SU056

Degenerate consensus sequences in the 30-untranslated regions of cellular mRNAs as specific motifs potentially involved in the YB-1-mediated packaging of these mRNAs

Alexander V. Gopanenko a, 1, Alexey A. Malygin a, b, 1, Olga A. Kossinova a, Alexey E. Tupikin a, Marsel R. Kabilov a, Galina G. Karpova a, b, *

A B S T R A C T

The multifunctional protein YB-1 has previously been shown to be the only protein of the cytoplasmic extract of HEK293 cells, which is able to specifically interact with imperfect RNA hairpins containing motifs that are often found in exosomal (e) RNAs. In addition, it has been revealed that similar hairpins formed by degenerate consensus sequences corresponding to three eRNA-specific motifs are responsible for the cooperative binding of YB-1 to RNA in vitro. Here, using the photoactivatable ribonucleoside- enhanced cross-linking and immunoprecipitation method applied to HEK293 cells producing FLAG- labeled YB-1, we identified mRNAs cross-linked to YB-1 in vivo and then carried out a search for the aforementioned sequences in the regions of the YB-1 cross-linking sites. It turned out that many of the mRNAs found cross-linked to YB-1 encode proteins associated with various regulatory processes, including responses to stress. More than half of all cross-linked mRNAs contained degenerate consensus sequences, which were preferably located in 30 -untranslated regions (UTRs), where most of the YB-1 cross-linking sites appeared, although not close to these sequences. Furthermore, YB-1 was mainly cross-linked to those mRNAs with degenerate consensus sequences, which could be classified as pack- aged because their translation levels were low compared to cellular levels. This suggests that the cooperative binding of YB-1 to mRNAs through the above sequences probably triggers the well-known multimerization of YB-l, leading to the packaging of these mRNAs. Thus, our findings indicate a previ- ously unknown link between the degenerate consensus sequences present in the 30 -UTRs of many cytoplasmic mRNAs and YB-1-mediated translational silencing.

Keywords:
RNA-protein interactions YB-1
Photoactivatable-ribonucleoside enhancing cross-linking and immunoprecipitation Regulation of gene expression
mRNA packaging
mRNA consensus sequences

1. Introduction

The Y-box-binding protein 1 (YB-1, YBX1), a member of a family of nucleic acid-associated proteins, is engaged in many DNA/RNA- dependent events, including DNA repair, transcription, pre-mRNA splicing, mRNA packaging, and regulation of mRNA stability and translation [1e3]. In humans, YB-1 participates in cellular processes such as proliferation, differentiation and malignant transformation, as well as responses to stress [2,4]. Moreover, it has been found that this protein is an important factor in the regulation of translation of specialized mRNAs involved in the Nodal signaling pathway in zebrafish [5,6]. Finally, YB-1 has been identified as a protein responsible for the sorting of specific miRNAs into exosomes [7,8]. Such a ubiquity of YB-1 is provided by its three main functional domains, which can be distinguished in its structure. These are the N-terminal arginine/proline rich domain (A/P), the cold shock domain (CSD) acting as a nucleic acid-binding module and the C- terminal domain (CTD). The CSD of YB-1 is well-structured and resembles CSDs of bacterial proteins [9], while both the CTD and the A/P domain are intrinsically disordered, which ensures the YB-1 ability to fold into multiple structures depending on its binding partners [10].
It is known that YB-1 binds to mRNAs in various YB-1/mRNA molar ratios, forming both translatable and untranslatable mRNPs [11,12]. The CTD provides the electrostatic binding of YB-1 to mRNA through clusters of positively charged amino acid residues, whereas the CSD is essential for the cooperative binding of the protein to mRNA [13,14]. Currently available information on sequences in RNAs previously proposed as those participating in YB-1 binding indicates that these sequences have practically nothing in common, although they are enriched in cytidine and adenosine residues (Table 1). Thus, A/C-rich sequences of exon splicing enhancers [15] and, in particular, the ACAAC motif [16] have been detected as el- ements recognized by YB-1 in pre-mRNAs. The sequence UCCA-G/ACAA has been identified as an YB-1 binding site on its own mRNA [17]. With the application of a method for the systematic analysis of RNA binding specificities based on the in vitro selection of short RNAs by a target protein and their subsequent identification using the microarray technique, it has been shown that YB-1 is an effective binder of RNA sites containing the CUGC motif [18]. In addition, using the method of systematic evolution of ligands by exponential enrichment, the CAYC tetramer has been defined as a consensus sequence of the YB-1 binding site in a splicing enhancer [19]. Later, based on the genome-wide analysis of YB-1-RNA in- teractions in U251-MG glioblastoma cells using the individual nucleotide resolution cross-linking and immunoprecipitation (iCLIP) technique, the UYAUC motif has been defined as a sequence- specific binding site of YB-1 on RNAs, including mRNA coding parts and several microRNAs. This motif has been found in 17.8% of 40 nt regions containing the cross-linking site [20]. Using the photo- activatable ribonucleoside-enhanced CLIP (PAR-CLIP) method applied to fused cells expressing the human and rat genes, the RNA binding site for YB-1 has been expanded to the sequence UCUUUNNCAUC, which is located mainly in the 30-untranslated regions (UTRs) and coding parts of mRNAs [21]. However, it has been unclear from the study which fraction of cross-linked mRNAs contained such an YB-1 binding site. Thus, despite the extremely important functional implications of YB-1 as an RNA-binding pro- tein, it remains poorly understood how YB-1 recognizes RNA.
Recently, we have demonstrated that YB-1 is the only protein, which can be pulled down from the cytoplasmic (S100) extract of human embryonic kidney 293 (HEK293) cells using short synthetic RNA hairpins containing one of the linear octanucleotide se- quences: ACCAGCCU (motif 1), CAGUGAGC (motif 2) or UAAUCCCA (motif 3) [22], often found in exosomal (e) RNA [23]. These RNA hairpins have been shown to bind YB-1 specifically [22], which has also been confirmed in experiments utilizing the respective re- combinant protein [24]. However, when examining the binding of YB-1 to the 30-UTR fragment of Sept14 mRNA (Sept14 RNA) containing all three motifs, we have found that YB-1 prefers to interact not with the motifs themselves, but with specific imperfect hairpin structures that include these motifs [24]. Besides, it has been shown that the same RNA with nucleotide substitutions preventing the motifs 1e3 from being drawn into the hairpin structures is also able to interact with YB-1, but unlike Sept14 RNA, in a non-cooperative manner. Therefore, it has been suggested that these are the hairpin structures formed with participation of the motifs 1e3 that are responsible for the cooperative binding of YB-1 to mRNAs [24]. Using bioinformatics analysis, we have revealed three degenerate consensus sequences (21e36 nt long) corresponding to the sur- roundings of each of motifs 1e3 in these hairpin structures, designating them as consensus sequences 1e3, respectively [24].
In this study, using the PAR-CLIP method applied to HEK293 cells transfected with a plasmid encoding YB-1 with a FLAG-containing peptide at the N-terminus (FLAGYB-1) and grown in a medium containing 4-thiouridine (s4U), we identified all mRNAs cross-linked to YB-1 in vivo and performed a search for sequences similar to those described in ref. 24 in these mRNAs. Analysis of next-generation sequencing (NGS) data revealed that most of the genes, to which reads with characteristic T/C transitions fell, were nuclear genes encoding proteins involved in various regulatory processes, including those related to cell responses to stress. The mapping of the T/C transition positions to mRNAs showed that more than 50% of the mRNAs cross-linked to YB-1 contained the above mentioned degenerate consensus sequences 1e3, which were preferably located in 30-UTRs, where most of the YB-1 cross-linking sites were detected at a distance of up to several hundred nt from these sequences. Besides, short sequences similar to those previously proposed as YB-1 binding sites (see above) and, in particular AACUCU, were also found at a distance of up to 20 nt from the protein cross-linking sites. For mRNAs containing degenerate consensus sequences 1e3, it was revealed that YB-1 was mainly cross-linked to those whose translation levels were low compared to cellular levels, which allowed these mRNAs to be classified as packed mRNAs. Thus, the cross-linking sites of YB-1 in mRNAs with degenerate consensus sequences 1e3 reflect the cor- responding RNA-protein contacts in ribonucleoproteins resulting from the packaging of mRNAs mediated by YB-1 multimerization [12,13]. Apparently, the cooperative binding of YB-1 to mRNAs through the above sequences contained in 30-UTRs initiates the multimerization of YB-1 on these mRNAs. This study also discusses the possible need for YB-1-mediated packaging when transferring mRNAs containing degenerate consensus sequences 1e3 into exosomes.

