The Bacterial Histone-like Protein HU Specifically Recognizes Similar Structures in All Nucleic Acids

HU, a major component of the bacterial nucleoid, shares properties with histones, high mobility group proteins (HMGs), and other eukaryotic proteins. HU, which participates in many major pathways of the bacterial cell, binds without sequence specificity to duplex DNA but recognizes with high affinity DNA repair intermediates. Here we demonstrate that HU binds to double-stranded DNA, double-stranded RNA, and linear DNA-RNA duplexes with a similar low affinity. In contrast to this nonspecific binding to total cellular RNA and to supercoiled DNA, HU specifically recognizes defined structures common to both DNA and RNA. In particular HU binds specifically to nicked or gapped DNA-RNA hybrids and to composite RNA molecules such as DsrA, a small non-coding RNA. HU, which modulates DNA architecture, may play additional key functions in the bacterial machinery via its RNA binding capacity. The simple, straightforward structure of its binding domain with two highly flexible β-ribbon arms and an α-helical platform is an alternative model for the elaborate binding domains of the eukaryotic proteins that display dual DNA- and RNA-specific binding capacities.

The Escherichia coli HU protein is a major component of the bacterial nucleoid (1)(2)(3). This small basic histone-like protein that can introduce negative supercoiling into a close circular DNA molecule in the presence of topoisomerase I is highly conserved and found in all bacterial species (4 -7). HU plays a role in DNA replication, recombination, and repair (8 -10). It participates in Mu transposition (11) and regulation of gene transcription (12). HU has been shown to be important for optimal survival of cells in the stationary phase and under various stress conditions (13).
HU belongs to the family of architectural nuclear proteins that control DNA topology by introducing bends into doublestranded (ds) 1 DNA and stabilize higher-order nucleoprotein complexes. HU resembles eukaryotic proteins of the high mobility group (HMG) class in its DNA binding properties because it binds dsDNA with low affinity and no sequence specificity. In contrast, it displays high affinity for some altered DNA structures such as junctions, nicks, gaps, forks, and overhangs even under stringent salt conditions (14 -18). The DNA structural motif for HU recognition consists of either two dsDNA modules with propensity to be inclined or one dsDNA module adjacent to a ssDNA binding module (19). X-ray crystallography and NMR studies have established the structure of HU dimer in the absence of DNA (20 -22). The two subunits are intertwined to form a compact ␣-helical hydrophobic core with two extended positively charged ␤-ribbon arms. Our recent studies suggest that HU contacts duplex DNA via the minor groove with its flexible arms, whereas the high affinity binding to its specific binding motif requires an additional contact with the HU body (19).
Similar to histones, HU has been shown to bind to poly(U) homopolymer, 2 but the role of this abundant protein in RNA binding was underestimated. Recently we have shown that HU binds with high affinity to mRNA from rpoS, encoding the stress sigma factor of RNA polymerase, and stimulates its translation (23). Interestingly, in parallel to this work it was shown that HBsu, the HU protein of Bacillus subtilis, specifically binds the Alu domain of a small cytoplasmic RNA (scRNA), a homologue of mammalian signal recognition particle RNA (24). In the eukaryotic field, a growing body of evidence shows that a number of proteins including transcriptional factors containing zinc finger or RNA recognition motifs are able to bind specifically to both DNA and RNA (25)(26)(27)(28)(29)(30)(31)(32)(33).
