Knockdown of Hnrnpa0, a del(5q) gene, alters myeloid cell fate in murine cells through regulation of AU-rich transcripts

Abstract

The control of mRNA stability plays a central role in orchestrating gene-regulatory networks in hematopoietic cell growth, differentiation and tumorigenesis. HNRNPA0, which encodes an RNA-binding protein shown to regulate transcript stability via binding to the AU-rich elements of mRNAs, is located within the commonly deleted segment of 5q31.2 in myeloid neoplasms with a del(5q), and is expressed at haploinsufficient levels in these patients. We show that HNRNPA0 is normally highly expressed in hematopoietic stem cells and exhibits dynamic changes in expression during the course of differentiation. To model HNRNPA0 haploinsufficiency, we used RNAi interference in primary murine cells and an experimental cell system, and found that reduced Hnrnpa0 expression leads to a shift from monocytic towards granulocytic differentiation. Microarray-based global expression profiling revealed that Hnrnpa0 knockdown disproportionally impacts AU-rich containing transcripts and alters expression of myeloid specification genes. In therapy-related myeloid neoplasms with a del(5q), AU-rich containing mRNAs are enriched in transcripts that encode proteins associated with increased growth and proliferation. Our findings implicate haploinsufficiency of HNRNPA0 as one of the key initiating mutations in the pathogenesis of myeloid neoplasms with a del(5q), and suggest that therapies that target AU-rich elements warrant consideration in efforts to develop new mechanism-based treatment strategies.

Introduction

Therapy-related myeloid neoplasms (t-MNs) are a late complication of cytotoxic therapy typically for a primary malignant disease, and are characterized by a poor prognosis.1–3 Notably, for patients treated with alkylating agents, deletions of the long arm of chromosome 5, del(5q), or unbalanced rearrangements leading to the loss of chromosomal material from 5q, occur in approximately 40% of cases.3–5 Previously, we defined a 970 kb commonly deleted segment (CDS) at 5q31.2 that is lost in all t-MN and acute myeloid leukemia (AML) de novo patients with abnormalities of chromosome 5.6 This region contains 19 genes and 1 micro-RNA sequence; however, none of the genes revealed inactivating mutations or silencing by DNA methylation of the remaining allele.6,7 For this reason, we advanced the hypothesis that AML with a del(5q) results from haploinsufficiency of one or more genes on 5q. One del(5q) candidate of interest is the heterogeneous nuclear ribonucleoprotein A0 (HNRNPA0) gene encoding a RNA-binding protein that binds to adenylate-uridylate (AU)-rich elements (AREs) and is believed to be a positive regulator of transcript stability.8–10 In t-MNs with a del(5q), HNRNPA0 transcript levels are approximately 50% that of control subjects,11 suggesting that a reduced or haploinsufficient dosage of HNRNPA0 may be relevant to this genetic subgroup.

AU-rich RNA binding proteins (AUBPs) provide the cell with a rapid and precise mechanism to alter gene expression patterns in response to extracellular stimuli.12 Although some AUBPs direct ARE mRNAs toward rapid decay, others increase stability of their mRNA ligands, and there is growing appreciation that many AUBPs serve both functions depending on the target gene and cellular context (e.g. HuR).13 Moreover, it has been recognized that ARE-mediated decay and translational roles of AUBPs are influenced by miRNAs, the other well-known regulator of mRNA stability.14 Several AUBP mRNA targets encode proteins that regulate cell growth and survival, such as cytokines, tumor suppressors and oncoproteins, and AUBPs have been identified as key regulators of both normal and malignant hematopoiesis, including AUF, HuR, KSRP/KHSRP, nucleolin, and members of the ZFP36 family.13,15

HNRNPA0 expression is up-regulated by p38-dependent signaling in response to LPS, mycobacterial proteins, heat shock and IL-3 stimulation, and the protein is phosphorylated at Ser84 by MAPKAP-K2 downstream of p38 signaling, enabling it to bind to AU-rich mRNA targets.10,16 Modulation of mRNA stability by HNRNPA0 has also been implicated in controlling cell cycle, especially with regard to DNA damage checkpoints.9 Little is known of the role of HNRNPA0 in hematopoiesis and leukemogenesis. Given its putative role of stabilizing ARE-containing transcripts, we hypothesized that reduced function of HNRNPA0 would lead to a decrease in the stability and effective expression of target genes that may have a profound impact upon hematopoiesis, and consequently contribute to the changes seen in t-MN patients.

