Metformin-induced suppression of NLK improves erythropoiesis in pre-clinical models of Diamond Blackfan Anemia through induction of miR-26a

Abstract

Diamond Blackfan Anemia (DBA) results from haploinsufficiency of ribosomal protein subunits in hematopoietic progenitors in the earliest stages of committed erythropoiesis. Nemo-like kinase (NLK) is chronically hyperactivated in committed erythroid progenitors and precursors in multiple human and murine models of DBA. Inhibition of NLK activity, or suppression of NLK expression, both improve erythroid expansion in these models. Metformin is a well-tolerated drug for type 2 diabetes mellitus with multiple cellular targets. Here we demonstrate that metformin improves erythropoiesis in human and zebrafish models of DBA. Our data shows that the effects of metformin on erythroid proliferation and differentiation is mediated by suppression of NLK expression through induction of miR-26a, which recognizes a binding site within the NLK 3’UTR to facilitate transcript degradation. We propose that induction of miR-26a is a potentially novel approach to treat DBA and could improve anemia in DBA patients without the potentially adverse side effects of metformin in a DBA patient population.

Graphical Abstract

Introduction

Diamond Blackfan Anemia (DBA) is one of the inherited bone marrow failure syndromes and presents with a macrocytic red blood cell aplasia usually within the first year of life (1). Over 70% of patients carry genetic mutations that lead to haploinsufficiency in one of at least 22 genes encoding ribosomal proteins (2). RPS19 and RPL11 genes are commonly mutated and account for approximately 25% and 5% of cases respectively (1). Ribosomal insufficiency leads to reduced translational efficiency of a subset of mRNA transcripts, including the master erythropoiesis transcription factor GATA1 (3). Irrespective of the driving ribosomal gene mutation, ribosomal insufficiency increases p53 protein stabilization (1, 4), contributing to aberrant activation of the serine-threonine protein kinase, nemo like kinase (NLK). Suppression of NLK expression or functional activity, significantly improves expansion of DBA erythroblasts in pre-clinical models, with minimal impact on other hematopoietic lineages (5).

NLK protein expression is significantly higher in erythroblasts than other hematopoietic lineages, and is not influenced by ribosome insufficiency. Rather, the increased activity in DBA is due to post-translational stimulation (5). Unlike in wild type hematopoietic lineages, NLK kinase activity increases in megakaryocyte and erythroid progenitors (MEPs) in DBA, and is maintained through the progenitor and precursor stages (5). The erythroid-specific effects of ribosome-insufficiency are partially due to the erythroid-specific expression of NLK, and may also explain why targeting NLK does not affect other lineages (5).

NLK is an orthologue of the Drosophila Nemo. It is an atypical member of the mitogen-activated protein kinases (MAPK) family. The kinase domain shares a high degree of sequence conservation with other MAPKs and cyclin-dependent kinases (Cdks) (6) which has hampered attempts to develop small molecules that specifically inhibit NLK. Off-target inhibition of NLK kinase activity with broad kinase inhibitors, such as SD208, have improved erythropoiesis in pre-clinical DBA models, but these compounds are not clinically useful due to low potency and poor solubility (data not shown). Therefore, alternative strategies to suppress NLK offer promise in the treatment of diseases affected by aberrant NLK activity.

Previous work has shown that NLK expression is significantly suppressed in response to a number of microRNAs (miRNAs), including miR-181 (7), miR-208 (8), miR-101 (9), miR-199 (10) and miR-221 (11). MicroRNAs are small RNAs, 21–24 base in length that are key regulators of post-transcriptional gene expression and RNA silencing. Hundreds of miRNAs have been identified with expression often highly tissue specific (12). Most miRNAs exert an inhibitory effect through binding a short 6–8 nucleotide sequence in the 3’ untranslated region (3’UTR) of a target gene transcript. Binding is facilitated by the complementary seed sequence within the miRNA, which triggers recruitment of the RNA-induced silencing complex (RISC) leading to mRNA degradation (13). Often miRNAs only moderately perturb expression of the targeted transcript (13), however the significant susceptibility of NLK mRNA offers the potential to facilitate targeted regulation of NLK expression by the modulation of specific miRNAs.

One strategy to modulate miRNA function and, in turn, NLK expression could involve metformin, a synthetic analog of guanidine and approved therapeutic for type 2 diabetes. Metformin use has increased dramatically for pediatric type 2 diabetes and has been used in the adult population for many years. The drug is well tolerated and remains the mainstay of therapy along with diet and exercise (14). In addition to treatment of type 2 diabetes, metformin is effective in polycystic ovary syndrome and is being explored as an antiviral and anticancer agent in adult populations (15). Notably, although the mechanistic effect of metformin on NLK activity has not been thoroughly investigated, one study has demonstrated that metformin does suppress NLK expression in lung carcinoma models (16). Furthermore, metformin has been reported to deregulate miRNA profiles in a myriad of tissues (17). Collectively, these data led us to hypothesize that metformin may suppress NLK expression through miRNA modulation. Indeed, many of the miRNA perturbations mediated by metformin have potential clinical implications including the modulation of disease and homeostasis in healthy individuals (18–26).

