A novel class of ZNF384 aberrations in acute leukemia

Insight into the genomic landscape of childhood B-cell precursor acute lymphoblastic leukemia (BCP-ALL) has significantly deepened over the past decade with the identification of novel genetic subtypes. This particularly applies to the “B-other” ALL subgroup (BCP-ALL negative for “classical” aberrations: hyper/hypo-diploidy, ETV6-RUNX1, TCF3-PBX1, BCR-ABL1 fusions, and KMT2A gene rearrangements [r]) comprising approximately one-fourth of childhood BCP-ALL.

The ZNF384 gene encoding transcription factor zinc-finger protein 384 is rearranged in 5% to 10% of B-other leukemias in the form of fusion to >10 different partners (most frequently TCF3, EP300, TAF15, or CREBBP).^1-7 Within these “canonical” fusions, the entire ZNF384 coding region is fused in-frame to the partner coding sequence and forms the 3' part of the fusion gene/C-terminus of the resulting fusion protein .

The ZNF384r leukemias have a predominant BCP immunophenotype, although typical immature B-lineage marker CD10 is only weakly expressed in some patients, and myeloid markers (CD13/CD33) are often present.^2,4,5,7,8  The immunophenotype of ZNF384r leukemia often fulfills the criteria for mixed phenotype acute leukemia (MPAL), and the ZNF384r leukemias comprise approximately half of all childhood B/myeloid MPAL cases.^6,9 

The unique gene expression profile (GEP) of ZNF384r leukemias is enriched for hematopoietic stem cell and immature myeloid lineage features and reflects upregulation of the JAK-STAT signaling pathway.^4,5,10  In studies using the GEP to cluster B-other ALL, the ZNF384r cluster has often included cases without detectable ZNF384 rearrangement.^6,10,11  In some of these “ZNF384r-like” leukemias, rearrangement of ZNF362 (paralog of ZNF384) was identified.¹⁰  However, in some “ZNF384r-like” cases, no established primary aberration has been found.

Here, we describe leukemias with the “ZNF384r-like” GEP and previously undiscovered ZNF384 aberrations disrupting the C-terminal portion of the encoded protein.

This study is based on a patient cohort reported in our previous paper⁶  and enlarged by recently diagnosed consecutive B-other patients. A total of 643 children were diagnosed with BCP-ALL or B/myeloid MPAL in the Czech Republic between December 2010 and December 2020, including 158 classified as B-other ALL. Hierarchical clustering analysis (HCA) was performed for 110 cases from this cohort that were diagnosed from December 2010 to December 2017, as published previously,⁶  as well as patients with the ZNF384 aberration diagnosed before (n = 3) and after (n = 4) this period. Analyses of the biological and clinical features of ZNF384 leukemias also included another 4 patients diagnosed before December 2010 using RT-PCR.

Routine diagnostics, whole-transcriptome sequencing (RNA-seq), including GEP and HCA, whole-exome sequencing, and single-nucleotide polymorphism array analyses were performed as described previously.^6,12,13  Variant calling was performed using VarScan¹⁴  and Samtools (http://samtools.sourceforge.net/). Fusion calling was performed using TopHat,¹⁵  deFuse,¹⁶  and Cicero.¹⁷ 

For more detailed methods, see the supplemental material.

Unsupervised HCA using the RNA-seq–based GEP data identified 3 patients that coclustered with ZNF384r leukemias but lacked canonical ZNF384 fusion (Figure 1A).

Routine analysis of RNA-seq data using in-house bioinformatic pipelines for single-nucleotide variants, indels, and fusion gene detection revealed noncanonical ZNF384 fusion (ZNF384 as the 5' partner) in 1 patient, but no aberration in ZNF384 or ZNF362 was found in the remaining 2 patients. We “manually” reanalyzed the reads mapping to the respective genes in these 2 patients and identified in-frame indels in the last exon of ZNF384 in both. Therefore, unlike canonical ZNF384 fusion, all the novel aberrations resulted in alteration of the C-terminal portion of the ZNF384 protein.

