The SH3-like Domain Switches Its Interaction Partners to Modulate the Repression Activity of Mycobacterial Iron-dependent Transcription Regulator in Response to Metal Ion Fluctuations*

Iron-dependent regulator (IdeR), a metal ion-activated pleiotropic transcription factor, plays a critical role in maintaining the intracellular iron homeostasis in Mycobacteria, which is important for the normal growth of the cells. This study was initially performed in an attempt to elucidate all potential interactions between the various domains of IdeR that occur in living mycobacterial cells. This led to a hitherto unidentified self-association for the SH3-like domain of IdeR. Further studies demonstrate that the SH3-like domain interacts with different partners in the dimeric forms of IdeR depending on the levels of metal ions in the environment: it undergoes inter-subunit self-association in the metal-free DNA-non-binding form, but interacts with the N-terminal domain in the metal-bound DNA-binding form in an intra-subunit manner to finely modulate the transcription repression activity of IdeR. Our more detailed mapping studies reveal that the SH3-like domain uses an overlapping surface to participate in these two interactions, which therefore occur in a mutually exclusive fashion. This novel mechanism would allow an effective and cooperative interconversion between the two functional forms of IdeR. Our data also demonstrate that a disturbance of the interactions involving the SH3-like domain impairs the transcription repression activity of IdeR and delays the growth of mycobacterial cells.

Ferrous ion (Fe 2ϩ ) is an essential cofactor for a large number of proteins, functioning in many vital physiological processes (1). Nevertheless, excessive free Fe 2ϩ is extremely harmful to cells under aerobic conditions by catalyzing the formation of highly toxic hydroxyl free radicals (2). As a result, strict cellular homeostasis of free Fe 2ϩ must be effectively maintained for all living organisms.
In mycobacteria, the iron-dependent regulator (IdeR) 3 has been identified as a pleiotropic transcription regulator that regulates the expression of a number of iron-acquisition genes, thus playing a critical role in maintaining the intracellular iron homeostasis (3)(4)(5)(6). IdeR has also been found to regulate the expression of genes functioning in the storage of iron and resistance to oxidative stress (4,5,7). Deletion of the ideR gene in the pathogenic Mycobacteria tuberculosis was found to be lethal for the strain (7). Similar deletion of the ideR gene in Mycobacteria smegmatis, a non-virulent and fast growing mycobacterium, engendered the cells defective in repressing the expression of the iron-acquisition genes even in the presence of excessive iron (4).
Homologues of IdeR have also been found in a variety of other Gram-positive bacteria (8 -12), with the best characterized being DtxR from Corynebacterium diphtheriae (12), which is highly similar to IdeR in structure and function (3). The binding of Fe 2ϩ to IdeR/DtxR transforms it from a metal-free DNA-non-binding form to a DNA-binding form that is able to specifically bind to a conserved "iron box" operator DNA sequence (5,13). Besides Fe 2ϩ , other transition metal ions such as Co 2ϩ , Ni 2ϩ , Zn 2ϩ , Cd 2ϩ , and Mn 2ϩ are also able to activate IdeR/DtxR to its DNA-binding form under in vitro conditions (13,14).
Data of crystal structure determination revealed that the DNA-binding form of IdeR/DtxR exists as homo-dimers with each monomer consisting of an N-terminal domain and a C-terminal domain connected via a highly flexible linker (15)(16)(17)(18). The highly conserved N-terminal domain with two metal binding sites (15)(16)(17)(18)19) has been considered responsible for the dimerization and DNA-binding processes and can be further divided into two corresponding sub-domains. The C-ter-* This work was supported by grants from the National Key Basic Research Foundation of China (Grants 2006CB806508 and 2006CB910300) and National Science Foundation of China (Grants 30570355 and 30670022). 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 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1 and S2 and Tables S1 and S2. 1  minal domain was also designated as the SH3-like domain, in view of its similar folding pattern but hardly any sequence similarity to that of the commonly existing SH3 domains of eukaryotic proteins (17,18,20). In contrast to that of the DNA-binding form, the structural information of the DNA-non-binding form of IdeR/DtxR is still very limited. Data from NMR studies revealed that the N-terminal domain of DtxR undergoes a significant conformation change during the transformation from its DNA-binding to DNA-non-binding form, changing from a well organized structure to a partially disordered "molten globule" state (21). It has been generally believed that the IdeR/DtxR dimers dissociate into monomers during this structural transformation. Nevertheless, dimers were also occasionally observed for the DNAnon-binding form of IdeR/DtxR (13,(21)(22)(23)(24), which was interpreted as such that the dimerization sub-domain in the molten globule state somehow is still able to weakly mediate the dimerization process (13,24). Whether regions other than the dimerization sub-domain also contribute to this dimerization process is a question that has not been addressed.
The SH3-like domain has been known to play a role in forming the DNA-binding form of IdeR/DtxR, mainly based on the determination of its crystal structure (17,18). Although hardly visible in the initially reported crystal structures of the DNAbinding forms of IdeR/DtxR (15,25), the SH3-like domain was later found to interact with the N-terminal domain, contributing to metal ion binding in the better refined structures (16 -18). In such structures, the orientation of the SH3-like domain was found to be slightly different in each of the IdeR/DtxR monomers, suggesting a dynamic nature of the interaction between the SH3-like and N-terminal domains, the meaning of which is hardly known.
In an attempt to examine whether the domain-domain interactions in IdeR/DtxR revealed by these in vitro studies (17,18,20) indeed occur in vivo, we unexpectedly revealed the hitherto unidentified self-association of the SH3-like domain of IdeR. Further in vitro studies demonstrate that this novel interaction largely occurs in the DNA-non-binding form of IdeR in the absence of metal binding, using a surface that overlaps with the one used by the SH3-like domain to interact with the N-terminal domain. The physiological significance of these interactions involving the SH3-like domain was also demonstrated in both M. smegmatis and M. tuberculosis.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Culture Conditions-All strains and plasmids used in this study are listed in supplemental Table S1. Escherichia coli DH5␣, Top10, and BL21(DE3) strains were cultured routinely in Luria-Bertani (LB) broth at 37°C. M. smegmatis ATCC607 was grown in Middlebrook 7H9 liquid media or on Middlebrook 7H11 agar plates (supplemented with 0.5% glycerol, 0.5% Tween 80, and the indicated antibiotics). M63 minimal media (supplemented with 0.5% glycerol, 1 mM MgSO 4 , and the indicated antibiotics) were used in survival assays for protein-protein interaction detection. M. tuberculosis H37Rv was grown in Middlebrook 7H9 liquid media or on Middlebrook 7H10 agar plates (supplemented with 10% oleic acid/albumin/dextrose/catalase, 0.5% glycerol, and 0.5% Tween 80 and the indicated antibiotics). For colony-forming unit counting, M. tuberculosis H37Rv cells were cultured on Löwenstein-Jensen medium. Concentrations for chemicals and/or antibiotics added in the culture media were as follows: ampicillin (Amp) at 100 g/ml, kanamycin (Kan) at 50 g/ml (for the E. coli) or 25 g/ml (for M. smegmatis), hygromycin (Hyg) at 200 -250 g/ml (for the Escherichia coli) or 50 g/ml (for Mycobacteria), trimethoprim (TMP) at 10 g/ml, anhydrotetracycline (Atc) at 50 ng/ml. The M. smegmatis strains were routinely cultured at 37°C except for protein-protein interaction detection, where the culture temperature was shifted to 30°C. M. tuberculosis H37Rv strains were routinely cultured at 37°C.