Although metformin is well tolerated in a general pediatric population and the promise of NLK suppression is exciting, the mechanism of suppression is not understood and potential side-effects may adversely impact DBA patients. For example, metformin is known to reduce mTOR signaling and protein synthesis; this effect would likely be deleterious in DBA where there is already reduced translation brought about by ribosomal-insufficiency. Understanding the mechanism through which metformin suppresses NLK will facilitate design of more NLK-specific inhibitors without the potentially adverse side-effects associated with metformin.

Here we report the impact of metformin on erythroid expansion in human, murine and zebrafish DBA models and define the mechanism through which metformin suppresses NLK to mediate this effect.

Methods

Cell culture –

Human CD34⁺ HSPCs were purified from cord blood (New York Blood Center) using magnetic-activated cell sorting (Miltenyi Biotec) and differentiated as described previously (5). Luciferase, or luciferase fused to various permutations of the NLK 3’ or 5’UTRs, were cloned in pcDNA3.1 expressing RFP and transfected into K562 or Kp53A1 cells using Lipofectamine® 2000 (Thermo Fisher).

Lentiviral transduction –

CD34⁺ cells were transduced as published (27) with lentivirus expressing shRNA against RPS19, RPL11, or luciferase (Luc). Virus co-expressed GFP, RFP, mCherry or puromycin to enable selection.

Metformin and miRNAs –

Metformin was purchased from SelleckChem and added to cells at indicated concentrations with a final DMSO concentration of 0.5%. MISSION® synthetic miRNAs, inhibitors and mimetics were purchased from Sigma Aldrich and transfected according to manufacturer’s instructions.

Colony assays –

Sorted HSPCs were seeded in cytokine-containing methylcellulose medium (H4434; STEMCELL Technologies) in triplicate, with 1000 cells per plate. Erythroid (burst-forming unit erythroid) and myeloid (colony-forming unit, granulocyte-macrophage) colonies were counted 14–18 days later.

Flow cytometry –

cells were incubated with human Fc receptor binding inhibitor (#14–9161-73; eBioscience) followed by primary antibodies CD235-APC (#306607; BioLegend) and CD11b-PE/Cy5 (#101209; BioLegend). Data were collected on a DxP10 flow cytometer (Cytek) and analyzed by using FlowJo Software, v.9.7.2.

Kinase Assays, Western blotting –

NLK kinase activity was performed as published (5). For Western blotting, antibodies against NLK (#AB97642; Abcam; 1:1000 dilution) and GAPDH (#MAB374; Millipore; 1:10000) were used according to manufacturer’s instructions.

qRT-PCR –

MessengerRNA was quantified as described (5). MicroRNA was quantified using TaqMan® Small RNA Assays (Applied Biosystems) as per manufacturer’s directions and normalized to snoRNA.

Luciferase Assay –

The NLK minimal promoter (1019 5’nucleotides) and or NLK 3’UTR (1885 3’ nucleotides) were cloned upstream, or downstream respectively, of firefly luciferase and transfected into K562 cells. Transfection efficiency was normalized by RFP expression and firefly luciferase activity determined by Luciferase Assay Reagent II (LAR II) from Dual-Luciferase® Reporter (DLR™) Assay System (Promega). Luminescence was assessed using a Synergy™ H1 hybrid multi-mode microplate reader (BioTek®). Mutations and truncations in NLK 3’UTR were introduced using QuikChange® II XL Site-directed Mutagenesis (Agilent).

Mice –

The RPS19- and RPL11-deficient mouse models have been described previously (5, 28–30). Kit⁺ progenitors were isolated from E14.5 liver cells of tet-shRPS19-expressing fetuses and differentiated in vitro in the presence or absence of doxycycline. Lin⁻Kit⁺ HSPCs were isolated from femur bone marrow of inducible RPL11 heterozygous deletion adult mice. All animal experiments were performed with consent from the Lund University animal ethics committee or the Ethical Committee of the Carlos III Health Institute, Madrid, Spain (#54–2013-v2) and in agreement with the recommendations of the Federation of European Laboratory Animal Science Association (FELASA).

