acquired disease examples demonstrate that the presence of a wild type (WT) allele may be protective, and biallelic inactivation has the strongest leukemogenic drive. Consequently, cases with apparent biallelic TP53 lesions are associated with poor overall survival (OS) in myeloid neoplasia (MN). However, clinically applied NGS techniques currently do not always clearly identify biallelic subclones within seemingly single hit mutants, as the latter may contain subclones of various sizes. Thus, poorer than expected prognosis may be present in some cases depending on the presence and size of cryptic biallelic clones (hemizygous, homozygous, biallelic two hits) vs truly monoallelic or biclonal TP53 mosaic cases. Assessing the impact of variables surrounding TP53 mutations (TP53^MT) is complex and not always feasible. Thus, the development of clinically viable algorithms may improve prognostic precision in clinical settings.

We collected clinical and molecular data from 7400 MN patients and applied a novel model to properly resolveTP53^MTallelic configuration of and assess prognosis. Overall, 1285 TP53^MT were found in 1010 patients. Missense mutations were the most common (74%), followed by truncating (15%), whereas 11% of the patients had concomitant missense/truncating configuration. Most missense mutations (69%) were in canonical hotspot locations. TP53^MTcarriers presented with a variety of phenotypes and displayed worse OS compared to TP53^WT (HR=2.7; 95%CI 2.5-3). Following the traditional criteria for estimating allelic hits¹,36% of cases were classified as single hits, while 64% exhibited double hit configuration. Overall, carriers of TP53 double hits (either missense and/or truncated) had a worse OS than those with single hits TP53 (HR= 2.5; 2-3). However, a greater difference in OS between single and double TP53 hits was observed in MDS (HR= 3.1; 2.43- 4.09) compared to AML (HR= 1.5; 1.2- 2). When we compared TP53 single hits (traditional criteria) to TP53^WT in both AML (HR: 1.8 [1.4-2.5]) and MDS (HR: 1.3 [1-1.7]), TP53^MTcases were associated with worse OS.

To further resolve the uncertainty of TP53^MT allelic status assessment, we applied ML-driven recursive partitioning on cumulative hazard objective benchmarked to external survival data. Obligatory biallelic cases were identified among patients with: i) single TP53 hit and VAF >50%, ii) double TP53 hits with a combined VAF>50%, iii) TP53^MT VAF+del17p clonality >50%. Cases with VAF <50% can be differentiated into those likely with truly monoallelic TP53^MTand/or likely containing subclonal cryptic biallelic TP53 hits. We then cross-validated VAF cutoff values (Fig1-A) and applied random forest regression² with survival as a surrogate marker for TP53 allelic status. A VAF cutoff of 23% was found to be optimal in separating monoallelic and biallelic TP53^MT based on the frequency of VAF splitting criteria and decrease in error. Thus, we classified TP53^MT into: A) “obligatory” biallelic (OBi), B) “probable biallelic” (PBi), and C) “probable monoallelic” (PMo). Based on this approach, 579 patients (57%) had OBi, 239 (24%) had PBi, and 192 (19%) had PMoTP53^MT (Fig1-B). Of note, the OS of patients with PMo TP53^MT (median OS: 29 [10-77]) was similar to that of the TP53^WT (median: 42 [15-103]), p=.070). However, patients with PBi TP53^MT (median OS: 14 [7-37]) had worse outcomes compared to TP53^WT (p<.001). Furthermore, OS was similar when we compared AML to MDS within the OBi TP53^MT. These results were recapitulated in our single cell DNA sequencing studies. In 4 patients (3 likely monoallelic , 1 likely biallelic TP53^MT) a combined single cell DNA and CN analysis was performed. We found that some cells contained monoallelic TP53^MT while others had indeed biallelic lesions in all cases studied. Finally, our new algorithm was able to further discriminate OS differences beyond the traditionally classified single and double TP53 hit groups. Validity of our algorithm was confirmed in an external confirmatory cohort of 92 patients.

In sum, our novel approach more accurately resolves TP53 genomic configuration and uncovers genetic mosaicism to improve the prognostic evaluation.

1

2