Structure of Escherichia coli Hfq bound to polyriboadenylate RNA

Todd M. Linka, Poul Valentin-Hansenb, and Richard G. Brennana,1

aDepartment of Biochemistry and , University of Texas, M.D. Anderson Cancer Center, Houston, TX 77030-4009; and bDepartment of Biochemistry and Molecular Biology, University of Southern Denmark, DK-5230 Odense M, Denmark

Edited by Carol A. Gross, University of California, San Francisco, CA, and approved September 24, 2009 (received for review August 3, 2009) Hfq is a small, highly abundant hexameric protein that is found in proteins that serve as iron reservoirs, including sodB (superoxide many and plays a critical role in mRNA expression and RNA dismutase), ftn (ferritin), and bfn (bacterioferritin). Hfq also stability. As an ‘‘RNA chaperone,’’ Hfq binds AU-rich sequences and functions in cell envelope integrity by binding the sRNAs MicA facilitates the trans annealing of small (sRNAs) to their target and RybB to down-regulate the expression of the outer mem- mRNAs, typically resulting in the down-regulation of expres- brane proteins (ompA, ompC, and ompW, respectively) during sion. Hfq also plays a key role in bacterial RNA decay by binding growth and environmental stress condition (23–26). tightly to polyadenylate [poly(A)] tracts. The structural mechanism Beyond its role as an RNA chaperone, Escherichia coli (Ec) by which Hfq recognizes and binds poly(A) is unknown. Here, we Hfq also functions in RNA stability by binding with nanomolar report the crystal structure of Escherichia coli Hfq bound to the to subnanomolar affinities to polyadenylate [poly(A)] tails that poly(A) RNA, A15. The structure reveals a unique RNA binding have been added to the 3Ј ends of RNAs by poly(A) polymerase mechanism. Unlike uridine-containing sequences, which bind to (PAP I) (27–29). Unlike cytosolic eukaryotic mRNA, the addi- the ‘‘proximal’’ face, the poly(A) tract binds to the ‘‘distal’’ face of tion of adenosine nucleotides to bacterial mRNAs enhances their Hfq using 6 tripartite binding motifs. Each motif consists of an degradation (30). The extent of poly(A) tailing of Ec mRNAs by adenosine specificity site (A site), which is effected by peptide PAP I has been estimated recently to be greater than 90% of all backbone hydrogen bonds, a purine nucleotide selectivity site (R ORFs of bacteria that are growing exponentially (31). Intrigu- site), and a sequence-nondiscriminating RNA entrance/exit site (E ingly, Hfq switches PAP I from a distributive to a processive site). The resulting implication that Hfq can bind poly(A-R-N) catalytic mode, allowing for the more extensive adenylation of BIOCHEMISTRY triplets, where R is a purine nucleotide and N is any nucleotide, was mRNAs (27). How Hfq causes this switch is unknown. Moreover, confirmed by binding studies. Indeed, Hfq bound to the oligori- Hfq protects poly(A) tails from degradation by the exoribo- bonucleotides (AGG)8, (AGC)8, and the shorter (A-R-N)4 sequence, nucleases RNase II and polynucleotide phosphorylase (PN- AACAACAAGAAG, with nanomolar affinities. The abundance of Pase), and the endoribonuclease RNase E (28). The latter two (A-R-N)4 and (A-R-N)5 triplet repeats in the E. coli genome suggests enzymes are members of the degradosome, a multicomponent additional RNA targets for Hfq. Further, the structure provides assembly that is responsible for RNA decay in bacteria (32). Hfq insight into Hfq-mediated sRNA-mRNA annealing and the role of has been shown to interact with polynucleotide phosphorylase Hfq in RNA decay. (PNPase), which adds AG-rich polynucleotide tails to the 3Ј ends of mRNAs that are terminated in a Rho-independent fashion, gene regulation ͉ poly(A) ͉ posttranscription regulation ͉ RNA binding ͉ and RNase E, the key endoribonuclease involved in RNA Sm/Lsm family destruction (27–29)—both presumably via protein-RNA-protein interactions (33). fq is a pleiotropic posttranscriptional regulator that is a A full understanding of the RNA binding mechanisms of Hfq Hmember of the Sm/Lsm superfamily and found in many and its role in the protection of poly(A) tails from degradation Gram-negative and Gram-positive bacteria (1–3). Hfq binds to from endo- and exoribonucleases has been hindered by the lack AU-rich stretches near stem loop structures of small noncoding of relevant high-resolution structures. Indeed, the structure of RNAs (sRNA) and their target mRNAs to facilitate the annealing Staphylococcus aureus (Sa) Hfq bound to the uridine-rich oligo- of their trans-encoded complementary sequences and hence to ribonucleotide (AU5G) is the only example of an Hfq-RNA effect message-specific translational regulation (4–8). Thus, Hfq complex (34). The uridines and adenosine of this stretch of has been called an ‘‘RNA chaperone.’’ In conjunction with these sRNA bind to the ‘‘proximal’’ face of Hfq circling the central sRNAs, Hfq appears to down-regulate the expression of the ma- pore and making base-specific contacts with multiple side chains. jority of targeted messages; however, some mRNAs, including rpoS, The uridine and adenosine binding sites are essentially identical. which encodes the stress response/stationary phase sigma factor ␴S, By contrast, binding and mutagenesis studies revealed that require Hfq for their efficient translation (4, 9, 10). poly(A) tails do not bind to the proximal face but rather to the opposite side of Hfq, the so-named ‘‘distal’’ face (35, 36). The In addition to its role in the ␴S stress response, Hfq regulates structural mode of poly(A) binding and recognition is com- the expression of multiple virulence factors and —which pletely unknown but critical for a full understanding of Hfq are located in pathogenicity islands in several pathogenic bacteria—and itself is considered a virulence factor (11–14).

