Myeloid cell–targeted miR-146a mimic inhibits NF-κB–driven inflammation and leukemia progression in vivo

MicroRNAs (miRNAs) are small noncoding RNAs that control expression of a broad set of target genes based on sequence complementarity. By binding to the 3′ untranslated (3′UTR) regions of the target messenger RNA (mRNA), miRNAs regulate gene expression and enable control of multiple gene targets within the same or distinct signaling pathways.^1,2  Many miRNAs are dysregulated in cancer and cardiovascular and autoimmune diseases.³  Genomic mutations, deletions, or changes in the key enzymes in miRNA biogenesis may all lead to alterations in miRNA levels.^4,5  Genome-wide miRNA screening of leukemia-associated loci identified miR-146a as a major mediator of the chromosome-5q deletion myelodysplastic syndrome [del(5q) MDS] and acute myeloid leukemia (AML).^6-8  The reduced miR-146a expression contributes to the development of del(5q) MDS and progression to AML through IRAK1- and TRAF6-dependent activation of NF-κB.^9,10  In nonmalignant myeloid cells (eg, monocytes), decreased miR-146a levels result in expression of IL-6 and other proinflammatory mediators implicated in the pathogenesis of autoimmune diseases and cancers.^11-13  The miR-146a dysregulation and IL-6 elevation in hematopoietic stem/progenitor and myeloid cells is also associated with many autoinflammatory diseases such as rheumatoid arthritis, systemic lupus erythematosus, type 2 diabetes, Sjögren’s syndrome, and endotoxemia-related cytokine storm.¹⁴ 

Although the role of miR-146a remains unclear, the NF-κB–mediated release of IL-6 from monocytes was also shown to be responsible for cytokine release syndrome (CRS), a serious adverse effect of CAR T-cell therapies.^12,15  Due to the lack of pharmacologic NF-κB inhibitors, synthetic miR-146a mimics are an attractive opportunity for immunomodulation or elimination of tumorigenic signaling. However, the effective delivery of miRNA therapeutic agents is challenging, complicated by safety concerns and potential off-target effects.^2,16  Several types of miRNA delivery vehicles, including liposomes, lipid nanoparticles, dendrimers, or hydrogels, had been tested previously.^17-19  Only a few of the synthetic miRNA mimics, including antifibrotic and transforming growth factor β–targeting miR-29/remlarsen, progressed to initial clinical testing.¹⁹  The current article describes an original approach for the targeted delivery of a chemically modified miR-146a mimic to myeloid cells and verifies miR-146a mimic activity in models of inflammatory and myeloproliferative diseases.

All animal experiments were conducted following institutional guidance and approved protocols from the Institutional Animal Care and Use Committee. C57BL/6, BALB/c mice 6 to 8 weeks of age were purchased from the National Cancer Institute; female C.B-Igh-1b/GbmsTac-Prkdc^scid-Lyst^bgN7 (SCID-Beige) mice were from Taconic, and NOD/SCID/IL-2RgKO (NSG) and NSG Tg(IL3,CSF2,KITLG)1Eav/MloySzJ (SGM3) mice were from The Jackson Laboratory. miR-146a^−/− and LyzM-Cre/miR-146a^fl/fl mice^9,20  were bred and housed in the Laboratory Animal Resources facility at the California Institute of Technology.

The C-miR146a conjugates were synthesized in the DNA/RNA Synthesis Core (City of Hope) by linking CpG-D19 to miR-146a passenger strands as previously described.²¹  These were hybridized with complementary guide strands of mature miR-146a, creating chimeric C-miR146a mimic. The single-stranded sequences are listed in the following text (x = C3-unit; asterisks = phosphorothioation; underline = 2′-O-methylation):

1. C-miR146a passenger:

2. 5′ G*G*TGCATCGATGCAGG*G*G*G*G-xxxxx-CCCAUGGAAUUCAGUUCUCAAA 3′.

3. miR-146a guide:

4. 5′ UGA​GAA​CUG​AAU​UCC​AUG​GGU​U 3′.