Construction of the Plasmids Used for Detecting Protein-Protein Interactions in Mycobacteria-In our developed mycobacterial mDHFR-protein fragment complementation assay system, the two fusion proteins were expressed from two compatible E. coli-Mycobacteria shuttle vectors, designated as pLC TetR 1/2 and pLC3, respectively (see Fig. 1A). pLC TetR 1/2 is an extra-chromosomal plasmid with four to eight copies in each mycobacteria cell; however, pLC3 is an integrative plasmid with only a single copy in each mycobacterial cell.
The pLC TetR 1/2 vector was constructed as follows. The DNA fragment containing P 1/tetO sequence was prepared as described (26), and then was digested with HindIII and EcoRV before being subcloned into the pAL5,000-originated E. colimycobacteria shuttle vector pSUM40 (kindly provided by Dr. Ainsa (27)). Subsequently inserted into this plasmid was an EcoRV and KpnI double-digested DNA fragment encoding yeast GCN4 protein, a (Gly-Gly-Gly-Gly-Ser) 2 flexible polylinker and the mDHFR F [1,2] fusion protein, as amplified from the plasmid pBluscript-mDHFR-F [1,2] (kindly provided by Prof. Michnick (28)). The region encoding GCN4 in this new plasmid was removed by digesting with EcoRV and ClaI and then replaced by an oligonucleotide fragment encoding the FLAG tag and a multiple cloning site (see Fig. 1A for its detailed sequence). The constructed plasmid was named as pDF myc 1/2. Finally, a DNA fragment encoding the tetR gene of the E. coli Tn10 transposon (amplified from the genome of E. coli XL1blue and the intrinsic HindIII restriction site of the tetR gene was synonymously mutated via overlapping PCR) was fused to the downstream of a DNA fragment containing the Hsp60 promoter of M. bovis bacille Calmette-Guérin (amplified from the genome of M. bovis bacille Calmette-Guérin) via overlapping PCR, before being digested with HindIII and inserted into pDF myc 1/2. The plasmid with transcriptions controlled by P Hsp60 and P 1/tetO promoters in opposite directions was selected and designated as pLC TetR 1/2.
The pLC3 vector was constructed as follows. The first step of the construction was with virtually the same procedure as that for generating pDF myc 1/2 except that the DNA fragment encoding yeast GCN4 protein, a (Gly-Gly-Gly-Gly-Ser) 2 flexible polylinker and the mDHFR-F [3] fusion protein was amplified from the plasmid pBluescript-mDHFR-F [3] (kindly provided by Dr. Michnick (28)). The new constructed plasmid was thus designated as pDF myc 3. The DNA fragment encoding test protein (designated as X here) was then inserted into the multiple cloning site (all were inserted between NotI and BamHI in this study) of pDF myc 3. To avoid potential incompatibility between the pLC tetR 1/2 and pLC myc 3, the whole expression cassette of DHFR-F [3] fusion proteins was unloaded by cleaving with HindIII and KpnI and then inserted into the pMV306 (HygR) vector (kindly provided by Dr. Clifton Barry, National Institutes of Health) to generate pLC3-X. The physical maps of pLC TetR 1/2 and pLC3 can be obtained upon request.
Detection of Protein-Protein Interactions in M. smegmatis-pLC TetR 1/2 and pLC3 carrying the DNA-fragment encoding the indicated test proteins (each ϳ300 ng) were co-transformed into M. smegmatis-competent cells by electroporation. Cotransformed bacteria were selected on Middlebrook 11 agar media (containing 0.5% glycerol, 50 g/ml Amp, 25 g/ml Kan, and 50 g/ml Hyg) at 37°C. The well separated colonies were then transferred into liquid Middlebrook 7H9 media (supplemented with 0.5% glycerol, 0.5% Tween 80, and the same amount of above three antibiotics) and grown at 37°C to late log phase. The cultured cells were washed twice with M63 medium, diluted to A 600 0.001, before being spotted onto M63 agar plates (supplemented with 0.5% glycerol, the three antibiotics, 10 g/ml TMP, and 50 ng/ml Atc) and cultured at 30°C for ϳ5-7 days. The test was performed at 30°C instead of 37°C to avoid any possible impairment of high temperature on the folding and subsequent interaction of the fusion proteins, as demonstrated by Pelletier et al. (28) when applying mDHFRprotein fragment complementation assay in E. coli.
Site-specific Mutagenesis, Expression, and Purification of Recombinant Proteins-DNA encoding full-length IdeR was amplified from the genome of M. smegmatis ATCC607 directly. The DNA fragments encoding all IdeR variants were digested with NcoI and NotI before being subcloned onto pET21bm (a modified pET21b plasmid in which the initial NdeI restriction site was mutated to a NcoI restriction site) and transformed into E. coli BL21(DE3) cells. His tags were fused to the C termini of all expressed proteins for the purposes of affinity purification and Western blot detection. The sequences of encoding genes were all verified by DNA sequencing. The detailed sequences of all primers used in this study can be found in supplemental Table S2.
For protein expression, all E. coli transformants were cultured in LB/Amp media at 37°C to A 600 ϳ0.5, induced with 0.25 mM isopropyl ␤-D-thiogalactopyranoside for 4 h before being harvested by centrifugation. All recombinant proteins were purified according to the following procedure. The cells were re-suspended in a buffer containing 50 mM phosphate, 150 mM NaCl, and 15 mM imidazole, at pH 7.4 (5 mM ␤-mercaptoethanol was also added for the purification of the cysteine-substituted proteins) before being lysed by ultra-sonication. The cell lysates were then clarified by centrifugation, loaded onto a nickel-nitrilotriacetic acid column before the bound IdeR proteins were eluted with 100 -500 mM imidazole. Collected protein samples were then dialyzed against a dialysis buffer containing 50 mM phosphate, 150 mM NaCl, pH 7.4, to remove the imidazole (5 mM dithiothreitol was also added in the dialysis buffer for cysteine-substituted IdeR mutant proteins) and stored at Ϫ80°C before use. Protein concentration was determined with the BCA-based method using bovine serum albumin as standard.
