Retrograde trafficking of Argonaute 2 acts as a rate-limiting step for de novo miRNP formation on endoplasmic reticulum-attached polysomes in mammalian cells

Abstract

microRNAs are short regulatory RNAs in metazoan cells. Regulation of miRNA activity and abundance is evident in human cells where availability of target messages can influence miRNA biogenesis by augmenting the Dicer1-dependent processing of precursors to mature microRNAs. Requirement of subcellular compartmentalization of Ago2, the key component of miRNA repression machineries, for the controlled biogenesis of miRNPs is reported here. The process predominantly happens on the polysomes attached with the endoplasmic reticulum for which the subcellular Ago2 trafficking is found to be essential. Mitochondrial tethering of endoplasmic reticulum and its interaction with endosomes controls Ago2 availability. In cells with depolarized mitochondria, miRNA biogenesis gets impaired, which results in lowering of de novo-formed mature miRNA levels and accumulation of miRNA-free Ago2 on endosomes that fails to interact with Dicer1 and to traffic back to endoplasmic reticulum for de novo miRNA loading. Thus, mitochondria by sensing the cellular context regulates Ago2 trafficking at the subcellular level, which acts as a rate-limiting step in miRNA biogenesis process in mammalian cells.

Figure 1.. Attachment of de novo–formed miRNAs with ribosomes attached with the rER membrane.

(A, B) Active miRNPs are membrane associated. FH-Ago2 was immunoprecipitated (IP) from total cell lysate as well as digitonin-soluble cytosolic and insoluble membrane fractions after fractionation of cells performed by treating with 50 μg/ml of digitonin. The IPed materials obtained from cell equivalent amount of each fraction were used in the in vitro RISC cleavage assays. Densitometric quantification of the cleaved RNA band was used for calculating specific activity of miR-122 and normalized against the amount of Ago2 IP following procedures described in the Materials and Methods section and plotted. Representative autoradiogram data obtained from one experimental set have been shown in panel (A) along with the Western blot data for IP FH-Ago2. The relative amount of Ago2 present in each reaction is shown in the Western blots showing levels of FH-Ago2 and GAPDH in panel (B). The FH-Ago2 detected with anti-HA antibody. RISC(-) is the reaction done without any addition of RISC. In Total (3×) reaction, threefold excesss of cytosolic fractions was used. (C, D) Subcellular distribution of miRNP components in human cells. FH-Ago2 stable HEK 293 cells transiently expressing miR-122 were lysed under isotonic conditions and subjected to ultracentrifugation on a 3–30% iodixanol gradient for separation of subcellular organelles. Hepatocyte growth actor-regulated tyrosine kinase substrate (HRS), calnexin, LAMP1, COX4, and Rab7 were used as markers of endosomes/multivesicular bodies (MVBs), rER, lysosomes, mitochondria, and late endosomes, respectively. Subcellular distribution of Dicer1 has been observed. FH-Ago2 was IPed from individual fractions using FLAG-specific antibody, and IPed materials were used for in vitro RISC cleavage assay. Specific activity of miR-122 RISC in each fraction has been plotted by normalizing the RISC activity present against the amount of Ago2 precipitated from each fraction. Pre-miR-122 levels in individual fractions have been quantified by qRT-PCR. Values for each fraction have been plotted as percentage of total miR-122 calculated by summing miR-122 present in all the fractions. Positions of ribosomal RNA bands are marked by * in the ethidium bromide–stained gel shown here. (E, F) Mature miRNAs are rER-enriched in HEK 293 cells. Relative enrichment of miR-16, miR-21, and let-7a as determined in isolated microsome by qRT-PCR when compared with equal amount of RNA from total cell lysate. Presence of Ago2 and DICER1 in microsome fractions was determined by Western blot. Absence of GAPDH (cytosolic marker) and β-Actin and presence of calnexin (rER marker) in the microsomal fraction confirmed the purity of isolated microsome as compared with protein equivalent amount of non-microsomal or total cell extract. (G) Schematic presentation of isolation of polysomes using KCl–puromycin based on HEK 293 cells. Microsomes isolated from HEK 293 cells were treated with 500 mM KCl and 1 mM puromycin before ultracentrifugation at 100,000g for 1 h to separate the ribosomal and non-ribosomal pool. (H) miRNP and target mRNAs are associated with the rER-attached ribosomes. Western blot data show the distribution of Ago2 in the soluble and pellet fractions after treatment and centrifugations. Calnexin was used as the ER marker protein. Mature miR-122 as well as its target mRNA RL-3xbulge-miR-122 distribution in the abovementioned pellet and soluble fractions were determined by qRT-PCR. (I) Exclusiveness of the Ago2 and miRNA extraction with KCl–puromycin. Treated and non-treated extracts were analyzed by Western blots using cell equivalent amounts for each fraction. ER (calnexin), ribosome (S3), and endosomes/MVBs (HRS) marker protein distributions were checked to observe the specificity and purity of the isolation method. Western blot data of HRS protein suggest that minimum contamination of endosomal fractions in the microsome that can account for the observed Ago2 that was found to be extracted with ribosome upon extraction with KCl and puromycin. (J) Relative amount of let-7a miRNA associated with soluble and pellet fractions with or without KCl–puromycin treatment was measured. Paired two-tailed t tests were used for all comparisons. P < 0.05 (*); P < 0.01 (**); P < 0.001 (***). In (B, D, E, H, J), values are means ± SEM from at least three biological replicates. Source data are available for this figure.

