S1 and KH Domains of Polynucleotide Phosphorylase Determine the Efficiency of RNA Binding and Autoregulation

Abstract

To better understand the roles of the KH and S1 domains in RNA binding and polynucleotide phosphorylase (PNPase) autoregulation, we have identified and investigated key residues in these domains. A convenient pnp::lacZ fusion reporter strain was used to assess autoregulation by mutant PNPase proteins lacking the KH and/or S1 domains or containing point mutations in those domains. Mutant enzymes were purified and studied by using in vitro band shift and phosphorolysis assays to gauge binding and enzymatic activity. We show that reductions in substrate affinity accompany impairment of PNPase autoregulation. A remarkably strong correlation was observed between β-galactosidase levels reflecting autoregulation and apparent K_D values for the binding of a model RNA substrate. These data show that both the KH and S1 domains of PNPase play critical roles in substrate binding and autoregulation. The findings are discussed in the context of the structure, binding sites, and function of PNPase.

INTRODUCTION

Polynucleotide phosphorylase (PNPase) is a conserved, widely distributed phosphorolytic 3′-5′ exoribonuclease that may also function under some circumstances as a template-independent RNA polymerase (1; reviewed in reference 2). Although it is not essential, deletions of its gene (pnp) are synthetically lethal when either RNase II or RNase R, both of which are hydrolytic 3′-5′ exoribonucleases, is deficient (3–5). Strains deficient in PNPase are also sensitive to cold shock and other stresses (6–9). Thus, PNPase is believed to play significant roles in mRNA turnover and other aspects of RNA processing and metabolism (4, 10–12). Partial or full structures of PNPase from several microorganisms (13–16), as well as structures of the related archaeal exosome (17, 18), have shed considerable light on its mechanism of action. Bacterial PNPase is composed of three identical subunits. Each subunit consists of two tandem globular domains (residues 8 to 210 and 312 to 541; see Fig. 1a) derived from RNase PH that form a core whose central channel is accessible from both the upper and lower surfaces of the core (14, 19). Each subunit also contains a C-terminal extension that consists of two additional small domains, KH and S1 (residues 551 to 591 and 622 to 691, respectively; Fig. 1), which are positioned on the upper surface of the core. The KH and S1 domains have been implicated in autoregulation (20), in resistance to cold shock (7), in pathogenesis (21), and in substrate binding (22). Although a solution structure of the S1 domain from Escherichia coli PNPase has been available (13), the positions of the S1 and KH domains relative to the core have been elucidated only recently from the structure of PNPase from Caulobacter crescentus (16). The S1 and KH domains do not appear to be required for the enzymatic activity of PNPase (20, 22, 23). Nonetheless, deletion of the S1, the KH or both domains results in significant loss of RNA binding and inefficient enzymatic turnover (22).

Fig 1.

(a) Domain organization of the E. coli PNPase monomer. The numbers of amino acid residues defining domain boundaries are taken from reference 22. The diagram is not to scale. (b) Structure of the E. coli PNPase KH domain as determined by homology modeling of the Nova antigen-2 KH3 domain (38) (Protein Data Bank [PDB] ID 1EC6a) with SWISS-MODEL (46–48). The conserved GxxG loop containing residues G570 and K571 is highlighted in red. The blue dot indicates the position of residue I576. (c) Structure of the E. coli PNPase S1 domain as determined by Bycroft et al. (13) (PDB ID 1SRO). Colored dots denote the positions of F635 (green), F638 (red), and H650 (yellow). The image in panel c has been modified from Cell (13) with permission of the publisher.

PNPase from E. coli and other organisms exhibits strong autoregulation that also depends on RNase III and conceivably on RNase E (24–27). A compelling model to rationalize the underlying observations has been developed by Jarrige et al. (26). PNPase is expressed from a bicistronic mRNA that encodes ribosomal protein S15 (rpsO) and pnp. These two cistrons are separated by an intercistronic region containing an extended stem-loop structure that is a target for RNase III. Cleavage of the stem by RNase III exposes a new 3′ terminus on the 5′ side of the cleavage site. Exonucleolytic degradation from this terminus removes one strand of the stem-loop structure and exposes the monophosphorylated 5′ untranslated region (UTR) and coding sequence of the pnp mRNA. Presumably, this is an efficient substrate for RNase E (28) and is rapidly degraded, accounting for the downregulation of pnp mRNA. Thus, regulation of PNPase expression requires three RNases: RNase III, PNPase, and RNase E. In this model, the availability of PNPase and its ability to recognize the RNase III cleavage product determine the rate at which the initial product of RNase III digestion is converted to a vulnerable substrate for RNase E. Prior work has shown that the activity of PNPase is required for autoregulation (20, 23, 26); in addition, the pnp-71 mutation, mapping to the KH domain, impairs autoregulation (29, 30). We have undertaken a systematic investigation to uncover residues in the KH and/or S1 domains that contribute to RNA binding, as we hypothesized that RNA binding would be essential for autoregulation. We have also tested the predictions of Bycroft et al., who suggested that Phe 635, Phe 638, and His 650 in the S1 domain would contact RNAs (13). Our data show that there is a strong correlation between the affinity of PNPase for unstructured RNA as measured with purified enzymes and autoregulatory capacity measured in vivo. These data fully support the model of Jarrige et al. (26).

MATERIALS AND METHODS

Bacterial strains and plasmids.

Strain IBPC7322 (thi-1 argE3 ΔlacX74 mtl-1 tsx-29 rpsL pnp::Tn5) from the collection of Claude Portier was obtained from Matthias Springer (Institut de Biologie Physico-chimique, Paris, France). It was lysogenized at 30°C with λGF2 containing a pnp::lacZ fusion (also from C. Portier via M. Springer) to create the β-galactosidase reporter strain IBPC7322(λGF2). Strain ENS134-3 {F⁻ ompT gal [dcm] [lon] hsdS_B pnp::Tn5 (λDE3)} was provided by M. Dreyfus (Ecole Normale Supérieure, Paris, France) and served as the host strain for purification of PNPase or its derivatives.