2. Materials and methods

2.1. Cells culturing, transfection and PAR-CLIP procedure

PAR-CLIP was performed using HEK293 cells producing FLAGYB- 1, according to Ref. [25] with minor modifications. Anti-FLAG M2 (Sigma, #F1804) antibodies were utilized for immunoprecipitation of proteins cross-linked to RNAs. In typical experiment, adherent HEK293 cells were cultured in 10-cm Petri dishes on Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (100 U/ml) (all from Thermo Fisher Scientific) in CO2-incubator (5% CO2) at 37 ◦C up to 80% of confluence. Cells were transfected with previously described plasmid pCAGGS-FLAG-YB1 [26], kindly gifted by Dr. Kawaguchi, using Turbofect (Thermo Fisher Scientific) according to the manufacturer’s protocol. After 24 h, s4U was added to the cells up to 250 mM, and after the next 24 h, the culture medium was aspirated, the cells were washed with ice-cold PBS, and cell culture dishes were put on ice and UV-irradiated (365 nm, 1000 mJ/cm2) in Bio-Link (Viber Lourmat). Then, cells were washed with ice-cold PBS and collected by centrifugation at 1000g (4 ◦C, 1 min). Next procedures were performed as described in Ref. [27]. Briefly, the cell pellet was lysed, triturated using the 29 G needle and clarified by centrifugation at 20 000 g for 10 min at 4 ◦C. The supernatant was treated with RNase T1 (1 U/ml) and incubated with anti-FLAG M2 pre-bound to magnetic Protein G beads (Dynabeads, Life Technologies). After the immunoprecipitation, beads were succes- sively treated with RNase T1 (1 U/ml) and calf intestinal phospha- tase (NEB) (0.5 U/ml), and after the respective washes, RNA on the beads was 32P-labeled by incubation with g-32P-ATP (2 mCi) and T4 polynucleotide kinase (NEB) (1 U/ml) for 30 min at 37 ◦C. Then ATP was added up to 1 mM to this mixture and the incubation was continued for more 5 min. After final washing, the beads were resuspended in sodium dodecyl sulfate polyacrylamide gel elec- trophoresis (SDS-PAGE) loading buffer, and the supernatant sepa- rated by centrifugation was subjected to 10% SDS-PAGE. Gel slices containing radioactive bands were then excised, dried under vac- uum and wetted with solution of Proteinase K (1 mg/ml) in the buffer (50 mM TriseHCl (pH 7.5), 75 mM NaCl, 6 mM EDTA and 1% SDS) for 4 h at 37 ◦C. The eluate containing the RNA fragments was sub- jected to phenolechloroform deproteination, RNA was precipitated using isopropanol and resuspended in nuclease-free water. PAR- CLIP experiments were conducted in three biological replicates.

2.2. cDNA library preparation from cross-linked RNA fragments and NGS data analysis

cDNA libraries for deep sequencing were prepared using the NEBNext Small RNA Library Prep Set for Illumina (NEB) according to the manufacturer’s protocol. To select the size of the cDNA frag- ments, the libraries were loaded onto 2% Agarose Dye Free Gel Cassettes (Sage Science) and fragments of 125e180 bp were iso- lated with the Blue Pippin DNA size selection system (Sage Science) according to the manufacturer’s instructions. Sequencing the li- braries was performed on the MiSeq (Illumina) genomic sequencer using the Reagent Kit v3 (150 cycles, Illumina) in the Genomics Core Facility of Siberian Branch of the Russian Academy of Sciences (SB RAS) working on the basis of Institute of Chemical Biology and Fundamental Medicine (ICBFM) of SB RAS. The respective read data were submitted to the GenBank under the study accession PRJNA310826 and the sample accessions SRR8061701, SRR8061703 and SRR8061808. Fastq reads were analyzed using tools of the CLC GW 11.0 software (Qiagen). The reads were filtered both by the quality of the adapters (ambiguous limit, 2; quality limit, 0.031) and by their sequences, and then were mapped to the human reference genome (hg38) with the Ensembl annotation GRCh38.89 using the RNA-Seq analysis tool (length fraction, 0.95; similarity fraction, 0.95; maximum number of hits for a read, 1; and other parameters by default). PAR-CLIP peaks were detected by the appearance of T/C and A/G transitions in reads corresponding to sense or antisense DNA strands, respectively, which were found using the basic variant detection tool (ploidy, 2; ignore broken reads, no; ignore non- specific matches, no; minimum coverage, 5; minimum count, 1; frequency, 0.5%; relative read direction filter, no; other parameters by default). After filtering the transitions, depending on which DNA strand they appeared in, the transition frequency was determined for each position, and only those transitions whose total frequency exceeded the frequency of the “spontaneous” background T/C transitions estimated in control experiments (see below), were taken for further consideration. Transitions in genes and different regions of genes were counted using BEDTools [28].

