|Home | About | Journals | Submit | Contact Us | Français|
Hin, a member of the serine family of site-specific recombinases, regulates gene expression by inverting a DNA segment. DNA inversion requires assembly of an invertasome complex in which a recombinational enhancer DNA segment bound by the Fis protein associates with the Hin synaptic complex at the base of a supercoiled DNA branch. Each of the four Hin subunits becomes covalently joined to the cleaved DNA ends, and DNA exchange occurs by translocation of a Hin subunit pair within the tetramer. We show here that although the Hin tetramer forms a bidirectional molecular swivel, the Fis/enhancer system determines both the direction and number of subunit rotations. The chirality of supercoiling directs rotational direction, and the short DNA loop stabilized by Fis-Hin contacts limit rotational processivity, thereby ensuring that the DNA strands re-ligate in the recombinant configuration. We identify multiple rotational conformers that are formed under different supercoiling and solution conditions.
Site-specific DNA recombination (SSR) systems function to control diverse biological reactions by cleaving and rejoining DNA strands at specific loci (Craig et al., 2002). In the case of flagellar antigen variation in Salmonella, an inversion of a ~1 kb segment of chromosomal DNA catalyzed by the Hin recombinase switches the orientation of a promoter that controls transcription of flagellin genes (Zieg et al., 1977). SSR reactions have also been employed as powerful tools for genetic engineering.
Each SSR system encodes a dedicated recombinase that binds to a specific DNA sequence and catalyses the chemical and mechanical steps required for DNA exchange. Most site-specific recombinases can be classified into 2 mechanistically and structurally distinct families, which are named after their key catalytic residues (Grindley et al., 2006). Tyrosine recombinases, exemplified by Cre and FLP, assemble square planar synaptic complexes with the exchanging DNA strands forming a Holliday structure within the tetramer core. By contrast, serine recombinases like Hin and the γδ or Tn3 resolvases assemble a tetramer with the recombining DNA segments located on opposite sides of the protein complex. Both DNA strands within the center of each recombination site are cleaved by a concerted nucleophilic attack that covalently joins the active site serine of each subunit to the respective 5′ ends of the cleaved DNAs. DNA exchange is proposed to be accomplished by a translocation of one pair of synapsed subunits, along with their covalently-linked DNAs, relative to the other synapsed subunit pair within the tetrameric structure (Figure 1A).
Whereas some recombinases, such as Cre and FLP, function alone in a largely promiscuous manner, many require additional DNA binding and bending proteins, and in some cases DNA supercoiling, to promote assembly of active recombination complexes. These additional factors impart critical regulatory controls, including insuring that recombination only occurs between loci on the same DNA molecule and that only the desired type of DNA rearrangement occurs. The DNA inversion reaction catalyzed by Hin requires DNA supercoiling and the regulatory protein Fis, which binds to two segments within a 65 bp recombinational enhancer sequence (Figure 1A) (Johnson, 2002). The enhancer segment can function at different locations and in either orientation relative to the “hix” recombination sites at the boundaries of the invertible segment. The active recombination complex (invertasome) contains a Hin tetramer assembled from Hin dimers bound to each of the two synapsed hix sites together with two Fis dimers bound to each end of the enhancer (Heichman and Johnson, 1990). The invertasome is assembled at the base of a supercoiled DNA branch, thereby creating a geometry of DNA strands that specifies inversion of the intervening DNA segment upon DNA exchange (Heichman et al., 1991; Kanaar et al., 1990). DNA supercoiling stabilizes the assembly of the tripartite invertasome complex where two negative DNA nodes are trapped (Kanaar and Cozzarelli, 1992).
Most recent molecular and structural studies of Hin and resolvases have employed hyperactive mutants that support recombination without the requirement for accessory factors or DNA supercoiling (Arnold et al., 1999; Dhar et al., 2004; Klippel et al., 1993; Li et al., 2005; Nollmann et al., 2004; Sanders and Johnson, 2004). These mutants can generate stable synaptic complexes with short linear DNA substrates in which all four DNA strands are cleaved and engaged in serine-5′-phosphodiester protein-DNA covalent linkages within the recombinase tetramer. Crystal structures of hyperactive γδ resolvases in DNA-cleaved synaptic complexes have been reported by Steitz, Grindley, and co-workers (Figure 1B) (Kamtekar et al., 2006; Li et al., 2005). These structures reveal that extensive conformational changes occur when the two DNA bound dimers are remodeled into the tetrameric structure and provide a snapshot of how DNA exchange by subunit rotation could occur. Importantly, they show that a flat and hydrophobic interface is created within the tetramer, which could allow pairs of synapsed subunits to rotate relative to each other without extensive exposure of the rotating surfaces to solvent (Figure 1C). Experimental support for rotation about the flat interface within the tetramer of hyperactive Hin and resolvase mutants has been obtained from formation of crosslinks at cysteines introduced at various positions within the interface (Dhar et al., 2004; Li et al., 2005; Kamtekar et al., 2006; this paper). Moreover, earlier studies that followed topological changes in the DNA after single and multiple exchanges are consistent with and initially lead to the proposal of DNA exchange mediated by the rotation of subunits within the synaptic complex of serine recombinases (Kanaar et al., 1988; Stark et al., 1989).
In the present work we ask how the Fis/enhancer control element and DNA supercoiling, two critical features of the wild-type Hin reaction, impact the subunit rotation reaction. Whereas the Hin tetramer assembles into a bidirectional protein swivel, we present evidence that the Fis/enhancer element imparts directionality and limits rotation to a single subunit exchange step, in addition to its established role in activating the initial chemical steps of the recombination reaction. We propose that mechanical energy trapped in the short DNA loop between the enhancer and Hin complex functions to prevent multiple (processive) subunit exchanges that would lead to a random distribution of recombinant (inverted) and parental DNA orientations as well as knotting of DNA duplexes. Finally, time-resolved protein crosslinking experiments combined with molecular modeling provide strong experimental support for intermediate conformers formed during the subunit rotation process whose abundance can be modulated by DNA supercoiling and solution conditions.
