The C-terminal domain of Escherichia coli Hfq increases the stability of the hexamer Ve ´ ronique Arluison 1 , Marc Folichon 1 , Sergio Marco 2 , Philippe Derreumaux 3 , Olivier Pellegrini 1 , Je ´ ro ˆ me Seguin 4 , Eliane Hajnsdorf 1 and Philippe Regnier 1 1 Institut de Biologie Physico-Chimique CNRS UPR 9073 conventionne ´ e avec l’universite ´ Paris 7, Paris, France; 2 Institut Curie CNRS UMR 168, Paris, France; 3 Institut de Biologie Physico-Chimique CNRS UPR 9080, Paris, France; 4 Service de Biophysique des Fonctions Membranaires, DBJC/CEA & URA 2096 CNRS, Gif/Yvette, France The Hfq (Host factor 1) polypeptide is a nucleic acid binding protein involved in the synthesis of many polypeptides. Hfq particularly affects the translation and the stability of several RNAs. In an earlier study, the use of fold recog- nition methods allowed us to detect a relationship between Escherichia coli Hfq and the Sm topology. This topology was further validated by a series of biophysical studies and the Hfq structure was modelled on an Sm protein. Hfq forms a b-sheet ring-shaped hexamer. As our previous study pre- dicted a large number of alternative conformations for the C-terminal region, we have determined whether the last 19 C-terminal residues are necessary for protein function. We find that the C-terminal truncated protein is fully capable of binding a polyadenylated RNA (K d of 120 p M vs. 50 p M for full-length Hfq). This result shows that the functional core of E. coli Hfq resides in residues 1–70 and confirms previous genetic studies. Using equilibrium unfolding studies, how- ever, we find that full-length Hfq is 1.8 kcalÆmol )1 more stable than its truncated variant. Electron microscopy ana- lysis of both truncated and full-length proteins indicates a structural rearrangement between the subunits upon trun- cation. This conformational change is coupled to a reduction in b-strand content, as determined by Fourier transform infra-red. On the basis of these results, we propose that the C-terminal domain could protect the interface between the subunits and stabilize the hexameric Hfq structure. The origin of this C-terminal domain is also discussed. Keywords: RNA binding protein; Sm-like (L-Sm); b-topol- ogy; urea equilibrium unfolding; electron microscopy. Hfq (Host factor 1) of Escherichia coli is an 11 kDa polypeptide which was originally discovered as a host factor required for the replication of bacteriophage Qb RNA [1]. However, by inactivating of the Hfq gene, it was later demonstrated that it is involved in a variety of other metabolic pathways [2–4]. In particular, Hfq has been implicated in the translation and the control of the stability of certain mRNAs. For example, Hfq has been shown to interfere directly with ribosome binding of the ompA transcript, exposing the transcript to ribonucleases [5,6]. It has also been implicated in the stimulation of the elongation of poly(A) tails by poly(A) polymerase, leading to poly(A)- dependent mRNA degradation [7]. Finally, it has been shown to be involved in the translation regulation of the rpoS transcript, encoding the r S subunit of RNA poly- merase and, as a consequence, influences the expression of many stationary phase genes whose transcription depends on r S [3]. This last effect was the first cellular role observed for Hfq and has since been the subject of much attention because Hfq influences rpoS translation by altering the binding of small RNAs (sRNAs) to their complementary target sequence [8–11]. The sRNAs involved in rpoS translation control are OxyS, DsrA, RprA. More recently, it has been also shown that many other sRNA can interact with Hfq, pointing to a global role of the protein in facilitating sRNA function [12,13]. Little is known about the mechanism of Hfq action. It has been shown to bind strongly to single-stranded RNAs that are A and U rich. Taking into account its ability to rescue a folding trap of a splicing defective intron [14] and its requirement for the activity of many sRNAs [11,15], it has been proposed to be an RNA chaperone. The interaction between Hfq and RNA may increase the propensity of RNA to interact with itself or other RNAs, but also its susceptibility to nucleases or poly(A) polymerase. Recently, the Sm-like nature of Hfq was proposed on the basis of weak sequence similarities between the N-terminal domain of Hfq and the Sm and Sm-like (L-Sm) proteins of eukaryotes and archaea [11,15,16]. These proteins are components of the spliceosome complex and are also involved in other RNA metabolism steps [17,18]. The relationship between Hfq and the Sm topology was further confirmed by using fold recognition methodology and by a series of biophysical and biochemical studies. The structure of Hfq from E. coli was modelled on an Sm protein [16] and Correspondence to Philippe Regnier, Institut de Biologie Physico- Chimique CNRS UPR 9073 conventionne ´ e avec l’universite ´ Paris 7, 13 rue P. et M. Curie, 75005 Paris, France. Fax: + 33 1 58 41 50 20, Tel.: + 33 1 58 41 51 32, E-mail: regnier@ibpc.fr Abbreviations: ATR-FTIR, attenuated total reflectance fourier transform infra-red; EM, electron microscopy; Hfq ec , E. coli Hfq; Hfq f , Hfq full-length; Hfq Nter , Hfq lacking the 19 last amino acid; sRNAs, small RNAs. (Received 19 December 2003, revised 2 February 2004, accepted 6 February 2004) Eur. J. Biochem. 