Although HU does not possess any sequence or structural homology to RNA recognition motifs (RRMs) or zinc finger motifs, its small DNA-binding domain formed from two ␤-ribbon arms and an ␣-helical core, is able to bind with high specificity to RpoS mRNA (20 -23). In this study we investigated the general RNA binding features of HU. We measured HU affinity to total cellular RNA and found that HU binds to RNA as strongly as to supercoiled DNA, formerly believed to be the main target for the nucleoid-associated HU (2). The characteristics of HU-nonspecific binding to double-stranded RNA and DNA as well as to simple linear DNA-RNA hybrids are shown to be similar. In contrast, HU binding to discontinuous DNA-RNA structures is much stronger than that to DNA and RNA duplexes and displays the same affinity as with nicked DNA, one of the structures that HU binds specifically. We searched for the structural determinant(s) of HU-RNA recognition using DsrA RNA, a small noncoding stable RNA that modulates the translation of several key transcriptional regulators such RpoS or H-NS (34 -38). We found that DsrA RNA is one of the specific targets for HU binding. Truncation of DsrA RNA showed that HU bound with high affinity an RNA structure similar to that of a DNA overhang, one of its specific targets on DNA (19).

EXPERIMENTAL PROCEDURES
DNA Templates and RNA Synthesis-The DNA fragments corresponding to DsrA RNA and DsrA deletion mutants RNA C, D, and E were amplified by PCR using the pDDS164 plasmid (provided by S. Gottesman, Ref. 34) as a template and appropriate primers. In the case of the smallest RNA-F, two complementary DNA oligonucleotides were annealed. The resulting DNAs were cloned under the control of the T7 promoter into the EcoRI and HindIII sites of pGem3Z (Promega). The [␣-32 P]RNAs were synthesized from a plasmid linearized with HindIII by in vitro transcription. Before use, the RNAs were renatured by incubating at 65°C for 5 min and cooling on ice. For construction of the RNA duplex the DNA synthetic oligonucleotides X1 and Y1 were annealed with their complementary DNA strands and cloned under the control of the T7 promoter into the EcoRI and HindIII sites of pGem3Z. The corresponding ␣-32 P-labeled X and non-labeled Y RNA were prepared from plasmid linearized with HindIII by in vitro transcription. These complementary RNAs were annealed in buffer A (20 mM Tris-HCl, pH 7.5, 100 mM NaCl) by incubation at 80°C for 3 min followed by slow cooling. For the construction of the DNA-RNA hybrid, the synthetic DNA oligonucleotide H was annealed with ␣-32 P-labeled X RNA in buffer A as described above. The following oligonucleotides X1, Y1, H, and X RNA are: X1, AATTCGGGTAGGAGCCACCTTATGAGGAATTCGCCCA; Y1, AATTCCTCATAAGGTGGCTCCTACCCGAATTCGCCCA; H, TGGGC-GAATTCCTC ATAAGGTGGCTCCTACCCGAATTCGCCC; X RNA, GG-GCGAATTCGGGTAGGAGCCACCTTATGAGGAATTCGCCCA.
DNA Construction-Duplex DNA was constructed from oligonucleotide H, and a complimentary oligonucleotide. DNA containing a nick was constructed from oligonucleotides X, C, and D. 3Ј-DNA overhang was constructed from oligonucleotides X and C. DNA-RNA nick and 3Ј-DNA-RNA overhang were constructed as DNA structures, but oligonucleotide C was replaced with the corresponding oligoribonucleotide. Oligonucleotides X and H were 5Ј-labeled and annealed with appropriate oligonucleotides or oligoribonucleotides in buffer A by incubation at 80°C for 3 min followed by slow cooling. The following oligonucleotides X, C, and D are: X, AGTCTAGACTGCAGTTGAGTCCTTGCTAGGACGGAT-CCCT; C, A CTCAACTGCAGTCTAGACT (5Ј-phosphorylated); D, AGGGA-TCCGTCCTAGCAAG G.
Gel Mobility Shift Assay-Binding assays were carried out as described previously (20) in high salt buffer (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 10% glycerol) or in low salt buffer, which had the same ingredients but contained only 10 mM NaCl. Electrophoresis was carried out either in 22.5 or 90 mM Tris borate buffer, pH 8.6, for low or high salt conditions, respectively.
