ABSTRACT
Upon exposure to amyloid-β oligomers (Aβ1–42), glial cells start expressing proinflammatory cytokines, despite an increase in levels of repressive microRNAs (miRNAs). Exploring the mechanism of this potential immunity of target cytokine mRNAs against repressive miRNAs in amyloid-β-exposed glial cells, we have identified differential compartmentalization of repressive miRNAs in glial cells that explains this aberrant miRNA function. In Aβ1–42-treated cells, whereas target mRNAs were found to be associated with polysomes attached to endoplasmic reticulum (ER), the miRNA ribonucleoprotein complexes (miRNPs) were found to be present predominantly with endosomes that failed to recycle to ER-attached polysomes, preventing repression of mRNA targets. Aβ1–42 oligomers, by masking Rab7a proteins on endosomal surfaces, affected Rab7a interaction with Rab-interacting lysosomal protein (RILP), restricting the lysosomal targeting and recycling of miRNPs. RNA-processing body (P-body) localization of the miRNPs was found to be enhanced in amyloid-β-treated cells as a consequence of enhanced endosomal retention of miRNPs. Interestingly, depletion of P-body components partly rescued the miRNA function in glial cells exposed to amyloid-β and restricted the excess cytokine expression.
This article has an associated First Person interview with the first author of the paper.
INTRODUCTION
MicroRNAs (miRNAs) are 20–22-nucleotide-long non-coding RNAs that represses target mRNAs in both plant and animal cells. miRNAs associate with different argonaute proteins (Ago1 to Ago4) and form miRNA-induced silencing complexes (miRISC), which bind with target mRNAs and either repress or degrade them to stop protein expression (Filipowicz et al., 2008). On average, miRNAs have a half-life of 11.4 h, which is almost four times that of mRNAs, and thus miRNAs are in general more stable than mRNAs (Reichholf et al., 2019). The high stability of miRNA is possibly required to avoid the complicated multistep procedure of miRNA biogenesis, and metazoan cells use each copy of a miRNA repeatedly for repression of different target messages instead of making a new copy of the miRNA for each target mRNA. Interestingly, at steady state, there are several copies of each miRNA present in a cell to fine-tune expression of specific mRNAs.
miRNA-mediated repression and target degradation are two spatiotemporally uncoupled processes in human cells, and the mRNA degradation events that happen on the late endosome (LE) membrane play the pivotal role in recycling of miRNA ribonucleoprotein complexes (miRNPs) and, thus, for the next round of repression of target mRNAs by the respective miRNAs (Bose et al., 2017). In this context, the endoplasmic reticulum (ER) membrane acts as the site where nucleation of miRNA or siRNA takes place before the miRNPs encounter the target mRNA (Li et al., 2013; Stalder et al., 2013; Barman and Bhattacharyya, 2015; Bose et al., 2020). While perturbation of endosomal trafficking can affect siRNA function (Lee et al., 2009), entrapment of miRNPs at the ER restricts its export and abundance (Chakrabarty and Bhattacharyya, 2017).
RNA processing bodies (P-bodies, referred to hereafter as PBs) are considered to be the subcellular structures responsible for reversible storage of miRNA-repressed messages, and depletion of PBs is associated with relocalization of the repressed messages to translating polysomes and, thus, is considered as the key mechanism for reversible regulation of the repressive activity of miRNAs (Bhattacharyya et al., 2006). In a neuronal context, storage of Ago2 in PBs is considered a prerequisite for reversal of miRNA activity during growth factor withdrawal and is essential for neuronal survival (Patranabis and Bhattacharyya, 2018).
In this study, we report how exposure of rat glioblastoma cells to amyloid-β oligomers (Aβ1–42) decreases the cellular miRNA activity by enhancing the sequestration of miRNP complexes to early endosomes (EEs) and PBs. This also causes uncoupling of cytokine mRNAs, including those encoding IL-1β and IL-6, from Ago2, which accounts for their higher expression in diseased cells. Aβ1–42 oligomers perturb LE maturation and subsequent fusion with lysosomes, which is found to be important for miRNP recycling and binding with newly formed target mRNAs. We propose that this mechanism explains the increased proinflammatory cytokine expression and related neuroinflammation due to defects in miRNA recycling and target mRNA repression in Aβ1–42-exposed glial cells.
RESULTS
Target mRNAs with reduced miRNA repression in Aβ1–42-treated cells are associated with polysomes
Treatment of rat C6 glioblastoma cells with Aβ1–42 for 24 h increased mRNA levels of proinflammatory cytokines, including IL-1β and IL-6. qPCR data showed that 24 h exposure to 2.5 μM Aβ1–42 resulted in maximum induction of IL-1β and IL-6 mRNA levels (Fig. 1A). Because the 2.5 μM Aβ1–42 concentration yielded the maximum induction of cytokine mRNA levels, the rest of our experiments were carried out using an Aβ1–42 concentration of 2.5 μM. Interestingly, neither monomeric Aβ1–42 nor oligomeric Aβ1–40 at 2.5 μM concentration elicited any response in glioblastoma cells (Fig. S1A,B). Expression of cytokines is regulated by several miRNAs in mammalian cells, for example IL-6 mRNA is a direct target of miRNA let-7a (Iliopoulos et al., 2009). Does the repressive activity of miRNAs become impaired in treated cells? To check the activity of miRNAs upon Aβ1–42 exposure, C6 glioma cells were transfected with a Renilla luciferase (RL) reporter containing three imperfect miRNA let-7a binding sites (RL–3×bulge-let-7a; Fig. 1B). An almost 2-fold reduction in let-7a activity was observed in Aβ1–42-treated cells compared to activity in the DMSO-treated control cells (Fig. 1C). However, the total cellular level of let-7a was found to be upregulated upon exposure to Aβ1–42 (Fig. 1D). No changes were found in either let-7a repression or cellular let-7a level when cells were treated with either monomeric Aβ1–42 or oligomeric Aβ1–40 peptides (Fig. S1C,D). An exogenously expressed liver-specific miR-122 in C6 cells also showed similar reduction in repressive activity and increased cellular miRNA level, indicating that reduced miRNA activity upon Aβ1–42 treatment is not a miRNA identity-specific event. The expression of miR-122 was ensured by using a pre-miR-122-encoding plasmid containing a U6 promoter transfected in C6 glial cells, and the change in miR-122 levels in transfected cells was normalized to the level of U6 snRNA. The observed increase in mature miR-122 level and absence of any change in pre-miR-122 level upon Aβ1–42 treatment suggests that the increase in miRNA level occurs primarily at a post-transcriptional level and is not due to a transcriptional surge of pre-miR-122 in treated cells (Fig. 1D,E). Because Aβ1–42 treatment did not result in any decrease in either miRNA or Ago2 protein levels, we hypothesized that a lack of interaction between miRNA and Ago2 might explain the reduced miRNA activity in amyloid-β-exposed cells. To verify this hypothesis, we used C6 glioma cells expressing FLAG–HA-tagged Ago2 (FH–Ago2) and treated them with DMSO or Aβ1–42 for 24 h. FH–Ago2 was immunoprecipitated (IP) using anti-FLAG beads, followed by qPCR-based measurement of Ago2-bound miRNAs. The qPCR data was normalized against the amount of Ago2 immunoprecipitated from each of the samples. The qPCR data revealed an increased association between Ago2 and both miR-146a and miR-155 (Fig. 1F), two key miRNAs that are important regulators of proinflammatory cytokines in mammalian macrophage and glial cells (Sheedy and O'Neill, 2008; Kurowska-Stolarska et al., 2011). In contrast to the increased association of miRNA with Ago2, the interaction between Ago2 and the target mRNAs was found to be decreased in Aβ1–42-treated cells, as Ago2-association of cytokine IL-1β and IL-6 mRNAs was reduced considerably upon Aβ1–42 treatment (Fig. 1G). This suggests that the miRNPs that accumulate in Aβ1–42-treated cells are ineffective in finding their targets.