5. control scrambled RNA (C-scrRNA) passenger:

6. 5′ G*G*TGC​ATC​GAT​GCA​GG*G*G*G*G -xxxxx-AUUUAGCCUUAAUACACGCCAA 3′

7. scrRNA guide:

8. 5′ GGC​GUG​UAU​UAA​GGC​UAA​AUC​U 3′

For internalization studies, the miR-146a guide strand was labeled on the 3′ end by using Cy3 fluorochrome.

RNA-binding protein immunoprecipitation was performed by using the Magna RNA Immunoprecipitation Kit (MilliporeSigma) according to the manufacturer’s instructions. Details are described in the supplemental Methods (available on the Blood Web site).

miR-146a^−/− mice or syngeneic Cbfb/Myh11/Mpl mouse leukemia–bearing C57BL/6 mice were injected retro-orbitally with 2.5 to 20 mg/kg of C-miR146a, C-miR146a^Cy3, or miR146a^Cy3 as previously described.²²  At indicated times, mice were euthanized to collect organs for flow cytometry, quantitative polymerase chain reaction (qPCR), or western blot analysis. EasySep PE-positive selection kits (Stemcell Technologies) were used for cell subset enrichment.

A total of 10⁵ colony-forming units of Listeria monocytogenes (strain-10 403/serotype-1) were injected to each wild-type (WT) C57BL/6 or LyzM-Cre/miR-146a^fl/fl mouse using retro-orbital injections. Mice were treated every day using 5 mg/kg of C-miR146a or C-scrRNA (negative control) for 3 days before and 3 days after infection. The mice were euthanized, and various organs were collected for further analysis.

For in vitro studies, mock or CD19 CAR T cells from 4 donors were cultured with target Nalm6 leukemia with or without CD14⁺ monocytes from the same donor. Healthy donors’ peripheral blood mononuclear cells were collected in accordance with the Declaration of Helsinki. The cocultured cells were treated with C-miR146a or C-scrRNA (500 nM) for 2 days. After 48 hours, supernatants were collected, and IL-1 and IL-6 levels were analyzed by using an enzyme-linked immunosorbent assay (ELISA) (Thermo Fisher Scientific). In vivo CRS studies used female SCID-Beige mice injected using 3 × 10⁶ Raji^GFP/Luc cells, followed by intraperitoneal injections of C-miR146a (5 mg/kg) or phosphate-buffered saline (PBS) daily for 3 days before transfer of 12.5 × 10⁶ mock-transfected or CD19-specific CAR T cells, as previously described.^15,23,24 

NSG-SGM3 mice were injected intravenously with 2 × 10⁶ of HL-60^LUC cells, and tumor progression was monitored by using BLI/AmiX (Spectral Instruments). Mice were injected retro-orbitally with 10 mg/kg of C-miR146a or C-scrRNA every other day. Total RNA was isolated from bone marrow cells and analyzed by using RT² Profiler PCR Arrays (Qiagen).

The overall survival analysis was conducted by using R package survival (version 2.44.1.1). Log-rank statistics were applied to identify the optimal cut-point for transforming the continuous variable of gene expression into categorical high and low expression groups in a survfit model, using the cutp function of the R package survMisc (version 0.5.5).

An unpaired Student t test was used to calculate 2-tailed P values to estimate statistical significance between 2 treatment groups. One- or 2-way analysis of variance plus the Bonferroni posttest assessed differences between multiple groups or in tumor growth kinetics, respectively. Statistically significant P values were indicated in figures compared with untreated or PBS groups. Data were analyzed by using Prism software version 7 (GraphPad Software).