Plasmid pAW101, which encodes wild-type (WT) PNPase under the control of the lac operator-promoter, was constructed in several steps (see Fig. S1 and Table S1 in the supplemental material). In effect, amino acid residues 547 to 711 encompassing the KH and S1 domains were modified by PCR so that they were flanked by engineered SalI and XbaI sites. Native or variant KH and S1 domains could be readily ligated to the catalytic core of PNPase (amino acid residues 1 to 546) (see Fig. S1 in the supplemental material). Plasmid pAW101 contains a reconstructed full-length pnp gene under the control of plac. This plasmid served as the prototype for all subsequent derivatives expressing pnp mutant forms to be assayed in IBPC7322(λGF2). Plasmid pGC400, renamed pAW001, which encodes WT PNPase in the pET-11 backbone (31), was the prototype for all constructions for overexpression in strain ENS134-3 and subsequent purification (see purification of PNPase and PNPase variants below). All subsequent derivatives of pAW001 (pAW013-pAW017, pAW019, pAW021, pAW022, and pAW024) encoded PNPase variants of interest. Site-directed mutant proteins (for the sequences of the primers used, see Table S2 in the supplemental material) were prepared by the quick-change method (Invitrogen, Inc.) and confirmed by sequencing analysis. Unexpectedly, we discovered a previously unidentified, benign point mutation in the pnp coding sequences of pGC400 and pAW001, resulting in a T261A substitution. Site-directed mutagenesis was used to reverse the mutation (PNPFIX primers; see Table S2 in the supplemental material). Experimentation did not reveal any differences between the T261A mutant and WT PNPases (data not shown).

Many transformations were performed with Library Efficiency DH5α Competent Cells [Invitrogen, Inc.; F⁻ ϕ80lacZ′M15 ′(lacZYA-argF)U169 recA1 endA1 hsdR17(r_K⁻ m_K⁺) phoA supE44 thi-1 gyrA96 relA1 λ⁻]. In addition, chemically competent KCM cells of the desired strains were made by using the Berglund KCM competent protocol (http://openwetware.org/wiki/Berglund:KCM_competent and http://openwetware.org/wiki/Transforming_chemically_competent_cells).

Construction of a mutagenic KH-S1 library.

DNA fragments containing the KH and S1 domains (amino acid residues 547 to 711) were amplified by the error-prone PCR method (32) (see Fig. S1 in the supplemental material). Plasmid p19khs1_nm (see Fig. S1) was used as the template along with primers KTfp1 and KTrp2 (see Table S1 in the supplemental material) to create a library of mutant KH-S1* fragments framed by SalI and XbaI sites. The concentration of MnCl₂ was optimized to increase the frequency of fragments containing an ideal number of point mutations (between one and three substitutions). This was done by titrating the amount of MnCl₂ added to the PCR mixture, cloning the resulting PCR fragments into pUC19, and sequencing a few purified clones to count the mutations within the KH and S1 domains manually. The optimized reaction mixture, totaling 50 μl, combined 5 U of Taq DNA polymerase, primers KTfp1 and KTrp2 at 0.3 μM each, deoxynucleoside triphosphates at 200 μM each, and approximately 90 ng of plasmid p19khs1_nm as the template in a mixture of regular PCR buffer (7.4 mM Tris-HCl [pH 8.3 at 25°C], 37 mM KCl, 1.1 mM MgCl₂, 0.074% [vol/vol] Triton X-100) and mutagenic PCR buffer (80 μM dTTP, 80 μM dCTP, 550 μM MgCl₂, 50 μM MnCl₂ [all reagents concentrations final]). The PCR products obtained were digested with SalI and XbaI and cloned into pUC19, creating a library of p19khs1* plasmids (see Fig. S1). These were transformed into competent DH5α, individual colonies were purified and grown to saturation, and plasmids were extracted. DNA sequencing was performed to identify induced mutations in the KH and/or S1 domains. Plasmids that encode interesting mutant proteins were cleaved with SalI and NdeI. The resulting ∼743-bp fragments containing mutated KH-S1 regions were subsequently cloned into p18phph to reconstruct full-length pnp containing the respective mutation(s) (see Fig. S1 in the supplemental material) to generate plasmids in the pAW101series. In all cases, following the transformation of the ENS134-3, IBPC7322(λGF2), or DH5α strain, individual colonies were purified by restreaking on selective medium.

β-Galactosidase assay.

Cultures of reporter strain IBPC7322(λGF2) (see bacterial strains and plasmids above) were transformed with pAW101 or its derivatives, and individual colonies were purified on selective medium. Cultures of purified transformants were grown at 30°C in Luria-Bertani medium supplemented with carbenicillin (100 μg/ml) and kanamycin (20 μg/ml). Samples were taken during mid-exponential growth (at an A₆₀₀ of 0.40 to 0.50) and assayed by Miller's method (33), with the following modifications. Samples of 0.3 ml were chilled on ice for 20 min and then added to 0.7 ml of Z buffer (60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄, 50 mM β-mercaptoethanol) with 2 drops of CHCl₃ and 1 drop of 0.1% (wt/vol) SDS solution and vortexed vigorously for 10 s. These mixtures were then incubated for 10 min at 30°C before the addition of 0.2 ml of 4 mg/ml o-nitrophenyl-β-d-galactopyranoside at time zero. The reaction mixtures were incubated for 15 min at 30°C before being quenched with 0.5 ml 1 M Na₂CO₃. Absorbances at 420 and 550 nm were subsequently recorded for the quenched reactions and used to determine β-galactosidase activity in Miller units (MU) (33). A minimum of three independent trials were performed for each strain assayed. The percent repression values were calculated with the following equation: % Repression = [(Mutant pAW1xx MU − Empty Vector pUC18 MU)/(WT pAW101 MU − Empty Vector pUC18 MU)] × 100.

Western blot analysis of PNPase levels in reporter strain IBPC7322(λGF2).