In this study, we now add HNRNPA0 to the growing list of AUBPs that are recognized for their vital roles in hematopoiesis and leukemogenesis. Herein, we show that HNRNPA0 is highly expressed in hematopoietic stem cells, and its expression exhibits dynamic changes during the course of differentiation. Deletion of a single allele of HNRNPA0 is associated with haploinsufficient transcript levels in CD34 cells from t-MN patients with a del(5q). Modeling HNRNPA0 haploinsufficiency in mouse cells alters myeloid lineage fate, in part through changes in the expression of ARE-containing genes, suggesting that a decreased dose of HNRNPA0 in t-MN patients may contribute to leukemogenesis. Moreover, we determined that ARE mRNAs in t-MN patients with a del(5q) are enriched in transcripts that encode proteins associated with increased growth and proliferation.

Methods

Retroviral transduction and in vitro differentiation of PUER and primary mouse hematopoietic cells

A short interfering RNA hairpin (shRNA), designed to match bases 288 to 306 of the open reading frame of Hnrnpa0 with a 9 nucleotide loop in the center, or a scrambled, irrelevant sequence,17 was cloned into pBanshee-GFP (an MSCV-based construct with P_CMV-driven GFP). The PUER cell line or mouse bone marrow cells from BALB/cAnNTac mice (Taconic, Hudson, NY, USA) were transduced by spinoculation with viral supernatants from HEK-293T cells co-transfected with the shRNA-containing pBanshee and pCL-ECO packaging plasmids (Imgenex, San Diego, CA, USA) using Effectene (Qiagen, Germantown, MD, USA). Two days post infection, the cells were sorted on a FACSAria (BD Biosciences, San Jose, CA, USA) for GFP positivity. All animal studies were approved by the University of Chicago Institutional Animal Care and Use Committee and mice were housed in a fully-AALAC-accredited facility.

GFP sorted PUER cells, expressing the Hnrnpa0 or control shRNA, were expanded in IL-3 for four days and then treated with 1–5 nM 4-hydroxytamoxifen (4-OHT, >98% (TLC) Z-isomer, Sigma, St. Louis, MO, USA) to induce macrophage differentiation. For gene expression, flow cytometry and morphological analyses, cells were sampled at 24 h, 4 days, and 7days, respectively. GFP bone marrow cells were plated in MethoCult GF M3434 (StemCell Technologies, Vancouver, BC, Canada). After 12 days of incubation, colonies of greater than 50 cells were scored for morphology by inverted light microscopy. Cells were then collected and stained with Mac-1 (CD11b) antibodies in combination with either F4/80 or Gr-1 (Ly-6G) antibodies, and analyzed on a FASCanto benchtop flow cytometric analyzer (BD Biosciences, San Jose, CA, USA).

Patient samples

All 38 t-MNs used for gene expression profiling analysis were obtained from patients treated with chemotherapy and/or radiation therapy at the University of Chicago who were diagnosed with t-MN between January 1984 and July 2007. Ten of the 38 samples had a del(5q) or unbalanced translocation leading to loss of 5q based on cytogenetic analysis. Informed consent was obtained from each patient in accordance with the Declaration of Helsinki. Study approval was obtained from the St. Jude Children’s Research Hospital and the University of Chicago Institutional Review Boards.

Gene expression profiling

GFP sorted PUER cells, transduced with retroviral contructs containing either the Hnrnpa0 shRNA or control vector, were expanded for four days and then harvested for total mRNA processing before 4-OHT induction (undifferentiated) and 24 h after treatment with 2nM 4-OHT (differentiated). Samples were hybridized to the 4x44K Whole Mouse Genome Oligo Microarray platform (Agilent, Santa Clara, CA, USA) according to the manufacturer’s protocols. Contingency analysis and gene set enrichment analysis (GSEA) were used to determine differential AU-rich gene expression. For human t-MN samples, total RNA was extracted from cryopreserved mononuclear cells from bone marrow or peripheral blood and gene expression profiling was performed using Affymetrix GeneChip HT HG-U133 PM microarrays according to the manufacturer’s instructions. Gene expression data have been deposited in Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/. Accession numbers GSE39934 (PUER) and GSE39991 (t-MN)).

Results

Expression analysis of HNRNPA0 in t-MNs with a del(5q) and during normal hematopoiesis

In myeloid neoplasms with a del(5q), expression profiling of CD34 cells revealed that HNRNPA0 transcript levels are on average 50% that of normal subjects and patients lacking abnormalities of chromosome 5 (Figure 1A), demonstrating that HNRNPA0 is expressed at haploinsufficient levels in these patients. However, there is some variability, with some patients showing up to an 80% loss of HNRNPA0 expression (Figure 1B and Online Supplementary Table S1). To determine whether HNRNPA0 may also play a role in normal hematopoiesis, we analyzed expression in human hematopoietic cells, using previously published data18 (GEO accession GSE24759). This analysis showed that HNRNPA0 is dynamically expressed during hematopoiesis (Figure 1C). HNRNPA0 is highly expressed in HSCs, CMPs and MEPs and suppressed as cells differentiate towards different hematopoietic lineages, most notable in the mature erythroid and lymphoid lineages. We found that Hnrnpa0 expression was also dynamically expressed during mouse hematopoiesis, e.g. Hnrnpa0 expression decreases as cells terminally differentiate (Online Supplementary Figure S1). Thus, HNRNPA0 is highly expressed in multipotent hematopoietic progenitors, exhibits dynamic dosage changes during the course of differentiation, and is expressed at haploinsufficient levels in myeloid neoplasms with a del(5q).