2.3. Estimation of the background level of T/C transitions

A DNA template for the synthesis of human 18S rRNA fragment 1e1207 was prepared by PCR using primers 50-aaattaa- tacgactcactatagggtacctggttgatcctg-30 (forward) and 50-ccgtcaattcctttaagtttc-30 (reverse) and 18S rDNA. Corresponding RNA samples with s4U residues, randomly replacing U ones, or without them, were obtained by T7 transcription, as described [27], using UTP and s4UTP in a 10:1 ratio to prepare s4U-containing RNA. To determine the background level of T/C transitions, three RNA samples were utilized. One of them was RNA with s4U residues, which was subjected to mild UV irradiation. To prepare this RNA sample, 0.25 A260 units of s4U-containing RNA in 50 ml of 20 mM Tris-HCl (pH 7.5) buffer containing 150 mM KCl and 10 mM MgCl2 were irradiated, as described above. Two other RNA samples (s4U- containing and without s4U) were not UV-irradiated. For the preparation of cDNA libraries, the NEBNext Ultra RNA Library Prep Kit (NEB) was used, and the RNA fragmentation step was omitted. Sequencing the libraries was performed on the MiSeq platform with the Reagent Kit v3 (600 cycles, Illumina) in the aforemen- tioned SB RAS Genomics Core Facility. The respective read data were submitted to the GenBank under the study accession PRJNA310826 and the sample accessions for s4U-containing UV- irradiated RNA (SRR8081971), s4U-containing non-irradiated RNA (SRR8081973) and non-irradiated RNA without s4U (SRR8081972). Analysis of Fastq reads and mapping of reads filtered in the same way as PAR-CLIP-reads were performed as described above. Data on the distribution of T/C transitions were obtained using the basic variant detection tool (ploidy, 2; ignore broken reads, no; ignore non-specific matches, no; minimum coverage, 10; minimum count, 3; frequency, 0.5%; relative read direction filter, no; other param- eters by default). The frequency of occurrence of T/C transitions in different positions in the DNA strands corresponding to the above RNA samples was averaged to find the background level of “spon- taneous” T/C transitions, and the respective value (1%) was used to determine the T/C transitions corresponding to FLAGYB-1 cross- linking sites (see above).

2.4. Analysis of motifs around T/C transition positions

Scanning motifs surrounding the T/C transition positions was performed using a desktop-based software of the MEME (Multiple EM for Motif Elicitation) Suite [29], ver. 5.05, with parameters (mode anr, max_number_of_motif 20, min_motif_width 8, max_motif_width 12, max_motif_sites 2000). The 41-mer se- quences (20 nt downstream and 20 nt upstream of the T/C transi- tion positions) derived for the T/C transition-containing genomic regions corresponding to mRNA 30-UTRs (according to our PAR-CLIP data) were used as input ones.

2.5. Analysis of the content of degenerate consensus sequences in HEK293 mRNAs

The degeneracy limits for the consensus sequences 1 (50- caggngtnnnnnacnnncctg-30), 2 (50-gagnntgcagtgnnnnnnnnnnnnnn- cactgcactc-30) and 3 (50-cctgtnntcccagnnnnttgggangcnnagg-30), previously revealed in the set of HEK293 mRNA 30-UTRs derived from the RNA-seq data in Ref. [24], were recalculated using the method described in the same work [24], with a resolution of up to 10 mismatches for each of the sequences instead of 4 ones allowed in Ref. [24]. For each of the above sequences, the number of resolved mismatches immediately preceding the number of those, at which the number of respective consensus sequences in mRNA 30-UTRs begins to grow exponentially, was taken as the limit of degeneracy. Given the found limits, the number of mRNA 30-UTRs containing consensus sequences 1e3 was redefined and retrieved with the application of the biomaRt package (2.34.2) utilizing useMart (“ensembl”, dataset “hsapiens_gene_ensembl”). Similarly, the 50- UTRs and CDSs of mRNAs were analyzed, whose sequences were recovered from the above RNA-seq data in the same way as those of the 30-UTRs [24].

2.6. Analysis of the distribution of degenerate consensus sequences and YB-1 cross-linking sites along the 30-UTRs

The pattern of distribution of degenerate consensus sequences and positions corresponding to T/C transitions along 30-UTRs of mRNAs were assessed and compared using the ENSEMBL hg38.89 annotation. The distribution density of the distances between the consensus sequences and the positions corresponding to the T/C transitions was represented in the form of histograms that were generated using standard R functions. A theoretical pattern of the distribution of distances between two random points located on the 10000-mer segment was used as a background control when obtaining the above histograms. For this control, two randomly generated 10000-mer vectors of numbers were created, and one vector was subtracted from the other one. The absolute values of the resulting vector were used as distances between two random points and as a source of points for plotting a theoretical histogram. The correlation between the 30-UTR length and the minimal dis- tance between the consensus sequences and the positions corre- sponding to the T/C transitions was assessed using the standard R function cor with default parameters. The above-described pipeline was also used to calculate the distribution of absolute distances between the consensus sequences and the positions corresponding to the T/C transitions in the 30-UTRs. The distribution pattern of 30- UTR lengths was plotted using standard R functions. All described manipulations were also implemented to perform the respective analyses for the 50-UTRs and CDSs of mRNAs.

2.7. Analysis of the translational activities of cellular mRNAs

To reveal a set of mRNAs translated in HEK293 cells, ribosome profiling was carried out as described [30] with minor changes. For typical experiments, one 10-cm Petri dish of cells was used. At 80e90% of cell confluence, harringtonine was added to the culture medium to a final concentration of 2 mg/ml for 2 min, and then cycloheximide was added to a final concentration of 100 mg/ml for 1 min. After removal of the culture medium, the cells were briefly washed with ice-cold PBS and then were collected by centrifuga- tion at 500g for 5 min at 4 ◦C. A third of the cell pellet was resus- pended in TRIzol (Invitrogen) and used to extract total RNA according to the manufacturer’s protocol for subsequent RNA-seq. The remaining 2/3 of the cell pellet was used for Ribo-seq. Cells were lyzed in 300 ml of buffer containing 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl2, 1 mM DTT, 100 mg/ml cycloheximide, 1% Triton-X100 and 25 U/ml Turbo DNaseI (Ambion). The lysate was then treated with RNase I (Ambion), loaded onto a sucrose cushion formed in a half-cut polycarbonate ultracentrifuge tube for use in the SW-60 rotor, and monoribosomes were pelleted by centrifugation at 45 000 rpm for 4 h at 4 ◦C. RNA fragments were recovered from pellet using the miRNeasy kit (Qiagen) and those whose size was in the range of 26e34 nt were selected. The resulting RNA fragments were 50-phosphorylated and 30- dephosphorylated. Based on this sample of fragments, a DNA li- brary was prepared using the 5500 SOLiD Fragment Library Core Kit (Life Technologies) in accordance with the manufacturer’s in- structions, which was sequenced on SOLiD 5500xl platform (Life Technologies) in SB RAS Genomics Core Facility (ICBFM, SB RAS). The resulting read length was 50 bp. The SOLiD read data were submitted to the GenBank under the study accession PRJNA310826 and the sample accessions for Ribo-Seq (SRR8061702, SRR8061704 and SRR8061705) and RNA-Seq (SRR8061706 and SRR8061707).
Since the Pearson product-moment correlation coefficient between these two RNA-Seq replicates (r 0.998) showed their extremely high identity, the third RNA-Seq replicate was omitted. To minimize sequencing errors, the color-space reads were subjected to error correction using the SOLiD Accuracy Enhancement Tool v.2.4 (Life Technologies). Corrected data were analyzed using CLC GW 11.0 (Qiagen) with default parameters. After adapter and quality trim- ming, SOLiD reads were mapped to the hg38 reference genome with the Ensembl GRCh38.89 annotation, ignoring non-specific matches. The resulted bam-files were sorted and indexed and then used for further analysis on the R Bioconductor platform. Read counts data for RNA-seq and Ribo-seq were obtained using the R Genomic Alignments package [31]. Mitochondrial genes were excluded from further consideration. Data on cytoplasmic mRNAs obtained this way were combined with data on mRNAs cross- linked to YB-1, the presence of consensus sequences in their 30- UTRs and their appearance in exosomes, and then according to these attributes, mRNAs were divided into groups. The scatter plots with a LOESS (LOcal regrESSion) curve for Ribo-seq counts (as X) ~ RNA-seq counts (as Y) were created using the R ggplot2 package with the stat_smooth function using parameters (method ¼ “loess”, formula ¼ y ~ x, span ¼ 0.75, se ¼ TRUE, fullrange ¼ FALSE, level ¼ 0.99).

