Characterization of the Dimerization Domain in the FNR Transcription Factor*

The global anaerobic regulator FNR from Escherichia coli is a dimeric Fe-S protein that is inactivated by O 2 through disruption of its [4Fe-4S] cluster and conver-sion to a monomeric form. As a first step in elucidating the molecular interactions that control FNR dimerization, we have performed alanine-scanning mutagenesis of a potential dimerization domain. Replacement of many hydrophobic residues (Met-143, Met-144, Leu-146, Met-147, Ile-151, Met-157, and Ile-158) and two charged residues (Arg-140 and Arg-145) with Ala decreased FNR activity in vivo . Size exclusion chromatography and Fe-S cluster analysis of three representative mutant proteins, FNR-M147A, FNR-I151A, and FNR-I158A, showed that the Ala substitutions produced specific defects in dimerization. Because hydrophobic side chains are known to stabilize subunit-subunit interactions between (cid:1) -helices, we propose that Met-147, Ile-151, and Ile-158 lie on the same face of an (cid:1) -helix that constitutes a dimerization interface. This alignment would also position Arg-140, Met-144, and Asp-154 on the same helical face. In support of the unusual positioning of a negatively charged residue at the dimer interface, we found that replacing Asp-154 with Ala repaired the defects caused by Ala substitutions of other residues located on the same helical face. These data also suggest that Asp-154 has an inhibitory effect on dimerization, which may be a key element in the control of FNR

The global anaerobic regulator FNR from Escherichia coli is a dimeric Fe-S protein that is inactivated by O 2 through disruption of its [4Fe-4S] cluster and conversion to a monomeric form. As a first step in elucidating the molecular interactions that control FNR dimerization, we have performed alanine-scanning mutagenesis of a potential dimerization domain. Replacement of many hydrophobic residues (Met-143, Met-144, Leu-146, Met-147, Ile-151, Met-157, and Ile-158) and two charged residues (Arg-140 and Arg-145) with Ala decreased FNR activity in vivo. Size exclusion chromatography and Fe-S cluster analysis of three representative mutant proteins, FNR-M147A, FNR-I151A, and FNR-I158A, showed that the Ala substitutions produced specific defects in dimerization. Because hydrophobic side chains are known to stabilize subunit-subunit interactions between ␣-helices, we propose that Met-147, Ile-151, and Ile-158 lie on the same face of an ␣-helix that constitutes a dimerization interface. This alignment would also position Arg-140, Met-144, and Asp-154 on the same helical face. In support of the unusual positioning of a negatively charged residue at the dimer interface, we found that replacing Asp-154 with Ala repaired the defects caused by Ala substitutions of other residues located on the same helical face. These data also suggest that Asp-154 has an inhibitory effect on dimerization, which may be a key element in the control of FNR dimerization by O 2 availability.
FNR is a global regulator that allows Escherichia coli to adapt to anaerobic growth conditions. In the absence of oxygen, FNR is in its active form and binds to specific DNA sites to either activate or repress transcription of target genes (for a review see Refs. 1 and 2). A novel feature of FNR is that its DNA binding activity is associated with its ability to dimerize, a property regulated by the presence of an O 2 -labile [4Fe-4S] cluster (3,4). When the protein contains a [4Fe-4S] cluster, FNR forms homodimers in solution (4,5). However, when O 2 is present, the [4Fe-4S] clusters are destroyed, and the ability of FNR to dimerize is decreased, therefore reducing its site-specific DNA binding and abolishing its ability to control transcription (5,6). Thus, for FNR, DNA binding and dimerization appear to be tightly coupled processes that provide a mechanism for controlling transcription of specific genes in response to changes in O 2 levels.
The molecular interactions that promote dimerization of FNR are unknown because there is currently no structural information for this protein. Nevertheless, amino acid residues 140 -159 of FNR have been proposed to function in subunitsubunit interactions because the position of these residues corresponds to the ␣-helix along which the closely related E. coli transcription factor, the cAMP receptor protein (CRP), 1 dimerizes via a coiled coil interaction (4, 7) (Fig. 1). Consistent with this region having a role in FNR subunit-subunit interactions, substitution of Asp-154 to Ala ( Fig. 1) increases FNR dimerization in the absence of a [4Fe-4S] cluster (3,8). However, the function of other residues within this putative dimerization domain has not been investigated. In addition, the amino acid side chains of CRP and FNR are not highly conserved in this region (9), raising the issue of whether this region of FNR will have unique features to attain regulated dimerization.