271, 1258–1265 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04026.x the model further confirmed by determination of the X-ray structure of Staphylococcus aureus and E. coli Hfq proteins [19,20]. Hfq forms a hexameric ring shaped structure [11,15,16,19], essentially b-sheet in character, with a cationic inner hole implicated in RNA binding [19]. However, the crystal structures available for two Hfq proteins are restricted to the N-terminal part of the protein (% 60 amino acids): the short C-terminal region of S. aureus Hfq was disordered and not resolved in its crystal structure and the C-terminal region of E. coli was genetically removed for its crystallization [19,20]. In addition, the last 26 amino acids were not present in the modelled E. coli Hfq structure, because the secondary structure prediction and fold-recog- nition methods proposed a large variety of conformations for this region. To investigate the impact of the C-terminal part on Hfq structure and function, we removed the 19 C- terminal residues proteolytically. RNA binding properties, electron microscopy and urea unfolding analysis of the truncated protein are presented in this paper. They indicate that the C-terminal region has a structural and thermo- dynamic role in stabilizing the hexameric form, but does not affect the binding of a polyadenylated RNA. Materials and methods Unless otherwise specified, all enzymes and chemicals were either from Sigma or Merck-Biochemicals. Purification of Hfq C-terminal His-tagged Hfq was purified from the BL21(DE3) strain transformed with plasmid pTE607 as follows: cells from the induced culture were resuspended in 20 mL of buffer containing 20 m M Tris/HCl, pH 7.8, 0.5 M NaCl, 10% (v/v) glycerol and 0.1% (v/v) Triton X-100 at 4 °C. The suspension was passed through a French press (1200 bar) and centrifuged for 30 min at 15 000 g. Imidazole-HCl (pH 7.8) was added to the supernatant to reach a final concentration of 1 m M . The resulting suspension was applied to a 1 mL Ni 2+ –nitrilotriacetic acid column (Qiagen). The resin was then washed sequentially with % 15 column volumes of: (a) 20 m M Tris/HCl, pH 7.8 buffer containing 0.3 M NaCl and 20 m M imidazole and (b) 50 m M sodium phosphate, pH 6.0 buffer containing 0.3 M NaCl, and Hfq was finally eluted with a buffer containing 50 m M sodium phosphate, pH 6.0, 0.3 M NaCl and 250 m M imidazole. The fractions containing Hfq were analysed by SDS/PAGE, pooled andheatedto80°C for 15 min. Insoluble material was removed by centrifugation and the supernatant was dialysed in buffer containing 50 m M Tris/HCl, pH 7.5, 1m M EDTA, 50 m M NH 4 Cl, 5% (v/v) glycerol and 0.1% (v/v) Triton X-100. The protein was stored at 4 °C. Protein concentrations were determined by measuring the absorption at 280 nm (e 280 at 1 mgÆmL )1 ¼ 0.34) or by using the Bradford assay (Bio-Rad) [21] with bovine serum albumin as a standard. Limited proteolysis and mass spectra analysis Digestion of Hfq by chymotrypsin (ROCHE biochemi- cals, Switzerland) was performed at 20 °Cin50m M Tris/HCl, pH 8 containing 100 m M NaCl. Chymotrypsin (1 ng) was used for 1 lg of Hfq (concentration 1mgÆmL )1 , 200 lL). Aliquots of 20 lL were withdrawn at different times and the reaction was stopped by adding the universal protease inhibitor a 2 -macroglobulin (Roche Biochemicals) at the same final concentration as that of the enzyme. Digestion was monitored on 16.5% Tris/tricine SDS/PAGE [22] without a ÔspacerÕ gel. The N-terminal fragment (Hfq Nter ) generated by chymotrypsin (% 9 kDa) was purified by extensive dialysis (molecular mass cut-off 3500 Da) against 10 m M Tris/HCl, pH 8, buffer containing 80 m M NaCl, 1% (v/v) glycerol, 0.01% (w/v) dodecyl-b- D -maltoside. The separation of the C-terminal fragment (% 2 kDa) from the the N-terminal fragment was verified by SDS/PAGE. For mass spectrometry analyses, samples were desalted using Zip Tips C4 (Millipore), as described in the technical manual. MALDI-TOF mass spectra were recorded with a Voyager STR-DE (Perspective Biosystems Inc., Framing- ham, MA, USA) mass spectrometer equipped with a delayed extraction device. Samples were recorded either in positive reflectron or linear mode. Calibration was performed using standards. For electrospray mass spectra, the sample was dissolved in a H 2 O/acetonitrile/acetic acid mix (49 : 50 : 1; v/v/v) and the spectrum was acquired in positive mode on a QTOF II mass spectrometer (Micromass). A reconstructed spectrum from multiply charged ions was obtained using the MaxEnt algorithm (Max Ent I, Micromass). Electron microscopy and image analysis Aliquots of full-length (Hfq f ) and proteolysed Hfq (Hfq Nter ) were adsorbed to 400 mesh carbon-coated grids and stained with 1% (w/v) uranyl acetate. Samples