Size-exclusion Chromatography-Analytical size-exclusion chromatography was performed on a Á CTA Purifier system using a pre-packed Superdex 75 10/300 GL column (Amersham Biosciences Biotech). Protein sample (100 l) was loaded and then eluted with buffer containing 50 mM sodium phosphate, 150 mM NaCl, pH 7.4, at a flow rate of 0.3 ml/min. The elution profile was recorded as the value of the light absorption at 220 nm (due to the absence of Trp, Tyr, and Phe residues in the SH3-like domain of IdeR).
Glutaraldehyde Cross-linking-Cross-linking of purified SH3-like domain of IdeR was carried out with glutaraldehyde by the following procedures. The purified proteins (1 mg/ml) were first treated with indicated concentration of glutaraldehyde at room temperature for 10 min, and then quenched on ice with 100 mM Tris-HCl (pH 8.0) for another 10 min. The crosslinked products were then analyzed by 12% SDS-PAGE.
Non-denaturing Pore Gradient PAGE-Non-denaturing pore gradient polyacrylamide gel with a gradient from 6% to 45% was prepared mainly according to the methods described by Ausubel et al. (29). Samples were electrophoresed at 100 V for indicated time at a temperature of 4°C.
Molecular Docking and Structure Analysis-Homo-multimeric docking of the SH3-like domains of IdeR was carried out utilizing the ClusPro program (30 -32), with one monomeric SH3-like domain of IdeR from M. tuberculosis (residues 151-230, PDB code number: 1U8R (17)) as the starting three-dimensional coordinates. The selection of residues for cysteine scanning mutagenesis described in Fig. 3 was based on following two criteria: (i) it should be located at or near the interaction surface of docked oligomers; (ii) C␣-C␣ distances between two same residue on assembled oligomers should be Ͻ10 Å. The analysis of crystal structures and the calculation of the distances between indicated residues were all performed with PyMOL.
Detection of Disulfide Bonds-E. coli BL21(DE3) cells were transformed with plasmid pET21m which carries the DNA fragment encoding the SH3-like domain with the indicated position being substituted with cysteine residue. The transformed cells were induced with 0.5 mM isopropyl ␤-D-thiogalactopyranoside for 4 h (37°C), collected, and re-suspended in phosphate-buffered saline containing 137 mM NaCl, 2.7 mM KCl, 10 mM NaH 2 PO 4 , 1.4 mM K 2 HPO 4 , and 1 mM phenylmethylsulfonyl fluoride, pH 7.4, before being lysed by ultrasonication and clarified by centrifugation. The protein concentration of the supernatant was determined by BCA method. Formation of disulfide bonds in cysteine-substituted SH3-like domains was induced by ambient oxygen during the ultrasonication treatment. The non-reducing condition was kept by the addition of 5ϫ Laemmli loading buffer (supplemented with 25 mM N-ethylmaleimide and 25 mM EDTA). The reducing condition was maintained by the addition of 50 mM DTT in replaced of N-ethylmaleimide and EDTA into the 5ϫ loading buffer. The samples separated with 12% SDS-PAGE (50 g each lane) were transferred to a polyvinylidene difluoride membrane and detected with monoclonal antibody against His tag. The bands were visualized immunochemically by using alkaline phosphatase-conjugated IgG.
Cysteine-specific Cross-linking-The cysteine-specific crosslinking was performed according to methods previously described (33, 34) with minor modifications. Briefly, protein samples were treated with o-or p-PDM (at 500 M, dissolved in Me 2 SO) and then incubated at room temperature for 10 min (Me 2 SO alone was added for negative control). The reaction was stopped by quenching with 10 mM DTT. The active or inactive forms of the IdeR proteins (of ϳ2 M) were maintained, respectively, by adding 200 M CoCl 2 or 3 mM DP for 10 min at room temperature before cross-linking treatment.
Spin Labeling and EPR Measurements-Spin labeling of the IdeR(C102D/H173C) protein and the following EPR measurement were performed as previously described (35) with minor modification. Briefly, purified IdeR(C102D/H173C) protein (at 5 mg/ml) was first dialyzed against 50 mM phosphate buffer (containing 150 mM NaCl, pH 7.4) at 4°C for 3 h to remove the trace amount of DTT before being labeled with 1 mM maleimide-PROXYL overnight at 4°C. The labeled samples were concentrated to a final concentration of 250 M by ultrafiltration after all free probes were removed by thoroughly dialysis (against with 20 mM Tris-HCl, containing 50 mM NaCl, pH 7.4). The EPR spectra were collected over an 80-Gauss scan width (at room temperature) or over 120-Gauss scan width (at 150 K) using the modulation amplitude of 2.5 Gauss and a microwave power of 8 milliwatts.
EMSA-Binding of IdeR to operator DNA was performed as described by Spiering et al. (23). The 33-bp mbtA operator DNA was prepared by directly annealing two synthesized complemented oligonucleotides (one was labeled with FAM at its 5Ј-end). The binding reaction mixture was incubated at 25°C for 10 min before being subjected to a 12% PAGE in 40 mM Tris acetate (pH 7.5)/2.5% glycerol and electrophoresed in the same buffer without glycerol. 200 M CoCl 2 was added in the reaction mixture, PAGE gel, and running buffer to maintain IdeR proteins in their active forms. Gels were visualized by scanning with Typhoon scanner utilizing exciting wavelength at 488 nm. Oxidized forms of IdeR mutants, in which certain kinds of interactions were immobilized via disulfide bonds (Fig. 5), were prepared by dialyzing the samples under the ambient oxidizing condition for 48 h (at 4°C). Reduction of disulfide bonds was achieved by addition of 5 mM DTT into oxidized samples and then incubated on ice for 30 min.
In Vitro Transcription/Translation Assay-The reporter plasmid, pPT-Luc, used in in vitro transcription/translation reactions, was constructed as follows. The region between the BglII and XbaI sites on plasmid pET28a was replaced by a DNA sequence containing the hybrid promoter tacP/toxO (36), to generate plasmid pPT. A DNA fragment encoding luciferase was then obtained from plasmid pGL3-Basic by digesting with NcoI and BamHI before being inserted into pPT to obtain pPT-Luc. The detailed physical map of pPT-Luc can be obtained upon request. In vitro transcription and translation reaction was performed as described by Love et al. (36 -38) by using Promega S30 Extract system (Promega Co.). Briefly, pPT-luc was added to 20 l of premix and 5 l of complete amino acid mixture. Indicated amount of IdeR proteins and/or DP were then added, and the reaction volumes were brought up to 50 l with diethypyrocarbonate-treated water. Reactions were incubated at room temperature for 10 min, and then 15 l of S30 extracts was added. Reaction mixtures were incubated at 37°C for 1 h before halted by incubating on ice for 10 min. The activ-ity of luciferase was analyzed with Turner luminometer (Terner 20/20 n ). All reactions were performed in triplicate.