Hnrnpa0 knockdown shifts differentiation from monocytes to granulocytes

To model HNRPA0 loss in mouse cells, we developed a short-hairpin interfering RNA (shRNA) directed at the open reading frame of Hnrnpa0. RT-PCR analysis of total mouse bone marrow cells infected with the hairpin-expressing construct showed that Hnrnpa0 transcript levels were only 15–25% of the level in cells expressing the control vector (Figure 2A). A colony forming assay showed that the number of BFU E, CFU GEMM and CFU GM colonies were not significantly different between the control vector and hairpin-expressing cells (P=0.3–1.0). However, hairpin-expressing cells showed a 38% increase in CFU G colonies (P=0.01) and a 35% decrease in CFU M colonies (P=0.0001) (Figure 2B). These results were confirmed with flow cytometric analysis. Hairpin-expressing cells showed a 38% increase in granulocytes (Mac-1Gr-1F4/80; P=0.0007) and a 27% decrease in monocytes (Mac-1Gr-1F4/80; P=0.07) (Figure 2C). Taken together, these results show that Hnrnpa0 knockdown (KD) causes a shift in differentiation of myeloid progenitors that favors the granulocytic lineage over the monocytic lineage. The fact that t-MNs with a del(5q) are characterized by trilin-eage dysplasia (dysgranulopoieis, dyserythropoiesis and dysmegakaryopoiesis) suggests that there is a defect in the function of the early myeloid progenitor cell. Our data raise the possibility that loss of HNRNPA0 may be involved in deregulating normal myeloid differentiation, a key component in the changes that drive leukemic growth.

Figure 1.Expression of HNRNPA0 in MDS and during normal hematopoiesis. A. Microarray expression analysis of CD34⁺ cells from healthy individuals and patients with MDS (data generated from GEO accession: GSE4619). The number of subjects in each group, and standard error bars are indicated. Patients with a del(5q) had significantly lower (P<1×10⁻¹²) expression than each of the other groups of patients, or control CD34⁺ cells. B. Microarray expression of mononuclear cells from t-MN patients with or without a del(5q) (data generated from GEO accession: GSE39991). HNRNPA0 expression in 10 t-MN patients with a del(5q) were expressed relative to the average HNRNPA0 expression in t-MN patients without a del(5q). This analysis revealed that, on average, HNRNPA0 is expressed at haploinsufficient levels, but some variability exists among patients. C. Gene expression analysis in human hematopoietic cells using previously published data (GEO accession GSE24759) from Novershtern et al.18 Cell populations (defined in Online Supplementary Methods) representing hematopoietic stem and progenitor cells, terminally differentiated cells and intermediate states were purified from 4–7 independent donors. For easier visualization, populations within the same lineage are aligned. Expression signals were normalized by RMA and batch-corrected by the ComBat method. The resulting probe expression signals were z-transformed for visualization.

Table 1.KEGG pathway enrichment analysis of genes showing a >1.5 fold difference between Hnrnpa0 hairpin-expressing and control PUER cells.

