Long-range function of secreted small nucleolar RNAs that direct 2′-O-methylation

Rpl13a snoRNAs accumulate in the culture medium of lipopolysaccharide (LPS)-stimulated macrophages

The Rpl13a snoRNAs U32a, U33, U34, and U35a accumulate outside the nucleus of fibroblasts and cardiomyoblasts in the setting of oxidative stress (9, 10, 22). To extend these findings to another cell type in which oxidative stress plays a central role in pathophysiologically important responses, we treated bone marrow–derived macrophages with LPS, a stimulus known to induce reactive oxygen species and inflammatory gene expression (23). Following exposure of macrophages to LPS, intracellular reactive oxygen species are increased, as evidenced by enhanced 2′,7′-dichlorodihydrofluorescein diacetate (DCF) staining (Fig. 1A) and expression of the well-characterized inflammatory response genes, cyclooxygenase 2 and tumor necrosis factor-α (COX2 and TNFα; Fig. 1B). Stimulation with LPS did not cause loss of cellular integrity, as assessed by LDH release (Fig. 1C), or apoptosis, as assessed by caspase 3/7 activity (Fig. 1D). Using snoRNA-specific stem loop primers for reverse transcription with qPCR amplification for detection, we found that the Rpl13a snoRNAs rapidly increased in the medium following 1 h of stimulation and were cleared over the subsequent 3 h (Fig. 1E). Increases in snoRNAs in the medium were not accompanied by increase in cytosolic or nuclear snoRNA abundance (Fig. 1, F and G). Absolute quantification of snoRNAs using standard curves for each species demonstrated that LPS-induced increases in snoRNAs in the medium are significant and specific, but small relative to nuclear and total cellular snoRNA abundance, pools that do not significantly change with treatment (Fig. 2, A–D). Control PCRs indicated that the target-specific stem-loop primed reverse transcription qPCR method detects mature snoRNAs, but not snoRNA sequences in DNA or pre-mRNA templates (Fig. 2E). These findings indicate that LPS induces macrophage snoRNA secretion. Rapid clearance of the snoRNAs from the medium suggests they may be degraded, and decreases in cytosolic snoRNAs at 4 h (significant for U33 and U34 only) may reflect the cumulative export over several hours. In addition to the Rpl13a snoRNAs, LPS also stimulates the secretion of SNORD82, SNORD92, and SNORA73b (Fig. 3), suggesting that LPS broadly stimulates secretion of box C/D and box H/ACA snoRNAs.

Rpl13a snoRNAs are released with extracellular vesicles

A growing body of literature provides evidence that RNAs are released from cells in membrane-bound structures, generally referred to as extracellular vesicles (EVs) (24). Release of Rpl13a snoRNAs from LPS-stimulated macrophages was unaffected by brefeldin A, suggesting that release of these RNAs did not require classical secretory pathways (Fig. 4, A and B). Appearance of Rpl13a snoRNAs in the medium was also stimulated by 2-hydroxypropyl-β-cyclodextrin and ionomycin, agents known to induce EV secretion though mechanisms that are distinct from LPS treatment (25, 26) and that do not lead to ROS generation (Fig. 4, C and D). Together, these observations suggested that release of Rpl13a snoRNAs into the medium might occur through EV secretion.

To determine whether snoRNAs in the medium are associated with EVs, we concentrated EVs from macrophage-conditioned medium following LPS treatment using either a commercial precipitation reagent that enriches for exosomes or differential centrifugation. Ultrastructural analysis of pellets isolated with both methods revealed membrane-bound structures with diameters of 70–200 nm (Fig. 5A), whereas these structures were absent from nonconditioned medium (not shown). In these quick-freeze deep-etch electron micrographs, either the true (outer) membrane surface or the smooth surface that results from fracturing through the two leaflets of the bilayer (p face) is visible at the perimeter of the vesicle. These data indicate that bone marrow–derived macrophages treated with LPS release EVs. Moreover, in EV-enriched fractions prepared by either method, Rpl13a snoRNAs were detected at baseline and increased following treatment of cells with LPS or ionomycin, indicating that macrophages constitutively release EV-associated snoRNAs but tightly regulate partitioning of snoRNAs for EV-mediated export in response to these stimuli (Fig. 5B). Recovery of EV-associated snoRNAs was more efficient following precipitation than ultracentrifugation, consistent with previous reports of higher yields from precipitation (27). EV preparations were enriched for the integral membrane proteins, flotillin and CD9, and relatively depleted for endoplasmic reticulum (calnexin) and nuclear (histone H3) markers (Fig. 5C). The absence of correlation between magnitude of stimulated EV-associated snoRNA secretion and EV-associated protein secretion between the two isolation methods and the lack of increase in EV-associated proteins between unstimulated and stimulated samples are observations that suggest snoRNA-containing EVs released upon LPS or ionomycin treatment represent a small fraction of the total secreted EVs.