Previous site-directed Hin crosslinking experiments were performed with 3′ radiolabeled oligonucleotide substrates using the hyperactive mutant Hin-H107Y, which does not require DNA supercoiling or the Fis/enhancer system to promote recombination (Dhar et al., 2004). In order to probe residues on Hin that are proximal to each other within Fis-activated invertasomes assembled on supercoiled DNA, a plasmid substrate was constructed in which each hix recombination site is flanked by EcoR1 sites (Figures 1D and S1). Digestion by EcoR1 separates the hix sites from the plasmid sequences and enables 3′ end-labeling. In initial experiments, the EcoR1 sites were positioned 50 bp from the center of each hix site, and the enhancer was located 155 bp from the closest hix site (hixL1). Hin, containing cysteines introduced at various positions, and Fis were incubated with the plasmid substrate in ethylene glycol Mg2+-free buffer conditions to generate invertasomes with each Hin subunit covalently-associated with the 5′ ends of the cleaved hix sites at Ser10. The reactions were then directly oxidized using diamide or exposed to bis-maleimide crosslinkers containing spacer lengths ranging up to 16 Å. After quenching the crosslinking and Hin reactions, the products were cleaved with EcoR1, 32P-end-labeled, and subjected to SDS-PAGE (Figure 1E). The radio-labeled low molecular weight bands correspond to a single Hin protomer covalently-linked to the cleaved 54 bp DNA segment at the 5′ end and 32P-labeled at the 3′ EcoR1 site. The higher molecular weight bands that are dependent upon crosslinker correspond to crosslinked Hin-DNA diprotomers.
Reactions employing cysteines at Hin residues 94, 99, 101, 129, and 134 generate large amounts of crosslinked diprotomer products with BMOE (8 Å) or BMH (16 Å) (Figure 1E). In some cases, particularly for Hin-S94C, a crosslinked Hin diprotomer species that migrates as predicted for containing only a single DNA fragment accumulates to significant levels. This would be the expected product if the crosslinked Hin complex cleaved only one of the hix sites; reactions employing Hin-S94C generate substantial levels of linearized plasmids. Hin-D96C produced a trace amount of crosslinked products with BMOE and a low amount with BMH, and Hin-S98C gave a trace amount with BMH. Cysteines that did not give detectable crosslinked products in Fis-activated reactions were located at residues 2, 3, 25, 26, 27, 28, 29, 30, 39, 54, 55, 56, 57, 58, 89, 132, within the catalytic domain and 150, 151, 154, 158, 182, and 183 within the DNA binding domain (data not shown).
The Fis-dependent crosslinking results parallel those obtained from tetramers assembled by the Fis-independent mutant Hin-H107Y on oligonucleotides (Dhar et al., 2004), indicating that the overall structure of Hin within Fis-containing invertasomes is similar to the tetramer assembled by the hyperactive mutant. As discussed further below, the profiles are consistent with Hin models based on the crystal structures of hyperactive resolvase tetramers, provided subunit pairs are allowed to rotate within the tetramer. In the following section we address differences in the kinetic profiles of the crosslinking reaction on supercoiled and non-supercoiled substrates and discuss the implications of these differences on the subunit rotation process.
Hin subunits bound to hixL1 and hixL2 within cleaved invertasomes were distinguished by different length DNA segments in order to follow their movements during the reaction. This was accomplished using a plasmid substrate (pRJ2330) in which the EcoR1 sites flanking hixL1 and hixL2 are 50 and 14 bp, respectively, from the Hin cleavage sites (Figure S1). The short DNA fragment generated after Hin and EcoR1 cleavage at hixL2 is unable to be 3′-labeled by Klenow, presumably because of interference from the Hin protomer covalently attached to its 5′ end. Thus, only the Hin-54 bp covalent complexes originating from hixL1 are 32P-labeled. The supercoiled substrate was pre-incubated with Fis in buffer containing ethylene glycol and EDTA, and the recombination reaction was initiated by addition of Hin. At various times after addition of Hin, the products were subjected to crosslinking with BMOE (8 Å spacer) for 20 sec. The presence of a Hin-(32P)54nt - Hin-(32P)54nt homo-diprotomer crosslinked product after SDS-PAGE is indicative of crosslinking between subunits of dimers originally bound at hixL1, whereas the presence of a Hin-(32P)54nt - Hin-18nt hetero-diprotomer crosslinked product is indicative of crosslinking between subunits originally bound at hixL1 and at hixL2, respectively.
Residues 134 or 129, located at the C-terminal ends of the E helices, are positioned too far apart in the tetramer conformer trapped by the γδ resolvase crystals to permit crosslinking. In a cleaved Hin tetramer model based on the resolvase crystal structure (2GM4), Cys134 Sγ atoms are separated by ~45 Å in subunits from the parental dimer and by ~65 Å from subunits corresponding to a recombinant dimer (Figures 1B, C, ,2A,2A, and Movie S1A, B). A clockwise ~105° rotation of synapsed subunits about the flat interface will align the E helices of recombinant dimers optimally to form the hetero-diprotomer crosslinked product, and a counterclockwise ~75° rotation will align the E helices of subunits from initially bound (parental) dimers to crosslink the homo-diprotomer species (see also Figure S6A).