were observed with a Philips CM120 electron microscope at an accelerating voltage of 120 kV and nominal magnification of 75 000·. Images with a final pixel size of 0.26 nm were recorded using a GATAN ssCCD camera. A total number of 3057 single particles from Hfq f , and 4833 from Hfq Nter , were windowed andalignedbyusing X - MIPP software [23] before classifica- tion by a self-organizing Kohonen neural network [24]. Average and standard deviation images of the major groups, from each protein sample, were computed and, after performing a rotational power spectra analysis, a sixfold symmetry was imposed. Resolution was estimated to 18 A ˚ using the SSNR method [25]. Once the image was filtered at the computed resolution and sixfold symmetry imposed, a differential map between the proteolysed and nonproteolysed Hfq average images was calculated by using the student-t algorithm as implemented in the X - MIPP software with a significance level a ¼ 0.05. Gel-shift assay Labelled RNA fragments (2.5 p M ) were incubated with Hfq for 20 min at 37 °Cin50lLof10m M Tris/HCl, pH 8 buffer containing 80 m M NaCl, 1% (v/v) glycerol, 0.01% (w/v) dodecyl-b- D -maltoside. The RNA used consisted of rpsO mRNA with a 18 nucleotide poly(A) tail [7]. The RNA was titrated with an excess of Hfq, its concentration ranging from 15 p M to 10 n M [26]. Complexes were separated on native PAGE as described in Zhang et al. [27]. Ó FEBS 2004 The role of the C-terminal domain on Hfq (Eur. J. Biochem. 271) 1259 Infrared spectra Attenuated total reflectance (ATR)-FTIR spectra of Hfq [200 l M , previously dialysed overnight against 10 m M Tris/ HCl, pH 8 containing 80 m M NaCl and 0.01% (w/v) dodecyl-b- D -maltoside] were measured with a Bruker vector 22 spectrophotometer equipped with a 45  ndiamondATR attachment at 20 °C. The spectra are the average of 125 scans. Spectra were corrected for the linear dependence of the absorption measured by ATR on the wavelength. The water signal was removed by subtraction of a buffer spectrum. Analysis of the protein secondary structure composition within each Hfq was performed by deconvo- lution of the absorption spectrum as a sum of Gaussian components [28]. Equilibrium unfolding study Unfolding of the hexameric Hfq protein in urea was monitored with a PTI A1010 fluorimeter using the intrinsic tyrosine fluorescence of Hfq. Protein concentration was 0.1 mgÆmL ) 1 . Tyrosine residues were excited at 275 nm and emission monitored at 303 nm using a cell path-length of 1 cm. A 5 min incubation in urea was required to reach equilibrium. Analysis of denaturation curves was performed as described in [29]. The free energy of unfolding, DG u ,fits to equation: DG u ¼ DG H 2 O u À m½urea; where, DG H 2 O u is the free energy of unfolding in water and m represents the effectiveness of the denaturant in destabi- lizing the protein. The effect of urea on the quaternary structure of Hfq was determined by rate zonal centrifugation. The centrifugation was performed at 20 °C, overnight, at 40 000 r.p.m., in the 70 Ti rotor of a Beckman LE-80 ultracentrifuge. Hfq was denatured in a 50 m M Tris/HCl pH 8 buffer containing 100 m M NaCl and 6 M urea prior to centrifugation. Samples (100 lL) of 10 l M Hfqwereappliedonthetop of 25 mL 1–5% (w/v) saccharose gradients prepared in the buffer with or without urea. Fractions (1 ml) were collected and Hfq polypeptides precipitated with 1 vol. of cold 10% trichloroacetic acid were detected by dot-blot using anti-Hfq Ig. Gradients were calibrated using mass standards indica- ted (Fig. 4 legend). Results Two peptides were generated upon limited proteolysis of Hfq by chymotrypsin, whose masses were determined by ESI- and MALDI-TOF spectrometry. These masses, together with the location of aromatic amino acids in the sequence, allowed us to deduce that chymotrypsin cleaves after Tyr83, while all other aromatic residues (Phe11, 39, 42 and Tyr25, 55) of the N-terminal domain are not accessible to the protease. This cleavage generates an 83 amino acid fragment containing the N-terminal domain conserved in bacterial Hfqs (% 65 amino acids in Hfq Ec ) and a short part of the C-terminal domain which is only found in a limited number of Hfqs (Fig. 1). We used the truncated polypeptide generated by chymotrypsin to investigate whether Hfq lacking the C-terminal domain exhibits identical structural and RNA binding properties than the full-length protein. Electron microscopy analysis of Hfq f and Hfq Nter was performed in order to determine the effect of the C-terminal domain truncation on Hfq structure. Rotational power spectra analysis performed on average images demonstrated a sixfold symmetry for each Hfq multimer (data not shown). This analysis indicates that the C-terminal domain is not necessary for hexamerization of the protein. Once the symmetry was imposed, the average of the images present a different shape, with a stain-excluding region with a % 70 A ˚ maximum diameter (Fig. 2A) and a central stained region having % 20 A ˚ diameter. Indeed, the shape of full-length