Hfq is also involved in the regulation of a quorum sensing in Author contributions: T.M.L. and R.G.B. designed research; T.M.L. and P.V.-H. performed Vibrio cholerae and Vibrio harveyi by eliciting the sRNA- research; T.M.L. and P.V.-H. contributed new reagents/analytic tools; T.M.L., P.V.-H., and mediated destabilization of the mRNAs that encode the R.G.B. analyzed data; and T.M.L., P.V.-H., and R.G.B. wrote the paper. quorum-sensing master regulators hapR and luxR, respectively The authors declare no conflict of interest. (15). Further, Hfq plays a key role in the cellular response to This article is a PNAS Direct Submission. phosphosugar toxicity and low cellular iron concentrations (16– Freely available online through the PNAS open access option. 22). To combat phosphosugar toxicity, ptsG, which encodes a Data deposition: Coordinates and structure factors have been deposited in the Research sugar transporter, is bound by the sRNA SgrS and degraded by Collaboratory for Structural Bioinformatics database, www.rcsb.org (RCSB ID code 3GIB). the principal cellular endoribonuclease RNase E in an Hfq- 1To whom correspondence should be addressed. E-mail: [email protected]. mediated process. In times of iron scarcity, Hfq and its sRNA This article contains supporting information online at www.pnas.org/cgi/content/full/ partner, RhyB, direct the destruction of mRNAs that encode 0908744106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0908744106 PNAS Early Edition ͉ 1of6 Downloaded by guest on October 1, 2021 Fig. 1. The structure of Ec Hfq bound to poly(A) RNA. (A) Electron density maps for the RNA tripartite module. Composite Fo–Fc omit map densities are shown in blue (0.8␴) and 2FoϪFc densities are shown in white (1␴). A stick model of the distal-side binding A15 RNA is shown with carbons colored slate. Hfq is shown as a gray surface. (B) A ribbon diagram of the biologically relevant Hfq hexamer and A18 oligonucleotide superimposed onto the apo Hfq hexamer (green). The Hfq trimers and A9 oligonucleotides of the 2-fold related asymmetric units are gray and black, respectively. One of the 6 tripartite binding repeats is labeled A (adenosine site), R (purine nucleotide site), and E (entrance/exit site). (C) Side view. A stick model of the bound RNA on the distal side is shown with carbons colored white, with one tripartite binding unit colored bluish gray. A CPK model of AU5G RNA with green carbons is docked in the proximal pore, as seen in the Sa-Hfq-AU5G structure. Hfq is shown as a gray surface. Figures were produced with PyMol (64).

function in RNA decay and posttranscription regulation. Here, The Tripartite (A-R-E) RNA Binding Motif. Hfq binds the poly(A) tail we report the structure of Ec Hfq bound to a poly(A) tail. The by sectioning the sequence into 6 trinucleotide repeats with the structure reveals a unique RNA binding mechanism that is well 5Ј-adenosine nucleotide binding to the adenosine specificity site suited for not only poly(A)-tract binding but also triplet repeats (A site), the following nucleotide to the purine nucleotide of the type (adenine-purine-any base). The mechanistic impli- selectivity site (R site; so named because of its likely ability to cations of this promiscuous binding site in the biological func- bind guanosine as well as adenosine), and the third nucleotide to tions of Hfq are discussed. a nondiscriminatory RNA entrance/exit site (E site; Fig. 1). Thus, the distal face of hexameric Hfq has the capacity to bind Results and Discussion 18 nucleotides. Global Structure of the Hfq-Poly(A) Complex. A C-terminal trunca- The specificity of Hfq for adenosine nucleotides is effected tion mutant of Ec Hfq, lacking residues 70–102, was crystallized primarily by the A site, which is a surface-exposed groove with the oligoribonucleotide A15. This truncation, although composed of residues from ␤-strands 2 and 4 (Fig. 2A). Com- reported to disrupt rpoS binding (37), has little effect on poly(A) plementary hydrogen bonds between the peptide backbone binding affinity but moderately affects the stability of Ec Hfq amide and carbonyl oxygen groups of residue Gln-33 and the N7 (38). The Hfq-poly(A) structure was solved by molecular re- and exocyclic N6 atoms of the base ensure adenine specificity placement using the apo Ec Hfq structure (PDB ID code 1hk9) and concomitant discrimination against guanine (Fig. 2B). Ad- (39) as the search model and refined to Rwork and Rfree values of ditional adenine-Hfq interactions include base stacking against 22.3% and 25.9%, respectively, at 2.40 Å resolution (Table S1). the side chain of residue Leu-32 and a polar interaction between The asymmetric unit of the Hfq-poly(A) crystal contains 3 Hfq subunits and 9 adenosine nucleotides. The biologically relevant hexamer (6, 8) and an A18 oligoribonucleotide are generated by a crystallographic 2-fold, resulting in individual nucleotide bind- ing sites having 83.3% (15/18) occupancies. Regardless, the electron density of each nucleotide is clear (Fig. 1A). Each Hfq subunit contains an N-terminal alpha helix (␣1, residues 8–18) followed by an extensively twisted 5-stranded ␤-sheet (␤1, residues 22–26; ␤2, residues 29–39; ␤3, residues 43–47; ␤4, residues 51–55; ␤5, residues 61–64). Superimposition of the refined apo and poly(A)-bound Ec Hfq hexamers reveals no significant structural changes (root mean square deviation ϭ 0.50 Å) indicating the poly(A) binding site is preformed (Fig. 1B). The poly(A) RNA tract binds to the distal face of Hfq on which the phosphodiester backbone traces a circular, weaving pattern (Fig. 1 B and C). This circular binding Fig. 2. The adenosine specificity site of Hfq. (A) The adenosine specificity conformation provides a ready explanation for data that showed pocket is labeled A and lies flat on the distal face of Hfq. The 3Ј-proximal covalently closed circularized A-tracts bound to Ec Hfq more purine nucleotide selectivity pocket (R) is shown with its adenine ring sand- tightly than their linear A-tract counterparts (40). Most of the wiched in a crevice between 2 adjacent subunits (shown as magenta and cyan RNA is solvent exposed and sits on Hfq much like a crown on ribbons). The RNA is depicted as a gray worm with bases and sugars shown as the head of a monarch (Fig. 1C). None of the riboses take the gray filled structures. (B) The interaction between the A-site adenosine and C3Ј-endo conformation that is typically associated with A-RNA. Hfq. The interacting residues are labeled and shown as sticks or CPK models Ј with carbons colored gray, oxygens red, and nitrogens blue. The adenosine Rather, 7 take the C2 -endo pucker, and the remaining 2 the nucleotide is shown as a stick model with carbons colored white, oxygens red, closely related C1Ј-exo or C3Ј-exo conformation. Each adenine nitrogens blue, and phosphorus orange. Hydrogen bonds are depicted as base is found in the anti conformation. black dashes with distances in Ångstrom.

