Exosomal miR-26a delivery increases erythroid expansion in models of Diamond-Blackfan anemia

Abstract

Diamond-Blackfan anemia (DBA) is a pure red blood cell aplasia with a predisposition toward cancer. In DBA erythroid progenitors, nemo-like kinase (NLK) is activated, and mammalian target of rapamycin complex 1 (mTORC1) activity is suppressed. Systemic leucine administration high enough to hyperstimulate mTORC1 improves erythroid expansion in patients and animal models of DBA. However, leucine at such high doses significantly increases malignant transformation of nonerythroid cells upon exposure to carcinogens, heightening cancer risk in these patients. Here we show that exosomal delivery of miR-26a restores basal mTORC1 activity in erythroid progenitors by targeting the NLK transcript for degradation. miR-26a delivery does not hyperstimulate mTORC1, but removal of NLK-mediated mTORC1 suppression primes erythroid cells for mTORC1 hyperstimulation at a 10-fold lower systemic dose of leucine. Delivery of exosomal miR-26a supports optimal mTORC1 hyperstimulation and erythroid expansion with the administration of a leucine dose that does not increase the risk of transformation in nonerythroid cells, thus offering a safe, novel, and effective treatment strategy for increasing erythropoiesis in DBA.

Introduction

Diamond-Blackfan anemia (DBA) is an inherited bone marrow failure syndrome associated with severe anemia, congenital malformations, and an increased risk of developing cancers. Many patients with DBA require frequent blood transfusions and chelation therapies, chronic steroid administration, or hematopoietic stem cell (HSC) transplantation.^1,2 Because all of these carry significant morbidity, improving patient care is an urgent unmet clinical need.

Leucine is an essential amino acid capable of stimulating mammalian target of rapamycin complex 1 (mTORC1) with demonstrated efficacy in cell and animal models of DBA at relatively high doses.^3-6 However, concerns of increasing systemic mTORC1 activity contributing to carcinogenesis warrant approaches that minimize systemic leucine exposure. Indeed, patients with DBA have a predisposition to a number of cancers, particularly osteosarcoma, colon, and breast cancer.⁷

Nemo-like kinase (NLK) is an evolutionarily conserved kinase with important cellular roles.⁸ NLK is chronically hyperactivated in the erythroid progenitors of patients with DBA, and NLK inhibition has been shown to improve erythropoiesis using in vitro models.^9-12

The pathogenic impact of hyperactivated NLK in DBA is suppression of mTORC1 activity.¹² This prevents the translational upregulation of critical mitochondrial biogenesis transcripts during early erythroid expansion.¹² Thus, inhibiting NLK activity in DBA is a potential therapeutic intervention. No small molecules that specifically target NLK are clinically available, but NLK can be inhibited as an off target by a number of small-molecule inhibitors of other kinases.⁸ SD208 inhibits NLK activity and improves ribosome-insufficient erythropoiesis.⁹ Metformin¹⁰ and ginsenoside Rb1¹¹ suppress NLK expression and improve erythropoiesis through the upregulation of miR-26a and miR-208, respectively. Because of challenges developing specific NLK inhibitors, targeting NLK expression with microRNAs (miRNAs) offers a promising therapeutic alternative.

Exosomes are a subset of extracellular vesicles capable of delivering nucleic acid cargoes, such as NLK-targeting miRNAs, to specific cell types. Multipotent mesenchymal stem cells (MSCs) are an effective source of exosomes and are largely non-immunogenic. Cell-type–specific cargo delivery can be achieved by the expression of ligands on the surface of exosomes that recognize receptors on target cells.

In this study, we demonstrate that exosomal delivery of miR-26a (Exo^miR-26a) inhibits NLK and prevents the NLK-mediated suppression of mTORC1 in ribosome-insufficient erythroid progenitors. Our results demonstrate that Exo^miR-26a allows optimal mTORC1 stimulation using a 10-fold lower dose of leucine. Of note, unlike high doses of leucine, a lower dose of leucine in combination with Exo^miR-26a does not significantly increase cellular transformation in nonhematopoietic cell types.

Methods

Cell culture and lentiviral transductions

Cell culture and lentiviral transductions were performed as previously published.^9-12 Ribosome-insufficient cells were generated by lentiviral transduction of short hairpin RNA against RPS19 coexpressing green fluorescent protein. Unless otherwise defined, cultures were presorted by fluorescence-activated cell sorting for high green fluorescent protein fluorescence 2 days after transduction or before analysis, depending on the parameters of the assay. To induce cellular transformation, indicated cell types were treated overnight at 37°C with 30 μL of a stock solution of 50 mg/mL of the carcinogen N-ethyl-N-nitrosourea (ENU; 50 μg/mL) to enhance genomic instability. Further details are provided in supplemental Methods.

Flow cytometry and western blotting

Detailed protocols and a list of antibodies and the concentrations used are provided in supplemental Methods. Kinase assays and quantitative reverse transcription polymerase chain reaction (qRT-PCR) were performed as previously published.^9-12

Mice

Inducible RPL11 heterozygous deletion mice were induced by tamoxifen (Teklad; Harlan Laboratories) to induce activation of the CreERT2 transgene. Animals were maintained at the Spanish National Cancer Research Centre under specific pathogen-free conditions, and all procedures were evaluated and approved by the Ethical Committee of the Carlos III Health Institute, Madrid, Spain (54-2013-v2).