Hfq appeared sharper than the proteolysed one (Fig. 2B). The absolute value of the difference calculated from the two average images (a ¼ 0.05) showed that Hfq f and Hfq Nter differ in 12 regions arranged in the form of two crowns (Fig. 2C, top). The difference corresponding to the external crown appeared positive when the full-length minus that proteolysed is calculated (Fig. 2C, center-left), indicating the presence of a stain-excluding region, which can be assigned to the C-terminal domain not present in the proteolysed protein. This domain is at the origin of the sharpest shape of full-length Hfq (Fig. 2C, bottom-left) and extend beyond the canonical Sm fold. It should however be noted that these represents almost 20% of the subunit mass and the amount of density in this feature is small. It is thus probable that most of the peptide is not seen, because it is not contrasted by the negative stain. Interestingly, another difference is located on the internal circumference that appears positive when the images of proteolysed minus full-length were calculated (Fig. 2C, center-right). This suggests a conformational change at the interface between the subunits (Fig. 2C, bottom-right), that have been described as the strand b4ofchainBandthe strand b5 of chain A in the crystal structure [19,20]. As the electron microscopy data pointed to a conformational change at the subunits’ interface upon truncation of Hfq, we measured the secondary structure composition of full- length and truncated Hfq using infrared spectroscopy. FTIR spectra showed that full-length Hfq has 34 ± 2 residues in b-strands and 10 ± 1 residues in a-helices, while truncated Hfq has 28 ± 2 residues in b-strands and 9 ± 1 residues in helical conformation. Thus, protease cleavage causes a loss of six residues in b-sheet residues in the truncated protein. The margin of error in determining secondary structures by FTIR being of the order of 2% in comparing two similars (the error comes only from the deconvolution method), the 6% overall difference in signal is thus significant. We expect that this reduction in b-sheet character affects six amino acids located in the N-terminal region (i.e. the Sm-like domain) for three reasons: (a) the C-terminal region of E. coli Hfq is flexible in the molecular model [16], (b) that of S. aureus Hfq is disordered in the crystal structure [19] and (c) the FTIR spectrum of the C-terminal peptide is strongly dominated by a broad band at 1641 cm )1 , indicating an unordered structure. The ability of the conserved N-terminal fragment to bind polyadenylated rpsO mRNA alone was tested using gel- shift assays. Figure 3 shows the binding curves of Hfq f and 1260 V. Arluison et al. (Eur. J. Biochem. 271) Ó FEBS 2004 Hfq Nter . The corresponding equilibrium dissociation con- stants (K d ) were found to be 50 ± 10 p M and 120 ± 15 p M for Hfq and Hfq Nter , respectively. The minor increase in K d upon truncation suggests that the C-terminal domain of Hfq does not contribute significantly, if at all, to poly- adenylated RNA binding. As no major effect on RNA binding could be attributed to the C-terminal region, we determined the contribution of this region to the thermodynamic stability of Hfq. Equilib- rium unfolding studies of Hfq were performed using protein intrinsic fluorescence. Prior to analysis of the denaturation curves, the effect of urea on the quaternary structure of Hfq was determined by rate zonal centrifugation. This allows discrimination between the unfolded hexameric state (U 6 ) and monomeric state (6 U). Figure 4A shows that the native form of Hfq has an apparent molecular mass of 50 ± 10 kDa, while the unfolded state of Hfq has an average apparent molecular mass of 12 ± 5 kDa. This indicates that the unfolded state of Hfq was monomeric in urea. Urea unfolding of Hfq is accompanied by an overall increase in the fluorescence intensity, but not by a shift in the peak maximum (data not shown). This probably results from a quenching of tyrosine fluorescence in the native state. As Hfq possesses only three tyrosines (Tyr25, 55 and 83) and Tyr55 is located in strand 4 [19,20] at the interface between monomers, exposure of this tyrosine is probably largely responsible for the increase in fluorescence signal. The denaturation curves are presented in Fig. 4B. The simplest model, where hexamer dissociation and protein unfolding occur in a single step, fits the experimental data well. We rule out the possibility of a reversible model with more than two-states because DG° ¼ f [urea] is linear (if we had a stable intermediate state, we would observe two transitions and this is not true in our case). Confirmation of this two-state process is sustained by the observation that the hexamer dissociation – observed by rate zonal centrifu- gation – and protein unfolding – exposition of tyrosine or secondary structure disruption (results not shown) – occur Fig. 1. Multiple sequence alignment of various bacterial Hfqs. The alignment was produced with T-COFFEE. Amino acids characteristic of Hfq are indicated