Competition Assay with Non-labeled Nucleic Acids-Total RNA was prepared by hot phenol extraction of E. coli K-12 C600 cells grown in LB medium to exponential phase. E. coli bulk tRNA was a gift from A. Rak. The supercoiled form of the plasmid pNB1 (Biolabs) was purified on an agarose gel and electroeluted. To prepare the linear DNA sample, the same plasmid was linearized with ScaI. HU protein at a final concentration of 13 nM was mixed with 2 fmol of 32 P-labeled nicked DNA, and varying concentrations of non-labeled nucleic acids were added in 15 l of high salt buffer. Samples were analyzed as described previously (23).
Determination of Dissociation Constant-At equilibrium, the dissociation constant of the complex formed by one nucleic acid molecule and one HU dimer is given by Equation  For a DNA that is N bp long, the number of binding sites for the ligand that covers L bp, is (N ϪL ϩ 1) per DNA molecule. The values f and b measured in the experiment are proportional to the concentration of free and bound nucleic acids, respectively, quantified in arbitrary units. The radioactivity of the bands corresponding to f (free) and b (bound) fractions was determined by PhosphorImager analysis of dried gels. Thus, the final equation for calculation of the dissociation constant of the first protein-nucleic acid complex is shown in Equation 2.
Notice that this equation is adequate for nonspecific binding when the affinity of the protein is the same for any L-bp binding site of an N-bp long nucleic acid. For a nucleic acid molecule containing one specific binding site, if nonspecific interaction is negligible, we have Equation 3.
The best fit over several protein concentrations was taken as K d . Equation 1 serves as an evaluation of HU dissociation constant for the first complex. Binding of the second HU dimer might be facilitated by the protein-protein interactions that can be measured by the factor of cooperativity, . The factor is determined as K d1 /K d2 , where K d1 and K d2 are the dissociation constants of the first and second complexes, respectively. For binding of HU to double-stranded DNA, RNA, and the DNA-RNA hybrid the has been determined as described previously (15,18).
The apparent dissociation constant of HU to non-labeled nucleic acids was calculated as in Equation 4, where [A free ] is the concentration of free non-labeled nucleic acid. Assuming that binding is nonspecific and that the nucleic acid is much longer than a protein binding site, The dissociation constant of HU with nicked DNA is K N ϭ [P free ] ϫ f/b, where f and b are proportional to the concentrations of free and bound nicked DNA, respectively, and the radioactivity of the bands corresponding to f (free) and b (bound) fractions was determined by PhosphorImager analysis of dried gels. Thus, [P free ] ϭ K N ϫ b/f, and we have Equations 5-7, where [P free ] ϭ K N ϫ b/f or as in Equation 7.
These equations were obtained assuming that the number of binding sites is equal to the length of the nucleic acid. This is the case when DNA is far from saturated with the protein. The saturation of the DNA can be evaluated according to the McGhee-von Hippel equation (39). We estimated that under the experimental conditions used, DNA is far from saturated. Indeed, in our experiments the amount of the competitor added is 100 -1000 bp per one HU dimer.

RESULTS
The Binding of HU to Supercoiled and Linear DNA Is Similar to Its Binding to Total RNA and tRNA-HU is one of the most abundant DNA-binding proteins in the bacterial cells. In contrast to its low affinity salt-sensitive binding to duplex linear DNA, HU binds with high affinity to DNA damage and repair intermediates under high salt conditions (19). We also have found recently that HU specifically recognizes the mRNA from rpoS and stimulates its translation (23). HU affinity to RpoS mRNA fragment is as high as that to nicked DNA, which is 1000-fold higher than that for double-stranded DNA. Because it appears that HU is able to bind DNA and RNA, both nucleic acids may serve as HU binding targets in the cell. This finding prompted us to further investigate the RNA binding properties of HU.
