Biochemical and genetic dissection of the RNA-binding surface of the FinO domain of Escherichia coli ProQ

The concave-face pocket of the FinO domain is conserved in E. coli ProQ

We previously used an in vivo B3H assay to examine the effects of mutations on ProQ–RNA interactions (Fig. 1A; see Materials and Methods for detailed description). One mutation that caused very strong defects in RNA interaction was R80A. To further explore the role arginine 80 plays in RNA binding, and whether its role goes beyond electrostatics, we created an R80K substitution, replacing arginine with the other basic residue, lysine. We tested RNA interactions of both R80A and R80K ProQ variants using our B3H assay with three RNA ligands: cspE-3′, SibB, and malM-3′. Surprisingly, even the very conservative substitution of arginine to lysine at position 80 eliminated RNA binding to a similar extent as an alanine for all RNAs tested (Fig. 1B,C), even as western blotting showed that all three variants of α-ProQ were expressed to similar extents (Fig. 1C; quantification and loading controls in Supplemental Fig. S3).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 1.

A positive charge at conserved position 80 is not sufficient for RNA interaction in vivo or in vitro. (A) Design of B3H system to detect the interaction between ProQ and an RNA (malM 3′UTR). Interaction between the protein moiety and RNA moiety fused, respectively, to the NTD of the alpha subunit of RNA polymerase (RNAP) (α) and to one copy of the MS2 RNA hairpin (MS2^hp) activates transcription from test promoter, which directs transcription of a lacZ reporter gene (Berry and Hochschild 2018). Compatible plasmids direct the synthesis of the α-fusion protein (under the control of an IPTG-inducible promoter), the CI-MS2^CP adapter protein (under the control of a constitutive promoter; pCW17) and the hybrid RNA (under the control of an arabinose-inducible promoter). (B) Results of β-galactosidase (β-gal) assays performed in Δhfq plac-O_L2–62 reporter strain cells containing three compatible plasmids: one (α-ProQ) that encoded α (−) or the α-ProQ^ΔCTD (pKB955; residues 2–176) fusion protein (WT or an R80A or R80K mutant), another (CI-MS2^CP) that encoded λCI (−) or the λCI-MS2^CP fusion protein (+), and a third (Bait) that encoded a hybrid RNA with the 3′-UTR of malM (pKB1210) following one copy of an MS2^hp moiety (+) or an RNA that contained only the MS2^hp moiety (−). Cells were grown in the presence of 0.2% arabinose and 50 µM IPTG (see Materials and Methods). Bar graphs show the averages of three independent measurements and standard deviations. (C) (Left) Results of B3H assays detecting interactions between α-ProQ^ΔCTD and three RNA baits. β-gal assays were performed as in (B) but with three-hybrid RNA constructs (MS2^hp-malM-3′, MS2^hp-cspE-3′, MS2^hp-SibB). The bar graph shows the fold-stimulation over basal levels as averages and standard deviations of values collected from three independent experiments conducted in triplicate across multiple days. (Right) Western blot to compare steady-state expression levels of mutant α-ProQ^ΔCTD fusion proteins. Lysates were taken from the corresponding β-gal experiment containing MS2^hp-malM-3′ and all other hybrid components at 50 µM IPTG. Following electrophoresis and transfer, membranes were probed with anti-ProQ antibody. See Supplemental Figure S3 for loading controls and quantification of results.

The inability of lysine to substitute for arginine at residue 80 is analogous to the strongly deleterious effect we previously observed of a conservative Y70F substitution on RNA binding in vivo (Pandey et al. 2020). We wondered whether the positive charge and aromatic ring were not the critical features of this particular arginine and tyrosine, respectively. To explicitly test whether any other amino acid could support RNA binding at either position 70 or 80 in the FinO domain of E. coli ProQ, we constructed saturation mutagenesis libraries at each of these positions and screened for colonies showing any amount of blue over negative controls with malM-3′ RNA as bait in the B3H assay. Sequencing results of isolated plasmids showed that all codons encoding arginine at position 80 and tyrosine at position 70 were recovered (Supplemental Table S1), but no other substitutions at either position were found to support detectable binding in vivo. This confirms that both Y70 and R80 are uniquely required for RNA interaction. Indeed, these residues are both highly conserved across FinO-domain sequences and are located near one another in a concave pocket in most solved structures of FinO domains (Supplemental Figs. S1, S2).

