Methods
Indeed , other putative targets identified by MAPS still need validation . 
Surprisingly , in Lalaouna et al. [ 1 ] , unusual binding partners were also co-purified with MS2-sRNA , such as transcribed spacer sequences of polycistronic tRNA transcripts ( glyW-cysT-leuZ and metZ-metW-metV ) . 
For a long time , these sequences were considered as non-functional `` junk RNA '' . 
Nevertheless , we demonstrated that , during glyW-cysT-leuZ pre-tRNA processing , the 30external transcribed spacer sequence ( 30ETSleuZ ) is excised and acts as a sRNAs sponge ( RyhB and RybB ) preventing them from binding their real targets [ 1 ] . 
Thus , 30ETSleuZ represents the first functional tRNA-derived RNA fragment ( tRF ) in bacteria ( see reviews [ 6 -- 8 ] for more details on tRFs ) . 
Evidence also showed that RybB sRNA interacts with another kind of tRFs : the internal transcribed spacers ( ITS ) of the metZ-metW-metV pre-tRNA [ 1,9 ] . 
We first noticed a strong enrichment of both ITSmetZ-metW/ITSmetW-metV with MS2-RybB [ 1 ] . 
Then , we performed the reverse experiment using each ITS as bait and were able to co-purify RybB sRNA confirming that ITSmetZ-metW/ITSmetW-metV and RybB interact in vivo [ 9 ] . 
However , the related mechanism of regulation is still under investigation . 
Taken together , these results demonstrated that , unlike other methods commonly used to identify targets ( e.g. microarray , Northern blot , RNAseq ) , MAPS can depict the whole interactome of a specific RNA as it identifies all kinds of interacting partners . 
In this report , we concisely describe adjustments made to MAPS to study various kind of RNA : RNA interactions . 
We also bring important modifications to MAPS protocol to be suitable for patho-genic or Gram-positive strains . 
Finally , we discuss the future of this technology 
Recently , we developed an in vivo technology , called MS2-affinity purification coupled with RNA sequencing ( MAPS ) , to tackle the problem of the identification of RNA : RNA interactions [ 1 ] . 
Briefly , the technique is based on the strong affinity of the MS2 protein for the MS2 RNA aptamer used to tag an RNA of interest ( see review [ 2 ] for more details on the history of RNA tagging ) . 
The MAPS technology was first used in two recent reports studying sRNA-mRNA interactions [ 1,3 ] . 
Regulatory small RNAs ( sRNAs ) are involved in the adaptation of bacteria facing environmental fluctuations by post-transcriptionally regulating a subset of mRNAs [ 4,5 ] . 
Using MAPS , we already drew the interacting map of RyhB , RybB and DsrA , three well-characterized sRNAs in Escherichia coli . 
Here , we confirmed previously known targets but also revealed new ones [ 1,3 ] . 
For example , we characterized grxD mRNA , involved in iron-sulfur cluster biosynthesis , as negatively regulated by RyhB sRNA [ 1 ] . 
In the same report , we validated yifE , encoding a conserved protein of unknown function , as a RybB positively regulated target . 
Moreover , we characterized a new atypical mRNA target of DsrA , coding for the ribose pyranase ( RbsD ) [ 3 ] . 
Contrary to the canonical mechanism of repression , DsrA inhibits translation of rbsD mRNA by interacting in the coding sequence and induces mRNA decay by an unknown ribonuclease ( RNase ) . 
However , we only touched the tip of the iceberg in these reports . 
⇑ Corresponding author at : Department of Biochemistry , Université de Sherbrooke , 3201 Jean Mignault Street , Sherbrooke , QC J1E 4K8 , Canada . 
E-mail address : eric.masse@usherbrooke.ca ( E. Massé ) . 
2. Materials and methods
2.1. MS2 tagging
We usually fused the 50-end of the RNA fragment of interest with the MS2 RNA aptamer . 
We recommend 50-end MS2 tagging in order to keep the endogenous terminator and to minimize secondary structure changes . 