Figure 2.. Target-driven biogenesis occurs on the rER-attached polysomes.

(A) Scheme of digitonin-based membrane isolation from HEK 293 cells. (B) Relative steady-state distribution of miR-122 in total, cytosol, and digitonin-insoluble membranes in HEK 293 cells exogenously expressing the liver miRNA-122 in the presence of its target RL-3xbulge-miR-122 and control RL-con mRNA. Northern blot (left panel) where ethidium bromide staining of the gel used for Northern blotting of miRNA and U6 are shown. Quantification of the same (right panel) against the U6 RNA has been shown. (C) In vitro RISC cleavage assay with FH-Ago2 IP from cytosolic and membrane fractions. Membrane-associated miRISC isolated from target RNA–expressing cells showed higher specific activity for substrate RNA cleavage. The cleaved RNA products are marked by an arrowhead. (D) Target mRNA–induced de novo miRNPs get enriched on microsomes. Northern blotting was used to detect mature miR-122 and U6. RISC cleavage assay was performed with protein equivalent amounts of FH-Ago2 IP. (E) Target mRNA–driven de novo–formed miRNAs are present in the ER fraction. Relative distribution of newly synthesized miR-122 as well as endogenous let-7a in the presence of miR-122 target mRNA. From the 3–30% OptiPrep gradient, fractions 7, 8, and 9 were pooled to get rER-associated miRNA, whereas fractions 1–3 and fractions 4–6 were enriched, respectively, for endosome/MVBs– and lysosome-associated miRNAs (as described in Fig 1C and D). Amounts of RNA used for quantification in RT-PCR reaction were identical. (F) Inducible miR-122 associates with microsome in the presence of target mRNA. Real-time quantification of mature miR-122 present in microsomal fraction and in total after 24 h of induction in the presence of RL-3xBulge-miR-122 and control RL-3xBulge-let-7a were measured and plotted. (G, H) Increase in de novo–formed mature miR-122 level with time in the presence of its target mRNA. Levels of miR-122 were quantified and plotted after 0, 12, 16, 20, and 24 h of induction. (G) Values obtained with total cellular RNA were plotted in panel (G). (H) Quantifications of microsome and polysome-associated miR-122 are represented in panel (H). In both cases, the level at 0 h is taken as one unit. (I) De novo–formed miR-122 associates with rER-bound Ago2 in the presence of target mRNA. Tet-On–stable HEK 293 cells were transiently transfected with FH-Ago2, inducible miR-122, and RL-3xbulge-miR-122–expressing plasmids. FH-Ago2 IP from cytosolic and microsome after induction were used for quantification of associated mature miR-122. In all cases, the level at 0 h is taken as one unit. (J) The newly formed miRNPs are functional. Repression kinetics of RL-3xbulge-miR-122 after induction of de novo synthesis of miR-122 was followed and quantified. Tet-On–stable HEK 293 cells were transfected with either RL-con or RL-3xbulge-miR-122 and inducible miR-122 expression plasmids. After 0, 2, 4, 12, and 24 h of induction for miR-122 expression, Renilla Luciferase expression in cells transfected with RL-con or RL-3xbulge-miR-122 was measured in the dual luciferase assay. Firefly luciferase–expressing plasmids were co-transfected in all cases, and relative luciferase activity in 0 h was taken as unit. In all cases, Renilla luciferase (RL) expression were normalized by firefly expression. Paired two-tailed t tests were used for all comparisons. P < 0.05 (*); P < 0.01 (**); P < 0.001 (***). For statistical analysis, minimum three sets of data were used. Source data are available for this figure.