Uncoupling of target mRNA from miRNPs in Aβ1–42-exposed cells. (A) qPCR data showing cellular IL-1β (left) and IL-6 (right) levels in C6 glioblastoma cells activated with increasing concentrations of Aβ1–42 oligomer (Aβ) for 24 h. qPCR data was normalized to GAPDH mRNA level (n=4). (B) Schematic design of different Renilla luciferase (RL) reporters used for scoring miRNA activity. The respective miRNA binding sites were cloned into the 3′ UTR of the RL mRNA, adding a perfect site and three bulge sites for the miRNA. Fold repression is the ratio of normalized reporter expression versus normalized control (RL–con). (C,D) Effect of Aβ1–42 treatment on cellular miRNA activity and level. Specific activity of each miRNA was calculated by normalizing the fold repression of the target reporter mRNA to the cellular miRNA level. The repression data was obtained by luciferase-based quantification (C), whereas expression level was measured using qPCR and was normalized against the expression of U6 snRNA (D) (n=6). (E) Levels of pre-miRNA (pre-miR-122) in cells treated with Aβ1–42 oligomers, value normalized against 18S rRNA (n=3). (F,G) Aβ1–42-treated cells show increased Ago2–miRNA association but decreased Ago2-associated cytokine mRNA levels. Immunoprecipitation (IP) of FH–Ago2 from control and Aβ1–42 oligomer-treated (2.5 µM for 24 h) C6 glioblastoma cells expressing FH–Ago2 were performed, and RT-qPCR-based quantification was done. Values are normalized to the amount of Ago2 immunoprecipitated for the indicated miRNAs (F) and mRNAs (G) (n=4). Representative blots of input and IP are shown. (H) Distribution of Ago2 and other relevant proteins among endosomal (EE/MVB) and ER fractions in control cells (con) and cells treated with Aβ1–42. C6 glioblastoma cells were activated with 2.5 µM Aβ1–42 for 24 h before lysis, and lysates were ultracentrifuged on a 3–30% iodixanol (Optiprep) gradient. Fractions 2–4 were designated as MVBs and EEs, and fractions 7–9 were designated as ER. Alix was used as a marker of endosomes and MVBs, and calnexin was used as a marker for ER (blots are representative of n=3 experiments). (I) Input of cell extracts used for iodixanol (Optiprep) gradient analysis shown in H were western blotted for EEA1, Ago2 and β-actin. (J) Western blot data showing expression of EEA1 and Ago2 protein in endosomal fractions isolated from both control and Aβ1–42-treated cell lysate. Graph depicts Ago2 intensity change in the endosome fraction (n=3). (K) Distribution of cytokine mRNAs in different subcellular fractions after Aβ1–42 treatment of C6 glioblastoma cells. The mean threshold cycle (Ct) values were plotted for IL-1β and IL-6 mRNAs to check the distribution of cytokine mRNAs in both MVB and ER fractions in control and treated cells (n=3). (L) Relative levels of IL-1β and IL-6 mRNA associated with the polysomal fraction isolated from control and Aβ1–42-treated C6 glioblastoma cells. The qPCR data was normalized against the GAPDH mRNA present in the polysome fraction (n=3). For statistical significance, a minimum of three independent experiments were considered in each case, unless otherwise mentioned, and error bars represent the s.e.m. *P<0.05; **P<0.01; ***P<0.0001; ns, not significant (two-tailed, paired Student's t-test).
Uncoupling of target mRNA from miRNPs in Aβ1–42-exposed cells. (A) qPCR data showing cellular IL-1β (left) and IL-6 (right) levels in C6 glioblastoma cells activated with increasing concentrations of Aβ1–42 oligomer (Aβ) for 24 h. qPCR data was normalized to GAPDH mRNA level (n=4). (B) Schematic design of different Renilla luciferase (RL) reporters used for scoring miRNA activity. The respective miRNA binding sites were cloned into the 3′ UTR of the RL mRNA, adding a perfect site and three bulge sites for the miRNA. Fold repression is the ratio of normalized reporter expression versus normalized control (RL–con). (C,D) Effect of Aβ1–42 treatment on cellular miRNA activity and level. Specific activity of each miRNA was calculated by normalizing the fold repression of the target reporter mRNA to the cellular miRNA level. The repression data was obtained by luciferase-based quantification (C), whereas expression level was measured using qPCR and was normalized against the expression of U6 snRNA (D) (n=6). (E) Levels of pre-miRNA (pre-miR-122) in cells treated with Aβ1–42 oligomers, value normalized against 18S rRNA (n=3). (F,G) Aβ1–42-treated cells show increased Ago2–miRNA association but decreased Ago2-associated cytokine mRNA levels. Immunoprecipitation (IP) of FH–Ago2 from control and Aβ1–42 oligomer-treated (2.5 µM for 24 h) C6 glioblastoma cells expressing FH–Ago2 were performed, and RT-qPCR-based quantification was done. Values are normalized to the amount of Ago2 immunoprecipitated for the indicated miRNAs (F) and mRNAs (G) (n=4). Representative blots of input and IP are shown. (H) Distribution of Ago2 and other relevant proteins among endosomal (EE/MVB) and ER fractions in control cells (con) and cells treated with Aβ1–42. C6 glioblastoma cells were activated with 2.5 µM Aβ1–42 for 24 h before lysis, and lysates were ultracentrifuged on a 3–30% iodixanol (Optiprep) gradient. Fractions 2–4 were designated as MVBs and EEs, and fractions 7–9 were designated as ER. Alix was used as a marker of endosomes and MVBs, and calnexin was used as a marker for ER (blots are representative of n=3 experiments). (I) Input of cell extracts used for iodixanol (Optiprep) gradient analysis shown in H were western blotted for EEA1, Ago2 and β-actin. (J) Western blot data showing expression of EEA1 and Ago2 protein in endosomal fractions isolated from both control and Aβ1–42-treated cell lysate. Graph depicts Ago2 intensity change in the endosome fraction (n=3). (K) Distribution of cytokine mRNAs in different subcellular fractions after Aβ1–42 treatment of C6 glioblastoma cells. The mean threshold cycle (Ct) values were plotted for IL-1β and IL-6 mRNAs to check the distribution of cytokine mRNAs in both MVB and ER fractions in control and treated cells (n=3). (L) Relative levels of IL-1β and IL-6 mRNA associated with the polysomal fraction isolated from control and Aβ1–42-treated C6 glioblastoma cells. The qPCR data was normalized against the GAPDH mRNA present in the polysome fraction (n=3). For statistical significance, a minimum of three independent experiments were considered in each case, unless otherwise mentioned, and error bars represent the s.e.m. *P<0.05; **P<0.01; ***P<0.0001; ns, not significant (two-tailed, paired Student's t-test).
For an effective miRNA-mediated repression process in mammalian systems, all three main components of miRNA-mediated repression – Ago2, miRNA and target mRNA – are required to nucleate before effective repression can occur. Our observations suggest that the loss of miRNA activity was caused by loss of the interaction between miRNPs and target mRNA. This possible de-linkage of miRNP and target mRNA may occur due to differential localization of miRNPs and target messages in distinct subcellular compartments. To investigate the subcellular localization of miRNPs and their target mRNAs, a 3–30% iodixanol gradient ultracentrifugation was performed with cell homogenates obtained from control and Aβ1–42-treated C6 cells to separate different subcellular organelles, including multivesicular bodies (MVB) and the ER (Fig. 1H,I). There was an enrichment of Ago2 in the lighter fractions along with the EE marker EEA1 in Aβ1–42-treated cells compared to the control cells (Fig. 1J). RNA extraction followed by qPCR-based quantification from both Alix (also known as PDCD6IP)-enriched (endosomes) and calnexin-enriched (ER) fractions showed higher levels of IL-1β and IL-6 cytokine mRNAs in ER-enriched fractions in Aβ1–42-treated cells (Fig. 1K). More specifically, increased association of IL-1β and IL-6 mRNAs was found with polysomal fractions, which explains the impaired repression of these mRNAs by miRNAs that leads to their increased expression in Aβ1–42-treated C6 glioblastoma cells (Fig. 1L).