Given that both Y70 and R80 were important for E. coli ProQ’s interaction with RNA, it was important to clarify if these residues are indeed located on opposite faces of the protein as previously suggested (Gonzalez et al. 2017; Pandey et al. 2020), especially because these residues are found on the same face of the FinO domain in the structures of other homologs (Supplemental Fig. S2; Ghetu et al. 2000; Chaulk et al. 2010; Immer et al. 2020; Kim et al. 2022). We reasoned that the chemical change in ProQ^R80K was minor enough that it should be possible to rescue with compensatory mutations of nearby residues—and that the location of these compensatory substitutions would provide evidence regarding the likely structural position for R80 in the structure of E. coli ProQ found in vivo. To create compensatory mutations that could rescue RNA binding of ProQ^R80K, we selected three regions of the primary sequence that were close to the proposed positions of R80 in the NMR structure of E. coli ProQ (convex face) or in other FinO domain proteins (concave face; Fig. 2A; Supplemental Fig. S2). We then used a saturation mutagenesis strategy (Supplemental Fig. S4A) to create a library of all possible single point-mutations at each of these amino acid positions on the pPrey-ProQ^R80K plasmid.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Saturation mutagenesis screen to identify compensatory mutants that rescue RNA-binding activity of ProQ^R80K. (A) Residues selected for saturation mutagenesis are highlighted in yellow, green, and cyan on the NMR structure of ProQ^NTD (Gonzalez et al. 2017). Arginine 80 is shown in orange sticks in the convex-face position found in the NMR structure and in pink sticks in a hypothetical concave-face position analogous to its position in other orthologs. (B) Compensatory mutations found in pPrey-ProQ^R80K plasmids that produced blue colonies on X-gal-indicator plates with pBait-malM-3′ are listed. Codons identified in multiple colonies are indicated by 2× and 3×. (C) Results confirming the effects of compensatory mutants in a plate-based B3H assay, detecting interactions between variants of ProQ^ΔCTD with malM-3′ RNA. β-gal assays were performed with Δhfq reporter strain cells containing three compatible plasmids: one that encoded the CI-MS2^CP fusion protein, another that encoded α or an α-ProQ^ΔCTD fusion protein (wild-type, WT, or the indicated mutant), and a third that encoded a hybrid RNA (MS2^hp-malM-3′) (see Materials and Methods). Quantification of color intensity shown in Supplemental Figure S4B. (D,E) The two strongest compensatory substitutions (V74K and L91V) are shown on (D) the NMR structure and (E) the AlphaFold Model (Varadi et al. 2022) of ProQ^NTD. Side chains were mutated in PyMol to the lowest-energy rotamer to show predicted structures of the compensatory substitutions. Distances between the terminal atom of each side chain (R80K, V74K, L71V) are shown in Angstroms and visualized with dashed lines.

To screen for compensatory mutations that could partially rescue RNA binding of ProQ^R80K, saturation mutagenesis libraries were transformed into B3H reporter cells containing pAdapter and pBait-malM plasmids. Transformants were plated on X-gal-indicator plates, and plasmids were purified and sequenced from colonies that were more blue than a pPrey-empty control. Mutations that partially rescued ProQ^R80K-malM binding were identified from each of the three mutagenesis libraries constructed (Fig. 2B); multiple mutations were identified at the same amino acid positions (V74 and L91) and several of the mutant plasmids were isolated and sequenced multiple times. When isolated plasmids were retransformed into fresh B3H reporter cells, many of the initial hits showed above-background interaction with malM-3′ RNA in qualitative plate-based assays (Fig. 2C); substitutions at V74 and L91 produced statistically significant effects when the colors from plate-based phenotypes were quantitatively analyzed (Supplemental Fig. S4B). The strongest of these substitutions also resulted in statistically significant but modest rescue of ProQ^R80K binding to malM-3′ in liquid β-gal assays (L91V, V74K; Supplemental Fig. S4C) and to cspE-3′ and SibB RNAs in plate-based assays (Supplemental Fig. S4D,E). The fact that compensatory effects of substitutions on RNA binding were more evident in plate-based assays with malM-3′ than in liquid β-gal assays or interactions with other RNA ligands is not surprising since they were identified in a plate-based screen with malM-3′ RNA and the conditions of the plate-based assay can be tuned to highlight small differences in β-gal levels (Stockert et al. 2022).