3. Results

3.1. Cells producing FLAGYB-1 and in-cell RNA-protein cross-linking

HEK293 cells ectopically producing FLAGYB-1 were obtained by their transient transfection with a plasmid encoding the respective protein. The relative production levels of FLAGYB-1 and endogenous YB-1 in transfected cells compared to the level of endogenous YB-1 in non-transfected cells were estimated by western blot analysis using anti-YB-1 antibodies after separation of the total protein isolated from the lysates of the corresponding cells by SDS PAGE. The results presented in Fig. 1A show that the levels of FLAGYB-1 and endogenous YB-1 in transfected cells are approximately the same and that the overall level of these proteins there is comparable to the level of YB-1 in non-transfected cells.
In-cell RNA-protein cross-linking was carried out by mild UV irradiation of transfected HEK293 cells producing FLAGYB-1 and grown in the presence of s4U. After RNase T1 treatment of lysate of the irradiated cells, FLAGYB-1 was immunoprecipitated with anti- FLAG antibodies, and RNA fragments cross-linked to FLAGYB-1 were 32P-labeled and resolved by SDS-PAGE. The non-irradiated cells grown in a medium containing s4U and subjected to the same manipulations as the irradiated ones were used in a control experiment. With irradiated cells, the gel autoradiograph showed a strong signal in the region of FLAGYB-1 migration, whereas with non-irradiated cells, no FLAGYB-1-specific cross-links were detected (Fig. 1B). To ensure that the above radioactive signal was associated with RNA fragments cross-linked to the target protein, the in-gel resolved cross-links were transferred onto a nitrocellulose mem- brane adsorbing proteins, which was then subjected to autoradi- ography. Again, a strong signal was revealed only for irradiated cells (Fig. 1C).
Several modified versions of the original PAR-CLIP method [25] exist, differing in manipulations related to the recovery of RNA fragments, preparation of cDNA libraries, NGS procedures and so on [8,32e34]. To determine the binding sites of YB-1 in RNAs, we chose the PAR-CLIP protocol including the recovery of RNA frag- ments cross-linked to FLAGYB-1 by immunoprecipitation from irradiated cells, followed by purification with SDS-PAGE. The NGS data obtained using this PAR-CLIP protocol by high-throughput sequencing of the corresponding cDNA libraries on the Illumina platform consisted of 5.5 million reads.

3.2. Determination of the profile of YB-1-targeted RNAs

As it is known, the characteristic T/C transitions in PAR-CLIP sequencing reads arise from occasional substitutions in the syn- thesized cDNA, when the reverse transcriptase reads through s4U residues cross-linked to peptide fragments, and their positions usually correspond to those of s4U residues in RNAs, which form cross-links with the protein of interest. However, the frequency of appearance of T/C transitions due to s4U-mediated RNA-protein cross-links can be low, and, besides, specific T/C transitions can overlap with background ones originating from a variety of other causes. To estimate the background level of T/C transitions, we compared the RNA-seq data obtained with a long T7 transcript consisting of only conventional nucleotides and those gained for the same transcript containing random substitutions of U residues with s4U ones, which was either irradiated with mild UV or non- irradiated. We found that the frequency of spontaneous T/C tran- sitions was less than 1% (Table S1). Therefore, in a further analysis, only T/C transitions that occurred in more than 1% of the reads covering the respective nucleotide positions were considered as specific ones.
When analyzing the NGS data obtained for PAR-CLIP with cells producing FLAGYB-1, we found 6544 positions of T/C transitions in sequencing reads falling to 3060 genes (Tables S2 and S3), most of which (2833 genes), as expected, were nuclear protein-coding genes (see Fig. 2A, left). Remarkably, the share of T/C transition- containing reads falling to mitochondrial genes (Fig. 2A right) was much larger than the fraction of mitochondrial genes among all the genes covered by reads with T/C transitions (Fig. 2A, left), which suggested that the RNAs encoded by the mitochondrial genome actively cross-linked to FLAGYB-1. In about 50% of the nuclear protein-coding genes (1375 ones), the T/C transition-containing reads covered exons (Table S3). The mapping of the T/C transition positions to mRNAs encoded by these genes showed that about 71% of them fell to 30-UTRs, whereas about 25% and 4% turned out to be in CDSs and 50-UTRs, respectively (Table S3, Fig. 2B, left). Comparing the distribution of T/C transitions between gene regions corre- sponding to different parts of mRNAs and the coverage of those with reads in RNA-seq (Fig. 2B, right), one can see that FLAGYB-1 does have a significant preference for cross-linking to 30-UTRs rather than other parts of mRNAs. All this indicates that YB-1 in- teracts with mRNAs, which represent almost half of all cellular RNA partners found for FLAGYB-1, mostly through their 30-UTRs.
Due to the pronounced RNA-binding properties of YB-1, it was expected that it would be cross-linked to many of the most abun- dant cellular mRNAs. However, as can be seen in Fig. S1, only 270 genes from 1.500 ones corresponding to such mRNAs according to RNA-Seq data (Table S4) are found in a set of 1375 nuclear-protein coding genes corresponding to mRNAs cross-linked to FLAGYB-1 (Table S3). This allowed us to conclude that a high level of mRNA in cells was not a determining factor for its binding and cross-linking to FLAGYB-1 and that FLAGYB-1 interacted with a set of selected mRNAs. Indeed, among the mRNAs identified as cross-linked to FLAGYB-1, there are many mRNAs that encode proteins involved in various regulatory processes according to the Gene Ontology annotation [35] (Table 2). Many of these proteins are known to be closely related to the cellular stress response. These include, for example, proteins EIF2AK1, EIF2AK2, EIF4E and EIF4EBP2, which are involved in the regulation of translation [36], and proteins that maintain mRNA stability, such as UPF1 [37], HNRNPC [38] and HNRNPD [39], as well as proteins SRSF1, SRSF3, SRSF5, SRSF6 and SRSF7, which participate in the processes of pre-mRNA splicing and mRNA transport [40]. Thus, YB-1 interacts with a large, but not random pool of cellular mRNAs. Through binding to particular mRNAs, it can be implicated in controlling the expression of specific genes associated with pro- cesses that undergo adaptive changes, allowing cells to survive and to maintain their functions under stress conditions.