Cultures of IBPC7322(λGF2) transformants that were assayed for β-galactosidase were also harvested for analysis by Western blotting when they achieved an A₆₀₀ of 0.40 to 0.50. Volumes of 1 ml were collected by centrifugation, and pellets were resuspended in 100 μl sample buffer (100 mM Tris-HCl [pH 6.8], 4% [wt/vol] SDS, 0.005% [wt/vol] bromophenol blue, 0.2 M dithiothreitol [DTT], 20% [wt/vol] glycerol). Cells were lysed by boiling for 5 min, and portions of each extract (10 μl) were separated by electrophoresis, electroblotted onto a nitrocellulose membrane, and probed with polyclonal rabbit anti-PNPase antibodies (1:10,000) in a manner similar to that described previously (34). Modifications included the use of IRDye 800CW-conjugated goat polyclonal anti-rabbit IgG (diluted 1:10,000; LI-COR Biosciences) as the secondary antibody and the use of a LI-COR Odyssey for detection.

Purification of PNPase and PNPase variants.

All purifications were performed with untagged versions of PNPase overexpressed in host strain ENS134-3 as described previously (35), with the following changes. A French pressure cell was used for lysis; after dialysis and clarification, the S15 supernatant containing PNPase was applied to a column of Q Sepharose Fast Flow resin (GE Healthcare). Subsequent hydrophobic interaction chromatography was performed with Phenyl Sepharose 6 Fast Flow (high sub) matrix (GE Healthcare).

Enzyme assays and preparation of SL9A RNA substrate.

Synthetic radiolabeled SL9A RNA (36) (see Fig. 4a) was used for all activity and band shift assays. SL9A RNA was prepared by in vitro transcription of a derivative of pTZ18U linearized with XbaI (36). Transcription was performed with this template in the presence of [α-³²P]CTP.

Fig 4.

Assays of SL9A RNA processing by WT PNPase and representative PNP mutant proteins. (a) Schematic diagram of a synthetic SL9A RNA whose secondary structure has been described previously (36). The X at the 5′ end represents the RNA sequence (5′-ppp)-GGGAAUUCGAGCUCGGUAC. The arrow indicates the point at which processing by WT PNPase stalls. (b) Time courses of PNPase activity showing pmol of SL9A processed by 100 fmol of the following enzymes (added at time zero): c, WT; d, K571L mutant enzyme; e, I576T mutant enzyme; f, F635A F638A H650A mutant enzyme (data for I576T F638A mutant enzyme not shown). Activity assays were performed as described in Materials and Methods and were repeated at least once. nt, nucleotides. The asterisk in panel c represents a sample lacking enzyme.

Assays of phosphorolytic activity were performed as previously described (22), except that the assay buffer contained 20 mM Tris-HCl (pH 7.5), 1.5 mM dithiothreitol, 1 mM MgCl₂, 10 mM Na-phosphate (pH 7.5), 100 mM KCl, and 50 nM labeled SL9A RNA. Activity assays were initiated by the addition of the appropriate PNPase to give a final enzyme concentration of 2.5 nM. Aliquots (4 μl) from each incubation were quenched at various times with 3 volumes of 90% formamide, denatured, and separated on 8% polyacrylamide gels containing 8 M urea in Tris-borate-EDTA buffer. Gels were fixed and dried. Substrates and products were visualized and quantified by phosphor imaging.

RNA binding was assessed by electrophoretic mobility shift assays largely as described by Stickney et al. (22), except that increasing concentrations of WT PNPase or its derivatives (0.25 to 150 nM) were incubated in 20 μl of buffer containing 10 mM Tris-HCl (pH 8.2), 0.1 mM EDTA, 80 mM NaCl, 1% glycerol, and 0.01% dodecyl maltoside and the labeled SL9A RNA concentration was lowered to 1 nM. Dilutions of purified PNPase or PNPase mutant proteins were made in the same buffer (without SL9A) containing in addition 1 mM DTT and 25 μg/ml bovine serum albumin. Following incubation, the same portion of each sample was separated by electrophoresis in a nondenaturing (7%) polyacrylamide gel in Tris-borate-EDTA buffer. Gels were fixed and dried. Substrates and products were visualized and quantified by phosphorimaging.

RESULTS

Both the S1 and KH RNA-binding domains facilitate PNPase autoregulation.

PNPase autoregulation has been shown to require enzymatic activity and, in particular, RNA binding, as mutations that impair either of these functions result in increased cellular levels of the enzyme (20, 26, 29, 37). We postulated that PNPase autoregulation would be affected by mutations of RNA contact residues in the S1 and KH domains that also impact substrate binding. We therefore constructed the pnp::lacZ reporter strain IBPC7322(λGF2) to measure autoregulation by the WT and various mutant PNPase proteins in vivo (see Materials and Methods). This strain contains a pnp::Tn5 insertion to inactivate the chromosomal pnp gene. As a result, the expression of β-galactosidase from the chromosomal reporter is sensitive to regulation by the pnp variant expressed in trans (i.e., from the pAW1xx plasmids listed in Table 1).

Table 1.

Autoregulation of pnp::lacZ reporter by WT and mutant PNPases

pnp mutation(s)	E. coli reporter strain IBPC7322(λGF2)
	Plasmid	Enzyme activity (MU)a	% Repressionb
None	pUC18	584 ± 44	0
WT	pAW101	24 ± 4	100
ΔKH ΔS1	pAW102	535 ± 18	9
ΔKH	pAW103	479 ± 24	19
ΔS1	pAW104	427 ± 28	28
I555T	pAW110	44 ± 8	96
G570C V679A	pAW111	342 ± 28	43
I576T T585A	pAW112	121 ± 9	83
G570C	pAW113	467 ± 32	21
K571L	pAW114	186 ± 23	71
K571Q	pAW115	80 ± 03	90
I576A	pAW116	119 ± 20	83
I576T	pAW117	260 ± 25	58
F635A	pAW118	80 ± 7	90
F638A	pAW119	97 ± 11	87
H650A	pAW120	136 ± 10	80
F635A F638A H650A	pAW121	309 ± 24	49
F635R F638R H650R	pAW122	135 ± 10	80
I576A F638A	pAW123	235 ± 40	62
I576T F638A	pAW124	358 ± 18	40

^aβ-Galactosidase activity expressed from the chromosomal pnp::lacZ fusion in strain IBPC7322(λGF2)/pAW1xx was measured as described in Materials and Methods and is the average of at least three independent trials.