Figure S1.. Early time kinetics of miRNA formation and its association with microsome.

(A, B) Relative levels of total cellular (A) and microsome-associated (B) miR-122 in HEK 293 cells after indicated time points. Paired two-tailed t tests were used for all comparisons. P < 0.05 (*); P < 0.01 (**); P < 0.001 (***). Values are means from at least three biological replicates ± SD. Later time points of the response.

Figure 3.. Target mRNA–dependent miRNA formation on isolated cellular membranes in vitro.

(A) Target-dependent miRNA biogenesis assay in vitro. Schematic representation of the experiment is shown. Digitonin-insoluble membrane was incubated with rabbit reticulocyte lysate (RRL), synthetic pre-miR-122, and in vitro–transcribed mRNAs (50 ng) and subjected to in vitro translation reaction for 30 min at 30°C. After the reaction, the membranes were reisolated and mRNA levels associated with the membrane were quantified. (B) Enrichment of target mRNA in the presence of cognate miRNA on digitonin-insoluble membrane in vitro. (C) Effect of target mRNA concentration on miRNA biogenesis in vitro. The translation reaction with RRL was performed with isolated digitonin-insoluble membranes and with increasing concentration of in vitro transcribed RL-3xbulge-miR-122 (0, 10, 50, 100, and 200 ng) mRNA and synthetic pre-miR-122. After the reaction, membrane-associated mature miR-122 formed was quantified. (D) Increased association of RL-3xbulge-miR-122 and mature miR-122 with microsome in vitro. In vitro translation was performed with isolated microsome and RRL, pre-miR-122, target RNA for different time points (0, 15, and 30 min) and levels of mature miR-122 formed and attached with microsome were quantified after reisolation of microsome from the reaction. (E) Polysomal enrichment of target mRNA and its cognate miRNA in the in vitro translation cum miRNA biogenesis assay system. Polysome association of mRNA and miR-122 after in vitro translation for 30 min carried out with polysome isolated from HEK 293 cells in the presence of synthetic pre-miR-122 and in vitro–transcribed RL-con or RL-3xbulge-miR-122. (F) Targeting of mRNA to polysome precedes cognate miRNA biogenesis and its association with polysome in vitro. Time-dependent polysome association of target mRNA and de novo–synthesized cognate miRNAs with time. Translation reaction was carried out as described in previous panels, in the presence of RL-3xbulge-miR-122 and pre-miR-122, in a reaction mixture containing polysomes and RRL for different time intervals, and amounts of miRNA and mRNA associated with polysomes reisolated after the reaction were quantified and plotted. All experiments were carried out three times. Paired two-tailed t tests were used for all comparisons. P < 0.05 (*); P < 0.01 (**); P < 0.001 (***).

Figure 4.. Defective de novo miRNP formation in mitochondria-depolarized cells.