Compartmentalization of miRNPs to early endosomes in glial cells treated with Aβ1–42 oligomers
The experiments described above revealed that Aβ1–42 causes reduced interaction between Ago2 and cytokine mRNAs, with derepressed mRNAs associating with the ER, while Ago2 is enriched in endosomal fractions (Fig. 1J,K). There have been previous studies suggesting that contact between ER and endosomes is required for endosomal maturation, and that the maturation process may not happen in isolation (Eden et al., 2010). Endosomes form contact sites with the ER that help both endosome maturation and exchange of cargo between endosomes and the ER (Rocha et al., 2009). We checked the possibility that Aβ1–42 treatment could perturb interaction between the ER and endosomes to lead to compartmentalization of miRNPs in treated cells. Confocal images were taken for both control and Aβ1–42-treated cells, where C6 glioblastoma cells were transiently transfected with ER–DsRed-encoding plasmid, which ensured tagging of the ER, whereas EEs and LEs were detected using anti-EEA1 and anti-Rab7a antibodies (Fig. 2A). We found a significant decrease in colocalization for both ER–EE and ER–LE pairs upon Aβ1–42 treatment (Fig. 2B). LEs are known to interact more with ER than EEs, and EEs show increased interaction with ER as endosome mature (Friedman et al., 2013). Interestingly, Aβ1–42 exposure diminished the increase in ER–EE and ER–LE colocalizations (Fig. 2B), which may affect the endosome maturation process as well. However, no such interaction loss between lysosomes and ER was observed in Aβ1–42-treated cells (Fig. 2B). Live-cell time-lapse microscopy of both control and Aβ1–42-treated C6 cells expressing YFP–Endo and ER–DsRed, which label EEs and ER, respectively, also showed reduced interaction upon Aβ1–42 exposure (Fig. S2A). To confirm the role of Aβ1–42 in regulating the interaction between the ER and EEs, an in vitro reaction was carried out with enriched EE and ER fractions in the presence of Aβ1–42 and ATP. An IP of EEA1 followed by western blotting revealed reduced interaction between EEs and ER upon amyloid-β exposure (Fig. S2B,C). To find out the cellular localization of non-functional miRNPs, a 3–15% iodixanol gradient ultracentrifugation was carried out with cell lysates taken from both the control and Aβ1–42-treated cells. The EEs and LEs were separated on the 3–15% gradient, and the presence of Ago2 in the EE fraction was noted (Fig. 2C,D). RNA extraction followed by qPCR-based quantification from both HRS (also known as HGS)-enriched (EE) and Rab7a-enriched (LE) fractions showed an abundance of both miR-146a and miR-155 in EE fractions in Aβ1–42-treated cells (Fig. 2E). Increased presence of Ago2 with endosomes separated by organellar IP further established the idea that Ago2 associates with the EE fraction upon Aβ1–42 treatment (Fig. 2F). Enrichment of Ago2-associated miRNA levels in the EE fraction in Aβ1–42-treated cells was also noted (Fig. 2G).
Loss of ER–endosome interaction leads to miRNPs being retained with early endosomes. (A,B) Decreased ER–endosome interaction upon Aβ1–42 (Aβ) treatment. (A) EEs and LEs were tagged with anti-EEA1 and anti-Rab7a antibodies, respectively, in control DMSO-treated cells (Con) and Aβ-treated cells. ER was tagged with ER–DsRed. Confocal images showing colocalized area in merge images (yellow) between EEs or LEs and the ER. Boxes indicate regions shown in magnified images on the right. Arrowheads show possible interactive regions. (B) Pearson's coefficient of colocalization between the ER and endosomes was measured for both DMSO and 2.5 µM Aβ1–42 treatment (left). Similarly, the interactions between lysosomes and the ER were also measured and plotted for control and Aβ1–42-treated cells (right). n=3, ≥30 cells used for quantification. Scale bars: 10 μm. (C,D) Distribution of Ago2 and miRNA on a 3–15% iodixanol (Optiprep) gradient to separate the early (EE) and late (LE) endosomes of control and Aβ-treated cells. (C) The Optiprep fractions were western blotted for HRS, Rab7a and Ago2. (D) Quantification of percentage Ago2 present in each fraction was performed. HRS and Rab7a were used as the markers of EE and LE, respectively (n=2). (E) Distribution of miR-146 and miR-155 in EE and LE fractions in Aβ1–42-treated cells. The mean threshold cycle (Ct) values were plotted for these miRNAs to check the distribution in both EE and LE fractions (n=3). (F) Western blot analysis showing level of Ago2 from organellar IP samples from both control and Aβ1–42-treated cells. EEA1 shows the presence of the immunoprecipitated early endosome fraction (blots are representative of n=3 experiments). (G) Estimation of Ago2-associated miR-122 level in EE fractions of control and Aβ1–42-treated C6 glioblastoma cells. The qPCR data (right) was normalized to the amount of Ago2 pulled down during IP (n=2). Error bars represent the s.d. A representative IP for the EE fraction is shown on the left. (H) Representative image showing tracks of EEs analyzed from time-lapse video microscopy. Color code depicts the track speed of individual EEs (blue, low speed; red, high speed). Scale bars: 10 μm. (I,J) Graphs representing the relative frequency distributions (top) and values (bottom) of mean track speed (I) and mean track length (J) of individual EEs of both control and Aβ1–42-treated glioblastoma cells (n=3, ≥20 cells used for quantification). For statistical significance, a minimum of three independent experiments were considered in each case unless otherwise mentioned, and error bars represent the s.e.m. *P<0.05; ***P<0.0001; ns, not significant (two-tailed, unpaired Student's t test in B; two-tailed, paired Student's t-test in I,J).
Loss of ER–endosome interaction leads to miRNPs being retained with early endosomes. (A,B) Decreased ER–endosome interaction upon Aβ1–42 (Aβ) treatment. (A) EEs and LEs were tagged with anti-EEA1 and anti-Rab7a antibodies, respectively, in control DMSO-treated cells (Con) and Aβ-treated cells. ER was tagged with ER–DsRed. Confocal images showing colocalized area in merge images (yellow) between EEs or LEs and the ER. Boxes indicate regions shown in magnified images on the right. Arrowheads show possible interactive regions. (B) Pearson's coefficient of colocalization between the ER and endosomes was measured for both DMSO and 2.5 µM Aβ1–42 treatment (left). Similarly, the interactions between lysosomes and the ER were also measured and plotted for control and Aβ1–42-treated cells (right). n=3, ≥30 cells used for quantification. Scale bars: 10 μm. (C,D) Distribution of Ago2 and miRNA on a 3–15% iodixanol (Optiprep) gradient to separate the early (EE) and late (LE) endosomes of control and Aβ-treated cells. (C) The Optiprep fractions were western blotted for HRS, Rab7a and Ago2. (D) Quantification of percentage Ago2 present in each fraction was performed. HRS and Rab7a were used as the markers of EE and LE, respectively (n=2). (E) Distribution of miR-146 and miR-155 in EE and LE fractions in Aβ1–42-treated cells. The mean threshold cycle (Ct) values were plotted for these miRNAs to check the distribution in both EE and LE fractions (n=3). (F) Western blot analysis showing level of Ago2 from organellar IP samples from both control and Aβ1–42-treated cells. EEA1 shows the presence of the immunoprecipitated early endosome fraction (blots are representative of n=3 experiments). (G) Estimation of Ago2-associated miR-122 level in EE fractions of control and Aβ1–42-treated C6 glioblastoma cells. The qPCR data (right) was normalized to the amount of Ago2 pulled down during IP (n=2). Error bars represent the s.d. A representative IP for the EE fraction is shown on the left. (H) Representative image showing tracks of EEs analyzed from time-lapse video microscopy. Color code depicts the track speed of individual EEs (blue, low speed; red, high speed). Scale bars: 10 μm. (I,J) Graphs representing the relative frequency distributions (top) and values (bottom) of mean track speed (I) and mean track length (J) of individual EEs of both control and Aβ1–42-treated glioblastoma cells (n=3, ≥20 cells used for quantification). For statistical significance, a minimum of three independent experiments were considered in each case unless otherwise mentioned, and error bars represent the s.e.m. *P<0.05; ***P<0.0001; ns, not significant (two-tailed, unpaired Student's t test in B; two-tailed, paired Student's t-test in I,J).