To overcome challenges in the efficient and cell-selective delivery of miR-146a, we modified the scavenger receptor (SR)/Toll-like receptor 9 (TLR9)–targeting platform originally developed for transfer of 25/27-mer Dicer-substrate small interfering RNA.²¹  The double-stranded miR-146a was conjugated through the 5′ end of the passenger strand to the 3′ end of a single-stranded, partly phosphorothioated oligodeoxynucleotide (CpG ODN) using a synthetic carbon linker. The specific type A CpG ODN sequence of the conjugate was selected for monocytes/myeloid cell specificity²⁵  and poor ability to activate NF-κB and its downstream IL-6 and IL-10 targets.²⁶  To ensure the maximum activity and target specificity, the miR-146a was chemically modified for nuclease resistance by a 2′-O-methyl-modification in the 3′ end of the passenger strand (Figure 1A). This design significantly improved serum stability of the C-miR146a conjugate without affecting activity. Compared with miR-146a (half-life, ∼1 hour), the conjugated C-miR146a had a 34-hour half-life in human serum (Figure 1B; supplemental Figure 1).

The C-miR146a^Cy3 was rapidly internalized by primary human immune cells and mouse RAW264.7 macrophages (Figure 1C) or splenocytes (supplemental Figure 2) within 1 hour of incubation, without any transfection reagents. Compared with the unconjugated miR146a^Cy3, C-miR146a^Cy3 was also efficiently taken up by human MDSL or HL60 myeloid leukemia cells and Raji Burkitt’s lymphoma cells but not human T-cell leukemia TAIL7 cells. The uptake of C-miR146a depended on the scavenger receptor A– and clathrin-mediated endocytosis, as verified by using internalization inhibitors (dextran sulfate, fucoidan, and cadaverine) (supplemental Figure 3). TLR9 is commonly expressed by myeloid and B-cell malignancies, including tested myeloid leukemia and B-cell lymphoma cells (supplemental Figure 4).²⁶  Confocal microscopy showed that C-miR146a localized to the cytosol of target cells within 1 hour, whereas miR146a alone was undetectable (Figure 1D). To inhibit target mRNAs, the miRNA guide strand must bind Argonaute-2 (Ago2) protein within the RNA-induced silencing complex (RISC). To assess whether the guide strand of C-miR146a was successfully loaded onto the RISC, we performed RNA-binding protein immunoprecipitation using Ago2-specific antibodies (Figure 1E). In fact, the level of Ago2-bound miR-146a guide strand increased several fold in mouse RAW264.7 macrophages incubated with the C-miR146a but not with miR146a. This effect was even more prominent in primary mouse splenocytes derived from miR-146a^−/− mice. These results indicate that the C-miR146a allows for the delivery of functional miR-146a to human and mouse myeloid target cells.

miR-146a targets IRAK1 and TRAF6, the critical upstream regulators of NF-κB signaling.¹⁰  We used 3′-UTR luciferase reporter assays to verify whether C-miR146a can alter IRAK1 and TRAF6 expression. As shown in Figure 2A, C-miR146a conjugate reduced IRAK1 and TRAF6 3′-UTR luciferase reporter activities in a sequence-specific manner, verified by mutating 3′-UTR sequences. Neither miR-146a alone nor negative C-scrRNA affected these 3′-UTR luciferase reporters. Moreover, C-miR146a, but not the control oligonucleotides, reduced IRAK1 and TRAF6 protein levels in mouse RAW264.7 macrophages but only IRAK1 in human MDSL cells (Figure 2B). Lack of TRAF6 inhibition by C-miR146a is likely related to a mutation in one of the miR-146a–binding sites in TRAF6 3′UTR in human MDSL cells (supplemental Figure 5). Activated NF-κB translocates to the nucleus; we therefore examined the effect of C-miR146a on the localization of NF-κB in lipopolysaccharide (LPS)-treated RAW264.7 cells expressing NF-κB/p65-eGFP fusion protein.²⁷  Confocal microscopy confirmed that C-miR146a, but not C-scrRNA, efficiently blocked the nuclear translocation of NF-κB (Figure 2C). Correspondingly, C-miR146a reduced the NF-κB DNA-binding activity in nuclear extracts from RAW264.7 and MDSL cells as measured by using gel shift assays (Figure 2D). These C-miR146a effects reduced LPS-induced transcriptional NF-κB activity in RAW-Blue macrophages (Figure 2E) and secretion of IL-6 by parental RAW264.7 cells (Figure 2F).

