HMGA1 and HMGA2: epigenetic regulators in myeloproliferative neoplasms and novel platelet transcriptomic biomarkers of myelofibrosis

We read with great interest two recent studies by Handa et al. [ 1 ] and Bartalucci et al. [ 2 ] reporting novel chromosomal deletions within the 3’ untranslated region (UTR) of HMGA2 (Δ3’ HMGA2 ) in myeloproliferative neoplasms (MPNs). Both teams identified submicroscopic deletions ranging from 1.3 kb to 2.7 Mb that encompass the MIRLET7 microRNA tumor suppressor response element. These alterations are predicted to induce HMGA2 expression by releasing MIRLET7-mediated repression of HMGA2 in myelofibrosis (MF). In light of this important work, we performed integrative analysis of single cell and platelet transcriptomic datasets from independent MPN patient cohorts to further define the potential impact and distinct function of HMGA1 and HMGA2 family members during MPN progression. HMGA1 and HMGA2 genes encode highly homologous oncofetal proteins and chromatin regulators that cause clonal expansion in blood cells [ 3 ]. In transgenic mouse models, hematologic phenotypes range from myeloproliferation following Hmga2 overexpression in hematopoietic stem cells (HSCs) [ 4 ] to leukemia induced by Hmga1a overexpression in T and B cell progenitors [ 5 ]. Importantly, HMGA1 is also required for JAK2 V617F mutant leukemogenesis in mouse models and Hmga1 haploinsufficiency in the hematopoietic compartment is sufficient to prevent progression to myelofibrosis (MF) in JAK2 V617F mice [ 6 ]. Hmga2 also cooperates with JAK2 V617F in the development of MPN in mouse models [ 7 , 8 ]. Moreover, HMGA1 and HMGA2 are overexpressed in CD34 + cells and differentiated progeny from patients with MPN, with highest expression levels in more advanced disease [ 6 , 7 , 8 , 9 , 10 , 11 ]. Using high-resolution array comparative genomic hybridization and single nucleotide polymorphisms (aCGH + SNP), Handa et al. discovered Δ 3’HMGA2 in 11% of patients with MF and advanced stage MF with increased blasts (denoted accelerated phase/blast phase (MF-AP/BP) [ 1 ]. Most importantly, Δ3’ HMGA2 deletions were associated with more aggressive clinical phenotypes and co-occurred with ASXL1 and CALR mutations in their cohort [ 1 ]. They also demonstrated HMGA2 overexpression in peripheral blood mononuclear cells from a cohort of MF patients ( n = 31) compared to healthy donors ( n = 10) with highest mean HMGA2 in samples harboring Δ 3’HMGA2 . Notably, HMGA2 overexpression was also detected in MF lacking Δ3’ HMGA2 , indicating that alternative mechanisms drive HMGA2 overexpression in MPN. This study [ 1 ] was the first to report 3’ UTR HMGA2 microdeletions as a molecular mechanism for HMGA2 overexpression in MPN. Given the potential pathogenic significance of Δ3’ HMGA2 , Bartalucci et al. subsequently developed an innovative detection assay [ 2 ]. Using targeted long-read sequencing with ultra-high sequencing depth and custom-designed primers, they compared the allelic ratio (AR) of Δ 3’HMGA2 variants relative to total reads amplified by the same primers. In a cohort of MPN patients with MF [ n = 108, including 96 Primary (P) MF, 12 post-essential thrombocytosis (post-ET) MF] or MF-AP/BP ( n = 79), Δ 3’HMGA2 variants were identified in 2.1% of cases, with ARs ranging from 0.1%-21.3%. This was validated in a second, independent cohort ( n = 136, MF = 96, MF-AP/BP = 40) in which 2.9% of patients harbored Δ 3’HMGA2 variants with ARs ranging from 0.5% to 70.5%. The authors postulated that higher ARs may promote MPN phenotypes. Across both cohorts, Δ 3’HMGA2 deletions were identified in 5 PMF, 2 post-ET MF, and 1 MF-AP/BP cases, all with normal karyotypes and no chromosomal alterations detected using conventional cytogenetics. Mutational status included CALR (Type 1-like; PMF), CALR and ASXL1 (PMF), triple-negative (PMF), and JAK2 V617F (Post-ET MF, PMF, MF-AP/BP). Among the JAK2 V617F cases, additional co-mutations included ASXL1 , GATA2 , and SH2B3 (MF-AP/BP) or SH2B3 , TP53 , and NRAS (PMF). Of the 8 cases with Δ 3’HMGA2 , one ( JAK2 V617F PMF) progressed to MF-AP/BP over a median follow-up of 38 months. Although HMGA2 expression was not available from all cohorts, the results from Handa et al. and Bartalucci et al. suggest that Δ 3’HMGA2 deletions are rare but potentially important drivers in MF and disease progression by inducing HMGA2 [ 1 , 2 ]. Handa et al. also reported HMGA2 overexpression by single cell RNA sequencing in CD34 + cells from untreated MF patients (GSE144568) [ 11 ] in an HSC-enriched cluster ( n = 6; 4 JAK2 V617F ; 2 CALR ; ages 54-73 years) compared to those from a subset of healthy