Engineering exosomes with miR-26a mimetics

The human miR-26a-5p mimics (5'-UUC AAG UAA UCC AGG AUA GGC U-3'; Sigma-Aldrich Co, LLC) and negative control (5'-GGU UCG UAC GUA CAC UGU UCA-3'; Sigma-Aldrich Co, LLC) were loaded into MSC exosomes using the Neon electroporation system (Life Technologies, Carlsbad, CA). After ultracentrifugation, exosomes were dissolved in R buffer to ensure a concentration of ≥1 mg/mL, and 10 μL of miR-26a mimic (0nM-50nM) was then added to 90 μL of electrotransfer solution at a mass ratio of 1:1. This mixture was electrically transformed (1000 V, 10 milliseconds, 2 pulses), and 1 mL of phosphate-buffered saline was added before incubation at 37°C for 1 hour to promote exosome repair. Engineered exosomes with miR-26a mimics (Exo^miR-26a) were obtained by ultrasonication. miR-26a levels were assessed using qRT-PCR.

Exosome preparation

Exosomes were isolated from the cellular supernatant of MSC lines cultured in serum-free Dulbecco modified Eagle medium (1% Glu-taMax; Invitrogen, Carlsbad, CA) for 48 hours. The supernatant was centrifuged at 2000g for 20 minutes and then at 10 000g for 30 minutes to remove cell debris. Exosomes were pelleted in small amounts by ultracentrifugation at 120 000g for 70 minutes and then washed once with phosphate-buffered saline. The protein content of exosomes was measured using a rapid protein measurement kit as per manufacturer instructions (Wako Pure Chemicals, Osaka, Japan). The average amount of exosomes in 100 mL of the supernatant (1 × 10⁵ cells/mL; n = 5) was 68.3 μg.

Statistics

The P values for statistical significance were obtained by using a paired Student t test. Significance was designated as P < .05. The data are representative of ≥3 independent experiments (n = 3). Where appropriate, experiments were performed in triplicate with the mean reported as a single data point.

Results

Administration of NLK inhibitors or leucine, independently or in combination, yields different erythropoiesis outcomes by regulating differential mTORC1 activity

The activity status of mTORC1 is not binary, instead fluctuating through a continuum between extreme hypo- and hyperactivation. Because NLK suppresses and leucine stimulates mTORC1 activity, we first examined the mTORC1 activity status in control and ribosome-insufficient erythroid progenitors and their response to treatment with small-molecule NLK inhibitors both independently and in combination with leucine.

The addition of 10mM leucine increased mTORC1 activity in both control and ribosome-insufficient erythroid progenitors. Addition of leucine to the control group increased phosphorylation of 4E-BP1 by 63.6% (P = .0184), and the ribosome-insufficient progenitors increased phosphorylation from 30.2% to 57.2% of controls (P = .009; Figure 1Ai). Similar results were obtained when S6K was the substrate (supplemental Figure 1A). In contrast, and consistent with the fact that NLK is not hyperactivated in control cells, treatment with NLK inhibitors SD208 (2μM), metformin (30mM), and ginsenoside Rb1 (10mM) had no impact on mTORC1 activity. However, in ribosome-insufficient cells, each inhibitor rescued mTORC1 activity from 28.2% to 102.4%, 84.4%, and 72.1%, respectively (Figure 1Ai). Furthermore, although leucine (10mM) did not affect NLK activity against raptor (Figure 1Aii) or 2 other verified substrates (supplemental Figure 1B) in ribosome-insufficient cells, SD208 (P = .0011), metformin (P = .0017), and ginsenoside Rb1 (P = .0015) all robustly reduced NLK activity, confirming that NLK functions upstream of mTORC1 suppression.

Figure 1. NLK suppression restores mTORC1 activity to basal levels, and high concentrations of leucine hyperstimulate mTORC1 to increase expansion of erythroid cells yet promote transformation in nonerythroid cells.

(A) Cord blood CD34⁺ HSPCs were transduced with short hairpin RNA (shRNA) against RPS19 or control and differentiated in the presence of vehicle (dimethyl sulfoxide), leucine (10mM), SD208 (2μM), metformin (30mM), and ginsenoside Rb1 (10mM) for 5 days before assessment of mTORC1 kinase activity toward 4E-BP1 (i), or NLK activity toward raptor (ii). (B) EC₅₀ was determined for SD208 (i), metformin (ii), ginsenoside Rb1 (iii), and leucine (iv), in differentiated CD34⁺ HSPCs transduced with shRNA against RPS19 or control. Erythroid expansion (CD235⁺) was assessed at day 12 by flow cytometry and viability counts. (C) NCM460 (i), BMSCs (ii), and MCF10As (iii) expressing shRNA against RPS19 or controls were cultured in the presence of 50 μg/mL ENU overnight before plating in soft agar containing the indicated concentrations of leucine. Colony numbers in soft agar were assessed after 10 to 12 days. (D) Parallel samples were cultured in liquid media for 5 days before assessment of mTORC1 kinase activity as determined by an mTORC1 in vitro kinase assay. Data are represented as mean ± standard deviation (SD). See also supplemental Figures 1 and 2. *Significant value defined as P < .05 (n = 3).