We next mapped the positions of the strongest confirmed compensatory substitutions to two available structures for the ProQ^NTD (Gonzalez et al. 2017; Jumper et al. 2021; Varadi et al. 2022). While the two models share an overall fold of the FinO domain, they differ in the region containing R80. V74 and L91, the sites of the strongest compensatory mutations, point away from the convex-facing R80 in the NMR structure (Fig. 2D; Gonzalez et al. 2017), but toward R80 and the concave face in the AlphaFold structural model (Fig. 2E; Jumper et al. 2021; Varadi et al. 2022); D92 and H95, which showed qualitative rescue of malM-3′ binding (Fig. 2C), further form a pocket with V74 and L91 around R80 in the AlphaFold model (Supplemental Fig. S5). Finally, when V74K and L91V substitutions are modeled into the respective structures as lowest-energy rotamers, the terminal atoms are 9–14Å away from that of a convex-facing R80K (Fig. 2D), but only 4–7Å away in the AlphaFold model (Fig. 2E). Indeed, the AlphaFold model rationalizes why the conservative R80K would be such a deleterious substitution: lysine’s shorter aliphatic side would pull the terminal polar/charged functional group into the hydrophobic core of the protein, disrupting the overall fold (Supplemental Fig. S5). The model also explains why a V74K substitution would rescue a lysine at position 80, as it provides a second amine in hydrogen-bonding distance of R80K in an otherwise hydrophobic environment (Fig. 2E).

Together, these results lend experimental validation to the AlphaFold model of E. coli ProQ’s FinO domain, in which all critical RNA-binding residues, including R80, are located on the concave face of the FinO domain (Fig. 2E; Supplemental Fig. S2). In this model, the concave face of the FinO domain looks quite like other structural homologs, with the highly conserved Y70 and R80 residues positioned near one another in a concave pocket.

Central concave-pocket residues are essential for RNA binding in vitro

With this refined structural model for E. coli ProQ^NTD in mind, we wished to further explore how specific FinO-domain residues contribute to RNA binding. In particular, we wanted to test whether conclusions from B3H studies with cspE-3′ and SibB RNAs about the importance of specific residues on RNA binding (Pandey et al. 2020) would also hold true with purified components and for other RNA ligands of ProQ. For this purpose, we compared RNA binding to ProQ^NTD mutants using electrophoretic mobility shift assays, as previously conducted (Stein et al. 2020). We selected seven RNAs for this study, each of which had been found to bind to ProQ in vivo (Supplemental Fig. S6; Holmqvist et al. 2018; Melamed et al. 2020). These RNAs included malM-3′, the top RNA ligand identified by RIL-seq of ProQ (Melamed et al. 2020), two versions of cspE 3′-UTR (a 52-nt fragment [cspE-3′] and an 81-nt fragment [cspE81-3′]), gapA-3′, SibA and SibB, all of which are specific ligands of ProQ in vivo, and RybB, which is bound by both ProQ and Hfq in vivo. With the exceptions of SibB and RybB, the binding of all these RNAs had already been studied with purified ProQ^NTD (Stein et al. 2020).

For binding studies, we used a 130-aa long version of ProQ’s NTD, the RNA-binding properties of which had been previously studied (Chaulk et al. 2011; Pandey et al. 2020; Stein et al. 2020). We first compared the binding affinities of the seven RNA ligands with WT ProQ^NTD. The observed K_d values for ProQ^NTD binding to cspE-3′, cspE81-3′, malM-3′, SibA, and SibB fell within in a similar range (4–10 nM); only gapA-3′ and RybB bound weaker with K_d values of 14 nM and 26 nM, respectively (Table 1; Figs. 3; Supplemental Figs. S7–S13). Overall, the seven RNAs bound WT ProQ^NTD with a relatively narrow range of affinities.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

Mutations in the central pocket have strong detrimental effects on the binding of seven RNAs. (A) The location of substitutions on the AlphaFold model of ProQ^NTD (Varadi et al. 2022). (B) The raw gel data for 1 nM ³²P-labeled malM-3′ binding to the eight proteins. (C) The plots of fraction bound versus protein concentration for the binding of cspE-3′, cspE81-3′, gapA-3′, malM-3′, RybB, SibA, and SibB RNAs to WT ProQ^NTD, and T65A, K54A, R62A, R58A, Y70F, R80A, and R80K mutants measured using the gel shift assay. The data sets in which the maximum fraction bound was above 40% were analyzed by fitting to the quadratic equation. Raw gel data for all RNAs are presented in Supplemental Figures S7–S13. The obtained K_d values are shown in Table 1.