Zebrafish –

Zebrafish were reared and injected with control or rps19-specific morpholino (MO) at the 1-cell stage as previously described (27), and treated with 20 mM metformin 4 to 5 hours post fertilization (hpf). At day 3, embryos were stained with o-dianizidine to detect hemoglobin. Embryos were obtained by natural spawning. UCLA Animal Committee approved the study.

Statistics –

P values for statistical significance were obtained by using a paired Student t test. Significance was designated as p<0.05. The data are representative of at least 3 independent experiments. When possible, variability between replicates was normalized for by designating values of controls to 100% (or 1-fold), and comparing variables against that.

Metformin suppresses NLK expression leading to improved erythropoiesis in human models of Diamond Blackfan Anemia.

Transduction of shRNA against RPS19 and RPL11 into CD34+ HSPCs have been demonstrated to induce erythroid defects as a model of DBA (27, 31, 32). CD235+ erythroid progenitors are reduced to 5.7% (p=0.0004) of control in RPS19-insufficiency and 16.4% (p=0.0032) in RPL11-insufficiency when expanded in erythroid liquid culture (Fig 1A upper). A mild reduction in the non-erythroid CD11b+ myeloid population occurs (13% (p=0.0436) and 7% (p=0.2758) reduction in RPS19- and RPL11-insufficiency respectively) (Fig 1B upper). When cultured in the presence of 50 mM metformin, CD235+ erythroid expansion increased from 5.7% to 33.3% (p=0.0008) and 16.4% to 32.7% (p=0.0951), which constituted an increase of 5.9-fold and 2.0-fold in RPS19- and RPL11-insufficiency respectively (Fig 1A lower). The presence of metformin had a negligible influence on CD11b+ expansion (Fig 1B – lower). Metformin EC₅₀ was 11.4mM and significant erythroid improvement was detected at concentrations above 8mM.

Figure 1. Metformin improves expansion of CD235+ erythroblasts and BFU-E erythroid colony formation in human RPS19- and RPL11-insufficiency from CB CD34+ progenitors.

(A). CD34+ HSPCs were transduced with shRNA against control (shLuc), RPS19 (shRPS19) or RPL11 (shRPL11) and after sorting, were differentiated for 14 days in the presence or absence of 50mM metformin. Cells were counted and the percentage expressing CD235+ erythroid (left) and CD11b+ myeloid (right) was determined by flow cytometry. Obtained values were multiplied to give an overall number that was normalized to the untreated control (grey columns). Values are presented as a percentage of the untreated control. (B) Direct comparison between metformin treated and untreated cultures is facilitated by normalizing the metformin-treated values to the untreated values in control, RPS19- and RPL11-insufficient groups. Values are expressed as a fold induction relative to untreated. (C) Transduced and sorted CD34+ progenitors were cultured in methylcellulose in the presence or absence of metformin for 16 days and BFU-E erythroid (left) and CFU-GM myeloid (right) colonies scored. Values are represented as the percentage of colonies induced in untreated controls (grey columns). (D). To directly compare the effect of metformin in each group, metformin-treated cultures were normalized to untreated cultures and are expressed as a fold induction of the untreated control. (E) Representative images of BFU-E colonies at day 14.Scale bar = 200 μM. Data are displayed as means +/− SD. Statistics: two-tailed Student’s t test, significant *P < 0.05

Metformin also improved erythropoiesis in colony assays. Compared to vector control (shLuc), BFU-E erythroid colonies were increased from 7.7% to 16.3% (p=0.0292) in RPS19-insufficiency and 8.2% to 14.3% (p=0.0257) in RPL11-insufficiency (Fig 1C upper), corresponding to 2.9- and 2.2-fold increases (Fig 1D lower). Metformin did not affect CFU-GM colonies (Fig 1D). RPS19-insufficient progenitors formed markedly smaller BFU-E colonies than controls. Metformin treatment appeared to only modestly improve the size of these colonies (Fig 1E).

As metformin influences NLK in other cell systems (16), we examined NLK activity and expression in RPS19-insufficiency in the presence and absence of metformin. NLK immunoprecipitated from 5000 differentiating RPS19-insufficient progenitor cells robustly phosphorylated NLK, c-Myb and raptor in vitro (Fig 2A). Metformin treatment reduced NLK activity from the same number of RPS19-insufficient progenitors, reducing phosphorylation of NLK, c-Myb and raptor by 48.5%, 39.8% and 40.2% respectively (Fig 2A). SD208 is a Transforming Growth Factor-beta receptor small molecule inhibitor that inhibits NLK activity as an off-target in these cells (5). The effects of SD208 and metformin were compared. Examination of NLK expression in metformin-treated cultures revealed significantly less NLK expression, independent of RPS19 status. Metformin treatment reduced NLK protein expression by 62.6% and 65.4% in control and RPS19-insufficient cultures respectively. In contrast, SD208 did not significantly influence NLK expression (Fig 2B).