Time courses of crosslinking reactions were performed with Fis-activated Hin-Q134C (Figures 2B, D, and S2) and Hin-A129C (Figure S3) on supercoiled pRJ2330. The hetero-diprotomer crosslinked product is 5-10 times more abundant than the homo-diprotomer at early time points. The overwhelming bias for the hetero-diprotomer product is consistent with a synapsed pair of Hin protomers within most of the Hin tetramers undergoing a single clockwise rotation about the flat interface after DNA cleavage. Moreover, the crosslinked product ratio remains constant over the course of the 20 min reaction, suggesting a barrier for multiple rotations that would lead to a processive recombination reaction and result in an equal mixture of products.
Kinetic profiles of crosslinked products from Fis-independent Hin-H107Y/Q134C reactions on open circular pRJ2330 (Figures 2C and D) or on oligonucleotide substrates in the absence of Fis (Dhar et al., 2004) are markedly different from the Fis-activated reactions. Initial crosslinks favor the homo-diprotomer product, which is most likely generated from a counterclockwise rotation requiring the smallest rotation, but an equal distribution of products rapidly develops. The profiles on non-supercoiled substrates are consistent with a bidirectional rotation of subunits and/or multiple rotations.
Fis-activated Hin-S94C crosslinking reactions on supercoiled DNA (Figures 3B and D) generate a kinetic profile which is opposite from that obtained with Hin-Q134C. In this case, the homo-diprotomer crosslinked product is highly overrepresented, although the homo-/hetero-diprotomer crosslinking ratio is somewhat less at early time points. This pattern also contrasts with the profile obtained in Fis-independent Hin-H107Y/S94C reactions on open circular DNA (Figures 3C and D) or oligonucleotide substrates (Figures S4A and and3D)3D) (Dhar et al., 2004) where at early reaction times crosslinked hetero-diprotomers are favored, but with increasing Hin reaction times the crosslinked products equilibrate to a near equal mixture.
Residue 94 on subunits of dimers bound to different hix sites and positioned diagonally across the synaptic interface are located closest to each other in the Hin tetramer model. However, the separation between these Cys94 Sγ atoms, ~15 Å, is too far to support the crosslinking observed for Hin-S94C with BMOE (8 Å spacer). Furthermore, most of the products generated on supercoiled DNA are between subunits originally bound to the same hix site. Clockwise rotations ranging from 90° to 155° by one pair of synapsed subunits about the flat interface would be required to obtain the crosslinked homo-diprotomers that are preferentially formed on supercoiled DNA (Figures 3A, S6B, and Movie S2A). As discussed below, we believe it is likely that most of the Cys94 crosslinks are formed after a ~90° clockwise rotation that would generate a rotational conformer (helix E-aligned) similar to that demanded by crosslinking at Cys134.
The hetero-diprotomer crosslinked products that are present early in reactions, particularly those employing non-supercoiled substrates, can be explained by at least two models. One model is a counterclockwise subunit rotation: a ~25° rotation would be sufficient for crosslinking by BMOE (8 Å) and a ~50° rotation would position Cys94 from subunits of parental dimers for direct crosslinking (Figure 3A, S6B, and Movie S2B). Disulfide bonds are formed in reactions employing oligonucleotide substrates (Dhar et al., 2004) but are not detectable using supercoiled DNA (Figure 1D). A potential alternative model was suggested by Kamtekar et al. (2006) based on the structure of the asymmetric resolvase tetramer that was crystallized in the absence of DNA (2GM5). In this structure, Sγ atoms of cysteines modeled at the equivalent of residue 94 of trans-diagonally oriented subunits are positioned suitably for crosslinking with BMOE (8 Å) without additional conformational changes. In a symmetric model of Hin in which each subunit is in an “open” structure similar to two subunits of 2GM5, cysteines at residue 94 also reside within crosslinking distance with BMOE. This conformer, which has been observed in a recent structure of the TP901 serine recombinase catalytic domain tetramer (Yuan et al., 2008), could represent an early intermediate in the assembly of the active synaptic complex.
Fis-activated Hin-M101C crosslinking reactions on supercoiled DNA generate a constant ratio of 65-70% crosslinked hetero-diprotomers over the 20 min reaction (Figures 4B and D). We posit that an equilibrium rapidly becomes established whereby 65-70% of the Hin-M101C invertasomes have undergone ~80° clockwise rotation into the helix E-aligned conformer in which the cysteines are positioned to form crosslinked hetero-diprotomers (Figure 4A and Movie S3). The other 30-35% of invertasomes have undergone a complete 180° subunit rotation where the DNA ends are aligned for ligation in the recombinant configuration and the subunits will crosslink as homo-diprotomers. We note that DNA ligation, induced by adding Mg2+ and diluting out the ethylene glycol after a 20 min cleavage reaction with Hin-M101C, generates >90% unknotted plasmids containing inversions (data not shown). The lack of knotted products indicates that multiple DNA exchanges did not occur over this time frame, consistent with subunit rotation being limited to a single translocation step.
The profile on supercoiled DNA differs from that obtained from reactions on open circular plasmids (Figures 4C and D) or oligonucleotides (Figures 4E and F). On non-supercoiled substrates, the crosslinked hetero-diprotomer form is overrepresented at early times, but the products then shift to a more equal ratio of hetero-and homo-diprotomers. The initial bias for hetero-diprotomers on non-supercoiled substrates fits well with the 2GM4 conformer in which the Cys101 Sγ atoms on subunits located trans-diagonally from each other are within crosslinking distance by BMOE (8 Å) and are predicted to remain within crosslinking distances up to an 80° clockwise rotation (Figures 4A, S6C, and Movie S3). The homo-diprotomer Cys101 crosslinked products that accumulate later in reactions on non-supercoiled DNA could form from either a complete 180° clockwise subunit rotation or a >100° counterclockwise rotation.