3.3. Analysis of YB-1 binding sites in mRNAs

Alignment of the nucleotide sequences of genomic regions corresponding to the 50-UTRs, CDSs and 30-UTRs of mRNAs relative to the positions of FLAGYB-1-specific T/C transitions did not reveal significant similarities in the environments of these positions for any of the regions (Fig. S2). This means that the YB-1 binding sites in mRNAs are highly degenerate, or the protein recognizes specific structures in mRNAs, rather than strict linear nucleotide sequences, or, being bound to mRNAs, it contacts regions that are distant from the sites of initial recognition. Therefore, at first in an attempt to identify any conserved motifs in the sequences surrounding the T/C transitions positions, we scanned segments between 20 nt up- stream and 20 nt downstream of these positions in regions corre- sponding to 30-UTRs, to which the main share of T/C transitions fell, using the MEME Suite software [29]. By limiting the maximum motif width to 12 nt when analyzing 2115 positions of T/C transi- tions in the above regions, we found 11 statistically significant
motifs (for about of 60% positions), including the AACTCT one that had the highest value of statistical significance (Fig. 3). Neverthe- less, it seems unlikely that the AACUCU motif plays an essential role in the binding of mRNAs to YB-1, since this motif is quite common in mRNAs, and YB-1 interacts only with certain mRNA species. In general, sequences in cross-linked mRNAs, which correspond to the motifs surrounding the T/C transitions positions in the analyzed gene regions, resemble and partially overlap those that have been proposed as RNA binding sites for YB-1 in previous reports [15e19,21] (see Introduction).
According to our previous findings [22,24], upon binding to RNA, YB-1 recognizes degenerate consensus sequences 1e3 folded into imperfect hairpins of a specific shape. In this regard, we tried to find out whether these sequences are present in mRNAs cross- linked to FLAGYB-1. During the analysis, we found that the limits of degeneracy for 36-mer consensus sequence 2 and 31-mer consensus sequence 3, at which they were still able to form imperfect hairpins, were even wider than had previously been determined [24] (6 and 7 mismatches instead of 4 and 4 ones, respectively) (Fig. 4 and Fig. S3-S5). We checked the occurrence of consensus sequences 1e3 in 30-UTRs, CDSs and 50-UTRs of the mRNAs encoded by genes with T/C transitions in exons, and discovered that more than half of the mRNAs (596 of 1109) contain such sequences in the 30-UTRs (Table S3). The numbers of the mRNAs with the consensus sequences in CDSs and 50-UTRs turned out to be much less (92 and 14, respectively) (Table S3) than the number of those with the ones in 30-UTRs, which is consistent with the prevalence of the aforementioned motifs 1e3 in mRNA 30- UTRs [22].
Analyzing the mutual locations of the degenerate consensus sequences 1e3 and of the positions corresponding to the T/C transitions in the 30-UTRs, we found that in a quarter of cases, the distance between them was less than 200 nt, and the median dis- tance was 482 nt, whereas the median length of the considered 30- UTRs exceeded 3300 nt (Fig. 5A). Thus, there is a correlation be- tween the locations of the above sequences and the YB-1 cross- linking sites in 30-UTRs of mRNAs. It is seen that, although YB-1 does not cross-link to the 30-UTRs in close proximity to degen- erate consensus sequences 1e3, there is convergence between the locations of these sequences and the positions corresponding to the T/C transitions (Fig. 5B). Considering the almost complete absence of YB-1 cross-linking to nucleotide residues in the above sequences, as well as the high affinity of YB-1 in vitro for short RNA hairpins containing similar sequences [22,24], one can assume that, when recognizing these sequences, the protein interacts with their ribose-phosphate backbones rather than nucleotide bases. How- ever, it should not be ruled out that the low content of U residues in degenerate consensus sequences 1e3 and/or the absence of a suitable target for s4U-mediated cross-linking in YB-1 itself could also be the reasons for the lack of YB-1 cross-linking sites in these sequences.

3.4. Analysis of the packaging of mRNAs with consensus sequences when cross-linking to YB-1

Given that the presence of degenerate consensus sequences 1e3 in RNA insures the cooperative binding of YB-1 to this RNA in vitro [24], one can suppose that the recognition of such sequences in mRNAs in vivo triggers the well-known protein multimerization, leading to the compaction and packaging of the respective mRNAs [12,13]. In this line, the next step was to find out whether the cellular mRNAs were packaged when cross-linked to YB-1. When addressing this issue, we took into account that, in the light of general considerations, if widely represented mRNAs are not translated, they should be packaged. To find out which mRNAs were preferably translated, the HEK293 cell genes were plotted as points with coordinates corresponding to the mean read counts of their mRNAs, determined by analyzing the respective Ribo-seq (abscissa) and RNA-seq (ordinate) data, on the same graph (Fig. 6, Table S5). The LOESS (locally estimated scatterplot smoothing) curve on the graph is a normalized regression line that divides the set of plotted points into two subsets located above and below this curve. Thus, genes corresponding to mRNAs whose translation levels are low (according to Ribo-seq) compared to their cellular levels (according to RNA-seq) could be considered as genes that are associated mostly with poorly translated (silenced) or packaged mRNAs (points located above the normalized regression line). Accordingly, genes whose mRNAs were translated with high efficiency relative to their cellular levels could be assigned to those that correspond mainly to unpacked mRNAs (points located below the normalized regression line). One can see that translated (according to Ribo-seq data) mRNAs with degenerate consensus sequences 1e3 in 30-UTRs are prone to packaging, although the number of unpackaged mRNAs of this type is quite large (696 vs. 329 species) (Fig. 6A). Among the translated mRNAs that were found in the set of those cross-linked to YB-1 via their 30-UTRs, the fraction of packaged ones was more than four fifth (Fig. 6B). Finally, the share of mostly packaged mRNAs, which contained the above sequences 1e3 in 30- UTRs and were among those cross-linked to YB-1, was even larger (about 93%) compared to predominantly unpackaged ones (Fig. 6C). It should be noted that only 287 mRNAs from the set of 1109 ones identified cross-linked to YB-1 through their 30-UTRs (Table S3) were found as being translated using Ribo-seq (Fig. 6B), which in- dicates that mRNAs whose translation levels were very low compared to their cellular levels (i.e., untranslated or packaged mRNAs) were mainly involved in cross-linking. All this allowed us to conclude that mRNAs with degenerate consensus sequences 1e3, which can exist both in unpackaged (translated) and packaged forms, cross-linked to YB-1, mainly when they were packaged into untranslated ribonucleoproteins. These sequences may be impli- cated in mRNA packaging through binding to YB-1 during its multimerization. Remarkably, a similar analysis of RNA-seq data previously obtained for total eRNA from HEK293 cells [22] showed that mRNAs containing degenerate consensus sequences 1e3 in their 30-UTRs are present in the exosomes exclusively as packaged ones (Fig. S6).