To obtain the corresponding chimeric DNA sequence cloned in a plasmid under the control of a pBAD promoter ( arabinose-inducible promoter ) , we have first inserted MS2 stemloops into a pNM12 vector using MscI and EcoRI restriction enzymes ( NEB ) . 
Then , the DNA sequence corresponding to the RNA of interest is added using EcoRI and SphI ( Fig. 1A ) . 
The secondary structure of the chimeric RNA is verified using mfold software ( http://unafold.rna.albany.edu/?q=mfold ) . 
In the case of 50UTR-mRNA ( untranslated region of mRNA ) , we added MS2 stemloops at the 30end ( commonly at the beginning of the coding sequence ) followed by a T7 transcriptional terminator ( Fig. 1B ) . 
The MS2 aptamer and the T7 terminator are added to the 50UTR of mRNA using nested PCR . 
A stop codon is inserted in frame , just before the MS2 aptamer sequence to interrupt protein translation . 
In all cases , we used the untagged RNA as a control . 
We performed Northern blot analysis to verify that the MS2-RNA construct is expressed at a level similar to the untagged RNA . 
For sRNA , we also tested the activity of the MS2-sRNA construct on known mRNA targets . 
This activity has to be equivalent to the control . 
2.2. Strains
Typically , we performed MS2-sRNA MAPS in a DsRNA rne131 ( RNA degradosome assembly mutant ) background . 
Both mutations maximize mRNA targets enrichment by eliminating interaction with the endogenous untagged sRNA and slowing down mRNA targets decay . 
Indeed , the use of a wild-type background reduces the enrichment ratio of negatively regulated targets ( data not shown ) . 
In the case of MS2-tRFs and 50UTR-mRNA-MS2 , we did not delete the endogenous gene to avoid potential disruptions of cellular metabolism . 
2.3. MS2-MBP protein purification
The plasmid pHMM , first published in Batey and Kieft [ 10 ] , encodes the DNA sequence required to produce the 59 kDa polypeptide His6-MS2-MBP . 
This hybrid protein is composed of an N-terminal hexahistidine-tag , a maltose-binding protein ( MBP ) domain , and a C-terminal MS2 coat protein . 
The MS2-MBP protein purification was performed as described by Batey and Kieft [ 10 ] with major modifications . 
BL21/pLysS E. coli cells containing pHMM plasmid were first grown in 1 L of rich medium ( Luria Bertani medium , LB ) supplemented with 30 lg/mL chloramphenicol and 10 lg/mL kanamycin at 37 C to an OD600nm of 0.7 . 
At this point , the construct was expressed by addition of 0.5 mM IPTG . 
After 3 h of induction , cells were harvested and centrifuged ( 3800g for 15 min at 4 C ) . 
Cells pellets were resuspended in 50 mL of Lysis buffer ( 50 mM NaH2PO4 pH 8.0 , 300 mM NaCl , 0.5 % Tween 20 , 10 mM imidazole and 10 % glycerol ) supplemented with 50 ll of protease inhibitor cocktail ( 1 mg/mL ; Sigma-Aldrich ) and 5 mL of lysozyme ( 10 mg/mL ) . 
Then , cells were lysed by sonication ( 25 % for 4 min 5 s sonication , 5 s pause on ice ; Branson Digital Sonifier ) and centrifuged ( 17,000 g for 45 min ) . 
A second step of sonication is required if the supernatant remains too viscous . 
After the lysis step , the His6-MS2-MBP protein was purified using a Ni-NTA agarose column ( 4 mL of Ni-NTA agarose resin ( Qiagen ) on a 25 mL Econo-Pac chromatography column ( Biorad ) ) equilibrated with 25 mL of lysis buffer . 
The supernatant was loaded onto the column and the column was washed with 25 mL of lysis buffer . 
Samples were eluted by addition of 4 mL of Lysis buffer supplemented with 250 mM imidazole . 
After a dialyze step in 1 L of Dialysis buffer ( 25 mM Na-MES pH 6.0 , 25 mM NaCl ) , samples were loaded on a amylose resin column ( 3 mL of amylose resin ( NEB ) on a 25 mL Econo-Pac chromatography column ) equilibrated with 25 mL of Column buffer ( 20 mM Tris-HCl pH 7.4 , 200 mM NaCl , 0.5 mM EDTA ) . 