We first measured the affinity of HU to the major DNA and RNA species present in the bacterial cell. HU binding to supercoiled and linear DNA as well as to total cellular RNA and tRNA was compared. Gel mobility shift assays can not be applied directly to separate free and bound high molecular weight nucleic acids because plasmid DNA is too bulky to be gel-shifted by HU, and total RNA is too disperse in size to form sharp bands in the gel. To tackle the problem we applied an assay on which one nucleic acid of interest competes with labeled 40-bp nicked DNA for HU binding. HU forms a single complex with the nicked DNA with an apparent dissociation constant (K d ) of 10 nM under stringent salt conditions (16 -18). HU was mixed with labeled nicked DNA, and the complex was challenged with increasing amounts of the non-labeled nucleic acid of interest. The progressive decrease in HU-nicked DNA complex upon increase in the concentration of non-labeled DNA or RNA reflects a decrease in the concentration of free HU in the solution as a result of binding to the non-labeled nucleic acid competitor. The comparison of the ratio of complexed and free nicked oligonucleotides in the absence or presence of the competitor provides the concentration of HU bound to the nonlabeled nucleic acid. Thus, the dissociation constant of the non-labeled nucleic acid can be calculated on the basis of the known parameters of the HU-nicked DNA interaction as explained under "Experimental Procedures." As a control of this experimental approach, the same nicked DNA was used as a competitor of the labeled nicked DNA (Fig.  1A). The apparent dissociation constant of HU to nicked DNA found in this experiment is 12 nM, a value very close to the K d of 10 nM derived from the classical protein titration experiment (16 -18). Using this methodology, we successively measured the dissociation constants for linear and supercoiled DNAs, tRNA, and total RNA (Fig. 1A). Based on the competition experiment, the apparent dissociation constant of HU to supercoiled DNA was estimated to be 450 nM under stringent conditions (200 mM NaCl, Fig. 1A). Supercoiled plasmid was used as an example of genomic DNA. The plasmid was used after linearization to estimate the affinity of HU for linear DNA, and a value of K d ϭ 1300 nM was obtained under the same high salt conditions. This apparent dissociation constant accounts for the cooperativity of HU binding and dissociation/association of HU monomers. Thus, HU protein binds supercoiled DNA 3ϫ more strongly than linear DNA of the same length. This is in agreement with previous studies with HU chemical nuclease, which cleaves supercoiled DNA 2.5ϫ faster than relaxed DNA (40). To investigate whether RNA may serve as a potential intracellular HU target, we isolated total bacterial RNA and estimated the affinity of HU using the same technique (Fig. 1A). The apparent dissociation constant of 2500 nM was calculated as described under "Experimental Procedures" (Fig. 1). Likewise, the affinity of HU affinity to bulk E. coli tRNA was estimated to be 2200 nM. Thus, HU binds the major species of nucleic acids present in bacterial cells with a similar affinity.
Finally the graph of bound nicked DNA versus the concentration of the different competitors (Fig. 1B) shows that the concentration of the nucleic acids species required to reduce the bound nicked DNA corresponds well with the K d (Fig. 1B).