Hnrnpa0 knockdown inhibits monocytic differentiation in PUER cells

To test the hypothesis that a gene signature, that regulates myeloid cell fate, is altered by Hnrnpa0 KD, we adopted the well-characterized PUER cell line as a model system for myeloid differentiation.19,20 We chose PUER cells because this cell line was previously used in elegant studies to describe the gene regulatory network directing macrophage and granulocyte cell fate. The PUER cell line is a PU.1 (Spi1) fetal liver progenitor cell line engineered to express Spi1/PU.1 (when treated with 4 OHT). We established that our PUER cells differentiate in a dose-dependent manner into F4/80 macrophages when treated with doses as low as 1–5 nM 4-OHT (Online Supplementary Figure S2). We and others (H. Singh et al., personal communication, 2010) have noted that these drug concentrations are lower than the original publication,19 suggesting that the PUER cell line has evolved and/or there were variations in drug purity. PUER cells were transduced with the control or Hnrnpa0-specific shRNA, and RT-PCR and Western analysis confirmed that Hnrnpa0 expression was approximately 33–40% the level of control cells (Figure 3A and B). Hnrnpa0 KD led to a significant reduction (P≤0.03) in monocytic (F4/80) differentiation at all doses of 4 OHT (Figure 3C); with decreasing doses of 4 OHT, the reduction in differentiation was more pronounced (Figure 3D) (P=0.002–0.03). There was no obvious change in PUER cell morphology upon Hnrnpa0 KD at 24 h; however, as anticipated, there were fewer macrophages observed upon Hnrnpa0 KD, most apparent 4 days after differentiation with 2 nM 4-OHT (approx. 40% vs. approx. 60% macrophages). For subsequent experiments, we examined the biological consequences of reduced Hnrnpa0 expression after treatment with 2 nM 4 OHT, since the effects of Hnrnpa0 KD were best seen at this concentration. At lower doses, a very small percentage of cells undergo differentiation, potentially masking the effects of Hnrnpa0 KD in subsequent assays. At higher doses, there is such a strong pro-differentiation signal that the phenotype resulting from Hnrnpa0 KD becomes overwhelmed. For example, reduced differentiation following Hnrnpa0 KD was confirmed by the reduced induction of Emr1, Il1r2, and Csf1r, after 24 h of 2nM 4-OHT treatment (P<0.01) (Online Supplementary Figure S3). Spi1 transcript levels (encoding PU.1) were comparable in control and Hnrnpa0 shRNA-expressing cells, confirming that changes in gene expression were not due to variations in Spi1 induction.

Figure 2.Hnrnpa0 knockdown shifts colony formation away from monocytic towards granulocytic lineage in primary murine cells. Total mouse bone marrow cells were retrovirally-transduced with an empty vector (Control) or an Hnrnpa0 shRNA construct, and sorted for GFP positivity. A. RT-PCR analysis of Hnrnpa0 expression in total mouse bone marrow cells infected with control or shRNA-expressing vectors. Expression was normalized against Gapdh expression and reported relative to expression in controls. B. Cells were plated in methylcellulose media with IL-3, IL-6, SCF and Epo for 10–14 days, and colonies were scored by morphology. Each colony type is reported as percent of total colonies scored. CFU-GEMM: granulocyte-monocyte-erythroid-megakaryocyte colony-forming unit; BFU-E: erythroid burst-forming unit; CFU-GM: granulocyte-monocyte colony-forming unit; CFU-G: granulocyte colony-forming unit; CFU M: monocyte colony-forming unit. Significant differences from the control are indicated. C. Colonies were harvested and cells were classified by flow cytometry, as either monocytes (Mac-1^hi Gr-1^lo F4/80⁺) or granulocytes (Mac-1^lo Gr-1^hi F4/80⁻). For all experiments, the average of 3 independent experiments is shown. Bars represent the standard error of the mean.

Figure 3.Knockdown of Hnrnpa0 in PUER cells reduces monocytic differentiation. PUER cells were transduced with either empty (Control) or Hnrnpa0 hairpin-expressing (shRNA) vector, and sorted for GFP positivity. Relative Hnrnpa0 expression was determined by RT-PCR (A) and Western (B) analysis. Expression was normalized against Gapdh or Tubulin expression and reported relative to expression in controls. A scrambled shRNA was used as the control. C and D. PUER cells were induced to differentiate using different doses of 4 OHT as indicated. After four days of induction, the cells were labeled to detect F4/80 expression (marker of monocytes) and analyzed by flow cytometry. The histogram depicts the percent (C) or fraction (D) of F4/80⁺ cells after shRNA treatment, relative to controls.

To rule out off-target effects, we compared the Hnrnpa0 shRNA to an irrelevant, scrambled shRNA (Online Supplementary Figure S4). The level of Hnrnpa0 KD in BM and PUER cells is comparable to the decrease seen in some t-MN patients with a del(5q), since many have a greater than 50% decrease in HNRNPA0 expression (Figure 1B). However, on average, most patients express haploinsufficient HNRNPA0 levels. Thus, we used a second Hnrnpa0-specific shRNA that resulted in 50% KD in undifferentiated PUER cells to model HNRNPA0 haploinsufficiency. This shRNA showed similar effects on monocyte differentiation, as confirmed by the suppression of Hnrnpa0, Emr1, and Egr2 (Online Supplementary Figure S5), suggesting that our system effectively tests the effects of Hnrnpa0 haploinsufficiency on myeloid differentiation. Moreover, it emphasizes that deregulated HNRNPA0 expression, be it a decrease of 50% or 80%, likely has similar effects on myelopoiesis.

Figure 4.The effect of Hnrnpa0 knockdown upon the expression of select myeloid genes during PUER differentiation. Microarray expression analysis was performed prior to differentiation and 24 h after induction by 2nM 4-OHT. Expression is reported as the fold difference in expression in the shRNA (n=3) and control samples (n=3). The genes are grouped broadly according to their effects upon myeloid lineage specification. Underlined gene symbols indicate transcripts known to contain AU-rich 3′ UTR elements. Significant differences between hairpin-expressing and control cells (P 0.05, corrected for multiple comparisons) are indicated: [1] undifferentiated only, [2] differentiated only, [3] both undifferentiated and differentiated. Bars indicate the estimated standard error of the fold difference as determined by linear modeling.