Figure 2. Metformin improves erythropoiesis through suppression of NLK expression.

(A) Control (shLuc – grey columns) or RPS19-insufficient (shRPS19 – black columns) progenitors were differentiated in erythroid media alone, vehicle or vehicle containing 50mM metformin or 5uM SD208 for 5 days. 5000 cells per treatment were lysed and immunopurified NLK was subjected to in vitro kinase assay to determine phosphorylation potential against 3 recognized NLK substrates; NLK itself (upper), c-Myb (middle) or raptor (bottom). (B) Simultaneously, qRT-PCR was performed to examine NLK (upper) and RPS19 (lower) mRNA expression. (C) Active NLK was purified from activated Kp53A1 cells and subjected to in vitro kinase assay in the presence of 0, 50 nM, 50 μM, or 50 mM metformin or SD208. The phosphorylation of NLK (upper), c-Myb (middle) and raptor (lower) was determined after 30 mins. (D) CD34+ progenitors were transduced with a combination of either shRNA against luciferase (shLuc – grey columns) or RPS19 (shRPS19 – black columns) and siRNA against a non-targeting sequence (NT) or NLK (siNLK). After 14 days differentiation in the presence or absence of metformin, cells were counted and the percentage of cells with surface expression of CD235 and CD11b were determined by flow cytometry, to yield the number of CD235+ erythroid (upper) and CD11b+ myeloid (lower) cells. The total number of cells is expressed as a percentage of the number of each cell type in the control (untreated/shLuc/NT). Data are displayed as means +/− SD. Statistics: two-tailed Student’s t test, significant *P < 0.05

SD208 inhibited the kinase activity of NLK (5) (Fig 2A), but did not reduce NLK expression (Fig 2B). We speculated that metformin reduced NLK kinase activity and improved erythropoiesis through suppression of NLK expression, rather than inhibiting the kinase activity directly. The addition of metformin to activated NLK in in vitro kinase assays did not influence NLK activity, supporting our hypothesis that reduced intracellular NLK activity is due to reduced NLK expression, rather than inhibition of kinase activity (Fig 2C). NLK activity was robustly inhibited by SD208 (Fig 2C). Expression of siRNA against NLK improved erythropoiesis in RPS19-insufficiency by 7.0-fold (4.9% – 34.2% of control; p=<0.0001), however metformin treatment did not significantly increase CD235+ erythroblast expansion (7.0- to 7.5-fold; p=0.6284) (Fig 2D) in control cultures. SD208 and NLK silencing similarly improve erythropoiesis in both RPS19 and RPL11-insufficient models (5). Because of this, we propose the observed effect of metformin in RPL11-insufficient cultures (Fig 1) is also the result of NLK suppression.

Metformin improves erythroid expansion in zebrafish but not murine models of DBA.

NLK activation has been observed to contribute to erythropoiesis defects in murine models of DBA (29, 33). However, metformin treatment did not rescue Ter119+ erythroblast expansion in RPS19- or RPL11-insufficiency (Fig 3A). In contrast to human models, metformin had no impact on NLK expression (Fig 3B) or NLK activity (Fig 3C) in control or either RPS19- or RPL11-insufficient mice. In zebrafish (Danio rerio) anemia reminiscent of DBA occurs with RPS19-insufficiency (34). We induced RPS19-insufficiency by morpholino in the presence or absence of 20 mM metformin and examined hematopoiesis/hemoglobin activity in embryos by O’dianisidine staining. In RPS19-insufficiency, loss of O’dianisidine staining was significant, along with accompanying heart failure (Fig 3D middle). Although heart conditions persisted, metformin dramatically restored O’diansidine staining, particularly along the midline (Fig 3D right).

Figure 3. Metformin improves erythropoiesis in zebrafish model of DBA but not murine models.

(A left) Murine fetal liver Kit+ progenitors expressing tetracylin-regulatable shRNA against RPS19 were cultured in the absence (grey columns) or presence (black columns) of doxycycline either alone, or treated with metformin. After 8 days, the number of ter119+ erythroblasts was calculated. (A right) RPL11+/lox mice were left untreated (grey columns) or treated (black columns) with tamoxifen for 5 weeks prior to isolation of bone marrow Lin-Kit+ progenitors. Progenitors were differentiated in the presence or absence of metformin for 8 days before assessing the number of Ter119+ erythroblasts. (B) In conjunction with flow cytometry, cultures were subjected to qRT-PCR for NLK expression and (C) kinase NLK kinase activity was assessed by in vitro kinase assay. (D) Zebrafish were reared and injected with control or rps19-specific morpholino and treated with 20 mM metformin 4 to 5 hours post fertilization (hpf). At day 3, embryos were stained with o-dianizidine to detect hemoglobin. Data are displayed as means +/− SD. Statistics: two-tailed Student’s t test, significant *P < 0.05

Metformin sensitivity is mediated through a miR-26a-binding site within the NLK 3’UTR.