2of6 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0908744106 Link et al. Downloaded by guest on October 1, 2021 Table 1. Equilibrium dissociation constants (Kd) for Ec Hfq and selected oligonucleotides* Oligonucleotide Hfq (nM)

A3 1,770 Ϯ 90 A6 160 Ϯ 30 A16 1.4 Ϯ 0.9 A27 1.6 Ϯ 0.9 (GGA)2 88 Ϯ 28 (GGA)9 16 Ϯ 1 (GCA)9 70 Ϯ 4 (GCCA)9 42 Ϯ 4 Fig. 3. The purine nucleotide selectivity site of Hfq. (A) Top view showing all AU5G64Ϯ 10 interacting residues as CPK. (B) Side view. Selected RNA interacting residues are labeled and shown with sticks with carbons colored gray, oxygens red, and G6 14,900 Ϯ 1,000 nitrogens blue. The bound adenosine is shown as sticks with carbons colored C6 6,200 Ϯ 600 white, oxygens red, nitrogens blue, and phosphorus orange. All other atoms DNA-A 9,500 Ϯ 2,800 6 in the pocket are shown as white spheres. Prime numbers represent residues † Ϯ AAYAA element 7.7 1 from a neighboring Hfq subunit. Hydrogen bonds are depicted as black dashes *Binding constants and errors are the average and standard deviation of at with distances in Ångstrom. least 5 measurements. †The affinity of the AAYAA element has been measured only once, and the error is associated with the fit of the curve to the polarization data. 2,800 nM, which is nearly 60-fold higher than that of A6 (Table 1). In addition to its role in poly(A) binding, the R site is the most likely binding pocket for ATP and ADP as residue Tyr-25 has the N1 atom and the N␧ amide of residue Gln-52. The 5Ј been implicated in such binding (43). However, ATP or ADP/ phosphate group of the A site-bound adenosine nucleotide AMP binding to the A site cannot be excluded as either a interacts with the peptide backbone amide of residue Lys-31. secondary or primary binding site. Although the closest approach of the side chain N␧ group of The E site lies mostly above the surface of the distal face and

residue Lys-31 to the RNA is Ϸ6 Å, substitution of Lys-31 with makes no adenosine-protein interactions (Fig. 1 A and C). Thus, BIOCHEMISTRY alanine results in a 100-fold loss of affinity for A18 (35, 36) and the exposed E site offers little in the way of sequence discrim- likely reflects its importance in providing a positive electrostatic ination and likely represents the entrance and exit points of RNA surface to complement the negatively charged RNA sugar tracts. Despite its lack of direct interaction with the protein, the phosphate backbone and in productive electrostatic steering of 3 independent E-site nucleotides show well-conserved structures polynucleotides to the binding pockets of the distal face (41, 42). due to geometric constraints placed on nucleotide binding to the Pyrimidines could reciprocate the donor-acceptor hydrogen A and R sites. bonding pattern of the peptide backbone in the A site, but would require the base to be in higher-energy syn conformation, and for Hfq-Poly(A) Binding Mechanism Is Novel. The poly(A) binding uridine, require a Ϸ3 Å shift. This shift would disrupt the mechanism of Ec Hfq differs completely from that which Sa Hfq phosphate backbone hydrogen bond to Lys-31 and a key ribose- uses to bind small U-rich RNAs (34). Indeed, Sa Hfq binds a protein hydrogen bond in the R site (vide infra). Both pyrimi- U-rich oligoribonucleotide near a basic pore on the proximal dines also display steric clash with the carbonyl oxygen of Lys-31 face using 6 essentially identical nucleotide binding pockets (Fig. (Fig. S1) that clearly would contribute to a low distal-side binding 1C and and Fig. S2). Similar binding by Ec Hfq to small U-rich affinity of the oligoribonucleotide C6 (Kd ϭ 6,200 nM), which is oligonucleotides is expected in light of the strong sequence Ϸ39-fold weaker than A6 (Kd ϭ 160 nM), and the strong conservation of the Sa and Ec Hfq proximal sites (Fig. S2) and preference of AU5G to bind to its high-affinity proximal side site mutational studies on Ec Hfq that show only proximal side (Table 1). residues involved in the binding of U-rich and sRNA sequences The R site is a crevice that is formed between the ␤-sheets of (35, 36). Superimposing the hexamers of the Sa Hfq-AU5G and neighboring subunits (Figs. 2A and 3A). The adenine ring inserts Ec Hfq-A18 complexes results in an rmsd of 0.84 Å and shows no into the crevice and stacks against the side chains of residues steric clash between the proximally and distally bound RNAs. Tyr-25, Leu-26Ј, Ile-30Ј, and Leu-32Ј (where the prime denotes Combined, these findings provide structural evidence that a residues from the other subunit). Replacement of residue Tyr-25 single Hfq hexamer could facilitate sRNA-mRNA annealing by with alanine results in a 100-fold decrease in binding affinity for simultaneously binding the sRNA to its proximal face and the A18, and changing residue Ile-30 to alanine or aspartate lowers mRNA to its distal face. the affinity for A27 nearly 10-fold, thereby confirming the Intriguingly, despite belonging to the same superfamily and importance of these interactions in poly(A) binding (35, 36). The showing