Figure 5.Venn diagram of differentially expressed genes with and without AU-rich elements. All transcripts that were differentially expressed 1.5 or 2- fold between the Hnrnpa0 shRNA-expressing cells and control PUER cells were classified as either up-regulated (orange/red) or down-regulated (blue/blue-gray), in both the undifferentiated and differentiated cells. On the Agilent microarray platform, 1868 transcripts contain AREs (8.9% of 21,087 unique transcripts). The overlap of these ARE-containing genes (yellow) and the genes differentially expressed upon Hnrnpa0 KD can be visualized by means of Venn diagrams. These classifications show that genes containing AREs are over-represented in the group of genes that are down-regulated upon Hnrnpa0 KD, especially in differentiated cells (lower panel). This conclusion is supported by contingency table analysis (Online Supplementary Table S2).

Hnrnpa0 knockdown alters gene expression consistent with a shift towards a pro-granulocytic program in PUER cells

To examine how a decrease in Hnrnpa0 transcript levels translates into a deregulation of the normal myeloid program, we used microarray-based expression profiling. PUER cells expressing the Hnrnpa0 shRNA were sorted for GFP-positivity, before (undifferentiated) and 24 h after differentiation with 2 nM 4-OHT. The network of genes involved in monocytic and granulocytic specification is very well characterized and has been mapped extensively.19,21,22 We grouped these genes broadly into two categories: genes that regulate differentiation in a pro-monocytic (Egr1, Egr2, Nab2, Spi1/PU.1, Csf1r, Maf, Irf8) or pro-granulocytic (Cebpa, Gfi1, Csf3r) manner, and genes that serve as markers of monocytic (Emr1, Mmp12, Itgam, Tlr2) or granulocytic (Ltf, Ncoa3, Prtn3, Il1r2) differentiation. It should be noted that many of these ‘marker’ genes play critical roles in mediating the function of mature myeloid cells. We also examined several genes that block myeloid differentiation either directly (Myc, Myb, Kit), or by promoting differentiation along other lineages at the expense of myeloid differentiation (Gata1, Gata2). Among these genes, Egr1, Egr2, Gata2, Gfi1, Irf8, Myc, Mmp12, and Nab2 transcripts contain AREs and, as such, represent potential direct targets of Hnrnpa0.23–26

Analysis of the expression of these genes by microarray (Figure 4) and subsequent validation by RT-PCR (Online Supplementary Figure S6) revealed a shift away from genes that regulate monocytic differentiation. For example, the microarray data confirmed significant reductions of the monocytic marker, Emr1 (F4/80), and pro-monocytic gene, Egr2, in Hnrnpa0 hairpin-expressing cells, both before and after differentiation. The pro-monocytic genes encoding Nab2, an Egr1/Egr2 binding partner, and Irf8 (Icsbp) were suppressed in differentiated and undifferentiated cells, respectively. Analysis of the pro-granulocytic genes showed changes consistent with a shift towards a pro-granulocytic program. Upregulation of the Gfi1 and Cebpa transcription factors induce granulocytic differentiation in a complementary manner.19 Cebpa, an antagonist of PU.1 activity, has been shown to promote granulocytic differentiation, and Gfi1 is a direct antagonist of the pro-monocytic activity of Egr2 and Nab2. Consistent with a shift towards a granulocytic program, the expression of the Cebpa, Csf3r, and Gfi1 genes was elevated in hairpin-expressing cells relative to control. Together, these data suggest that Hnrnpa0 is involved in myelopoiesis by suppression of pro-monocytic genes (Emr1, Egr2, Nab2, Irf8) and induction of pro-granulocytic genes (Gfi1, Cebpa, Csf3r), some of which contain AREs (underlined genes in Figure 4).

Table 2.KEGG pathway enrichment analysis of AU-rich genes differentially expressed in t-MN del(5q) patients versus non-del(5q) patients.