We next examined the impact of increasing concentrations of NLK inhibitors, metformin, or ginsenoside Rb1 in ribosome-insufficient cells and observed concentration-dependent increases in mTORC1 activity with 50% effective concentration (EC₅₀) values of 442nM, 11.7mM, and 2.4mM, respectively. Maximal mTORC1 activity aligned with maximal NLK inhibition,^9-11 indicating that the effect of these compounds on mTORC1 in ribosome-insufficient cells is because of NLK suppression (Figure 1Bi-iii). As shown in Figure 1A, NLK inhibition had no impact on mTORC1 activity in healthy controls where NLK is not activated. In contrast, increasing concentrations of leucine (iv) increased mTORC1 activity in both control and ribosome-insufficient cells, with EC₅₀ in controls of 0.16mM and EC₅₀ in ribosome-insufficient cells of 2.4mM. Additionally, although NLK inhibitors improved mTORC1 activity in ribosome-insufficient cells compared with that observed in controls, increasing leucine raised mTORC1 activity above that observed in healthy controls, with maximal mTORC1 stimulation at 1.9-fold above controls at concentrations >30mM. Rescue of mTORC1 to basal activity was ~2.5mM (Figure 1B).

Collectively, our data are consistent with a model that during healthy erythropoiesis, mTORC1 activity is moderately activated, whereas in DBA, NLK hyperactivation results in mTORC1 hypoactivity. Inhibition of NLK and doses approaching 2.5mM leucine restore basal mTORC1 activity, but higher doses of leucine hyperactivate mTORC1.

Hyperactivation of mTORC1 increases transformation of nonerythroid cells

Because mTORC1 hyperactivation is common in many cancers,¹³ we sought to correlate doses that affect erythropoiesis in DBA models with cellular transformation in nonerythroid cells. Because spontaneous cellular transformation is a rare event, we treated cell models with the carcinogen ENU.

Similar to patients with DBA, our results show that in the presence of ENU, ribosome insufficiency increased the transformation frequency of colon epithelial cells (NCM460) (i) 3.0-fold (P = .0121), bone marrow stromal cells (BMSCs) (ii) 4.3-fold (P = .0029), and breast epithelial cells (MCF10A) (iii) 3.04-fold (P = .0003; Figure 1C). Importantly, as observed in patients with DBA, colony formation is not significantly increased in ribosome-insufficient cells in the absence of ENU (supplemental Figure 2), indicating that DBA does not spontaneously cause cancer but predisposes cells to transformation when exposed to transforming agents.

The addition of 1mM leucine did not significantly increase transformation of control or ribosome-insufficient NCM460s (i), BMSCs (ii), or MCF10As (iii). In contrast, 10mM leucine increased the observed transformation rates in control and ribosome-insufficient cultures reaching statistical significance in control BMSCs (ii) 2.8-fold (P = .003), MCF10As (iii) 2.38-fold (P = .0004), and ribosome-insufficient NCM460s (i) (P = .0003), 1.7-fold in BMSCs (ii) (P = .0003) and 1.56-fold in MCF10As (iii) (P = .0024; Figure 1C).

The activity of mTORC1 was also assessed. Presumably because NLK expression is low in epithelial and fibroblast cells, ribosome insufficiency did not affect mTORC1 activity in these cells (Figure 1D). The addition of 1mM leucine increased mTORC1 activity 1.6-fold (P = .0009), 1.42-fold (P = .0021), and 1.7-fold (P = .0007) in NCM460, MCF10A, or BMSCs, respectively. The addition of 10mM leucine increased mTORC1 activity 4.0-fold (P ≤ .0001) in NCM460, 3.69-fold in MCF10As (P ≤ .0001), and 3.9-fold (P ≤ .0001) in BMSCs (Figure 1D). These results demonstrate that increasing leucine concentrations from 1mM to 10mM significantly increases transformation in ENU-treated NCM460 (P = .0014), MCF10As (P ≤ .0001), and BMSCs (P = .0008), suggesting the benefit of increased erythropoiesis with 10mM leucine must be weighed against the negative impact of increased transformation potential when considering treatment of patients with DBA.

Impact of NLK inhibitors alone or in combination with low-dose leucine on erythroid expansion in DBA models

As 1mM leucine did not significantly increase transformation in ribosome-insufficient nonerythroid cells exposed to carcinogens, we tested this lower concentration on erythroid expansion and observed a modest 1.5-fold increase in mTORC1 activity (P = .0047) in controls and 2.2-fold in ribosome-insufficient progenitors (P = .0026), with corresponding increases in erythroid expansion of 1.4-fold (P = .0351) and 1.9-fold (P = .003), respectively. Coadministration of 1mM leucine with 2μM SD208 (a concentration that maximally inhibits NLK) did not further increase mTORC1 activity or erythropoiesis in controls but did increase mTORC1 activity from 2.2-fold with leucine alone to 5.4-fold (P = .0021) in combination with SD208. This increased mTORC1 activity corresponded with an increase in erythropoiesis from 1.9-fold with leucine alone to 5.8-fold in combination (P = .0015; Figure 2A).