Because the in vivo B3H data showed the importance of universally conserved residues tyrosine 70 and arginine 80 for the binding of cspE-3′, malM 3′, and SibB (Figs. 1, 2; Pandey et al. 2020), we next compared the effects of mutations at these positions on the binding of the seven RNAs. The Y70F mutant had a dramatic effect on RNA binding across all RNAs, with the fraction of RNA bound reaching no higher than 20% at the highest concentration of ProQ^NTD (1 µM) for cspE-3′, cspE81-3′, and malM-3′ RNAs, and no binding detected at all for gapA-3′, RybB, SibA, and SibB RNAs (Table 1; Fig. 3; Supplemental Figs. S7–S13). Because this and subsequent mutations affected the maximum fraction of RNA bound to ProQ, Table 1 reports the maximum fraction bound for each mutant as well as K_d values for experiments where the fraction bound reached at least 40%. Substitutions at R80 were even more deleterious than those at Y70: no binding was detected for any of the seven RNAs to either R80A or R80K mutants (Table 1; Fig. 3B,C; Supplemental Figs. S7–S13), consistent with the strong effect of both mutants in vivo (Fig. 1C). The fact that substitutions of Y70 and R80 were so deleterious for the binding of all seven RNAs supports the conclusion that these amino acid residues are universally important for the binding of natural RNA ligands of ProQ.

To test if the compensatory effects of double mutations observed in the B3H assay can also be observed with an in vitro binding assay, we constructed and purified the R80K-L91V double mutant, which showed restoration of binding to malM-3′ in both plate-based and liquid B3H assays (Fig. 2C; Supplemental Fig. S4B,C). However, we did not observe improvement in binding compared to the single R80K mutant (Supplemental Fig. S14). This suggests that despite an observable effect in the in vivo B3H assay, the compensatory role of the L91V substitution is too subtle to be detected using a gel shift assay. The lack of an effect could reflect that the complex is not sufficiently stable during the time required for electrophoretic separation. Alternatively, it is possible that the double mutant ProQ might be stabilized inside the cell by interactions with cellular RNAs or an additional factor that could support its function in the three-hybrid assay.

Next, we analyzed the role of threonine 65, which is close to Y70 on the concave face of the FinO domain (Fig. 3A). While T65 was not identified in our in vivo mutagenesis screens (Pandey et al. 2020), we hypothesized that it could be important for RNA binding based on its proximity to the side chain of Y70 as well as its conservation as a threonine or a structurally similar serine or cysteine (Supplemental Fig. S1). Despite these factors, the T65A mutation had rather small effects on RNA binding by ProQ^NTD, ranging from no effect on the binding of gapA-3′ and RybB, to twofold weaker binding than WT ProQ^NTD for cspE81-3′, malM-3′, and SibA, and fourfold weaker binding for cspE-3′ and SibB (Table 1; Fig. 3B,C; Supplemental Figs. S7–S13). In keeping with the modest effects on K_d, the T65A mutant did not alter the maximum fraction bound of any RNA compared to WT ProQ^NTD. These data suggest that T65 makes only small contributions to RNA binding despite its partial conservation and positioning adjacent to the essential Y70 residue.