Figure 6.Model of the impact of 5q deletion upon myeloid cell fate. A schematic diagram showing some of the factors that regulate myeloid cell fate (adapted from P. Laslo et al., 200619). PU.1 normally induces EGR2 and NAB2 to induce monocytic differentiation. CEBPA and GFI1 promote granulocytic differentiation, partially through suppression of EGR1, EGR2 and NAB2. Knockdown of Hnrnpa0 in murine cells leads to the suppression of pro-monocytic transcripts (Egr2, Nab2, Irf8, and Emr1 (F4/80)), many of which contain AREs (light gray boxes), and induction of pro-granulocytic genes (Cebpa, Gfi1, Csf3r (G-CSFR)). MAPK-induced phosphorylation of CEBPA has been shown to inhibit granulopoeisis37. KD of Hnrnpa0 leads to a decreased expression of many genes within the MAPK signaling pathway, consistent with the observed shift towards granulocytic differentiation. Loss of EGR1 (also mapped to human 5q) also favors granulocytic over monocytic differentiation36. Egr1 and Egr2 function in a redundant manner. In del(5q) patients, haploinsufficiency for both EGR1 and HNRNPA0 may synergistically disrupt myeloid differentiation.

Hnrnpa0 knockdown preferentially alters the expression of AU-rich genes in PUER cells

To determine whether Hnrnpa0 KD leads preferentially to differential expression of AU-rich genes, we used the AU-rich Element Database (ARED 3.0) to identify a signature of transcripts known to contain AREs26 and classified them by the fold difference between the Hnrnpa0 hairpin-expressing cells versus the controls, for both the undifferentiated and differentiated PUER samples. The overlap between all differentially expressed genes and those containing AREs was visualized using Venn diagrams (Figure 5). From these classifications, it becomes apparent that genes containing AREs are over-represented in the group of genes whose expression is lower in the hairpin-expressing cells. Although this overrepresentation is observed in both undifferentiated and differentiated sample, the difference is most striking in differentiated samples, where ARE-containing genes make up 12–14.7% of the genes that show a 1.5–2 or more fold decreased expression, but only 9.3–9.5% of the genes that show a 1.5–2 or more fold increased expression. Contingency table analysis also confirmed that there are significantly more down-regulated ARE-containing genes upon Hnrnpa0 KD than expected for differentiated cells (Online Supplementary Table S2). These data suggest that KD of Hnrnpa0 results in a preferential decrease in the transcript levels of ARE-containing transcripts.

GSEA analysis reveals a set of potential Hnrnpa0 target transcripts

To study the possible enrichment of AU-rich genes at the extremes of the differentially expressed genes, we implemented a simplified GSEA algorithm.27 GSEA demonstrated a significant (P<0.001) enrichment of AU-rich genes at both extremes of differential expression, with a greater enrichment of down-regulated genes (Online Supplementary Figure S7). An analysis of genes showing at least a 1.5-fold difference between the hairpin-expressing cells and control cells revealed 172 genes in undifferentiated cells (72 up-regulated, 100 down-regulated) and 222 genes in differentiated cells (80 up-regulated, 142 down-regulated) (Online Supplementary Table S3). The genes that showed altered expression in both undifferentiated and differentiated cells (11 were up-regulated and 30 were down-regulated) represent potential Hnrnpa0 targets. Of particular interest are Cbfb and Sp3 given their roles in hematopoiesis,28,29 as well as Creb5, Egr2, and Tcfap2e, given their role in positive regulation of transcription. Thus, upon acute decrease in Hnrnpa0, AU-rich genes made up a disproportionately large fraction of the under-expressed genes and, as a group, were found to be more dysregulated than was seen globally for all genes interrogated.

To infer biological meaning from the list of differentially expressed ARE genes, we employed the Database for Annotation, Visualization and Integrated Discovery (DAVID).30,31 Of significance, the JAK-STAT signaling pathway (P=0.004) was enriched in the up-regulated genes in undifferentiated cells; the MAPK pathway (P=0.005) was enriched, and the chemokine signaling (P=0.057) pathway showed a trend toward enrichment in the down-regulated genes in differentiated cells. Moreover, the cytokine-cytokine receptor interaction path was enriched in both of these gene lists (P=0.004 and P=0.003) (Table 1). The control of mRNA stability is a key means of regulating cytokine production,13 suggesting that Hnrnpa0 may be a positive or negative regulator of cytokine transcript stability, depending on the gene and cellular context. Alternatively, it remains possible that Hnrnpa0 KD may have had indirect effects on cytokine expression, independent of transcript stabilization. To determine if any of these pathways could predict functional changes, we focused on the MAPK signaling pathway, given its importance in myeloid differentiation.32 A comparison of phospho-p42/44 MAPK (Erk1/2) in differentiated versus undifferentiated PUER cells revealed that Erk1/2 was less activated upon differentiation in Hnrnpa0 shRNA cells, compared to the scrambled controls (Online Supplementary Figure S8), consistent with our model for how HNRNPA0 contributes to the regulation of myeloid differentiation (see Discussion section and Figure 6).

