Page 42 - 2021_06-Haematologica-web

M. Novakova et al.
liferation-activating gene CRLF2 (lower in patients with monocytic switch than in those without monocytic switch), the cell cycle regulator CCNA1 (cyclin A1, higher in patients with monocytic switch) and androgen receptor (AR, higher in patients with monocytic switch), as well as six genes with a CD marker designation: CD371, CD301, CD1E, CD125 (all higher in patients with monocytic switch), CD158K and CD20 (both lower in patients with monocytic switch). Consistent with our previously pub- lished data, patients with switch had increased CEBPA expression (P<0.0001), which was also the case when only DUX4r patients were analyzed (P<0.0001).
We asked whether these changes were primarily driven by monocytic switch itself or by the underlying genotype. Both the unsupervised hierarchical clustering and the UMAP analyses showed that the patients primarily clus- tered according to genotype, including the DUX4r, PAX5- P80R and ZNF384r subtypes, which were enriched for monocytic switch, and whether the patients eventually experienced monocytic switch seemed to be secondary (Figure 2A; Online Supplementary Figure S4).
As genotype-associated transcriptome differences may override monocytic switch-associated differences, we repeated the analysis in the most prevalent (accounting for >60% of the patients with switch) genotypic subset (DUX4r). The Online Supplementary Figure S5 shows that the clustering of the switching patients was not clear. Among the top-ranking most differentially expressed genes between switching (n=44) and nonswitching (n=11) patients were the previously described CEBPA,1 hematopoi- etic regulator FLT3 and Toll-like receptor TLR10, which were all higher in patients with the DUX4r subtype with switch than in patients with DUX4r without switch (Online Supplementary Table S4).
Genotype also influences immunophenotype associated with subsequent monocytic switch
We next sought to determine whether the diagnostic immunophenotype predicts, similarly to transcriptome, monocytic switch behavior and whether it can be associat- ed with underlying genetic aberrations.
For these purposes we analyzed the diagnostic immunophenotype of the blast population in 745 patients. The overall picture appeared to be analogous to that of
the transcriptome data: diagnostic immunophenotypes were grouped based on the genetic subtype as seen in the UMAP analysis rather than by the monocytic switch status (Figure 2B).
In our previous study all patients with monocytic switch harbored CD2 expression, although belonging to different genetic subtypes. As we have shown recently,28 CD2 is expressed in approximately 75% of DUX4r patients where- as a newly described marker CD371 (CLL-1) is found in all patients with DUX4r (Online Supplementary Figure S8).
We also observed the expression of CD2 and CD371 in five of seven and two of five patients with PAX5-P80R, respectively. Interestingly, all seven patients with PAX5- P80R had homogeneous expression of CD66c. Expression of CD4, a rare aberrant marker in BCP-ALL, was present in four of seven PAX5-P80R patients, all of which had cells with a phenotype of CD34neg33pos (in contrast to the other three of seven CD4negCD34pos patients). Interestingly, CD66c expression was retained on switched monocytoid cells on d+15 in all six patients with available data.
Patients with the ZNF384r subtype were often classified
as having acute leukemia of ambiguous lineage (ALAL) using the EGIL classification (six of 11 patients).
The typical immunophenotypes of the main three sub- types with switch (DUX4r: CD10posCD20negCD34posCD2posCD371pos; PAX5-P80R: CD10neg/posCD66c>75%CD2<50%CD4negCD34posCD33<50% or CD10neg/posCD66c>75%CD2posCD4posCD34negCD33pos; ZNF384r: CD10<50%CD13posCD66cnegCD34posCD135posCD24<60%), are shown in the Online Supplementary Figure S8.
Discordance between flow cytometry- and polymerase chain reaction-determined minimal residual disease reflects different switching dynamics in individual genetic subtypes
Loss of B-cell markers during monocytic switch interferes with B-cell-oriented MRD analysis by FC. As we showed previously in our pilot study1 and now in a significant cohort of patients, switched monocytoid blasts maintain leukemic IG/TR rearrangements despite completely losing the B-cell phenotype. We thus analyzed the influence of switch on MRD detection by comparing FC and PCR MRD quantitation results in the genotype subsets most prone to switch.
We had only limited data for d+8 PB samples from PAX5- P80R (n=3) and ZNF384r patients (n=2). Spearman R for DUX4r (n=31) and patients with monocytic switch outside these three genetic subtypes (n=11) was 0.7 and 0.72, respectively (Online Supplementary Figures S6 and S7).
But, as shown in Figure 3, FC underestimation of MRD led to pronounced discordance at d+15 in the PAX5-P80R patients. The concordance of the FC and PCR MRD positiv- ity/negativity at the 10-3 level was 91% for the DUX4r sub- type and 82% for the ZNF384r subtype but only 17% for the PAX5-P80R subtype. Concordance at the 10-1 level, which was determined at this time point as the FC cut-off value for stratifying patients for undergoing high-risk treat- ment according to the BFM protocols, was 78% for the DUX4r subtype and 78% for the ZNF384r subtype but only 43% for the PAX5-P80R subtype.
At d+33, all three subtype groups showed poor correla- tion between the FC and PCR MRD values (Figure 3) com- pared to previously published data.18,29 The concordance of the FC- and PCR-determined MRD at the level of 10-3 was 55% for the DUX4r subtype, 56% for the ZNF384r subtype and 33% for the PAX5-P80R subtype.
Monocytic switch rarely caused MRD discrepancies in the other genetic subtypes. When analyzed separately, among 23 patients in the DUX4rnegPAX5-P80RnegZNF384rneg subset and in which monocytic switch was observed, FC MRD at the appropriate level of sensitivity (defined in the Methods) was measured in 21 and 16 patients on d+15 and d+33, respectively (Online Supplementary Figure S7). The concordance at d+15 for the 10-1 and 10-3 levels was 86% and 90%, respectively. The concordance at d+33 at the 10-3 level was 76%.
At week plus 12 majority of samples were PCR (86%) and FC (97%) MRD negative. Concordance for DUX4r (n=44), PAX5-P80R (n=5), ZNF384r (n=8) and cases with switch out- side these subtypes (n=21) was 82%, 80%, 100% and 90%, respectively (Online Supplementary Figures S6 and S7).
Prognostic relevance of monocytic switch
From a clinical perspective, it is important to know whether monocytic switch is of prognostic relevance and whether the phenotype changes during relapse. All but one
2070
haematologica | 2021; 106(8)