donor controls ( n = 4; ages 55–67 years). Notably, the status of Δ 3’HMGA2 in these samples was unknown. We previously demonstrated that HMGA1 is overexpressed in MPN CD34 + cells compared to healthy controls, with the highest HMGA1 expression in advanced disease (MF, leukemia) across multiple cohorts [ 6 ]. When comparing all CD34 + cells from MF patients ( n = 15) from the same cohort (GSE144568), including untreated ( n = 6) and previously treated MF ( n = 9) to all healthy controls ( n = 6; ages 52–67), HMGA1 is significantly overexpressed compared to controls (Fig. 1A ) [ 6 ]. Subset analysis restricted to untreated patients and healthy controls analyzed by Handa et al. also shows HMGA1 overexpression (Fig. 1B ) Similarly, HMGA2 is overexpressed in CD34 + cells in MF patients compared to controls (Fig. 1A, B ). Fig. 1: HMGA1 and HMGA2 are enriched in different MPN CD34 + single cell clusters and linked to activation of distinct pathways. Full size image HMGA1 and HMGA2 transcripts in single CD34 + cells (GSE144568) from healthy, age-matched controls ( n = 6) and MPN MF patients ( n = 15) with JAK2V617F ( n = 12) or CALR mutations ( n = 3) show overexpression of HMGA1 and HMGA2 , although fewer cells express HMGA2 and transcript abundance for HMGA2 is lower compared to HMGA1 ( A ). Subset analysis of age-matched controls ( n = 4) and untreated MPN MF patients ( n = 6) with JAK2V617F ( n = 4) or CALR ( n = 2) shows overexpression of HMGA1 and HMGA2 ( B ). Uniform manifold approximation and projection maps (UMAPs) show clusters based on distinct transcriptomes of single cells from control, healthy donors (HD; blue), JAK2V617F MF (red) or CALR MF (green) ( C ; left ). We identified 12 distinct clusters (Seurat; resolution 0.25), and imputed cell identities from established markers ( C ; right ). Feature plots for each HMGA gene ( HMGA1- upper; HMGA2– lower ) show that more cells express HMGA1 and transcript abundance is higher compared to HMGA2. HMGA1 is most abundant in clusters C-Mk, C-MEP, and J-MyE2. Compared to its expression in other clusters, HMGA2 is most abundant in J-My, J-MyE1, and HD-HSPC ( D ). Dot plots show relative HMGA1 and HMGA2 transcript abundance (intensity) and frequency of cells expressing each transcript (circle size) in each cluster ( E ). Violin plots show the distribution of single cells expressing each HMGA transcript by cluster; central dots represent the median and error bars denote the interquartile range (25th–75th percentile) ( F ). Enrichment plots show the normalized enrichment score (NES), normalized p value (NOM p-val), and false discovery rate q value (FDR q-val) for the most significant pathways ( G ). Fewer cells express HMGA2 and fewer pathways are significant for HMGA2- expressing cells. **** p < 0.0001. To better understand HMGA1 and HMGA2 function in MPN, we compared their expression at the level of single cell transcriptomes from MF and control CD34 + cells. Intriguingly, HMGA1 is detected in a greater proportion of MPN CD34 + cells (~87%) as compared to HMGA2 (~29%) and HMGA1 transcripts are more abundant per cell compared with HMGA2 (Fig. 1A, B ). To explore the stem and progenitor context for each gene, we used clustering (Seurat; resolution 0.25) to identify CD34 + stem and progenitor subsets (Fig. 1C ). We also generated feature plots, dot plots, and violin plots to compare HMGA1 and HMGA2 expression to each other within each cluster (Fig. 1D-F ). While most clusters from CD34 + cells express HMGA1 , HMGA2 expression was lower and more spatially restricted, suggesting distinct functions for HMGA1 and HMGA2 (Fig. 1D-F ). To further explore these differences, we performed gene set enrichment analysis (GSEA) comparing pathways associated with CD34 + cells expressing HMGA1 or HMGA2 to those lacking HMGA1 or HMGA2 , respectively. Strikingly, HMGA1- expressing cells were enriched for pathways previously linked to HMGA1 downstream networks, including those involved in proliferation (E2F, cMYC, G2M) and oxidative phosphorylation [ 6 , 12 , 13 ]. By contrast, HMGA2- expressing cells were enriched for inflammatory signaling pathways (TNF-α, TGF-β, Epithelial-Mesenchymal Transition) (Fig. 1G ). When analyzing CD34 + cells from healthy controls via GSEA, we also identified proliferative pathways associated with HMGA1 and inflammatory signaling pathways associated with HMGA2 . The association of HMGA1 with cell cycle progression and metabolic pathways promoting HSC cycling suggests that HMGA1 may drive clonal expansion in MPN. By contrast, the inflammatory pathways (TNF-α, TGF-β) linked to HMGA2 expression are consistent with a role in cytokine signaling to