Opposing views have been proposed regarding the role of tau, the principal microtubule-associated protein in axons. On the one hand, tau forms cross-bridges at the interface between microtubules and induces microtubule bundling in neurons. On the other hand, tau is also considered a polymer brush which efficiently separates microtubules. In mature axons, microtubules are indeed arranged in parallel arrays and are well separated from each other. To reconcile these views, we developed a mechanistic model based on in vitro and cellular approaches combined to analytical and numerical analyses. The results indicate that tau forms long-range cross-bridges between microtubules under macromolecular crowding conditions. Tau cross-bridges prevent the redistribution of tau away from the interface between microtubules, which would have occurred in the polymer brush model. Consequently, the short-range attractive force between microtubules induced by macromolecular crowding is avoided and thus microtubules remain well separated from each other. Interestingly, in this unified model, tau diffusion on microtubules enables to keep microtubules evenly distributed in axonal sections at low tau levels.
RNA-protein interactions (RPIs) are promising targets for developing new molecules of therapeutic interest. Nevertheless, challenges arise from the lack of methods and feedback between computational and experimental techniques during the drug discovery process. Here, we tackle these challenges by developing a drug screening approach that integrates chemical, structural and cellular data from both advanced computational techniques and a method to score RPIs in cells for the development of small RPI inhibitors; and we demonstrate its robustness by targeting Y-box binding protein 1 (YB-1), a messenger RNA-binding protein involved in cancer progression and resistance to chemotherapy. This approach led to the identification of 22 hits validated by molecular dynamics (MD) simulations and nuclear magnetic resonance (NMR) spectroscopy of which 11 were found to significantly interfere with the binding of messenger RNA (mRNA) to YB-1 in cells. One of our leads is an FDA-approved poly(ADP-ribose) polymerase 1 (PARP-1) inhibitor. This work shows the potential of our integrative approach and paves the way for the rational development of RPI inhibitors.
Deletion of murine Smn exon 7, the most frequent mutation found in spinal muscular atrophy, has been directed to either both satellite cells, the muscle progenitor cells and fused myotubes, or fused myotubes only. When satellite cells were mutated, mutant mice develop severe myopathic process, progressive motor paralysis, and early death at 1 mo of age (severe mutant). Impaired muscle regeneration of severe mutants correlated with defect of myogenic precursor cells both in vitro and in vivo. In contrast, when satellite cells remained intact, mutant mice develop similar myopathic process but exhibit mild phenotype with median survival of 8 mo and motor performance similar to that of controls (mild mutant). High proportion of regenerating myofibers expressing SMN was observed in mild mutants compensating for progressive loss of mature myofibers within the first 6 mo of age. Then, in spite of normal contractile properties of myofibers, mild mutants develop reduction of muscle force and mass. Progressive decline of muscle regeneration process was no more able to counterbalance muscle degeneration leading to dramatic loss of myofibers. These data indicate that intact satellite cells remarkably improve the survival and motor performance of mutant mice suffering from chronic myopathy, and suggest a limited potential of satellite cells to regenerate skeletal muscle.
Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Appendix 1 Appendix 2 Appendix 3 Appendix 4 Appendix 5 Data availability References Decision letter Author response Article and author information Metrics Abstract RNA-protein interactions (RPIs) are promising targets for developing new molecules of therapeutic interest. Nevertheless, challenges arise from the lack of methods and feedback between computational and experimental techniques during the drug discovery process. Here, we tackle these challenges by developing a drug screening approach that integrates chemical, structural and cellular data from both advanced computational techniques and a method to score RPIs in cells for the development of small RPI inhibitors; and we demonstrate its robustness by targeting Y-box binding protein 1 (YB-1), a messenger RNA-binding protein involved in cancer progression and resistance to chemotherapy. This approach led to the identification of 22 hits validated by molecular dynamics (MD) simulations and nuclear magnetic resonance (NMR) spectroscopy of which 11 were found to significantly interfere with the binding of messenger RNA (mRNA) to YB-1 in cells. One of our leads is an FDA-approved poly(ADP-ribose) polymerase 1 (PARP-1) inhibitor. This work shows the potential of our integrative approach and paves the way for the rational development of RPI inhibitors. Editor's evaluation A novel approach is introduced to modulate and/or inhibit protein-RNA interactions, based upon integration of computational techniques with cellular assays. https://doi.org/10.7554/eLife.80387.