In summary, crosslinking experiments from three distinct regions of Hin generate profiles of Fis/enhancer-activated reactions that are most consistent with a single-round clockwise rotation of subunits, whereas profiles from Fis-independent reactions on non-supercoiled substrates are consistent with bidirectional or multiple (processive) rotations (see below). The role of negative supercoiling and the Fis/enhancer control element in modulating the subunit rotation reaction are investigated further below.
Assembly of an invertasome complex at a DNA branch on negatively supercoiled DNA is expected to trap torsional energy that could direct the clockwise rotation of the DNA-linked subunits (Kanaar and Cozzarelli, 1992). Counterclockwise rotation within an invertasome structure would require energy because it would introduce twist resulting in additional negative supercoils. Thus, the strong bias in the crosslinked products observed on negatively supercoiled substrates is consistent with DNA supercoiling determining the direction of subunit rotation. A direct test of this idea is to perform similar experiments using positively supercoiled DNA with the expectation that the direction of Hin subunit rotation will be reversed. We find that Hin reactions on positively supercoiled DNA substrates are inefficient; however, the crosslinking products within cleaved invertasomes are consistent with predominantly counterclockwise rotations of Hin subunits.
For these experiments we used the plasmid substrate pRJ2385, which contains the enhancer located at the native distance of 99 bp from the center of hixL1 (Figure S1). Positively supercoiled preparations of pRJ2385 with mean superhelical densities (sigma values) exceeding +0.034 were generated using reverse gyrase (Figure 5A). Cleavage products generated by Hin in the presence of Fis on positively supercoiled pRJ2385 under Mg2+-free, ethylene glycol reaction conditions increased over the course of 60 min (Figure 5B), but the rate corresponds to only about 10% of cleavage rates on negatively supercoiled DNA, a rate only 3-4 times greater than reactions performed in the absence of Fis. Fis-dependent inversion also increased linearly, but the rate was <3% of that measured on negatively supercoiled pRJ2385 (Figure 5C). Low inversion rates by Hin on positively supercoiled plasmids have been reported previously (Lim et al., 1997).
Positively and negatively supercoiled pRJ2385 were subjected to 20 sec crosslinking reactions at various times after addition of Fis and Hin-Q134C or S94C. As shown in Figure 5D, the homo-diprotomer crosslinked species representing the parental dimeric form of Hin-Q134C is overrepresented at each time point in reactions on positively supercoiled DNA. This profile contrasts with the large bias for hetero-diprotomer crosslinked products obtained with negatively supercoiled pRJ2385 (Figures 5E and F), as was observed for pRJ2330 (Figure 2B). The distribution of crosslinked products generated on the positively supercoiled DNA is consistent with most of the subunits undergoing counterclockwise rotation whereby residues 134 from subunits on the same dimer will initially localize (Figures 2A, S6A, and Movie S1B). Reactions performed with Hin-S94C on positively supercoiled pRJ2385 preferentially generated hetero-diprotomer crosslinked products, particularly at early time points, whereas on negatively supercoiled DNA the products rapidly favor the homo-diprotomer form (Figures 5G-I), as observed with pRJ2330 (Figure 3C). The profile with Hin-S94C on positively supercoiled DNA again suggests preferential counterclockwise rotation, whereby rotations from 25° -90° are predicted to localize Cys94 from subunits of the parental dimers for hetero-diprotomer crosslinking by BMOE (8 Å) (Figures 3A, S6B, and Movie S2B).
To evaluate whether Fis influences the single-step nature of Hin subunit exchange, crosslinking reactions were performed on Hin-Q134C and S94C in the presence of the Fis mutant D20N. Asp20 is one of the critical triad of Hin-activating residues located near the tip of the β-hairpin arm region of Fis (Figure 6F) (Safo et al., 1997). The Fis-D20N mutation reduces rates of Hin-catalyzed DNA inversion to about 15% of Fis-wt but has no effect on other activities of Fis (Osuna et al., 1991; Safo et al., 1997). Twenty minute cleavage reactions with Hin-Q134C and S94C in the presence of Fis-wt and D20N were performed using supercoiled pRJ2330 and subjected to Hin crosslinking with BMOE. Hin-Q134C reactions (Figure 6A) generated nearly equivalent numbers of homo-and hetero-diprotomer crosslinked products and Hin-S94C reactions (Figure 6B) resulted in only a small bias for the homo-diprotomer form when activated by Fis-D20N, unlike profiles obtained with Fis-wt. A likely explanation for this result is that the subunit rotation reaction is no longer limited to a single translocation in the presence of the mutant Fis protein. This is consistent with previous observations showing Hin reactions performed in the presence of Fis-D20N or other Fis proteins containing mutations in the region that interacts with Hin generate a greater proportion of knotted products than reactions with wild-type Fis (Merickel and Johnson, 2004).
We asked whether the location of the Fis/enhancer segment relative to the hix sites could influence subunit rotation processivity. Near the native spacing, where the enhancer is located 99 bp from the center of hixL, there is a strong rotational phasing relationship between Hin-activation and the length of DNA separating hixL and the enhancer (Haykinson and Johnson, 1993). pRJ2330 contains the enhancer spaced 155 bp from hixL1, which corresponds to an additional five DNA helical turns on negatively supercoiled DNA (11.2 bp/turn; (Haykinson and Johnson, 1993; Law et al., 1993)) relative to the native spacing. pRJ2123 is identical to pRJ2330 except that the enhancer is located 6 bp closer to hixL1. Even though this would position the enhancer a half-helical turn closer to hixL1 as compared to pRJ2330, at this distance Fis activates the Hin cleavage reaction similarly on the two plasmids. Nevertheless, Hin-Q134C and S94C crosslinking experiments demonstrate that subunit rotation on pRJ2123 is highly processive as shown by the near equivalent amounts of homo-and hetero-protomer crosslinked products formed (Figures 6C-E and S4B). Subunit rotation processivity in reactions with pRJ2123 correlates with the formation of large numbers of DNA knots in cleavage-ligation experiments, which are indicative of iterative rounds of DNA exchange (Figure S5). We conclude from the crosslinking and DNA knotting analyses that the rotational phasing of the enhancer relative to the proximal hix site influences the processivity of the Hin subunit rotation reaction, probably by affecting the stability of Fis-Hin interactions (see below).