4. Discussion

Although YB-1 interactions with RNAs in vivo have already been studied by CLIP methods [20,21], no unifying binding site has been proposed. In this study, using the PAR-CLIP method applied to HEK293 cells producing FLAGYB-1, we revealed the main binding RNA partners of YB-1, which were primarily cytoplasmic mRNAs, although transcripts encoded by the mitochondrial genome were also found. We showed that FLAGYB-1 is cross-linked to a large set of cytoplasmic mRNAs that poorly overlaps with a set of those that have the highest cellular levels. It turned out that genes associated with cellular stress response encode many of the cross-linked mRNAs. The FLAGYB-1 cross-linking sites were preferably located in 30-UTRs of mRNAs, and regions limited by 20 nt around these sites often included short conserved sequences from 6 to 12 nt evenly distributed along these parts of mRNAs. Three long degen- erate consensus sequences previously described as specific RNA hairpins that have high affinity for YB-1 in vitro [24], were found in 30-UTRs of more than half of the mRNA partners of the protein, but these sequences were located at a median distance of 482 nt from the FLAGYB-1 cross-linking sites, and not near them as expected. We determined that the mRNAs carrying the above sequences were predominantly packaged when they formed cross-links with FLA- GYB-1. The obtained data provide new information on the interactions of YB-1 with cellular RNAs, including the involvement of the protein in the regulation of expression of specific protein- coding genes at the translation level, and indicate a possible link between the recognition of degenerate consensus sequences in 30- UTRs of mRNAs by YB-1 and the packaging of these mRNAs.
Our results on multiple contacts of YB-1 with cellular RNAs are mostly consistent with data from earlier studies of YB-1-RNA in- teractions using iCLIP [20] or s4U-utilizing PAR-CLIP [21] mentioned in the Introduction. In the above works, YB-1 cross-linking sites have been found in both 30-UTRs and CDSs of mRNAs, although in the report [20] the number of these sites in the 30-UTRs prevails over the number of those in CDSs, as in our study. The multiple interactions of YB-1 with mRNAs are not surprising, because in addition to the above-mentioned ability to package mRNAs [12], it integrates into P-bodies and stress granules (SGs) containing translationally silent mRNAs and RNA-binding proteins [41e44] and regulates the formation of SGs [44]. Among the mRNAs cross- linked to YB-1, we did identify many mRNAs that encode proteins involved in specific cellular processes related to the regulation of mRNA metabolism, including transport, translation and stability, as responses to stress. However, we do not exclude that the stress in the cells caused by the experimental conditions, namely, transfection with a plasmid encoding YB-1, incorporation of s4U residues into RNAs or UV irradiation, could also activate some of the genes associated with stress responses, thereby resulting in the appear- ance of the corresponding mRNAs in the set of those cross-linked to YB-1.
As noted in the Introduction, earlier attempts to identify the YB- 1 recognition site on mRNAs have not led to any unambiguous sequence that could act as such a site. Our data, in general, support this point by showing short motifs around the FLAGYB-1 cross- linking sites, which are similar to those suggested in previous studies (see refs. in the Introduction) as YB-1 binding sites. Apparently, all these motifs can be considered as those coming occasionally in contact with YB-1 when a positively charged CTD of YB-1 electrostatically interacts with the mRNA ribose phosphate backbone [13]. Recognition by YB-1 of much longer motifs, such as degenerate consensus sequences 1e3 found in 30-UTRs of cross- linked mRNAs, possibly with the aid of some other RNA-binding proteins, could initiate its multimerization along the 30-UTRs, which also involves the CSDs, and thereby lead to the compaction and packaging of the respective mRNAs, making them inactive during translation. This conclusion is in line with our previous data that the above sequences are present in a large pool of human mRNA 30-UTRs [24] and that model RNA hairpins containing similar sequences have a high specific affinity for YB-1, and, being a part of RNA, they provide the cooperative binding of YB-1 to the RNA in vitro [22,24]. The findings that mRNAs with degenerate consensus sequences 1e3 are less actively translated (i.e., pre- dominantly packed) compared to other mRNAs that are cross- linked to YB-1, are also consistent with the above statement.
Our data allow us to suggest a way for the transition of trans- lationally active mRNAs containing the degenerate consensus se- quences 1e3 in 30-UTRs into packaged untranslated ribonucleoproteins. When mRNA is translated, the YB-1 molecules interact through their CTDs with its ribose phosphate backbone, while the CSDs of YB-1 recognize degenerate consensus sequences 1e3 folded into specific structures mostly in the mRNA 30-UTR, occupying it, and the more such sequences is associated with YB-1, the more likely multimerization of YB-1 on this mRNA. Going along mRNA, this process, in accordance with the model proposed in Ref. [13], extends to its CDS, so that the mRNA ceases to be trans- lated, which further results in its compaction and packaging. It should be noted that the above model suggests that the predomi- nant multimerization of YB-1 on certain mRNAs could be realized due to the existence of some specific sites [13]. In this line, degenerate consensus sequences 1e3 present in the 30-UTRs of many human mRNAs are very suitable candidates for these sites. One of the possible fates of mRNAs packaged in this way may be their sorting into exosomes, since, according to our analysis of previously obtained RNA-seq data [22], the fraction of mRNAs with the above sequences in the total eRNA from HEK293 cells is rep- resented exclusively by packaged ones. We believe that mRNAs that do not have similar sequences are packaged to a lesser extent than those that contain them, although we do not claim that degenerate consensus sequences 1e3 are the only ones involved in YB-1- mediated mRNA packaging.
The high density of cross-links between FLAGYB-1 and mito- chondrial transcripts was somewhat surprising, since YB-1 is usu- ally not considered as a protein with mitochondrial localization (see the MitoCarta [45]). However, according to the MitoMiner database, YB-1 has been found in mitochondria of humans and mice in several mass spectrometry studies [46]. Mitochondrial localization of YB-1 has also been demonstrated for stressed HeLa cells [47]. In particular, it has been shown that the mismatch-repair activity detected in HeLa mitochondrial extracts is significantly reduced with a decrease in the level of YB-1 in these extracts. Our data that, in addition to mitochondrial mRNAs, mitochondrial rRNAs and several mitochondrial tRNAs also present in a set of RNAs cross-linked to FLAGYB-1, prompted us to link the appearance of mitochondrial mRNAs in this set with the binding of YB-1 to mitochondrial polycistronic H- and L-strand transcripts in mean read counts obtained in RNA-seq and Ribo-seq for their mRNAs (in log2 scale). RNA-seq mean read counts correspond to the total mRNA content in cells, and Ribo- seq mean read counts reflect the amount of mRNA fragments protected by ribo- somes, i.e. translation levels of mRNAs. The normalized regression (LOESS approxi- mation) line plotted for black dots is shown as a blue curve with a gray area indicating regions with a p-value> 0.01. The points above and below the curve are considered as those corresponding to genes with mostly packaged and unpackaged mRNAs, respectively; m and n, the number of points above and below the regression line, respectively, excluding those in a gray area. (A) The distribution of genes containing degenerate consensus sequences 1e3 in the regions corresponding to the 30 -UTRs of their mRNAs (yellow points). (B) The distribution of genes corresponding to mRNAs cross-linked to YB-1 via their 30 -UTRs (cyan points). (C) The distribution of genes corresponding to mRNAs that contain consensus sequences in the 30 -UTRs and are present in the set of those cross-linked to YB-1 via their 30 -UTRs (green points). mitochondrial RNA granules [48], rather than to mature mRNAs. This assumption is consistent with a recent report that YB-1 prac- tically does not bind to mitochondrial mRNAs, such as ND1, ND4 and ATP8 [49], which are among those cross-linked to FLAGYB-1 in this study. Perhaps YB-1 is also involved in the packaging of pri- mary mitochondrial RNA transcripts, but since the latter do not have degenerate consensus sequences 1e3, the mechanism of this packaging should be different from how YB-1 packages cytoplasmic mRNAs containing such sequences. All of the above may indicate YB-1 as a possible regulator of mitochondrial biogenesis, although this issue certainly requires a separate study.

