The EMBO Journal vol.10 no.3 pp.687-696, 1991
HU and IHF, two homologous histone-like proteins of Escherichia coil, form different protein - DNA complexes with short DNA fragments
Eliette Bonnefoy and Josette Rouviere-Yaniv Laboratoire de Physiologie Bacterienne. Institut de Biologie PhysicoChimique, 13 rue Pierre et Marie Curie, 75005 Paris, France Communicated by P.Chambon
Using the gel retardation technique we have studied the protein-DNA complexes formed between HU-the major histone-like protein of Escherichia coli -and short DNA fragments. We show that several HU heterodimers bind DNA in a regularly spaced fashion with each heterodimer occupying about 9 base pairs. The a2 and (2 HU homodimers form the same structure as the ac heterodimer on double stranded DNA. However when compared to the heterodimer, they bind single stranded DNA with higher affinity. We also show that HU and the Integration Host Factor of E.coli (UEF) form different structures with the same DNA fragments. Moreover, HU seems to enhance the DNA-binding capacity of IHF to a DNA fragment which does not contain its consensus sequence. Key words: curved DNA/histone-like proteins/HU/IHF/ protein -DNA interactions
Introduction The physical properties of DNA allow for the existence of curvature, resulting in loops, hair-pins and the DNA
wrapping around a protein core. Such local structures play important role in modulating the expression of the information contained in the sequence of DNA (Ptashne, 1986; Huo et al., 1988; Richet et al., 1986). Specific DNA conformations are either induced or stabilized by a wide variety of DNA-binding proteins. These proteins either recognize a specific base sequence and by binding to it induce a new conformation like bending (Wu and Crothers, 1984; Robertson and Nash, 1988) or they bind to DNA without recognizing a particular sequence stabilizing, at least transiently, nucleoprotein complexes (Raibaud, 1989; Bramhill and Komberg, 1988). HU is the most abundant DNA-binding protein of Escherichia coli with around 30 000 dimers per cell (for review see Drlica and Rouviere-Yaniv, 1987). It is classified as a histone-like protein since it shares some of the characteristics of eukaryotic histones: small size, basic pl, an amino acid composition rich in lysine and alanine residues and a capacity to introduce in vitro negative supercoils into relaxed circular DNA in the presence of topoisomerase I, probably through the formation of nucleosome-like structures (Rouviere-Yaniv et al., 1979; Broyles and Pettijohn, 1986). an
HU exists in the cell as a dimer and at least 90% of dimers are heterodimers (Rouviere-Yaniv and Kjeldgaard, 1979) composed of two subunits, a subunit (HU2) and subunit Oxford University Press
(HU 1). These subunits are highly homologous with 70 % of their amino acids identical. The synthesis of these two subunits appears to be regulated to maintain a balance between the respective amounts of a and ,B subunits in the cell (Rouviere-Yaniv et al., 1990). Each subunit is encoded by a separate gene, the hupA gene located at 90 min of the bacterial map codes for the ca subunit (Kano et al., 1987) and the hupB gene located at 10 min codes for the a subunit (Kano et al., 1986). In vitro as well as in vivo studies have shown that HU plays an important role in several processes in the cell: replication from oriC (Dixon and Kornberg, 1984; Ogawa et al., 1989), transposition of TnJO (Morisato and Kleckner, 1987) as well as of bacteriophage Mu (Craigie et al., 1985; Huisman et al., 1989) and DNA inversion reaction in Salmonella typhimurium (Johnson et al., 1986). There is no consensus sequence for the binding site of HU to DNA, just as there is no consensus sequence for eukaryotic histones. Even though we do not know how HU participates in the processes described above, it was proposed that HU may stabilize the DNA in a particular conformation. It has been suggested that HU could be involved in the coiling of DNA during the initiation of replication at oriC (Bramhill and Kornberg, 1988) or in the stabilization of DNA loops during the DNA inversion reaction associated with flagellar phase variation in Salmonella (Johnson et al., 1986). IHF has also been classified as a histone-like protein mainly because it shares amino acid sequences homology with HU. It is also a small, basic, dimeric protein composed of two different subunits with 30% of their amino acids identical: the ca subunit encoded by the himA gene at 38 min in the E. coli chromosome (Miller, 1981; Mechulam et al., 1985) and the 1 subunit encoded by the hip or himD gene at 25 min (Flamm and Weisberg, 1985). This protein although 10-fold less abundant than HU (Bonnefoy et al., 1989), is essential for the integration of bacteriophage lambda (Nash, 1981) and plays a role in several molecular events in the cell (reviewed by Friedman, 1988). Even though, as opposed to HU, a consensus sequence for the binding of IHF to DNA has been determined (Craig and Nash, 1984), it has been proposed by these authors that like HU, IHF stabilizes DNA segments in particular conformations. The extensive sequence homology between IHF and HU and especially in the putative DNA-binding domain of HU, led Yang and Nash to suggest that both proteins bind DNA in a very similar way (Yang and Nash, 1989). Both proteins would encircle DNA with a pair of, sheet arms (Tanaka et al., 1984) which occupy the minor groove. This is different from most regulatory DNA-binding proteins that occupy the major groove. More recently a model has been proposed in which HU and IHF not only bind DNA in the same way, but that they also bend DNA after binding to it (White et al., 1989). It was of interest to examine if HU and IHF interact with
687
E.Bonnefoy and J.Rouvi&re-Yaniv.