RNA Extraction and Real-time PCR-To analyze the expression level of fxbA in M. smegmatis cells, the bacteria transformed with indicated plasmids were cultured in Middlebrook 7H9 medium (supplemented 25 g/ml kanamycin, 0.5% glycerol, 0.5% Tween 80, and 50 ng/ml Atc) at 37°C until the late log phase. The collected cells were re-suspended into TRNzol and broken via ultrasonication. RNA extraction was performed according to the manufacturer's instructions (Tiangen Co.). A mixture of the reverse primers of fxbA and mysA (with their sequence to be found in supplemental Table S2) was added in each extracted RNA sample to synthesize the cDNAs of the two genes at the same time. mysA, as a housekeeping gene in mycobacteria (5), was used as the internal standard to normalize the expression level of fxbA in each RNA sample. The real-time PCR reactions were performed on DNA-Engine Opticon-continuous Fluorescence detection system (MJ research) by using a SYBR ExScript RT-PCR kit (Takara Co.). Reaction mixtures were subjected to PCR with the following program: 95°C for 5 s and then 60°C for 20 s for 45 cycles.
M. tuberculosis Infection and Pathogenicity Testing-M. tuberculosis H37Rv was transformed with pLC3-IdeR(⌬1-140) or pLC3-GCN4 by electroporation and then selected onto Middlebrook 7H10 agar plates (supplemented with 0.5% glycerol, 10% oleic acid/albumin/dextrose/catalase, and 50 g/ml Hyg) at 37°C for 4 -6 weeks. Well separated colonies were picked into Middlebrook 7H9 liquid media (supplemented with 0.5% glycerol, 0.5% Tween 80, 10% oleic acid/albumin/dextrose/ catalase, and 50 g/ml Hyg) and grown at 37°C until the late log phase. Two groups of BALB/c mice (6 -8 weeks old) were then infected via tail vein injection with 1 ϫ 10 7 of the cultured bacteria. Mice (n ϭ 3) from each group were sacrificed at the end of 1, 2, and 4 weeks after the initial infection. Lungs and spleens were weighed and homogenized. Bacterial loads in organs were assessed by serial dilution of homogenates on Löwenstein-Jensen media before being cultured at 37°C for 4 -6 weeks.

RESULTS
The SH3-like Domain of IdeR Self-associates in Living Mycobacterial Cells-In an attempt to systematically examine all the possible pairwise interactions occurring between the domains of IdeR, including those observed by crystal structure determination, we developed a system that is able to detect proteinprotein interactions in living mycobacterial cells. This system was designed as follows. On the one hand, the murine dihydrofolate reductase (mDHFR) is rationally split into two fragments (designated as F [1,2] and F [3]), which are unable to interact with each other by themselves to restore the enzymatic activity unless both are fused to two interacting proteins. On the other hand, the restored enzymatic activity of mDHFR would be essential for the growth of the M. smegmatis cells under a selection condition, where the endogenous DHFR is specifically suppressed by TMP, an inhibitor known to be highly effective toward prokaryotic DHFR but far less effective toward the eukaryotic ones such as mDHFR (28).
The expression of both fusion proteins from the pair of plasmids, designated as pLC TetR 1/2 and pLC3 (schematically presented in Fig. 1A), is directed by the P 1/tetO hybrid promoter (26), toward which the TetR protein acts as a repressor. The TetR protein, being constitutively expressed from the pLC TetR 1/2 plasmid, will be inactivated by binding with its specific inactivator Atc (26) and thus allowing the expression of the fusion proteins to be initiated. The feasibility of this designed system was demonstrated by fusing yeast GCN4, which is known to self-associate (28), with the two mDHFR fragments: the M. smegmatis cells co-transformed with the pair of plasmids expressing GCN4-F [1,2] and GCN4-F[3] grew on a selection medium containing both the inhibitor of prokaryotic DHFR (TMP) and the inactivator of TetR repressor (Atc), but not on that containing the former but not the latter (Fig. 1B). It should be noted that our protein-protein interaction system described above is similar to what was recently reported by Singh et al. (40) in overall strategy, but different in details. For instance, the expression of the fusion proteins will be initiated only by induction in our system, but being constitutive in theirs.
Results of analysis of pairwise interactions of various forms of IdeR (including the full-length and the indicated truncation forms) in living M. smegmatis cells are displayed in Fig. 2. Consistent with previous observations of in vitro studies (15,17,18,41), the full-length IdeR homodimerized in living M. smegmatis, as indicated by the growth of cells co-expressing IdeR-F [1,2] and IdeR-F[3] on the selection medium (spot bB in Fig. 2B); the N-terminal domain alone, containing the dimerization sub-domain characterized earlier, also underwent homodimerization in the cells (spot dD in Fig. 2B).
Totally unhinted from previous studies are the following observations. On the one hand, a truncation form of IdeR lacking the dimerization sub-domain so far defined not only selfassociated (spot cC in Fig. 2B) but also interacted with the full-

interactions in living mycobacterial cells (A) and a feasibility test of the system (B).
A, schematic representation of the two plasmids, pLC TetR 1/2 and pLC3, carrying the DNA sequences encoding F [1,2] and F [3] fragments of mDHFR, respectively. Linker, DNA encoding the flexible (Gly 4 Ser) 2 fragment; MCS, a multiple cloning site; P 1/tetO , a hybrid promoter having two tandem tetO sequences (solid diamonds) integrated into the strong mycobacterial promoter P 1 ; P hsp60 , a strong mycobacterial promoter derived from the M. bovis bacille Calmette-Gué rin groEL2 gene; tetR, DNA sequence encoding the E. coli Tn10 Tet repressor (TetR) protein; oriM and oriE, the replication origins derived from mycobacteria and E. coli, respectively; AttP, the phage attachment site that allows the integration of pLC3 plasmid into the genome of mycobacteria; int, DNA sequence encoding the integrase of the mycobacteriophage L5. The detailed sequences of the FLAG tag and the multiple cloning site are shown at the bottom. B, the M. smegmatis cells were co-transformed with pLC TetR 1/2-GCN4 and pLC3-GCN4 (the three spots at the top) or with pLC TetR 1/2-GCN4 and pLC3-IdeR (the three spots at the bottom) before growing in the indicated selection media. Atc will bind to and inactivate the transcription repression activity of TetR; TMP will inhibit the enzymatic activity of prokaryotic DHFR. The cells were grown on agar plates made of M63/glycerol minimal media, containing 100 g/ml Amp, 25 g/ml Kan, and 50 g/ml Hyg. TMP and Atc were added to a final concentration of 10 g/ml and 50 ng/ml, respectively. smegmatis cells co-expressing the indicated pairs of fusion proteins on the selection media. Proteins or protein fragments expressed from the pLC TetR 1/2 plasmid (being fused to mDHFR-F [1,2]) are indicated at the top, and those expressed from the pLC3 plasmid (being fused to mDHFR-F [3]) are indicated on the left. The selection media contained TMP (10 g/ml), Atc (50 ng/ml), Amp (100 g/ml), Kan (25 g/ml), and Hyg (50 g/ml), all added in the M63/glycerol minimal agar media. The yeast GCN4 protein was used here for both positive and negative controls, for its ability to self-associate (spot aA) but inability to interact with any form of IdeR (spots aB, aC, aD, aE, bA, cA, dA, and eA).