5. Conclusions

Our findings provide a new insight into the interactions of YB-1 with cellular RNAs, highlighting mRNAs that encode mainly pro- teins involved in stress response, as the major mRNA partners of YB-1. These findings point to three types of degenerate consensus sequences present in the 30-UTRs of the large pool of human mRNAs as potential motifs that may be responsible for the YB-1- mediated packaging of these mRNAs and hint at the possible role of YB-1 in the formation of ribonucleoproteins suitable for transfer into exosomes. Therefore, the next frontier of research may be obtaining direct evidence of the participation of degenerate consensus sequences 1e3 in the above events, along with deter- mining the respective mechanisms.

References

[1] I.A. Eliseeva, E.R. Kim, S.G. Guryanov, L.P. Ovchinnikov, D.N. Lyabin, Y-box- binding protein 1 (YB-1) and its functions, Biochem. Mosc 76 (2011) 1402e1433.
[2] D.N. Lyabin, L.P. Ovchinnikov, I.A. Eliseeva, YB-1 protein: functions and regulation, Wiley Interdiscip. Rev. RNA 5 (2014) 95e110.
[3] E.E. Alemasova, O.I. Lavrik, At the interface of three nucleic acids: the role of RNA-binding proteins and poly(ADP-ribose) in DNA repair, Acta Naturae 9 (2017) 4e16.
[4] J.A. Lindquist, J.R. Mertens, Cold shock proteins: from cellular mechanisms to pathophysiology and disease, Cell Commun. Signal. 16 (2018) 63.
[5] P. Kumari, P.C. Gilligan, S. Lim, L.D. Tran, S. Winkler, P. Philp, K. Sampath, An essential role for maternal control of Nodal signaling, eLife 2 (2013), e00683.
[6] A. Zaucker, A. Nagorska, P. Kumari, N. Hecker, Y. Wang, S. Huang, L. Cooper, L. Sivashanmugam, S. VijayKumar, J. Brosens, J. Gorodkin, K. Sampath, Translational co-regulation of a ligand and inhibitor by a conserved RNA element, Nucleic Acids Res. 46 (2018) 104e119.
[7] M.J. Shurtleff, M.M. Temoche-Diaz, K.V. Karfilis, S. Ri, R. Schekman, Y-box protein 1 is required to sort microRNAs into exosomes in cells and in a cellfree reaction, eLife 5 (2016), e19276.
[8] M.J. Shurtleff, J. Yao, Y. Qin, R.M. Nottingham, M.M. Temoche-Diaz, R. Schekman, A.M. Lambowitz, Broad role for YBX1 in defining the small noncoding RNA composition of exosomes, Prot. Nat. Acad. Sci. 114 (2017) E8987eE8995.
[9] C.P. Kloks, C.A. Spronk, E. Lasonder, A. Hoffmann, G.W. Vuister, S. Grzesiek, C.W. Hilbers, The solution structure and DNA-binding properties of the cold shock domain of the human Y-box protein YB-1, J. Mol. Biol. 316 (2002) 317e326.
[10] S.G. Guryanov, V.V. Filimonov, A.A. Timchenko, B.S. Melnik, H. Kihara, V.P. Kutyshenko, L.P. Ovchinnikov, G.V. Semisotnov, The major mRNP protein YB-1: structural and association properties in solution, Biochim. Biophys. Acta 1834 (2012) 559e567.
[11] V.M. Evdokimova, C.L. Wei, A.S. Sitikov, P.N. Simonenko, O.A. Lazarev, K.S. Vasilenko, V.A. Ustinov, J.W. Hershey, L.P. Ovchinnikov, The major protein of messenger ribonucleoprotein particles in somatic cells is a member of the Y-box binding transcription factor family, J. Biol. Chem. 270 (1995) 3186e3192.
[12] M.A. Skabkin, O.I. Kiselyova, K.G. Chernov, A.V. Sorokin, E.V. Dubrovin, I.V. Yaminsky, V.D. Vasiliev, L.P. Ovchinnikov, Structural organization of mRNA complexes with major core mRNP protein YB-1, Nucleic Acids Res. 32 (2004) 5621e5635.
[13] D.A. Kretov, P.A. Curmi, L. Hamon, S. Abrakhi, B. Desforges, L.P. Ovchinnikov, D. Pastre, mRNA and DNA selection via protein multimerization: YB-1 as a case study, Nucleic Acids Res. 43 (2015) 9457e9473.
[14] D.A. Kretov, M.-J. Clement, G. Lambert, D. Durand, D.N. Lyabin, G. Bollot, C. Bauvais, A. Samsonova, K. Budkina, R.C. Maroun, L. Hamon, A. Bouhss, E. Lescop, F. Toma, P.A. Curmi, A. Maucuer, L.P. Ovchinnikov, D. Pastre´, YB-1, an abundant core mRNA-binding protein, has the capacity to form an RNA nucleoprotein filament: a structural analysis, Nucleic Acids Res. 4 (2019) 3127e3141.
[15] E. Stickeler, S.D. Fraser, A. Honig, A.L. Chen, S.M. Berget, T.A. Cooper, The RNA binding protein YB-1 binds A/C-rich exon enhancers and stimulates splicing of the CD44 alternative exon v4, EMBO J. 20 (2001) 3821e3830.
[16] N. Skoko, M. Baralle, E. Buratti, F.E. Baralle, The pathological splicing mutation c.6792C > G in NF1 exon 37 causes a change of tenancy between antagonistic splicing factors, FEBS Lett. 582 (2008) 2231e2236.
[17] O.A. Skabkina, D.N. Lyabin, M.A. Skabkin, L.P. Ovchinnikov, YB-1 auto regu- lates translation of its own mRNA at or prior to the step of 40S ribosomal subunit joining, Mol. Cell. Biol. 25 (2005) 3317e3323.
[18] D. Ray, H. Kazan, E.T. Chan, L. Pena Castillo, S. Chaudhry, S. Talukder, B.J. Blencowe, Q. Morris, T.R. Hughes, Rapid and systematic analysis of the RNA recognition specificities of RNA-binding proteins, Nat. Biotechnol. 27 (2009) 667e670.
[19] W.-J. Wei, S.-R. Mu, M. Heiner, X. Fu, L.-J. Cao, X.-F. Gong, A. Bindereif, J. Hui, YB-1 binds to CAUC motifs and stimulates exon inclusion by enhancing the recruitment of U2AF to weak polypyrimidine tracts, Nucleic Acids Res. 40 (2012) 8622e8636.
[20] S.-L. Wu, X. Fu, J. Huang, T.-T. Jia, F.-Y. Zong, S.-R. Mu, H. Zhu, Y. Yan, S. Qiu, Q. Wu, W. Yan, Y. Peng, J. Chen, J. Hui, Genome-wide analysis of YB-1-RNA interactions reveals a novel role of YB-1 in miRNA processing in glioblas- toma multiforme, Nucleic Acids Res. 43 (2015) 8516e8528.
[21] F. Hinze, P. Drewe-Boß, M. Milek, U. Ohler, M. Landthaler, M. Gotthardt, Expanding the map of protein-RNA interaction sites via cell fusion followed by PAR-CLIP, RNA Biol. 15 (2017) 359e368.