The effects of Fis mutants and enhancer locations on the processivity of Hin subunit exchange suggests a mechanism in which physical linkage between the Fis-bound enhancer and the Hin synaptic complex limits Hin subunit rotations to a single-round translocation. We tested this model by quantitatively assaying the association between Fis and Hin under conditions where single or processive rotations predominate. A cysteine was substituted for Gln21, located between the critical Hin-activating residues Asp20 and Val22 on Fis (Figure 6F), to probe for interactions between Fis and Hin by protein crosslinking. Hin cleavage reactions were incubated for 20 min with either Fis-Q21C, Fis-wt (no cysteine), or Fis-R71C (a cysteine near the DNA binding domain) on pRJ2372 and subjected to 30 sec crosslinking reactions with heterobifunctional agents containing variable spacer lengths between the maleimide (reacts with sulfhydryl on residue 21 of Fis) and succinimidyl (reacts with a primary amine on Hin) groups. After quenching, the plasmid DNA was cleaved with EcoR1, labeled with 32P-dATP in a similar manner as for the Hin-Hin crosslinking, and the products were separated by SDS-PAGE. Reactions with Fis-Q21C generated a prominent band whose molecular weight shift corresponds to a Fis monomer crosslinked to the radio-labeled Hin-54 nt complex with AMAS (4.4 Å spacer) and less prominent band with GMBS (6.8 Å spacer) (Figure 6G). The 32P-label ensures that the crosslinking occurred within a Fis-activated Hin cleavage complex (invertasome); the absence of a detectable Fis-Hin product with Fis-wt or R71C further demonstrates the specificity of the crosslinking.
Efficiencies of Fis-Q21C crosslinking to Hin using AMAS on pRJ2372 (enhancer in-phase) were then compared with the out-of-phase substrate pRJ2118 where processive subunit rotation dominates. Reactions using pRJ2118 generated about 1/3 the level of Fis-Hin crosslinked products as pRJ2372 (Figures 6G and H), implying that the Fis/enhancer segment is less stably associated with Hin when the enhancer is in an out-of-phase location. The inverse correlation between Fis-Hin crosslinking efficiency and processivity of subunit exchange is consistent with a physical association between the enhancer and the Hin synaptic complex in the invertasome limiting subunit rotations.
If the Fis/enhancer system is limiting processivity of subunit rotation by a DNA torsion mechanism, long DNA segments separating the enhancer and hix sites should relieve this effect and allow for multiple rotations. To test this model, Hin-Q134C crosslinking was performed on pRJ2340, which contains 730 bp between the closest hix and the enhancer (Figure S1). Hetero-and homo-diprotomer crosslinked products were formed at nearly equivalent amounts, consistent with multiple subunit rotations (Figures 6I and S4C). Although standard inversion reactions with pRJ2340 generate primarily unknotted inversion products, ligation of products after long-term cleavage reactions with pRJ2340 also generate large numbers of DNA knots (Figure S5). Thus, a substrate with long DNA segments between the hix and enhancer sites readily undergoes processive rotations when ligation is inhibited for long time periods.
Identity between the two base pair core nucleotides where the cleavage-ligation chemistry occurs is essential for single-round DNA exchange leading to inversion (Johnson and Bruist, 1989; Johnson and Simon, 1985). It is possible that base pairing over the core nucleotides may also function to limit the number of subunit rotations after a single round of exchange by stalling the reaction in the recombinant configuration. This pause could encourage ligation to occur, thereby generating the unknotted inverted molecules that are the primary reaction products. By this model, if base pairing cannot form due to a mutation in the core nucleotides (Figure 7A), multiple rounds of exchange occur, consistent with the formation of processive recombination products (DNA knots) that are observed in reactions employing plasmid substrates containing hix sites with non-identical core sequences (Heichman et al., 1991; Merickel and Johnson, 2004). We tested the importance of core base pair complementarity on the processivity of Hin subunit rotation by performing Hin crosslinking experiments on pRJ2383, which contains a wild-type hixL1 site (AA core sequence) and a mutant hixL2(AT) site (Figures 7A and S1). Hin-Q134C generated nearly exclusively homo-diprotomer products (Figure 7B), and Hin-S94C generated nearly exclusively hetero-diprotomer products (Figure 7C), even after 15 min reaction times. Since the crosslinking profiles are essentially identical to reactions performed with substrates containing wild-type hix sites, we conclude that base pairing within the core nucleotides does not play a dominant role in limiting subunit rotation to a single cycle of DNA exchange. This conclusion is also supported by the results with pRJ2340 containing long DNA segments between the enhancer and hix sites where processive rotations readily occur even when the DNA ends are compatible for ligation after a single exchange (Figure 6I above). The requirement for complementary core nucleotides for recombination must therefore be primarily at the chemical step of the ligation reaction.
The high concentrations of ethylene glycol plus absence of metal combine to trap Hin complexes in a DNA-cleaved intermediate state (Johnson and Bruist, 1989; Sanders and Johnson, 2004). The site-directed crosslinking profiles discussed above for each of the Hin cysteine mutants (Figures 2--4)4) suggests that subunit rotation within the DNA cleaved Hin tetramer may preferentially stall in a conformer in which the four E helices are aligned in a parallel/anti-parallel configuration. The following experiments provide evidence that the ethylene glycol Mg2+-free reaction condition preferentially stabilizes a helix E-aligned conformer, thereby providing a partial explanation for why these solution conditions trap cleaved reaction intermediates.