Expression of the AU-rich transcriptome in t-MN del(5q) patients

The dysregulation of regulatory proteins affecting ARE mRNA stability can lead to abnormalities in many critical cellular processes. In addition to Hnrnpa0, which is hap-loinsufficient in del(5q) t-MN patients, several other genes encoding AUBPs were differentially expressed in t-MN patients with a del(5q) versus non-del(5q) t-MN patients (Online Supplementary Table S1). Since AUBPs can increase and/or decrease mRNA stability, depending on the cellular context,13 their differential expression may indicate an intricate regulation of mRNA stability. Given the importance of AREs in disease,12,26 we examined ARE-mRNAs that are differentially expressed in del(5q) versus other t-MN cases. Using the ARED 3.0 database, we identified 1674 unique ARE transcripts in the Affymetrix t-MN array. Among these, 515 genes showed a significant differential association, with 237 down-regulated and 278 up-regulated (FDR <0.3) (Online Supplementary Table S4 and Figure S9). Using the KEGG pathway analysis tool within DAVID, we identified the cell cycle pathway as being enriched (P=0.008) along with the MAPK (P=0.030) and chemokine (P=0.044) signaling pathways (Table 2). Extending this analysis of biological significance with the Ingenuity software, we found that del(5q) patients have gene expression changes consistent with an increase in cellular growth and proliferation, as compared to non-del(5q) t-MN patients (Online Supplementary Table S5). This analysis raises the intriguing possibility that stabilization of transcripts, through regulation of the AU-rich transcriptome, may be a contributing factor to the dysregulation of critical growth pathways in t-MN del(5q) patients.

Discussion

To date, none of the individual 5q genes show homozygous loss-of-function, or recapitulate all of the features of myelodysplastic syndrome (MDS) or AML when expressed at haploinsufficient levels in hematopoietic cells in mouse models. Rather, current studies support a haploinsufficiency model, in which loss of a single allele of more than one gene on 5q contributes to the pathogenesis of t-MN.33 The current challenge is identifying which of the numerous genes in the deleted chromosome 5 segment are involved in the pathogenesis of human t-MN. In this paper, we show that decreased HNRNPA0 expression is characteristic of leukemias with a del(5q), and that recapitulating loss of Hnrnpa0 expression in mouse hematopoietic cells alters myeloid differentiation. Our work suggests that KD of Hnrnpa0, a gene encoding an AUBP, alters myeloid fate by shifting differentiation away from the monocytic in favor of the granulocytic lineage. This shift is accompanied by the suppression of pro-monocytic and induction of pro-granulocytic transcriptional regulators, several of which contain AREs, e.g. Irf8, Nab2, and Egr2.

One interesting phenomena that this work, as well as our previous work on EGR1, raises is that decreasing expression of del(5q) genes by half (i.e. haploinsufficiency) may be as detrimental as complete loss of the gene when placed in an environment conducive to malignant transformation. For example, we previously showed that reducing Egr1 expression in mice by 50% (Egr1) or knocking it down completely (Egr1) was equally effective at cooperating with secondary mutations induced by the alkylating agent ENU to induce a myeloproliferative disease with ineffective erythropoiesis. Herein, we show that, on average, t-MN patients with a del(5q) express half of the levels of HNRNPA0, but there is some variability with some patients showing greater loss. Moreover, loss of Hnrnpa0 expression, ranging from 50–80%, appears to have similar effects on myeloid differentiation in primary and PUER mouse cells.

Numerous studies have established that a core set of genes act as ‘master regulators’ of myeloid differentiation.19,21,22,34,35 Monocytic differentiation is promoted by the PU.1-induced transcription factors EGR2 and NAB2, whereas granulocytic differentiation is promoted by GFI1 and CEBPA (Figure 6). Our data show that loss of Hnrnpa0 leads to the suppression of pro-monocytic genes (Egr2, Nab2, Irf8, Emr1) and induction of pro-granulocytic genes (Cebpa, Gfi1, Csfr3), within this network. EGR1, like HNRNPA, is located within the CDS of 5q31.2, and is expressed at reduced levels in CD34 cells from patients with a del(5q). Similar to HNRNPA0, loss of EGR1 expression favors granulocytic over monocytic differentiation.36 Thus, loss of a single allele of EGR1 and HNRNPA0, as a result of a del(5q), may lead to a synergistic disruption of EGR1, and the functionally redundant family member, EGR2, leading to aberrant myeloid differentiation, a hallmark of t-MN. Any disruption to the gene program that regulates the transition from the CMP to the GMP and beyond can uncouple lineage commitment and proliferation control during myeloid differentiation and lead to malignant transformation. Although, t-MN is thought to be due to the simultaneous loss of multiple 5q genes together with secondary cooperating mutations, the observed shift towards granulocytic differentiation and ensuing change in gene expression, following Hnrnpa0 KD, may be one piece of the puzzle explaining the dys-granulopoiesis observed in patients.