Having demonstrated that metformin improved erythropoiesis in ribosomal-insufficiency through NLK suppression in human models, we sought to determine the mechanism of action. The influence of miRNA on gene expression is often subtle (13), but it has been demonstrated that NLK expression can be extensively suppressed by miR-181 (7), miR-208 (8), miR-199 (10), miR-101 (9) and miR-221 (11). As miRNA influence is typically through binding to elements within the 3’UTR (13) we asked if metformin induced NLK degradation through the NLK 3’UTR.

Fusion of the NLK 3’UTR to the luciferase gene, but not 5’ promoter sequence, resulted in a dose-dependent, metformin-mediated degradation of luciferase (67.5% decrease at 50 mM) similar to endogenous NLK suppression (Fig 4A). Expression of luciferase alone, or fused with an alternative 3’UTR from SATB1, did not result in metformin sensitivity. In parallel with measuring luciferase activity, the response of endogenous NLK protein expression was monitored by Western blot analysis with a high degree of correlation.

Figure 4. NLK 3’UTR facilitates metformin-mediated NLK suppression.

(A) K562 cells stabling expressing luciferase alone (top panel), luciferase expressed behind a minimal promoter and the NLK proximal promoter (2^nd panel), luciferase immediately upstream of the human NLK 3’UTR (3^rd panel), , or luciferase with the SATB1 3’UTR (bottom panel) were grown in the presence of 0, 0.5, 5.0 or 50.0mM of metformin for 72 hrs. Cell were pelleted and either processed for luciferase assay (left) or western blot analysis of endogenous NLK (upper right) or GAPDH (lower right). (B) Luciferase immediately upstream of the human (upper), murine (middle) and zebrafish (lower) NLK 3’UTR were transiently transfected into human K562 (left) and cultured with indicated concentrations of metformin for 72 hrs. Lysates were assessed for luciferase activity and endogenous NLK and GAPDH protein expression by Western blotting. Data are displayed as means +/− SD. Statistics: two-tailed Student’s t test, significant *P < 0.05

Having determined that the suppressive effect of metformin on NLK in human hematopoietic cells is mediated by the 3’UTR, we asked if the lack of metformin sensitivity in mice was due to differences in the murine 3’UTR sequence. When expressed in human K562 cells, human, murine and zebrafish 3’UTR sequences all facilitated a dose-dependent reduction in luciferase activity (Fig 4B), indicating all 3 species retain a conserved metformin-responsive element within the 3’UTR. The human 3’UTR facilitated a 65.4% (p=0.0056) decrease while the murine and zebrafish sequence reduced luciferase activity by 61.6% (p=0.008) and 43.6% (p=0.031) respectively (Fig 4B).

The NLK 3’UTR sequence contains approximately 30 potential miRNA binding sequences. We generated a series of truncations in the human NLK 3’UTR sequence and analyzed the metformin-responsiveness (Fig 5A). Luciferase fused to 3’UTR sequence lacking only the smallest deletions (truncations 1 and 2) retained metformin-sensitivity, while deletion of more nucleotides lost sensitivity (Fig 5B). We conclude that the 261 nucleotides between truncations 2 and 3 are required for metformin-mediated suppression. This includes potential binding sites for 4 miRNA species; let-7, miR-30, miR-181 and two sequential binding sites for miR-26a (Fig 5C). While comparison of human and mouse nucleotide sequence in this region reveals a high degree of sequence conservation (92.7% identical), the zebrafish sequence shares little conservation with the exception of a small region containing a miR-181 and one copy of the miR-26a binding sequences. Let-7, miR-181 and both miR-26a binding sequences are shared between human and mouse, while the miR-30 binding sequence found in the human sequence is lost in mice (Fig 5D).

Figure 5. NLK 3’UTR metformin-response element contains 4 potential miRNA binding sites and is highly conserved between human and mouse but divergent in zebrafish.