Fis-activated crosslinking reactions on Hin-M101C under ethylene glycol Mg2+-free conditions generate both hetero-diprotomer products, which likely reflect a helix E-aligned conformer, and homo-diprotomer products, which likely reflect a recombinant DNA-aligned conformer (Figure 4). Using pRJ2383, where ligation in the recombinant DNA-aligned conformer is inhibited because of the non-complementary core nucleotides, the ratio of hetero-to homo-diprotomer crosslinked products is 1.9 (Figure 7D, lane 2), similar to reactions with pRJ2330 (Figure 4C). Cleaved invertasomes generated on pRJ2383 under 25% ethylene glycol Mg2+-free conditions were then chased with 4 volumes of buffer containing 10 mM Mg2+, conditions that would lead to near 100% of the cleaved DNA ligating into the inverted configuration with the wild-type substrate (Johnson and Bruist, 1989). This resulted in the distribution of crosslinked products shifting to a hetero/homo-diprotomer ratio of 0.5 (lane 8), consistent with a dominant DNA-aligned conformer. When 25% ethylene glycol was re-added to the reaction, the ratio shifted back to 1.7 (lane 12), or a dominant helix E-aligned conformer. Taken together, these results suggest that the two conformers can oscillate between each other and that high ethylene glycol preferentially stabilizes the helix E-aligned conformation. The presence of Mg2+ had little effect on the distribution of Hin-M101C crosslinked products, leading us to suspect that Mg2+ may influence the chemistry of the ligation step, perhaps indirectly by stabilizing base pairing over the core nucleotides.
In this study we investigated the role of the Fis/enhancer regulatory element and DNA supercoiling on the mechanics of the Hin subunit rotation process in cleaved synaptic complexes. We find that the Fis/enhancer element imparts directionality to the rotation of an otherwise bidirectional protein swivel. The direction of subunit rotation is set by the chirality of DNA supercoils that are trapped by association of the Fis-bound enhancer with the Hin-DNA complex. The enhancer is normally located 99 bp from hixL, and we show that at this and 155 bp spacings it functions to restrict subunit rotation to a single translocation. Moreover, we provide evidence that physical association of Fis with the Hin-DNA complex is required for this additional function of the enhancer. Subunit rotation is not restricted when long DNA segments separate the enhancer from the hix sites, strongly implying that torsional constraints generated from short DNA loops are responsible for controlling subunit rotation. Identity of the core nucleotides that must base pair in order to ligate does not appear to play a significant role in limiting subunit rotations when ligation is inhibited. As elaborated below, many of the features revealed here by analyzing movements of the proteins are consistent with changes in global DNA structure that were elucidated in earlier topological studies where movements of DNA strands were inferred from single-round and processive exchange reactions. Altogether, the evidence points to a remarkably consistent view of the function of the Fis/enhancer system on the Hin subunit rotation reaction.
The cleaved DNA segments in the resolvase tetrameric crystal structures are positioned in a manner that would likely represent the recombinase structure immediately after DNA cleavage (or before ligation). In this conformer, which we refer to as DNA-aligned, the C-terminal ends of the E helices of all subunits are too far from each other (>45 Å) to support crosslinking. Nevertheless, Fis-activated reactions on negatively supercoiled DNA result in robust crosslinking between cysteines at residues 134 or 129 on subunits that were initially bound to different hix sites. These cysteines are modeled to become sufficiently close to crosslink into hetero-diprotomers upon a 75-105° clockwise rotation, depending on the residue and the curvature of the E helices (Figures 1C, ,2A,2A, S6A, and Movie S1A). Rotational conformers in this range have the E helices from each of the four subunits approximately aligned in a parallel/antiparallel configuration within the tetramer. Crosslinks between cysteines at residue 94 located on the opposite side of Hin generate a different kinetic profile from the C-terminal helix E cysteines, but one that is also entirely consistent with supercoil-directed Hin subunit rotation in the clockwise direction. Although the closest approach of Cys94 occurs after a 110-140° clockwise rotation, subunits are within crosslinking distance in helix E-aligned rotational conformers (Figures 1C, ,3A,3A, S6B, and Movie S2A). Fis-activated reactions employing positively supercoiled DNA preferentially generate crosslinking patterns that are consistent with counterclockwise rotation of the Hin subunits (Figures 1C, ,2A,2A, ,3A3A and Movie S1B, S2B). In contrast, Fis-independent reactions performed on non-supercoiled plasmids or DNA fragments generate both the crosslinked homo-and hetero-diprotomers, consistent with bidirectional and/or multiple (processive) rotations.
Residue 101 is unique because a cysteine at this position can support crosslinking in both helix E-aligned and DNA-aligned rotational conformers (Figures 1C, ,4A,4A, and Movie S3). Crosslinking profiles between Cys101 provide strong evidence for multiple conformational states and for the dynamic nature of subunit rotation. These experiments also provide insights into a long standing question of how the presence of ethylene glycol inhibits the ligation step. On negatively supercoiled substrates the hetero/homo-diprotomer ratio remains constant at ~2 at all time points. Reactions with less ethylene glycol (not shown) or dilution of ethylene glycol produce ratios that are shifted in favor of the homo-diprotomer, which for residue 101 is modeled to be in the DNA-aligned conformer. Re-addition of ethylene glycol shifts the distribution back to a bias for the helix E-aligned conformer. These experiments suggest that the Hin tetramer adopts different preferred rotational conformers depending on solution conditions with high ethylene glycol favoring the helix E-aligned conformer where the DNA ends are rotated out of position for ligation. We note that rotational conformers around the helix E-aligned and DNA-aligned structures are predicted to exhibit the largest area of contacts between synapsed subunits of Hin or resolvase (Figure S7), and thus may be normal pause sites in the rotation reaction.