Arginine 62 is located one helical turn away from T65 in helix H3 (Fig. 3A). We were intrigued that among the six FinO-domain proteins with solved structures and/or natural RNA ligands identified, Ec ProQ is the only one that contains arginine in this position in the concave-face pocket (Supplemental Fig. S1). While arginine is present in homologous positions of ProQs from related enterobacteria (Smirnov et al. 2016; Olejniczak and Storz 2017; Pandey et al. 2020), it is absent from the equivalent position of Lp RocC, which suggested that R62 could be a good candidate for an RNA contact that is different in Ec ProQ than Lp RocC. Indeed, gel shift data showed that the R62A mutation had a strongly detrimental effect on the binding of all seven RNAs, as the maximum fraction bound was below 40% for most RNAs (Table 1; Fig. 3; Supplemental Figs. S7–S13). The mutant protein’s interaction with cspE-3′ was the strongest of all the RNAs with a K_d value only fourfold weaker than WT ProQ^NTD, although with a fraction bound just above 40% at the highest concentration tested of R62A ProQ^NTD (Table 1). The maximum fractions bound observed for other RNAs were lower still—around 25% for malM-3′ and SibA, and below 20% for cspE81-3′ and gapA-3′—while no binding was detected for RybB and SibB. Although there seems to be a range of effects of R62A on different RNAs, these differential effects are impossible to precisely quantify because of the overall strongly weakened binding. Overall, the R62A mutation is strongly detrimental to the binding of all seven RNAs, although not to the same degree as mutations of Y70 or R80.

We next turned our attention to two basic residues that were found to be important for in vivo ProQ–RNA interactions and were hypothesized to interact with the double-helical portion of a terminator hairpin (Pandey et al. 2020). These residues—arginine 58 and lysine 54—are located one or two additional helical turn(s) further along H3 from R62 on the concave face of ProQ (Fig. 3A). We first analyzed the effects of an alanine substitution of the R58 residue, the closer of the two residues to Y70, and found that an R58A mutation was strongly deleterious for binding of all seven RNAs. As with R80A, no binding of any RNA was detected up to concentrations of 1 µM of the R58A mutant (Table 1; Fig. 3; Supplemental Figs. S7–S13), demonstrating an essential role of R58 in RNA binding by the ProQ FinO domain. On the other hand, the effects of a K54A mutation were more moderate: the maximum fraction bound only reached about 60% for cspE-3′, 30% for malM-3′, and cspE81-3′, and <20% for gapA-3′ and RybB, while no binding was detected for SibA and SibB (Table 1; Fig. 3; Supplemental Figs. S7–S13). These effects are comparable to those of the R62A mutation described above. Together, these data show that residues K54 and R58 on helix H3 of the FinO domain play important roles in RNA binding.

RNA ligands differ in their sensitivity to amino acid substitutions on the periphery

In addition to the central part of the concave-face pocket explored above, our previous B3H studies indicated that some residues on the periphery of the concave face, such as lysine 35 and aspartate 41, could also contribute to RNA binding (Pandey et al. 2020). Interestingly, B3H studies showed that the K35A mutation had more detrimental influence on the binding of SibB than cspE, which suggests that K35 could contribute differentially to the binding of distinct RNAs (Pandey et al. 2020). When the binding of our panel of seven RNAs was tested in vitro with the K35A ProQ^NTD mutant, there were indeed large differences in the maximum fraction bound reached by each RNA. While the maximum fraction of cspE-3′ and cspE81-3′ bound to the K35A mutant remained similar to that of WT ProQ^NTD, the K35A mutant only reached ∼70% RNA bound with malM-3′, and 35%–50% with gapA-3′, SibA and RybB. The effects on SibB were the strongest, with no binding detected up to 1 µM protein. For several RNAs that reached a much lower fraction bound with K35A than for WT ProQ^NTD, the K_d values calculated from the fits were still within three- to fourfold of the K_d for WT ProQ^NTD. That the differences in binding were more apparent from the maximum fractions bound rather than the K_d values could suggest that complexes formed by the K35A mutant with some RNAs were less stable during electrophoresis in native gels, leading to the underestimation of calculated K_d values. Alternatively, such a result could be explained if the tight binding of some RNAs required a conformational change that was only efficiently induced by ProQ^NTD in the presence of K35. Importantly, the relative effects of the K35A mutation on the binding of cspE-3′ and SibB are in agreement with the results obtained by the B3H assay in vivo, which also showed a stronger negative effect of this mutation on the binding of SibB (Pandey et al. 2020).