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Targeting RNA:protein interactions (RPIs) critically involved in pathological mechanisms is a promising strategy to find novel classes of drug candidates that remains largely unexploited (Einstein et al., 2021). RPIs in cells are highly diverse encompassing interactions with messenger RNA (mRNA; Baltz et al., 2012), ribosomal RNA (rRNA; Simsek et al., 2017), and non-coding RNA (ncRNA; Lu et al., 2019), which are critical to fine tune the spatiotemporal gene expression. As revealed by genomic approaches (Van Nostrand et al., 2020; Castello et al., 2012), the human genome contains more than 1000 transcripts encoding RNA-binding proteins (RBPs), thus providing a large variety of interactions with coding or non-coding RNAs. However, while the diversity of RNA:Protein interfaces may allow the development of RPIs inhibitory molecules (Wu, 2020), only scarce studies have already been undertaken and were restricted to few complexes such as LIN28/let-7 (Roos et al., 2016; Wang et al., 2018), MUSASHI (MSI)/RNA (Minuesa et al., 2019) and heterogeneous nuclear ribonucleoprotein A18 (hnRNP A18)/RNA (Solano-Gonzalez et al., 2021). Several challenges arise from the drug discovery process such as finding a druggable pocket in RNA-binding interfaces (Minuesa et al., 2019), the quality of the computational models, the strategies used in the in silico screening, and the lack of experimental feedback and validation of computationally predicted inhibitors essential to orient the rational drug design procedure toward the most relevant molecules. Besides the above-listed issues, new experimental assays must be developed to screen molecules targeting RPIs which ideally would work in a cellular context and be amenable to high content screening (HCS) (Mattiazzi Usaj et al., 2016; Julio and Backus, 2021). Indeed, to find potent inhibitors of RNA:protein interfaces, previous approaches used in vitro assays such as fluorescence polarization assay complemented by pull-down experiments with cell lysates or RNA enzyme-linked immunosorbent assay (ELISA) to test the effectiveness or selectivity of few hits (Roos et al., 2016; Minuesa et al., 2019). While in vitro approaches are important to define putative hits and lead to the validation of effective compounds, deciphering whether the selected molecules are effective in a cellular context generally relies on indirect measurements using techniques such as cellular engagement thermal shift assay (CETSA) or functional assays where the putative consequences of disrupting RPIs on cellular function bear a considerable uncertainty. Indeed, multiple functions are associated to RBPs, which renders the interpretation of the results of functional assays tricky. In addition, toxicity and off-target effects are putative biases which are always difficult to get rid of, notably when using small molecules with a Kd in the low micromolar range, which is generally the case for RPI inhibitors. To fill the gap between in vitro and functional assays, cellular approaches initially used to detect protein:protein interactions (PPIs) such as fluorescence resonance energy transfer (FRET) or proximity ligation assay (PLA) have been adapted to detect RPIs (Jung et al., 2013; Camborde et al., 2017) in cells but several technical issues have hampered their application such as the requirement of an adapter to RNA in FRET and PLA, the proximity of the donor and acceptor proteins in FRET, and the use of antibodies in PLA. The aim of this paper is to tackle these challenges by introducing an experimental assay amenable to HCS to score RPIs in cells and a drug screening approach that integrates chemical, structural, and cellular data from both advanced computational and experimental techniques for the development of small molecules that target RPIs. As an application model we chose to target Y-Box binding protein 1 (YB-1) of the YBX1 gene. As other abundant nucleic acid binding proteins, YB-1 