Both conformational and torsional properties of supercoiled DNA are utilized by the Fis-activated Hin reaction. Supercoiling is required to achieve contacts between Fis and Hin to assemble the tripartite invertasome complex at the base of a DNA branch (Heichman and Johnson, 1990; Heichman et al., 1991). The branched invertasome structure (Figure 1A) traps thermodynamic energy which is released upon DNA exchange by subunit rotation. A single DNA exchange has been measured in the Gin-and Hin-catalyzed inversion reactions to release four supercoils (Kanaar et al., 1988; Merickel and Johnson, 2004). The linking number change of +4, together with the chirality of knots produced from multiple exchanges on negatively supercoiled DNA (Heichman et al., 1991; Kanaar et al., 1990), are consistent with the clockwise direction of subunit rotation determined in the present work. In this work we also demonstrate the converse; subunit rotation on positively supercoiled DNA preferentially occurs in the counterclockwise direction. It is important to emphasize, however, that supercoiling energy is not required for subunit rotation since experiments measuring DNA exchange (Klippel et al., 1988; Klippel et al., 1993; Sanders and Johnson, 2004) or subunit rotation (this paper and Dhar et al., 2004) in Fis-independent reactions show that the reaction can occur on non-supercoiled DNA. Moreover, the results with Cys101 indicate that some subunit rotation can occur in the direction disfavored by supercoiling. Indeed, some of the crosslinked products observed at the very early time points with Cys94 and Cys134 on supercoiled DNA may reflect reverse rotations.
Several factors may contribute to the poor Fis-dependent cleavage reaction and even poorer inversion reaction on positively supercoiled DNA. Our positively supercoiled DNA preparations probably have a lower superhelical density than negatively supercoiled plasmids isolated from cells. Previous studies have shown that physiological densities of negative supercoiling are required for efficient inversion (Lim and Simon, 1992). The geometry of a DNA branch formed on positively supercoiled DNA may not be optimal for stabilizing Fis-Hin contacts. Indeed, quantifying the numbers of DNA cleaved, inverted, and knotted products formed on positively supercoiled DNA indicate that most of the cleaved Hin complexes that engage in subunit rotation do so in a processive manner, implying that the Fis/enhancer segment has been released (data not shown). Processive rotations will also decrease the number of inversion products because only ligation after odd numbers of exchanges will generate the inverted DNA orientation.
A key feature of the Fis-activated subunit rotation reaction is that most events are limited to a single translocation. Earlier DNA topological analysis also implied that most reactions initiated by wild-type Hin are restricted to a single exchange; less than 5% of reactions in vivo or under standard in vitro conditions generate DNA knots indicative of multiple exchanges (Heichman et al., 1991; Merickel and Johnson, 2004). In addition to activating DNA cleavage, several lines of evidence suggest that the Fis/enhancer element also functions to limit further Hin subunit rotations by forming a stable loop between Fis at the enhancer and Hin bound to the proximal hix site. This loop is only 99 bp in the native substrate, and the half-turn of twist that is introduced into the tethered DNA upon each 180° rotation of the DNA will limit the number of potential subunit rotations. We tested two experimental conditions that destabilize the Fis-Hin loop: a Fis mutant that is partially defective in Hin activation and a substrate in which the enhancer is located out-of-helical phase with the proximal hix site. Both conditions resulted in increased subunit rotation processivity as assayed by crosslinking and multiple DNA exchanges as assayed by knotting (this work) (Merickel and Johnson, 2004). Moreover, we showed that crosslinking between Fis and Hin is markedly reduced with the out-of-phase substrate, consistent with the enhancer being released from Hin when processive reactions occur. Finally, when long (>700 bp) DNA segments separate the enhancer from the hix sites, Hin crosslinking experiments show that processive rotations readily occur within invertasomes held under cleavage conditions. The increased processive subunit rotation as a function of hix-enhancer DNA length correlates with an increase in complex knots in the plasmid when DNA-cleaved invertasomes are then allowed to ligate (pRJ2330 vs. pRJ2340, Figure S5). Likewise, knotting of substrates containing non-identical core nucleotides was found to be much more efficient with long hix-enhancer separations both in vitro and in vivo (Heichman et al., 1991; Merickel and Johnson, 2004). Multiple rotations will be more easily accommodated since the twist introduced upon each exchange is dispersed throughout the longer DNA segment.
In vitro topological analysis of the related Gin-catalyzed DNA inversion reaction, which controls the synthesis of alternative tail fibers in phage Mu, suggests that subunit rotation is more processive than in the Hin reaction (Kanaar et al., 1990). This led to the proposal that the Gin enhancer normally functions by a “hit-and-run” mechanism, unlike what we find with Hin. Our results suggest that this difference could be explained by a lower affinity between Fis and Gin as compared to Hin, leading to a less stable DNA loop. A more processive reaction may be advantageous for the phage because the resulting mixture of tail fibers enables the population to infect multiple host species. The homologous Cin DNA invertase system, which controls expression of phage tail fiber genes in phage P1, has its enhancer located about 500 bp from the closest recombination site and thus would be expected to support processive exchanges (Huber et al., 1985).