promote progression and fibrosis. Together, these results suggest different stem and progenitor cell contexts and functions for HMGA1 and HMGA2 in MPN. To further investigate HMGA1/2 in JAK2 V617F MPN, we compared expression of each gene in CD34 + cells harboring JAK2 V617F with matched cells from the same patients lacking the JAK2 V617F mutation (GSE122198) [ 14 ]. HMGA1 is overexpressed in JAK2 V617F cells compared to JAK2 wildtype (WT) cells from matched patients (Fig. 2A ; left ). HMGA2 also increases in the JAK2 V617F CD34 + cells compared to WT CD34 + cells (Fig. 2A ; right ), though the difference did not reach statistical significance. Similar to the cohort GSE144568, HMGA2 was detected in fewer cells and with lower transcript abundance compared to HMGA1 . Given that overexpression of HMGA genes is linked to tumor progression and adverse outcomes in diverse settings [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 ], we also examined platelet transcripts [ 15 ]. Moreover, platelet transcriptomes provide an accessible context as an MPN biomarker [ 15 ]. HMGA1 transcripts increase progressively in ET, PV, and MF platelets compared to platelets from healthy controls with significant enrichment in MF compared to controls (p < 0.0001) and/or ET/PV (Fig. 2B, C ). HMGA2 also increases with higher mean expression restricted to MF ( p = 0.0045, Fig. 2B ). Similar to CD34 + cells, HMGA2 transcripts were detected in fewer platelets and at lower abundance than HMGA1 transcripts (Fig. 2B ). There was also greater heterogeneity in HMGA2 transcript levels across diagnostic groups as demonstrated by higher coefficients of variations (CV = standard deviation/mean x 100%) compared to HMGA1 . The CV for HMGA1 ranged from 6.53% to 9.16%, whereas the CV for HMGA2 was 34.43% to 53.13%, reflecting a lower signal-to-noise ratio with HMGA2 . Fig. 2: HMGA1 and HMGA2 are overexpressed in JAK2V617F CD34 + single cells and platelet transcripts from MF MPN patients. Full size image Both HMGA1 and HMGA2 transcripts increase in JAK2V617F mutant CD34 + single cells compared to expression in wildtype (WT) CD34 + cells from a cohort of matched patient WT and CD34 + cells (GSE122198). Only cells expressing HMGA1 or HMGA2 are shown. A greater percentage of cells express HMGA1 and transcripts are more abundant per cell compared to HMGA2 ( A ). Platelet transcriptomes show HMGA1 in healthy controls (CTRL, n = 21) and MPN, with increasing levels in ET ( n = 24), and PV ( n = 33), and highest levels in MF ( n = 42) using one-way ANOVA ( p < 0.0001) followed by multiple students t tests ** p < 0.01, *** p < 0.001, **** p < 0.0001). Coefficients of variation (CV) are shown by gene below each diagnostic category ( B, left ). By contrast, HMGA2 expression is similar in CTRL, ET, and PV platelets with significantly higher mean levels in MF platelets (one-way ANOVA, p = 0.03) followed by planned comparison of CTRL + ET + PV versus MF ( p = 0.0045) ( B, right ). Proportion stacked bar plots demonstrate progressive enrichment for both high HMGA1/HMGA2 or high HMGA1 alone with increasing disease severity ( C ). Comparison of biomarkers (both high HMGA1/HMGA2 or high expression of either alone) shows that high HMGA1 expression predicts MF with greatest sensitivity ( D ). To determine whether elevated platelet transcripts of HMGA1 and/or HMGA2 could serve as a biomarker for MF in this cohort, we dichotomized HMGA1 /2 expression to high (≥median) or low (<median). Strikingly, high HMGA1 demonstrates a strong association with MF, with the greatest sensitivity (85.7%) for MF compared with ET, PV, and all non-MF cases (Odds ratio 13.5, Fisher’s exact test p < 10 −8 , Fig. 2C, D ). Since high HMGA2 expression was detected in control platelets, it showed limited discriminatory value. Combined high expression of HMGA1 and HMGA2 was also strongly associated with MF (Odds ratio 11.7, Fisher’s exact test p < 10 −6 ) and demonstrated the highest specificity (89.7%), although the sensitivity (57.1%) was lower than that observed for HMGA1 alone (85.7%). While these findings suggest that HMGA genes within platelet transcriptomes could provide clinically useful biomarkers for MF, the current analysis was derived from a single cohort with limited sample sizes across diagnostic groups (healthy donors, n = 21; ET, n = 24; PV, n = 33; MF, n = 42) and demonstrated heterogeneity in gene expression, particularly for HMGA2 . Given the lack of reliable biomarkers to predict progression to MF, however, further validation of HMGA genes alongside existing prognostic markers in larger, independent cohorts is warranted. Together, these studies illuminate