In the experiments described above, we found non-functional miRNPs to remain with the EE fraction in Aβ1–42-treated cells, whereas in untreated cells, they are known to get shuttled to LEs along with the repressed mRNAs for degradation, miRNP recycling or MVB entrapment of miRNAs for extracellular export (Mukherjee et al., 2016; Bose et al., 2017). Subsequently, we studied the endosomal maturation process in Aβ1–42-treated cells. The number and surface area of both EEs and LEs were measured. On exposure to Aβ1–42, even though EEs were found to be enlarged in glial cells, the number of LEs was found to be decreased, suggesting a possible retardation of maturation from early to late endosomes (Fig. S2D,E). We hypothesized that Aβ1–42 may affect the mobility of enlarged EEs resulting in reduced EE–ER association and EE–LE maturation. We compared the frequency distributions of EE track speed and track length for control and Aβ1–42-treated cells, as assessed using time-lapse live-cell microscopy (Fig. 2H). Although the majority of both control and Aβ1–42-treated EEs showed track speeds greater than 1.0 μm/s, Aβ1–42-treated EEs were significantly slower (Fig. 2I). Furthermore, we found a significant reduction in mean track length of EEs in Aβ1–42-treated cells as compared to those of EEs in control cells (Fig. 2J). Hence, from the above data we could conclude that Aβ1–42 may cause a decrease in mobility of EEs.
Perturbation of endosomal maturation in Aβ1–42-treated cells causes defects in miRNP recycling and de novo target mRNA repression
Defects in the endosomal pathway have previously been reported in different neurological disorders, such as Alzheimer's disease (AD), and in Niemann–Pick disease type C (Nixon, 2005; Maxfield, 2014). Also, there are reports that suggest that EEs are the site where amyloid precursor protein (APP) localizes to form pathogenic amyloid-β protein (Grbovic et al., 2003). The onset of defects in the endosomal pathway has been noted in a murine model of AD, and researchers have also reported development of AD in Down syndrome-affected patients at a very early age due to a defect in the endosomal pathway (Cataldo et al., 2000).
Reduced mobility of EEs upon Aβ1–42 exposure and loss of endosome–ER interactions could play an important role in endosomal cargo delivery to lysosomes. To assess the impact of EE dynamics on lysosomal cargo delivery, we assayed LEs and observed a slight reduction in track speed and track length upon amyloid exposure (Fig. S3A–C). However, a significant decrease in colocalization between LEs and lysosomes substantiated the idea of an altered endosome maturation process altogether in Aβ1–42-treated cells (Fig. 3A,B; Fig. S3D). A reduced interaction between the LE marker Rab7a and lysosomal protein RILP (Rab-interacting lysosomal protein) was found in the organellar co-immunoprecipitation assay. This confirmed the reduced fusion of LEs with lysosomes upon treatment with Aβ1–42 (Fig. S2F,G). Therefore, it is possible that the endosomal maturation defect is linked with the miRNP inactivation observed in glial cells.
Endosomal maturation is important for miRNA activity regulation and miRNP recycling. (A) Reduced LE–lysosome interaction upon Aβ1–42 exposure in C6 glioblastoma cells. Confocal images showing colocalization between LEs and lysosomes. LEs were visualized using anti-Rab7a antibody (green) and lysosomes were stained with LysoTracker (red) in control cells (Con) and cells treated with Aβ1–42 (Aβ). Colocalized regions are shown in yellow. Boxes indicate regions shown magnified on the right. Scale bars: 10 μm. (B) Extent of colocalization between LEs and lysosomes was measured by calculating the Pearson's coefficient of colocalization between the green (LE) and red (lysosome) pixels (n=3, ≥45 cells used for quantification). (C) Levels of miRNA-interacting and related proteins in HEK293 cells where endosome maturation was blocked. Representative western blot showing the levels of cellular Ago2 in HEK293 cells treated with siRNA targeting RILP (siRILP) or control siRNA (siCon). β-actin was used as loading control. Levels of RILP and the endosomal component Rab5a were also estimated (n=3). (D–F) Effect of the endosome maturation defect on miRNA levels (D), activity (E) and Ago2 association (F). Levels of let-7a were measured in cellular lysates, or in Ago2 IPs in siCon- and siRILP-treated cells. RNA was recovered and RT-qPCR estimation was performed. The value was normalized against U6 snRNA for cellular samples, whereas the amount of immunoprecipitated Ago2 was used for normalization of the amount of let-7a present in the IP samples. Activity of let-7a was measured by quantifying the repression of a reporter mRNA in control and RILP-knockdown cells (n=4). (G,H) Defective recycling of miRNPs and poor re-binding to de novo synthesized mRNAs in mammalian cells defective for endosome maturation. Control and siRILP-treated TET-ON HEK293 cells were co-transfected with inducible iRL–3×bulge-miR-122, FH–Ago2 and miR-122-expressing pre-miR-122 plasmids. After 36 h of transfection, iRL-3×bulge-miR-122 was induced using doxycycline for 14 h and 24 h, followed by IP of FH–Ago2 to check Ago2-associated reporter mRNA levels. (G) The scheme of the experiments is shown. (H) Levels of total (bottom left) and Ago2-associated (bottom-right) induced RL mRNA (iRL–3×bulge-miR-122) were measured and quantified at 14 h and 24 h post induction. RL mRNA levels are normalized to the amount of Ago2 pulled down in the IP experiments. Relative cellular RL mRNA levels confirm positive induction of reporter target mRNA in both control and RILP-knockdown cells. This data was normalized to 18S rRNA (n=4). A representative blot showing the results of Ago2 IP is shown (top). For statistical significance, a minimum of three independent experiments were considered in each case, unless otherwise mentioned. Error bars represent the s.e.m. *P<0.05; **P<0.01 (two-tailed, unpaired Student's t-test in B; two-tailed, paired Student's t-test in D–F,H).
Endosomal maturation is important for miRNA activity regulation and miRNP recycling. (A) Reduced LE–lysosome interaction upon Aβ1–42 exposure in C6 glioblastoma cells. Confocal images showing colocalization between LEs and lysosomes. LEs were visualized using anti-Rab7a antibody (green) and lysosomes were stained with LysoTracker (red) in control cells (Con) and cells treated with Aβ1–42 (Aβ). Colocalized regions are shown in yellow. Boxes indicate regions shown magnified on the right. Scale bars: 10 μm. (B) Extent of colocalization between LEs and lysosomes was measured by calculating the Pearson's coefficient of colocalization between the green (LE) and red (lysosome) pixels (n=3, ≥45 cells used for quantification). (C) Levels of miRNA-interacting and related proteins in HEK293 cells where endosome maturation was blocked. Representative western blot showing the levels of cellular Ago2 in HEK293 cells treated with siRNA targeting RILP (siRILP) or control siRNA (siCon). β-actin was used as loading control. Levels of RILP and the endosomal component Rab5a were also estimated (n=3). (D–F) Effect of the endosome maturation defect on miRNA levels (D), activity (E) and Ago2 association (F). Levels of let-7a were measured in cellular lysates, or in Ago2 IPs in siCon- and siRILP-treated cells. RNA was recovered and RT-qPCR estimation was performed. The value was normalized against U6 snRNA for cellular samples, whereas the amount of immunoprecipitated Ago2 was used for normalization of the amount of let-7a present in the IP samples. Activity of let-7a was measured by quantifying the repression of a reporter mRNA in control and RILP-knockdown cells (n=4). (G,H) Defective recycling of miRNPs and poor re-binding to de novo synthesized mRNAs in mammalian cells defective for endosome maturation. Control and siRILP-treated TET-ON HEK293 cells were co-transfected with inducible iRL–3×bulge-miR-122, FH–Ago2 and miR-122-expressing pre-miR-122 plasmids. After 36 h of transfection, iRL-3×bulge-miR-122 was induced using doxycycline for 14 h and 24 h, followed by IP of FH–Ago2 to check Ago2-associated reporter mRNA levels. (G) The scheme of the experiments is shown. (H) Levels of total (bottom left) and Ago2-associated (bottom-right) induced RL mRNA (iRL–3×bulge-miR-122) were measured and quantified at 14 h and 24 h post induction. RL mRNA levels are normalized to the amount of Ago2 pulled down in the IP experiments. Relative cellular RL mRNA levels confirm positive induction of reporter target mRNA in both control and RILP-knockdown cells. This data was normalized to 18S rRNA (n=4). A representative blot showing the results of Ago2 IP is shown (top). For statistical significance, a minimum of three independent experiments were considered in each case, unless otherwise mentioned. Error bars represent the s.e.m. *P<0.05; **P<0.01 (two-tailed, unpaired Student's t-test in B; two-tailed, paired Student's t-test in D–F,H).