We next examined the effects of mutating aspartate 41, another residue located at the base of helix H2 (Fig. 4A) that was suggested by previous B3H results to be important for RNA binding (Pandey et al. 2020). In vitro binding data showed that the D41A substitution does not affect the K_d value of ProQ^NTD for gapA-3′, but does have a detrimental effect on the binding affinity for other RNAs; the size of its effect ranged from approximately twofold (cspE81-3′ and RybB) or threefold (cspE-3′ and malM-3′) to more than fivefold in the strongest cases (SibA and SibB) (Table 1; Fig. 4; Supplemental Figs. S7–S13). Interestingly, the binding of all seven RNAs to the D41A mutant reached the same maximum fraction bound as with WT ProQ^NTD (Table 1), in contrast to the effects of K35A above. In summary, both D41 and K35 contribute modestly to RNA binding by ProQ in vitro, with K35 having more varied effects across RNAs, perhaps mediated through kinetics or facilitating conformational changes in the RNA.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

Mutations on the periphery of the central pocket have differential effects on the binding of seven RNAs. (A) The location of substitutions on the AlphaFold model of ProQ^NTD (Varadi et al. 2022). (B) The raw gel data for 1 nM ³²P-labeled malM-3′ binding to the five proteins. (C) The plots of fraction bound versus protein concentration for the binding of cspE-3′, cspE81-3′, gapA-3′, malM-3′, RybB, SibA, and SibB RNAs to WT ProQ^NTD, and R32A, D41A, K35A, and R69A mutants measured using the gel shift assay. The data sets in which the maximum fraction bound was above 40% were analyzed by fitting to the quadratic equation. Raw gel data for all RNAs are presented in Supplemental Figures S7–S13. The obtained K_d values are shown in Table 1.

Next, we moved further away from the concave-face pocket of ProQ, and asked whether positively charged amino acids on the convex face of the FinO domain could also affect RNA binding. We first examined the effects of arginine 32, which is conserved either as arginine or lysine in other FinO-domain proteins (Supplemental Fig. S1). In vitro binding data showed that an R32A mutation does not affect the K_d value of ProQ binding to gapA-3′ and malM-3′ and has only modest two- to threefold detrimental effects on the binding of cspE-3′, cspE81-3′, RybB, and SibA (Table 1; Fig. 4; Supplemental Figs. S7–S13). The only RNA that was markedly affected by this mutation was SibB; while its K_d value was only weakened approximately threefold by the R32A mutation, its maximum fraction bound was reduced to just above 40% (Table 1). The fact that the R32A substitution has only small effects on the binding of most of the RNAs tested suggests that R32 does not make a universal key contact with RNA.

Another basic side chain exposed on the convex face of the FinO domain is arginine 69. Unlike R32, R69 is in an evolutionarily variable position, but its location next to Y70 in the primary amino acid sequence suggested it could affect RNA binding (Supplemental Fig. S1). In vitro binding data showed that an R69A substitution had a three- to fivefold detrimental effect on the binding of malM-3′ and cspE-3′, and an eight- to 10-fold effect on interactions with SibA and gapA-3′ (Table 1; Fig. 4; Supplemental Figs. S7–S13). The R69A mutation had an even stronger detrimental effect on the binding of other RNAs: the maximum fraction bound reached by RybB was below 40%, while the binding of SibB was not detected. The large range of effects of the R69A mutation was reminiscent of the effects of K35A, though perhaps qualitatively slightly stronger for each RNA. It is interesting to note that the rank-order of effects caused by K35A and R69A mutations on the maximum fraction bound reached by each RNA were similar, with cspE-3′ being the least affected by either mutation and SibB being the most affected (Table 1).

The terminator hairpin drives the differential dependence of cspE-3′ and SibB on R69