Figure 2. Leucine and NLK suppression stimulate mTORC1 activity by different mechanisms and synergize when administered in combination.

(A) Differentiating HSPCs transduced with shRNA against RPS19 or control were assessed for mTORC1 (i) and NLK (ii) activity at day 5 and expansion of CD235⁺ erythrocytes at day 12 (iii). Cultures were differentiated in the presence of leucine (1mM) or SD208 alone (2μM) or in combination. (B) Lin-Kit⁺ HSPCs from RPL11^+/+ or RPL11^+/Δ mice were differentiated for 5 days and assessed for mTORC1 (i) and NLK (ii) activity or for 10 days and expansion of Ter119⁺ erythrocytes (iii) assessed by flow cytometry. (C) Mononuclear cells from bone aspirates of a control and patients with DBA were differentiated as above, with mTORC1 (i) and NLK (ii) activity and CD235⁺ erythroid expansion (iii) determined. (D) EC₅₀ of leucine alone or in combination with SD208, metformin, or ginsenoside Rb1 was calculated for RPS19-insufficient erythrocytes differentiated for 12 days. (E) Schematic describing that NLK suppression increases mTORC1 activity by allowing the complex to localize more efficiently at the lysosomal membrane, whereas leucine acts by recruiting more Rheb (an upstream activator of mTORC1) to the lysosomal membrane. Control erythroblasts (upper left) and DBA erythroblasts with mTORC1 localization disrupted (lower left). Control erythroblasts with excess leucine recruiting more Rheb to the lysosomal membrane (upper right). DBA erythroblasts with excess leucine in which mTORC1 remains mislocalized (lower right). Data are represented as mean ± SD. *Significant value defined as P < .05 (n = 3).

Leucine (1mM) and SD208 (2μM) similarly modulated mTORC1 and NLK and erythropoiesis in differentiating mouse hematopoietic stem and progenitor cells (HSPCs). Leucine (1mM) alone did not significantly improve expansion of Ter119⁺ RPL11^+/Δ erythrocytes (P = .1449), yet coadministration of leucine and SD208 increased erythropoiesis from 27.5% to 59.8% (P = .0091) of healthy controls (Figure 2B). We also observed the same trend in erythroid differentiation from the bone marrow aspirates of control patients and patients with DBA (Figure 2C). Leucine and SD208 modulated activity of mTORC1 (i) and NLK (ii). Leucine (1mM) alone did not significantly increase erythropoiesis (P = .1391), but CD235⁺ erythrocytes increased from 14.4% to 35.9% (P = .016) of controls with 1mM leucine and SD208 combined (iii).

Synergy was observed when ribosome-insufficient cells were cotreated with leucine and any NLK inhibitor (from 2.4mM with leucine alone to 0.1mM with 2μM SD208, 0.11mM with 30mM metformin, and 0.11mM with 10mM ginsenoside Rb1; Figure 2D). The combination index values were 0.2, 0.25, and 0.28 for leucine combined with SD208, metformin, and ginsenoside Rb1, respectively, at 50% of maximal erythroid rescue. These data are consistent with a model where NLK-mediated phosphorylation of raptor suppresses the ability of mTORC1 to localize to the site of leucine action, the outer lysosomal membrane. Thus, preventing raptor phosphorylation allows mTORC1 to localize to the lysosomal membrane where leucine can act at full potency (Figure 2E).

Together, these results suggest that leucine treatment alone requires a concentration of 2.5mM to restore basal levels of mTORC1 activity. This can similarly be achieved by inhibiting NLK with SD208, metformin, or ginsenoside Rb1 at concentrations >2μM, 30mM, and 10mM, respectively (Figure 2D). Although increasing mTORC1 stimulation above basal levels in these cells further improves erythropoiesis, it also increases transformation in nonerythroid cells.

Exosome-delivered miR-26a inhibits NLK and improves erythropoiesis in ribosome-insufficient hematopoietic progenitors

The presented data indicate that a preferred alternative to giving high doses of leucine would be to target NLK expression specifically in erythroid progenitors to prime these cells for mTORC1 hyperstimulation with a lower dose of leucine given systemically. A major limitation of this approach is that no specific NLK inhibitors have been developed, and knockout mouse models suggest that systemic NLK inhibition may generate serious toxicities. To address this, we have generated exosomes that deliver NLK-targeting miR-26a to hematopoietic progenitors (Exo^miR-26a).

Because NLK is expressed in HSPCs and retained only in erythroid progenitors,⁹ targeting CD34⁺ HSPCs should only affect NLK in erythroid progenitors. Expression of E-selectin on the membrane can target these exosomes specifically to CD34 receptors,¹⁴ thereby targeting CD34⁺ HSPCs.