For an additional verification that the observed effects were sequence-specific, we used an analogical approach to deliver the miR-146a–targeting antisense oligonucleotide, C-anti-miR146a (supplemental Figure 6A).²⁸  C-anti-miR146a exhibited similar cell selectivity as the C-miR146a mimic (supplemental Figure 6B) but opposite biological activity. C-anti-miR146a knocked-down miR-146a in mouse splenocytes and myeloid cells (supplemental Figure 6C), thereby upregulating IRAK1 and TRAF6 3′UTR activities (supplemental Figure 6D-E), NF-κB DNA binding (Figure 2D), and nuclear translocation (supplemental Figure 6F). Taken together, the flexibility of our strategy permits positive or negative regulation of NF-κB activity by using SR/TLR9-mediated miR-146a mimic or antisense delivery, respectively.

Treatment of NF-κB–related disorders requires efficient delivery of miR-146a mimic to target myeloid cells in various organs. To directly quantify the dose- and time-dependent biodistribution of C-miR146a mimic in the absence of endogenous miR-146a, we used the real-time qPCR to detect miR-146a guide-strand in various organs of miR-146a–deficient mice.⁹  Our initial pharmacokinetic analysis found that C-miR146a reaches the maximum plasma level ∼30 minutes after IV injection, with 144 minutes of circulatory half-life (supplemental Figure 7). Consistently, we found dose-dependent accumulation of miR146a guide-strand in bone marrow, spleens, and lymph nodes of miR-146a^−/− mice after IV injection of C-miR146a (Figure 3A). In the 2.5 to 20 mg/kg dose range, C-miR146a delivered miR146a-5p guide-strand preferentially to CD11b⁺ myeloid cells from bone marrow (Figure 3B). For more precise identification of target cells, we injected IV fluorescently labeled C-miR146a^Cy3 or miR146a^Cy3 (5 mg/kg) into AML-bearing mice and analyzed their cellular uptake after 3 hours by using flow cytometry. C-miR146a^Cy3 was internalized by ∼30% to 50% of various myeloid cells and leukemic cells, ∼20% of long-term hematopoietic stem cells, and minimally by B cells and natural killer cells but not by T cells (Figure 3C). As expected, cellular uptake of miR146a^Cy3 alone was negligible.

We next compared the miR-146a levels in oligonucleotide-treated miR-146a^−/− mice vs those in the untreated WT mice to quantify the relative miR-146a restoration in specific immune cell populations. As shown in Figure 3D, the single injection of 5 mg/kg C-miR146a to miR-146a^−/− mice replaced ∼61% and ∼53% of WT miR-146a levels in bone marrow and spleens, respectively. miR146a was restored to ∼47%, ∼28%, and ∼5% of the WT levels in splenic CD11b⁺ myeloid cells, CD19⁺ B cells, and CD3⁺ T cells, respectively (Figure 3E), confirming preferential myeloid cell targeting. Finally, we tested whether partial restoration of miR-146a will have sufficient biological activity to inhibit target IRAK1 and TRAF6 proteins. In fact, within 12 to 24 hours after a single IV injection of C-miR146a (5 mg/kg), both IRAK1 and TRAF6 were reduced close to the WT levels, returning to baseline after 2 to 3 days (Figure 3F). These results indicate that systemic delivery of C-miR146a using relatively low dosing can transiently restore levels and full biological activity of miR-146a in target myeloid cells in vivo.