^bPercent repression was calculated as described in Materials and Methods.

A series of derivatives of IBPC7322(λGF2) was prepared by transformation with pUC18 or plasmids in the pAW1xx series (Table 1). The resulting bacteria were grown individually from stock cultures and assayed for β-galactosidase activity. The activities measured are reported in MU (33) in Fig. 2a and b and Table 1. These MU values inversely reflect the activity and strength of autoregulation by the plasmid-encoded PNPase variant. To facilitate the comparison of repression efficiencies, data are also reported as percent repression in Table 1 (see Materials and Methods for the calculation method used). The reporter strain transformed with the empty vector, pUC18 [IBPC7322(λGF2)/pUC18], generated the highest levels of β-galactosidase activity (584 ± 44 MU; Fig. 2a and Table 1). This is interpreted to reflect a lack of repression of PNPase (0% repression). In contrast, strain IBPC7322(λGF2)/pAW101, expressing full-length WT PNPase, produced consistently low levels of β-galactosidase activity (24 ± 4 MU). This activity defines 100% repression, as calculated by the equation described in Materials and Methods.

Fig 2.

Repression of pnp::lacZ in strain IBPC7322(λGF2) by KH and S1 mutants. (a) Activity of S1 domain mutant constructs shown by crosshatched bars was assessed by β-galactosidase assays as described in Materials and Methods. The names of the plasmids (pAW1xx) from which mutant PNPases were expressed are shown on the x axis. Data are the averages of at least three determinations. (b) Autoregulation by KH domain mutants. Details are the same as for panel a. (c, d) Western blot assays showing levels of S1 (c) and KH (d) domain mutant enzymes expressed in reporter strain IBPC7322(λGF2) were performed as described in Materials and Methods.

The S1 domain is required for efficient PNPase autoregulation (20, 26), and we previously proposed a role for the S1 domain in initial substrate recognition and subsequent product displacement (22). Thus, removal of the S1 domain from PNPase would likely impede autoregulation in current models (26, 27). Strain IBPC7322(λGF2)/pAW104, which encodes PNPase completely lacking the S1 domain (ΔS1; Δ605-711), generated 427 ± 28 MU (Table 1 and Fig. 2a) or only 28% repression. This impairment of autoregulation parallels previous in vitro findings of loss of substrate affinity by purified ΔS1 PNPase (22).

Although Bycroft et al. (13) predicted that F635 and F638 in the S1 domain were likely candidates for RNA contact, the results obtained by Jarrige et al. (20) were equivocal. These authors reported that an F638G substitution reduced catalytic activity but had little effect on the autoregulation of PNPase. Curiously, the F635G F638G double mutant protein exhibited almost normal activity but showed a partial loss of autoregulation (20). We endeavored to clarify these findings by assessing autoregulation by S1 domain mutant proteins with substitutions at these sites and by investigating the properties of purified PNPase variants containing single or multiple point mutations in the S1 domain. Single mutations to alanine were introduced by site-directed mutagenesis at positions 635, 638, and 650 in the S1 domain (Fig. 1c), creating pAW118, pAW119, and pAW120, respectively (Table 1). Derivatives of IBPC7322(λGF2) containing plasmids pAW118 to pAW120 were assayed for β-galactosidase activity. All of these strains exhibited similar levels of activity, ranging from 80 to 136 MU (Table 1 and Fig. 2a). These values represent small defects in autoregulation equivalent to 80 to 90% of the repression activity of WT PNPase.

Plasmid pAW121 expressing a mutant PNPase with all three of the putative aromatic contact residues simultaneously mutated to alanine (F635A F638A H650A; triple-A mutant protein) was transformed into reporter strain IBPC7322(λGF2), and purified clones were assayed for β-galactosidase activity. This strain produced 309 ± 24 MU of activity (Table 1 and Fig. 2a), over three times the activity of the F638A single mutant [IBPC7322(λGF2)/pAW119; 97 MU, 87% repression]. The activity of the triple-A mutant strain represents 49% repression. Interestingly, pAW122 expressing PNPase with all three aromatic residues simultaneously mutated to arginine (F635R F638R H650R; triple-R mutant) produced 135 ± 10 MU of activity in IBPC7322(λGF2) (Table 1 and Fig. 2a). This reflects relatively efficient autoregulation (80% of WT repression activity), in contrast to the triple-A mutant (49% repression). At 80% repression, the triple-R mutant (pAW122; Table 1 and Fig. 2a) repressed the reporter to a level comparable to those of the single S1 domain point mutants (pAW118 to pAW120, 80 to 90% repression), effectively reversing the impact of the triple-A mutations on autoregulation (see Discussion).

To assess the expression level of each mutant PNPase in reporter strain IBPC7322(λGF2), Western blotting of crude cell extracts was conducted (Fig. 2c). Almost identical levels of expressed WT or mutant PNPase were observed for each strain, confirming that relative PNPase abundance was not contributing to the differences in assayed β-galactosidase activity.

Random mutagenesis implicates the KH domain in efficient PNPase autoregulation.