[22] O.A. Kossinova, A.V. Gopanenko, S.N. Tamkovich, O.A. Krasheninina, A.E. Tupikin, E. Kiseleva, D.D. Yanshina, A.M. Malygin, A.G. Ven’yaminova, M.R. Kabilov, G.G. Karpova, Cytosolic YB-1 and NSUN2 are the only proteins recognizing specific motifs present in mRNAs enriched in exosomes, Biochim. Biophys. Acta Protein Proteonomics 1865 (2017) 664e673.
[23] A.O. Batagov, V.A. Kuznetsov, I. Kurochkin, Identification of nucleotide SU056 pat- terns enriched in secreted RNAs as putative cis-acting elements targeting them to exosome nano-vesicles, BMC Genomics 12 (2011) S18.
[24] D.D. Yanshina, O.A. Kossinova, A.V. Gopanenko, O.A. Krasheninina, A.M. Malygin, A.G. Venyaminova, G.G. Karpova, Structural features of the interaction of the 3’-untranslated region of mRNA containing exosomal RNA- specific motifs with YB-1, a potential mediator of mRNA sorting, Biochimie 44 (2018) 134e143.
[25] M. Hafner, M. Landthaler, L. Burger, M. Khorshid, J. Hausser, P. Berninger, A.Rothballer, M. Ascano Jr., A.C. Jungkamp, M. Munschauer, A. Ulrich, G.S. Wardle, S. Dewell, M. Zavolan, T. Tuschl, Transcriptome-wide identi cation of RNA-binding protein and microRNA target sites by PAR-CLIP, Cell 141 (2010) 129e141.
[26] A. Kawaguchi, K. Matsumoto, K. Nagata, YB-1 functions as a porter to lead influenza virus ribonucleoprotein complexes to microtubules, J. Virol. 86 (2012) 11086e11095.
[27] A.V. Gopanenko, A.M. Malygin, A.E. Tupikin, P.P. Laktionov, M.R. Kabilov, G.G. Karpova, Human ribosomal protein eS1 is engaged in cellular events related to processing and functioning of U11 snRNA, Nucleic Acids Res. 45 (2017) 9121e9137.
[28] A.R. Quinlan, I.M. Hall, BEDTools: a flexible suite of utilities for comparing genomic features, Bioinformatics 26 (2010) 841e842.
[29] T.L. Bailey, M. Bode´n, F.A. Buske, M. Frith, C.E. Grant, L. Clementi, J. Ren, W.W. Li, W.S. Noble, Meme suite: tools for motif discovery and searching, Nucleic Acids Res. 37 (2009) W202eW208.
[30] N.T. Ingolia, S. Ghaemmaghami, J.R.S. Newman, J.S. Weissman, Genome-wide analysis in vivo of translation with nucleotide resolution using ribosome profiling, Science 324 (2009) 218e223.
[31] M. Lawrence, W. Huber, H. Page`s, P. Aboyoun, M. Carlson, R. Gentleman, M.T. Morgan, V.J. Carey, Software for computing and annotating genomic ranges, PLoS Comput. Biol. 9 (2013), e1003118.
[32] C. Sievers, T. Schlumpf, R. Sawarkar, F. Comoglio, R. Paro, Mixture models and wavelet reveal high confidence RNA-protein interaction sites in MOV10 PAR- CLIP data, Nucleic Acids Res. 30 (2012) e160.
[33] Q. Li, Y. Uemura, Y. Kawahara, Cross-linking and immunoprecipitation of nuclear RNA-binding proteins, in: S. Nakagawa, T. Hirose (Eds.), Nuclear Bodies and Noncoding RNAs, Methods in Molecular Biology, Humana Press, New York, NY, 2015, p. 1262.
[34] D. Benhalevy, S.K. Gupta, C.H. Danan, S. Ghosal, H.-W. Sun, H.G. Kazemier, K. Paeschke, M. Hafner, S.A. Juranek, The human CCHC-type zinc finger nucleic acid-binding protein binds G-rich elements in target mRNA coding sequences and promotes translation, Cell Rep. 18 (2017) 2979e2990.
[35] The Gene Ontology consortium, Expansion of the gene Ontology knowl- edgebase and resources, Nucleic Acids Res. 45 (2017). D1 D331-D338.
[36] N. Sonenenberg, A.G. Hinnebusch, Regulation of translation initiation in eu- karyotes: mechanism and biological targets, Cell 136 (2009) 731e745.
[37] Y.K. Kim, L.E. Maquat, UPFront and center in RNA decay: UPF1 in nonsense- mediated mRNA decay and beyond, RNA 25 (2019) 407e422.
[38] M.N. Hossain, M. Fuji, K. Miki, M. Endoh, D. Ayusawa, Downregulation of hnRNP C1/C2 by siRNA sensitizes HeLa cells to various stresses, Mol. Cell. Biochem. 296 (2007) 151e157.
[39] E.J.F. White, G. Brewer, G.M. Wilson, Post-transcriptional control of gene expression by AUF1: mechanisms, physiological targets, and regulation, Bio- chim. Biophys. Acta 1829 (2013) 680e688.
[40] L. Twyffels, C. Gueydan, V. Kruys, Shuttling SR proteins: more than splicing factors, FEBS J. 278 (2011) 3246e3255.
[41] N. Kedersha, P. Anderson, Mammalian stress granules and processing bodies, Methods Enzymol. 431 (2007) 61e81.
[42] W.H. Yang, D.B. Bloch, Probing the mRNA processing body using protein macro arrays and “auto antigenomics”, RNA 13 (2007) 704e712.
[43] H. Onishi, Y. Kino, T. Morita, E. Futai, N. Sasaqawa, S. Ishiura, MBNL1 associates with YB-1 in cytoplasmic stress granules, J. Neurosci. Res. 86 (2008) 1994e2000.
[44] S.P. Somasekharan, A. El-Naggar, G. Leprivier, H. Cheng, S. Hajee, T.G. Grunewald, F. Zhang, T. Ng, O. Delattre, V. Evdokimova, Y. Wang, M. Gleave, P.H. Sorensen, YB-1 regulates stress granule formation and tumor progression by translationally activating G3BP1, J. Cell Biol. 208 (2015) 913e929.
[45] S.E. Calvo, C.R. Klauser, V.K. Mootha, MitoCarta 2.0: an updated inventory of mammalian mitochondrial proteins, Nucleic Acids Res. 44 (2015) 1e7.
[46] A.C. Smith, A.J. Robinson, MitoMiner v3.1, an update on the mitochondrial proteomics database, Nucleic Acids Res. 44 (D1) (2016) D1258eD1261.
[47] N.C. De Souza-Pinto, P.A. Mason, K. Hashiguchi, L. Weissman, J. Tian, D. Guay, M. Lebel, T.V. Stevnsner, L.J. Rasmussen, V.A. Bohr, Novel DNA mismatch- repair activity involving YB-1 in human mitochondria, DNA Repair (Amst.) 8 (2009) 704e719.
[48] S.F. Pearce, P. Rebelo-Guiomar, A.R. D’Souza, C.A. Powell, L. Van Haute, M. Minczuk, Regulation of mammalian mitochondrial gene expression: recent advances, Trends Biochem. Sci. 42 (2017) 625e639.
[49] E. Kwon, K. Todorova, J. Wang, R. Horos, K.K. Lee, V.A. Neel, G.L. Negri, P.H. Sorensen, S.W. Lee, M.W. Hentze, A. Mandinova, The RNA-binding protein YBX1 regulates epidermal progenitors at a posttranscriptional level, Nat. Commun. 9 (2018) 1734.