Expansion of myeloid cells is one of the critical phenotypic features of the miR-146a–deficient mice.²⁹  Compared with bone marrow–derived macrophages (BMDMs) from WT mice, BMDMs from miR-146a–deficient mice exhibit enhanced proliferation accompanied by elevated expression of colony-stimulating factor–1 receptor (CSF1R).³⁰  We therefore examined whether delivery of C-miR146a to miR-146a^−/− BMDMs prevents aberrant myeloproliferation. In fact, the C-miR146a but not C-scrRNA treatment reduced the proliferative rate of miR-146a^−/− BMDMs close to the WT BMDMs (Figure 4A). Together with the reduced proliferation, the elevated cell surface levels of CSF1R on miR-146a^−/− BMDMs were reduced by C-miR146a to baseline WT levels (Figure 4B). Consistently, repeated 2-week C-miR146a treatment of aged miR-146a^−/− mice reduced the expansion of immature CD11b⁺Gr1⁺ myeloid cells, together with their proliferation and CSF1R expression on macrophages (Figure 4C-E).

We next examined whether miR-146a mimic delivery can attenuate the exaggerated inflammatory response to a systemic bacterial endotoxin challenge resulting from the miR-146a ablation.³⁰  Age- and sex-matched WT and miR-146a^−/− mice were injected with a sublethal dose of LPS, and the plasma levels of proinflammatory cytokines IL-6 and tumor necrosis factor-α (TNF-α) were measured by using ELISA. As expected, LPS induced a higher and more prolonged cytokine response in miR-146a^−/− mice than in WT mice, peaking at 2 to 6 hours before returning to baseline at 24 hours. In contrast, C-miR146a treatment reduced IL-6 and TNF-α upregulation in miR-146a^−/− mice at both 2 and 6 hours compared with C-scrRNA control treatment (Figure 4F). The effect of C-miR146a on the inflammatory cytokine secretion in WT mice was minimal, indicating adequate negative control of cytokine production by the endogenous miR-146a.

Following on these results, we used mice with the myeloid cell–specific miR-146a ablation (LyzM-Cre/miR-146a^fl/fl) to assess the effect of miR-146a restoration on the IL-6–controlled response to a bacterial infection with L monocytogenes.^20,31,32  Both WT and miR-146a^fl/fl mice were injected daily with C-miR146a or control C-scrRNA (5 mg/kg) starting 3 days before the sublethal infection with L monocytogenes. The C-miR146a treatment interfered with the bacterial clearance in miR-146a–deficient and WT mice, increasing bacterial burden in the liver and spleen (Figure 4G; supplemental Figure 8A). The reduced bacterial load in control-treated miR-146a^fl/fl mice correlated with weight loss (Figure 4H) and elevated plasma levels of IL-6 (Figure 4I). All these effects of miR-146a deletion were completely alleviated by C-miR146a treatment. Finally, miR-146a restoration decreased the expansion of CD11b⁺ myeloid cells, including Gr1⁺ granulocytes, in blood and spleen of control miR-146a^fl/fl mice (Figure 4J; supplemental Figure 8B). Taken together, our results indicate that myeloid cell–targeted miR-146a restoration using C-miRNA mimic can reverse the key phenotypic features of miR-146a deletion in mice. It is worth noting that we observed no significant effect of C-miRNA mimic on other immune cell populations in WT mice, such as dendritic cells, B cells, or natural killer cells, even after repeated 2-week daily IV injections (supplemental Figures 9-10). Finally, our C-miR146a two-week dose escalation study (5-20 mg/kg per day) found no evidence of toxicities to hematopoietic cells or organ gross abnormalities (supplemental Figure 11).