similar pinwheel-like binding of nucleotides to their exocyclic N6 of this adenine engages in a hydrogen bond to respective proximal faces, the 2-sided RNA binding mechanisms the O␧ of residue Gln-52Ј, thereby linking A and R sites. The of Ec Hfq and the Sm proteins of the U1 snRNP diverge with adenines bound in the A and R pockets are also connected by a respect to their distal sides (44). Indeed, an RNA stem-loop nondiscriminating water-mediated hydrogen bond between the structure extends above the Sm distal face and appears to have A-site N3 atom and R-site N7 atom. The exocyclic N6 of the minimal interaction with this surface of the heptamer. More- R-site adenine is also proximal to the N3 atom of the A-site over, the RNA of the spliceosomal U1snRNP complex threads adenine (d ϭ 3.6 Å) but is offset. The bases are not engaged in through the central pore of the Sm heptamer. Although Hfq stacking interactions. Additional Hfq-R-site adenine hydrogen might use similar pore threading with longer sRNAs, no evi- bonds are made between the adenine N1 and O␥ of residue dence currently supports this possibility. The divergent distal Thr-61 and the N3 and N␦ of residue Asn-28Ј via a water face binding mechanisms of the Hfq and Sm proteins are likely molecule (Fig. 3B). The R-site ribosyl 2Ј hydroxyl group makes a reflection of their different oligomer assembly mechanisms a hydrogen bond to the carbonyl oxygen of residue Gly-29, likely and differently evolved functions in RNA processes (45). contributing to the preference of Hfq for RNA over DNA. The poly(A) binding mechanism of Hfq also differs entirely Indeed, the Hfq dissociation constant for DNA-A6 is 9,500 Ϯ from that used by the human poly(A) binding protein (hPABP),

Link et al. PNAS Early Edition ͉ 3of6 Downloaded by guest on October 1, 2021 which instead uses RRM (RNA recognition motif) motifs, the exocyclic N2 would be able to make a hydrogen bond to the ␤-sheets of which form a long trough to bind an extended, rather hydroxyl group of residue Ser-60. This finding implies that RNA than circular, RNA (46). For the greater part, hPAPB uses side binding to the distal face requires only 2 consecutive purines, of chain/base contacts to ensure specificity for adenines. One which only one must be an adenine. Given the stereochemical adenine is read by the peptide backbone, but beyond this there constraints of the A site, A-site binding adenosine nucleotides is little in common between the poly(A) binding mechanisms of must be separated by at least 2 nucleotides and adjacent to either Hfq and PABP. Indeed, even the peptide backbone readout of a3Ј adenosine or guanosine nucleotide. Hence, RNA with adenine differs between them; PABP contacts the N6 and N1 embedded triplets (5Ј-A-R-N-3Ј)i, (where R is a purine nucle- atoms, and Hfq, the N6 and N7 atoms. Thus, eukaryotes and otide and N is any nucleotide) should bind the distal face of Hfq. prokaryotes exploit different mechanisms to bind RNA poly(A) In accord, Hfq binds (GGA)9, which contains 8 successive AGG tails. triplets, with a Kd ϭ 16 nM—an only 10-fold lower affinity than The multitriplet purine binding mode of Hfq-poly(A) RNA Hfq binding to (AAA)9 (Table 1 and Fig. S4). The less ener- does have some parallels with the RNA binding mechanism of getically favorable binding of guanosine to the R site originates TRAP, the Bacillus subtilis trp RNA binding attenuation protein, in part by the required relocation of the O␧ oxygen of residue whereby 11 GAG/UAG triplets bind to the upper perimeter of Gln-52 to below the adenine ring in the A site. (GCA)9, which that undecameric protein (47). Unlike Hfq-poly(A) binding, contains 8 successive AGC triplets, binds Hfq with a Kd ϭ 70 nM. each GAG triplet binds to the edges of the ␤-sheet of TRAP. In The higher Kd of (GCA)9 might result from its ability to form a addition, base-base stacking of the GAG triplet is an integral stabile stem-loop structure (as predicted by mFold) (55) that part of the binding mechanism, and each guanine is recognized would be unsuitable for binding the distal face and require an specifically by engaging in one or more hydrogen bonds to additional energy input to melt this RNA secondary structure. proximal side chains. Neither of these is found in the Hfq- From the Ec Hfq-A15 structure we also reasoned that an extra poly(A) complex. Like TRAP, the B. subtilis antiterminator nucleotide added 3Ј to the E-site nucleotide should not interfere protein HutP also binds UAG triplets, but again the RNA with distal-side RNA binding. Remarkably, (GCCA)9, in which binding mechanism of this hexamer shows little resemblance to a second C is inserted to form 8 consecutive AGCC quartets and that used by Hfq (48). Indeed, each subunit of HutP uses residues disrupts the predicted RNA secondary structure of (GCA)9, from two loops to engage in side chain-purine ring hydrogen binds Hfq with Ϸ1.7-fold higher affinity (Kd ϭ 42 nM) than bonds to the adenosine and guanosine nucleotides. Moreover, (GCA)9. the adenine and guanine bases of each UAG triplet are stacked. The effect of (A-R-N) tract length on affinity was also tested. One similarity is found: like the adenosines bound to the E-site (A-R-N)2 tracts bind Hfq less well than longer tracts but still with of Hfq, the uridines of the HutP-bound UAG triplets are not considerable strength, whereby the affinity for Hfq drops only engaged in sequence-specifying contacts. 