In order to sample the impact of additional residues in both the KH and S1 domains of PNPase, we turned to random-mutagenesis PCR (see Materials and Methods for the strategy used). More than 60 independent clones from mutagenic PCR were examined by DNA sequencing. Three interesting KH domain mutant constructs (pAW110 to pAW112 in Fig. 2b and Table 1) were identified and examined further. The first mutant construct investigated contained an Ile-to-Thr substitution within the KH domain at residue I555. Plasmid pAW110, which encodes the I555T variant, was transformed into reporter strain IBPC7322(λGF2), which was then assayed for β-galactosidase activity. This strain generated 44 ± 8 MU, essentially autoregulating the pnp::lacZ message at WT levels (96% repression; Table 1). The second pnp mutant, encoded by plasmid pAW111, expressed full-length PNPase with substitutions at two residues: G570C, a Gly-to-Cys mutation in the first conserved G of the important GxxG loop (colored red in Fig. 1b), within the KH domain and V679A within the S1 domain. This mutant generated 342 ± 28 MU of β-galactosidase activity in strain IBPC7322(λGF2) (pAW111 in Table 1 and Fig. 2b), equivalent to 43% of WT repression activity. Interestingly, a similar impairment of autoregulation has been described for the pnp-71 mutant containing a G570D substitution (29). The third mutant protein identified in the mutagenic library, encoded by pAW112, contained substitutions at two residues within the KH domain (I576T and T585A; Table 1). The IBPC7322(λGF2)/pAW112 transformant produced 121 ± 9 MU of β-galactosidase activity (Table 1 and Fig. 2b), corresponding to 83% repression. This represents a small but obvious defect in autoregulation similar to what was found for the S1 domain point mutant constructs (pAW118 to pAW120; Table 1 and Fig. 2b). Coincidentally, in their mutational analysis of E. coli PNPase, Jarrige et al. (20) also found an I576N mutant protein to be impaired in autoregulation. This, in conjunction with the properties of pnp-71 (29) and modeling of the KH domain (Fig. 1b), prompted us to postulate that the G570C and I576T residues were responsible for the observed effects on autoregulation while V679A and T585A were likely benign. To test this idea, we effectively eliminated the V679A and T585A mutations by constructing single point mutations in full-length pnp containing G570C, I576A, or I576T (pAW113, pAW116, and pAW117, respectively; Table 1). Two of these KH domain substitutions, G570C and I576T (Fig. 2b), resulted in further reductions of repression (pAW113 and pAW117; Table 1) compared to the respective original double mutants identified in the mutagenic library (G570C V679A and I576T T585A; Fig. 2b and Table 1). Strain IBPC7322(λGF2)/pAW113, which encodes the G570C variant, produced 467 ± 32 MU, which is equivalent to only 21% of the WT repression activity, while IBPC7322(λGF2)/pAW117, which encodes I576T, was assayed at 260 ± 25 MU, which is equivalent to 58% of WT repression (Table 1). In contrast, I576A in IBPC7322(λGF2)/pAW116 had a smaller effect on autoregulation, producing 119 ± 20 MU—comparable to the 121 ± 9 MU generated by the original I576T T585A variant (pAW112; Table 1 and Fig. 2b).

An inherent feature of KH domains, the canonical GxxG sequence, has been shown to interact with RNA in C. crescentus PNPase (16) and to affect RNA affinity in E. coli PNPase (29, 30). These observations, together with the positioning of the loop in the predicted structure of the KH domain (Fig. 1b), suggested that the lysine in the second position of GxxG could also be important for autoregulation by E. coli PNPase. To investigate this possibility, we constructed pAW114 and pAW115, which respectively express PNPase K571L and K571Q variants (Table 1). Both plasmids were transformed into reporter strain IBPC7322(λGF2), which was then assayed for β-galactosidase activity. The strain expressing the K571L mutation [IBPC7322(λGF2)/pAW114] produced 186 ± 23 MU of activity, equivalent to 71% of WT repression activity (Fig. 2b and Table 1). The strain expressing the K571Q mutation [IBPC7322(λGF2)/pAW115] was less affected, generating an average of 80 ± 3 MU; equivalent to 90% of WT repression. This small deficiency in autoregulation is also similar to those seen for the three aromatic S1 point mutation and I576A variants (ranging from 80 to 136 MU of activity; pAW118 to pAW120 and pAW116; Table 1 and Fig. 2a and b).

To investigate the idea that the tandem KH and S1 domains could form a single, continuous RNA-binding domain (22), we asked if separate mutations in each domain interact. We used site-directed mutagenesis to construct pAW123 and pAW124, each of which encodes simultaneous mutations in the KH and S1 domains (Table 1 and Fig. 2b). The I576T F638A double mutant (IBPC7322(λGF2)/pAW124) exhibited 358 ± 18 MU of β-galactosidase activity, nearly 1.5 times the activity of the singly mutated I576T variant (260 ± 25 MU; pAW117 in Table 1 and Fig. 2b). This was equivalent to 40% of WT repression activity in I576T F638A, compared to 87% and 58% for the single point mutant proteins (F638A and I576T, respectively; Table 1, fourth column). Likewise, IBPC7322(λGF2)/pAW123, which contains I576A F638A, was also impaired in autoregulation, albeit less severely, repressing at 62% of WT activity. This suggests that the two mutations (one in the KH domain and one in the S1 domain) act synergistically to impair autoregulation rather than acting independently and redundantly.

As was done for the S1 domain variants, the expression of each KH variant in reporter strain IBPC7322(λGF2) was assessed by Western blotting of crude cell extracts (Fig. 2d). Similarly, almost identical levels of expressed WT or mutant PNPase were observed for all of the strains, confirming that relative PNPase abundance was not contributing to the differences in assayed β-galactosidase activity.

RNA binding by S1 and KH domain mutants.

Mutant PNPases warranting further investigation were purified to near homogeneity (see Materials and Methods). We used electrophoretic mobility shift assays with a radioactive model RNA, SL9A (see Materials and Methods and Fig. 4a), to assess RNA binding by purified, full-length PNPases containing the various KH and S1 domain mutations. K_D values were estimated by quantifying the fraction of shifted SL9A (see Materials and Methods) and are listed in Table 2 (third column). Images from typical mobility shift experiments are shown in Fig. 3. WT PNPase complexed with SL9A at relatively low concentrations with an apparent K_D of 0.6 ± 0.07 nM (n = 4) (Table 2 and Fig. 3a and b). In contrast, the F638A singly mutated PNPase was roughly 4.5-fold less efficient in RNA binding than the WT (Table 2), with an apparent K_D of 2.7 ± 0.05 nM (n = 3). Consistent with the apparent defect in autoregulation in vivo (pAW121; Table 1), higher concentrations of the F635A F638A H650A triple-A mutant PNPase were required to complex with SL9A (Fig. 3a and d). With an apparent K_D of 7.7 ± 1.3 nM (n = 3), the triple-A mutant was almost three times less efficient in RNA binding than the singly mutated F638A PNPase (2.7 ± 0.05 nM) and more than 12-fold less efficient than the WT PNPase (0.6 ± 0.07 nM; Table 2). Also mirroring findings from in vivo assays (pAW122; Table 1), the F635R F638R H560R triple-R mutant PNPase appeared to partially reverse the RNA binding defect observed in the triple-A mutant protein, with an apparent K_D of 2.7 ± 0.6 nM (n = 3) (Table 2). This represents an affinity for SL9A nearly identical to that of the F638A mutant (pAW019) PNPase.