HU Binding to Linear dsDNA, dsRNA, and DNA-RNA Hybrids-Double-stranded RNA and DNA as well as a DNA-RNA hybrid of 40 bp (all of the same sequence) were constructed. Their binding to HU was studied using the classical gel mobility shift assay. Fig. 2 shows first that the gel mobilities of the duplexes of the three different types of nucleic acids (dsDNA, dsRNA, and DNA-RNA) are rather different, probably reflecting the different conformation of the three nucleic acids. These conformations seem to be sensitive to the salt conditions. At high salt the hybrid DNA-RNA migrates the fastest, whereas in low salt dsDNA is faster. Under both sets of conditions, duplex RNA is the most retarded matrix. In terms of binding to HU, under high salt conditions (200 mM NaCl) no complex was detectable with any duplex studied ( Fig. 2A). This was expected for dsDNA because under high salt conditions only a smear is visible because of the dissociation of nonspecific HU linear DNA complexes during their migration in the polyacrylamide gel (19). In contrast, a defined complex could be isolated under low salt conditions with dsDNA having an apparent dissociation constant (K d ) of 450 nM for the first complex and a cooperativity of ϭ K d1 /K d2 ϭ 30. One HU dimer occupies 9 bp so that four HU complexes are formed with the 40-bp dsDNA ( 2B) (41). Fig. 2B also shows that four HU dimers can be accommodated similarly by 40-bp dsRNA and by the hybrid DNA-RNA. Thus, each HU dimer would cover 9 -10 bp of dsRNA in a manner similar to the HU binding on dsDNA. The apparent dissociation constant of the first HU-dsRNA complex is 1800 nM. It is 4-fold higher than that for dsDNA, but the cooperativity of binding is 2.5ϫ higher, ϭ 70. Dissociation constants for the second complexes K d2 for dsDNA and RNA are 15 and 25 nM, respectively. The opposing differences in the dissociation constants for the first complexes and cooperativity suggest in fact that HU binds 1.6-fold stronger dsDNA than dsRNA. This result is in agreement with the ratio of affinities obtained in our competition studies performed under stringent salt conditions, which indicated that HU binds linear DNA twice as strongly as total RNA (Fig. 1). A 40-bp DNA-RNA hybrid of the same sequence was also checked for HU binding. Formation of four complexes with the 40-bp hybrid (and three complexes with 30-bp hybrid, data not shown) corresponds also to the binding model of one HU dimer to every 9 -10 bp. The apparent dissociation constant of the first complex K d ϭ 450 nM is the same as for dsDNA, the cooperativity being slightly higher, ϭ 40.
Thus, HU is able to bind dsDNA, dsRNA, and DNA-RNA hybrids with very similar affinities. The conformations of B-DNA and A-RNA are known to differ significantly, while DNA-RNA hybrids were shown to assume an intermediate A/B conformation (42). Nevertheless, HU seemed to recognize both the A and B conformations of RNA and DNA duplexes. This wide spectrum of binding can be explained by the high flexibility of HU arms (21).
Specific Interaction of HU with DNA-RNA Hybrid Structures-HU does not recognize any particular DNA sequence, but it binds (under stringent conditions) with high affinity to some altered DNA structures. The simplest DNA structure that HU binds strongly even in the presence of excess of dsDNA competitor is either a DNA containing a singlestranded break (nicked DNA) or a DNA 3Ј-overhang (19). The specificity of HU DNA binding was explained by the simultaneous interaction of the HU arms with the 5Ј doublestranded part of the molecule and the interaction of HU body with the flexible 3Ј-branch, which can be either a double-or single-stranded DNA (19). Because we have seen that under low salt conditions HU binds an RNA-DNA hybrid with a similar affinity to which it binds DNA duplex (Fig. 2), we asked whether HU also could specifically recognize both nick and 3Ј-overhang structures in which one of the DNA strands is replaced with RNA. Both structures are of particular interest because they are DNA replication intermediates (43). In addition, the DNA-RNA 3Ј-overhang is involved in the priming repair of the double-stranded breaks and priA replication (44). Fig. 3 shows that HU binds these structures under stringent conditions and forms a single complex with each. The apparent dissociation constants (K d ) are 10 and 16 nM for nicked RNA and 3Ј-overhang RNA, respectively, identical to the values found for nicked DNA and DNA 3Ј-overhang and 100ϫ stronger than that for double-stranded nucleic acids. Thus, this binding is structure-specific.