CAR T-cell immunotherapy has emerged as a revolutionary treatment of various hematologic cancers; however, many patients experience severe side effects of CRS.^23,33  Recent studies linked CRS to CAR T-cell–dependent activation of CD40 signaling in monocytes/macrophages, thereby inducing IL-1 and IL-6, the key cytokine storm mediators. Although CD40 triggers canonical and noncanonical NF-κB signaling, it does not upregulate miR-146a levels.²⁹  Thus, we tested whether delivery of miR-146a mimic will dampen the proinflammatory effects of human monocytes induced by the CD19-specific CAR T cells. Predictably, CD19 CAR T cells exhibited low uptake of fluorescently labeled C-miR146a^Cy3 compared with monocytes (supplemental Figure 12A). C-miR146a did not affect CAR T-cell viability or cytotoxicity against target CD19⁺ Nalm6 leukemia cells (Figure 5A). We next tested C-miR146a in a 3-component CRS model coculturing in vitro CD19⁺ Nalm6 cells with CD19-specific CAR T cells (generated from peripheral blood mononuclear cells from 4 different donors) and donor-matched CD14⁺ monocytes (Figure 5B). The upregulation of IL-1 and IL-6 depended on the presence of monocytes and was induced only by CD19⁺ leukemia-specific and not mock-transfected CAR T cells. Importantly, C-miR146a reduced IL-1 secretion by one-half, while bringing IL-6 levels close to baseline compared with control treatments. Similar results were obtained by using THP-1 monocyte-like cells instead of primary monocytes (supplemental Figure 12B-C).

To further confirm that the combination of C-miR146a with CD19 CAR T cells has potential for alleviating CRS-related toxicities, we adopted a model of CAR T-cell–induced CRS in human CD19⁺ Raji B-cell lymphoma in partly immunodeficient SCID-Beige mice, similar to that recently published.¹⁵  Mice with heavy intraperitoneal lymphoma burden developed CRS and acute inflammation within 3 days after CD19 CAR T-cell injection (Figure 5C). Mice were treated by using C-miR146a or control vehicle 3 days before CAR T-cell transfer. As shown in Figure 5D-E, the C-miR146a did not interfere with the CAR T-cell–mediated inhibition of lymphoma progression. However, the C-miR146a mimic doubled the amount of endogenous miR146a in target peritoneal myeloid cells (Figure 5F). As a result, C-miR146a dramatically reduced major CRS-related cytokines, IL-6 and granulocyte colony-stimulating factor (G-CSF), likely derived from mouse monocytes in this model as suggested by others (Figure 5G). Our findings underscore the potential of using C-miR146a mimic to alleviate adverse effects of CD19 CAR T-cell therapy without impeding on-target therapeutic effects.

Low expression of miR-146a located on chromosome 5q has been reported in both MDS and AML as del(5q) aberrations are common in high-risk leukemia.⁷  However, constitutive activation of NF-κB survival signaling is not limited to del(5q) MDS/AML and occurs in up to 40% of AML cases.³⁴  We tested C-miR146a mimic effect on viability of HL-60, MDSL del(5q) leukemia cells, and also FLT3ITD⁺/N(5q) MV4-11 AML cells, which have all been found to be sensitive to small-molecule NF-κB inhibitors (supplemental Figure 13). As shown in Figure 6A, C-miR146a induced cell death in all 3 models, although it was most effective against miR-146a–deficient MDSL and HL-60 cells. In initial testing, we confirmed that local delivery of C-miR146a into subcutaneously engrafted leukemia models effectively inhibited growth of both HL-60 (supplemental Figure 14A) and MDSL (supplemental Figure 14B). We next examined the efficacy of systemic C-miR146a administration (10 mg/kg) against disseminated HL-60 leukemia. As shown in Figure 6B-D, compared with control mice, C-miR146a reduced AML progression and thereby significantly prolonged mice survival. Focused gene expression analysis of a panel of 84 NF-κB–related genes confirmed that C-miR146a downregulated key elements of NF-κB signaling in HL-60 cells, such as RELA, MYD88, NFKB1/2, TRAF3, and TRAF6, or mediators IL1B and TNF (Figure 6E; supplemental Figure 15). Altogether, these results illustrate the direct antitumor effect of C-miR146a, both in vitro and in vivo, against miR-146a–deficient MDS/AML, likely through the inhibition of NF-κB–mediated survival signaling. These conclusions are also supported by analysis of The Cancer Genome Atlas dataset from 164 patients with AML. As expected, low miR-146a expression was associated with worse overall patient survival (Figure 6F). The gene set enrichment analysis indicated significant inverse correlation of miR-146a levels in AML with 2 different sets of NF-κB–related genes, including IRAK1, NFKB1, and NFKB2 but not TRAF6 (Figure 6G-H; supplemental Figure 16). Overall, these results underscore the therapeutic potential of C-miR146a for the treatment of del(5q) MDS/AML and potentially other NF-κB–dependent types of leukemia.