To investigate a possible role of the endosomal maturation pathway in regulation of miRNA activity, HEK293 cells were knocked down for RILP, which is required for the fusion of endosomes with lysosomes. RILP is a Rab effector protein that facilitates cargo delivery from LEs to lysosomes (Cantalupo et al., 2001; Progida et al., 2007). Knockdown of RILP increased both total cellular Ago2 level (Fig. 3C) and let-7a miRNA level (Fig. 3D). To check the effect of downregulation of RILP on let-7a activity, a luciferase assay was performed by transfecting the RL–3×bulge-let-7a reporter, which has three imperfect let-7a-binding sites. The assay revealed low miRNA activity when endosome maturation was compromised in RILP siRNA-treated HEK293 cells (Fig. 3E). An Ago2 pulldown assay revealed increased miRNP levels in RILP-compromised cells (Fig. 3F). The data obtained using RILP-compromised non-neuronal HEK293 cells had a similar trend of miRNA activity alteration as was observed in Aβ1–42-treated glioblastoma cells. Additionally, exposure to bafilomycin A1, which blocks endo-lysosomal fusion by inhibiting V-ATPases also increases total Ago2 level along with cellular miRNAs (Fig. S3E,F). We also found a reduction in exogenously expressed miR-122 activity (Fig. S3G) and an increase in Ago2-associated miR-122 level in HEK293 cells upon bafilomycin A1 treatment, similar to that observed upon RILP knockdown (Fig. S3H). This further establishes the idea that endo-lysosomal fusion has a role in regulation of miRNA activity in mammalian cells.
To further test the effect of the endosome maturation defect in RILP-compromised cells on de novo target recognition by miRNPs, miR-122-expressing TET-ON HEK293 cells were used to express the miR-122 target reporter RL–3×bulge-miR122 in an inducible manner. It was found that Ago2 in RILP-knockdown cells showed a reduced association with the de novo synthesized mRNA, which may account for the reduced repressive activity we observed in cells depleted for RILP (Fig. 3G,H).
Endosomal retention-induced P-body entrapment of miRNPs inhibits miRNA activity in amyloid-β-exposed cells
Using expression of a constitutively active GTPase-deficient Rab5a mutant, Rab5Q79C (hereafter referred to as Rab5-CA), which is known to disrupt early to late endosome maturation (Wegner et al., 2010), we documented excess expression of the proteins Dcp1a and RCK/p54 (also known as DDX6), along with Ago2, which are known to accumulate in PBs in neuronal cells undergoing differentiation (Fig. S4A) (Patranabis and Bhattacharyya, 2018). We also noted an increased number of PBs in Rab5-CA-expressing C6 cells (Fig. S4A,B). It has been reported previously that endosome maturation and degradation of PB component proteins are linked (Siomi and Siomi, 2009). C6 cells exposed to bafilomycin A1 also showed increased numbers of RCK/p54 bodies (Fig. S4C,D). Therefore, the defect in endosome maturation has an effect on PB components.
Increased aggregation of RNA granules has been reported in different neurological contexts (Fan and Leung, 2016). Re-localization of RNA-binding protein TDP43 (also known as TARDBP) from nucleus to cytoplasm is a significant phenomenon in amyotrophic lateral sclerosis (ALS) (Chen-Plotkin et al., 2010). Researchers have also reported the presence of huntingtin protein (Htt) in PBs, which co-purified with Ago2 protein (Savas et al., 2008). So, the abundance of PBs in Aβ1–42-exposed cells could be the direct cause of pathogenesis and, by inactivating the miRNPs, may account for the reduced miRNA activity (but not abundance) in the disease context. To investigate the possible role of PBs in regulation of miRNA activity, rat glioblastoma cells were stained for different markers of PBs and confocal imaging was performed to measure the size and number of different bodies positive for Dcp1a and RCK/p54. Colocalization between these two proteins was also estimated (Fig. 4A). We found significant increase in Dcp1a-positive body size and also increased colocalization of Dcp1a and RCK/p54 in PBs after Aβ1–42 treatment (Fig. 4B). This was accompanied by an almost 2.5-fold increase in both Dcp1a- and RCK/p54-positive body number (Fig. 4C). We also noted ∼2-fold increase in colocalization between Dcp1a- and RCK/p54-positive bodies, indicating a correlation between PB size and number with Aβ1–42 treatment (Fig. 4D). In that context, expression of PB components also increased in Aβ1–42-exposed cells (Fig. 4E). To investigate the possibility of Ago2 sequestration in PBs in Aβ1–42-treated cells, C6 glioma cells were transiently transfected with GFP–Ago2 and then counter-stained for RCK/p54 as a marker of PBs (Fig. 4F). The results showed a significant increase in Ago2-positive bodies (Fig. 4G) in treated cells. We also found a 3-fold increase in colocalization coefficient between Ago2-positive bodies and RCK/p54-positive bodies indicating that, in Aβ1–42-treated cells, endosome-sequestered Ago2 miRNPs translocate to the PBs, which might contribute to the reduced miRNP activity observed in treated cells (Fig. 4G).
Aβ1–42-activated cells show increased targeting of Ago2 to PBs, and deactivation of PBs restores miRNA activity. (A–D) Increased number of PBs in cells treated with Aβ1–42 oligomers. (A) Colocalization of endogenous Dcp1a (red) with endogenous RCK/p54 (green) upon 24 h treatment of C6 glioblastoma cells with 2.5 μM Aβ1–42 oligomers (Aβ). Scale bars: 10 μm. (B) Size of the individual Dcp1a-positive and RCK/p54-positive bodies, and of P-bodies positive for both proteins, were measured and plotted for both control (Con) and Aβ1–42-treated C6 glioblastoma cells. The plot was generated from three individual experiments (minimum of 10 cells/experiment). (C) Number of individual bodies was calculated for each cell for both control and Aβ1–42-treated C6 glioblastoma cells. The plot was generated from three individual experiments (minimum of 10 cells/experiment). (D) Colocalization of Dcp1a with RCK/p54-positive bodies was measured by calculating Pearson's and Mander's coefficient in control and Aβ1–42-treated C6 glioblastoma cells (n=4, ≥50 cells used for quantification). (E) Levels of expression of PB components in control and Aβ1–42-treated C6 glioblastoma cells. β-actin was used as a loading control (blots are representative of n=3 experiments). (F,G) Increased colocalization between Ago2 and RCK/p54-positive bodies. (F) Transiently expressed GFP–Ago2 (green) with endogenous RCK/p54 (red) in control and Aβ1–42-treated C6 glioblastoma cells. GFP–Ago2 and Dcp1a colocalized bodies are in yellow. Scale bars: 10 μm. (G) Number of Ago2-positive bodies per cell was calculated and was plotted for both control and Aβ1–42-treated C6 glioblastoma cells (left). Translocation of Ago2 to the RCK/p54-positive PBs was measured by calculating the Pearson's coefficient of colocalization between transiently transfected GFP–Ago2 with the endogenous RCK/p54 protein (right) (n=3, ≥30 cells used for quantification). (H–J) Effect of depletion of PB components on Aβ1–42-induced miRNA and cytokine expression. (H) Effect of siRNA treatment on expression of Dcp1a in control siRNA (si-con) and Dcp1a siRNA (si-Dcp1a)-treated cells. Ago2 and β-actin blots are shown as controls. Blots are representative of three experiments. (I) Relative mRNA level of expression of IL-6 (right) and IL-1β (left) in control and in PB-depleted cells that were either non-treated or treated with Aβ1–42 oligomers. PBs were depleted by treatment with a combination of si-Dcp1a and si-Lsm1, an siRNA that targets Lsm1. (J) miR-146a levels were similarly measured. For I and J, quantification was done by RT-qPCR-based estimation. For mRNA, GAPDH levels were used for normalization. U6 snRNA was used for normalization of miRNA levels. Values for untreated control cells were used as unit (n=4). (K) Relative cytokine mRNA (right) and miR-146a association (middle) with Ago2 in Aβ1–42-treated control (si-con_Aβ) and PB-depleted (si-Dcp1a+si-Lsm1_Aβ) cells. The amount of Ago2 isolated by IP was determined (see representative western blot, left) and used for isolation and quantification of associated mRNA and miRNA levels. RT-qPCR based methods were used for RNA estimation (n=3). For statistical significance, a minimum of three independent experiments were considered in each case, unless otherwise mentioned. Error bars represent the s.e.m. *P<0.05; **P<0.01; ***P<0.0001; ns, not significant (two-tailed, unpaired Student's t-test in B–D,G; two-tailed, paired Student's t-test in I).