Specific Binding Motif of HU to RNA-We have demonstrated recently that HU is able to bind RpoS mRNA with an affinity similar to that with which it binds nicked DNA, one of its specific substrates (23). The folding of this RNA is not yet characterized in detail. Therefore, another RNA target with a simple and well characterized secondary structure was necessary to better understand the HU-RNA interaction. We investigated whether HU could bind to the regulatory DsrA RNA of E. coli, which has both positive and negative action on global transcriptional regulators such as H-NS or RpoS, the sigma factor controlling the nutrient starvation and stress responses (34,35). DsrA was cloned under the control of T7 promoter, in vitro transcribed, and used for gel mobility shift assay with HU protein. The structure of this small RNA, which consists of three stem loops (34,38), is presented in Fig. 4A. The configuration shown is derived from data based on ds-or ss-specific nuclease susceptibility and phylogenetic data (38). Fig. 4B shows that HU binds this DsrA RNA and forms three specific complexes under high salt conditions with an apparent dissociation constant of 20 nM for the first complex. To find the structural motifs recognized by HU within this small RNA, we truncated DsrA RNA from its 3Ј-end to two stem loops connected by a single-stranded RNA of 12 nucleotides. HU formed two complexes with this RNA, RNA-C, containing stem loops 1 and 2 (Fig. 4C). Further truncation resulted in the structure RNA-D, which contained only one stem loop connected to a ssRNA of 12 nucleotides. This structure was still recognized by HU, forming one complex (Fig. 4D). As indicated at the bottom of Fig. 4, RNA-C can be considered the superposition of two putative HU-binding motifs, one RNA 3Ј-overhang and one 5Ј-overhang, whereas RNA-D has only the RNA 3Ј-overhang. We have shown recently that the corresponding DNA structure, the DNA 3Ј-overhang, is HU-specific in contrast to the DNA 5Ј-overhang, a much less powerful target for HU (19). The fact that the RNA 3Ј-overhang (RNA-D) as well as the DNA-RNA 3Ј-overhang (Fig. 3) were recognized with a similar high affinity to the DNA 3Ј-overhang (Fig. 3) indicates that HU recognizes the same motif in both RNA and DNA molecules. To investigate the possibility that HU also specifically binds to RNA 5Ј-overhangs, RNA-E, an RNA 5Ј-overhang, was constructed by truncation of RNA-C from its 5Ј-end. Surprisingly, HU binds this structure with the formation of one complex as well as it binds the RNA 3Ј-overhang (Fig. 4E). Finally, truncation of RNA-D from its 3Ј-end that leaves only the first stem loop (RNA-F) abolishes HU binding. Thus, there is one difference between the binding of HU to DNA and RNA. In contrast to binding to DNA, HU recognizes RNA 3Ј-as well as 5Ј-overhangs.
Because HU has a high affinity for RNA 5Ј-overhangs in contrast to the corresponding DNA structure, we reconsidered the DNA-RNA hybrid analysis shown in Fig. 3 and compared the binding of HU to a DNA-RNA 5Ј-overhang to that of the DNA 5Ј-overhang. Using this approach, we confirmed that HU binds poorly to the DNA 5Ј-overhang (19) but binds strongly to the DNA-RNA 5Ј-overhang (Fig. 5). Under high salt conditions HU forms one complex with the DNA-RNA 5Ј-overhang with an apparent dissociation constant (K d ) of 16 nM. It is interesting to note that this DNA 5Ј-overhang structure is involved in DNA replication, as is its DNA-RNA 3Ј-overhang counterpart.

DISCUSSION
We report here that the DNA-binding protein HU is also an RNA-binding protein. We found that under physiological conditions HU affinity to total bacterial RNA is only 4.5-fold lower than that for supercoiled DNA, which is representative of the bacterial chromosome DNA. Based on these values and those for the DNA and RNA content of the bacterial cell, we suggest that the distribution of HU between DNA and RNA should be roughly 2:1. In addition, using gel shift retardation assays that directly measure the affinity constants, we could show that the characteristics and patterns of HU binding to short DNA or RNA duplexes as well as to DNA-RNA hybrids are very similar, with one dimer packed every 9 -11 bp. This indicates that HU is able to adopt both the A and B conformations of nucleic acids, probably because of the high flexibility of its arms, which have been shown to contact the minor groove of DNA (19,21). However, a first difference is observed in the binding of HU to dsDNA and dsRNA. The cooperativity value for the RNA is higher than was found for HU binding to dsDNA (Fig. 2B). This probably reflects a more advantageous orientation of tandemly bound HU molecules along the A-helix for dimer-dimer interactions. It is also noteworthy that HU exhibited in its binding to DNA-RNA hybrids more resemblance to dsDNA than was seen for dsRNA, even though the DNA-RNA heteroduplex assumed the intermediate A/B conformation (42). Interestingly, the Z␣ domain of dsRNA adenosine deaminase (ADAR1) was demonstrated to bind left-handed Z-RNA specifically, as well as Z-DNA, and to stabilize the Z-conformation (45). The helixturn-helix motif was used by Z␣ for conformation-specific interaction with Z-DNA (46). Z␣ probably is able to bind to both nucleic acids, since Z-DNA and Z-RNA adopt similar Z-conformations. It remains to be elucidated whether HU with its rather different DNA-binding motif could possess a Z-conformation binding activity.