To distinguish between co-fractionation and true association of Rpl13a snoRNAs with EVs, we separated EVs isolated by ultracentrifugation from potentially nonvesicular co-sedimenting components by floating the EVs into a continuous sucrose gradient. Both Rpl13a snoRNAs and EV membrane protein markers were enriched at a density between 1.14 and 1.20 g/cm², the characteristic density for EVs (Fig. 6A). Rpl13a snoRNAs in these EVs were resistant to RNase treatment but degraded when treated with RNase in the presence of detergent (Fig. 6B), suggesting that the snoRNAs are protected by a membrane. Together, our findings support a model in which macrophages release EVs containing snoRNAs.

EV-associated Rpl13a snoRNAs are internalized by recipient cells

Secreted RNAs can impact the biology of target cells by engaging cell surface receptors or by uptake through endocytosis or macropinocytosis. To determine whether RNA from macrophage-derived EVs is incorporated into recipient cells, EVs were enriched from the medium of donor cells that had been metabolically labeled with ethynyl uridine and incubated for 45 min with unlabeled recipient cells. Following washing to remove surface-bound material, recipient cells were probed with an Alexa Fluor 594–linked click reagent. RNA from donor cell EVs was detected in the nucleus, where it co-localized with fibrillarin (Fig. 7A). The dye-only control indicates that any small amount of free dye that might remain after labeling and cleanup does not cause detectable nucleolar staining. In an alternate approach, RNA and lipids of isolated EVs were stained with Syto RNAselect and 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine perchlorate (DiD), respectively, and incubated with recipient cells. In the recipient cells, labeled RNA was distributed throughout the cytoplasm and nucleus and concentrated in nuclear structures suggestive of nucleoli (arrowheads), a pattern that was distinct from background autofluorescence in the dye-only control (Fig. 7B). DiD was also transferred to recipient cells and was distributed in cytosolic punctae. Together, these findings indicate that both RNA and lipid constituents of EVs are transferred.

Approaches in which RNA is followed by a metabolic label or dye cannot formally distinguish between uptake of exogenous EV components and transfer of free label or dye that has partitioned into the EVs. To specifically assess whether secreted snoRNAs are taken up into recipient cells, we isolated cells from WT mice and from mice with germ line knockout of all four Rpl13a snoRNAs (KO) (28). KO mice are viable and fertile with no overt pathophysiological phenotype under standard housing and feeding conditions. However, pancreatic islets from these mice have decreased ROS tone, resistance to oxidative stress stimuli, and enhanced insulin secretion and mitochondrial proton leak.

WT and KO macrophages were co-cultured in a transwell apparatus that maintains the cells in separate compartments separated by a membrane with 1-μm pores. In this setting, medium and associated EVs were freely shared between the cells of different genotypes, but PCR analysis of DNA confirmed that cellular material was not transferred (Fig. 8A). In situ hybridization to detect snoRNA U33 revealed scattered nuclear puncta and larger areas of nuclear accumulation consistent with nucleoli in singly cultured WT cells (Fig. 8, B and C). Compared with KO macrophages that had been co-cultured with KO macrophages, KO macrophages co-cultured with WT macrophages demonstrated significantly increased nuclear punctae, consistent with transfer of U33. Whereas the signal for this single snoRNA species is small compared with the signal for total labeled RNA (Fig. 7), the co-culture approach (Fig. 8) provides further evidence of snoRNA transfer under more physiological conditions.

LPS increases Rpl13a snoRNAs in the serum of murine model and human volunteers

Because LPS triggers EV release of snoRNAs from cultured macrophages, we asked whether this inflammatory stimulus precipitated secretion of snoRNAs in vivo by treating mice with LPS. Compared with PBS-treated mice, animals that received LPS demonstrated a rapid increase in EV-associated Rpl13a snoRNAs in the serum over 4 h, concomitant with an increase in hepatic inflammatory gene expression (Fig. 9, A and B). We extended these findings by quantifying serum snoRNAs in healthy human volunteers undergoing intrabronchial instillation of LPS for a study of lung inflammation (29). Although the dose of LPS was small and delivered locally in the lungs, an increase in C-reactive protein, white blood cell count, and percentage of polymorphonuclear leukocytes 6 h following the procedure indicated the presence of systemic inflammation (Fig. 9C). Compared with pretreatment conditions, levels of U32a, U33, and U35a in serum EVs were significantly increased, and levels of U34 trended up (Fig. 9D). Thus, in mice and in humans, as we previously observed in cultured macrophages, LPS induces secretion of snoRNAs, most likely in EVs.