Aβ1–42-activated cells show increased targeting of Ago2 to PBs, and deactivation of PBs restores miRNA activity. (A–D) Increased number of PBs in cells treated with Aβ1–42 oligomers. (A) Colocalization of endogenous Dcp1a (red) with endogenous RCK/p54 (green) upon 24 h treatment of C6 glioblastoma cells with 2.5 μM Aβ1–42 oligomers (Aβ). Scale bars: 10 μm. (B) Size of the individual Dcp1a-positive and RCK/p54-positive bodies, and of P-bodies positive for both proteins, were measured and plotted for both control (Con) and Aβ1–42-treated C6 glioblastoma cells. The plot was generated from three individual experiments (minimum of 10 cells/experiment). (C) Number of individual bodies was calculated for each cell for both control and Aβ1–42-treated C6 glioblastoma cells. The plot was generated from three individual experiments (minimum of 10 cells/experiment). (D) Colocalization of Dcp1a with RCK/p54-positive bodies was measured by calculating Pearson's and Mander's coefficient in control and Aβ1–42-treated C6 glioblastoma cells (n=4, ≥50 cells used for quantification). (E) Levels of expression of PB components in control and Aβ1–42-treated C6 glioblastoma cells. β-actin was used as a loading control (blots are representative of n=3 experiments). (F,G) Increased colocalization between Ago2 and RCK/p54-positive bodies. (F) Transiently expressed GFP–Ago2 (green) with endogenous RCK/p54 (red) in control and Aβ1–42-treated C6 glioblastoma cells. GFP–Ago2 and Dcp1a colocalized bodies are in yellow. Scale bars: 10 μm. (G) Number of Ago2-positive bodies per cell was calculated and was plotted for both control and Aβ1–42-treated C6 glioblastoma cells (left). Translocation of Ago2 to the RCK/p54-positive PBs was measured by calculating the Pearson's coefficient of colocalization between transiently transfected GFP–Ago2 with the endogenous RCK/p54 protein (right) (n=3, ≥30 cells used for quantification). (H–J) Effect of depletion of PB components on Aβ1–42-induced miRNA and cytokine expression. (H) Effect of siRNA treatment on expression of Dcp1a in control siRNA (si-con) and Dcp1a siRNA (si-Dcp1a)-treated cells. Ago2 and β-actin blots are shown as controls. Blots are representative of three experiments. (I) Relative mRNA level of expression of IL-6 (right) and IL-1β (left) in control and in PB-depleted cells that were either non-treated or treated with Aβ1–42 oligomers. PBs were depleted by treatment with a combination of si-Dcp1a and si-Lsm1, an siRNA that targets Lsm1. (J) miR-146a levels were similarly measured. For I and J, quantification was done by RT-qPCR-based estimation. For mRNA, GAPDH levels were used for normalization. U6 snRNA was used for normalization of miRNA levels. Values for untreated control cells were used as unit (n=4). (K) Relative cytokine mRNA (right) and miR-146a association (middle) with Ago2 in Aβ1–42-treated control (si-con_Aβ) and PB-depleted (si-Dcp1a+si-Lsm1_Aβ) cells. The amount of Ago2 isolated by IP was determined (see representative western blot, left) and used for isolation and quantification of associated mRNA and miRNA levels. RT-qPCR based methods were used for RNA estimation (n=3). For statistical significance, a minimum of three independent experiments were considered in each case, unless otherwise mentioned. Error bars represent the s.e.m. *P<0.05; **P<0.01; ***P<0.0001; ns, not significant (two-tailed, unpaired Student's t-test in B–D,G; two-tailed, paired Student's t-test in I).
How the Aβ1–42 oligomers affect endosome maturation and lead to accumulation of miRNPs and their entrapment in PBs is an interesting question. By microscopic analysis we documented proximal localization of Aβ deposition with EEs and PBs (Fig. S5A–D). We further noted, using STED microscopy, proximal localization and contact between EEs and PBs in Aβ1–42-treated cells (Fig. S5E). To investigate the connection between miRNP entrapment by PBs and excess cytokine production, we depleted PBs by treating C6 cells with specific siRNAs before exposing them to Aβ1–42 oligomer and measuring the cytokine content. Cells pre-treated with siRNAs against Dcp1a and Lsm1, two important PB components, showed a reduction in expression of IL-1β and IL-6 compared to expression in cells transfected with control siRNA (si-Con). This result suggests the importance of miRNP entrapment in PBs for the regulation of cytokine expression in Aβ1–42-treated cells (Fig. 4H,I). In this context, the levels of cellular miRNA and miRNA associated with Ago2 were also reduced, whereas Ago2-associated cytokine mRNA levels increased in cells depleted of PB components (Fig. 4J,K). Our data indicate that PB depletion can potentially decrease the neuroinflammation process by rescuing the miRNP recycling defect in Aβ1–42-treated glioblastoma cells, restoring cytokine levels to those seen in the inactivated cells (Fig. 5).
Amyloid-β exposure causes accumulation of inactive miRNPs, leading to elevated proinflammatory cytokine production in glial cells. Amyloid-β masks the Rab7a–RILP interaction to reduce endosome–lysosome interactions. Accumulated miRNPs fail to be targeted to lysosomes in amyloid-exposed cells due to the loss of endosome–lysosome interactions. Lysosomal compartmentalization of miRNPs is required for miRNP recycling and for repression of de novo targets. Accumulated miRNPs are stored in P-bodies (PB), and depletion of P-bodies (siDcp1a/siLsm1, siRNA-mediated knockdown of PB components) rescues miRNA function in amyloid-exposed glial cells.
Amyloid-β exposure causes accumulation of inactive miRNPs, leading to elevated proinflammatory cytokine production in glial cells. Amyloid-β masks the Rab7a–RILP interaction to reduce endosome–lysosome interactions. Accumulated miRNPs fail to be targeted to lysosomes in amyloid-exposed cells due to the loss of endosome–lysosome interactions. Lysosomal compartmentalization of miRNPs is required for miRNP recycling and for repression of de novo targets. Accumulated miRNPs are stored in P-bodies (PB), and depletion of P-bodies (siDcp1a/siLsm1, siRNA-mediated knockdown of PB components) rescues miRNA function in amyloid-exposed glial cells.
DISCUSSION
The role of different cellular organelles in regulating miRNA-mediated gene expression in mammalian cells is an interesting but seldom explored area. Membrane- and non-membrane-bound cell organelles divide the cell cytoplasm into different compartments, and this compartmentalization is important for various biochemical processes such as transcription, translation and protein degradation, among others. This compartmentalization also plays a crucial role in fine-tuning miRNA activity. Previous reports have highlighted the ER membrane as the site where miRNP nucleation and target mRNA repression take place (Li et al., 2013; Stalder et al., 2013). Although the ER membrane is believed to be the site where target mRNA repression occurs, our previously published work indicates that LEs are the site of mRNA degradation (Bose et al., 2017). In another work it has been highlighted how miRNA-free Ago2 accumulation affects MVB to ER translocation and miRNP biogenesis (Bose et al., 2020).
Several neurodegenerative diseases, like AD, Parkinson's disease and ALS, are prominently linked with EE dysfunction. Enlargement of Rab5-positive vesicles has been shown to be one of the early symptoms in AD patients (Cataldo et al., 2000). In ALS, problems in GDP–GTP exchange causes hyperactivation of Rab5 leading to EE accumulation (Otomo et al., 2003; Lai et al., 2009). How does EE dysfunction affect this wide variety of diseases? It is possible that EE dysfunction also affects downstream LE maturation and LE–lysosome fusion to have wider implications in health and diseases. The lysosome is one of the most important organelles, removing cellular waste and aberrant proteins to control protein homeostasis in eukaryotic cells. Niemann–Pick disease type C is a pathologic condition involving accumulation of different lipid molecules, including sphingolipids and cholesterol, in the lysosomal lumen (Butler et al., 1993; Vance, 2006). Extensive works have studied the role of autophagy in protein homeostasis and organelle turnover in neurodegenerative diseases. But one key question has remained unanswered: how does EE dysfunction or a problem in endo-lysosomal fusion trigger the innate immune system? In this study we have tried to explain how impaired endo-lysosomal fusion leads to accumulation of miRNPs within EEs and PBs, which fail to recognize the newly formed target mRNA. This mechanism causes a global decrease in miRNA activity and leads to an inability to repress cytokine mRNA in glial cells. Here, we have shown the importance of cellular organelles that affect the miRNA recycling process, suggesting a mechanism by which cellular structures fine-tune the delicate gene expression changes during pathological events.