HU is well characterized as a protein that binds to DNA with low affinity and without sequence specificity but that recognizes with high specificity, on the other hand, DNA structures such as nicked DNA or 3Ј-overhangs. Because HU binds DNA-RNA hybrids as strongly as it binds dsDNA it was of interest to see if it also interacts with nicked and 3Ј-overhang DNA-RNA hybrids. For this, we compared the binding of HU to short DNA and RNA duplexes and to DNA-RNA hybrids. Effectively, HU does bind such DNA-RNA structures under stringent conditions with an affinity similar to that for related DNA substrates. Interestingly these hybrid structures, the nicked RNA and the overhang RNA, appear as intermediates of DNA replication via the Okazaki fragments in which the short RNA primers hybridize to the lagging DNA strand. Furthermore, the DNA-RNA 3Ј-overhang is also involved in priA replication and in priming repair of DNA ds breaks (44). Because the HU concentration in E. coli was estimated to be ϳ30,000 copies/ cell, which corresponds roughly 30 -60 M, the concentration of free HU therefore is well above its dissociation constants (K d ) for these structures, which are in the low nanomolar range. This suggests that specific binding could occur in vivo. This finding may explain the role of HU in DNA replication and repair (8,9).
Although HU is the only bacterial protein that clearly exhibits a dual specific DNA and RNA binding activity, such proteins have been characterized in eukaryotes. Whereas most of these proteins possess separate DNA-and RNA-binding domains, some of them contain multifunctional domains. Although the RRMs are found commonly in RNA and single-strand-binding proteins, at least two transcription factors use an RRM for preferential binding to double-stranded DNA (31)(32)(33), and the RRM of murine IPEB protein recognizes both dsDNA and pre-mRNA (33). The Cys 2 -His 2 zinc fingers represent the canonical DNA-binding motif, which is able to interact specifically with RNA. The zinc finger domains of TFIIIA, PEP, MOK2, and WT-1 proteins mediate specific binding to both RNA and dsDNA (25)(26)(27)(28)(29). One of the best studied is the TFIIIA transcription factor, which contains zinc finger modules specialized for either DNA or RNA recognition. Some of these modules provide the majority of the protein binding affinity for DNA, whereas the others form specific complexes with their RNA targets (47)(48)(49). Thus, zinc fingers differ somewhat in ability to bind tightly to DNA and RNA.
The HU protein provides an extension to this bifunctional DNA and RNA binding module. The structure of HU consists solely of the concave, positively charged surface made up of ␤-ribbon arms and an ␣-helical hydrophobic core (20 -22). Hence, HU RNA-and DNA-binding domains should be equivalent. HU, thus, should bind dsRNA in a way similar to how it binds dsDNA, namely through its ␤-arms. Although DNA-and RNA-binding domains are generally distinct, ribosomal protein S7, a primary 16S rRNA-binding protein, possesses striking structural similarity with the HU family of DNA-binding proteins (50,51). Both have a very similar ␤-hairpin architecture grasping the double-stranded nucleic acid and a pair of S7 ␣-helices that superimpose quite well onto those of the HU monomer. The structure of the of S7 ␤-hairpin bound to 16S rRNA via the minor groove (52) allows us to suggest the possibility that the HU arms contact dsRNA in a similar way. It also has been shown that TFIID, the TATA-box-binding protein, shares homologies with integration host factor and HU (53). The highly symmetric structure of TFIID contains a DNA binding module resembling a molecular saddle that sits astride the DNA (54), very similar to the saddle formed by the flexible arms and the platform of the dimeric HU (20 -22). It will be interesting to see whether TFIID could also, as HU, interact with RNA or DNA-RNA hybrids.