The role of the [4Fe-4S] cluster in promoting subunit-subunit interactions is even less well understood. The [4Fe-4S] cluster is ligated by three Cys residues located within the N-terminal end of FNR, a region that is absent in CRP. The fourth ligand of the cluster (Cys-122) is predicted to fall within a proposed ␤-roll structure of FNR (10 -12) that precedes the putative dimerization domain (Fig. 1). Given the location of these four cysteine residues, the [4Fe-4S] cluster of FNR is not expected to be very close to the dimerization helix (13). Thus, it is an open question as to how the presence of a [4Fe-4S] cluster facilitates subunit-subunit interactions within FNR.
To understand the molecular interactions that promote FNR dimerization, we have examined the role of residues 140 -159 of FNR. Alanine scanning mutagenesis was initially used to identify residues of this region required for FNR activity in vivo. Size exclusion chromatography and Fe-S cluster analysis was subsequently used to characterize the in vitro properties of representative mutant proteins. Our findings suggest that residues 140 -159 define a dimerization domain of FNR. In addition, the properties of these mutant FNR proteins support the notion that the [4Fe-4S] cluster of FNR promotes subunit interactions by altering the monomer-dimer equilibrium.

EXPERIMENTAL PROCEDURES
Construction of Alanine Mutants-Plasmids and strains are listed in Table I. Single alanine substitutions in the dimerization helix (residues 140 -159) were constructed by site-directed mutagenesis of pPK821 using oligonucleotide primers (Operon Technologies) as described previously (3). Alanine mutants with a second alanine substitution at Asp-154 were constructed in the same manner except using pPK822 (3) as the starting template. The FNR coding sequence of the pPK821 or * This work was supported by Postdoctoral Training Grant GM19792 from the National Institutes of Health (to L. J. M.) and Grant GM45844 from the National Institutes of Health (to P. J. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  pPK822 derivatives was cloned into pET-11a using the NdeI and BamHI restriction sites. For some mutants, the fnr gene with the normal fnr promoter region was cloned into pACYC184 using the HindIII and BamHI restriction sites. The fnr gene of each plasmid was sequenced using either a 33 P thermosequenase terminator cycle sequencing kit (Amersham Pharmacia Biotech) or a Big Dye sequencing kit at the University of Wisconsin Biotechnology facility to verify the presence of only the introduced mutation.
In Vivo Assays for FNR Activity-pET-11a constructs expressing FNR or its mutant derivatives from fortuitous vector sequences were utilized for most experiments because Western blot analysis showed that the FNR protein levels produced from the pET-11a-derived plasmids are 3-5 times lower than that expressed from the chromosome (data not shown). Thus, the low levels of protein expressed from these plasmids would avoid masking of potential dimerization defects at higher protein concentrations that would result if expressed from its own promoter on this plasmid. Additionally, because FNR synthesis is not under the control of its own promoter, these constructs have the added benefit of avoiding any differences in protein concentration that might arise through a defect in negative autoregulation of FNR synthe-sis (14,15). We also assayed the activity of some mutant proteins at higher protein concentrations by using pACYC184-derived plasmids. FNR protein produced from the pACYC184-WT fnr construct was 3-5 times higher in concentration than that produced from WT fnr on the chromosome (data not shown).
FNR activity was determined by assaying ␤-galactosidase activity from cultures of E. coli strains (see Table I) that have lacZ under the transcriptional control of either the narG, dmsA, or ndh promoter as described previously (3). Cell cultures were grown under anaerobic conditions in M9 minimal medium containing 0.2% glucose, and ampicillin (25 g/ml) or chloramphenicol (20 g/ml) was added to the media when the strain contained either pET-11a or pACYC184 derived plasmids, respectively. In addition, 1.4 mM potassium nitrate was added to the media when assaying expression of narG with the pET-11a plasmids to stimulate phosphorylation of the coactivator, NarL (16). Whereas the trends observed both with and without nitrate are the same, the increased levels of ␤-galactosidase observed in the presence of nitrate allowed us to better distinguish those mutants that retained some activity over that found in the absence of FNR.