5-fold for (GGA)2, as compared with (GGA)9, and Ϸ10-fold in Hfq-poly(A) binding does bear a superficial resemblance to comparison with A6 (Table 1). The addition of a third G to the the RNA trinucleotide binding mechanism of the RsmA/CsrA triplet, i.e., (GGG)2 or G6, results in a Ϸ170-fold loss of affinity (carbon storage regulator protein) family (49, 50). RsmA/CsrA as compared with (GGA)2, and supports the premise that a binds as a dimer to 2 stem-loop structures that contain ANG- guanine cannot bind the A site strongly. Indeed, even the single GAN cores to regulate posttranscriptionally virulence factors, AAA triplet binds better than G6 (Table 1). However, caution carbohydrate metabolism, biofilm production, and motility (51). must be used in interpreting these latter data, as A3 and A6 might The GGA triplets are highly conserved amongst sRNAs that are bind the proximal site as well as the distal site. Regardless, the bound by RsmA/CsrA family members (52). The conserved GG significant gain in affinity of Hfq for longer A-R-N tracts can be pair of each bound oligoribonucleotide forms part of a 3-nucle- attributed to an increase in the local concentration of these otide loop that is recognized by a ␤-strand and loop via triplets, thus allowing the tract to zipper cooperatively into place, noncontiguous peptide backbone base-specific hydrogen bonds and the positive electrostatic nature of the distal face (56), which and an arginine side chain hydrogen bond to the N7 of the more through electrostatic steering (41, 42) can attract longer nucle- 3Ј guanine. Also different from the poly(A) binding mechanism otides more effectively. Although the distal face can accommo- of Hfq, these guanines are coplanar and form a hydrogen bond date 6 (A-R-N) triplets, once 5 A-R-E sites are filled, the between the N2 amino group of one base to the N7 of the other. presence of additional triplets does not affect binding affinity The conserved adenine of the GGA triplet is part of the RNA (compare A16 and A27, the values of which parallel those stem and read by the main-chain NH and CO group of ␤-strand reported previously) (refs. 35 and 40, and Table 1). residue Ile-3. Like Hfq binding to the A-site adenine, RsmA/ CsrA utilizes the Hoogsteen edge of the base (Fig. 2B). Given the Functional Implications of Hfq Binding to Poly(A-R-N) Tracts. The presence of at least 2 functionally important and distinct RNA finding that Hfq binds poly(A-R-N) sequences with high affinity binding surfaces, which can be filled simultaneously (8, 53), Hfq has wide-ranging functional consequences. Indeed, these data is qualitatively similar to the small RNA quality control protein and the finding that Hfq inhibits PNPase (29), a component of Ro, which also uses separate surfaces to bind its multiple RNA the degradosome that adds AG-rich polynucleotide tails to ligands (54). mRNAs undergoing degradation, suggest that Hfq modulates PNPase activity by blocking access of this exoribonuclease to the Hfq Binds Poly(A-R-N) and Poly(A-R-N-N؅) Tracts. At first glance, the 3Ј end of the A-R-N tracts through its ability to bind such tracts A and R sites seem ideal for adenine base-only binding. How- to its distal face. Hfq has been shown already to protect poly(A) ever, modeling studies suggested that the R site could also tails from exonucleolytic degradation by PNPase, and similar accommodate a guanine base but neither pyrimidine, which are protection of A-R-N tracts is logical (28). The Hfq-poly(A) unable to reach deeply enough into the pocket to engage in direct structure also suggests how Hfq changes PAP I from a distrib- interactions with Hfq (Fig. S3). The inability of pyrimidines to utive to a processive enzyme, but only after the addition of Ϸ20 contact R-site residues would essentially preclude tight binding adenosine nucleotides to the mRNA, i.e., the complete filling of by [A-(pyrimidine)-N] tracts to the distal face. By contrast, the Hfq poly(A) binding site (27). Hfq binding to the A-tract guanosine binding to this site requires only a simple rotation of would remove the growing poly(A) tail from the PAP I catalytic the Gln-52 side chain to make an N␧-O6 hydrogen bond. Such pocket, thereby aiding product dissociation and preventing prod- rotation is facilitated by the lack of any hydrogen bonds between uct inhibition or backtracking. Intriguingly, in the presence of the N␧ and the rest of the protein. Moreover, the guanine Hfq, PAP I produces poly(A) tails in vivo that are on average