(A) Schematic representing the many potential miRNA binding sites within the human NLK 3’UTR (upper). Diagrammatic representation of the full length and 5 truncated constructs fused to the luciferase gene and engineered to determine the metformin-responsive element within the NLK 3’UTR (lower). (B) After transient transfection of plasmids carrying the various 3’UTR fragments, K562 cells were treated with indicated concentrations of metformin for 72 hrs and assessed for luciferase activity and endogenous NLK and GAPDH protein expression by Western blot. (C) A schematic indicating the potential miRNA binding sites contained within the metformin-responsive element. (D). The corresponding nucleotide sequences of the human, mouse and zebrafish NLK 3’UTRs across the metformin-responsive element. Data are displayed as means +/− SD. Statistics: two-tailed Student’s t test, significant *P < 0.05

NLK suppression is mediated by miR26a induction by metformin in human, but not murine models of DBA.

To identify which miRNA was responsible for metformin-induced NLK suppression, we compared the expression of a number of miRNA species between untreated and metformin-treated CD34+ progenitors differentiating in erythroid media. As the metformin-responsive region of the NLK 3’UTR contained predicted binding sites for let-7, miR-30, miR-181 and miR-26a, we initially focused on these. With the exception of miR-26a, metformin did not alter the expression of any miRNA species. In contrast, miR-26a was upregulated 2.4-fold in response to metformin (Fig 6A upper). Comparison of the entire 3’UTR between human, mouse and zebrafish revealed that, apart from miR-181 and miR-26a within the metformin-response element, the only other miRNA-binding sites conserved between all 3 species was miR-199 and miR-144. However, metformin did not alter expression of either of these miRNAs in human erythropoiesis (Fig 6A upper). As the murine 3’UTR sequence of NLK is metformin-responsive when expressed in human cells, we examined the miRNA expression profile in response to metformin in differentiating Lin-Kit+ murine progenitors. In contrast to the human system, none of the miRNA species examined demonstrated significant upregulation (Fig 6A lower).

Figure 6. Metformin-mediated upregulation of miR-26a in human progenitors suppresses NLK expression.

(A) Human CD34+ (upper) and murine Lin-Kit+ (lower) progenitors were differentiated for 8 days in the presence or absence of metformin and the levels of indicated miRNAs were assessed by qRT-PCR. (B) K562 cells stably expressing the luciferase gene coupled to the NLK 3’UTR were mock transfected, transfected with indicated miRNA mimetics, or treated with metformin, and cultured in the presence or absence of metformin for 72 hrs. After lysis, luciferase activity was determined and endogenous NLK and GAPDH protein expression was analyzed by Western blot. (C) K562 cells were mock transfected, or transfected with either miR-34 or miR-26 sponges, prior to being left untreated or treated with metformin for 72 hrs. Lysed cells were subjected to luciferase assay and western blotting to examine NLK and GAPDH protein expression. Data are displayed as means +/− SD. Statistics: two-tailed Student’s t test, significant *P < 0.05

Having observed upregulation of miR-26a in response to metformin, and the presence of a miR-26a binding site within the metformin-responsive element of the NLK 3’UTR, we asked whether the metformin response is due to miR-26a upregulation. In human K562 cells expressing luciferase fused to a wild-type human NLK 3’UTR, recombinant expression of a miR-26a mimetic, but not miR-30 or miR-34 mimetics, reduced luciferase activity and endogenous NLK protein expression by 58.8% (p=0.0155) and 55.1% respectively (p=0.0311) (Fig 6B). The effect of miR-26a mimetics reduced luciferase and endogenous NLK expression similarly to that observed with metformin treatment (72.5% and 68.6%). A striking suppression of NLK has been reported in response to miR-181(7). Suppression of luciferase activity and endogenous NLK by miR-181 mimetics was more extensive (85.6% and 88.0%) than was observed in response to metformin or miR-26a mimetics (Fig 6B). Furthermore, expression of a miR-26a inhibitor reduced both the luciferase and endogenous NLK metformin-response by 74.1% (p=0.0416) and 70.3% (p=0.0449) respectively, whereas a miR-34 inhibitor had a negligible effect (Fig 6C).

The preceding data support a model in which metformin induces miR-26a expression in differentiating human hematopoietic cells. MiR-26a binds a sequence within the NLK 3’UTR facilitating NLK mRNA degradation. The mouse NLK 3’UTR sequence does include miR-26a binding sites, however in differentiating hematopoietic cells no miR-26a induction was observed in response to metformin in mice.

As metformin increases erythropoiesis in RPS19-insufficient human CD71+ progenitors, we compared CD235+ erythroid expansion in response to metformin, with expression of miR-26a mimetics, or a combination of both. RPS19-insufficiency decreased production of CD235+ erythroblasts to 8.3% of controls, however miR-26a mimetics, metformin treatment, and the combination all improved erythropoiesis similarly (3.4-fold; p=0.0324, 3.6-fold; p=0.0394 and 3.5-fold; 0.0279 respectively). No significant difference was observed between miR-26a mimetic alone compared to metformin alone (p=0.8544) or when treated together with metformin (p=0.9615). Negligible effects on erythropoiesis were observed in control (shLuc) with any treatment (Fig 7A upper).