participates in many DNA/RNA-dependent processes such as mRNA translation, splicing, transcription, long ncRNA (lncRNA) functions, and DNA repair (Lyabin et al., 2014). However, YB-1 is mostly a core component of untranslated messenger ribonucleoprotein particles (mRNPs) in the cytoplasm (Singh et al., 2015) which, according to crosslinking immunoprecipitation coupled to sequencing (CLIP) analysis (Wu et al., 2015), preferentially binds coding sequences and 3’-UTRs across most transcripts with a weak specificity. Since YB-1 binds to and regulates the activation of dormant mRNAs (Budkina et al., 2021) which are particularly enriched in gene controlling transcription (Roos et al., 2016), YB-1 is possibly involved in cellular decisions; and consistently, YB-1 was recently identified as one of the few key genes that control gene expression plasticity in rats subjected to caloric restriction (Ma et al., 2020). Interestingly, YBX1 is also one of the genes whose gene-protein expression is the most correlated in cancers vs. normal tissues (Kosti et al., 2016), and YBX1 was identified among the few genes in a clustered regularly interspaced short palindromic repeats (CRISPR) screen showing the highest sensitivities with broad proteome co-expression in cancer cell lines (Nusinow et al., 2020, Figure S4 of this reference), pointing toward a possible role for YBX1 in cancer. The involvement of YB-1 in the progression and resistance to stress and chemotherapy (Kang et al., 2013; Yang et al., 2010; El-Naggar et al., 2019), notably after its translocation in the nucleus in certain cancers (Bargou et al., 1997), has also been documented. Together, these data make YB-1 a relevant target for cancer treatment (Lasham et al., 2013) and a subject of ongoing research to identify YB-1 inhibitors (Khan et al., 2014; Tailor et al., 2021). Moreover, YB-1 is one of the host proteins implicated in viral replication of human immunodeficiency virus (HIV) (Jung et al., 2018; Poudyal et al., 2019) and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (Schmidt et al., 2021) and hence targeting it along with targeting specific viral proteins can help reduce viral replication to a higher extent than just targeting the viral proteins. Our choice in targeting YB-1 was also guided by the availability of structural data on RNA:YB-1 complexes to probe in vitro whether small molecules can interact with the cold-shock domain (CSD) of YB-1 (Kretov et al., 2019; Yang et al., 2019). We started this work by addressing the drug screening challenge and developing an integrative approach that uses in synergy advanced computational and experimental techniques in a concerted manner (as illustrated in Figure 1). Based on our discovery of a druggable pocket by molecular dynamics simulations (MD) located on the outside surface of the CSD β-barrel (which is also part of the RNA binding interface Yang et al., 2019), we implemented a large-scale computational approach that balances accuracy and computational cost to virtually screen potent compounds from small molecule libraries containing more than 7 million molecules. Next, we addressed the abovementioned lack of methods able to score RPIs in a cellular context. To this end, we adapted the microtubule bench (MT bench) assay to score protein interactions with endogenous mRNAs in cells and implemented a robust HCS-based detection scheme. The MT bench was first introduced in 2015 to probe PPIs in cells with conventional fluorescence microscope by using microtubules as intracellular nanoplatforms (Boca et al., 2015; Rengifo-Gonzalez et al., 2021). Figure 1 with 2 supplements see all Download asset Open asset Schematic representation of an integrative approach for screening RPI inhibitors. This approach combines information from three data sources: computational (in silico, top left), cellular (in vivo, bottom), and structural (in vitro, top right). Blue arrows indicate the data flow. In silico: Starting by a large-scale computational approach that uses Docking (static approach), Molecular Dynamics and Free Energy Simulations (dynamic approach), using a computational model to virtually screen large libraries of small molecules (here, Molport and FDA-approved drugs) with the prior knowledge of a validated pocket and where several filters are used to reduce the selection to the most pertinent ligands that are then proposed as hits to be tested experimentally. Filters are represented as funnels. In vivo: In cell validation of the efficiency of the proposed hits in blocking RPIs with the MT Bench assay. This technology can quantify RPIs at the single cell level by using microtubule