in black. Amino acids in light grey are conserved in most Hfqs and located in the b-strands. Acidic amino acids at the end of Hfq sequences are indicated in white and in dark grey boxes. The alignment clearly indicates that the N-terminal domain is very conserved between Proteobacteria, Firmicutes, Thermotogales and Aquificales. On the contrary, the C-terminal fragments are variable in length and amino acid composition. The position of Helix H1 and of the five b-strands are indicated as H1 and E1-E5.Hfq from Photobacterium profundum, Microbulbifer degradans and Geobacter sulfurreducens are fragments (fr). Firmicutes, Bacillus/clostridium group – Clope, Clostridium perfrin- gens;Thetn,Thermoanaerobacter tengcongensis. Bacillus/staphylococcus group – Bachd, Bacillus halodurans;Stau,Staphylococcus aureus; Bacsu, Bacillus subtilis. Proteobacteria c subdivision – Haein, Haemophilus influenzae;Haeso,Haemophilus somnus;Pasmu,Pasteurella multocida; Phopr, Photobacterium profundum;Vibch,Vibrio cholerae;Xylfa,Xylella fastidiosa;Pseae,Pseudomonas aeruginosa; Pssyr, Pseudo- monas syringae;Micde,Microbulbifer degradans;Yerpe,Yersinia pestis; Yeren, Yersinia enterocolitica;Salty,Salmonella typhimurium;Shifl, Shigella flexneri;Ecoli,Escherichia coli;Erwca,Erwinia carotovora; Xanac, Xanthomonas axonopodis;Xancp,Xanthomonas campestris.Pro- teobacteria d subdivision – Geosu, Geobacter sulfurreducens. Proteobacteria b subdivision – Neima, Neisseria meningitidis;Ralso,Ralstonia solanacearum. Proteobacteria a subdivision – Caucr, Caulobacter crescentus;Bruab,Brucella abortus;Brume,Brucella melitensis;Azoca, Azorhizobium caulinodans;Agrtm,Agrobacterium tumefaciens; Rhilo, Rhizobium loti. Thermotogales – Thema, Thermotoga maritima.Aquifi- cales – Aquae, Aquifex aeolicus. Ó FEBS 2004 The role of the C-terminal domain on Hfq (Eur. J. Biochem. 271) 1261 simultaneously. For the same reason, we never observed a hexameric unfolded (U 6 ) form nor oligomeric forms of Hfq during the dissociation process. The unfolding process can thus be described by N 6 Ð 6U. From these curves, the [urea] 1/2 parameter (midpoint of the transition region) was measured as 2.9 M for Hfq f and 2.4 M for Hfq Nter , indicating that the truncated form of Hfq is less stable than the full-length protein. Based on the analysis of the transition region, the conformational stability of the proteins can be calculated [29]: the free energy when Fig. 2. Image analysis of Hfq oligomers. (A) Average images computed from 3057 projections for Hfq f (left) and 4833 for Hfq Nter with imposed sixfold symmetry. (B) The corresponding level curves are represented. A sharper shape is seen for the full-length Hfq projection compared to the proteolysed; scale bar 2 nm. (C) Significant difference images (a ¼ 0.05) computed from full-length and proteolysed Hfq averages. Absolute value of the difference shows the existence of 12 significant regions arranged into two crowns (top). The external crown corresponds to the positive difference between full-length minus pro- teolysed Hfq (center left). This difference indicates the absence of densities in the most external peaks of the full-length oligomer. Pro- teolysed minus full-length Hfq differences (center right) are significant at the internal region of the oligomer, as demonstrated by the super- position of differences in the average image of proteolysed Hfq (bottom right). Fig. 3. Affinity of Hfq for polyadenylated rpsO mRNA. The labelled RNA fragments were incubated with various concentrations of Hfq. Complexes were separated on native polyacrylamide gels as described in Materials and methods. Intensities were quantified using a Phos- phorImager. Fig. 4. Urea equilibrium unfolding of Hfq. (A) Rate zonal centrifuga- tion was performed as described above. The final concentration of Hfq was 5 l M . Samples were subjected to a 15–55% (w/v) saccharose density gradient and detected with anti-Hfq Igs. Cytochrome c (12.4 kDa), ovalbumin (44 kDa), bovine serum albumin (67 kDa), aldolase (158 kDa) and catalase (232 kDa) were used as standards to calibrate the gradient and are indicated as squares in the gradient. (B) Denaturation of Hfq monitored by the increase of fluorescence emission at 303 nm (excitation at 275 nm). Fluorescence emission is expressed as F app to facilitate comparison. n,Hfq f ; s,Hfq Nter . Fluorescence emission was plotted as a function of urea concentration. From these curves, DG H 2 O u describing the transition and extrapolated in water was calculated as 11.5 kCalÆmol )1 and 9.3 kCalÆmol )1 for Hfq and Hfq Nter . 1262 V. Arluison et al. (Eur. J. Biochem. 