(A) Scheme of the experiment used to determine the de novo rate of Ago2-miR-122 miRNP formation using doxycycline-inducible Tet-On HEK 293 cells. The Tet-ON HEK 293 cells were transfected with doxycycline-inducible pre-miR-122 plasmid construct (ipmiR-122). Doxycycline was added to the cells for 48 h after transfection. The cells were harvested after doxycycline induction of indicated time periods, and immunoprecipitation was performed for endogenous Ago2 protein. miRNA estimation and relative quantification were performed for IPed Ago2-associated miRNA by qRT-PCR. (B, C) Values of Ago2-associated miR-122, mature miR-122, or precursor-miR-122 in Tet-on HEK 293 cells induced with doxycycline for specified time periods. Shown are the mean and SEM values from at least three independent experiments. (B, C) All cells were co-transfected with Renilla-3xbulge-miR-122 reporter, Tet-inducible pre-miR-122 expression plasmid ipmiR-122, and FH-Ucp2 (panel B) or siMfn2 (panel C). IPed Ago2 level, U6 small nuclear RNA level, and 18S ribosomal RNA level were used as internal controls for the estimation of Ago2-associated miR-122, mature miR-122, and precursor-miR-122, respectively. All values have been normalized against values at 0 h. (D) Shown are the mean and SEM from at least four independent experiments with values of Ago2-associated miR-122 in an iodixanol OptiPrepR (3–30%) gradient fractions positive for early endosome (EE) and ER markers. Lysate was obtained from TET-ON HEK 293 cells induced with doxycycline for 24 h. All cells were co-transfected with Renilla-3xbulge-miR-122 reporter, Tet-inducible pre-miR-122 expressing ipmiR-122, and either FH-Ucp2 or siMfn2. miR-122 present in ER fractions from each set were normalized against corresponding EE-associated miR-122 levels. (E, F) Target mRNA–driven miRNA biogenesis is hampered in mitochondria–ER uncoupled condition in Huh7 cells. (E) Scheme of the experiment is shown in panel (E). qRT-PCR data have confirmed no up-regulation of miR-122 biogenesis in the presence of its target under Ucp2 over-expression condition during amino acid re-feeding (F, left panel). CAT-1 mRNA level was also increased after 4 h of amino acid starvation on HA-Ucp2–expressing cells (F, right panel). (G) Western Blot data confirm HA-Ucp2 over-expression on Huh7 cells. (H) Pre-miR-122 expression in starved versus re-fed conditions in HA-Ucp2–expressed cells. Reduced pre-miRNA processing was observed upon re-fed of Ucp2 over-expressed cells compared with control. Paired two-tailed t tests were used for all comparisons. P < 0.05 (*); P < 0.01 (**); P < 0.001 (***). In (A, B, C, D, E, F, G, and H), values are means from at least three biological replicates ± SEM. Source data are available for this figure.

Figure S2.. Effect of mitochondrial depolarization on mitochondrial morphology in HeLa cells.

(A) Expression of HA-Ucp2 and depletion of Mfn2 by siRNA on mitochondrial morphology in HeLa cells. The depletion of Mfn2 was achieved by siRNA treatment (as shown in the Western blot), whereas the expression of HA-Ucp2 was also confirmed by Western blot analysis. Beta Actin blots were used as loading controls (B) Mitochondrial morphology in Mito-GFP–expressing HeLa cells, which also express FH-Ucp2 or treated with siMfn2.

Figure 5.. Increased target RNA trafficking to rER in Mfn2 negative cells.