Anaerobic Isolation of FNR and Size Exclusion Chromatography-FNR was purified from cells overexpressing mutant or WT FNR under anaerobic conditions as described previously (17). The iron and S 2Ϫ occupancy of each preparation was evaluated using procedures from Kennedy et al. (18) and Beinert (19), respectively. Size exclusion chromatography was performed on a Amersham Pharmacia Biotech HR-12 Superose column (10 ϫ 300 mm) attached to a Beckman high pressure liquid chromatography system enclosed within a Coy anaerobic chamber with an atmosphere of 90% N 2 , 5% CO 2 , and 5% H 2 . The elution experiments were performed in 50 mM KPO 4 (pH 6.8), 400 mM KCl, and 10% glycerol at a flow rate of 0.5 ml/min. The FNR samples were injected in 200-l aliquots at a concentration of [4Fe-4S] FNR protein ranging from 5 to 40 M. The absorption profile of the eluent was continuously monitored from 200 to 600 nm using a photodiode array detector (Beckman 168 System Gold). The column was calibrated with cytochrome c, bovine serum albumin, and carbonic anhydrase.
Absorption and CD Spectroscopy-Absorption spectra (600 -200 nm) of the mutant proteins (5-10 M) in 50 mM KPO 4 (pH 6.8) and 400 mM KCl were obtained in sealed quartz cuvettes (1-cm pathlength) on a PerkinElmer Life Sciences Lambda2 spectrometer. Circular dichroism spectra over a range of 300 -198 nm were taken in 1-mm sealed quartz cuvettes in an Aviv model 202SF CD spectropolarimeter. The protein samples were 5 M FNR in a buffer of 50 mM KPO 4 (pH 6.8) and 100 mM KCl. Each spectrum was recorded in 1 nm wavelength increments, and the signal was acquired for 3 s at each wavelength.

Alanine Scanning Mutagenesis of the Proposed Dimerization Domain Indicates That a Hydrophobic Surface Is Required for
FNR Function-Amino acid residues 140 -159 of FNR are analogous to the residues of the CRP C-helix that promotes subunit interactions through formation of a coiled coil. To test if this 20-amino acid region performs an analogous function in FNR, we examined the effect of replacing each residue with Ala on FNR function in vivo. If our hypothesis is correct, we would predict that at a minimum, Ala substitution of hydrophobic residues with a spacing typical of a coiled coil interaction (3-4 residues apart) should decrease FNR activity.
FNR activity was assessed by measuring the level of ␤-galactosidase from anaerobically grown strains containing lacZ under control of the FNR-dependent narG promoter (P narG ). For this analysis, pET-11a containing fnr or its mutant derivatives cloned behind the T7 promoter were utilized because in the absence of T7 RNA polymerase, we found that FNR levels produced from this plasmid do not exceed single copy levels of FNR (data not shown). The majority of the amino acid substitutions that were found to decrease FNR activity contained an Ala in place of a hydrophobic residue (FNR-M143A, FNR-M144A, FNR-L146A, FNR-M147A, FNR-I151A, FNR-M157A, and FNR-I158A; Fig. 2a). In addition, when two charged residues were replaced by Ala (FNR-R140A and FNR-R145A) FNR activity was also reduced. Many of the residues that were found to be important for FNR activity are spaced three to four residues apart (Arg-140, Met-144, Met-147, Ile-151, and Ile-  158), suggesting that they fall on the same side of an ␣-helix ( Fig. 2b; the residues of the helix are fitted to a coiled coil model; see "Discussion"). Similar results were observed using strains containing lacZ fused to another FNR-dependent promoter, P dmsA (data not shown) indicating that the phenotypes of these mutant proteins are not promoter-specific. Western blot analysis demonstrated that the mutant proteins are present at levels similar to that of wild type (data not shown), indicating that none of these defects can be simply attributed to differences in FNR levels.