HMGA1 and HMGA2 genes as plausible drivers of MPN progression via distinct mechanisms and within different stem and progenitor cell contexts. HMGA genes were first recognized as oncogenes over 25 years ago in cell-based transformation assays [ 3 ]. Subsequent transgenic mouse studies revealed that overexpression of Hmga1 or Hmga2 induces clonal expansion, leukemogenesis, or myeloproliferation depending on the cellular context of transgene expression [ 4 , 5 , 6 , 7 , 8 ]. We reported that loss of just a single Hmga1 allele in JAK2 V617F mutant mice attenuates erythrocytosis and thrombocytosis while preventing splenomegaly and progression to MF, whereas Hmga2 deficiency had no impact on MPN phenotypes in this model [ 6 ]. Beyond hematologic malignancies, HMGA overexpression is also linked to metastatic progression and adverse outcomes in diverse solid tumors [ 12 , 13 ]. Importantly, the novel Δ 3’HMGA2 deletion events can be detected by cCGH + SNP [ 1 ] or an innovative and sensitive, long-read sequencing assay [ 2 ]. This work also reveals a new potential mechanism for HMGA2 overexpression during MPN progression, although HMGA2 is also up-regulated in samples lacking the Δ 3’HMGA2 deletion. Similar to HMGA1 , diverse oncogenic signals induce HMGA2 expression, such as inflammatory growth factors or loss of tumor suppressor microRNAs that normally repress their expression [ 3 , 6 ]. Additional mechanisms, including activation by oncogenic transcription factors or epigenetic de-repression of silenced chromatin, may also up-regulate HMGA1 /2 expression in specific contexts. Surprisingly, although HMGA1 and HMGA2 are highly homologous proteins that occupy similar DNA sequences, they associate with distinct pathways known to drive myelofibrosis pathogenesis. The cell cycle progression pathways linked to HMGA1 could foster clonal expansion, while inflammatory signaling via TGF-β associated with HMGA2 may promote fibrosis. Emerging evidence also suggests that clonal expansion in the setting of inflammatory signals provides an optimal environment for HSCs to acquire additional mutations. Thus, it is plausible that these divergent pathways associated with HMGA1 and HMGA2 cooperate in promoting advanced stage disease and/or progression in MPN patients. Since high HMGA1 expression was absent in platelet transcriptomes from healthy controls, and HMGA1 gradually increases with more advanced MPN phenotype in this cohort, high HMGA1 alone was able to distinguish MF from less advanced disease. Combined high expression of HMGA1 /2 also associated with MF, although HMGA2 transcripts were lower and heterogenous. Importantly, neither the platelet nor single cell transcriptomes included analyses for Δ 3’HMGA2 deletions, and the relative paucity of cases with these deletions may have contributed to lower HMGA2 levels in these cohorts. Results from Handa et al. and Bartalucci et al. suggest that Δ 3’HMGA2 deletions are more specific to MF than high HMGA2 levels. Because platelet transcriptomes are readily accessible and amenable for biomarker development, further studies identifying signatures that predict MF would be clinically relevant and could complement existing clinical and molecular prognostic markers. Chromosomal alterations affecting the HMGA2 3’ UTR have been reported in acute myeloid leukemia and myelodysplastic syndromes, indicating that this mechanism may impact diverse hematologic malignancies [ 1 , 2 ]. Translocations involving HMGA1 were also identified in myeloid malignancy and variants in the 3’ noncoding region associate with increased risk for developing clonal hematopoiesis or MPN [ 6 ], although translocations and chromosomal alterations are less common at the HMGA1 locus compared to that of HMGA2 . While the molecular underpinnings remain incompletely understood, HMGA genes encode DNA binding proteins that modulate chromatin structure to activate or repress gene expression, including genes involved in self-renewal, clonal expansion, plasticity, differentiation, and inflammation. Collectively, the findings described here shed light on distinct functions and cellular milieus for HMGA1 and HMGA2 epigenetic regulators in MPN. Given their fundamental roles in stem cell and MPN biology, further studies are warranted to identify therapeutic approaches to target these proteins or their downstream pathways in MPN. Although direct pharmacological inhibition of oncogenic transcription factors has been challenging, these studies suggest that HMGA proteins could provide impactful therapeutic opportunities and biomarkers relevant not only to MPN, but also across other hematologic malignancies and solid tumors.