length IdeR in the living cells (spots bC and cB in Fig. 2B). On the other hand, a truncation form of IdeR solely containing the SH3-like domain also self-associated in mycobacterial cells (spot eE in Fig. 2B). These observations taken together strongly suggest that the SH3-like domain represents one additional region, other than the previously characterized dimerization sub-domain, to be also able to mediate dimerization for IdeR in mycobacterial cells.
Interaction between the SH3-like domain and the N-terminal domain, despite its observation in the resolved crystal structures of full-length IdeR (17,18), was not detected using our in vivo assay, where the two domains were separately expressed and fused to the mDHFR fragments F [1,2] and F [3], as indicated by the failure of growth of mycobacterial cells co-expressing these proteins (spots dE and eD in Fig. 2B). This result apparently helps to resolve an unanswered question of crystal structure studies, where it was not known whether the interacting SH3-like domain and the N-terminal domain come from the same or different subunits in the dimeric forms of IdeR, because of the invisibility of the highly flexible linker region between these two domains (17,18). The failure of detecting interactions between these two domains when expressed separately implicates that coming from the same subunit and being connected together in one polypeptide chain are essential for the interaction of these two domains to occur (data presented below in Fig.  6 further demonstrate that the cleavage of the linker region almost completely abolished the interaction between the SH3like and N-terminal domains).
The Self-association of the SH3-like Domain and Its Interaction with the N-terminal Domain of IdeR Are Mediated by an Overlapping Surface-The capacity for the separately expressed SH3-like domain of IdeR to undergo self-association under in vitro conditions was then examined via both size-exclusion chromatography (Fig. 3A) and chemical cross-linking (Fig. 3B) analyses, revealing the presence of both monomeric and dimeric forms. This somehow suggests a dynamic and/or weak nature of the self-association for the SH3-like domain. This conclusion is apparently supported by results of non-denaturing pore gradient PAGE analysis (29), presented in Fig.  3C, where a dimeric form of the separately expressed SH3-like domain was visible only after a shorter time of electrophoresis but disappeared after a longer time of electrophoresis, as indicated by the gradual decrease in intensity of the upper bands (asterisk-labeled) from lanes 1 to 3. Such results reflect a typical undergoing of dynamic dissociation of oligomeric proteins as explained by us before (42)(43)(44).
To map the surface on the SH3-like domain mediating its self-association, cysteine scanning mutagenesis (45) was performed, for which the residue located at the interaction surface will be identified by its allowing the formation of disulfide bonds after being replaced by a cysteine residue, as reflected by the detection of cross-linked dimers under oxidized conditions (46). For this purpose, each of 21 positions on the SH3-like domain was substituted by a cysteine residue. The substituting positions were chosen as such that they are largely located at the surface of the SH3-like domain according to the determined structure of full-length IdeR (17,18), with some having a predicted location at the self-association surface based on the models generated from homo-oligomeric docking (for more details see "Experimental Procedures").
Results presented in Fig. 3D demonstrate the formation of covalently linked dimers for the SH3-like domain under oxidizing (top panel) but not reducing (bottom panel) conditions when the residue at position 173, 192, 201, 202, 203, 218, 219, or 220 was replaced by a cysteine residue. The mapping of these positions (red colored in Fig. 3E) on the tertiary structure of the SH3-like domain reveals a remarkable clustering of them. Markedly, this surface nicely overlaps with that the SH3-like domain uses to interact with the N-terminal domain in the DNA-binding form of the full-length IdeR (17,18) (also see the right side in Fig. 3E). This apparently explains why the selfassociation of the SH3-like domain was not observed in the reported crystal structure of IdeR, in which the SH3-like domain interacts with the N-terminal domain (17,18).
The SH3-like Domain Self-associates in the Metal-free DNAnon-binding Form of Full-length Dimeric IdeR-It was then examined whether the SH3-like domain in the full-length IdeR protein indeed undergoes self-association using the surface characterized above. Given that the SH3-like domain interacts with the N-terminal domain in the DNA-binding form of IdeR (17,18), it is conceivable that the self-association of the SH3like domain, if it does happen in the full-length IdeR, might occur in its metal-free DNA-non-binding form, in which the N-terminal domain is believed to exist as a partially disordered molten globule state (21). As a result, the interaction between the SH3-like domain and the N-terminal domain would be unlikely to occur, and thus free the surface for the SH3-like domain to self-associate.
To test such a hypothesis, the cysteine-specific cross-linking method (33,34) was applied to analyze the IdeR(C102D/ H173C) mutant, where replacement of the single intrinsic Cys-102 by aspartate would only allow the specific cross-linking to occur via the cysteine introduced at position 173, which was above revealed to be located at the self-association surface for the separately expressed SH3-like domain (see Fig. 3, D and E). The cysteine to aspartate substitution at position 102 has been constructed in a few previous reports, with the mutant IdeR/ DtxR protein retaining much of its activity in metal-dependent DNA binding and transcription repression (14,21,25,(47)(48). Our own data of in vitro transcription/translation analysis (pre- sented below in Fig. 7C) also demonstrate that IdeR(C102D) indeed retains its metal-dependent repression activity against specific operator DNA sequences. Similarly, the IdeR(C102D/ H173C) protein was also shown to exhibit a metal ion dependent binding activity toward the specific operator DNA fragment (see Fig. 5A below).
Results presented in Fig. 4A demonstrate that the self-association of the SH3-like domain in the full-length IdeR indeed occurs but only in its metal free DNA-non-binding form. This is indicated by the detection of a cross-linked dimeric form of IdeR(C102D/H173C) predominantly when all free metals were chelated by the addition of 2,2Ј-dipyridyl (DP) but hardly any in the presence of Co 2ϩ (compare lanes 8 and 7 in Fig. 4A) or when IdeR(C102D) was subject to the same analysis (lanes 2-4 in Fig. 4A). It should be noted that a second cross-linked product (star-labeled in Fig. 4A), present in far less abundance and migrated at a slower rate than the dimeric form described above, is occasionally observed for both IdeR(C102D) and IdeR(C102D/H173C) (see lanes 3 and 7, Fig.  4A). These second cross-linked products are likely the dimeric forms that are cross-linked via the amino groups at the N-terminal or internal lysine residues, in view that the functional group maleimide of cross-linking agent p-PDM used here may also generate cross-linking between amino groups, although in a far less effective manner (49).