Results of the current study show the feasibility of systemic delivery of miRNA mimic specifically to human and mouse myeloid cells for therapeutic modulation of their immune activity or neoplastic growth. The C-miR146a injected intravenously restored miR-146a-5p in target myeloid cells to levels sufficient for complete elimination of exacerbated NF-κB activity in miR-146a^−/− mice, thereby preventing the exaggerated inflammatory responses and aberrant myeloproliferation. Our miR-146a mimic delivery strategy proved effective also in human models of NF-κB–driven inflammation and in myeloid leukemia (MDS/AML). Importantly, unlike standard anti-inflammatory strategies (eg, steroid hormones), the myeloid cell–specific miR-146a mimic delivery proved effective without interfering with antitumor activity of CAR T cells or any signs of toxicity even with repeated, long-term administration.

Despite broad clinical potential, only a few miRNA therapeutic agents entered clinical trials, and the majority represented antisense molecules (anti-miRs or antagomirs), which as single-stranded oligonucleotides proved easier to optimize forin vivo use.^19,35  One of the best examples is targeting lymphoma cell addiction to oncogenic miR-155.³⁶  In contrast, the therapeutic replacement/restoration of tumor suppressor miRNAs proved challenging. The chemical modification of miRNA to ensure nuclease resistance can interfere with intracellular processing, RISC loading, and/or mRNA targeting. Pharmacologic formulation of miRNA mimics for systemic administration can result in severe immune toxicities.³⁷  Cell-selective delivery of chemically stabilized, naked miRNA mimics is gaining attention with the success of small interfering RNA delivery using hepatocyte-specific GalNAc-conjugates,³⁸  and a recent study revisited folate as targeting moiety to miRNA delivery to breast and lung cancer cells.³⁹ 

Our previous studies focused on myeloid cell–selective delivery of oligonucleotides to disrupt immunosuppression in the tumor microenvironment. The conjugation of CpG ODNs with inhibitors of STAT3, a master immune checkpoint regulator, resulted in potent antitumor immune responses. Although it seems counterintuitive, the SR/TLR9-mediated delivery strategy can be adapted for dampening excessive immune activity of myeloid cells in autoinflammatory diseases or myeloid malignancies. Unlike type B CpG sequences, the type A CpG ODNs are poor activators of NF-κB signaling and cytokines such as IL-6 and IL-10.⁴⁰  Although a recent study showed that injections of CpG(A) alleviated severity of sepsis in mice by reducing blood clotting,⁴¹  the control CpG(A)-scrRNA did not exhibit anti-inflammatory activity in our experiments. The presence of a functional miR-146a-5p guide strand in the CpG conjugate was key to the biological activity of this oligonucleotide. The more extensive chemical modifications of miR-146a guide strand affect its activity, likely interfering with unwinding of the duplex while in RISC.⁴²  Importantly, even partial restoration of miR-146a-5p levels in target immune cells in vivo was sufficient for near complete and durable inhibition of target IRAK1 and TRAF6. The systemic administration of C-miR146a effectively reversed key features of the miR-146a deletion related to excessive NF-κB signaling in both mouse and human myeloid cells.