Mutant FNR Proteins Show a Decrease in Dimerization in
Vitro-To test if the defects of the Ala substitution mutants resulted from altered subunit-subunit interactions, representative proteins (FNR-M147A, FNR-I151A, and FNR-I158A) were isolated anaerobically and characterized. The visible spec-tra ( Fig. 3) of the mutant proteins were shown to have an absorption band centered at 410 nm that is characteristic of a [4Fe-4S] cluster (5). In addition, the amount of iron and S 2Ϫ in the protein preparations was quantified to determine whether the occupancy of the [4Fe-4S] cluster in the mutant proteins was similar to that of WT-FNR. This analysis showed that 50, 70, and 80% of isolated FNR-M147A, FNR-I151A, and FNR-I158A, respectively, contained a [4Fe-4S] cluster, which is within the range of 50 -85% normally observed with WT-FNR (17). Taken together, these data suggest that the mutant proteins contain a [4Fe-4S] cluster at levels similar to that of WT-FNR.
To access directly whether any of the mutant proteins had defects in dimerization, size exclusion chromatography of isolated proteins was carried out under anaerobic conditions. To distinguish the elution profile of the clusterless FNR (apoFNR) from the [4Fe-4S] form, the absorption profile of the column eluate was monitored on line from 200 to 600 nm. As expected for WT FNR protein (4) Fig. 4). Furthermore, at all column onput concentrations tested (5-20 M), all of the [4Fe-4S] WT-FNR eluted as a dimer (Fig. 5a).
In contrast, the elution behavior of some mutant FNR proteins was dependent upon their concentration. ization defect than FNR-I158A. The concentration dependence of the elution profiles and the non-symmetrical and broad nature of the peaks suggest that some fraction of the dimeric [4Fe-4S] mutant protein population was dissociating as it moved through the column. These data indicate that replace-ment of Met-147 and Ile-158 with Ala decreases but does not eliminate FNR dimer formation.
In comparison, FNR-I151A had a more severe defect because the majority of I151A-FNR eluted largely as a monomer up to an onput concentration of 20 M [4Fe-4S] protein (Fig. 5d). This monomer species had a visible absorption spectrum characteristic of a [4Fe-4S] protein, indicating that a [4Fe-4S] cluster can be found in the FNR protein even when the protein is not a dimer. Because FNR-I151A did not appear to dimerize well at even very high concentrations of protein, we evaluated the overall structural features of this protein by comparing the CD of the protein bands of FNR-I151A and WT-FNR. As shown in Fig. 6, the CD spectra of the two proteins are almost identical, suggesting that the substitution did not cause a major change in the secondary structure of the protein. There is a small decrease in the ratio of the band at 208 nm versus the band at 220 nm, which may be attributed to the fact that the WT can form a dimer, whereas under these conditions I151A does not. Thus, taken together, the in vitro data suggest that the in vivo phenotypes of FNR-M147A, FNR-I151A, and FNR-I158A are most easily explained by specific defects in dimerization.
The in Vivo Defect of Many FNR Ala Substitution Mutants Is Corrected by Increasing Their Cellular Abundance-Because the in vitro data indicated that dimerization of FNR is reduced but not abolished by some Ala substitutions of the dimerization domain, we tested whether higher cellular concentrations of the mutant proteins would increase the amount of FNR activity in vivo. Western blot analysis showed that at least 20-fold higher protein levels were achieved by expressing mutant and WT-FNR from a multicopy plasmid (pACYC184) under control of the normal fnr promoter region than by expressing the proteins from pET-11a (data not shown). The amount of ␤-galactosidase activity measured from anaerobically grown strains containing lacZ under control of the FNR-dependent narG promoter (P narG ), with FNR supplied from a pACYC184 plasmid, showed that, at these higher protein levels, most of the FNR mutants (R140A, M143A, R145A, L146A, M147A, and I158A) activated transcription at a level similar to WT-FNR (Fig. 7a). Such a result is most easily explained if the overall effect of the Ala substitutions is to decrease a FNR monomer-dimer equilibrium. Then, by mass action, producing FNR mutant proteins at a higher level increases the amount of dimeric FNR protein to levels sufficient to fully occupy the promoter.