The column was wash with 36 mL of Column buffer . 
Samples were eluted with 6 mL of Column buffer supplemented with 10 mM maltose . 
The output was dialyzed in 1 L of buffer A ( 20 mM Tris-HCl pH 8.0 , 150 mM KCl , 1 mM MgCl2 ) supplemented with 10 % glycerol . 
Finally , the protein concentration was calculated using OD280nm and a molar extinction coefficient of 83.310 M 1 cm 1 . 
2.4. MS2-affinity purification coupled with RNA sequencing
A schematic representation of MAPS technology is available in Fig. 2 . 
Cells were grown in LB supplemented with 50 lg/mL ampicillin ( diluted 1/1000 from an overnight culture grown ) . 
Note that MAPS was successfully performed in other media ( e.g. M63 minimal medium ) ( data not shown ) . 
Cells were harvested in ( A ) exponential ( OD600nm = 0.5 ; 100 mL ) and/or ( B ) stationary phase of growth ( OD600nm > 1 ; 100 mL ) after induction of MS2 construct ( or control ) with 0.1 % arabinose for 10 min and then chilled on ice for 10 min . 
For input samples ( collected before affinity purification ) , RNA was extracted from 600 lL of culture using the hot-phenol procedure [ 11 ] . 
The remaining cells were washed with 1 mL of buffer A ( supplemented with 1 mM DTT and 1 mM PMSF ) and centrifuged . 
Cells were resuspended in ( A ) 2 or ( B ) 3 mL of buffer A and lysed using a French Press Cell Disrupter ( Thermo electron corporation ) with the following parameters : 430 psi , ( A ) three or ( B ) four times . 
Next , the lysat was cleared by centrifugation ( 17,000 g for 30 min at 4 C ) . 
For protein samples ( input ) , 20 lL of the soluble fraction were collected and mixed with 20 lL of protein sample buffer ( 125 mM Tris-HCl pH 6.8 , 1 % SDS , 20 % glycerol , 0.02 % bromophenol blue and 100 mM DTT ) . 
The remaining supernatant was subjected to affinity chromatography ( all the following steps were performed at 4 C ) . 
The column was prepared by adding ( A ) 75 or ( B ) 100 lL of amylose resin ( New England Biolabs ) to Bio-Spin disposable chromatography columns ( Bio-Rad ) . 
We washed the column with 3 mL of buffer A. Next , ( A ) 100 or ( B ) 200 pmol of His6-MS2-MBP protein were immobilized on the amylose resin . 
Again , we washed the column with 2 mL of buffer A. Here , the supernatant was loaded onto the column , before being washed with ( A ) 5 or ( B ) 8 mL of buffer A . 
This crucial step requires some adaptation and adjustment in function of cell density . 
Bound RNA was eluted using 1 mL of buffer A supplemented with 15 mM maltose . 
Eluted RNA was extracted with phenol-chloroform ( V/V ) and precipitated by the addition of ethanol ( 2 vol ) and 20 mg of glycogen . 
At the same time , the organic phase was subjected to acetone precipitation to recover proteins . 
RNA samples were then analyzed by Northern blot and protein samples by Western blot as described in Lalaouna et al. [ 1 ] 
After all these steps , RNA samples ( output ) were treated with TURBOTMDNase ( 4 units ; Ambion ) for 30 min at 37 C to eliminate remaining genomic DNA . 
TURBO DNase was then removed using phenol-chloroform extraction and RNA was again precipitated . 
RNA quality and quantity were analyzed on Agilent Nano Chip on the bioanalyzer 2100 . 
cDNA libraries were prepared with ScriptSeqTM v2 RNA-Seq Library Preparation Kit ( Illumina ) for MS2-sRNA samples and with NEBNext Small RNA Library Prep set E7330S kit for MS2-ITS , MS2-ETS and 50UTR-mRNA-MS2 . 
Libraries were sequenced using Illumina MiSeq . 