Figure 7. MiR-26a mimetic substitutes for metformin in improving erythropoiesis in RPS19-insufficiency.

(A) Human CD34+ cells were transduced with shRNA against a control (shLuc) or RPS19 (shRPS19) and transfected with a miR-26a mimetic or a mock control. After differentiation in the absence or presence of metformin for 12 days, cells were counted and the percentage of CD235+ erythroblasts determined by flow cytometry. The calculated CD235+ population of each treatment goup is represented as a percentage of the untreated control group (upper). In parallel, qRT-PCR was performed to determine mRNA expression of NLK (middle) and miR-26a (lower). (B) Lin-Kit+ progenitors for RPL11+/+ and RPL+/lox mice were mock transfected or transfected with miR-26a mimetic and differentiated for 8 days. Cells were subjected to counting and the percentage of Ter119+ erythroblasts determined (upper) in conjunction with qRT-PCR examining NLK (middle) and miR-26a (lower) mRNA expression. Data are displayed as means +/− SD. Statistics: two-tailed Student’s t test, significant *P < 0.05

NLK expression was similarly suppressed by metformin, miR-26a, and the combination in both RPS19-insufficient and control cells (Fig 7A middle). Metformin increased miR-26a in both RPS19-insufficiency (2.5-fold, p=0.0001) and control (2.4-fold, p=0.0088), while miR-26a mimetics elevated levels to 3.6-fold (p=0.005) and 3.5-fold (p=0.0012). Combined treatment did not significantly increase miR-26a expression relative to either treatment alone (Fig 7A lower). Despite not responding to metformin (Fig 3A), the expression of miR-26a mimetics in murine Lin-Kit+ induced a moderate (1.75-fold; p=0.0166) increase in ter119+ erythroid expansion (Fig 7B upper), with corresponding NLK suppression (Fig 7B middle).

Collectively, we demonstrate that metformin improves erythroid expansion in RPS19- and RPL11-insufficient human (Fig 1) and zebrafish (Fig 3) models of DBA. The effect appears to be exclusively mediated through the suppression of NLK expression (Fig 2). Our results showed that metformin induces miR-26a (Fig 6) that binds a consensus sequence in the NLK 3’UTR (Figs 4 & 5). Furthermore, the expression of miR-26a mimetics had no significant difference with metformin treatment regarding improvement of erythropoiesis in human models of DBA (Fig 7). While metformin did not improve erythropoiesis in murine models of DBA (Fig 3) due to a failure to induce miR-26a (Fig 6), expression of miR26a mimetics improved erythroid expansion in the murine model (Fig 7).

Discussion

Here we report that metformin treatment improves erythropoiesis in human and zebrafish models of DBA. We hypothesized NLK-independent metformin effects would inhibit human erythropoiesis in ribosomal insufficiency. We detected no evidence to support this in vitro and metformin was well tolerated at the indicated doses. In human cells, metformin-induces upregulation of miR-26a, which targets NLK for degradation. In murine models, metformin failed to induce miR-26a, however inducing miR-26a expression with transduced cDNA effectively suppressed NLK expression and improved erythropoiesis.

In addition to miR-26a, the 3’UTR of NLK is particularly susceptible to miR-181 binding, a feature that has been documented in hepatocellular carcinoma (35) and natural killer (NK) cell development (7). It is intriguing that the NLK 3’UTR is more sensitive to miRNA-mediated degradation than others. While a complementary miRNA seed sequence is critical, numerous other factors contribute to defining the efficacy with which a miRNA will bind and/or degrade a transcript. These include GC content, sequence and secondary structures that dictate stability of both miRNA and RISC complex binding, as well as intrinsic diversity of 3’UTR length and structure across copies of some transcripts (36). While mouse and human NLK 3’UTR sequences are highly conserved, the zebrafish 3’UTR is highly divergent. Nevertheless, one copy of the miR-26a binding site and the miR-181-binding site is conserved. The upregulation of miR-181 after the MEP stage is critical for megakaryocyte lineage differentiation (37). Indeed, upregulation of miR-181 is induced in all non-erythroid lineages (37–41). Mutation of the miR-181 binding site in the NLK 3’UTR increased NLK expression in non-erythroid progenitors resulting in adverse lineage expansion (5). The importance of NLK expression in lineage differentiation may contribute to the evolution of the highly sensitive NLK 3’UTR. It is also possible that the highly conserved role of miR-181 in the regulation of hematopoiesis provides evolutionary pressure to conserve the region surrounding miR-181, thus maintaining the miR-26a binding site by association.