DNA similarly and so substantiate the idea of identical binding mechanisms for the two proteins. Another biological reason to compare the non-specific binding of these two proteins to the DNA was the fact that neither HU nor IHF are essential (Wada et al., 1988 and Huisman et al., 1989). Nevertheless we have shown that E. coli strains lacking HU acquire compensatory mutations (Huisman et al., 1989), IHF is evidently a possible candidate. We therefore set up, using the gel retardation technique, a system which allows us to compare several parameters of the binding of these two proteins to DNA. We used short DNA fragments in order to avoid the topological contraints imposed by long, circular or superhelical DNA molecules. The DNA fragments used here are short synthetic fragments predicted to be curved or uncurved on the basis of their nucleotide sequence. We show in this work that HU and IHF interact with DNA differently. Several HU dimers are able to bind to a linear double stranded DNA fragment side by side with a binding length of around 9 bp. Whereas four HU dimers bind to a 42 bp DNA segment, only one IHF dimer binds to the same fragment under similar conditions. We also show that HU does not exhibit strong preferential binding to curved DNA compared to uncurved DNA. On the other hand IHF binds curved DNA much more tightly than uncurved DNA. Another difference described in this work is the ability of HU, but not IHF, to bind single stranded DNA. Moreover we show that the HU a2 or 02 homodimers bind single stranded DNA with higher affinity than the heterodimer (HU o43). Finally, we show that HU, ofi as well as HU 12 and to a lower extent a2, enhances the binding of IHF to a DNA fragment that does not carry the IHF DNA-binding consensus sequence.
Results In order to study and compare the interactions of HU and IHF with DNA we have used the gel-retardation technique which permits the separation of DNA-protein complexes from free DNA on polyacrylamide gels (Garner and Revzin, 1981). The protein bound DNA is retarded in its electrophoretic migration as compared to free DNA. The degree of retardation depends not only upon the molecular weight of the bound protein but also on the conformation acquired by the DNA molecule in the complex; for example it was shown that complexes with bent DNA are retarded more than complexes with unbent DNA (Wu and Crothers, 1984). The DNA fragments used in this work (see Table I) are double stranded or single stranded synthetic oligonucleotides of different sizes. The sizes of the fragments used in this work vary between two and four double helical turns of the B form.