filaments as intracellular nanoplatforms (lower left inset, the bait, here YB-1, is shown in cyan and mRNA in orange). Lower right inset: enlarged view on mRNAs (orange) brought on microtubules using YB-1 as bait (cartoon representation: YB-1 in dark cyan with a violet surface mesh is complexed with RNA (orange ribbon)). In vitro: Experimental validation of binding the target pocket using solution NMR spectroscopy. A zoom in on pocket residue signals in a 2D 1H-15N-SOFAST-HMQC of YB-1 alone (black) and in the presence of Quercetin F1 (magenta). The top right 3D structure shows the binding of Quercetin (green stick) to YB-1 (cartoon representation in cyan combined with a transparent surface). Residues showing chemical shifts upon F1 binding are colored in magenta and depict what we identified as the Quercetin-pocket. The results presented here, show that the physics-based in silico approach allowed the identification of 22 potential hits that we subsequently tested in vitro by nuclear magnetic resonance (NMR) spectroscopy and in cells using the adapted MT bench assay by scoring the interaction of YB-1 with mRNA in the cytoplasm. Of these 22 potential YB-1 inhibitors, 15 compounds were found to bind YB-1 in vitro and 11 of them were found to efficiently interfere with the interaction of YB-1 with mRNA in cells at low micromolar concentrations; and with a notable specificity when compared with two other RBPs, Human antigen R (HuR) and fused in sarcoma (FUS). The potency of the selected compounds was further demonstrated by in depth MD and NMR analyses. The results also validate that the MT bench allows to robustly and automatically score RBP-specific interactions with endogenous mRNAs by using high-resolution HCS imagers. Interestingly, compound P1, an FDA-approved poly(ADP-ribose) polymerase 1 (PARP-1) inhibitor (Zandarashvili et al., 2020), was found to interact with YB-1 with higher selectivity compared to the other hits. Whether P1 interferes with YB-1 cellular functions in cells therefore merits further investigations. Together, these results demonstrate the validity of our integrative approach and the efficacy of the MT bench assay that critically complements computational and structural approaches to identify compounds targeting RPIs in cells. Results A druggable pocket found in YB-1 CSD, a conserved RNA-binding domain The first challenge was to find a druggable pocket in the structured cold-shock domain of YB-1 located at the RNA-binding interface. We started by taking into consideration small molecules that were reported to target YB-1 in the literature. The only molecule for which a structural validation was available, though only in silico, is the flavonoid, Fisetin (Khan et al., 2014). In this paper, using refined docking, Fisetin was found to inhibit YB-1 activation by Akt-mediated phosphorylation at S102 with a binding pocket located inside the β-barrel structure of YB-1 CSD (51–129 aa). Having in hand the longest YB-1 fragment (1–180 aa) amenable to NMR spectroscopy (Kretov et al., 2019), we then analyzed the interaction between Fisetin and YB-1 fragment in vitro. Significant chemical shift perturbations (CSPs) were indeed observed but not within the previously predicted pocket (Khan et al., 2014). The observed CSPs implicated residues located in a hydrophobic pocket on the outside surface of the β-barrel; these are W65, V84, F85, V86, G116, K118, G119, and A120 (pocket residues shown on the top right of Figure 1). Quercetin, a Fisetin analog with an additional hydroxyl group capable of forming new H-bond interactions with YB-1, was also tested. Since it showed higher CSPs with the same pocket, compared to Fisetin, we decided to subsequently name it the ‘Quercetin-pocket’ (average CSP of 0.032 for Quercetin (F1) compared to 0.028 for Fisetin (F4)). To delineate the characteristics of the Quercetin-pocket, we used extensive MD simulations of YB-1 CSD either in its unbound or RNA-bound form (Figure 1—figure supplement 1 and Figure 1—figure supplement 1, respectively; detailed MD analysis can be found in Appendix 1). Results show that the Quercetin-pocket in its unbound form presents an open and a closed state. This pocket is located at the third β-hairpin and is monitored by K118 and F85 side chains. The opening mechanism is controlled by an electrostatic cation-π interaction formed between the cationic side chain of K118 (NH3+) and the π-electron ring system of F85 (Figure 1—figure supplement 1). The sampled structures of both open and closed states of CSD were also captured by NMR in the published 3D solution structure of Kloks et al., 2002 which is consistent with