The self-association of the SH3-like domain in the DNAnon-binding form of full-length IdeR was further confirmed by detecting the occurrence of spin-spin interaction between a pair of spin probes specifically attached to the cysteine residue of each of the subunits of IdeR(C102D/H173C). Such spin-spin interaction can only be detectable as the two probes are very close to each other (within a distance of 8 -25 Å (35, 50)). For this purpose, Cys-173 in IdeR(C102D/H173C) was labeled with spin probe PROXYL, and the labeled protein was then subject to EPR measurement. The EPR spectrum of the frozen PROXYL-labeled IdeR(C102D/H173C) protein in the presence of metal chelator (DP), presented in Fig. 4B, reveals a strong spin-spin interaction effect, as indicated by having a d1 to d ratio (d1/d) of 0.58 (a value of d1/d larger than 0.4 is commonly taken to indicate a spin-spin interaction between two probes (35,50)).
It was subsequently examined whether the lack of association between the SH3-like domains of the two subunits in the metalbound DNA-binding form (as indicated by results presented in lane 7, Fig. 4A) is really accompanied by the occurrence of interaction between the SH3-like and N-terminal domains of the same subunit. For this purpose, two cysteine residues were introduced into IdeR(C102D), substituting Lys-185 in the SH3like domain and Glu-95 in the N-terminal domain respectively, to generate the IdeR(C102D/E95C/K185C) mutant protein.
These two positions are expected to be proximal to each other in the DNA-binding form of IdeR, as guided by the data of crystal structure determination (17,18).
The cysteine-specific cross-linking produced a form of the IdeR(C102D/E95C/K185C) protein that migrated at a faster rate than the uncross-linked monomeric form and that is most predominant with the addition of Co 2ϩ (compare results presented in lanes 7 and 8, Fig. 4C). The position of the cross-  7). The samples in lanes 6 and 5 represent the purified IdeR(C102D/ H173C) protein with or without being treated with the cross-linking agent (neither Co 2ϩ nor DP was added in both samples). Analyzed in lanes 1-4 is the cysteine-free IdeR(C102D) protein after being subjected to a parallel treatment. Position of the dimeric form of IdeR(C102D/H173C) resulted from the cross-linking reaction, as indicated on the right. The position of a second minute amount of cross-linked product, which was most likely resulted from the amino-targeted cross-linking, is indicated with an asterisk on the right. B, EPR spectrum of IdeR(C102D/H173C) (at 250 M) that was spin-labeled with PROXYL as recorded at a temperature of 150 K in the presence of DP (at 10 mM). The ratio of 0.58 (being larger than 0.4, a threshold value) for the amplitude of d1 to d is taken to demonstrate the occurrence of spin-spin interaction between the pair of PROXYL groups. The length of the bar shown at the bottom right represents the width of 10 Gauss. Shown in the inset is the EPR spectrum of the same sample recorded at a temperature of 300 K. C, SDS-PAGE analysis of a parallel cross-linking treatment for IdeR(C102D/E95C/ K185C) with o-PDM (at 500 M). Positions of the uncross-linked and intrasubunit cross-linked monomeric forms of IdeR(C102D/E95C/K185C) are indicated on the right. linked product unequivocally indicates intra-subunit nature of the cross-linking. This also demonstrates that the interacting SH3-like and N-terminal domains come from the same chain, as suggested by our results of in vivo protein-protein interaction assays, where the interaction between the separately expressed SH3-like and N-terminal domains was not detected (see spots dE and eD in Fig. 2B). The detection of minute amount of intra-subunit cross-linked product in the presence of the chelator (lane 8 in Fig. 4C) most likely reflects the difficulty of a complete removal of the protein-bound metals by the chelators. The cross-linking agent o-PDM was chosen here because the length of its cross-linking arm (7.7 Å (33)) fits ideally the estimated spatial distance between positions 95 and 185 (7.4 Å) in the DNA-binding form of IdeR by referring to the determined crystal structure (17,18).
Immobilization of the Self-association of the SH3-like Domain Weakens but That of the Interaction between the SH3-like and N-terminal Domains Enhances the DNA-binding Capacity of IdeR-To understand its biological role, the self-association of the SH3-like domain in IdeR was immobilized by disulfide bond formation before examining its capacity to bind to specific mycobacterial mtbA operator (O mtbA (5,18)) DNA fragment by EMSAs, again using the IdeR(C102D/H173C) protein.
The EMSA results, presented in Fig. 5A, clearly demonstrate that the DNA-binding capacity of IdeR(C102D/H173C) was significantly weakened after the self-association of its SH3-like domain was immobilized even in the presence of Co 2ϩ (compare lanes 2 and 3, 4 and  5, 6 and 7, and 8 and 9, respectively, in Fig. 5A). In contrast, a similar immobilization of the intra-subunit interaction between the SH3-like and N-terminal domains significantly strengthened the capacity of IdeR(C102D/E95C/K185C) to bind to the same operator DNA sequence (compare lanes 2 and 3, 4 and 5, 6 and 7, and 8 and 9, respectively, in Fig. 5C). Formation of disulfide bonds in these interaction-immobilized protein samples was confirmed by SDS-PAGE analyses (Fig.  5, B and D).

Cleaving or Lengthening the Linker between the SH3-like Domain and the N-terminal Domain Impairs Their Interaction as Well as the Metal Ionresponsive Transcription Repression
Activity of IdeR-The failure to detect the interaction between the separately expressed SH3-like and N-terminal domains of IdeR under in vivo conditions (Fig. 2B) suggests a relatively weak interaction between them and thus a need for a linker of proper length to connect them together for the intra-subunit interaction to occur in full-length IdeR. Two strategies were taken to test this hypothesis. One is to make a cleavage at the linker region; the other is to lengthen the linker region. The effects of such disturbances on the IdeR proteins were then examined by measuring the DNA-binding activity via EMSA or the transcription repression activity by using the in vitro transcription/translation assay (51).
To make the cleavage, a six-residue thrombin-cleavage site was first inserted into the linker region of IdeR(C102D), generating the IdeR(C102D, thromA) protein (Fig. 6A). The results of EMSA, which is always performed in the presence of Co 2ϩ , presented in Fig. 6C demonstrate that the cleaved products indeed exhibited a significantly decreased DNA-binding activity (compare lanes 2-5 with lanes 7-10) with the DNA probes being shifted by the cleaved products to a position lower than that by the intact proteins (also compare lanes 2-5 with lanes 7-10) but the same position as that by IdeR(C102D)⌬SH3 (compare lanes 7-10 with lanes 12-15). These results strongly suggest that the SH3-like domain no longer interacts with the N-terminal domain in the cleaved products (even in the presence of Co 2ϩ ) and only the N-terminal domain complexes with the operator DNA. Effective thrombin cleavage was confirmed by SDS-PAGE analysis (lanes 3-6 in Fig. 6B).