However, two mutants, FNR-M157A and FNR-I151A, only activated P narG to 30 and 40% of WT-FNR under comparable conditions (Fig. 7a), respectively, suggesting that the in vivo defect cannot be totally overcome by these increased protein levels. In the case of FNR-I151A, we found that this protein was also defective in repression of P ndh (Fig. 7b) at these FNR levels indicating that the in vivo phenotype reflected a decrease in DNA binding (20), an outcome expected for a dimerization defect. The partial restoration of FNR-I151A activity indicates that some dimeric FNR-I151A protein must form but not at levels sufficient to saturate the FNR sites at the respective promoters. This result is consistent with the large defect in dimerization observed by size exclusion chromatography implying that the monomer-dimer equilibrium is perturbed the most in FNR-I151A. In contrast, the properties of FNR-M157A could be most easily explained by a defect in transcription activation because FNR-M157A repressed P ndh as well as WT-FNR and the other mutants that fully activated P narG (Fig. 7b), indicating that the ability of FNR-M157A to dimerize and bind DNA is not defective. This conclusion is supported by the fact that size exclusion analysis of the isolated FNR-M157A protein showed that this mutant dimerized as well as WT-FNR at an onput concentration of 5 M (data not shown).
An Inhibitory Role for Asp-154 at the Dimerization Helix Interface-If FNR dimerizes via a coiled coil interaction (see "Discussion"), then Asp-154 is predicted to be on the same side of an ␣-helix as many of the hydrophobic residues that we found to be important for FNR activity (Fig. 2b). However, negatively charged residues like aspartate are rarely found at a coiled coil subunit interface because the carboxylate group would make the interaction energetically unfavorable. Thus, it was possible that Asp-154 may actually weaken interactions between the hydrophobic side chains within the dimerization interface of FNR. To test this hypothesis, we determined whether removing the negative charge at 154 would improve the activity of FNR proteins that had a defect in dimerization. Indeed, we found that replacement of Asp-154 for Ala in FNR-R140A, FNR-M143A, FNR-M144A, FNR-M147A, and FNR-I158A increased ␤-galactosidase activity under anaerobic conditions to greater than 70% of WT FNR levels (Table II) (Table II). This lack of recovery is similar to what was observed when the D154A substitution was assayed in combination with substitutions that cause defects in either DNA binding (3) or transcription activation (21). In addition, the Ala substitutions that did not cause large decreases in FNR activity showed only slight increases in activity when the D154A substitution was included. Therefore, removal of the carboxyl group at position 154 preferentially improves the activity of the mutants containing Ala substitutions proposed to be at the dimer interface. A simple interpretation of this observation is that Asp-154 is also present at the dimerization interface in WT FNR but has an inhibitory role. DISCUSSION Although dimerization of the transcription factor FNR is known to be regulated by the presence of a [4Fe-4S] cluster (13), we know very little about the protein-protein interactions that promote dimerization. To begin to address this issue, we analyzed the in vivo and in vitro properties of FNR mutants containing alanine substitutions along a putative dimerization helix. Assuming residues 140 -159 of FNR form an ␣-helix like its counterpart in CRP, our simplest interpretation of the properties of these mutants is that this region promotes subunit interactions by forming a largely hydrophobic dimerization interface. A small patch of residues located on the opposite face of this putative ␣-helix was also found to be important for FNR activity and may indicate a second function for this ␣-helix. In addition, our finding that Asp-154 appears to have an inhibitory role in dimerization may provide insight into understanding how dimerization of FNR is regulated by the presence of an oxygen-labile [4Fe-4S] cluster.
