Evidence for a Monomeric Intermediate in the Reversible Unfolding of F Factor TraM*

F factor TraM is essential for efficient bacterial conjugation, but its molecular function is not clear. Because the physical properties of TraM may provide clues to its role in conjugation, we have characterized the TraM oligomerization equilibrium. We show that the reversible unfolding transition is non-two-state, indicating the presence of at least one intermediate. Analytical ultracentrifugation experiments indicate that the first phase of unfolding involves dissociation of the tetramer into folded monomers, which are subsequently unfolded to the denatured state in the second phase. Furthermore, we show that a C-terminal domain isolated by limited proteolysis is tetrameric in solution, like the full-length protein, and that its loss of structure correlates with dissociation of the TraM tetramer. Unfolding of the individual domains indicates that the N- and C-terminal regions act cooperatively to stabilize the full-length protein. Together, these experiments suggest structural overlap of regions important for oligomerization and DNA binding. We propose that modulating the oligomerization equilibrium of TraM may regulate its essential activity in bacterial conjugation.

Bacterial conjugation is a plasmid-mediated mode of lateral gene transfer. F factor was the first conjugative plasmid described (1). For successful F factor conjugation, the mating signal, which initiates DNA transfer upon formation of a stable mating pair, must be transmitted (2). TraM is a plasmid-encoded protein believed to be involved in mating signal transmission (3)(4)(5). TraM mutants form stable mating pairs and nick the plasmid at oriT (origin of transfer), but DNA transfer does not occur (4). Although the identity of the mating signal is unknown, the signal is mediated by the conjugative pilus and requires cell-cell contact (6). Furthermore, signal transmission requires a step in addition to the interaction of the pilus with the recipient cell (2).
F factor TraM can interact with other F factor proteins and plasmid DNA, and its DNA-binding activity correlates with its in vivo function. TraM binds to three sites in oriT (7)(8)(9)(10). The two highest affinity sites, sbmA and sbmB, overlap with the two traM promoters and negatively autoregulate the expression of TraM (11)(12)(13). Deleting the oriT region that includes sbmA and sbmB from a mobilizable plasmid decreases its transfer efficiency by 100-fold (14). Further deletion of a region including sbmC, the lowest affinity site, from this plasmid reduces the efficiency of mobilization by another 100-fold (14).
In similar experiments with the oriT of F-like plasmid R100, mutations in the site analogous to sbmC inhibited transfer (15). These experiments indicate that TraM binding at sbmC may be required for its essential role in conjugal DNA transfer. In addition to binding DNA, TraM associates with the plasmidencoded inner membrane protein TraD in vitro (16). TraD is believed to be involved in transferring the single-stranded plasmid through the transfer pore to the recipient cell (5). The ability to interact with oriT DNA and a membrane protein suggests that TraM may function to tether the plasmid near the site of DNA transfer.
It is not clear how the putative tether function of TraM is important in transmitting the mating signal. The TraM tether may be formed in response to the mating signal, implying that the ability of TraM to interact with oriT or TraD is modulated by the mating signal. In vitro, TraM binds to DNA as a tetramer (17) and to the three oriT sites cooperatively (9). Therefore, small changes in the cellular concentration of tetrameric TraM in response to the mating signal could alter the occupancy of the lowest affinity binding site, sbmC. It is unlikely that the concentration of TraM in the cell is increased transcriptionally because expression is negatively autoregulated. Furthermore, conjugation efficiency is not affected by inhibitors of RNA or protein synthesis (4), suggesting that transcription and translation are not required for mating signal transmission.
We have performed a thermodynamic characterization of TraM oligomerization and stability, a necessary first step toward understanding how TraM might respond to the mating signal. We report the first reversible equilibrium unfolding of TraM; and based on these and other data, we propose that TraM unfolds via a folded monomeric intermediate. Moreover, we show that the DNA-binding and oligomerization activities of TraM cannot be structurally separated. The thermodynamic characteristics of TraM and the interaction between structures that are important for DNA binding and oligomerization suggest a mechanism for regulating the activity of TraM in vivo.