In the case of MS2-30ETSleuZ ( Table 1 ) , we extracted RNA from exponential ( OD600nm = 0.5 ; 100 mL ) and stationary phase ( OD600nm = 1.3 ; 100 mL ) cultures ( WT background ) . 
We then followed the ( B ) procedure as described above . 
GEO accession numbers of MAPS data are : MS2-30ETSleuZ ( GSE79278 ) , MS2-ITSmetZW/MS2-ITSmetWV ( GSE66517 ) , MS2-RyhB ( GSE66519 ) , MS2-RybB ( GSE66518 ) , MS2-DsrA ( GSE67605 ) and MS2 control ( GSE67606 ) . 
2.5. MAPS with pathogenic and/or Gram positive strains
Recently , we adapted MAPS technology to pathogenic strains manipulation . 
Interestingly , described modifications are also effective for Gram-positive strains ( data not shown ) . 
Here , the medium used for cell growth will depend on the studied strain ( e.g. LB med-ium for Salmonella enterica serovar Typhimurium , brain-heart infusion ( BHI ) broth for Staphylococcus aureus ) and is supplemented by the appropriate antibiotic in function of the overexpression system used . 
Generally , the extraction is performed on larger volume ( 300 mL ) . 
The lysis is a crucial step and requires adaptations for safety purposes . 
Here , the Precellys 24 bead beater was used to break cells instead of French press . 
After washing , harvested cells were resuspended in 1 mL of buffer A . 
The cell suspension was then transferred to screw cap tubes containing 200 lL of glass beads ( 0.1 mm in diameter ; BioSpec ) . 
Cells were disrupted using the bead beater ( three cycles : 6500 rpm for 20 s , followed by 1 min in ice ) . 
The lysate was then centrifuged at 17,000 g for 30 min at 4 C and the supernatant was apply to the affinity purification column . 
Following steps are identical to those described in Section 2.3 . 
For input samples , two different procedures were used depending on bacterial classification . 
For Gram-negative bacteria , RNA was extracted from 1 mL of culture using hot-phenol procedure [ 11 ] . 
For Gram-positive bacteria , 1 mL of culture was centrifuged and resuspended in 500 lL of lysis solution ( 0.5 % SDS , 1 mM EDTA , 20 mM sodium acetate ) and transferred into a screw cap tube containing 200 lL of glass beads and 500 lL of phenol ( pH 4 ) . 
To break the cell wall , cells were vortexed using the bead beater ( three cycles : 6500 rpm for 20 s , followed by 1 min in ice ) . 
The sample was then centrifuged and RNA was then precipitated as described in Section 2.3 . 
2.6. Data processing
Data processing procedure is described in details in Fig. 3 . 
We used bioinformatics tools freely available on Galaxy Project platform ( https://galaxyproject.org/ ) [ 13 ] . 
The first workflow enables to align reads to the corresponding genome ( e.g. E. coli K12 ) and to visualize them using the UCSC Gen-ome Browser ( Fig. 3A ) . 
First , we used FASTQ Groomer ( Galaxy version 1.0.4 ) to convert FASTQ files . 
Second , the quality of raw sequences data was assessed using FastQC ( Galaxy version 0.52 ) . 
Third , checked sequences were aligned to the corresponding genome assembly using Map with Bowtie for Illumina ( Galaxy version 1.1.2 ) . 
To finish , we used Create a BedGraph of genome coverage ( Galaxy version 0.1.0 ) to obtain data files in a format compatible with UCSC Microbial GenomeBrowser ( http : / / microbes.ucsc.edu / ) . 
The second workflow is useful to assign reads to gene . 
All required steps are indicated in Fig. 3B . 
Briefly , we compare mapped regions with gene positions ( extracted from NCBI/GenBank ) . 
For this purpose , we formatted a gene bank file as indicated by the following example : chr 190 255 þ thrL 
After all , read counts were normalized by coverage . 
Data are presented in a tab-delimited text file which includes normalized reads and MS2-RNA/RNA ( Control ) ratio . 