4of6 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0908744106 Link et al. Downloaded by guest on October 1, 2021 Ϸ20 nucleotides longer, again consistent with filling the 6 ation. Crystals were grown by hanging-drop vapor diffusion from equal ␮ ␮ ␮ ϩ ␮ (A-R-N) binding sites on Hfq (29). volume mixtures of 200 M Hfq (hexamer) and 200 MA15 (2 L 2 L) added The broader biological significance of high-affinity Hfq-(A- to an equal volume of the crystallization buffer [40% MPD and 0.1 M CHES (pH 9.5); 4 ␮L Hfq-A15 to 4 ␮L crystallization buffer]. The crystals took the orthor- R-N)i tract binding can be seen by the recent identification of functionally important Hfq binding-A-R-N tracts in the up- hombic space group P21212 and diffract to 2.40-Å resolution. X-ray intensity data were collected at the Advanced Light Source (ALS) using beamline 8.2.2, stream leader sequences of rpoS mRNA, including the AAYAA and merged and scaled with MOSFLM and SCALA (Table S1) (59). element, which contains an (AAN)4 tract. The AAYAA element is critical for rpoS interaction with its regulatory sRNA, DsrA Structure Determination and Refinement. The Hfq-A15 structure was solved by (57). Indeed, we found the AAYAA element (GGGAACAA- molecular replacement using MOLREP (60) and the apo Ec Hfq (39) as the CAAGAAGUUA) binds truncated Ec Hfq very tightly (Kd ϭ 7.7 search model (PDB ID code 1hk9). This model was then subjected to rigid body nM). Given that the AAYAA element in the 5Ј UTR of rpoS refinement followed by simulated annealing in CNS (61), and positional and begins at nucleotide Ϫ370, a number of potential (A-R-N)i thermal factor refinement with REFMAC in the CCP4 package (60). The final tract-Hfq binding sites might exist in other sRNA-regulated composite omit map was generated by CNS (61). An Hfq-A9 complex model messages that contain extended upstream regions. Moreover, the was refined to Rwork and Rfree values of 22.3% and 25.9%, respectively. Refinement statistics are given in Table S1. The final Hfq-A9 model includes an functional importance of such Hfq-binding (A-R-N)i tracts within mRNA coding sequences (CDS) is likely, and supported Hfq trimer [residues 6–69 (chain A), 4–68 (chain B), 6–67 (chain C)], one RNA by the recent finding that the Hfq-associated sRNA, MicC, molecule 9-mer, 3 CHES buffer molecules, and 9 water molecules; it has excellent stereochemistry (Table S1) with but one Ramachandran outlier (62) targets the CDS of ompD mRNA and that the target site, which per subunit (Asn-48) as was seen in the original apo Ec Hfq model (39). contains an AACAAA stretch, is bound (albeit weakly) by Hfq (58). Finally, a simple search of the E. coli genome for poly(A- Fluorescence Polarization. Fluorescence polarization measurements were col- R-N)i sequences (http://genolist.pasteur.fr/Colibri/ using lected with a PanVera Beacon Fluorescence Polarization System (PanVera). ‘‘Search Pattern’’) revealed 69 unique (A-R-N)5 tracts in the Samples were excited at 490 nm, and emission was measured at 530 nm. mRNA translation initiation regions (spanning nucleotides 5Ј-fluoresceinated oligonucleotides were purchased from Oligos, Etc. or Ϫ100 to ϩ1) with the average start site at position Ϫ44 and 14 Sigma-Genosys. The binding buffer used for all measurements contained 20 (Ϸ20%) overlapping the Shine-Dalgarno (SD) site (Ϫ23 to ϩ1). mM sodium phosphate (pH 6.5), 150 mM NaCl, and 0.5 mM EDTA. Hfq (in 50 mM Tris 8.0, 150 mM NaCl, and 0.5 mM EDTA) and was serially titrated into the A similar search for (A-R-N)4 tracts yielded 394 unique examples with 72 (Ϸ18%) overlapping the SD box and an average start site cuvette, which contained 0.15 nM 5Ј-fluoresceinated oligoribonucleotide. The measurements were performed at 4 °C. Samples were incubated 15 s at nucleotide Ϫ39. Thus, Hfq is expected to have a more BIOCHEMISTRY before each polarization measurement, ensuring equilibrium binding. The expansive role as a riboregulator than already described. data were plotted using KaleidaGraph (Synergy Software), and the generated Materials and Methods curves were fit by nonlinear least squares regression assuming a bimolecular model such that the Kd values represent the protein concentration at half- Crystallization and Data Collection. The truncated E. coli Hfq protein (residues maximal oligonucleotide binding (63). 2–69) was overexpressed in an E. coli ⌬hfq derivative of the ER2566 strain using the Impact-CN system (New England Biolabs) and purified as described ACKNOWLEDGMENTS. We thank the Advanced Light Source (ALS) and their (8). Before crystallization, multiple buffer exchanges were carried out with support staff. ALS is supported by the Director, Office of Science, Office of 100 mM NaCl, 25 mM Tris (pH 8.0), and 0.5 mM EDTA to remove excess DTT. A Basic Energy Sciences, Materials Sciences Division of the U.S. Department of 2-mM solution of A15 was made by dissolving solid oligoribonucleotide in 10 Energy at the Lawrence Berkeley National Laboratory. This work was sup- mM Na cacodylate (pH 6.5) and used without further purification or denatur- ported by the Robert A. Welch Foundation Grant G-0040 (to R.G.B.).