Table 2.

RNA binding by purified mutant PNPases

pnp mutation(s)	E. coli expression strain ENS134-3
	Plasmida	Apparent KDb, nM (fold WT KD)
WT	pAW001	0.6 ± 0.07 (1)
G570C	pAW013	6.9 ± 2.6 (11.5)
K571L	pAW014	2.4 ± 0.3 (4.0)
K571Q	pAW015	1.6 ± 0.2 (2.7)
I576A	pAW016	2.0 ± 0.2 (3.3)
I576T	pAW017	7.0 ± 1.0 (11.7)
F638A	pAW019	2.7 ± 0.05 (4.5)
F635A F638A H650A	pAW021	7.7 ± 1.3 (12.8)
F635R F638R H650R	pAW022	2.7 ± 0.6 (4.5)
I576T F638A	pAW024	6.2 ± 1.3 (10.3)

^aEach mutant PNPase was purified to near homogeneity after overexpression from the respective plasmid (pAW0xx) as described in Materials and Methods.

^bApparent dissociation constants (K_Ds) were calculated by quantifying the fraction of shifted SL9A as a function of the enzyme concentration and determining the enzyme concentration required to shift 50% of the input RNA. Electrophoretic mobility shift assays were performed as described in Materials and Methods.

Fig 3.

In vitro RNA-binding activities of WT PNPase and representative PNP mutant proteins. Electrophoretic mobility shift assays were performed with 1 nM SL9A substrate as described in Materials and Methods. (a) Representative binding curves of the WT PNPase (solid black line), the I576T mutant PNPase (dashed line), and the F635A F638A H650A mutant PNPase (dotted line). Respective apparent dissociation constants are listed in Table 2 and were calculated from plots of the fraction of SL9A shifted versus increasing concentrations of WT PNPase or PNP mutant protein in three independent trials. (b) Representative mobility shift of WT PNPase. (c) Representative mobility shift of I576T mutant PNPase. (d) Representative mobility shift of F635A F638A H650A mutant PNPase. The concentrations of the respective PNPases assayed are shown above the lanes. The positions of unbound substrate (S) and retarded complexes (C) are indicated in the center margin. The asterisk shows the position of a putative aggregate in panels b and c.

At concentrations of 7.5 nM or higher, PNPase was observed to form a second complex in some cases (asterisks in Fig. 3b and c) that has been reported for band shifts in other work, also at higher concentrations of PNPase (complex II-R in reference 30 and * intermediate in reference 22). This may represent two PNPase trimers complexed to a single RNA but has not been characterized.

A similar pattern demonstrating loss of affinity for SL9A was observed for the purified KH domain mutants. Mutations at two KH residues, K571 and I576, were introduced by site-directed mutagenesis into full-length PNPase, and the purified enzymes were assayed by band shift (see Materials and Methods). At comparable enzyme concentrations, smaller proportions of SL9A were shifted by I576T mutant PNPase than by WT PNPase (Fig. 3c). Interestingly, different alleles of I576 resulted in significantly different apparent K_D values (Table 2), consistent with in vivo assays (pAW116 and pAW117; Fig. 2b and Table 1). I576T mutant PNPase yielded an apparent K_D of 7.0 ± 1.0 nM (n = 3) (Table 2), compared to an apparent K_D of 2.0 ± 0.2 nM (n = 3) for I576A mutant PNPase, a >3-fold difference. Other KH domain mutations were also assessed similarly. G570C mutant PNPase exhibited an apparent K_D of 6.9 ± 2.6 nM (n = 3), nearly 12-fold that of WT PNPase (pAW013 versus pAW001 in Table 2). At the second position of the GxxG sequence (red loop in Fig. 1b), the introduction of a leucine (K571L mutant PNPase) resulted in an apparent K_D of 2.4 ± 0.3 nM (n = 3), 4-fold higher than that of WT PNPase (Table 2). A glutamine substitution at the same site (K571Q mutant PNPase) was milder, resulting in an apparent K_D of 1.6 ± 0.2 nM (n = 3).

To investigate the relative contributions of the KH and S1 domains, full-length PNPase simultaneously containing I576T in the KH domain and F638A in the S1 domain (pAW024) was purified. The I576T F638A double mutant PNPase exhibited an apparent K_D of 6.2 ± 1.3 nM (n = 3), >10-fold that of WT PNPase (Table 2). The reductions in affinity for SL9A in the double mutant and the other various KH mutants compared to that of WT PNPase (Table 2) correlate well with the loss of autoregulation in vivo (see Discussion).

Enzymatic activities of KH and S1 domain mutants.

We assayed purified WT and mutant PNPases for phosphorolytic activity (see Materials and Methods) against SL9A, an RNA containing a 3′ unstructured tail (Fig. 4a) (36). Shortening of the poly(A) tail of 30 adenosine residues to the limit product results in a product of ∼55 nucleotides (Fig. 4a and center margin of panels c to f). The rate of conversion of SL9A to this limit product was taken as a measure of PNPase enzymatic activity (22, 36). Typical time courses of digestion of SL9A are shown in Fig. 4c to f. The reaction time in all cases was 30 min, as further incubation resulted in negligible additional product accumulation (data not shown). WT PNPase (100 fmol) shortened SL9A RNA at an average initial rate of 0.45 pmol/min. This value was taken as 100%; the rates measured for mutant enzymes are expressed relative to the WT rate.

For S1 domain mutants, the effects on phosphorolytic activity were either moderate (greater than 50% loss of the initial rates compared to the WT) or severe (greater than 75% loss). The two triply mutated (F635A F638A H650A and F635R F638R H650R) S1 domain mutant enzymes were only moderately impaired in phosphorolysis compared to the WT, with both enzymes exhibiting 29% of the WT rate (compare the WT in Fig. 4c with the triple-A mutant in Fig. 4f). Interestingly, the only S1 domain mutant that was severely impaired in enzymatic activity was the F638A (pAW019) mutant PNPase. This mutant PNPase, containing a single point substitution at one of the Phe residues proposed to contact RNA (Fig. 1c) (15), exhibited only 15% of the WT rate. A similar result was reported for a F638G mutant PNPase, which presented only 20 to 30% of the WT activity level in crude extracts (20).