271) Ó FEBS 2004 extrapolated in water ðDG H 2 O u Þ is estimated to be 11.5 ± 0.3 kCalÆmol )1 for Hfq f and 9.3 ± 0.25 kCalÆ mol )1 for Hfq Nter . As indicated previously, the His-tag is located at the C-terminus end of the protein. Thus, the truncated protein does not carry the His-tag. However, we do not think that the His-tag is responsible for the stabilization of the protein because the measured stability ðDG H 2 O u Þ of His-tagged and nonHis-tagged Hfq are almost identical. Discussion As no quantitative data describing the affinity of C-terminal truncated Hfq for RNA were available, we measured the equilibrium dissociation constant K d of Hfq f and Hfq Nter for polyadenylated rpsO RNA. The small change observed in K d (50 ± 10 p M for Hfq f vs. 120 ± 15 p M for Hfq Nter ) indicates that the RNA binding function of the protein is nearly exclusively located in the N-terminal domain. Our results agree with previous in vivo studies: indeed, it has been shown that the 82 amino acid Hfq from Pseudomonas aeruginosa (92% sequence identity to the N-terminal domain of Hfq Ec ) and the C-terminally truncated Hfq Ec (lacking the 27 last amino acids) can functionally replace full-length Hfq Ec protein in vivo for phage Qb replication, rpoS and ompA expression [30]. Our results also agree with in vitro studies, particularly with the crystal structure of the 77 amino acid S. aureus Hfq bound to a small oligoribo- nucleotide (AUUUUUG) [19]. This structure indicates that the first 66 amino acids are sufficient to obtain a complex between an oligoribonucleotide and Hfq. However, despite all of these in vitro and in vivo studies, it could not be excluded that the C-terminal region influences Hfq affinity for RNA. This is the first direct biochemical evidence that removal of the C-terminal region does not affect binding of the polyadenylated rpsO RNA. We have shown that the presence of the remainder of the acidic tail results in the thermodynamic stabilization of Hfq by 1.8 kcalÆmol )1 . A comparison of the average electron microscopy images of Hfq Nter and Hfq f (Fig. 2) also indicates that proteolysis causes a structural rearrangement within the protein. The ring of the full-length Hfq is sharper (not larger) than the ring of the truncated Hfq. This effect in the truncated form could be explained by a motion of the segment Ser65–Tyr83 relative to the b5-strand (Ile59– Pro64). Electron microscopy also indicates that a conform- ational change probably occurs at the interface between two consecutive monomers (which involve strands b4ofchainB and b5 of chain A, Fig. 5). As determined by FTIR, this rearrangement is coupled to a reduction in the b-content. On the basis of these observations, we propose that the C-terminal domain (84–102) protects the interface between monomers and thus could contribute to the thermodynamic stabilization of the hexameric Hfq structure. The absence of this domain results in a reduction of the b-sheet character within the Sm domain and could perturb the hydrogen bonding interactions between the interchain strands. Indeed, it should be noted that FTIR analysis, and particularly the Amide I vibration that is associated with the C¼O stretching mode, is a good probe for detecting variation in the pattern of hydrogen bonding interactions. The difference D(DG) value found (1.8 kCalÆmol )1 between full-length and truncated Hfq) represents % 20% of the total free energy of unfolding DG H 2 O u . We emphasize that the truncation of Hfq does not break the interface because we still observe the hexamer in the truncated protein (Fig. 2), but may destabilize the network of hydrogen bonds. The crystal structure of the first 71 amino acids of E. coli Hfq indicates that the C-terminal domain is probably located at the top of the ring, because the last residues (66– 71) form a short tail pointing towards the a-helices [20]. Taking these results and the electron microscopy data (Fig. 2) into account, we can assign this region at the periphery of the ring, to be placed above the Sm-ring, i.e. on the Ôa-helices faceÕ. This positioning of the C-terminal protuberance probably protects the hydrophobic part of the Sm-ring from solvent, particularly at the interfaces between the subunits, and reinforces our hypothesis of a stabilization function for this domain. In addition, it should be noted that the C-terminal domain of Hfq seems to be located at the same position as that of the Sm-like SmAP3 of the archaea P. aerophilum [31] and, as in the case of SmAP3, the possibility of two conformations for the C-terminal domain could not be excluded. The multiple sequence alignment of 23 Hfqs using T-COFFEE [32] presented in Fig. 1 shows that the first 67 amino acids are highly conserved in all species, whereas the length and sequence composition of the C-terminal region varies from one species to another. The difference between the Hfqs results from the insertion of a fragment of variable length between the first 67 amino acids and the end of the acidic C-terminal domain. Determining the exact location of this inserted fragment is not an easy task, as its length is Hfq-dependent. For example, T-COFFEE fails to detect a gap for Bacillus subtilis, Pseudomonas aeruginosa and Brucella melitensis species, because of alignment errors. Using the phylogenetic tree shown in Fig. 6, based on evolutionary distances between 16S rRNA and the multiple Fig. 5. Comparison of the X-ray structure of Hfq Ec (Protein Data Bank entry 1HK9 [20]) with the EM projection. (A) Projection of the volume computed from the X-ray filtered at 18 A ˚ resolution (orange). The cartoon representation of the X-ray structure has been superimposed. Each monomer is represented by a different colour. (B) Grey level curves of the sixfold EM projection. Arrows point to the structural rearrangement within the protein at the monomers interface. The shape of this projection is similar to that published by Zhang et al.