1. Sun X, Zhulin I, Wartell RM (2002) Predicted structure and phyletic distribution of the 17. Lenz DH, et al. (2004) The small RNA chaperone Hfq and multiple small RNAs control RNA-binding protein Hfq. Nucleic Acids Res 30:3662–3671. quorum sensing in Vibrio harveyi and Vibrio cholerae. Cell 118:69–82. 2. Valentin-Hansen P, Eriksen M, Udesen C (2004) The bacterial Sm-like protein Hfq: A key 18. Morita T, Maki K, Aiba H (2005) RNase E-based ribonucleoprotein complexes: Mechan- player in RNA transactions. Mol Microbiol 51:1525–1533. ical basis of mRNA destabilization mediated by bacterial noncoding RNAs. Genes Dev 3. Tsui HC, Leung HC, Winkler ME (1994) Characterization of broadly pleiotropic pheno- 19:2176–2186. types caused by an Hfq insertion mutation in Escherichia coli K-12. Mol Microbiol 19. Go¨rke B, Vogel J (2008) Noncoding RNA control of the making and breaking of sugars. 13:35–49. Genes Dev 22:2914–2925. 4. Storz G, Opdyke JA, Zhang AX (2004) Controlling mRNA stability and translation with 20. Vanderpool CK (2007) Physiological consequences of small RNA-mediated regulation small, noncoding RNAs. Curr Opin Microbiol 7:140–144. of glucose-phosphate stress. Curr Opin Microbiol 10:146–151. 5. Zhang A, et al. (1998) The OxyS regulatory RNA represses rpoS translation and binds the 21. Aiba H (2007) Mechanism of RNA silencing by Hfq-binding small RNAs. Curr Opin Hfq (HF-I) protein. EMBO J 17:6061–6068. Microbiol 10:134–139. 6. Zhang A, et al. (2002) The Sm-like Hfq protein increases OxyS RNA interaction with 22. Vanderpool CK, Gottesman S (2004) Involvement of a novel transcriptional activator target mRNAs. Mol Cell 9:11–22 (see comment). and small RNA in post-transcriptional regulation of the glucose phosphoenolpyruvate 7. Wassarman KM, et al. (2001) Identification of novel small RNAs using comparative phosphotransferase system. Mol Microbiol 54:1076–1089. genomics and microarrays. Genes Dev 15:1637–1651. 23. Masse´E, Escorcia FE, Gottesman S (2003) Coupled degradation of a small regulatory 8. Møller T, et al. (2002) Hfq: A bacterial Sm-like protein that mediates RNA-RNA inter- RNA and its mRNA targets in Escherichia coli. Genes Dev 17:2374–2383 (see comment). action. Mol Cell 9:23–30. 24. Masse´E, Gottesman S (2002) A small RNA regulates the expression of genes involved 9. Gottesman S (2004) The small RNA regulators of Escherichia coli: Roles and mecha- in iron metabolism in Escherichia coli. Proc Natl Acad Sci USA 99:4620–4625. nisms. Ann Rev Microbiol 58:303–328. 25. Udekwu KI, et al. (2005) Hfq-dependent regulation of OmpA synthesis is mediated by 10. Muffler A, et al. (1997) The RNA-binding protein HF-I plays a global regulatory role an antisense RNA. Genes Dev 19:2355–2366. which is largely, but not exclusively, due to its role in expression of the sigmaS subunit 26. Rasmussen AA, et al. (2005) Regulation of ompA mRNA stability: The role of a small of RNA polymerase in Escherichia coli. J Bacteriol 179:297–300. regulatory RNA in growth phase-dependent control. Mol Microbiol 58:1421–1429. 11. Figueroa-Bossi N, et al. (2006) Loss of Hfq activates the sigmaE-dependent envelope stress response in Salmonella enterica. Mol Microbiol 62:838–852. 27. Hajnsdorf E, Re´gnier P (2000) Host factor Hfq of Escherichia coli stimulates elongation 12. Johansen J, Rasmussen AA, Overgaard M, Valentin-Hansen P (2006) Conserved small of poly(A) tails by poly(A) polymerase I. Proc Natl Acad Sci USA 97:1501–1505. non-coding RNAs that belong to the sigmaE regulon: Role in down-regulation of outer 28. Folichon M, et al. (2003) The poly(A) binding protein Hfq protects RNA from RNase E membrane proteins. J Mol Biol 364:1–8. and exoribonucleolytic degradation. Nucleic Acids Res 31:7302–7310. 13. Sittka A et al. (2008) Deep sequencing analysis of small noncoding RNA and mRNA 29. Mohanty BK, Maples VF, Kushner SR (2004) The Sm-like protein Hfq regulates poly- targets of the global post-transcriptional regulator, Hfq. PLoS Genet 4:e1000163. adenylation dependent mRNA decay in Escherichia coli. Mol Microbiol 54:905–920. 14. Ding YP, Davis BM, Waldor MK (2004) Hfq is essential for Vibrio cholerae virulence and 30. Wilusz CJ, Wilusz J (2008) New ways to meet your (3Ј) end oligouridylation as a step on downregulates sigma(E) expression. Mol Microbiol 53:345–354. the path to destruction. Genes Dev 22:1–7. 15. Meibom KL, et al. (2009) Hfq, a novel pleiotropic regulator of virulence-associated 31. Mohanty BK, Kushner SR (2006) The majority of Escherichia coli mRNAs undergo genes in Francisella tularensis. Infect Immun 77:1866–1880. post-transcriptional modification in exponentially growing cells. Nucleic Acids Res 16. Fantappie`L, et al. (2009) The RNA chaperone Hfq is involved in stress response and 34:5695–5704. virulence in Neisseria meningitidis and is a pleiotropic regulator of protein expression. 32. Carpousis AJ (2007) The RNA degradosome of Escherichia coli: An mRNA-degrading Infect Immun 77:1842–1853. machine assembled on RNase E. Annu Rev Microbiol 61:71–87.