In contrast, mutations of residues in the KH domain were found to affect phosphorolytic activity either mildly (up to 50% loss compared to the initial WT rate) or moderately (50 to 75% loss) and in one case resulted in WT activity (K571L mutant PNPase; Fig. 4). The K571L mutant PNPase exhibited essentially WT activity (compare WT in Fig. 4c with panel d). Both the I576A and I576T mutant PNPases exhibited initial rates and time courses of digestion that were mildly affected compared to those of the WT (compare the WT in Fig. 4c with the I576T mutant in panel e). The G570C mutant PNPase was moderately impaired compared to WT activity (42% ± 11% of the initial WT rate; time course not shown). The I576T F638A doubly mutated PNPase also exhibited a moderately impaired initial rate of phosphorolysis (31% ± 4% of the WT rate) in addition to generally lowered rates of activity over a 30-min time course (Fig. 4b, blue curve).

DISCUSSION

PNPase autoregulation activity strongly correlates with in vitro RNA-binding activity.

Past investigations have linked PNPase autoregulation to RNA binding (20). In this study, we found a remarkably strong correlation between the in vitro affinity of mutant PNPases for SL9A and autoregulation (Fig. 5), with a Pearson r value of 0.892 for 10 sets of data. Because RNA binding correlates so well with autoregulatory efficiency, binding is likely the limiting step. In the current model of autoregulation of PNPase expression, RNase III cleavage generates a new 3′ end in the pnp mRNA 5′ leader (25). PNPase then binds to the 3′ termini of the cleavage products in the initial step of autoregulation. Following binding, PNPase degrades RNA 5′ to the RNase III cleavage site, eliminating a duplex structure and exposing a 5′-monophosphorylated terminus on the 3′ side of the RNase III cleavage (26, 27). Our data fully support the importance of RNA binding by PNPase in this model and show that the efficiency of RNA binding by PNPase determines its autoregulatory capacity. Moreover, the ability of the KH and S1 domains to engage substrates is likely rate determining for the other roles played by PNPase in cellular RNA metabolism. Finally, the strong correlation shown in Fig. 5 demonstrates that the pnp-lacZ reporter coupled to PNPase expression in trans could potentially be used to investigate the properties of any single-stranded RNA-binding domain. This would entail the expression of a chimeric PNPase consisting of its catalytic core (residues 1 to 541; Fig. 1a) appended to a C-terminal RNA-binding domain replacing the native KH and S1 domains.

Fig 5.

(a) Correlation between RNA binding and in vivo PNPase autoregulation. Repression by PNPase and PNP mutant proteins, reflected by β-galactosidase activities measured as described in Materials and Methods, is plotted against their apparent in vitro affinities for SL9A RNA. (b) Statistics of measure of correlation.

The dynamic range of activity observed between full repression (WT PNPase, 100%) and near lack of autoregulation (ΔKH ΔS1, 9% repression) allowed us to categorize the impairment of RNA binding conferred by different mutations. This is best exemplified by the differences in repression observed for substitutions at the same site (i.e., substitutions I576T and I576A and substitutions K571L and K571Q; Fig. 2b). These apparently subtle changes may alter the local environment affecting the electrostatic interaction between a base or ribonucleotide and an amino acid. Along the same lines, the reporter could also distinguish differences in folding at the domain or even quaternary level. For example, deletion of both domains (ΔKH ΔS1) resulted in a ≥90% decrease in repression activity (pAW102, Table 1), whereas a single mutation at a putative RNA contact residue (F635A) resulted in a 10% decrease in activity (pAW118, Table 1).

Plasticity of RNA contacts in the S1 domain.

Our data show that F635, F638, and H650 (Fig. 1c) individually play modest roles in contacting RNA. However, when combined (triple-A mutation), these mutations produce a strong phenotype whether measured by loss of autoregulation or by loss of affinity for RNA. This finding confirms their collective importance in RNA binding (22). The replacement of these exposed residues in the S1 domain (13) with three Arg residues (F635R F638R H650R) resulted in partial reversal of the triple-A mutation (F635A F638A H650A). It is surprising that the native interaction, apparently including stacking of two Phe residues on single-stranded A residues, can be replaced by electrostatic interactions involving Arg residues and the RNA backbone. This degree of plasticity was unexpected. We also believe that F635, F638, and H650 do not significantly contribute to structural integrity. During purification, the triple-R and triple-A mutant enzymes eluted in the expected fractions at high yields and were of the expected size, suggesting proper folding of these mutant proteins and the absence of aggregation (data not shown). In corroboration, a trypsin sensitivity assay of these mutant enzymes and WT PNPase resulted in similar cleavage patterns, one band corresponding to the N-terminal core PNPase fragment with the KH and S1 domains removed. At lower ratios of trypsin to mutant PNPase, we obtained truncated PNPase with just the S1 domain removed. These data suggest that, at the very least, the triple mutations do not affect the overall folded structures of the KH and core PH′-PH domains (data not shown).

RNA-binding regions within the KH domain and insights into mechanisms of RNA binding impairment.

Screening of mutant enzymes from random mutagenesis identified two regions within the KH domain that are involved in autoregulation and RNA binding (red loop and blue dot in Fig. 1b). Although we obtained G570C and I576T without selection, it is remarkable that the same residues have independently been identified as important for autoregulation (20, 29). Substantial impairment of autoregulation in the G570C pnp mutant (pAW113) is consistent with the properties of the autoregulation-deficient pnp-71 allele (G570D) (29, 30). As well, our data for the I576T and I576A mutant constructs (pAW116 and pAW117) agree with the observations for the previously described I576N mutant construct (20). We believe, therefore, that G570 and I576 are key residues in the RNA-binding surface of the KH domain; other residues could also contribute.