Do Aβ1–42 oligomers have any low-complexity sequence or any short linear motifs required for liquid–liquid phase separation? This has not been explored. PBs are an example of a separated phase within the cytoplasm, where RNA-binding proteins along with repressed mRNA can be stored (Filipowicz et al., 2008). Aβ1–42 may play a role in increasing local concentration of RNA and RNA-binding proteins in the cytoplasm to form the PBs that we found frequently in glioblastoma cells exposed to Aβ1–42. Researchers have made similar observations in Huntington's disease (HD), where Ago2 is found in stress granules (SGs), which may account for its inactivation (Pircs et al., 2018). Past investigations have also revealed a strong correlation between RNP hyperaggregation and different neurodegenerative diseases like ALS or HD (Savas et al., 2008; Chen-Plotkin et al., 2010), but whether and how these dynamic RNP granules add to the pathogenesis of these remains an interesting topic to understand. Finally, another important aspect is whether these phase-separated RNP droplets have other biological roles to play rather than serving as RNA storage sites. As these droplets are very dynamic in nature and constantly exchange cargo between the soluble cytosol and the insoluble granules, they might have a role to play in different rapid physiological changes, such as cell signaling or ion exchange processes.
How much of the cytokine mRNA upregulation is caused by increased transcription activity and how much by the suggested impaired miRNA repression is an interesting question. We have treated cells with Aβ1–42 in the presence and absence of α-amanitin, a known inhibitor of RNA polymerase II and III. Subsequent analysis revealed that α-amanitin exposure during amyloid treatment can prevent the cytokine mRNA surge (D.D., unpublished data). Therefore, a transcriptional surge of cytokine mRNA also occurs upon amyloid exposure. However, lack of miRNA repression also contributes to the accumulation of cytokine mRNA. Even after the transcriptional surge, the newly synthesized mRNAs avoid existing miRNA-mediated repression and degradation. This is achieved by preventing miRNP trafficking to newly synthesized target mRNAs located on rough-ER-attached polysomes. The PB depletion experiments shown in Fig. 4 suggest a full suppression of Aβ1–42-induced expression of cytokines when miRNPs are functional to repress their targets after miRNP reactivation due to PB depletion. Therefore, it may be assumed that Aβ1–42 not only activates the transcription of the cytokine mRNAs, but also ensures their stability and translation by inactivating the miRNPs to stop the degradation of the cytokine mRNAs.
Inflammation is associated with AD, and this mainly involves two of the most important cells, the astrocytes and microglia. Despite the population of astrocytes exceeding that of microglia, the two cell types coordinate in secreting various complementary factors and chemokines during neurodegeneration. Furthermore, in post-mortem AD brain, reactive gliosis is common in the region of amyloid plaques (Vehmas et al., 2003). Our findings have identified the mechanistic aspect of an endosome maturation defect and its connection to miRNP recycling and de novo target mRNA–miRNP interactions. Our data obtained with HEK293 and C6 cells suggest that a similar kind of miRNP dysfunction might also be observed in microglia. This is an interesting question that needs attention before nurturing the idea of targeting the miRNP inactivation process for therapeutic intervention to curtail neuroinflammation in AD.
Another key aspect to consider is the fate of the accumulated miRNPs. We do know that some of these miRNPs translocate to PBs. This increase in PBs might initiate pathogenesis by inducing some signaling process or by regulating mRNA repression. Another fraction of the accumulated miRNAs, along with miRNPs, can be packaged into extracellular vesicles (EVs) or exosomes, and can be exported out of cells (D.D., unpublished data). These extracellular miRNAs could act as an exocrine signal as they are taken up by recipient glial cells and neurons. The transfer of miRNAs and miRNPs could thus serve as a signal for disease initiation before these cells are exposed to amyloid-β. As such, this process could serve as a protection for the cells that are not exposed to pathogenic proteins. The presence of exosomes packed with miRNAs in cerebrospinal fluid or in blood could be used as biomarkers for early disease detection as well. It is also noteworthy that HuR protein (also known as ELAVL1), which can reverse miRNA-mediated repression (Bhattacharyya et al., 2006), is also responsible for EV-mediated transfer of miRNAs (Mukherjee et al., 2016), and HuR could be targeted to develop a potential therapeutic strategy against these diseases.
MATERIALS AND METHODS
Cell culture and transfection
Both C6 glioblastoma and HEK293 cells were cultured in high glucose DMEM medium (Life Technologies) containing 2 mM L-glutamine and 10% heat inactivated FCS (Gibco). All the plasmids and siRNAs were transfected using Lipofectamine 2000 and RNAiMax (Life Technologies), respectively, following the manufacturer's protocol. FH–Ago2 and GFP–Ago2 plasmids were kind gifts from Tom Tuschl of Rockefeller University, NY, USA, and Witold Filipowicz of FMI, Basel, Switzerland, respectively. All the SMARTpool ONTARGETplus siRNAs were bought from Dharmacon (siCon, Dharmacon D-001810-10-20; human siRILP, Dharmacon L-008787-01-0005; rat siDcp1a, Dharmacon L-088099-02-0005; rat siLsm1, Dharmacon L-089369-02-0005).
TET-ON HEK293 cells were used to carry out experiments using inducible constructs. TET-ON HEK293 cells were cultured in DMEM supplemented with 10% TET-approved FCS (Clontech). Specific genes were induced using 300 ng/ml doxycycline (Sigma) for the desired time points.
Preparation of Aβ1-42 oligomers
HPLC-purified lyophilized Aβ1–42 (American Peptide) was reconstituted in 100% 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP). HFIP was removed by evaporation in a SpeedVac, and the Aβ1–42 was resuspended in 5 mM DMSO. This stock was diluted with phosphate-buffered saline (PBS) to 400 mM, and SDS was added to a final concentration of 0.2%. The resulting solution was incubated overnight and diluted again with PBS to 100 mM, followed by incubation at 37°C for 18–24 h before use.
RNA isolation and real-time PCR
Total RNA was isolated using TRIzol reagent (Life Technologies) following the manufacturer's instructions. cDNA was prepared, taking 50 ng and 200 ng of total RNA for miRNA and mRNA, respectively. For real-time analysis of specific mRNAs, cDNA was prepared using random nonamer (Eurogentec reverse transcriptase core kit) followed by real-time PCR using Mesa Green quantitative PCR (qPCR) master mix plus (Eurogentec), following the manufacturer's protocol. For quantification of specific miRNAs, a Taqman reverse transcription kit (Applied Biosystems) was used for cDNA preparation, followed by real-time PCR using TaqMan universal PCR mix (Applied Biosystems). qPCR was done using specific Taqman-based miRNA primers (Table S3) following the manufacturer’s instructions. All PCR reactions were done in a 7500 Applied Biosystems real-time system or a Bio-Rad CFX96 real-time system. 18 s rRNA and GAPDH mRNA levels were used as a loading control for mRNA quantification, whereas U6 snRNA was used as loading control for miRNA quantification. Primers for reverse transcription-qPCR (RT-qPCR) are listed in Table S2.
Luciferase assay
Renilla luciferase (RL) and firefly luciferase (FF) activities were measured using a dual luciferase assay kit (Promega). Cells were transfected with 20 ng of control reporter (RL–con), let-7a reporter (RL–3×bulge-let-7a) or miR-122 reporter (RL–3×bulge-miR122) along with 200 ng of firefly luciferase. Cells were lysed with 1× passive lysis buffer (Promega) after 48 h of transfection, and luciferase activity was measured on a VICTOR X3 Plate Reader following the manufacturer’s protocol. FF normalized RL values were used to score miRNA repression level. Specific activity of miRNAs was calculated by normalizing miRNA repression level with total miRNA level. Details of plasmids are provided in Table S4.