The results above indicated that several residues, especially K35 and R69, made differential contributions to the binding of the seven RNAs we tested. To explore the basis for these differences, we focused on cspE-3′ and SibB RNAs, because the differential effect of a K35A mutation on ProQ binding to these two RNAs was also observed in vivo (Pandey et al. 2020), and because they fell on the opposite ends of the spectrum of susceptibility to ProQ^NTD mutants, which increased the likelihood of identifying specific binding determinants. Because the FinO domain of ProQ is known to specifically recognize the intrinsic terminators of its RNA ligands (Holmqvist et al. 2018; Melamed et al. 2020; Stein et al. 2020), we focused on this region of cspE-3′ and SibB. The two most striking differences between the terminator regions in these two RNAs are that SibB possesses a shorter oligoU tail that ends with two cytidines and that the base of cspE’s terminator hairpin contains several A–U base pairs missing from SibB (Fig. 5). We introduced changes into SibB, the weaker binding RNA, to make its 3′ terminator more similar to cspE-3′ and compared the binding of these mutant RNAs to WT and R69A ProQ^NTD. The first SibB mutant (SibB-5U) replaced the two terminal cytidines with uridines (Fig. 5A,B), resulting in an RNA with a total of five U’s in the tail, four of which are predicted to be single-stranded. While the SibB-5U mutation had no effect on the binding of WT ProQ^NTD, it improved binding to the R69A mutant, with the mutant protein reaching a fraction bound of >30% when binding was undetectable for unmutated SibB RNA (Table 2; Fig. 5B; Supplemental Fig. S15). We next introduced additional mutations to make the double-stranded base of the SibB terminator hairpin more like that of cspE-3′. This SibB–cspE chimera now had four A–U base pairs at the base of the terminator and an extended 8-uridine 3′ tail to match that of cspE-3′ (Fig. 5C,D). While the mutations in the SibB–cspE chimera had only slight effects on the binding of WT ProQ^NTD, they were sufficient to substantially restore RNA binding to the R69A mutant, which reached a fraction bound of 60% with a K_d value of 31 nM (Table 2; Fig. 5C; Supplemental Fig. S15). These are considerable improvements from the original SibB RNA, which showed no detectable binding to the R69A mutant, though they do not represent as strong of a binding as seen with cspE-3′ RNA, which reached a maximum fraction bound of 88% to the R69A mutant with a K_d of 20 nM.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 5.

The sequence and structure of the lower part of the terminator hairpins explain differences in the binding of ProQ^NTD and its mutants by cspE-3′ and SibB RNAs. (A–D), The secondary structures of RNA molecules are shown above the corresponding plots of fraction-bound data for WT ProQ^NTD (NTD) or its R69A mutant. The RNA secondary structures were predicted using RNAStructure software (Reuter and Mathews 2010). The changes introduced into the sequence of SibB are marked in red font. The data sets in which the maximum fraction bound was above 40% were analyzed by fitting to the quadratic equation. Data for SibB and cspE-3′ are the same as in Figure 3. The obtained K_d values are shown in Table 2, and raw gel data are presented in Supplemental Figure S15.

K35 and R69 indirectly contribute to RNA binding at the central RNA-binding site

We hypothesized that the differential effects of R69A and K35A on the binding of ProQ^NTD to distinct RNAs could result either from disruptions of direct contacts formed by K35 and R69 with specific RNAs or from indirect effects on alterations of contacts with other amino acids. Because R69 and K35 are located outside of the central concave-face pocket that has been proposed to interact with dsRNA (Pandey et al. 2020; Kim et al. 2022), we wondered whether R69 and K35 could form contacts with single-stranded regions on the 5′ and 3′ side of the terminator hairpin, respectively. We reasoned that if K35 or R69 directly contacted a single-stranded region outside of the hairpin, a shortened RNA would show reduced affinity with WT ProQ, and the mutant protein (K35A or R69A) would show reduced affinity with full-length RNA because it would be sufficient to remove only one of the elements taking part in the interaction to disrupt the contact that they formed. However, no additional loss of affinity would be observed when the mutant protein bound the shortened RNA because this contact would have already been disrupted when either element was removed. Alternatively, if the interaction between the mutant protein and shortened RNA resulted in additional loss of binding energy, the additive effects of these mutations would suggest that they arose from distinct contacts. To conduct these double-mutant experiments, we designed model RNA constructs derived from cspE-3′ and compared the effects of their 5′ and 3′ truncations on binding to WT ProQ^NTD and its R69A and K35A mutants (Table 3; Fig. 6; Supplemental Fig. S16). First, we constructed a model RNA (cspE-mini) by removing the minimally structured 17-nt sequence from its 5′ end and introducing guanosines in three locations within this model construct to enable efficient in vitro transcription of further 5′ truncated constructs—at the 5′ end, the base of the terminator hairpin, and above the A–U base pairs at the base of the hairpin (Fig. 6A,B). The cspE-mini construct was found to bind WT ProQ^NTD with the same affinity as unmodified cspE-3′ (Table 3; Fig. 6B), suggesting that the 17-nt stretch of nucleotides on the 5′ end of cspE-3′ does not markedly contribute to RNA binding by WT ProQ^NTD. Similarly, the loss of these 5′-terminal nt did not decrease the affinity of the R69A variant. Interestingly, however, the cspE-mini RNA bound to the K35A mutant with an approximately twofold weakened binding affinity (Table 3; Fig. 6B; Supplemental Fig. S16), suggesting that loss of a longer 5′ single-stranded tail makes RNA binding more dependent on the K35 side chain. While there were only modest effects of removing the first 17 5′ nt, removal of the remaining single-stranded sequence on the 5′ side (Fig. 6C) had a dramatic effect on binding across the board: the resulting cspE-mini-5′-blunt construct only achieved about 20% fraction bound with WT ProQ^NTD and resulted in no detectable binding to either R69A or K35A mutants (Table 3; Fig. 6C; Supplemental Fig. S16). To explore if releasing the 3′-oligoU tail of cspE-3′ would compensate for the contacts to ProQ^NTD lost on the 5′ of the hairpin, we further removed four adenosines opposite the 3′-oligoU tail (cspE-mini-5′-truncated) (Fig. 6D; Table 3; Supplemental Fig. S16). Despite its longer single-stranded oligoU tail, no binding of this construct was detected even for WT ProQ^NTD. Together, these data show that single-stranded nucleotides 5′-adjacent to the terminator hairpin are essential for the RNA binding by ProQ^NTD, while nucleotides further upstream of the terminator hairpin do not markedly affect RNA binding to WT ProQ but may make additional contacts to the protein that help compensate for the loss of K35. Importantly, these data are not consistent with either R69 or K35 directly contacting the 5′ single-stranded RNA, since substitution of either residue led to additional loss of interaction of 5′-truncated RNA.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 6.