Using a delivery system developed by Mahati et al,¹⁵ we loaded miR-26a into exosomes generated by MSCs transduced with E-selectin (Exo^miR-26a). Expression of E-selectin was validated by flow cytometry (Figure 3A), and the abundance of miR-26a contained within the exosomes was determined by qRT-PCR (Figure 3B) and calculated to be ~19.8 miRNA/exosome. To ensure the effective delivery of exosome contents, CD34⁺ HSPCs were exposed to 20 μg of exosomes loaded with the fluorescent dye carboxyfluorescein succinimidyl ester (CFSE), washed, and visualized by confocal microscopy. The presence of CFSE was detected only in cells treated with Exo^CFSE, confirming exosomes were targeted to and internalized by CD34⁺ cells with an increase from 4.0% to 53.3% of cells expressing fluorescence after exosome treatment (P = .0068; supplemental Figure 3A).

Figure 3. Validating specificity of Exo^miR-26a on NLK expression in CD34⁺ HSPCs.

(A) MSCs stably expressing E-selectin (red) or not (black) (i) or exosomes derived from naïve (gold) or E-selectin–expressing MSCs (blue) (ii) were analyzed for E-selectin expression by flow cytometry. (B) MSC-derived exosomes were electroporated with or without miR-26a, and incorporated miR-26a was quantified by qRT-PCR. (C) CD34⁺ HSPCs, RFP-expressing K562 cells, colon epithelial cells, and human skin fibroblasts were cocultured and exposed to 5, 10, 20, and 50 μg of exosomes containing the fluorescent dye CFSE for 30 minutes. (D) After washing, the percentage of each cell type with intracellular CFSE staining was assessed by flow cytometry. (E) Exo^CFSE was added to a mixed coculture of cells for 24 hours before being washed off. After 7 days of differentiation, cells were separated, and the percentage of each cell population demonstrating internal fluorescence was determined by fluorescence-activated cell sorting. See also supplemental Figure 3. GFP, green fluorescent protein. Data are represented as mean ± SD. *Significant value defined as P < .05 (n = 3).

Next, we cocultured HSPCs in the presence of colon epithelial and skin fibroblast cells and red fluorescent protein (RFP)-expressing leukemic K562 cells. Cultures were treated with CFSE-containing exosomes expressing E-selectin (Figure 3C). At concentrations <20 μg, only CD34⁺ HSPCs readily took up exosomes, with ~50% of cells emitting fluorescence. As exosome concentrations approached 50 μg, off-target uptake increased in the other cell types (Figure 3D). 30 μg of exosome (1 × 10⁸ particles) was selected because of an optimal balance of maximal uptake vs low off-target uptake.

To ensure that the preferential uptake of exosomes in HSPCs was because of E-selectin expression, we cultured CD34⁺ HSPCs in the presence of 30 μg CSFE-labeled exosome for 24 hours. Exosomes were naive or transduced to express E-selectin, and a 46.3-fold (P = .003) increase in fluorescence was observed when E-selectin was expressed (supplemental Figure 3B). Analysis of cocultures treated with Exo^CFSE after 8 days of differentiation indicated that exosome contents are retained in differentiated progeny of CD34-expressing progenitors (Figure 3E).

To examine the impact of Exo^miR-26a on NLK expression in different cell types, CD34⁺ HSPCs, K562 leukemic cells expressing RFP, and MCF10A epithelial cells were cultured together and exposed to 30 μg Exo^miR-26a for 24 hours (Figure 4Ai). NLK expression was reduced by 62.4% (P = .0114) in HSPCs that were exposed to Exo^miR-26a, whereas K562 (P = .7639) and MCF10As (P = .9652) did not demonstrate a significant reduction (Figure 4Aii). Moreover, Exo^miR-26a-induced NLK suppression was maintained during differentiation as CD235⁺ erythrocytes differentiated for 10 days in the presence of Exo^miR-26a and demonstrated 62.6% less NLK expression (P = .0121), whereas CD41⁺ megakaryocytes and CD11b⁺ myeloid cells expressed minimal NLK, irrespective of exosome exposure (Figure 4Aiii). NLK expression was not altered when cultures were exposed to empty exosomes or those containing miR-34a-5p (supplemental Figure 4). Our data conclusively conclude that Exo^miR-26a treatment targets CD34⁺ HSPCs, resulting in a meaningful reduction in NLK expression that is restricted to erythroid progenitors.

Figure 4. Exo^miR-26a suppresses NLK and improves expansion of ribosome-insufficient erythroid progenitors without promoting transformation in nonerythroid cells.