Topological studies on the resolvase class of serine recombinases provide strong evidence that these enzymes predominantly catalyze a single-round clockwise rotation of DNA strands (Grindley, 2002; Rowland et al., 2002; Stark et al., 1989). Although the structures of the recombination complexes are different from those assembled by DNA invertases, accessory resolvase proteins, together in some cases with DNA bending proteins, are believed to control rotations of the enzymatically-active resolvase protomers. Auxiliary factors that participate in tyrosine recombination reactions also regulate steps in the DNA strand exchange reaction. For example, the programmed set of sequential DNA exchanges that occur in the phage λ site-specific recombination reactions is directed by the functions of accessory proteins that organize the intasome structures (Radman-Livaja et al., 2006). In addition, the ATP-dependent translocase FtsK directs the order of DNA exchanges catalyzed by XerCD recombinase, ensuring that the reaction proceeds forward to recombinant products (Aussel et al., 2002). Like Hin, in each of these cases energetic forces that are introduced or captured within the DNA by accessory proteins are transduced to the recombinase to profoundly affect the product of the reaction.
Most Hin mutants used in this work were purified from renatured inclusion bodies (Sanders and Johnson, 2004); some were native preparations obtained from low temperature inductions (Merickel et al., 1998). Fis mutants and purification have been described (Safo et al., 1997).
The basic structure of the plasmid substrates is shown in Figure 1C, and details of the individual substrates are given in Figure S1. Open-circular plasmid DNA was prepared by nicking with DNase I in the presence of ethidium bromide. Positively supercoiled DNA was prepared by incubating plasmids with recombinant Archaeglobus fulgudis reverse gyrase, purified from pRGY1N, as described (Rodriguez and Stock, 2002), except that 15% DMSO was included in the reaction (G. Chaconas, personal communication) and the reaction temperature was 85°C. pRJ2385 (4615 bp) preparations contained at least 15 positive supercoils (σ >0.034), as determined by band counting; positive supercoiling was confirmed on chloroquine gels.
Hin (typically 30-40 ng) was added to 25 μl reaction mixes containing 20 mM HEPES (pH 7.5), 80 mM NaCl, 4 mM CHAPS, 2 mM EDTA, 30% ethylene glycol, 10 μg poly-glutamate, 0.05 pmole plasmid DNA, and 15 ng Fis. HU (100 ng) was included for reactions using pRJ2385. After 20 min at 37°C, bis-maleimidoethane (BMOE, 8 Å) or bis-maleimidohexane (BMH, 16 Å) (Pierce, dissolved in DMSO) was added to 0.4 mM and then quenched after 1 min by addition of DTT to 40 mM plus 0.05% diethylpyrocarbonate and then ethanol precipitated. Direct S-S crosslinking employed 0.4 mM diamide (Sigma) for 1 min, and DTT was not added. For kinetic experiments, the Hin reaction time was typically 0.5 – 20 min, and crosslinking with BMOE was reduced to 20 sec. The precipitate was dissolved in 40 μl EcoR1 buffer and digested for 1 h with 20 u EcoR1. A 5 μl mixture containing 1 nmole dTTP, 5 pmole dATP, 1 μCi α-32P-dATP (800 Ci/mmole MP Biomed), 40 mM DTT, and 1u Klenow polymerase (NEB) in 0.5× EcoR1 buffer was then added and incubated at 30°C for 30 min, followed by another 5 min after addition of 1 μl 1 mM dATP. After heat inactivation of the Klenow (65°C, 10 min), the samples were digested with 20 u BamH1, which cleaves within the invertible segment to reduce a contaminating labeled DNA band in the SDS gel. SDS loading dye (without reducing agent when diamide used) was added, samples heated at 74°C for 5 min, and aliquots were electrophoresed through SDS-polyacrylamide gels (8-10% depending on the Hin mutant and DNA substrate). 74°C was used to avoid the presence of hydrolysis products but in some cases resulted in incomplete denaturation of Hin and the appearance of multiple bands. Hin crosslinking using 3′-32P-labeled 112 and 40 bp hixL substrates followed the method described in (Dhar et al., 2004). For Hin-Fis crosslinking, reactions were set up as detailed above using pRJ2372 or pRJ2118, Hin-wt, and Fis-Q21C. After 20 min at 37°C the reactants were crosslinked with 0.4 mM N-succinimidyl iodoacetate (SIA, 1.5 Å), N-[α-maleimidoacetoxy]succinimide ester (AMAS, 4.4 Å), or N-[γ-maleimidobutyryloxy]sulphosuccinimide ester (GMBS, 6.8 Å) (Pierce). Crosslinking was typically quenched after 30 sec with 20 mM lysine and 40 mM DTT. The reaction mixture was ethanol precipitated, digested with EcoR1, radiolabeled as above, and subjected to 10% SDS-PAGE.
The Hin tetramer model was derived from the γδ resolvase crystal structure 2GM4 (Kamtekar et al., 2006) except that the Hin DNA binding domain structure (1IJW) was used for residues 143-190 (see Supplementary Information). Residues within the Hin and γδ resolvase catalytic domains (residues 1-134) share 40% identity and 63% similarity, and secondary structure prediction programs show close correspondence between the Hin and resolvase structures based on primary sequence and the X-ray structures of resolvase. The 2GM4 X-ray structure is from a resolvase-Hin chimera in which the Hin sequence is substituted in place of resolvase residues 96-105. Phyre (Bennett-Lovsey et al., 2008) was used to generate the initial model of a Hin subunit based on 2GM4 and the modeled tetramer was subjected to energy minimization using CNS (Brunger et al., 1998). The rotating interfaces of Hin and resolvase appear remarkably chemically similar with both containing exclusively hydrophobic residues, but the Hin model interface is not quite as flat after minimization as found in the resolvase X-ray structures. Structure figures were prepared using PyMOL (DeLano, 2002).
We thank George Chaconas for reverse gyrase and advice on its use, and Daniela Stock for pRGY1N. This work was supported by NIH grant GM038509.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.