2.7. Primer extension
The cleavage site that releases 30ETSleuZ was determined using primer extension ( Fig. 4 ) . 
This result was required to remove RNase E cleavage site from the MS2-30ETSleuZ construct and avoid the loss of the MS2 aptamer . 
Briefly , 20 lg of total RNA were incubated with 0.5 pmol of 32P-radiolabelled oligonucleotide ( EM3204 , CCGAAGGTGGTTT-CACGACAC ) and 1 mM dNTPs . 
After 5 min of incubation at 65 C , followed by 1 min on ice , 5X reaction buffer , 0.1 M DTT , RNase Inhibitor Murine ( 40 units , NEB ) and ProtoScript II Reverse Transcriptase ( 200 units , NEB ) were added to the reaction . 
Reverse transcription was performed for 60 min at 42 C before the enzyme was inactivated at 90 C for 10 min . 
Samples were then precipitated and migrated on a denaturing 8 % polyacrylamide gel . 
As a control , we performed the same experiment without reverse transcriptase . 
The sequencing ladder was obtained with a DNA template ( PCR with oligonucleotides EM3205 ( CTCCGGGTACCATGG-GAAAG ) and EM3206 ( CCTATCTTACATGCCGGTCCG ) ) and the same radiolabelled primer ( EM3204 ) . 
Here , we added ddNTP to stop the reaction performed by the Vent DNA polymerase ( 2 units , NEB ) as described below . 
For the G lane , we used 0.36 mM ddGTP , 0.037 mM dGTP , 0.03 mM dATP , 0.1 mM dCTP and 0.1 mM dTTP . 
For the A lane , we used 0.9 mM ddATP , 0.03 mM dATP , 0.1 mM dGTP , 0.1 mM dCTP and 0.1 mM dTTP . 
For the T lane , we used 0.72 mM ddTTP , 0.033 mM dTTP , 0.1 mM dGTP , 0.03 mM dATP and 0.1 mM dCTP . 
For the C lane , we used 0.42 mM ddCTP , 0.041 mM dCTP , 0.1 mM dGTP , 0.03 mM dATP and 0.1 mM dTTP . 
We also added 0.1 % Triton to the reaction . 
Then , we performed the following PCR steps : 1 min at 95 C , 1 min at 52 C and 1 min at 72 C ( 25 cycles ) . 
The reaction was stopped by addition of formamide loading dye and migrated next to primer extension samples 
3.1. Evolution of MAPS technology
Over the years , we modified and improved key parameters of the MS2-affinity purification [ 1,3,9,14 ] . 
We exhaustively expose these modifications in Section 2 . 
To perform MAPS with higher cellular concentration ( exponential and stationary phases or stationary phase only ) , we modified the lysis step and increased the loading capacity of the column . 
We also adapted MAPS procedure to pathogenic strains manipulation by modifying cell breakage . 
We demonstrated that the same protocol can be used for Gram-positive bacteria . 
The adjustment of MAPS technology to pathogenic and/or Gram-positive bacteria will certainly facilitate the targets identification in strains remarkably difficult to genetically manipulate . 
Deep characterization of the interactome of sRNA involved in virulence in pathogenic strains will also bring to light targets with a therapeutic potential . 
To study sRNA interacting with a 50UTR-mRNA or a tRF , we also used another library preparation kit to specifically enrich small RNA fragments . 
This approach has already been successfully used with both MS2-ITSmetZW/MS2-ITSmetWV [ 9 ] and MS2-30ETSleuZ ( Fig. 4A ) . 
Thus , we confirmed results previously obtained with MS2-RyhB and MS2-RybB [ 1 ] and hence the efficiency of MAPS technology to determine binding partners of tRNA-derived RNA fragments . 
In this case , one of the major limitations is the presence of enzymatic cleavage sites that naturally enable the maturation of pre-tRNAs and , as a consequence , the release of tRFs . 
To counter this problem and avoid loss of the MS2 aptamer , we added an additional step : we first identify the +1 of each tRF using primer extension ( see Section 2 ) . 
Next , we fuse the MS2 aptamer to the determined 50-end . 