(A) Decreased association of newely formed miR-122 with Ago2 on mitochondria-defective cells. qRT-PCR data showing Ago2-associated de novo formed miR-122 level decrease over time on Mfn2^−/− cells compared with WT cells was plotted. (B) Decreased de novo–formed target mRNA in Mfn^2−/− MEF cells. qRT-PCR data show reduced target mRNA levels in the presence of its cognate miRNA on Mfn2^−/− cells. (C) The relative amount of the reporter target mRNAs in the absence and presence of miR-122 has been shown in both cell types. All cells were co-transfected with Tet-ON expression plasmid and inducible Renilla-3xbulge-miR-122 reporter plasmid and FF reporter plasmid. In the right panel, relative fold repression of the miR-122 reporter has been studied in the presence and absence of miR-122 expression in wild-type and mitochondria-defective cells. (D, E) Higher microsomal association of target mRNA in Mfn2^−/− MEF cells. Microsomal sequestration of target mRNA in the presence and absence of its cognate miRNA has been plotted (D). The relative level of microsome/rER-associated target mRNA between normal and Mfn2-negative cells expressing pmiR-122 has also been shown (E). Western blot data of ribosomal protein S3 show equal amounts of microsome associated ribosomes were used for estimation of associated mRNAs. (F) Schematic representation of in vitro translation assay showing reisolation of rER used for the assay performed in the presence and absence of wild-type and mutant cell derived mitochondria. Calnexin, Ago2, and COX IV Western blots confirmed equivalent amount of microsome along with mitochondria after reisolation. (G) Relative estimation of RNA from reisolated microsome after incubation in the presence of mitochondria obtained from wild-type and Mfn2^−/− cells. Increased target mRNA sequestration in the presence of cognate miRNA in assays performed with detethered mitochondria was observed. (H) Measurement of miR-122 being formed and became associated with microsome showed reduced miR-122 association for per unit of RL-3xbulge-miR-122 mRNA when rER from wild-type cells were incubated with Mfn2^−/− mitochondria (left panel). Reduced microsomal association of miR-122 has been also observed when normalization was done with endogenous control Ago2 (right panel). In both panles while bars represent WT and black bars denote Mfn2^−/−. Paired two-tailed t tests were used for all comparisons. P < 0.05 (*); P < 0.01 (**); P < 0.001 (***). In (A, B, C, D, E, and F), values are means from at least three biological replicates ± SEM. Source data are available for this figure.

Figure S3.. Interaction of Dicer1 with endosomes in wild-type and Mfn2 null cells.

(A) Scheme of the procedure used to separate cellular organelles using iodixanol-based OptiPrep 3–30% gradients. (B, C) Distribution of Ago2 and Dicer1 in different subcellular compartments in wild-type and Mfn2^−/− MEF cells. Isotonic cell lysate extracts of WT or Mfn2 knockout MEFs were analyzed on a 3–30% OptiPrep gradient. Representative Western blot of indicated proteins was performed with individual fractions (B). Relative distribution of Ago2 and Dicer as percentage of total amount has been plotted in panel (C). HRS and Alix, calnexin, and LAMP1 were used as markers of endosomes/MVBs, rER, and lysosomes, respectively. (D, E, F) Representative combined frames showing endogenous Dicer1 (as shown in green or purple) protein localization with the ER (panel D) or endosomes (panel E) in MEFs of depicted genotypes. In panel (D), cells expressing an ER-targeting variant of DsRed (ER-DsRed, red) were stained for detection using indirect immunofluorescence for endogenous Dicer1 (green). Likewise, in panel (E), cells expressing an endosomal targeting variant of YFP (Endo-YFP, green) were stained for its visualization by indirect immunofluorescence for endogenous Dicer1 (purple). Marked areas are zoomed and ER regions of interest in the zoomed images have been enclosed by a dotted perimeter. White arrows depict foci with either elevated or reduced colocalization between ER or endosomal structures with Dicer1. Scale bars, 10 μm. Also shown are the mean ± SEM from at least four independent experiments (10 cells per experiment) of colocalization data between ER-Dicer1 (panel D) and endosome-Dicer1 (panel E). Mander’s coefficient was used to depict the colocalization between Dicer1 and DsRED-ER and YFP-endosome in respective cell types (panel F).

Figure 6.. Retarded Ago2 retro transport to rER-attached polysome in cells with defective ER–endosome interaction.