Several HU a,8 dimers bind to a linear DNA fragment side by side with a binding length of - 9 bp Purified HU a43 was mixed with double stranded radioactively labeled oligonucleotides of different sizes as detailed in Materials and methods. The mixture of free and protein bound DNA was analysed on a polyacrylamide gel. After migration the gel was dried and autoradiographed. Figure 1 shows that with saturating amounts of HU two retarded HU -DNA complexes are formed with a 21 bp (NC2 1) as
688
well as with a 26 bp oligonucleotides (NC26); three HU -DNA complexes are formed with a 29 bp oligonucleotide (NC29) and four complexes are formed with a 42 bp oligonucleotide (NC42). We observed that the more retarded complexes gradually appear as the concentration of HU is raised and as they increase in intensity the first less retarded complex gradually disappear. Since the formation of the different protein-DNA complexes is correlated with the concentration of HU used, we believe that the first less retarded complex corresponds to the binding of one HU co4 dimer, the second complex to the binding of two HU a13 dimers, the third complex to the binding of three HU ao1 dimers and the fourth most retarded complex to the binding of four HU caf dimers. Similar results have recently been described for the binding to DNA of the bacteriophage SPO1-encoded TFl protein which is another member of the HU family (Schneider and Geiduscheck, 1990). Several conclusions can be drawn from these experiments: Firstly, it appears from these results that several HU af3 dimers can orderly bind, side by side, to a linear DNA fragment as predicted by the model of Tanaka (Tanaka et al., 1984). The correlation of the number of HU -DNA complexes obtained with a given double stranded DNA fragment with the number of base pairs present in this fragment suggests a binding length of 9 bases per bound HU af3 dimer. Secondly, the binding side by side of several HU a: dimers to a linear double stranded DNA fragment does not appear to be cooperative. Figure 2 shows the scanning of a gel representing the formation of the four HU-NC42 complexes. We plotted the intensity of each complex (Cl, C2, C3 and C4) against the amount of HU protein used and we calculated the slope (A) of each curve which indicates the rate of formation of each complex. As represented in Figure 2, the rate of formation of C2 is slower than the rate of formation of C1, the rate of formation of C3 is slower than the rate of formation of C2 and the rate of formation of C4 is slower than the rate of formation than C3 (ACl > AC2 > AC3 > AC4). This indicates a non-cooperative phenomenon in which the binding of one HU dimer does not enhance the binding of the next HU dimers. Thirdly, the shorter the DNA fragment, the greater the retardation upon binding of a second HU ao4 dimer (see Figure 1 lanes f,h and j). Since all the complexes with this 2: 1 stoichiometry have roughly the same molecular weight, this phenomenon could be a consequence of the DNA adopting a bent conformation inside the complex and this conformation being more pronounced for the shorter DNA fragments. Bending of DNA as a consequence of HU binding has already been proposed (White et al., 1989) and recently demonstrated (Hodges-Garcia et al., 1989). The particular effect we describe here is probably reflecting different possible arrangements of two HU dimers according to the length of the DNA fragment they are bound to. Nevertheless since all the fragments do not have the same base sequences, we cannot completely rule out the possibility that the above described phenomenon is the result of HU inducing different conformations on the different DNA fragments in a sequence dependent manner. This however, seems to us quite unlikely since (i) such an effect is not observe after the binding of the first HU dimer and (ii) HU DNA binding is not sequence -
dependent.
l_h_.
1. SCL1LC.
HU and IHF, two homologous histone-like proteins of Escherichia coil, form different protein - DNA complexes with short DNA fragments
Eliette Bonnefoy and Josette Rouviere-Yaniv Laboratoire de Physiologie Bacterienne. Institut de Biologie PhysicoChimique, 13 rue Pierre et Marie Curie, 75005 Paris, France Communicated by P.Chambon
Using the gel retardation technique we have studied the protein-DNA complexes formed between HU-the major histone-like protein of Escherichia coli -and short DNA fragments. We show that several HU heterodimers bind DNA in a regularly spaced fashion with each heterodimer occupying about 9 base pairs. The a2 and (2 HU homodimers form the same structure as the ac heterodimer on double stranded DNA. However when compared to the heterodimer, they bind single stranded DNA with higher affinity. We also show that HU and the Integration Host Factor of E.coli (UEF) form different structures with the same DNA fragments. Moreover, HU seems to enhance the DNA-binding capacity of IHF to a DNA fragment which does not contain its consensus sequence. Key words: curved DNA/histone-like proteins/HU/IHF/ protein -DNA interactions
Introduction The physical properties of DNA allow for the existence of curvature, resulting in loops, hair-pins and the DNA