A Hydrophobic Surface Characteristic of Coiled Coils Is Required for FNR Dimerization-Many of the hydrophobic residues found to be required for FNR function are located three to four residues apart, suggesting that they fall on the same side of an ␣-helix. Because the requirement for a hydrophobic surface is a property common to ␣-helices that dimerize as a coiled coil, we modeled this 20-amino acid region of FNR as a repeating 7-residue helical unit characteristic of ␣-helices that form coiled coils (22,23). Many of the important hydrophobic resi-  a Experimental conditions are the same as in Fig. 2a. The results are normalized for comparison with WT-FNR supplied on pET-11a. Those mutant proteins in which the single substitution has less than 50% the activity of WT-FNR (column 1) are shown in bold. dues can be fit to the first (Met-144, Ile-151, Ile-158) and fourth (Met-147) position, referred to as a and d, respectively, of the 7-residue helical unit (Fig. 2b). The a and d positions are typically hydrophobic because they provide the energetic driving force for stabilizing the dimer interface (24 -26). Therefore, we propose that Met-144, Met-147, Ile-151, and Ile-158 form a hydrophobic interface that promotes FNR dimerization.
Charged Residues Within the Dimerization Interface-The proposed alignment of the ␣-helices would also place two charged residues, Arg-140 and Asp-154, within the dimerization interface. The positive charge of Arg-140 is important for FNR function at this position because, in addition to R140A decreasing FNR activity, replacing Arg-140 with either the neutral Leu or negatively charged Glu residue results in an FNR mutant with little anaerobic activity. 2 The finding of a positively charged arginine residue at the subunit interface is not uncommon, particularly when it is at the beginning of an ␣-helix (27,28) as expected for FNR. Therefore, this side chain may be making an important contribution to dimerization through the formation of interhelical hydrogen bonds as has been predicted with other proteins (27).
Unlike many well studied coiled coils, our data suggest that a negatively charged residue, Asp-154, is at the dimerization interface and likely has an inhibitory role in dimerization. This inhibitory role of Asp-154 may explain why Ile-151 is particularly critical for dimerization. In a coiled coil model, the side chain of Ile-151 is on the same face of the ␣-helix and in closest proximity to Asp-154. This position makes it possible that Ile-151 functions to shield the negative charge of Asp-154 away from the interface by sterically hindering certain conformations of its acidic side chain (Fig. 8a). Therefore, we propose that when the hydrophobic side chain of Ile-151 is removed, dimerization is impaired not only because of the loss of the hydrophobic packing energy but also because the negative charges of Asp-154 in opposing subunits interact repulsively (Fig. 8b). Consistent with this notion, we found that the defect caused by I151A is partially corrected when Asp-154 was replaced with Ala (Fig. 8c).
Thus, when Asp-154 is replaced by Ala, Ile-151 may no longer need to function as a shield for the negative charge. This proposal may explain why replacing Asp for Ala preferentially increases the activity of mutant proteins that show defects in dimerization and have substitutions along the proposed interface. It is possible that when Asp-154 is replaced with Ala, Ile-151 may have a larger effect on hydrophobic packing, increasing the stability of the coiled coil and making dimerization more favorable (29). Although one might expect that replacement of Asp-154 with Ala would also increase the activity of an otherwise WT FNR protein under anaerobic conditions, previous results (3) show that the promoter used in our studies appears to be at its maximum for transcription activation. Therefore, other experimental approaches are necessary to study the effects of substitutions that may increase dimerization under anaerobic conditions.
Residues Located at the e and g Positions of the Coiled Coil Model Do Not Play a Major Role in FNR Dimerization-The coiled coil model places the residues, Gln-141, Met-143, Ser-148, Glu-150, Gln-155, and Met-157, at the e or g positions of an ␣-helix (Fig. 2b) which in a typical coiled coil further stabilize the protein-protein interactions by forming interhelical hydrogen bonds and salt bridges (26). In the majority of the cases we did not observe significant defects when replacing the e and g residues with Ala. The only exception was with the Ala substitution at Met-143, a residue that is predicted to be in a g position. Surprisingly, this mutant protein has much less activity than the mutant with the Ala substitution at the neighboring a residue, Met-144. In vitro dimerization studies show that the FNR-M143A protein does not dimerize as well as WT-FNR (data not shown). Additionally, the presence of a hydrophobic residue in this position is highly conserved in the FNR homologs (Fig. 9). Whereas the data suggest that Met-143 has an important function, its exact role remains unclear. Nevertheless, the overall results are consistent with the observation that the e and g positions are usually not as important as the hydrophobic residues in the a and d position (30). However, these results may also suggest that the coiled coil of FNR may not be as energetically stable as other previously studied proteins, a desirable property for a regulated system.