The ability of CEBPA to induce granulocytic differentiation is negatively regulated by ERK1/2- and p38MAPK-mediated phosphorylation.32,37 Since many genes in the MAPK signaling pathway were down-regulated upon Hnrnpa0 KD in PUER cells (Table 1), these results raise the possibility that HNRNPA0 haploinsufficiency may shift differentiation towards granulopoiesis, not only through a deregulation in the expression of the myeloid specification genes, but also through an inhibition of MAPK activity (Figure 6). Consistent with this, Erk1/2 was less activated upon differentiation in PUER cells expressing the Hnrnpa0 shRNA as compared to scrambled shRNA (Online Supplementary Figure S8). Together, our data raise the possibility that HNRNPA0 haploinsufficiency may lead to numerous gene changes that may be small in magnitude, yet coordinately de-regulate myeloid differentiation, possibly together with a reduced dose of EGR1.

Consistent with the proposed role of Hnrnpa0 as a positive regulator of transcript stability, this work determined that AU-rich genes made up a disproportionately large fraction of the under-expressed genes in cells expressing an Hnrnpa0 shRNA. However, like many other AUBPs, its regulatory ability may be more complex and regulated in a cell-type and activation-dependent manner.38–40 The observed changes in transcript profiles upon Hnrnpa0 KD were numerous, yet small in magnitude for AU-rich genes, but also for some non-AU rich genes, suggesting that HNRPNA0 likely ‘fine-tunes’ gene expression, and that some of its effects on the pathogenesis of disease may be secondary to its post-transcriptional regulation. We attempted to explore the effects of Hnrpa0 loss in vivo; however, suppression of Hnrpa0 was not maintained past 12 weeks in the bone marrow of mice (data not shown). This may suggest that without a clonal advantage, such as TP53 mutations which are frequent in del(5q) patients, hematopoietic cells may not tolerate low levels of Hnrnpa0. An analysis of ARE genes revealed that the MAPK, cell cycle and chemokine signaling pathways were enriched in t-MN with del(5q) versus non-del(5q) patients. Most of these changes are consistent with pathway activation and increased cell cycling and growth (Table 2 and Online Supplementary Table S5). This is not entirely surprising given that activation of these pathways has been observed in the spectrum of myeloid malignancies. However, it is of interest that deletion of chromosome 5 alters gene expression of ARE-containing genes in a direction that predicts enhanced cellular growth, since cells from non-del(5q) myeloid neoplasms would also be predicted to be actively proliferating. Whereas, Hnrnpa0 led to the immediate decreased expression of some MAPK genes in PUER cells, in t-MN patients with a del(5q) there was an upregulation of similar MAPK genes. This is likely due to the complex genetic alterations in del(5q) patients and emphasizes that the use of genetic knockdown in the PUER cell line provided an informative model to pinpoint gene changes resulting from deregulated Hnrnpa0 expression. We recognize that this is a preliminary investigation into the relationship between dysregulation of AU-rich transcript stability and growth pathways in t-MNs; however, it is interesting to speculate that the post-transcriptional control of ARE-mRNA stability may be a critical factor influencing leukemia cell proliferation in t-MNs with a del(5q).

Our study shows that HNRNPA0, normally highly expressed in multipotent hematopoietic progenitors, is expressed at reduced levels in CD34 cells from patients with a del(5q), suggesting that it is a candidate gene in malignant myeloid disorders with abnormalities of chromosome 5. We modeled HNRNPA0 loss in murine cells and showed that a decreased dose leads to a dysregulation of myeloid cell differentiation, through suppression of pro-monocytic genes and induction of pro-granulocytic genes, mediated in part by alterations in the expression of AU-rich genes. Moreover, our analysis of differentially expressed AU-rich genes in t-MN patients reveals that post transcriptional control of mRNA stability warrants consideration especially with regard to HNRNPA0’s role in the dysregulation of myeloid differentiation in the pathogenesis of t-MNs with a del(5q).

Acknowledgments

We would like to thank Drs. Lucy Godley, Amittha Wickrema, John Crispino, and Barbara Kee, University of Chicago, for helpful advice, Dr. Kee for providing the pBanshee-GFP vector and Anthony Fernald, Elizabeth Davis and Jianghong Wang for technical assistance.

Footnotes

• DJY and AS equally contributed to this work.
• The online version of this article has a Supplementary Appendix.
• Funding Funding was provided by the MSTP program, NIH P01 CA40046 (MML, JRD) and the Cancer Center Support Grant of the University of Chicago Comprehensive Cancer Center (P30 CA14599) through use of the Shared Resources (Cytometry and Antibody Technologies Facility).
• Authorship and Disclosures Information on authorship, contributions, and financial & other disclosures was provided by the authors and is available with the online version of this article at www.haematologica.org.

• Received September 25, 2013.
• Accepted February 13, 2014.