Patient 1 harbored a novel ZNF384-TEX41 fusion consisting of the large 5' portion of ZNF384 (including the coding sequence for all 6 zinc fingers) fused to the 3' portion of TEX41, a gene not shown to be recurrently involved in leukemia (see supplemental material). The out-of-frame fusion led to an early stop codon with only 9 amino acids (AAs) encoded by TEX41 intron 1 (ZNF384 p. 481_516 > SFQDYCACQ). In patients 2 and 3, the indels led to replacement of 5 and 2 AAs by 3 and 8 new AAs, respectively, with the remaining 119 and 32 C-terminal AAs maintained from the wild-type ZNF384. In patient 2, the indel (c.1179_1192del14insGGGCAGGG; ZNF384 p.393_397KHNPP>RAG) had an allelic frequency 47% in RNAseq and was located just at the end of the last zinc finger. In patient 3, the indel (c.1448_1451del4insATTTTGAGTGGGACCCGAACGA; ZNF384 p.483_484CS>YFEWDPND) had an allelic frequency 52% and 43% in RNAseq and whole-exome sequencing, respectively (Figure 1B). More detailed results of the molecular analyses are provided in the supplemental material.

To exclude the presence of similar mutations (missed by variant or fusion callers employed in our bioinformatic pipeline) in other patients, we manually reanalyzed the reads mapping to ZNF384 in all remaining B-other cases included in our HCA (n = 105), but no further ZNF384 aberration was found. Combined with the high-allele frequency of ZNF384 indels, the findings suggest that the novel ZNF384 aberrations represent primary genetic hits, mutually exclusive with other known primary aberrations.

In accordance with the immunophenotypic features of ZNF384r leukemias, all 3 patients expressed myeloid markers, and CD10 expression was weak in 2. Two of the 3 patients were classified as MPAL. Notably, besides other myeloid markers, the 3 patients with novel ZNF384 aberrations expressed CD66c, which is rarely expressed in ZNF384r ALL (Table 1).⁸ 

ZNF384r leukemias represent 6% (10/158) of B-other ALL and 1.6% (10/643) of BCP-ALL cases in the cohort; when combining these with the novel ZNF384 aberrations, the frequency is 8% (12/158) and 1.9% (12/643), respectively. The leukemias with novel ZNF384 aberrations comprise 17% (2/12) of consecutively diagnosed cases with ZNF384 aberrations. The basic characteristics of the cases with ZNF384 aberrations are given in supplemental Table 1.

The chimeric ZNF384 oncoproteins studied thus far exhibit perturbed DNA binding and drive transcriptional deregulation.^2,9  DNA binding is likely preserved in ZNF384 with novel aberrations, as the 6 zinc fingers are mostly intact. However, as the more distal C-terminal portion of the protein affected by the novel aberrations lacks well-defined functional domains, functional studies would be necessary to directly demonstrate the biological consequences of these aberrations and to compare them with canonical ZNF384 fusions.

Despite indisputable progress in the analysis of genomic data, some “blind spots” insufficiently covered by widely used sequencing/analytical approaches may still exist in the leukemia mutational landscape. In our experience, particularly the detection of larger or more complex indels remains challenging when using “standard” variant callers. We cannot exclude that some other variant callers, which we did not test, may have found these mutations. However, the fact that similar ZNF384 mutations were not reported in any genomic study published thus far, or in publicly available databases, supports our belief that this type of mutation is easily missed in sequencing data by most automated approaches.

In conclusion, we presented childhood ALL/MPAL patients with a novel class of ZNF384 aberrations that phenocopy leukemias defined by canonical ZNF384 fusions and, thus, represent the same biological subtype. We think that the novel ZNF384 aberrations can recur enough to not be neglected in future leukemia diagnostics. Therefore, targeted reanalysis of available sequencing data in samples with the ZNF384r ALL gene expression signature and seemingly unaffected ZNF384/ZNF362 is highly encouraged in order to further assess the frequency of ZNF384 aberrations other than canonical fusion, and to provide additional data on the biological and clinical features that could aid in understanding the genomic landscape of childhood leukemia.

This study was supported by grants from the Czech Health Research Council (NU20-07-00322) and Charles University (UNCE 204012) and by the project (Ministry of Health, Czech Republic) for conceptual development of research organization 00064203 (University Hospital Motol, Prague, Czech Republic). Research infrastructure was supported by the NCMG - LM2018132.

Contribution: M. Zaliova and J.T. designed the study; M. Zwyrtkova performed NGS; M. Zaliova, L.W., J. Stuchly, K.F., and P.T. analyzed transcriptomic and genomic data; O.H., J. Starkova, L.S., J. Stary, J.T., and J.Z. provided diagnostic and clinical data; all authors participated on data integration, interpretation, and presentation; M. Zaliova and J.Z. wrote the draft; and all authors revised the draft and contributed to the final manuscript.

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

Correspondence: Marketa Zaliova, Department of Paediatric Haematology and Oncology, Second Faculty of Medicine, Charles University and University Hospital Motol, V Uvalu 84, 150 06 Prague, Czech Republic; e-mail: marketa.zaliova@lfmotol.cuni.cz.