(A) CD34⁺ HSPCs, RFP-expressing K562 cells, and MCF10A cells were exposed to 30 μg Exo^miR-26a and cultured for 48 hours (i). Suspension cells were sorted into RFP⁺ (K562) and RFP⁻ (CD34 ⁺ derived) populations, and attached cells (MCF10As) were collected before lysis and western blot analysis for NLK and GAPDH protein expression. Representative western blot image (ii) and collated qRT-PCR data (iii) are displayed. (iv) CD34⁺ HSPCs were treated with 30 μg Exo^miR-26a and cultured for 10 days before isolation of CD235⁺ erythroid, CD41⁺ megakaryocyte, and CD11b⁺ myeloid populations. Expression of NLK across hematopoietic cell populations was assessed by flow cytometry. (B) Cord blood CD34⁺ HSPCs transduced with shRNA against RPS19 and MCF10s were cultured in the presence or absence of 30 μg Exo^miR-26a for 10 days. MCF10A (i), CD235 ⁺ erythroid (ii), CD41⁺ megakaryocyte (iii), and CD11b⁺ myeloid (iv) populations were compared after cell viability counting and flow cytometry. (C) NCM460 (i), BMSCs (ii), and MCF10As (iii) expressing shRNA against RPS19 or controls were cultured in the presence of 50 μg/mL ENU O/N before plating in soft agar containing 30 μg of Exo^miR-26a. Colony numbers in soft agar were assessed after 10 to 12 days. (D) Parallel samples were cultured in liquid media for 5 days before assessment of mTORC1 kinase activity as determined by an mTORC1 in vitro kinase assay. Data are represented as mean ± SD. See also supplemental Figures 4 and 5. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. *Significant value defined as P < .05 (n = 3).

Having demonstrated that we could downregulate NLK expression in erythroid progenitors, we examined the impact of Exo^miR-26a treatment on the expansion of cell populations in ribosome insufficiency. Cocultures of HSPCs and MCF10A epithelial cells were transduced with lentivirus to induce ribosome insufficiency. CD34⁺ HSPCs and MCF10As with or without Exo^miR-26a were analyzed for the expansion of the lineages (Figure 4B). Exo^miR-26a did not affect expansion of any ribosome-competent lineage. Similarly, expansion of nonerythroid ribosome-insufficient lineage was not affected (Figure 4Bii,iv-v). In contrast, erythroid expansion improved from 20.7% to 46.6% of controls (P = .0097) when ribosome-insufficient CD34⁺ HSPCs were exposed to Exo^miR-26a (Figure 4Biii). The raw percentages of the lineages before and after normalization are presented in supplemental Figure 5. The addition of Exo^miR-26a did not influence transformation frequency (Figure 4C) or mTORC1 activity (Figure 4D). Importantly, we document that Exo^miR-26a-induced NLK suppression improves erythropoiesis in ribosome insufficiency to approximately the same extent as 2μM of the small molecule inhibitor SD208 (P = .072) but can be administered to a much more specific cell population than a systemically administered drug.

Exo^miR-26a delivery primes hematopoietic progenitors to hyperactivate mTORC1 and improve erythropoiesis at lower leucine concentrations

Releasing NLK-mediated mTORC1 suppression by NLK inhibition or low doses of leucine improves ribosome-insufficient erythropoiesis, but even greater levels of erythroid expansion can be achieved through mTORC1 hyperstimulation with higher doses of leucine. However, leucine at these doses increases transformation risk in nonerythroid cells. As Exo^miR-26a did not significantly increase the risk of transformation of nonerythroid cells, we sought to determine if we could remove the NLK-induced suppression of mTORC1 activity that occurs in erythroid progenitors using Exo^miR-26a to prime these cells to become susceptible to mTORC1 hyperactivation at a safer leucine concentration.

To investigate the synergy of erythroid-specific inhibition of NLK and leucine on proliferation and mTORC1 activity, we administered Exo^miR-26a alone or with 1mM or 10mM leucine to cocultures of differentiating ribosome-insufficient hematopoietic cells and breast epithelial cells (Figure 5Ai). Neither Exo^miR-26a nor leucine had a significant impact on the proliferation of nonerythroid cells. CD235⁺ erythrocytes were increased by 2.86-fold (P = .0295) and 1.85-fold (P = .0422) by Exo^miR-26a alone and 1mM leucine alone, respectively. In the presence of both Exo^miR-26a and 1mM leucine, erythrocytes increased 4.22-fold (P = .007). No additional increase was observed at 10mM leucine in the presence or absence of Exo^miR-26a (Figure 5Aii). These data support a model whereby supplementing 1mM leucine with Exo^miR-26a improves erythroid expansion in DBA models to the same extent as mTORC1 hyperactivation by high doses of leucine.

Figure 5. Targeting NLK expression primes erythroid progenitors to maximize mTORC1 activation and erythropoiesis when exposed to lower concentrations of leucine while limiting nonerythroid cell transformation.

(A) MCF10As and CD34⁺ HSPCs expressing shRNA against RPS19 were differentiated alone or in the presence of 30 μg Exo^miR-26a (i). After 8 days, control (no GFP) and ribosome-insufficient (GFP) CD235⁺ CD41⁺ and CD11b⁺ populations were sorted by fluorescence-activated cell sorting, whereas MCF10As were scraped from the plate and population quantified (ii), and mTORC1 kinase activity was analyzed by in vitro kinase assay against 4E-BP1 as substrate (iii). (B) Cord blood CD34⁺ HSPC and MCF10A epithelial cells were transduced with shRNA against RPS19 (co-expressing GFP) and treated with indicated concentrations of leucine in the presence or absence of 30 μg Exo^miR-26a for 8 days. After fluorescence-activated cell sorting for GFP expression, MCF10s were scraped and suspended in dimethyl sulfoxide containing 10% fetal bovine serum. The kinase activity of mTORC1 was assessed by in vitro kinase assay from part of each sample (ii), whereas the remaining cells were treated with 50 μg/mL ENU overnight and 2000 cells per plate were plated into soft agar. After 10 days, the number of colonies was assessed (iii). Data are represented as mean ± SD. See also supplemental Figure 6. *Significant value defined as P < .05 (n = 3).