Secreted snoRNAs direct new 2′-O-methylation in vivo

To determine whether secreted snoRNAs can function in distant tissues in vivo, we employed a parabiosis model in which WT mice expressing GFP in all tissues were surgically joined to Rpl13a snoRNA KO mice (Fig. 10A). PCR analysis of DNA from tail blood of the parabionts revealed the presence of both the WT and KO alleles in the circulation of each parabiont by 1 week following surgery (Fig. 10B), consistent with reports of rapid development of shared circulation in this model (30). To test for the function of the Rpl13a snoRNAs, we utilized a qPCR-based method for detection of 2′-O-methylation of RNA, the modification that these snoRNAs direct on their canonical rRNA targets (31, 32). In this assay, the covalent modifications of the template cause reverse transcriptase to stall at low deoxyribonucleotide concentrations, such that larger amounts of qPCR product indicate less modification (Fig. 11A). This approach revealed the expected increases in amplification across Rpl13a snoRNA–directed 2′-O-methylation sites on the 18S and 28S rRNAs in KO compared with WT macrophages (Fig. 11B).

We applied this method for analysis of Rpl13a snoRNA–directed rRNA modifications to tissues harvested 3 weeks following surgery from KO and WT parabionts. Because new rRNA transcription correlates with cell replication (33) and 2′-O-methylation occurs on nascent rRNAs (34), we chose for study the intestinal epithelium, where cells turn over every 2–3 days (35). Liver tissue served as a negative control, because hepatocytes divide approximately once in 150 days in the adult mouse (36), and therefore new rRNA synthesis and 2′-O-methylation are expected to occur at much slower rates. Compared with KO mice that had not undergone parabiosis, enterocytes of KO parabionts showed a significant decrease in amplification (relative quantification; RQ), indicating an increase in modification of all four Rpl13a snoRNA target sites (lower RQ) and phenocopying the WT parabiont and WT animal control (Fig. 12A). By contrast, in liver tissue, we observed no significant difference in modification (RQ) at target sites for U32a and U33 between KO controls and KO parabionts, despite a clear difference between KO controls and either WT controls or WT paraboints (Fig. 12B). Unexpectedly, there was no difference between WT and KO controls in apparent methylation in liver tissue at the U34 or U35a snoRNA target sites, raising the possibility that these target sites are unmodified in WT livers or that other snoRNAs could be up-regulated specifically in the KO livers to compensate for the loss of the Rpl13a snoRNAs. Importantly, GFP expression remained undetectable in the KO parabiont (Fig. 12, C and D), indicating the lack of transfer of cellular material. Taken together, our data provide evidence that Rpl13a snoRNAs travel through the circulation and are taken up by distant tissues in which they function to modify their established RNA targets.

Reagents

LPS, BFA, IO, CD, and staurosporine were from Sigma-Aldrich. α-Fibrillarin (ab4566), α-nucleophosmin (ab15440), α-flotillin 1 (ab50673), and α-calnexin (ab22595) antibodies were from Abcam (Cambridge, MA). α-Histone H3 (catalog no. 9717) antibody was from Cell Signaling (Danvers, MA). α-CD9 (catalog no. 564236) was from BD Biosciences.

Mice

FVB and FVB.Cg-Tg(CAG-EGFP)B5Nagy/J were obtained from the Jackson Laboratory (Bar Harbor, ME), and Rpl13a-snoless (KO) mice (28) were maintained in facilities under specific pathogen-free conditions at Washington University (St. Louis, MO). Bone marrow was isolated as described below. Blood samples were obtained from male and female mice 4 h following intraperitoneal injection of 10 mg/kg LPS. All experimental procedures were approved by the Washington University Animal Studies Committee and were conducted in accordance with the United States Department of Agriculture Animal Welfare Act and the Public Health Service Policy for the Humane Care and Use of Laboratory Animals.