In RNA-binding proteins of other species, KH domain residues equivalent to E. coli PNPase I576 have also been identified as contributors to RNA binding. Leu-28, the residue equivalent to I576 in the KH3 domain of Nova-2, provides an aliphatic stacking interaction with a base of the bound RNA. It has been designated a contact residue in part of an “aliphatic” binding platform (38). An I304N substitution (equivalent to I576) in the second KH domain of fragile X mental retardation protein (FMRP) perturbs KH domain function, leading to a rather malevolent form of fragile X mental retardation syndrome (reviewed in reference 39). These observations suggest that PNPase I576 and equivalent positions in the KH domains of other proteins play important roles in either direct substrate contact or stabilization of the KH fold.

Our identification and analysis of the G570C mutant are consistent with the key role played by the conserved GxxG loop in RNA binding (Fig. 1b; reviewed in reference 39). The crystal structure of the Nova antigen-2 KH3 domain shows that four bases of an RNA tetraloop are accommodated in a binding cleft formed by the GxxG motif in conjunction with a platform of hydrophobic residues and the variable loop region (observed in reference 38 and reviewed in reference 39). Mutating the first glycine of the conserved motif may interfere with the binding interface between the KH domain and the RNA substrate. We found the G570C mutation to have a pronounced effect on autoregulation and in vitro RNA binding (pAW113 in Table 1 and Fig. 2b and pAW013 in Table 2). Although the effects were less pronounced, both K571L and K571Q also resulted in losses of repression in vivo and weakened K_D values in vitro. Hollingworth et al. reported that a GDDG double mutation in KH-type splicing regulatory protein (KSRP) impairs nucleic acid binding without compromising KH domain stability (40). The absence or augmentation of the canonical GxxG sequence appears to be a hallmark of impairment of RNA binding (41, 42). Our data provide further support for this hypothesis.

It remains unclear whether the observed deficiencies in RNA binding brought about by these substitutions are the result of localized structural effects or loss of RNA contact residues. While mutations equivalent to I576T in other proteins (i.e., FMRP) could result in destabilization of the KH domain, in other contexts, it is just as likely that the particular residue directly contacts RNA (39). Indeed, Lewis et al. maintain that the I576 equivalent in Nova-2 KH3 (Leu-28) contacts RNA via an aliphatic stacking interaction (38). In our study, the impact of replacing I576 with two different residues on RNA binding in vitro supports their interpretation that I576 contacts RNA. Moreover, mutation of Ile is position specific, as a protein with I555T in the KH domain behaves like WT PNPase (pAW110 in Fig. 2b and Table 1). In addition, the behavior of the I576T mutant protein during purification implies that it folds natively, even though it is more impaired than the I576A mutant protein (data not shown). Nonetheless, to settle this issue, it would be desirable to solve the structure of appropriate KH domain mutant enzymes and to investigate their behavior by biophysical methods.

Model of substrate binding by the KH and S1 domains of PNPase.

An initial model of how the S1 and KH domains would facilitate RNA binding by PNPase was based on the structure of the archaeal exosome (18), as neither the KH nor the S1 domain was resolved by X-ray diffraction of E. coli or Streptomyces antibioticus PNPase (14, 15). Thus, Stickney et al. visualized the KH and S1 domains as promoting the initial, reversible interaction between PNPase and single-stranded RNA (22). Transient RNA-PNPase complexes would be stabilized when the 3′ end of the RNA passed into the central pore of PNPase, where it would make tight contacts near the catalytic site (15, 22, 43, 44). On the basis of the architecture of the exosome, Stickney et al. placed the S1 domain adjacent to the central pore and the KH domain more peripherally. The complete structure of C. crescentus PNPase with bound RNA has recently been solved under conditions where both the KH and S1 domains can be resolved (16). This structure strongly supports the concept that the S1 and KH domains function in the initial steps of recruitment of RNA to PNPase. However, the C. crescentus structure shows that the S1 domains are located peripherally while the three KH domains flank the opening to the central pore of PNPase. Moreover, the structure suggests that all three KH domains simultaneously contact an RNA substrate, although only a single S1 domain is presumed to be needed for the initial PNPase-RNA interaction.

We hypothesize that the combination of KH and S1 domains in PNPase to form an arrayed or continuous binding platform provides high affinity without significant sequence specificity (Fig. 6). These domains enable PNPase to promote the initial reversible, capture of single-stranded RNA spanning the full range of natural substrates. A low-affinity, reversible binding mechanism is also consistent with the “hands gripping a rope” model proposed by Hardwick et al. to describe ratcheting by the KH domains as they form nonequivalent interactions with the RNA substrate to propagate it toward the core (16). The requirement for a sufficiently long 3′ overhang to thread an RNA into the cleavage channel of human PNPase is also consistent with the simultaneous engagement of RNA 3′ termini by both S1 and KH domains and the “hands on a rope” model (45). We predict that KH domains with enhanced affinity for RNA may prove inhibitory to PNPase by trapping RNA on the surface and inhibiting ratcheting.

Fig 6.

Model of the roles of the KH and S1 domains in PNPase-RNA interactions. This is based on an earlier model by Stickney et al. (22) that takes into account structures determined in references 13, 16, 19, and 45). (I) An RNA substrate with a 3′ single-stranded extension is recognized by the peripheral KH and S1 domains (striped and white circles, respectively). This initial, reversible binding requires both domains to act synergistically (i.e., KH and S1 do not act redundantly in RNA binding). Contact points are minimally defined by F635, F638, and H650 in the S1 domain, the GxxG loop in the KH domain, and I576 also in the KH domain. (II) Upon binding, the KH and/or S1 domains modulate aperture size at the entrance to the central pore for substrate accommodation (bidirectional arrows) (15). Ratcheting (double-bent lines) by the KH domains propagates the RNA strand toward the active site (16) (red asterisks denote putative RNA contact points). (III) En route, the RNA strand makes nonequivalent contacts with the KH domains (16), in addition to making contact within a putative core binding site (red asterisks).

From recent structural determinations of PNPase and its homologues, it appears that the KH and S1 domains communicate with the conformationally dynamic catalytic core (15, 19, 45). Accordingly, an additional role for the KH domain may be the modulation of the size of the narrow “gateway” aperture to regulate substrate accommodation within the channel upon RNA binding.