our findings. MD and NMR analysis of YB-1 in complex with 5-nt long poly(C) RNA (C5) show that some of the CSD key residues implicated in RNA binding are located in the Quercetin-pocket; these residues include W65, Y72, F74, F85, H87, K118, and E121 (Figure 1—figure supplement 1). These residues are evolutionary conserved as shown by the ConSurf (Ben Chorin et al., 2020; Goldenberg et al., 2009) analysis reported in Appendix 1-section III and illustrated in Appendix 1—figure 1. Together, MD and NMR analysis evidence the validity of the Quercetin-pocket as a potential target for the development of small molecules interfering with RNA:YB-1 interactions. Prediction of potent inhibitors of mRNA:YB-1 interactions using a large scale computational approach Having identified a druggable pocket at the RNA:YB-1(CSD) interface, we next sought to target it pharmacologically. Therefore, based on these atomistic and structural data, we implemented a large-scale computational strategy to propose putative inhibitors of RNA:YB-1 interactions. This approach is illustrated in Figure 1 and detailed in the Computational Methods section and in Appendix 2. We started by using a pharmacophore approach to virtually screen a database composed of 208 million pharmacophores representing the conformers of around 7.3 million distinct commercially available molecules from MolPort: (i) a ‘pocket-’ased” pharmacophore screening built from the prediction of a pseudo-ligand in the binding site of the MD refined structure of the open-state pocket and (ii) a distinct ‘ligand-based’ pharmacophore built on the 3D structure of Quercetin (F1) with YB-1. The 3D structure of the YB-1:F1 complex was obtained by docking followed by refinement MD simulations and the binding site was confirmed by NMR spectroscopy (Figure 1). 249 and 407 distinct molecules were selected from the ‘pocket-based’ and the ‘ligand-based’ screening, respectively. Next these molecules were reduced to a final selection by predicting ADME-T (absorption, distribution, metabolism, excretion, and toxicity) endpoints and using computed molecular docking in the Quercetin-pocket (details in Appendix 2-section I). At the end, 111 molecules were retained from this static virtual screen after visual inspection and rational selection of structurally promising candidates. In a second step, we applied physico-chemical filters to keep only molecules belonging to a drug-like chemical space (molecular weight, number of rotational bonds, number of proton donors and acceptors, lipophilicity and solubility). Purchasability filters were also applied based on availability, purity and price in order to facilitate and optimize the conditions for the in vitro and in vivo assays. From the 7.3 million MolPort molecules, 78 molecules were finally retained. In parallel, we executed an automated blind docking of 4700 FDA-approved drugs (Drugs-lib library Lagarde et al., 2018) using the MTiOpenScreen web server (Labbé et al., 2015) which lead to the selection of 62 molecules that may target the Quercetin-pocket and may be suitable for a repositioning strategy (details in Appendix 2-sections II and III). In the last step, the pre-selected molecules using the above static approach, 140 in total (62 FDA-approved and 78 molPort molecules), were subject to a statistical mechanics-based filter that relies on MD and free energy simulations (dynamic approach). First, the docked poses were chosen after visual inspection of the docking results (Fischer et al., 2021). Second, short 10 ns MD simulations were run, in the presence of explicit water molecules, in order to refine the poses and check the stability of the ligands in the targeted pocket. Only ligands that stayed in the pocket during the short MD were retained for the next step (87 out of the 140), where a weighted score (S) based on two observables that describes the ability of the ligand to bind and reside in the pocket was derived (this is detailed in the computational methods section). Ligands with a positive S were considered as hits, and ligands with S<0 were only considered as ‘possible’ if S becomes positive when we take into account the statistical error. From the 87 molecules tested, only 26 potential hits were retained (of which 6 ‘possible’). Finally, absolute free energy simulations (ABFE) were used to compute the protein-ligand binding free energies (ΔGbind) and rank the ligands in terms of affinity (in kcal.mol– 1). ABFE simulations were done using the all-atom point charge CHARMM force field (MacKerell et al., 1998) and BAR (Bennett, 1976) for ΔG estimation. Here potential hits were selected for having a ΔG value