Cytokine storm or CRS occur in response to various conditions, from bacterial infections to antibody-based therapies to immunomodulatory drugs. In context of CAR T-cell cancer therapies, especially CD19-specific CAR T cells, CRS is one of the most common and potentially fatal adverse effects. The lower incidence of CRS correlates with prolonged survival of patients receiving CAR T-cell therapy.⁴³  Despite the routine use of an IL-6 receptor antagonist (tocilizumab), high-dose steroid hormones are frequently required to control severe CRS but at the same time they potentially curb the antitumor efficacy of CAR T cells.^43,44  Thus, the need exists for more precise and safer immunomodulatory strategies addressing the complexity of immune responses. Myeloid cell–selective miR-146a mimic acts at the nodal point of the immune cell network, preventing the release of cytokines driving CRS from monocytes and macrophages without interfering with CAR T-cell activity.^12,15 

The simplicity and flexibility of the C-miRNA mimic design, adaptable to delivery of both miRNA mimics and anti-miRs, provide an opportunity for the development of miRNA therapeutic agents for immunomodulation and/or therapy of myeloproliferative diseases and leukemia. The pleiotropic effect of a single miRNA on multiple protein targets benefits their potency and reduces development of drug resistance, whereas the myeloid cell–targeted delivery can improve the safety of miRNA therapeutic agents. The growing list of myeloid cell–specific anti-inflammatory miRNAs extends beyond miR-146a, including recently described miR-125b and miR-203b or tolerance-inducing miR-221 and miR-222.^45,46  In addition to the immunoregulatory role, the miR-146a is a well-established tumor suppressor, and the genetic loss of miR-146a or downregulation is common in patients with MDS and AML.⁶  miR-146a replacement using SR/TLR9–targeted delivery provides therapeutic opportunity in patients with MDS/AML. However, given the genetic instability of AML, C-miR146a therapy could likely benefit from combination with targeting additional oncogenic or tumor suppressor miRNAs. The upregulation of oncogenic miR-155 is common in FMS-like tyrosine kinase 3 internal tandem duplication–associated AML.⁴⁷  Noteworthy, miR-155 can potentially antagonize the inhibitory effects of miR-146a on NF-κB in mouse myeloid cells.²⁰  It is not known whether miR-155 plays a similar role, lessening the tumor suppressive effect of miR-146a in human leukemia, but our strategy can easily be adapted to delivery of anti-miR155 to AML cells and leukemia stem cells.⁴⁸  Further studies should explore the possibility of personalizing miRNA mimic/antisense combinations to an AML patient–specific miRNA profile using the same SR/TLR9 delivery platform. The emerging CAR T-cell strategies for AML, such as CD123-specific CAR design, underscore the potential to combine AML-specific CAR T cells with C-miR146a, thereby alleviating potential immunotoxicities as well as reducing AML cell survival.⁴⁹  We believe that SR/TLR9–targeted delivery of miR-146a mimic provides an outline for the development of miRNA therapeutic agents for a variety of inflammatory disorders and myeloid malignancies.

The authors thank and acknowledge the dedication of staff members at the DNA/RNA Synthesis, Analytical Cytometry, Light Microscopy Cores and Animal Resources Center (City of Hope).

This work was supported in part by the National Cancer Institute/National Institutes of Health awards R01CA213131 (M.K.), Lymphoma Specialized Program of Research Excellence P50CA107399 (S.J.F.), and P30CA033572 (to the City of Hope). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Contribution: Y.-L.S. and M.K. were responsible for study design; Y.-L.S., X.W., M.M., T.P.A., D.F.M., Z.Z., D.W., P.M.S., and M.K. performed data acquisition; Y.-L.S., M.M., T.P.A., D.F.M., D.W., and C.O. analyzed and interpreted the data; Y.-L.S., X.W., X.H., B.Z., P.M.S., S.J.F., D.B., L.L., G.M., M.P.B., and M.K. were responsible for administrative, technical, or material support; and Y.-L.S. and M.K. wrote the manuscript.

Conflict-of-interest disclosure: M.K., P.M.S., and G.M. have a patent on C-miRNA conjugates and uses thereof compositions for the treatment of cancers and other diseases. The remaining authors declare no competing financial interests.

Correspondence: Marcin Kortylewski, Beckman Research Institute at City of Hope Comprehensive Cancer Center, 1500 East Duarte Rd, Beckman Center, Room 3111, Duarte, CA 91010; e-mail: mkortylewski@coh.org.