Effects of individual base-pairs on in vivo target search and destruction kinetics of small RNA

Base-pairing interactions mediate intermolecular target recognition in many biological systems and applications, including DNA repair, CRISPR, microRNA, small RNA (sRNA) and antisense oligo therapies. Even a single base-pair mismatch can cause a substantial difference in biological activity but presently we do not yet know how the target search kinetics in vivo are influenced by single nucleotide level changes. Here, we used high-throughput sequencing to identify functionally relevant single point mutants of the bacterial sRNA, SgrS, and quantitative super-resolution microscopy to probe the mutational impact on the regulation of its primary target, ptsG mRNA. Our super-resolution imaging and analysis platform allowed us to further dissect mutational effects on SgrS lifetimes, and even subtle changes in the in vivo rates of target association, kon, and dissociation, koff. Mutations that disrupt binding of a chaperone protein, Hfq, and are distal to the mRNA annealing region still decreased kon and increased koff, providing an in vivo demonstration that Hfq directly facilitates sRNA-mRNA annealing. Single base-pair mismatches in the annealing region reduced kon by 24-31% and increased koff by 14-25%, extending the time it takes to find and destroy the target mRNA by about a third, depending on whether an AU or GC base-pair is disrupted. The effects of disrupting contiguous base-pairing are much more modest than that expected from thermodynamics, suggesting that Hfq also buffers base-pair disruptions.


Sort-Seq reveals SgrS nucleotides important for target regulation
We employed a high-throughput Sort-Seq approach to identify SgrS regions important for ptsG 119 regulation. We created a low copy number reporter plasmid containing a partial ptsG sequence heatmap grid (Fig. 2d) show red or dark red for G176C and G178C, and white or light blue for 151 C174G and G170C, validating our Sort-Seq results. We also see that SgrS regulation is 152 hampered if there are mutations in the small stem-loop region (nts 183-196 (Fig 1b)), the 153 terminator stem-loop region (nts 199-219) and the poly-U tail (nts 220-227). The largest effect is 154 8 seen in the stem region of the terminator stem-loop, C199 to G205 and C213 to G219, where we 155 see the darkest red, highlighting the importance of this stem-loop. These stem-loop regions and 156 the poly-U tail play a role in Hfq binding 27,28 , and our Sort-Seq analysis therefore confirms that 157 Hfq interaction is important for SgrS function in the cell. 158 Based on Sort-Seq results, we picked nine single point mutations for further investigation. These  SgrS mutation effects on regulation of ptsG reporter 164 To examine the effect of the selected SgrS mutations on ptsG regulation, we monitored the effect 165 of wild-type and seven of the SgrS mutants (plasmid-encoded and expressed from an inducible 166 promoter) on the activity of a chromosomal ptsG'-'lacZ translational fusion (Fig. 3a). The wild-167 type SgrS almost completely eliminated β-galactosidase activity whereas the mutants showed 168 regulation defects of various degrees consistent with the Sort-Seq data. SgrS G215A, which 169 disrupts the terminator stem-loop structure, showed the largest defect. To test if the regulatory 170 defects can be explained by a reduction of SgrS levels, for example, due to shorter cellular 171 lifetimes associated with impaired Hfq binding, we performed Northern blot analysis. We found 172 that SgrS abundance is not affected for four of the mutants (A177U, G178U, G178A and 173 G184A) and is reduced by 40-50% for mutations in the terminator stem-loop or poly-U tail 174 (G215A, U224G and U224A) (Fig. 3c). Interestingly, the latter three mutants showed large 175 increases in readthrough transcription, suggesting that transcription termination is defective (  Hfq. These mutant alleles in the background of strains with wild-type RNase E or a C-terminally 191 truncated RNase E were grown, and glucose-phosphate stress was induced using αMG for a 192 varied amount of time before cell fixation and permeabilization. We performed two-color 3D 193 super-resolution imaging of the SgrS sRNAs labeled with up to 9 FISH probes conjugated to 194 Alexa Fluor 647 and the ptsG mRNAs labeled with up to 28 FISH probes conjugated to CF568. 195 ∆sgrS and ∆ptsG strains were also examined to correct for the background arising from 196 nonspecific binding of FISH probes. The wild-type strain showed an increase in SgrS copy 197 number over time after sugar stress induction (Fig. 4, Supplementary Fig. 3). At the same time, 198 the copy number of ptsG mRNA showed a decrease (Fig. 4, Supplementary Fig. 3). We used a 199 density-based clustering algorithm 34 to determine the copy numbers of RNAs along with the 200 copy number of SgrS-ptsG mRNA complexes. Super-resolution imaging was especially 201 important for quantifying sRNA-mRNA complexes because at conventional microscopy 202 resolution there was too much false co-localization between sRNA and mRNA.