gt5.50 kcal/mol. However, the 6 ‘possible’ potential hits evaluated using S were considered as hits if they have a ΔG>6.5 kcal/mol (this is the case of F3: low S (6.15) and high ΔGbind (–10.82 kcal/mol); C11 and C12 represent a similar case). The selection of the hits at the end took into account both evaluation methods (S and ΔGbind) and their corresponding selection criteria. For example: A3 that was not considered a hit by S, was considered a "possible" potential hit due to its high ΔG. Based on these criteria, 22 potential inhibitors were selected to be tested in vitro and in cells where their efficiency to inhibit mRNA:YB-1 interactions can be measured. To this list, 18 molecules, predicted inefficient, were also added as negative controls (CTRL) in order to have a total number of 40 molecules which is convenient for the experimental assays. However, these 18 molecules were rationally selected from the 87 molecules that stayed in the pocket and for which we have calculated and applied the statistical mechanics-based filter described above and computed their ΔGbind. The selection criteria was based on their structural similarity to F1 (hit validated by NMR spectroscopy) in order to generate an initial QSAR that will help us rationally optimize these molecules later. As for the FDA-approved drugs, we chose all PARP inhibitors, in order to compare with P1; the other 2 non-PARP inhibitors (D2 and D3) were chosen for their scaffold. Figure 2 and Figure 2—source data 1 show the classification and chemical structures of these selected 40 molecules along with their resulting scores and free energy values. Figure 2 Download asset Open asset Chemical structures of the selected 40 molecules separated into 22 potential hits and 18 CTRL. Colored dashed boxes assemble hits by subclass and the black dashed box regroups the 18 CTRL. Labels and colored boxes are color coded as function of the family classification: Flavonoids (red) are divided into Flavonols (light red), Flavones (dark red) and Chalcones (green); Flavonoid Analogs in orange, FDA-approved drugs (blue) of which PARP-1 inhibitors (cyan). Figure 2—source data 1 Classification of the 40 molecules selected using the computational approach. The scoring function S and the free energy of binding ΔGbind from ABFE simulations used to identify potential hits are reported, along with the average pocket from NMR and the average from MT bench assay. The structures of the molecules are illustrated in Figure 2. Confirmed hits in vitro and/or in vivo are emphasized using a bold font and a color code: red for hits confirmed in vitro and in vivo, black for in vitro only, green for in vivo only and blue for a negative control that was found potent in vivo but not in vitro. Errors on the computed ΔGbind range from 0.34 to 0.97 kcal.mol– 1. https://cdn.elifesciences.org/articles/80387/elife-80387-fig2-data1-v2.docx Download elife-80387-fig2-data1-v2.docx In summary, this computational approach allowed us to identify 22 potential hits from ∼7 million molecule candidates. Robust HCS Scoring of endogenous mRPIs in cells with the MT bench assay In order to score the interaction between mRNAs and YB-1 in cells with an HCS imager, we adapted a method that we recently developed, the MT bench (Boca et al., 2015). Briefly, an RBP is brought to the microtubules (MTs) after its fusion to a microtubule binding domain (MBD) so it can be used as a bait for a prey (here, mRNA). In our constructs, an RBD was fused via its C-terminus to a GFP-tag itself fused to the MBD (MBD-GFP-RBP). As MBD, we used the longest isoform of MAPT gene (2N4R-tau), which allows the binding of microtubules in a non-cooperative manner (Butner and Kirschner, 1991) and enables the bait protein, for example YB-1, to protrude outward the MT surface several nm away from the microtubule surface, which increases the bait accessibility to ligands Boca et al., 2015; the RBP brought on MTs subsequently interacts with mRNAs in the cytoplasm which results in an enrichment of endogenous mRNAs along the MT network in cells (Figure 3a). To measure the enrichment of poly(A)-mRNA on microtubules, we used in situ hybridization with a cy3-labeled poly(dT) probe in fixed U2OS cells (Lubeck et al., 2014) which have a well-extended MT network. Importantly, an HCS imager equipped with a water immersed lens (40 x, NA = 1.1) operating in confocal mode was necessary to reach a sufficiently high lateral resolution and thus clearly distinguish the microtubule network in fluorescence microscopy images (Figure 3b and Figure 3—figure supplement 2.3). To detect the presence of baits on MTs, an automatic detection