The substitutions of peripheral residues K35 and R69 have indirect effects on the RNA binding in the central pocket of the ProQ^NTD. The RNA secondary structures were predicted using RNAstructure software (Reuter and Mathews 2010). The data sets in which the maximum fraction bound was above 40% were analyzed by fitting to the quadratic equation. Data for cspE-3′ are the same as in Figures 3 and 4. The average equilibrium dissociation constant (K_d) values are shown in Table 3, and raw gel data are presented in Supplemental Figure S16.

To explore whether R69 or K35 could affect RNA binding via contacts with the 3′ oligoU tail, we designed another construct (cspE-mini-3′-6U), the 3′ tail of which was shortened from eight to six Us, only two of which are predicted to be single-stranded (Fig. 6E). This 3′ truncation weakened the binding of WT ProQ^NTD an additional 2.5-fold beyond cspE-mini (Table 3), confirming previous observations that the length of the 3′ U-tail is important for the RNA binding to the FinO domain (Stein et al. 2020). The 3′ truncation also further reduced the binding affinity of both R69A and K35A mutants (by ∼2.4- and ∼1.4-fold, respectively), suggesting that neither of these residues primarily functions by contacting the terminal portion of the 3′ U-tail.

We next examined the effects of a more extensive 3′ truncation, removing an additional two terminal uridines to create cspE-mini-3′-4U (Fig. 6F). Binding of all three ProQ^NTD variants (WT, R69A, and K35A) was further weakened to this RNA construct, with strong effects seen on the maximum fraction bound for R69A and K35A proteins (Table 3). The fact that substitution of either R69 or K35 had additional detrimental effects on the binding of cspE-mini-3′-4U beyond those of cspE-mini-3′-6U suggests that these two nucleotides do not directly contact K35 or R69. We note that the effect of the 4-uridine truncation in cspE-mini-3′-4U is smaller than previously observed for a different cspE-3′ construct with the same truncation (Stein et al. 2020). While the affinities of these two constructs indeed differ in side-by-side binding assays (Supplemental Fig. S17), the result that a 4-uridine truncation is detrimental for RNA binding is consistent across both constructs.

In summary, these mutational studies suggest that R69 and K35 both likely contribute to contacts to the core terminator hairpin of the RNA rather than the adjacent 5′ or 3′ sequences. This is consistent with the above observation that the heightened susceptibility of SibB to R69A and K35A mutations is explained by the properties of its terminator hairpin (Table 2; Fig. 5). In addition, K35 may contribute in part to contacts that become especially important when RNAs possess a shorter extension on the 5′ side of their terminator hairpin.