[11] with exception of the central region, probably because of a stain penetration difference. This could also explain that the pore size of the level curve representation looks higher than that for the X-ray data. Scale bar 2 nm. Ó FEBS 2004 The role of the C-terminal domain on Hfq (Eur. J. Biochem. 271) 1263 sequence alignment, we find that long Hfqs are found exclusively in c-andb proteobacteria. In contrast, the a proteobacteria, the Firmicutes, the Aquificales and the Thermotogales lack this inserted fragment. This suggests that Hfq from c-andb-proteobacteria evolved from the ancestral sequences by gene fusion. This C-terminal domain may bring new properties to the protein: for E. coli Hfq, greater stability. It is also possible that this extension changes the folding kinetics of Hfq (this is however, beyond the scope of the present study). Like Hfq, the C-terminal extension of eukaryotic Sm proteins was shown to form a protuberance in the Sm core complex [33]. Several roles were attributed to these protuberances in Sm. They can, for example, interact with proteins involved in small nuclear ribonucleoprotein complex (SnRNP) biogenesis, mediate nuclear localization or stabilize interactions with RNA [33–36]. In contrast, our study shows that this extension has rather a structural role in Hfq. In addition, two major differences between the C-terminal extension of Sm and Hfq are observed: the C-terminal domains of Sm proteins are basic and longer than those of Hfq proteins, which contain mostly acidic residues. In addition, this extension is highly structured in some archaeal Sm-like proteins [31], while it is likely disordered in Hfq. However, this C-terminal extension may also have a specialized targeting function (either protein or RNA), in addition to its stability role that we have invoked. Acknowledgements This work was supported by CNRS (UPR 9073 and 9080) and University Denis Diderot-Paris 7. We are indebted to J. P. Le Caer (Ecole polytechnique, Palaiseau, France) and B. Robert (CEA, Saclay, France) for their help in performing mass spectra and FTIR experiments. The BL21(DE3) strain transformed with pTE607 plasmid was kindly provided by T. Elliot. We thank C. Condon for a careful reading of the manuscript. V. A. was supported by a fellowship from University Denis Diderot (Paris 7) and M. F. is recipient of a Ph.D. training grant from MEN. Fig. 6. Phylogenetic tree of the bacteria based on the evolutionary distance of the procaryotic small subunit rRNA phylogenetic. Hfqs har- bouring a long C-terminal end are underlined. Hfq of P. profundum, M. degradans and G. sulfurreducens are fragments and cannot be classified as long or short Hfq (listing available at http://rdp.cme.msu.edu/download/ SSU_rRNA/SSU_Prok.phylo). 1264 V. Arluison et al. (Eur. J. Biochem. 271) Ó FEBS 2004 References 1. Franze de Fernandez, M.T., Hayward, W.S. & August, J.T. (1972) Bacterial proteins required for replication of phage Qb ribonucleic acid. J. Biol. Chem. 247, 824–821. 2. Muffler, A., Fischer, D. & Hengge-Aronis, R. (1996) The RNA- binding protein HF-I, known as a host factor for phage Qb RNA replication, is essential for rpoS translation in E. coli. Genes Dev. 10, 1143–1151. 3. Muffler,A.,Traulsen,D.D.,Fischer,D.,Lange,R.&Hengge- Aronis, R. (1997) The RNA-binding protein HF-1 plays a global regulatory role which is largely, but not exclusively, due to its role in expression of the r S subunit of RNA polymerase in E. coli. J. Bacteriol. 179, 297–300. 4. Tsui,H C.T.,Leung,H C.,E.&Winkler,M.E.(1994)Char- acterization of broadly pleiotropic phenotypes caused by an Hfq insertion mutation in E. coli K-12. Mol. Microbiol. 13, 35–49. 5. Vytvytska, O., Jakobsen, J.S., Balcunate, G., Andersen, J.S., Baccarini, M. & von Gabain, A. (1998) Host-factor I, Hfq, binds to E. coli ompA mRNA in a growth rate-dependent fashion and regulates its stability. Proc. Natl Acad. Sci. USA 95, 14118–14123. 6. Vytvytska, O., Moll, I., Kaberdin, V.R., von Gabain, A. & Bla ¨ si, U. (2000) Hfq (HFI) stimulates ompA mRNA decay by interfering with ribosomes binding. Genes Dev. 14, 1109–1118. 7. Hajnsdorf, E. & Re ´ gnier, P. (2000) Host factor Hfq of E. coli stimulates elongation of poly (A) tails by poly (A) polymerase I. Proc.NatlAcad.Sci.USA97, 1501–1505. 8. Sledjeski, D.D., Whitman, C. & Zhang, A. (2001) Hfq is necessary for regulation by the untranslated RNA DsrA. J. Bacteriol. 183, 1997–2005. 9. Majdalani,N.,Chen,S.,Murrow,J.,StJohn,K.&Gottesman,S. (2001) Regulation of RpoS by a novel small RNA: the char- acterization of RprA. Mol. Microbiol. 39, 1382–1394. 