wrapping around a protein core. Such local structures play important role in modulating the expression of the information contained in the sequence of DNA (Ptashne, 1986; Huo et al., 1988; Richet et al., 1986). Specific DNA conformations are either induced or stabilized by a wide variety of DNA-binding proteins. These proteins either recognize a specific base sequence and by binding to it induce a new conformation like bending (Wu and Crothers, 1984; Robertson and Nash, 1988) or they bind to DNA without recognizing a particular sequence stabilizing, at least transiently, nucleoprotein complexes (Raibaud, 1989; Bramhill and Komberg, 1988). HU is the most abundant DNA-binding protein of Escherichia coli with around 30 000 dimers per cell (for review see Drlica and Rouviere-Yaniv, 1987). It is classified as a histone-like protein since it shares some of the characteristics of eukaryotic histones: small size, basic pl, an amino acid composition rich in lysine and alanine residues and a capacity to introduce in vitro negative supercoils into relaxed circular DNA in the presence of topoisomerase I, probably through the formation of nucleosome-like structures (Rouviere-Yaniv et al., 1979; Broyles and Pettijohn, 1986). an
HU exists in the cell as a dimer and at least 90% of dimers are heterodimers (Rouviere-Yaniv and Kjeldgaard, 1979) composed of two subunits, a subunit (HU2) and subunit Oxford University Press
(HU 1). These subunits are highly homologous with 70 % of their amino acids identical. The synthesis of these two subunits appears to be regulated to maintain a balance between the respective amounts of a and ,B subunits in the cell (Rouviere-Yaniv et al., 1990). Each subunit is encoded by a separate gene, the hupA gene located at 90 min of the bacterial map codes for the ca subunit (Kano et al., 1987) and the hupB gene located at 10 min codes for the a subunit (Kano et al., 1986). In vitro as well as in vivo studies have shown that HU plays an important role in several processes in the cell: replication from oriC (Dixon and Kornberg, 1984; Ogawa et al., 1989), transposition of TnJO (Morisato and Kleckner, 1987) as well as of bacteriophage Mu (Craigie et al., 1985; Huisman et al., 1989) and DNA inversion reaction in Salmonella typhimurium (Johnson et al., 1986). There is no consensus sequence for the binding site of HU to DNA, just as there is no consensus sequence for eukaryotic histones. Even though we do not know how HU participates in the processes described above, it was proposed that HU may stabilize the DNA in a particular conformation. It has been suggested that HU could be involved in the coiling of DNA during the initiation of replication at oriC (Bramhill and Kornberg, 1988) or in the stabilization of DNA loops during the DNA inversion reaction associated with flagellar phase variation in Salmonella (Johnson et al., 1986). IHF has also been classified as a histone-like protein mainly because it shares amino acid sequences homology with HU. It is also a small, basic, dimeric protein composed of two different subunits with 30% of their amino acids identical: the ca subunit encoded by the himA gene at 38 min in the E. coli chromosome (Miller, 1981; Mechulam et al., 1985) and the 1 subunit encoded by the hip or himD gene at 25 min (Flamm and Weisberg, 1985). This protein although 10-fold less abundant than HU (Bonnefoy et al., 1989), is essential for the integration of bacteriophage lambda (Nash, 1981) and plays a role in several molecular events in the cell (reviewed by Friedman, 1988). Even though, as opposed to HU, a consensus sequence for the binding of IHF to DNA has been determined (Craig and Nash, 1984), it has been proposed by these authors that like HU, IHF stabilizes DNA segments in particular conformations. The extensive sequence homology between IHF and HU and especially in the putative DNA-binding domain of HU, led Yang and Nash to suggest that both proteins bind DNA in a very similar way (Yang and Nash, 1989). Both proteins would encircle DNA with a pair of, sheet arms (Tanaka et al., 1984) which occupy the minor groove. This is different from most regulatory DNA-binding proteins that occupy the major groove. More recently a model has been proposed in which HU and IHF not only bind DNA in the same way, but that they also bend DNA after binding to it (White et al., 1989). It was of interest to examine if HU and IHF interact with
687
E.Bonnefoy and J.Rouvi&re-Yaniv.
DNA similarly and so substantiate the idea of identical binding mechanisms for the two proteins. Another biological reason to compare the non-specific binding of these two proteins to the DNA was the fact that neither HU nor IHF are essential (Wada et al., 1988 and Huisman et al., 1989). Nevertheless we have shown that E. coli strains lacking HU acquire compensatory mutations (Huisman et al., 1989), IHF is evidently a possible candidate. We therefore set up, using the gel retardation technique, a system which allows us to compare several parameters of the binding of these two proteins to DNA. We used short DNA fragments in order to avoid the topological contraints imposed by long, circular or superhelical DNA molecules. The DNA fragments used here are short synthetic fragments predicted to be curved or uncurved on the basis of their nucleotide sequence. We show in this work that HU and IHF interact with DNA differently. Several HU dimers are able to bind to a linear double stranded DNA fragment side by side with a binding length of around 9 bp. Whereas four HU dimers bind to a 42 bp DNA segment, only one IHF dimer binds to the same fragment under similar conditions. We also show that HU does not exhibit strong preferential binding to curved DNA compared to uncurved DNA. On the other hand IHF binds curved DNA much more tightly than uncurved DNA. Another difference described in this work is the ability of HU, but not IHF, to bind single stranded DNA. Moreover we show that the HU a2 or 02 homodimers bind single stranded DNA with higher affinity than the heterodimer (HU o43). Finally, we show that HU, ofi as well as HU 12 and to a lower extent a2, enhances the binding of IHF to a DNA fragment that does not carry the IHF DNA-binding consensus sequence.