(A, B, C) Representative combined frames showing endogenous Ago2 protein localization with ER (A) or endosomes (B) in MEFs of depicted genotypes. In panel (A), cells expressing an ER-targeting variant of DsRed (ER-DsRed, red) were stained by indirect immunofluorescence for endogenous Ago2 (green). Likewise, in panel (B), cells expressing an endosomal targeting variant of YFP (Endo-YFP, green) were stained by indirect immunofluorescence for endogenous Ago2 (purple). Endo-GFP or ER-dsRED signals were detected by direct fluorescence of the tagged protein. Marked areas are zoomed and ER region of interest in the zoomed images has been enclosed by a dotted perimeter. White arrows depict foci with either elevated or reduced colocalization between ER or Endosomal structures with Ago2 bodies. Scale bars, 10 μm. Also shown are the mean ± SEM from at least four independent experiments (10 cells per experiment) of colocalization data between ER-Ago2 (C, top) and Endosome-Ago2 (C, bottom). Mander’s coefficient was used to represent the colocalization between organelles under consideration and Ago2 foci. (D, E, F) Scheme of the in vitro inter-organellar Ago2 transfer assay is shown in panel (D). Microsome isolated form WT or Mfn2^−/− MEF cells (designated as WT and KO) were incubated with MVBs isolated form FH-Ago2–expressing HEK 293 cells and subsequently the microsome was re-isolated. MVB isolation was performed by immunoprecipitation using anti-Rab7 antibody, and the subsequent incubation with microsome was performed on the beads bound to MVBs. Western blot data indicated transfer of FH-Ago2 from MVBs (HEK 293) to the microsome isolated from WT MEF cells, whereas Mfn2 knockout prevents transfer of FH-Ago2. Quantification of the same is shown in panel (F). Marker protein calnexin was used to indicate equality of re-isolated microsome content (E). (G) Level of Ago2-associated doxycycline-induced miR-122 at specified time points in MEFs of indicated genotype. All cells have been co-transfected with Tet-ON–expressing Renilla-3xBulge-miR-122 reporter and Tet-inducible pre-miR-122–expressing ipmiR-122 and complemented with either HA-Ago2 or HA-Ago2Y529E expression plasmids as indicated. miR-122 levels were normalized against corresponding values at 0 h. Shown are the mean and SEM from at least three independent experiments. P-values were calculated by t test, and one, two, and three asterisks represent P-values less than 0.05, 0.01, and 0.001, respectively. (H) Phosphorylation level of Ago2 was determined in subcellular fractions from MEFs of indicated genotypes. OptiPrep (3–30%) density gradients were used for separation of fractions positive for EE and ER and were used for endogenous Ago2 protein immunoprecipitation. The level of phosphorylated-Ago2 (4G10) present in IP Ago2 from ER and EE positive fractions was determined by Western blotting. (I, J) Immunoprecipitation of FH-Ago2 was performed with anti-HA antibody to pull down Ago2 from MEFs of indicated genotype transfected with Flag-HA tagged Ago2 (FH-Ago2). Western blot analysis was performed for indicated proteins to check their association with IPed Ago2 (panel I). In panel (J), antibodies against endogenous Ago2 was used to immunoprecipitate Ago2 to confirm its Dicer1 dissociation in MEF cells. Paired two-tailed t tests were used for all comparisons. P < 0.05 (*); P < 0.01 (**); P < 0.001 (***). In (A, B, C, D, E, F, G, H), values are means from at least three biological replicates ± SEM. Source data are available for this figure.

Figure 7.. Ago2 recycling determined by mitochondria controls Ago2–miRNA complex formation on the polysome attached to the rER.

In this schematic model, the mode of miRNA biogenesis regulation by mitochondria is shown. Polysome attached to the ER serves as the cellular sites where this event occurs. This is an unique example of organellar control of a posttranscriptional gene regulatory process where the components of miRNP complexes are differentially localized to the ER and endosome-associated compartments, and internal exchange of Ago2 between compartments by mitochondria-driven inter-organellar interactions controls the rate-limiting step of miRNP formation process.