203
The accumulation of the mutant SgrS sRNAs was lower than for wild-type SgrS with an 204 accompanying impairment in ptsG mRNA degradation for all single point mutants examined, 205 showing that their regulatory functions are perturbed (Fig. 4, Supplementary Fig. 4-9, 11-13).

206
The single cell distribution of RNA copy numbers also showed a decreased accumulation of 207 SgrS, with the histogram peaking at lower copy numbers 20 min after αMG induction, and the 208 histograms for ptsG mRNA peaked at higher copy numbers per cell compared to the wild-type 209 (Fig. 5b, d, h, Supplementary Fig. 52). The lowest accumulation of SgrS was seen for G184A 210 and G215A, and they also showed the most impaired mRNA degradation (Fig. 4, 5a, c, g,  These imaging data by themselves cannot tell us whether regulatory defects are due to changes in 218 target binding kinetics or due to changes in the SgrS stability. Therefore, we next determined the 219 lifetimes of wild-type and mutant SgrS molecules.      Fig. 14-24), we used a previously developed deterministic kinetic 267 model to describe the SgrS-ptsG regulation kinetics (Fig. 1a) 34 . We used the experimentally-  for the wild-type strain was (1.9 ± 0.2) x 10 5 M -1 -s -1 and was 0.22 ± 0.02 s -1 giving a 273 of 1.16 ± 0.14 µM, comparable to the previously published results (Fig. 6b-d) 34 . was 274 lower for all single point mutants compared to the wild-type and the reduction ranged from 24% 275 for A177U to 53% for G184A. was higher for all the mutants and the increase ranged from 276 14% for A177U to 33% for G184A giving a dissociation constant, of 1.67 ± 0.13 µM and   its targets, and RNA secondary structures 27,33,48,49 . Hfq has also been shown to promote structural 347 changes to the RNAs, which in turn helps in the annealing and, consequently, regulation 50-52 .

348
Our study provides a quantitative description of the process of target search and off-target reporter sequence into the low copy plasmid pAS05 between the XhoI and XbaI restriction sites 413 and the expression of the reporter system was under the control of PLlac-O1.

414
The SgrS sRNA sequence was inserted in between the NdeI and BamHI restriction sites of the 415 medium copy plasmid pZAMB1 and its expression was under the control of PLtet-O1. The sgrS 416 mutation library was prepared by using the plasmid pZAMB1 as a template for mutagenesis PCR 417 and also as a vector to insert the sgrS mutation sequence.  The E. coli strain to be sorted was cultured overnight in LB Broth with appropriate antibiotics.

464
The following day, the liquid culture was diluted 200-fold and cultured with antibiotics until   that of the plasmid used as error-prone PCR template. 513 We normalized the raw number of reads of each group by both the total number of reads and the 514 fraction of cells falling under each gate. As such, we defined a weighted average intensity to 515 each individual mutant along this 106 base long SgrS segment that we probed. Referred to as the 516 "intensity moment" from now on, we calculated the following quantity:    Table 2). Bacterial strains were cultured and β-galactosidase assay performed as described above.

567
Simultaneously, aliquots of the same culture were taken and total RNA was extracted as  The raw data was acquired using the Python-based acquisition software and it was analyzed 687 using a data analysis algorithm which was based on work published previously by Zhuang's 688 group. 75,76 The peak identification and fitting were performed using the method described 689 before. 34 The z-stabilization was done by the CRISP system and the horizontal drift was  Fig. 41b). The coefficient, a, was used a 718 correction factor for colocalization calculation as, final colocalization = calculated colocalization 719 -a*SgrS copy number. The ptsG mRNA degradation rates were calculated using a rifampicin-chase experiment. The The mass-action equations used for the wild-type E. coli cells and the chromosomal mutations 747 are shown below:  The transcription rate of ptsG mRNA was calculated using = × [ ] 0 and in this equation 766 [ ] 0 is the concentration of ptsG mRNA before the induction of sugar stress in all of the cases.

767
This was done because it was observed previously 34    G184A-C195U mutant strain, (g) G215A mutant strain. Average copy numbers per cell are plotted against time. Rate constants obtained for these mutants are shown in Figure 6 and in