Immunoprecipitation
For immunoprecipitation (IP) cells were lysed with lysis buffer [20 mM Tris-HCl pH 7.5, 150 mM KCl, 5 mM MgCl2 and 1 mM dithiothreitol (DTT)] containing 0.5% Triton X-100, 0.5% sodium deoxycholate and 1× EDTA-free protease inhibitor cocktail (Roche) for 30 min at 4°C followed by three 10 s pulses of sonication. Lysate was cleared at 16,000 g for 10 min. Protein G–agarose beads were blocked in lysis buffer containing 5% BSA for 1 h then incubated with the required antibody (final dilution 1:100; antibody details provided in Table S1) for 3 h at 4°C. Cell lysate was incubated with the antibody-attached beads overnight at 4°C. Beads were washed thrice with 1× IP buffer (20 mM Tris-HCl pH 7.5, 150 mM KCl, 5 mM MgCl2 and 1 mM DTT) and separated into two halves for RNA and protein estimation.
Optiprep density gradient centrifugation
For subcellular organelle fractionation, a 3–30% or 3–15% continuous gradient was prepared using Optiprep (Sigma-Aldrich, USA) in a buffer constituting 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2,10 mM EGTA and 50 mM HEPES (pH 7.0). Cells were rinsed with ice cold PBS, and a Dounce homogenizer was used to homogenize the cells in a buffer containing 0.25 M sucrose, 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES (pH 7.0) supplemented with 100 μg/ml cycloheximide, 5 mM vanadyl ribonucleoside complex (VRC; Sigma Aldrich), 0.5 mM DTT and 1× protease inhibitor. The lysate was subjected to centrifugation twice at 1000 g for 5 min for clarification and placed on top of the prepared gradient. Ultracentrifugation was performed for 5 h at 133,000 g on an SW60 (Beckman Coulter) rotor to separate each gradient. Ten fractions were collected by aspiration and further analyzed for RNA and proteins.
Polysome isolation
In order to isolate total polysomes, buffer constituting 10 mM HEPES pH 8.0, 25 mM KCl, 5 mM MgCl2, 1 mM DTT, 5 mM vanadyl ribonucleoside complex, 1% Triton X-100, 1% sodium deoxycholate and 1× EDTA-free protease inhibitor cocktail (Roche) supplemented with cycloheximide (100 μg/ml; Calbiochem) was used to lyse the cells. The clearance of lysate was performed at 3000 g for 10 min followed by a pre-clearance shift at 20,000 g for 10 min at 4°C. The loading of the clear lysate was done on a 30% sucrose cushion and ultracentrifuged at 100,000 g for 1 h at 4°C. The washing of the sucrose cushion was done with a buffer (10 mM HEPES pH 8.0, 25 mM KCl, 5 mM MgCl2 and 1 mM DTT), followed by ultracentrifugation for an extra 30 min with final resuspension of the pellet in polysome buffer (10 mM HEPES pH 8.0, 25 mM KCl, 5 mM MgCl2, 1 mM DTT, 5 mM vanadyl ribonucleoside complex and 1× EDTA-free protease inhibitor cocktail) for further isolation of RNA and protein.
Immunoblotting
Protein samples from whole cell lysate, immunoprecipitated proteins or proteins from cell fractionation were subjected to SDS–PAGE analysis. Western blotting was done on a PVDF membrane overnight at 4°C. Membranes were blocked with 3% BSA for 1 h then probed with specific antibodies (Table S1). Images of western blots were taken with a UVP BioImager 600 system equipped with VisionWorks Life Science software, version 6.80 (UVP). Band intensities were calculated using ImageJ software (NIH, Bethesda, MD).
Immunofluorescence
Cells were grown on 18 mm round coverslips and transfected as described above. Cells were fixed with 4% paraformaldehyde for 20 min at room temperature in the dark. For indirect immunofluorescence, fixed cells were blocked and permeabilized with PBS containing 3% BSA and 0.1% Triton X-100 for 30 min at room temperature. Coverslips were probed with specific antibodies (Table S1) for 16 h at 4°C. Coverslips were washed thrice with 1× PBS and then probed with specific Alexa Fluor-conjugated secondary antibody tagged with fluorochrome (dilution 1:500) for 1 h at room temperature. Lysosomes were labeled using LysoTracker Red DND-99 (Thermo Fisher Scientific).
Confocal imaging and post-capture image analysis
Confocal fixed-cell images were taken using a Zeiss LSM800 confocal microscope with a Plan-Apo 63×/1.4NA oil immersion objective (Zeiss) and analyzed using Imaris7 (Oxford Instruments) and ImageJ software. Pearson's coefficient of colocalization was calculated using the Coloc plug-in of Imaris7 software. 3D reconstructions of specific bodies were performed using the Surpass plug-in of Imaris7 software. Numbers of individual bodies or vesicles were measured using particle generator of the Surpass plug-in.
Live-cell imaging and endosome dynamics
For live-cell microscopy, cells were transiently transfected with YFP–Endo, GFP–Rab7 (kind gift from Edouard Bertrand, IGMM, Montpellier, France) or pDsRed2–ER (ER–DsRed; Clontech). Imaging was performed 48 h after transfection using a Leica DMI6000 B inverted microscope equipped with Plan Apo100×/1.40 oil objective (Leica TCS SP8 confocal system). Endosome dynamics were calculated after assigning each vesicle as one particle using Imaris 7 Surpass plug-in. Endosome track speed and track length were calculated by applying a particle tracking algorithm and gap close algorithm. Details of plasmids are provided in Table S4.
In vitro organellar interaction
For in vitro EE–ER interaction studies, C6 cells were lysed using a Dounce homogenizer in a buffer containing 0.25 M sucrose, 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES pH 7.0 supplemented with 100 μg/ml cycloheximide, 5 mM vanadyl ribonucleoside complex (Sigma Aldrich), 0.5 mM DTT and 1× protease inhibitor. Cell lysate was loaded on top of a 3–30% Optiprep gradient, as described above, and ultracentrifuged at 133,000 g for 5 h. Fraction numbers 2–4, which were enriched in endosomes, and 7–9, which were enriched in ER, were further ultracentrifuged for 2 h at a speed of 217,000 g and 133,000 g to isolate endosome and ER, respectively. Endosomes were suspended in a buffer containing 0.25 M sucrose, 78 mM KCl, 4 mM MgCl2, 8.4 mM CaCl2, 10 mM EGTA and 50 mM HEPES pH 7.0 supplemented with 1× protease inhibitor along with 2.5 μM Aβ1–42 and kept for 1 h at 37°C. The ER fraction was added after 1 h, along with 1 mM ATP, and the in vitro interaction was carried out for 1 h at 37°C. EEs were isolated by immunoprecipitating EEs with Protein G–agarose beads tagged with EEA1 antibody, and western blotting was performed to check the amount of calnexin interacting with EEA1.
Statistical analysis
All graphs and statistical analyses were done using Graphpad prism 5.0 (GraphPad, San Diego, CA, USA). Student's t-test was done to determine P values. P<0.05 was considered as significant. All the experiments were done at least three times. Error bars indicate mean±s.e.m.
Acknowledgements
We acknowledge Witold Filipowicz, Gunter Meister, Edouard Bertrand, Tom Tuschl and J. M. Backer for different plasmid constructs. Subhas Biswas and Nakul Maiti helped us with primary cells and reagents.
Footnotes
Author contributions
Conceptualization: D.D., S.N.B.; Methodology: D.D., S.N.B.; Validation: D.D.; Formal analysis: D.D., S.N.B.; Investigation: D.D., S.N.B.; Resources: S.N.B.; Data curation: D.D., S.N.B.; Writing - original draft: S.N.B.; Visualization: S.N.B.; Supervision: S.N.B.; Project administration: S.N.B.; Funding acquisition: S.N.B.
Funding
S.N.B. is supported by The Swarnajayanti Fellowship (DST/SJF/LSA-03/2014-15) from the Department of Science and Technology, Ministry of Science and Technology, India. D.D. received support from the Council of Scientific and Industrial Research, India. The work also received support from a High Risk High Reward Grant (HRR/2016/000093) awarded by the Department of Science and Technology, Ministry of Science and Technology, India, and by Indo-French Centre for the Promotion of Advanced Research (CEFIPRA) project grant 6003-J.
Peer review history
The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258360
References
Competing interests
The authors declare no competing or financial interests.