Analysis of mTORC1 activity in nonerythroid cells revealed that 1mM leucine approximately doubled activity and 10mM approached double that again (Figure 5Aiii). In contrast, mTORC1 activity in ribosome-insufficient erythroblasts increased 1.88-fold with the addition of Exo^miR-26a alone and significantly increased to 4.92-fold (P = .0011) with Exo^miR-26a in the presence of 1mM leucine. At 10mM leucine, erythroblasts expanded 3.9-fold and 5.21-fold (P = .0068) compared with controls, which is not significantly different from 1mM leucine in combination with Exo^miR-26a (P = .1028), indicating mTORC1 hyperactivation can be achieved using a low combination of leucine if accompanied by Exo^miR-26a (Figure 5Aiii).

As leucine can potentially have many cellular effects other than mTORC1 stimulation, we assessed whether treatment with mTOR inhibitors could affect the effects of leucine and miR-26a. The addition of rapamycin or torin prevented any rescue effects of leucine or exosomes in controls (supplemental Figure 6A) or ribosome insufficiency (supplemental Figure 6B). It should be noted that both inhibitors significantly reduced erythropoiesis in both conditions (supplemental Figure 6C). Although miR-26a increased erythropoiesis in RPS19 insufficiency by 2.92-fold, rapamycin reduced this to 1.11-fold (P = .0096), and torin reduced erythroid expansion to 0.9-fold of controls (P = .0051).

Finally, control and ribosome-insufficient MCF10A cells were subjected to ENU treatment and were cultured in leucine and Exo^miR-26a mixtures (i). The activity of mTORC1 did not differ significantly between control and ribosome-insufficient samples, with incremental increases that correlated with leucine concentration. The presence or absence of Exo^miR-26a did not affect mTORC1 activity (iii). Controls (ENU treated but exposed only to vehicle) formed an average of 9.4 colonies per plate (c/p). The addition of leucine only significantly increased transformation at the highest (10mM) dose with 26.2 c/p (P = .0029). Ribosomal insufficiency in vehicle-treated controls increased transformation to 40.4 c/p (P = .003). The addition of 1mM leucine and the presence of Exo^miR-26a did not significantly further increase the transformation rate. However, 10mM leucine increased transformation to 68.4 c/p (P = .0008) (ii), further demonstrating the preferential safety of NLK inhibition in combination with a low concentration of leucine as compared with a high leucine dose.

MSC-derived exosomes are nontransformative and non-immunogenic, but miR-26a has multiple target transcripts in HSPCs.^16,17 As miR-26a has been linked to the regulation of a number of oncogenes, we assessed the suppression of 6 major miR-26a targets linked to transformation or other hematopoietic disorders in CD34⁺ HSPCs exposed to Exo^miR-26a. All 6 were suppressed, ranging from a 19.5% (P = .0196) reduction in PTEN to 56.9% (P = .0019) in BAK1 (Figure 6A). Exposure to Exo^miR-26a did not alter the lineage composition of differentiated HSPCs or rates of apoptosis (Figure 6B); however, because both MCL1 and BAK1 are downregulated miR-26a targets and have been linked to survival in hematopoietic stress conditions, we analyzed the impact of Exo^miR-26a when HSPCs were exposed to etoposide. Hematopoietic stress reduced expansion of CD235⁺ cells to 64.2% of controls (P = .0057) and CD11b⁺ cells to 68.7%. Empty exosomes had no impact, whereas Exo^miR-26a modestly improved erythropoiesis to 76.1% (P = .0427) and myelopoiesis to 78.9% (P = .0116) of controls (Figure 6Ci). Apoptosis was detected in only 5.9% of the untreated CD34⁺ HSPCs but increased to 44.3% upon etoposide treatment (P = .0089). Empty exosome had no effect, but Exo^miR-26a reduced apoptosis to 31.4% (P = .0385; Figure 6Cii). Although we did not see any changes in differentiation or proliferation during hematopoiesis, and we did not see any increases in transformation of nonhematopoietic cells with Exo^miR-26a treatment, we wanted to examine whether exosomes and leucine affected transformation predisposition in HSPCs. Treatment with multiple carcinogens did not generate colonies in soft agar. However, induced pluripotent stem cells that were differentiated into CD34⁺ HSPCs (induced pluripotent HSPCs) did form colonies upon exposure to ENU. Cell cotreatment with 10mM leucine modestly increased transformation (P = .0301), but Exo^miR-26a had no effect in control or ribosome insufficiency (Figure 6D).

Figure 6. Examination of miR-26a targets reveals no impact on normal hematopoiesis but may affect stress hematopoiesis.