This procedure has been validated with 30ETSleuZ ( Fig. 4B ) . 
Here , the cleavage site occurs 15 nt after the 30CCA end of leuZ gene , releasing a 53 nt-long RNA fragment ( Fig. 4C ) . 
MAPS technology was performed with pBAD-MS2 and pNM12 ( Ctrl ) . 
Cells were harvested in exponential ( OD600nm = 0.5 ; 100 mL ) and in stationary phase ( OD600nm = 1 ; 100 mL ) ( WT background ) . 
After that , the ( B ) procedure described in Section 2 was followed . 
The 50 most enriched genes are represented ( with more than 50 reads ) . 
The GEO accession number is GSE67606 . 
3.2. Removal of false positives
As previously mentioned in the Introduction , a lot of putative targets were revealed by MAPS but not verified [ 1,3 ] . 
To facilitate the analysis , we performed MAPS using MS2 aptamer only as bait to discriminate the subpopulation of genes that interact with the MS2 tag rather than the gene of interest [ 3 ] . 
The Top 50 most enriched genes are represented in Table 2 . 
The complete list is also available ( GEO accession number GSE67606 ) . 
This list is useful to eliminate false positives and reduce the number of candidates . 
For example , ompW gene is often co-purified with MS2-sRNA without being regulated by them ( data not shown ) . 
3.3. Perspectives
In bacteria , the MS2-affinity purification was successfully used to co-purify the RNA chaperone protein Hfq with sRNAs , 50UTR-mRNA and tRFs [ 1,3,15 ] . 
Others proteins are known to be involved in sRNA-mediated regulation like various ribonucleases [ 16 ] . 
Due to transitory interaction , MS2-pulldown assay failed to pick up RNase E with a sRNA : mRNA complex reproducibly ( data not shown ) . 
To improve RNase recovery , an additional step of UV crosslink should be performed before cell lysis . 
Especially , application of the UV crosslink to MS2-affinity purification coupled with mass spectrometry should enable the identification of still unknown ribonucleases as in the case of rbsD mRNA [ 3 ] or the characterization of sRNA sequestering proteins by mimicking their recognition site . 
Recently , McaS sRNA , first described as regulator of multiple mRNAs through base pairing , was also shown to directly interact with the pleiotropic regulatory protein CsrA and alleviate its activity [ 17 ] . 
Hence , we can easily assume that other well-characterized sRNAs could have a dual function . 
Certainly , UV crosslink MS2-affinity purification coupled with mass spectrometry will be complementary to recently published method where authors opted for the opposite approach by using protein as bait [ 18 -- 20 ] . 
Recently , Wade 's group adapted ribosome profiling technology to the identification of sRNA targets [ 21 ] . 
Using RNA sequencing , they compared ribosome-protected mRNA fragments profiles in presence or absence of sRNA . 
Thus , ribosome profiling could be performed in parallel to MAPS . 
The combination of these two methods will represent a perfect screen for sRNA targetome characterization . 
Finally , in Desnoyers and Masse [ 14 ] we demonstrated that MS2-affinity purification can be used to perform co-variational mutagenesis instead of using classical experiments ( e.g. in vitro probing or b-galactosidase ) to prove direct interaction . 
Mutations were introduced at either the sRNA pairing site or the Hfq pairing site in the mRNA sequence , which allowed us to characterize the Hfq-mRNA ribonucleoprotein complex . 
During the last few years , we performed MAPS with various kind of RNA : sRNA [ 1,3 ] , 50UTR-mRNA [ 3 ] and tRFs [ 1,9 ] . 
We overcame technical limitations of MAPS to apply it to all kind of RNA : RNA interaction regardless of the studied organism . 
Especially , extension of MAPS technology to pathogenic strains will help to explore the targetome of sRNAs involved in bacterial virulence and , therefore , increase the reservoir of potential molecules for antibacterial design . 
Moreover , the combination of MAPS with other high-throughput technologies will pave the way for an easier sRNA functional screening . 
This work has been supported by an operating grant from the Canadian Institutes of Health Research ( CIHR ) to EM .