The transcription repression activity of these cleaved products of IdeR was then examined using the in vitro transcription/translation system (51), being much closer to the real situation inside a living cell and quantitative in nature. In this system, the higher the level of transcription repression activity for IdeR, the lower the level of activity for the luciferase reporter whose expression is directed by a modified promoter of pTac integrated with two tandem IdeR-specific operator O tox sequences (36 -38, 52).
Consistent with the results of EMSA, the cleaved products of IdeR(C102D, thromA) exhibited a transcription repression activity much lower than that of the intact protein (compare columns 1 and 3, Fig. 6D) in such an in vitro transcription/translation system, itself containing a certain level of metal ions (36,38). Additionally, such repression activity of the cleaved products of IdeR was hardly brought to any further decrease when an excess amount of metal chelator (DP) was added (compare columns 3 and 4 in Fig. 6D), indicating an almost complete abolishment of repression activity for IdeR when the intra-subunit interaction between the SH3-like and N-terminal domains was eliminated. It should be noticed that, although the N-terminal domain alone was apparently able to bind to the specific operator DNA sequences (as shown by data presented in Fig. 6C), it was unable to repress the transcription as assayed with the in vitro transcription/translation system. These observations together suggest that interacting with the SH3-like domain is far more important than previously speculated for the N-terminal domain of IdeR to bind to the specific operator DNA sequences and in turn to exhibit the transcription repression activity in cells. In other words, the capacity of binding to the specific operator DNA sequences in vitro (as usually detected by EMSA) for the N-terminal domain of IdeR should not be simply taken as to indicate its capacity to repress transcription in vivo.
It is conceivable that optimal occurrence of the relatively week intra-subunit interaction between the SH3-like and N-terminal domains is highly dependent on the connecting linker of proper length. It follows that extending the length of the linker between the SH3-like and N-terminal domains would weaken the interaction, thus decreasing the DNA-binding and transcription repression activities of IdeR. Our data indeed demonstrate that the transcription repression activity was decreased proportionally, as shown by data presented in Fig.  7C, when the length of the linker was increased by 2-or 4-fold (illustrated in Fig. 7A and confirmed by data shown in Fig. 7B).

Heterologously Overexpressing the SH3-like Domain of IdeR Up-regulates the Expression of IdeR-controlling Gene in M.
smegmatis Cells-The physiological importance of the interactions involving the SH3-like domain of IdeR was then exam-  (lanes 2, 7, and 12), 0.05 nmol (lanes 3, 8, and 13), 0.075 nmol (lanes 4, 9, and 14), and 0.1 nmol (lanes 5, 10, and 15). Here, the cleavage of IdeR(C102D, thromA) was performed by incubating the protein with thrombin at a concentration of 2 units/mg of protein at 37°C for 4 h. D, relative activities of the luciferase expressed from the pPT-luc reporter plasmid (0.5 g in each reaction) in the presence of the various forms of IdeR proteins (each 25 nmol) as measured by using the in vitro transcription/translation system. Metal-free conditions were maintained by adding the metal chelator DP (50 M) (open columns). To exclude the possible influence of DP on the in vitro transcription/translation reaction, the level of the relative luciferase activity was calculated as the percentage of luciferase activity (in the presence of IdeR and 50 M DP) in comparison with the luciferase activity measured in the presence of 50 M DP but without adding IdeR. Each value represents the mean of triplicate independent experiments with standard deviations indicated. The cleaved IdeR(C102D, thromA) was prepared as in C. Existence of the trace amount of thrombin alone hardly affected the expression level of the luciferase via the in vitro transcription/ translation (data not shown). was included here as a control. C, metal ion-responsive repression activities of IdeR(C102D), IdeR(C102D, 2XL), and IdeR(C102D, 4XL) as measured by using the in vitro transcription/translation system against an increased amount of DP (performed as that for Fig. 6D). The concentrations of the IdeR proteins as well as the plasmid pPT-Luc were added at the same concentrations as described in Fig. 6 (D). Level of the luciferase activity was calculated as the percentage of luciferase activity (in the presence of IdeR and each indicated concentration of DP) in comparison with the luciferase activity measured in the presence of the same concentration of DP but without adding IdeR.
ined. For this purpose, the SH3-like domain of IdeR (as a fusion protein with mDHFR-F [1,2]) was heterologously overexpressed in M. smegmatis. Our data of in vivo protein-protein interaction studies (see Fig. 2B) have demonstrated that the separately expressed SH3-like domain predominantly interacts with the SH3-like domain instead of the N-terminal domain of IdeR in living mycobacterial cell. It follows that the heterologously overexpressed SH3-like domain would act in a dominant-negative manner to interact with the SH3like domain of the endogenous IdeR protein in the cells. As a result, the interaction between the SH3-like domain and the N-terminal domain in the endogenous IdeR protein would be prevented (schematically illustrated in Fig. 8), and in turn the formation of the DNA-binding form of IdeR would be impaired, because the SH3-like domain uses an overlapping surface to mutually exclusively interact with its two alternative partners (Fig. 3).
The repression activity of the endogenous IdeR in M. smegmatis cells overexpressing the SH3-like domain was then detected by measuring the transcription level of the fxbA gene via quantitative PCR. This gene encodes an enzyme needed for iron acquisition (53) and is known to be negatively regulated by IdeR (54).
The significant increase (column C, Fig. 8) in the transcription level of the fxbA gene in these M. smegmatis cells, in comparison with the control cells overexpressing an unrelated protein GCN4 or the wild-type IdeR (columns A and B, Fig. 8), demonstrates an impairment of the transcription repression activity of the endogenous IdeR. Serving as a positive control is the data showing the de-repression of the fxbA gene transcription in M. smegmatis cells overexpressing IdeR⌬-(1-74), a truncation form of IdeR with the DNAbinding sub-domain deleted (column D, Fig. 8), and thus its interaction with the endogenous IdeR would form a heterodimer (as demonstrated by results of in vivo assays presented in supplemental Fig. S1), which lost the DNA-binding capacity for exhibiting the transcription repression activity, because one of the two subunits in the hetero-dimer lacks a DNA-binding sub-domain.
A Heterologous Overexpression of the SH3-like Domain of IdeR Delays the Growth of M. tuberculosis Both in Vitro and in Vivo-We subsequently examined the role of the SH3-like domain on the physiology and pathogenicity of M. tuberculosis, also via the dominant-negative strategy described above.