scheme has been implemented using specific criteria such as a low width-to-length ratio of the detected GFP-rich spots (<0.22) keeping only MT-shaped spots (Figure 3b). Details on image acquisitions and statistical analysis are provided in Appendix 3. Figure 3 with 4 supplements see all Download asset Open asset MT bench assay scores mRPIs in cells in a 96-well plate. (a) Left panel: Schematic view of the MT bench technology. A GFP-labeled RBP fused to MBD (Microtubule-Binding domain, yellow) was brought to microtubules in U2OS cells to attract endogenous mRNAs (in red) on the microtubule network (grey). Middle panel: Image of a 96-well plates seeded with U2OS cells. Right panel: Image of a single well processed by HCS imager showing the expression of MDP-GFP-YB-1 in U2OS cells (green). (b) U2OS cells expressing MBD-GFP-YB-1 (bait in green, GFP). mRNAs in red (in situ hybridization, poly(dT) probe). Nuclei in blue (DAPI). The images were obtained with an HCS imager (40 x, water immersed objective operating in confocal mode). (i) DAPI and the red channel (mRNA) were used to detect automatically the nuclei and cytoplasm, respectively. (ii) Using HARMONY “find spots” procedure, elongated spots along the microtubules were detected using the green channel (the bait, RBP). Spots were selected owing to their width-to-length ratio (<0.22) and their enrichment in GFP (YB-1). Scale bar: 20 μm. (c) Left panels: The enrichment of mRNAs in single selected spots (spot/cytoplasm intensity ratio, red channel) and spot bait intensity on microtubules (green channel) show a linear relationship when YB-1 was used as bait. The slope of the regression line reflects the affinity of an RBP for mRNAs. A large number of cells can be analyzed by HCS (>500 cells per well with in average 10–50 spots per cell). Slopes from linear regression were measured for each well with a 95% confidence interval. Right panel: SSMD value estimated by measuring the normalized slopes in 48 negative controls (MBD-GFP used as bait) and 48 positive controls (YB-1 was used as bait). The SSMD value is 8.1 for a 96-well plate. Spot data from all wells are shown in Figure 3—figure supplement 2.3a. (d) Bar diagram representing the enrichments of 13 different mRNAs measured by RT-PCR after two different purification procedures, co-sedimentation (MT pellet) and immunoprecipitation (Beads), and for 2 different RBPs, YB-1 and HuR; the purification procedures are illustrated in Figure 3—figure supplement 2.3, data and correlation analysis are provided in Appendix 5—table 5 for 3 RBPs (YB-1, HuR, and FUS). (continued). Results indicate an accurate detection of MBD-GFP-YB-1-decorated MTs in U2OS cells. In the selected spots, the mean bait intensity and enrichment in mRNA (ratio of the mean intensity of cy3 in the spots to that in the cytoplasm) were measured (Figure 3c). In contrast to MBD-GFP spots, the enrichment of mRNA in MBD-GFP-YB-1 spots located on MTs increased linearly with GFP spot fluorescence. This result demonstrates the positive correlation between the number of YB-1 brought on MTs and the relative enrichment of mRNAs on the same MTs. Interestingly, the slope thus depends directly on the binding affinity of the bait for mRNAs. We therefore considered the slope as a mRNA affinity score for RBPs brought on MTs. We next estimated the sensitivity of this scoring method by measuring the slopes of 48 positive (MBD-GFP-YB-1) and 48 negative (MBD-GFP) controls from a 96-well plate (Figure 3c; data from all wells are given in Figure 3—figure supplement 2.3). The measured SSMD value (strictly standardized mean difference) for this assay is 8.1, which is the difference of the mean values of the positive and negative controls divided by the standard deviation. A SSMD value of 8.1 corresponds to an efficient assay whatever the estimated strength of the positive controls (Bray and Carpenter, 2017). The SSMD value also indicates the sensitivity of the MT bench assay. Here, only molecules that decrease the slope by more than 1/8 of the positive control can be detected. Additional negative control experiments were also conducted using, as baits, 3 different DNA-binding proteins that should not bring mRNAs onto microtubules in the MT bench assay. These proteins are DNA topoisomerase 1 (TOP1), Apurinic/apyrimidinic endonuclease 1 (APE1), and DNA ligase 1 (LIG1). The results represented in Figure 3—figure supplement 2.3 confirm that DNA-binding proteins indeed fail to bring mRNA onto the microtubules. In summary, the automatic image analysis that we implemented for the MT bench assay can reliably detect and score the interaction of YB-1 with