10. Majdalani, N., Hernandez, D. & Gottesman, S. (2002) Regulation and mode of action of the second small RNA activator of RpoS translation, RprA. Mol. Microbiol. 46, 813–826. 11. Zhang,A.,Wassarman,K.M.,Ortega,J.,Steven,A.C.&Storz, G. (2002) The Sm-like Hfq protein increases OxyS RNA inter- action with target mRNAs. Mol. Cell 9, 11–22. 12. Wassarman, K.M., Repoila, F., Rosenow, C., Storz, G. & Got- tesman, S. (2001) Identification of novel small RNAs using com- parative genomics and microarrays. Genes Dev. 15, 1637–1651. 13. Masse, E., Majdalani, N. & Gottesman, S. (2003) Regulatory roles for small RNA in bacteria. Cur. Opinion Microbiol. 6, 120–124. 14. Moll, I., Leitsch, D., Steinhauser, T. & Blasi, U. (2003) RNA chaperone activity of the Sm-like Hfq Protein. EMBO Reports 4, 284–289. 15. Moller, T., Franch, T., Hojrup, P., Keene, D.R., Bachinger, H.P., Brennan, R.G. & Valentin-Hansen, P. (2002) Hfq. A bacterial Sm-like protein that mediates RNA–RNA interaction. Mol. Cell 9, 23–30. 16. Arluison, V., Derreumaux, P., Allemand, F., Folichon, M., Hajnsdorf, E. & Regnier, P. (2002) Structural modelling of the Sm-like protein Hfq from E. coli. J. Mol. Biol. 320, 705–712. 17. Seraphin, B. (1995) Sm and Sm-like proteins belong to a large family: identification of proteins of the U6 as well as the U1, U2, U4 and U5 snRNPs. EMBO J. 14, 2089–2098. 18. He, W. & Parker, R. (2000) Functions of Lsm proteins in mRNA degradation and splicing. Curr. Opinion Cell Biol. 12, 346–350. 19. Schumacher, M.A., Pearson, R.F., Moller, T., Valentin-Hansen, P. & Brennan, R.G. (2002) Structures of the pleiotropic transla- tional regulator Hfq and an Hfq- RNA complex: a bacterial Sm- like protein. EMBO J. 21, 3546–3556. 20. Sauter, C., Basquin, J. & Suck, D. (2003) Sm-like proteins in Eubacteria: the crystal structure of the Hfq protein from E. coli. Nucleic Acids Res. 31, 4091–4098. 21. Bradford, M. (1976) A rapid and sensitive method for the quan- tification of microgram quantities of protein utilizing the principle of protein-dye binding. Anal. Biochem. 72, 248–254. 22. Schagger, H. & von Jagow, G. (1987) Tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis for the separation of proteins in the range from 1 to 100 kDa. Anal. Biochem. 166, 368–379. 23. Marabini,R.,Masegosa,I.M.,SanMartin,C.,Marco,S.,Fer- nandez, J.J., de la Fraga, L.G., Vaquerizo, C. & Carazo, J.M. (1996) Xmipp: An image processing package for electron micro- scopy. J. Struct. Biol. 116, 237–240. 24. Marabini, R. & Carazo, J.M. (1994) Pattern recognition and classification of images of biological macromolecules using artifi- cial neural networks. Biophys. J. 66, 1804–1814. 25. Unser, M., Trus, B.L. & Steven, A.C. (1987) A new resolution criterion based on spectral signal-to-noise ratios. Ultramicroscopy 23, 39–51. 26. Folichon, M., Arluison, V., Pellegrini, O., Huntzinger, E., Reg- nier, P. & Hajnsdorf, E. (2003) The poly (A) binding protein Hfq protects RNA from RNase E and exoribonucleolytic degradation. Nucleic Acids Res. 31, 7302–7310. 27. Zhang, A., Altuvia, S., Tiwari, A., Argaman, L., Hengge-Aronis, R. & Storz, G. (1998) The oxyS regulatory RNA represses rpoS translation and binds the Hfq (HF-1) protein. EMBO J. 17, 6061–6068. 28. Byler, D.M. & Susi, H. (1986) Examination of the secondary structure of proteins by deconvolved FTIR spectra. Biopolymers 25, 469–487. 29. Pace, N.R. (1986) Determination and analysis of urea and guanidine hydrochloride denaturation curves. Methods Enzymol. 131, 266–280. 30. Sonnleitner, E., Moll, I. & Blasi, U. (2002) Functional replacement of the E. coli Hfq gene by the homologue of Pseudomonas aeru- ginosa. Microbiology 148, 883–891. 31. Mura, C., Phillips, M., Kozhukhovsky, A. & Eisenberg, D. (2003) Structure and assembly of an augmented Sm-like archaeal protein 14-mer. Proc. Natl Acad. Sci. USA 100, 4539–4544. 32. Notredame, C., Higgins, D.G. & Heringa, J. (2000) T-Coffee: a novel method for fast and accurate multiple sequence alignment. J. Mol. Biol. 302, 205–217. 33. Bordonne, R. (2000) Functional characterization of nuclear localization signals in yeast Sm proteins. Mol. Cell Biol. 20, 7943– 7954. 34. Mouaikel, J., Verheggen, C., Bertrand, E., Tazi, J. & Bordonne, R. (2002) Hypermethylation of the cap structure of both yeast snRNAs and snoRNAs requires a conserved methyltransferase that is localized to the nucleolus. Mol. Cell 9, 891–901. 35. Brahms, H., Meheus, L., de Brabandere, V., Fischer, U. & Luhrmann, R. (2001) Symmetrical dimethylation of arginine residues in spliceosomal Sm protein B/B’ and the Sm-like pro- tein LSm4, and their interaction with the SMN protein. RNA 7, 1531–1542. 36. Zhang, D., Abovich, N. & Rosbash, M. (2001) A biochemical function for the Sm complex. Mol. Cell 7, 319–329. Ó FEBS 2004 The role of the C-terminal domain on Hfq (Eur. J. Biochem. 271) 1265 . The C-terminal domain of Escherichia coli Hfq increases the stability of the hexamer Ve ´ ronique Arluison 1 , Marc. to the thermodynamic stabilization of the hexameric Hfq structure. The absence of this domain results in a reduction of the b-sheet character within the