An overexpression of the SH3-like domain in M. tuberculosis (designated as Mtb-SH3) resulted in a significant delay of cell growth in comparison with that overexpressing GCN4 (designated as Mtb-Ctrl) when cultured on solid Middlebrook 7H10 medium (Fig. 9A). This observation somehow suggests that the interactions involving the SH3-like domain of IdeR are important for the normal physiology of M. tuberculosis.
The pathogenicity of the M. tuberculosis cells overexpressing the SH3-like domain was then examined using a mouse infection model. The survival curves, presented in Fig. 9B, showed that the median time to death (i.e. the time when 50% of the mice had died) of the BALB/c mice infected with Mtb-SH3 was significantly longer than that of the mice infected with the control strain Mtb-Ctrl (being 39 and 25.5 days, respectively, t test, p Ͻ 0.05). Correspondingly, the mean weight of the mice infected with Mtb-SH3, measured at the 28th day after the initial infection, was significantly higher than that of the mice infected by Mtb-Ctrl (see the inset in Fig. 9B). These observations suggest that the interactions involving the SH3-like domain are important for M. tuberculosis to exhibit its pathogenicity in mice.
Further examination of the bacterial loads in the infected mice demonstrated a 10-fold reduction of M. tuberculosis cells in the spleens (Fig. 9C) but no significant reduction in the lungs (Fig. 9D) for the mice infected with Mtb-SH3 in comparison with those infected with Mtb-Ctrl, as counted at day 28 after infection. This might be explained as such that the function of IdeR is more important for M. tuberculosis to survive in spleen than in lung as a result of the differences in niches between these two organs (e.g. in levels of reactive oxygen species and free ferrous iron).

DISCUSSION
This study was conducted in an effort to elucidate how the IdeR protein, a pleiotropic transcription regulator that plays a critical role in maintaining the intracellular iron homeostasis in mycobacteria (3)(4)(5)(6), interconverts between its DNA-binding and DNA-non-binding forms in response to the fluctuations of iron levels in the cells. For this purpose, both in vitro and in vivo studies were performed. Our data for the first time revealed that the SH3-like domain is able to undergo self-association to help to maintain the DNA-non-binding form of IdeR also in a dimeric instead of monomeric form under conditions where metal ions are lacking. Our results also demonstrate that the inter-subunit selfassociation of the SH3-like domain in IdeR protein will give way to the intra-subunit interaction between the SH3-like and N-terminal domains, which plays an important role for IdeR to bind to the specific operator sequences and thus repress the transcription of the dozens of regulated genes. Furthermore, these two types of interactions involving the SH3like domain occur via an overlapping surface on the SH3-like domain and thus in a mutually exclusive manner.
In light of our findings on the new roles played by the SH3-like domain, a new working model was proposed (schematically shown in Fig. 10) to explain how IdeR effectively interconverts between its two forms in response to the fluctuations of iron levels in mycobacterial cells. The most appealing feature of this model is that the inter-subunit self-association of the SH3-like domains would prevent its intra-subunit interaction with the N-terminal domain, and vice versa. It follows that the interconversion between the DNA-binding and DNA-non-binding forms of IdeR would occur in a cooperative nature, allowing an effective interconversion and sensitive response toward the physiological fluctuations of the iron levels to strictly maintain the critical cellular iron homeostasis.
Our data reported here (see Figs. 2, 3, 4, 6, and 7), in combination with observations of structure determination studies where the two SH3-like domains were found to be either invisible (15,25) or each interacting with an N-terminal domain in a different orientation in the dimeric DNA-binding forms of IdeR/DtxR (16 -18), strongly suggest that the two alternative interactions involving the SH3-like domain are all relatively weak and occur in a dynamic nature. Nevertheless, the interaction between two SH3-like domains seems stronger than that between the SH3-like and N-terminal domains, with the former but not the latter, which was detected by our in vivo proteinprotein interaction studies (see Fig. 2B). Connection together by a linker region of proper length and high flexibility is needed to increase the strength of the interaction between the SH3-like and N-terminal domains to a level that is comparable to that of interaction between two SH3-like domains, thus to ensure an effective switch of the SH3-like domain from interacting with one partner to another during the interconversion between the  DNA-binding and DNA-non-binding forms of IdeR. In view of these findings, the overlapping surface on the SH3-like domain used for its self-association and its interaction with the N-terminal domain might serve as a promising target for developing anti-tuberculosis drugs.
Earlier models presumed that the SH3-like domain acts to stabilize the DNA-non-binding form of DtxR by interacting with a proline-rich region in its N-terminal domain, mainly based on NMR spectroscopy analysis of the structure of a truncation form of DtxR, in which the whole DNA-binding subdomain and much of the dimerization sub-domain no longer exist (55). Given that the proline-rich region is an internal sequence that is an inseparable part of the N-terminal domain, interaction between the proline-rich region and the SH3-like domain would have to overcome an extraordinary spatial hindrance in the intact IdeR protein. In view of this obstacle, whether such interaction indeed occurs in the full-length DtxR/ IdeR has to be questioned.
In contrast to our observations, self-association was not observed for the separately expressed SH3-like domain of the DtxR protein when examined using non-denaturing PAGE (20). In view of this, the SH3-like domain of DtxR was subject to similar studies as we have performed for its counterpart in IdeR. Results of our chemical cross-linking analysis (see supplemental Fig. S2A) and protein fragment complementation assays (see supplemental Fig. S2B) demonstrate that the SH3-like domain of DtxR is also able to undergo self-association under both our in vivo and in vitro conditions. One likely reason for their failing to detect the occurrence of self-association for the SH3-like domain of DtxR is that the samples applied were purified via a denaturation and renaturation process; such artificial handling might have undermined the successful self-association. It should be pointed out that the protein fragment complementation assay did not reveal any occurrence of interaction between the SH3-like domain of DtxR and that of IdeR in living M. smegmatis (as indicated by the failure of growth of co-transformed cells on spot 7, supplemental Fig. S2B), suggesting such self-association is highly species-specific.
The eukaryotic SH3 domains, which share few similarities in amino acid sequences with the SH3-like domain of IdeR, have been known to interact with a proline-rich motif to mediate protein-protein interactions (56). The properties of the SH3-like domain unveiled here drove us to examine whether there is any literature reporting the occurrence of similar self-association for the SH3 domains of eukaryotic proteins. The effort dug out at least two likely cases, the Isletbrain 1 (a scaffold protein that participates in the organization of the JNK signaling pathway) and Eps8 (a substrate protein of several receptor and non-receptor tyrosine kinases), where the separately expressed SH3 domains were reported to undergo self-association (57,58). Whether the SH3 domains in these two and likely other proteins also play the similar interaction switch (i.e. such self-association probably only occurs under certain unique physiological conditions for the intact proteins) as those revealed for the SH3-like domain in IdeR warrants further investigation.