Results In order to study and compare the interactions of HU and IHF with DNA we have used the gel-retardation technique which permits the separation of DNA-protein complexes from free DNA on polyacrylamide gels (Garner and Revzin, 1981). The protein bound DNA is retarded in its electrophoretic migration as compared to free DNA. The degree of retardation depends not only upon the molecular weight of the bound protein but also on the conformation acquired by the DNA molecule in the complex; for example it was shown that complexes with bent DNA are retarded more than complexes with unbent DNA (Wu and Crothers, 1984). The DNA fragments used in this work (see Table I) are double stranded or single stranded synthetic oligonucleotides of different sizes. The sizes of the fragments used in this work vary between two and four double helical turns of the B form.
Several HU a,8 dimers bind to a linear DNA fragment side by side with a binding length of - 9 bp Purified HU a43 was mixed with double stranded radioactively labeled oligonucleotides of different sizes as detailed in Materials and methods. The mixture of free and protein bound DNA was analysed on a polyacrylamide gel. After migration the gel was dried and autoradiographed. Figure 1 shows that with saturating amounts of HU two retarded HU -DNA complexes are formed with a 21 bp (NC2 1) as
688
well as with a 26 bp oligonucleotides (NC26); three HU -DNA complexes are formed with a 29 bp oligonucleotide (NC29) and four complexes are formed with a 42 bp oligonucleotide (NC42). We observed that the more retarded complexes gradually appear as the concentration of HU is raised and as they increase in intensity the first less retarded complex gradually disappear. Since the formation of the different protein-DNA complexes is correlated with the concentration of HU used, we believe that the first less retarded complex corresponds to the binding of one HU co4 dimer, the second complex to the binding of two HU a13 dimers, the third complex to the binding of three HU ao1 dimers and the fourth most retarded complex to the binding of four HU caf dimers. Similar results have recently been described for the binding to DNA of the bacteriophage SPO1-encoded TFl protein which is another member of the HU family (Schneider and Geiduscheck, 1990). Several conclusions can be drawn from these experiments: Firstly, it appears from these results that several HU af3 dimers can orderly bind, side by side, to a linear DNA fragment as predicted by the model of Tanaka (Tanaka et al., 1984). The correlation of the number of HU -DNA complexes obtained with a given double stranded DNA fragment with the number of base pairs present in this fragment suggests a binding length of 9 bases per bound HU af3 dimer. Secondly, the binding side by side of several HU a: dimers to a linear double stranded DNA fragment does not appear to be cooperative. Figure 2 shows the scanning of a gel representing the formation of the four HU-NC42 complexes. We plotted the intensity of each complex (Cl, C2, C3 and C4) against the amount of HU protein used and we calculated the slope (A) of each curve which indicates the rate of formation of each complex. As represented in Figure 2, the rate of formation of C2 is slower than the rate of formation of C1, the rate of formation of C3 is slower than the rate of formation of C2 and the rate of formation of C4 is slower than the rate of formation than C3 (ACl > AC2 > AC3 > AC4). This indicates a non-cooperative phenomenon in which the binding of one HU dimer does not enhance the binding of the next HU dimers. Thirdly, the shorter the DNA fragment, the greater the retardation upon binding of a second HU ao4 dimer (see Figure 1 lanes f,h and j). Since all the complexes with this 2: 1 stoichiometry have roughly the same molecular weight, this phenomenon could be a consequence of the DNA adopting a bent conformation inside the complex and this conformation being more pronounced for the shorter DNA fragments. Bending of DNA as a consequence of HU binding has already been proposed (White et al., 1989) and recently demonstrated (Hodges-Garcia et al., 1989). The particular effect we describe here is probably reflecting different possible arrangements of two HU dimers according to the length of the DNA fragment they are bound to. Nevertheless since all the fragments do not have the same base sequences, we cannot completely rule out the possibility that the above described phenomenon is the result of HU inducing different conformations on the different DNA fragments in a sequence dependent manner. This however, seems to us quite unlikely since (i) such an effect is not observe after the binding of the first HU dimer and (ii) HU DNA binding is not sequence -
dependent.
l_h_.
1. SCL1LC.