Link et al. PNAS Early Edition ͉ 5of6 Downloaded by guest on October 1, 2021 33. Worrall JA, et al. (2008) Reconstitution and analysis of the multienzyme Escherichia coli 49. Mercante J, et al. (2006) Comprehensive alanine-scanning mutagenesis of Escherichia RNA degradosome. J Mol Biol 382:870–883. coli CsrA defines two subdomains of critical functional importance. J Biol Chem 34. Schumacher MA, et al. (2002) Structures of the pleiotropic translational regulator Hfq 281:31832–31842. and an Hfq-RNA complex: A bacterial Sm-like protein. EMBO J 21:3546–3556. 50. Schubert M, et al. (2007) Molecular basis of messenger RNA recognition by the specific 35. Sun XG, Wartell RM (2006) Escherichia coli Hfq binds A(18) and DsrA domain II with bacterial repressing clamp RsmA/CsrA. Nat Struct Mol Biol 14:807–813. similar 2:1 Hfq(6)/RNA stoichiometry using different surface sites. Biochemistry 51. Babitzke P, Baker CS, Romeo T (2009) Regulation of translation initiation by RNA 45:4875–4887. binding proteins. Annu Rev Microbiol 63:27–44. 36. Mikulecky PJ, et al. (2004) Escherichia coli Hfq has distinct interaction surfaces for DsrA, 52. Valverde C, Lindell M, Wagner EG, Haas D (2004) A repeated GGA motif is critical for rpoS and poly(A) RNAs. Nat Struct Mol Biol 11:1206–1214. the activity and stability of the riboregulator RsmY of Pseudomonas fluorescens. J Biol 37. Vecerek B, et al. (2008) The C-terminal domain of Escherichia coli Hfq is required for Chem 279:25066–25074. regulation. Nucleic Acids Res 36:133–143. 53. Lease RA, Woodson SA (2004) Cycling of the Sm-like protein Hfq on the DsrA small 38. Arluison V, et al. (2004) The C-terminal domain of Escherichia coli Hfq increases the regulatory RNA. J Mol Biol 344:1211–1223. stability of the hexamer. Eur J Biochem 271:1258–1265. 54. Fuchs G, et al. (2006) Structural and biochemical basis for misfolded RNA recognition 39. Sauter C, Basquin J, Suck D (2003) Sm-like proteins in Eubacteria: The crystal structure by the Ro autoantigen. Nat Struct Mol Biol 13:1002–1009. of the Hfq protein from Escherichia coli. Nucleic Acids Res 31:4091–4098. 55. Zuker M (2003) Mfold web server for nucleic acid folding and hybridization prediction. 40. de Haseth PL, Uhlenbeck OC (1980) Interaction of Escherichia coli host factor protein Nucleic Acids Res 31:3406–3415. with oligoriboadenylates. Biochemistry 19:6138–6146. 56. Brennan RG, Link TM (2007) Hfq structure, function and ligand binding. Curr Opin 41. Lawrenson ID, Stockley PG (2004) Kinetic analysis of operator binding by the E. coli Microbiol 10:125–133. methionine repressor highlights the role(s) of electrostatic interactions. FEBS Lett 57. Soper TJ, Woodson SA (2008) The rpoS mRNA leader recruits Hfq to facilitate annealing 564:136–142. with DsrA sRNA. RNA 14:1907–1917. 42. Selzer T, Schreiber G (1999) Predicting the rate enhancement of protein complex 58. Pfeiffer V, et al. (2009) Coding sequence targeting by MicC RNA reveals bacterial mRNA formation from the electrostatic energy of interaction. J Mol Biol 287:409–419. silencing downstream of translational initiation. Nat Struct Mol Biol 16:840–846. 43. Arluison V, et al. (2007) Sm-like protein Hfq: Location of the ATP-binding site and the 59. Leslie AGW (1992) Recent changes to the MOSFLM package for processing film and image effect of ATP on Hfq–RNA complexes. Protein Sci 16:1830–1841. plate data. Joint CCP4 ϩ ESF-EAMCB Newsletter on Protein Crystallography, No. 26. 44. Pomeranz Krummel DA, et al. (2009) Crystal structure of human spliceosomal U1 snRNP 60. Collaborative Computational Project, Number 4 (1994) The CCP4 suite: Programs for at 5.5 A resolution. Nature 458:475–480. protein crystallography. Acta Crystallographica D Biol Crystallogr 50:760–763. 45. Wilusz CJ, Wilusz J (2005) Eukaryotic Lsm proteins: Lessons from bacteria. Nat Struct 61. Bru¨ nger AT, et al. (1998) Crystallography & NMR system: A new software suite for Mol Biol 12:1031–1036. macromolecular structure determination. Acta Crystallogr D Biol Crystallogr 54:905–921. 46. Deo RC, Bonanno JB, Sonenberg N, Burley SK (1999) Recognition of polyadenylate RNA 62. Laskowski RA, MacArthur MW, Moss DS, Thornton JM (1993) PROCHECK: A program to by the poly(A)-binding protein. Cell 98:835–845. check the stereochemical quality of protein structures. J App Crystallogr 26:283–291. 47. Antson AA, et al. (1999) Structure of the trp RNA-binding attenuation protein, TRAP, 63. Lundblad JR, Laurance M, Goodman RH (1996) Fluorescence polarization analysis of bound to RNA. Nature 401:235–242. protein-DNA and protein-protein interactions. Mol Endocrinol 10:607–612. 48. Kumarevel T, Mizuno H, Kumar PK (2005) Structural basis of HutP-mediated anti- 64. Delano WL (2002) The PyMOL Molecular Graphics System (DeLano Scientific, San termination and roles of the Mg2ϩ ion and L-histidine ligand. Nature 434:183–191. Carlos, CA). Available at http://pymol.sourceforge.net/.

6of6 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0908744106 Link et al. Downloaded by guest on October 1, 2021