Role of HMGA2 Overexpression Caused By Specific Genetic HMGA2 Aberrations in Disease Progression in Calreticulin-Mutated Myelofibrosis Patients

High mobility group AT hook 1 and 2 (HMGA1 and HMGA2) have been implicated as drivers of disease progression in patients with JAK2^V617F+ myeloproliferative neoplasms (MPNs) (Andrieux et al. Genes Chromosomes Cancer 2004, Li et al. Blood 2022, Dutta et al. Blood 2017, Ueda et al. Blood Adv 2017, Guglielmelli et al. Stem Cells 2007). Others have linked HMGA2 upregulation to loss-of-function mutations in the epigenetic modifiers EZH2 and ASXL1 (Sashida et al. J Exp Med 2016, Yang et al. Blood 2016, Shimizu et al. J Exp Med 2016, Ueda et al. Blood Adv 2017). We determined that absolute transcript levels of HMGA1 exceeded those of HMGA2 in myelofibrosis (MF) patients and normal donors (NDs). The relative levels of HMGA2 but not HMGA1 transcripts measured by qRT-PCR were significantly increased in MF mononuclear cells (MNCs) as compared to ND (27.4-fold, p=0.003 vs 2.6-fold, p=0.5). We validated these findings by mining publicly available single cell (sc) RNA-seq data (GSE144568, Psaila et al. Mol Cell 2020) where elevated HMGA2 expression was seen in MF as compared to ND CD34⁺ cells with expression being highest in hematopoietic stem cells (HSCs; 1.23-fold, p.adj=5.3x10^-8). Immunohistochemical analysis of bone marrow biopsies failed to detect HMGA1 and HMGA2 expression in NDs. Both proteins were seen in MF but with distinct expression patterns. HMGA1 showed strong nuclear expression in both small lymphoid-like cells and atypical megakaryocytes (MKs). In contrast, HMGA2 failed to label MK nuclei but stained MK cytoplasm with weak intensity and rare small lymphoid-like cells strongly.

We performed additional studies to further define the genetic events responsible for HMGA1 and HMGA2 overexpression (OE) in MF. HMGA2OE in MPNs has been previously attributed to balanced translocations involving chromosome 12 at breakpoints 12q13-15 (Marquis et al. Blood Cancer J 2018). We have identified novel genomic deletions at 12q14.3 (named Δ3’HMGA2) in 9% of our cohort of 169 MF patients, which were not found in PV, ET, MDS/MPN or MPN-U patients. Such deletions result in loss of most of the 3’UTR in HMGA2 exon 5 that includes MIRLET7 binding sites leading to HMGA2OE. By performing scRNA-seq, we were able to identify the cells harboring Δ3’HMGA2 in MNCs from MF patients with such deletions but not from patients with normal HMGA2. Structural abnormalities involving HMGA1 at 6p21.31 were not observed. The relative levels of HMGA2 transcripts were significantly higher in MNCs of MF patients with HMGA2 aberration (Δ3’HMGA2 or HMGA2 rearrangement but not Δ3’HMGA2) than those lacking such changes (2.3-fold, p=0.008).

While HMGA2 expression level was not influenced by the MPN driver mutation (p=0.5), HMGA2 abnormalities occurred more frequently in patients with non-JAK2 driver mutations (63%), predominantly CALR (58%), which contrasts with their general distribution (57% JAK2 and 28% CALR). In fact, 23% (11/48) of all CALR⁺ MF patients carried HMGA2 aberrations.

ASXL1 mutations frequently co-occurred in MF patients with an HMGA2 aberration (84%; none had EZH2 mutations) as compared to only 28% (42/150) in patients lacking an HMGA2 abnormality (p≤0.001). We found no difference in HMGA2 levels in MF patients lacking an HMGA2 aberration based on the presence or absence of ASXL1 mutation/deletion (p=0.95). However, HMGA2 levels were significantly higher in MF patients with co-occurring HMGA2 and ASXL1 aberrations as compared to the prior two groups (3.3-fold, p=0.0002; 4.1-fold, p=0.00001). Furthermore, 42% (8/19) patients with HMGA2 abnormalities developed >5% peripheral blood blasts and 5 (26.3%) progressed to MPN-accelerated phase/blast phase (AP/BP) during their clinical course. Of these, four patients developed MPN-AP/BP within 3 years of detection of Δ3’HMGA2. Conversely, only 7.1% (3/42) MF patients who lacked Δ3’HMGA2 progressed.

We conclude that HMGA1 and HMGA2 each play a role in MF progression, with structural abnormalities involving MIRLET7 binding sites of HMGA2 occurring more frequently in MF patients with CALR mutations. These abnormalities result in HMGA2OE independent of concurrent mutations in ASXL1 or EZH2, and their frequency differs depending on the type of MPN driver mutation. Mining scRNA-seq data from patients with and without Δ3’HMGA2 will enable us to identify possible lineage skewing, differentially expressed genes and regulatory networks driven by HMGA2OE that control MF progression.