mRNAs in the cytoplasm with HCS capacity. MT bench assays measure RBP-specific interactions with mRNAs in cells Although mRNAs can be detected on microtubules in a 96-well plate setting with an HCS imager, it is critical to estimate whether fusion proteins that confine RBPs to microtubules do not lead to artificial interactions with non-specific transcripts. To this end, we designed an experiment to estimate the enrichments of mRNAs on microtubules in cells expressing MBP-GFP-RBP (mRNA brought on the microtubule with the bait protein). Briefly, cell lysates were incubated with purified MTs reconstituted in vitro from sheep brains (Figure 3—figure supplement 2.3a). Therefore, mRNAs were brought onto MTs owing to the presence of MBP-GFP-RBP in cell extracts and subsequently detecte
TDP-43 and FUS are two mRNA-binding proteins associated with neurodegenerative diseases that form cytoplasmic inclusions with prion-like properties in affected neurons. Documenting the early stages of the formation of TDP-43 or FUS protein aggregates and the role of mRNA stress granules that are considered as critical intermediates for protein aggregation is therefore of interest to understand disease propagation. Here, we developed a single molecule approach via atomic force microscopy (AFM), which provides structural information out of reach by fluorescence microscopy. In addition, the aggregation process can be probed in the test tube without separating the interacting partners, which would affect the thermodynamic equilibrium. The results demonstrate that isolated mRNA molecules serve as crucibles to promote TDP-43 and FUS multimerization. Their subsequent merging results in the formation of mRNA granules containing TDP-43 and FUS aggregates. Interestingly, TDP-43 or FUS protein aggregates can be released from mRNA granules by either YB-1 or G3BP1, two stress granule proteins that compete for the binding to mRNA with TDP-43 and FUS. Altogether, the results indicate that age-related successive assembly/disassembly of stress granules in neurons, regulated by mRNA-binding proteins such as YB-1 and G3BP1, could be a source of protein aggregation.
The immunomodulatory effects of opiates can modify host defenses against infection. We investigated the mechanisms involved in these effects by studying the influence of morphine on the pathogenesis of murine Friend retrovirus infection. The response to this opiate varied greatly according to the treatment schedule. Daily intraperitoneal administration of morphine (50 mg/kg) for 16 to 27 days attenuated pathological manifestations in infected animals without modifying the mortality rate. The protective effect increased proportionately with the duration of treatment and depended on the time of treatment initiation relative to inoculation. Naloxone (100 mg/kg/day i.p.) inhibited the morphine-induced decrease in both splenomegaly and viral titer. Mifepristone--a glucocorticoid receptor inhibitor--had no significant effect on the morphine-induced attenuation of splenomegaly. The influence of the infection on acute morphine toxicity was also analyzed using a nonlethal dose in noninfected mice (200 mg/kg). Susceptibility to morphine increased in parallel to the development of the infection, with mortality rates ranging from 20% on day 14 to 90% on day 21. Simultaneous administration of naloxone (20-100 mg/kg) reduced the mortality rate and postponed death. Administration of mifepristone, terfenadin, phentolamine or propranolol did not modify mortality at the doses used. These findings show that the influence of morphine on the development of Friend virus infection in mice depends on the conditions of administration. The transient protective effect seen in certain conditions of administration appears to be due essentially to the direct effects of morphine on its specific receptors.
To examine the role of nitric oxide (NO) in murine AIDS (MAIDS) pathogenesis, we determined NO production and inducible NOS (iNOS) mRNA expression in the macrophages of LP-BM5-infected mice, together with the in vivo effects of l-NAME, a competitive inhibitor of NO synthase. LP-BM5 infection induced neither spontaneous nitrite production nor iNOS mRNA expression. No differences in IFNγ + LPS-induced nitrite production or iNOS mRNA expression were observed in macrophages from non-infected or infected mice. Spleen weight, ecotropic MuLV replication, the blood lymphocyte phenotype and proliferative response of splenocytes were not modified by l-NAME. LP-BM5 infection did not increase macrophage NO production and NO production did not appear to protect against LP-BM5-induced immunodeficiency.
