To the Editor

In a recent issue of the journal, Handa S. et al. reported novel 12q14.3 deletions in Myelofibrosis (MF) and Myeloproliferative Neoplasms (MPN)-accelerated/blast phase (AP/BP) patients [1], leading to loss of most exon 5 non-coding regions and MIRLET7 binding sites in 3’UTR of HMGA2, resulting in HMGA2 transcript overexpression. These deletions were found in 19 of 169 patients (11.24%), caused by either translocation (n = 4), inversion (n = 1), combined deletion and translocation (n = 1) or in most instances (14/19 cases) caused by submicroscopic 3’UTR mutations, ranging 1.3–3.1 Kbp (12/14 cases), or larger deletions of 87 Kbp and 2.7 Mbp (2/14 cases). As reported by authors, the HMGA2-associated 12q alterations correlated with more aggressive disease and MPN-AP/BP progression, especially when co-occurring with ASXL1 [2] and CALR [3] mutation. Given the relevance of those findings, we sought to develop a friendly laboratory approach that might facilitate detection of HMGA2 deletions (Δ3’HMGA2) in clinical practice. To this end, we designed a comprehensive workflow that integrates a targeted Long-Read Sequencing assay with ultra-high sequencing depth and an automated analytical pipeline. The procedure is based on custom primers (FW:5’-TCCCTTCACAGTCCCAGGTTTAG-3’, Rev:5’-GGTTCTGGCAGTGCAGCATTC-3’) to generate a DNA amplicon spanning HMGA2 exon 5, including the 3’UTR (GRCh38:chr12:65963494-65968229). Resulting HMGA2 amplicons sizing 4735 bp are obtained using LongAmp Taq-Polymerase (New England Biolabs) and further barcoded (Oxford Nanopore Technology, SQK-NBD114.96) to allow parallelization of multiple samples in a single sequencing run. The samples are sequenced by ONT-PromethION using live basecalling option (ONT-Dorado, v0.9.1). The bioinformatic pipeline (freely available at https://github.com/blackSquare225/RdD) involves FASTQ demultiplexing and quality filtering through a Snakemake-based pipeline and alignment to GRCh38 using Minimap2 (v2.28). The pipeline’s core component is constituted by a custom, user-configurable, Python tool that enables the identification of all the most represented isoforms (≤4.7Kbp) of Δ3’HMGA2 [1]. Mapped reads (if quality ≥25) are analyzed by a union-find-based clustering algorithm that groups deletions based on overlapping or nearby start and end coordinates, and then are filtered by the number of supporting reads (Δ3’HMGA2-reads threshold: ≥50). The allelic ratio (AR) of Δ3’HMGA2 is calculated as the percentage of variant-supporting reads on total reads collected and the results are output as an interpretable Δ3’HMGA2-calling report. Sequencing data were visually inspected through Integrated Genome Viewer (IGV) and validated by Sanger method (Fig. 1). The procedure was first tested on a discovery cohort (Cohort-1) of 187 samples including 108 MF (96 Primary (P)-MF and 12 post-ET MF) and 79 BP (15 ET-BP, 21 PV-BP and 43 MF-BP) and further validated in the Cohort-2 comprising 136 patients of which 96 were MF (66 PMF and 30 post-ET MF) and 40 were BP (4 ET-BP, 5 PV-BP, 6 MF-BP and 25 secondary-AML from Myelodisplasia (sAML)). All patients included had provided informed signed agreement (local IRB approval #14560) as part of the AIRC-MYNERVA project (#21267). The peripheral blood or bone marrow mononuclear samples were available in our institutional repository. The median on-target read depth was 91,284x (95%CI:15,691-291,322) and 72,864x (95%CI:12,544-164,352x) in Cohort-1 and Cohort-2, respectively. In Cohort-1 we found a deletion event corresponding to Δ3’HMGA2 in 4 (2.14%) cases. The detected Δ3’HMGA2 were in size 1666 bp, 1475 bp, 2133 bp and 2612 bp, with an AR of 21.3%, 7.9%, 0.6% and 0.1%, respectively. Among the 4 Δ3’HMGA2-mutated patients, 2 were PMF, 1 post-ET MF and 1 MF-BP; all had normal karyotype with no chr12q alteration at conventional cytogenetics. The PMF patients were one CALRT1-like mutated, presenting with splenomegaly and grade III BM fibrosis and no additional myeloid mutation, the second a triple-negative presenting with normal spleen and grade II BM fibrosis, who harbored an additional NF1 mutation. The post-ET MF case was JAK2V617F mutated showing splenomegaly and grade II BM fibrosis. The MF-BP patient was a JAK2V617F mutated male with additional mutations of ASXL1, GATA2 and SH2B3. Similar findings were obtained in Cohort-2, where we detected a Δ3’HMGA2 in 4 patients (2.94%) with size of 3375 bp, 1879bp, 991 bp, and 981 bp, and AR of 0.4%, 70.5%, 28.6%, 0.5%, respectively. Among the 4 mutated patients, 3 had PMF and 1 had post-ET MF all with a normal karyotype. Notably, 2 PMF patients had mutations in both CALR and ASXL1, one of whom also carried additional mutations in IDH1, CBL, and NRAS. The other PMF patient harbored JAK2V617F and mutations in SH2B3, TP53, and NRAS, while the post-ET MF patient had a JAK2V617F mutation. Considering all the Δ3’HMGA2-mutated patients in both cohorts, none showed disease progression (median follow-up: 38 months) except for a JAK2V617F PMF patient (Cohort-2), who progressed to MF-AP. Table 1 shows the clinical and molecular characteristics of Δ3’HMGA2 patients of Cohort-1 and 2.