Metformin treatment did improve o-dianizidine in RPS19-insufficient zebrafish. However, the staining was predominantly along the midline and did not restore cardiac defects, as have been seen upon p53 suppression (42). This may be due to the incomplete rescue by NLK suppression, or that heart defects are an NLK-independent effect of RPS19 insufficiency in fish. Zebrafish carry two nlk homologues, nlk1 and nlk2. Nlk2 is implicated in similar signaling pathways as human and murine homologues, but a role in hematopoiesis has not been examined (43). We have no direct evidence that nlk1 or 2 is activated, nor that metformin suppresses NLK expression in RPS19-insufficient zebrafish erythroid cells. However, given that metformin improved erythropoiesis in RPS19-insufficient embryos and metformin treatment degraded zebrafish NLK 3’UTR transcripts, it is highly probable that NLK is similarly deregulated in and metformin-sensitive as human erythroblasts.

Another interesting result was the failure of miR-26a induction in response to metformin in murine erythroblasts. The deregulation of miRNAs in response to metformin has been evaluated in numerous cell and animal models. Effects of metformin have ranged from no miRNA deregulation (44) to hundreds (45). As miRNA expression profiles commonly differ across cell type and species, it is not surprising that miR-26a is not upregulated in response to metformin in RPS19-insufficient hematopoietic cells of all species. Metformin-mediated upregulation of miR-26a has been previously reported, inducing apoptosis in oral cancer cells (46) and contributing to reduced incidence and increased survival in breast cancer patients (47).

MiR-26a has been primarily linked to apoptosis, serving as pro-apoptotic (48–50) and anti-apoptotic (51), depending on the cell context. Other cellular roles attributed to miR-26a include activation of mTOR signaling by suppression of PTEN (52) (which would be potentially beneficial in DBA), suppression of ERBB2 (53), and suppression of HMGA1 (54).

Suppression of another MAPK-family kinase by miR-26a has been previously reported (55) in rats but the miR-26a-binding site present in the rat MAPK6 3’UTR sequence is not retained in the human sequence. Analysis of 3’UTR sequences across the human MAPK/cdk kinase family indicate no predicted miR-26a-binding sites within MAPK1–10 or MAP2K1–6 and 1 in each of MAP3K1 and MAP3K2 but none in MAP3K3–8. There were three miR-26a-binding sites within the 3’UTR of cdk8 but none were identified in cdk1–7, 9 or 10.

As NLK expression can have both tumor suppressive (9, 35, 56–59) and oncogenic (10, 16, 60–67) functions, depending on the origin of the malignancy, systemic NLK suppression may have significant risks. In a disease such as DBA in which most patients present early in life and some require life-long therapy (1), minimizing effects beyond the affected tissue becomes even more prudent. Many miRNAs demonstrate differential expression between tissues (13), so miRNA mimetics that suppress NLK in tissues of interest, but preserve NLK activity in tissues where NLK expression is favorable, may be advantageous. While the mechanism of metformin is through upregulation of miR-26a, improved NLK suppression and/or tissue-specificity may be engineered by targeting other miRNA-binding sites within the 3’UTR.

P53 suppression, mTOR stimulation and steroid regimes, impact regulators of multiple critical cellular processes. Targeting NLK may offer less potential side-effects. NLK null mice are viable and grow normally (68). NLK expression is not ubiquitous and enzymatic activity, like other kinases, is highly regulated (69). NLK likely plays no role during normal hematopoiesis and NLK expression is suppressed in all non-erythroid hematopoietic lineages. In normal erythroblasts, NLK activity is low and silencing NLK has no apparent impact on erythropoiesis. An influence of NLK on erythropoiesis is only established upon aberrant activation, which occurs downstream of p53-stabilization in DBA (5). These properties make agents that specifically target NLK expression in hematopoietic cells highly attractive.

Targeting miR-26a mimetics to hematopoietic cells in a murine system has already been effectively demonstrated using aptamer-based target delivery (70). Murine HSPCs highly express c-kit and Tanno and a miRNA-aptamer chimera containing miR-26a mimic and c-kit-targeting aptamer was successfully delivered miR-26a into HSPCs to attenuate the toxicity of 5’fluorouracel and carboplatin. Specific targeting of NLK in RP-insufficient hematopoietic cells coupled with naturally limited expression of NLK within non-erythroid lineages, offers a more selective therapeutic approach for DBA patients.

Highlights.

• Metformin improves erythropoiesis in human and zebrafish models of DBA.

• Metformin suppresses nemolike-kinase (NLK) expression to rescue erythroid expansion.

• Metformin induces miR-26a that binds the NLK 3’untranslated region to promote transcript degradation.