Bone marrow–derived macrophages

Bone marrow was harvested from the tibia and femur of male and female mice. Cells were seeded and differentiated for 6 days in 10% CMG14.12 conditioned medium (42). CMG14.12 cells were a gift from the laboratory of Deborah Veis (Washington University). Cells were treated with 50 ng/ml LPS for the indicated times, 10 μg/ml BFA for 90 min, 0.5 μm IO for 15 min, or 2% CD for 60 min. Cell death was measured using the Cytotox96 nonradioactive cytotoxicity assay (Promega, Madison, WI). LDH⁺ controls contained LDH from the kit (1:5,000 dilution of 1,600 units/ml solution), and the maximum signal was determined by measuring LDH activity in the medium of cells cultured under identical conditions and lysed by quickly freezing and thawing the plate. Apoptosis was quantified by a caspase 3/7 Glo assay (Promega, Madison, WI). ROS was detected by measuring fluorescence on a TECAN Infinite 200 PRO multimode reader (Tecan, Männedorf, Switzerland) after cells were loaded with 10 μm DCF (Thermo Fisher Scientific, Carlsbad, CA). Sequential detergent extraction was performed as described previously to generate nuclear and cytosolic fractions (10).

EV enrichment, labeling, and transfer

EVs were isolated from cell culture medium using Total Exosome isolation reagent (Thermo Fisher Scientific) (43), or EVs were isolated by differential centrifugation (44). Serum EVs were isolated with Exoquick (System Biosciences, Mountain View, CA) according to the manufacturer's instructions (38). Culture medium used for EV labeling/transfer and floatation/RNase treatment experiments was depleted of EVs before use by centrifugation at 120,000 × g for 90 min. Conditioned medium was cleared by centrifugation for 10 min at 1,000 × g, followed by centrifugation for 20 min at 2,000 × g, and then concentrated using an Amicon Ultra-15 centrifugal filter unit with Ultracel-10 membrane (3,500 × g, 20 min; Millipore, Billerica, MA). EVs were floated into a continuous sucrose gradient as described previously (44). A small sample was taken from each fraction, and density was measured at room temperature using an Atago (Bellevue, WA) three-range sugar hand refractometer (0–90%).

For EV transfer experiments, EVs were labeled using Syto RNA select and DiD (Thermo Fisher Scientific) according to the manufacturer's instructions, and unincorporated dye was removed either by washing in Amicon filters or with Exosome spin columns (Thermo Fisher Scientific) with similar results. Dye controls were treated in an identical manner. Ethynyl uridine–labeled EVs were collected from cells cultured overnight with 0.5 mm ethynyl uridine (45). Following incubation of recipient cells with EVs for 45 min, cells were washed twice with 3% BSA (Sigma-Aldrich) in PBS, fixed with 3.7% paraformaldehyde, (Electron Microscopy Science, Hatfield, PA), and permeabilized with 0.5% Triton X-100 (Sigma-Aldrich). Ethynyl uridine was detected using the Click iT RNA imaging kit (Alexa Fluor 594, Thermo Fisher Scientific). For transwell transfer, 10⁶ KO and WT bone marrow cells were grown for 7 days or until confluent in conditioned medium in chambers separated by Thincert 1-μm pore membranes (Greiner Bio-one) in 6-well plates and stained using the Quantigene ViewRNA Cell Plus assay with Alexa Fluor 546 detection, modified to include an additional fixation with 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride and protease treatment with reagents from the miRNA assay kit. Control wells were assayed for transfer of genomic material using genotyping primers for WT and KO Rpl13a loci (28).

qRT-PCR

RNA was purified using TRIzol. Relative quantification of RNA abundance was calculated using the ΔΔCt method on an ABI 7500 fast real-time PCR system. Primer sequences for COX2 were TGA GCA ACT ATT CCA AAC CAG C (forward) and GCA CGT AGT CTT CGA TCA CTA TC (reverse); for TNFα, they were CAT CTT CTC AAA ATT CGA GTG ACA A (forward) and TGG GAG TAG ACA AGG TAC AAC CC (reverse); and for Rplp0, they were ATC CCT GAC GCA CCG CCG TGA (forward) and TGC ATC TGC TTG GAG CCC ACG TT (reverse). The method for qRT-PCR for Rpl13a snoRNAs used target-specific stem loop RT primers, a target-specific forward PCR primer, and a universal reverse PCR primer (46). This method can detect as few as 10³ copies of target RNAs, is able to distinguish between small RNA species that differ by only 1 nucleotide, is not affected by genomic contamination, distinguishes mature snoRNAs and miRNAs from their precursor small RNA species, and has been validated against RNA-Seq for snoRNAs (10, 47, 48). mRNA quantification was calculated using Rplp0 as an endogenous control. For quantification of EV RNA, Caenorhabditis elegans miR39 (500 fmol) or luciferase mRNA (0.3 fmol; Promega) was added to EV samples before RNA isolation and used as a reference control. Amplification of reference genes varied by <1 cycle across samples. A plate of 2 × 10⁶ cells was used for cellular fractionation by sequential detergent extraction, and medium was pooled from three plates.