We have demonstrated here that HU binds strongly to DsrA RNA, the secondary structure of which has been characterized as three stem loops (34,38). To localize the structural determinants of HU binding, we analyzed HU interaction with truncated DsrA RNA fragments. Two HU dimers could be bound onto two stem loops connected together by a 12-nt ssRNA linker, and one HU dimer was bound to one stem loop with 12 nucleotides ssRNA from the 3Ј-end. These RNAs can be considered as the superposition of two RNA overhang motifs and one single RNA 3Ј-overhang, respectively. We have shown that relative DNA structure, that is, a DNA 3Ј-overhang, is HUspecific. It seems plausible that HU recognizes the same motif in RNA molecules. We proposed recently a model for specific binding of HU to DNA 3Ј-overhangs (19). The specificity of binding was explained by the simultaneous interaction of HU flexible arms with a 3Ј-dsDNA branch in the region of the discontinuous point and of HU body with the second ssDNA branch. Our finding that HU binds double-stranded DNA and RNA with the similar affinity as well as the capability of HU to bind analogous RNA structures allows us to propose a similar model for the interaction of HU with RNA 3Ј-overhangs. The interaction of HU arms with the double-stranded part of RNA and the interaction of HU body with the single-stranded part of the molecule renders the binding specific and high-salt resistant. Interaction of HU arms with the DNA minor groove prevents HU binding to DNA 5Ј-overhangs because the 5Ј-ss branch must cross the negatively charged double helix to reach the HU body (19). Because HU is capable of forming complexes with both RNA 3Ј-and 5Ј-overhangs, we conclude that HU is able to interact not only with the minor groove (as it does for DNA) but also with the narrowed RNA major groove. Binding of both minor and major RNA grooves is an unusual feature among nucleic acid-binding proteins. We demonstrated that HU binds to DNA-RNA 5Ј-overhangs. This specific interaction is possible because of the intermediate A/B conformation of DNA-RNA hybrid of the ds part of the structure, which is closer to the A-form RNA helix.
HBsu protein, the HU in B. subtilis, has been identified as a component of the signal recognition particle-like particle that interacts specifically with the Alu domain of scRNA (24). The Alu domain of scRNA is composed of two stem loops followed by an extended double-stranded region. It appears likely that the three-way junction is an element required for HBsu binding to the Alu domain. We have shown that HU has a high affinity to the corresponding DNA structure, namely the three-way DNAjunction (19). This is further evidence that HU applies the same strategy for structure-specific recognition with its RNA and DNA targets and that the universal DNA/RNA motif for HU binding consists of two double-stranded or double-and single-stranded modules apt to be inclined. RNA adopts many complex tertiary structures, and it is conceivable that HU also recognizes this rather widespread structural motif within a great number of RNAs. One example of such an RNA is DsrA, a member of the sRNA (small RNA) family, considered as mediators of cellular processes in bacteria. DsrA was first characterized as an untranslated small RNA that activates transcription of RcsA when overproduced by counteracting H-NS silencing (34). RcsA is a positive transcription regulator of cps genes necessary for capsid synthesis. It is interesting to recall that overproduction of HU also stimulates RcsA synthesis (55). In addition, HU-like DsrA stimulates RpoS translation (23,(35)(36)(37)(38). By binding strongly to the mRNA-encoding RpoS (23) and to DsrA (this study), HU may regulate major programs in the bacterial cell including the starvation phase, stress response, and bacterial envelope synthesis.