How   9. Alignment of the dimerization helix of FNR with several homologous proteins. FNR and CRP are from E. coli; FnrA is from Pseudomonas stutzeri; HlyX is from Actinobacillus pleuropneumoniae; FnrL is from Rhodobacter sphaeroides; FnrN is from Rhizobium leguminosarum; FixK is from Rhizobium meliloti). The sequences are aligned as reported previously (34,36), and the boxes represent the residues that align at the designated a and d positions.
was suggested that the presence of the [4Fe-4S] cluster within each subunit could produce a conformational change in the protein, which then favors dimerization (4 Whereas these data support a mechanism by which the [4Fe-4S] cluster produces a structural change within the protein that enhances its ability to dimerize, the pathway of this conformational change remains to be elucidated. Nevertheless, the presence of the negatively charged Asp-154 within the dimerization interface makes it one reasonable candidate for a residue whose position may need to change to increase dimerization when the protein contains an [4Fe-4S] cluster. Leu-146 and Arg-145 are two other possible candidates because of all the mutant substitutions that were found to decrease FNR activity, they are located on the opposite face of the helix, a position of expected closer physical proximity to Cys-122, a ligand of the [4Fe-4S] cluster. Therefore, Leu-146 and Arg-145 may be in a unique position to transduce the presence of the [4Fe-4S] cluster to the dimerization helix, in particular, to Asp-154. It is interesting to note that a substitution that increased FNR activity, Q142A, is also located on this face of the helix. However, we were unable to assess whether Gln-142 has a direct effect on dimerization or rather acts at a later step in transcription activation. Clearly, additional studies are necessary to define the role of all of the critical residues from 140 to 159, particularly those that may function in facilitating a conformational change in FNR. A key role of Asp-154 in the control of FNR dimerization is also supported by our previous observations that FNR-D154A dimerizes weakly and activates transcription under aerobic conditions when the [4Fe-4S] cluster is absent (3,8,12). The fact that this small increase in dimerization can occur in the absence of the [4Fe-4S] cluster suggests that Asp-154 plays an important regulatory role in dimerization. From these observations, it seems likely that at least one role of Asp-154 is to prevent dimerization of FNR under aerobic conditions, when it lacks a [4Fe-4S] cluster. This is an important area for future investigation.
Do Other FNR/CRP Family Members Contain a Similar Dimerization Domain?-Comparison of this putative dimerization helix of FNR to that of other FNR/CRP family members suggests that many of the hydrophobic residues within the proposed dimerization interface are conserved. A BLAST alignment of the 55 closest FNR homologs reveals that positions 151 and 158 are large aliphatic side chains in ϳ90% of the homologs, whereas position 147 is hydrophobic in 75% of the homologs (a representative sample of homologs is shown in Fig.  9). This indicates that there may be a common interface for the homologs, and it suggests that positions 151 and 158 may be especially important for dimerization. Interestingly, the Asp at position 154 is conserved only in the subset of FNR family members that has the same spacing of the Fe-S cluster ligands in the N-terminal region (Cys-X 2 -Cys-X 5 -Cys) as the E. coli protein (34). This conservation suggests that dimerization is coupled to the presence of a [4Fe-4S] cluster in these other orthologs. In contrast, other family members such as FixK (which lacks the N-terminal Cys ligands) and FnrN and FnrL, which have a Cys-X 2 -Cys-X 7 -Cys Fe-S cluster binding motif, have an Ala at this position (Fig. 9). Further experiments are necessary to determine whether dimerization is regulated in these other FNR family members.
In summary, our data defined residues 140 -159 as an important region for FNR dimerization. In addition, the requirement for many hydrophobic residues within this domain can be most easily explained if this region of FNR forms a coiled coil like CRP. Nevertheless, our data suggest that this proposed coiled coil is far from the most energetically stable coiled coils that are observed with other proteins. This feature may be the key property for allowing regulation of FNR dimerization in response to oxygen availability.