(A) CD34⁺ HSPCs were cultured in the presence or absence of 30 μg Exo^miR-26a for 6 hours before analysis of 6 validated miR-26a target transcripts in HSPCs by qRT-PCR. (B) (i) CD34⁺ HSPCs were cultured in the presence or absence of 30 μg Exo^miR-26a for 8 days, and the composition of derived hematopoietic progenitors was assessed by flow cytometry. (ii) The extent of apoptosis was determined by annexin V staining after CD34⁺ HSPCs were cultured in the absence or presence of 30 μg Exo^miR-26a for 3 hours. (C) (i) Stress hematopoiesis was induced by treating CD34⁺ HSPCs for 3 hours with 20μM etoposide. Cultures were cocultured with vehicle alone, empty exosomes, or Exo^miR-26a, and untreated HSPCs were cultured concurrently as a baseline. After 3 hours, cells were switched to fresh media without etoposide and allowed to differentiate for 8 days before assessment of CD235⁺ erythroid and CD11b⁺ myeloid population expansion. (ii) Cultures were treated as above except that after incubation with etoposide for 3 hours, cells were washed and apoptosis was assessed by annexin V staining. (D) Induced pluripotent stem (iPS) cells differentiated to iP-HSPC and were treated with 50 μg/mL ENU and the indicated combinations of leucine and Exo^miR-26a overnight before 2000 cells per plate were plated into soft agar. After 10 days, the number of colonies was assessed (iii). Data are represented as mean ± SD. iP-HSPC, induced pluripotent HSPC. *Significant value defined as P < .05 (n = 3).

Collectively, our data indicate that the activation of NLK in ribosome insufficiency suppresses mTORC1, and leucine can overcome NLK-mediated suppression. Increased leucine concentrations further hyperactivate mTORC1, further improve erythropoiesis, and promote transformation in nonerythroid cells. However, NLK inhibitors allow the administration of lower concentrations of leucine to effectively hyperactivate mTORC1. Moreover, specifically targeting NLK with Exo^miR-26a overcomes the NLK-mediated suppression in hematopoietic progenitors. This, in turn, primes the hematopoietic progenitors for mTORC1 hyperstimulation with a 10-fold lower dose of leucine than would be required otherwise. Effective anemia-alleviating concentrations are low enough that systemic delivery does not significantly promote transformation in other cell types.

Discussion

Given that activated NLK suppresses mTORC1, this study was initiated with the expectation that NLK inhibitors and leucine would function essentially as facsimiles in improving erythropoiesis in DBA models. Our observation that hyperstimulating mTORC1 beyond the basal state produces a superior and NLK-independent erythroid rescue was unexpected. It is currently unclear how mTORC1 hyperstimulation improves erythroid expansion, but it is reasonable to hypothesize that the critical role mTORC1 plays in translational regulation would be a key area.

Patients with DBA usually present in early childhood with anemia,² but the increased potential for cancer is a consideration that should not be overlooked. As NLK is not abundantly expressed in the stroma and epithelial cells that regularly give rise to cancer, we accurately hypothesized that NLK inhibition would be preferable to stimulating the protransformative mTORC1 with higher doses of leucine. As the cancer risk remains elevated in patients with DBA, even after successful HSC transplantation or other hematological therapies, clinicians and researchers alike must stay cognizant and avoid options that may improve anemia at the risk of exacerbating carcinogenesis.

Because metformin-induced induction of miR-26a has been previously demonstrated to suppress NLK expression and improve erythropoiesis in DBA models,¹⁰ these studies were restricted to this miRNA. Our studies did not uncover alterations to hematopoiesis under homeostasis conditions. However, Exo^miR-26a treatment did increase the expansion of erythroid and myeloid lineages under hematopoietic stress. This correlates with our observed downregulation of the miR-26a targeting MCL1 and BAK1, contributing to the initiation of progenitor apoptosis under these conditions.^16-18 Mitochondrial and endoplasmic reticulum stress of HSPCs is transiently experienced during HSC transplantation and other DBA therapies that may be affected by increased miR-26a expression. It should be noted that upregulation of miR-208,¹¹ miR-199,¹⁹ miR-3184,²⁰ and miR-101²¹ has been demonstrated to suppress NLK expression, and, as the contents of exosomes are not restricted to single miRNAs, various combinations of NLK-targeting miRNAs could be packaged into exosomes for different clinical applications.

Exosomes hold promise as delivery vehicles that can carry toxic cargos and deliver them to specific cells of interest. In DBA, we have demonstrated that exosomes can be used to deliver miRNAs to ameliorate the erythroid-specific suppression of mTORC1 and sensitize these cells to mTORC1 hyperactivation. This increases red blood cell production in these cells to improve anemia while reducing cancer risk in patients with DBA.

Key Points.

• Suppression of NLK restores mTORC1 activity, but hyperstimulation increases erythropoiesis in DBA and also increases cancer risk.

• Exosomes delivering miR-26 prime erythroblasts to hyperstimulate mTORC1 with 10-fold less stimulant to improve anemia without cancer risk.