Microscopy

Freeze fracture/deep etch EM was performed according to published protocols with minor modifications (49). A 4-μl droplet of sample was placed on top of a 4 × 4-mm acid-cleaned and air-dried coverslip and then topped with a 3-mm round sapphire disc. Samples were immediately frozen by abrupt application of the sample against a liquid helium–cooled copper block with a Cryopress freezing machine. Frozen samples were transferred to a liquid nitrogen–cooled Balzers 400 vacuum evaporator, fractured at −100 °C by removal of the sapphire disc with the instrument's nitrogen cooled knife. Samples were etched for 2 min at −100 °C and rotary replicated with ∼2-nm platinum deposited from a 20º angle above the horizontal, followed by an immediate ∼4-nm stabilization film of pure carbon deposited from an 85º angle. Replicas were floated onto a dish of concentrated hydrofluoric acid for coverslip removal and then transferred through several rinses of distilled H₂O with a loopful of Photoflo, picked up on Formvar-coated grids, and photographed on a JEOL 1400 microscope with attached AMT digital camera. Confocal images were taken on a Zeiss LSM 880 Airyscan confocal microscope with a 0.8-μm pinhole through a focal plane in the center of the nuclei and processed using Zen software (Zeiss, Oberkochen, Germany).

RNase protection

EVs enriched by ultracentrifugation were suspended in 250 μl of PBS and treated with 0.3% Triton X-100, 1 unit of RNase A, 40 units of RNase T1 (Ambion RNase mixture, Thermo Fisher Scientific), and 100 units of SUPERase-In RNase inhibitor (Thermo Fisher Scientific) for 30 min at 20 ºC, and then RNase inhibitor was added to all samples. RNA was isolated using TRIzol LS.

Human subjects

Blood samples were collected from human volunteers for a study of lung inflammation before and 6 h following endobronchial instillation of 4 ng/kg LPS (29). This study was approved by the Washington University Human Research Protection Office and complies with the Declaration of Helsinki principles. Informed consent was obtained from all subjects.

Parabiosis

The Washington University Mouse Cardiovascular Phenotyping Core performed parabiosis surgery using female WT mice expressing GFP and Rpl13a snoRNA KO mice as described (30). Tail vein blood was sampled serially from the parabionts for PCR analysis of DNA to detect the WT and KO Rpl13a loci with primers GAC AGG TTG CTG CTC AGG AAG TAA ATGG (forward) and CCA GAC CTG CTT TCA GAC TTT AGC CTG (reverse). Three weeks following surgery, parabionts were perfused with PBS to clear blood from tissues and euthanized. Enterocytes and liver were recovered and flash-frozen in liquid nitrogen. For PCR analysis, tissues were homogenized in TRIzol.

Primer extension qRT-PCR

Primer extension PCR to detect rRNA methylation was adapted from Dong et al. (31). For each target, 20 ng of total RNA was used in cDNA synthesis using the Super Script III kit (Thermo Fisher Scientific) with 0.3 μl of SSIII enzyme per reaction and dNTP at either the normal stock concentration (50 μm) or diluted 1:1,000 (50 nm). The following primers were used for each RT reaction: m18S RT (U32a, G1328), CAA CTA AGA ACG GCC ATG CA; m18S RT (U33, U1326), AGA ACG GCC ATG CAC C; m28S RT (U34, U2590), GAC TTC CCT TAC CTA C; and m28S RT (U35a, C4188), AAC CTG TCT CAC GAC G. qPCR was done on a 7500 fast real-time PCR system with SYBR Green real-time PCR master mix (Thermo Fisher Scientific). PCR primers used were as follows: m18S F2 (U32a/33), GGC CTG CGG CTT AAT TTG AC; m18S R2 (U32/33), ATG CAC CAC CAC CCA CG; m28S F1 (U34), TCT CGC TGG CCC TTG AAA AT; m28S R1 (U34), ACC TAC ATT GTT CCA ACA TGC C; m28S F1 (U35a), CTT CGA TGT CGG CTC TTC CT; m28S R1 (U35a), TCA CGA CGG TCT AAA CCC AG. RT efficiency was calculated by normalizing the Ct value for the low dNTP concentration to the high as an endogenous control (RQ). For each modification site within a specific tissue, values are reported relative to the mean WT value.