Fig. 1: Visualization of Δ3’HMGA2 by Integrated Genome Viewer and Sanger method.
figure 1

The image shows the Δ3’HMGA2 as visualized by long-read and Sanger sequencing. The Δ3’HMGA2-supporting reads generated by the long-read assay were sorted and imaged through Integrated Genome Viewer (IGV). The snapshots show also chromosome cytoband, reference coordinates, the coverage track and Δ3’HMGA2 size. Panel (A, B) show the deletion of HMGA2 observed in the post-ET MF (Cohort1_4.23) and MF-BP (Cohort1_enb.01) patients of Cohort-1 with GRCh38 coordinates chr12:65964148-65965814 (AR:21.3%) and chr12:65964116-65966249 (AR:0.6%), respectively. Panel (C, D) show Δ3’HMGA2-supporting reads belonging to the post-ET MF (Cohort2_2.56) and PMF (Cohort2_1.9) patients of Cohort-2 with coordinates chr12:65964074-65965055 (AR: 0.5%) and chr12:65964404-65966283 (AF:10.5%), respectively. Panel (E) shows a representative sequence of Δ3’HMGA2, as determined by Sanger sequencing of the PMF patient Cohort2_1.9.

Full size image
Table 1 Clinical, follow-up and molecular information on Δ3’HMGA2-mutated patients.
Full size table

HMGA2 is crucial for fetal hematopoietic stem cells (HSCs) self-renewal and is suppressed by MIRLET7 and PRC2 in adult HSCs [4]. Our group previously reported HMGA2 dysregulation in MF patients through gene expression analysis of CD34+ cells [5]. Its role in MF pathogenesis is further supported by evidence that 3’UTR truncation gives HSCs a clonal advantage in-vivo [6]. The results reported by Handa and colleagues ultimately point the attention to HMGA2 mutations as a predictive biomarker of more advanced/aggressive disease, as well as a potential therapeutic target to halt disease progression in MF. In our cohort of 323 patients, we detected a Δ3’HMGA2 in 2.48% of cases, which was associated with CALR or ASXL1 mutation in half of them. The frequency of Δ3’HMGA2 is comparable in PMF (3.1%) and post-ET MF (4.8%) whereas only 1 mutated case out of 119 BP/sAML patients (0.8%) was found mutated. Among all mutated cases, we observed heterogeneity in the AR of Δ3’HMGA2, ranging from 70.5% to 0.1%. In the latter cases, the Δ3’HMGA2 clone is unlikely to represent the one carrying the driver mutation, highlighting the need for single-cell level studies. Overall, the frequency of Δ3’HMGA2 cases observed in our cohorts is lower than that reported by Handa et al., which may reflect unique characteristics of series and highlights the need for larger studies in well-characterized cohorts of MPN patients and other hematologic malignancies [7]. To this end, the procedure we developed might be instrumental, allowing rapid (within ≤12 hours), simultaneous analysis of up to 96 samples (with affordable cost) and an estimated Limit of Detection (LOD) of 0.05% AR (determined by serial dilution of samples) for the presence of submicroscopic Δ3’HMGA2 whose identification is challenging using cytogenetic techniques or short-read sequencing.