EXPERIMENTAL PROCEDURES
Materials and Bacterial Strains-Guanidine hydrochloride (GndHCl) 1 (Mallinckrodt Chemical Works) solutions were prepared fresh and filtered (0.45 m). Chromatography supplies and the Gradifrac system were from Amersham Biosciences. Oligonucleotides were synthesized by Integrated DNA Technologies and used without further purification. DNA sequencing was performed by the DNA Analysis Facility, mass spectrometry by the Applied Biosystems-Mass Spectrometry Facility, and N-terminal peptide sequencing by the Synthesis and Sequencing Facility, all of The Johns Hopkins University School of Medicine. Standard molecular biology procedures were performed as described (18).
Clones-The TraM open reading frame was amplified from JM109 genomic DNA by PCR with primers that incorporated appropriate restriction sites (22) and cloned into NdeI/XhoI-digested pET21aϩ. For pET-TraM, the reverse PCR primer included a stop codon; but for pET-TraM-His 6 , the primer lacked a stop codon, which allowed for expression with a C-terminal hexahistidine tag. The N-terminal peptide (pET-NTP) and C-terminal domain (pET-CTD) clones were constructed by removing unwanted sequence by PCR mutagenesis (22). This procedure resulted in the addition of an N-terminal methionine to the CTD open reading frame. 2 Protein Overexpression and Purification-Plasmids were transformed into BL21(DE3) or BL21(DE3)/pLysS. To test for protein overexpression, several colonies were individually transferred to 1 ml of LB medium with 60 g/ml ampicillin (LB/ampicillin) and grown to A 600 ϳ 0.4. The culture was split, and protein expression was induced in one-half for 3 h by adding isopropyl-1-thio-␤-D-galactopyranoside to 1 mM, and the other half was kept on ice. Protein expression was assayed by SDS-PAGE, and the reserved portion of the culture that demonstrated the best expression was diluted to 20 ml with LB/ampicillin, regrown to A 600 ϳ 0.5, and then used to inoculate 2 liters of LB/ ampicillin. When this culture reached mid-log phase, isopropyl-1-thio-␤-D-galactopyranoside was added to 1 mM. After 3 h, cells were harvested by centrifugation and stored at Ϫ80°C. For protein purification, 500-ml cell pellets were resuspended in 25-50 ml of 50 mM Tris-HCl (pH 7.5), 0.1 mM EDTA, 50 mM NaCl, 5 mM dithiothreitol (DTT), and 1 mM phenylmethylsulfonyl fluoride and disrupted by sonication. The soluble fraction was cleared by centrifugation; ammonium sulfate (0.25 g/ml) was added; and precipitated protein was removed by centrifugation. The remaining soluble protein was precipitated by adding ammonium sulfate to saturation and collected by centrifugation. Protein was resuspended in 35 ml of PNE buffer (25 mM NaPO 4 (pH 7.5), 25 mM NaCl, and 0.1 mM EDTA) with 1 mM DTT and dialyzed overnight against 4 liters of the same buffer.
TraM and NTP were further purified by ion-exchange chromatography. Buffer was generally kept on ice, especially for the purification of NTP, to minimize degradation of the protein. Dialyzed samples were centrifuged at 15,000 ϫ g for 10 min and loaded onto a 5-ml HiTrap heparin column equilibrated in PNE buffer with 5 mM DTT (PNED buffer). A 100-ml linear gradient to 2 M NaCl was applied. Fractions enriched in the desired protein were pooled, diluted 1:5 with PNED buffer, and applied to a 5-ml HiTrap Blue column equilibrated in PNED buffer. A 75-ml linear gradient to 2 M NaCl was applied, and the purity of peak fractions was assessed by SDS-PAGE. When necessary, one or both of these two chromatography steps were repeated. The protein in the final pooled fractions was concentrated by ammonium sulfate precipitation when appropriate. Purified protein was dialyzed extensively against PIPES buffer (25 mM PIPES (pH 7.0), 150 mM NaCl, and 0.1 mM EDTA) or phosphate buffer (25 mM NaPO 4 (pH 7.5), 150 mM NaCl, and 0.1 mM EDTA).
For purification of CTD, dialyzed cell extracts were cleared by filtering (0.45 m) and loaded onto a 5-ml HiTrap Q column equilibrated in PNED buffer. Protein was eluted with a 100-ml linear gradient to 2 M NaCl. Enriched fractions were combined and run over a 5-ml HiTrap heparin column followed in series with a 5-ml HiTrap Blue column. The flow-through was collected and dialyzed overnight against 4 liters of PNED buffer. The HiTrap Q column step was repeated, and the concentrated protein (3-5 ml) was loaded onto a Sephacryl S-200 HiPrep 16/60 size-exclusion column equilibrated in PIPES buffer at room temperature and eluted at 0.2 ml/min.
TraM-His 6 was purified by batch loading the soluble fraction of sonicated cells onto Ni 2ϩ -NTA-agarose (QIAGEN Inc.). The column was washed with 5 mM and then 20 mM imidazole in phosphate buffer with 1 mM DTT, and protein was eluted with 600 mM imidazole in phosphate buffer with 1 mM DTT. Imidazole was removed from the protein solution by dialysis. Protein concentration was measured by absorbance at 280 nm with extinction coefficients of 3960 M Ϫ1 cm Ϫ1 for TraM and TraM-His 6 , 2560 M Ϫ1 cm Ϫ1 for NTP, and 1400 M Ϫ1 cm Ϫ1 for CTD (23).
GndHCl-induced Unfolding-Stock buffer solutions were filtered (0.45 m) before use. Concentrated protein solutions were used within 2 weeks of purification and were centrifuged on the day of the experiment for 1 h at 55,000 rpm in a Beckman Optima TL ultracentrifuge using a TLA-55 rotor at 4°C. Native and unfolded (in 4 M GndHCl) protein samples were diluted to the indicated concentrations immediately prior to the titration. When indicated, sbmA DNA (5Ј-CGCTAGGGGCGCTGCTAGCGGTGCGT-3Ј, annealed to its complement) (3,9) was included at an equal molar ratio to the TraM tetramer. All solutions were allowed to equilibrate for 5 min before the first measurement was recorded. Protein and DNA concentration remained constant throughout the titration. All data were collected at 37°C, and titrant was kept at 35 Ϯ 2°C during the experiment.
For titrations monitored by intrinsic tyrosine fluorescence, protein was in PIPES buffer to prevent potential interactions between phosphate and tyrosine (24,25). Titrations were monitored with a SLM-AMICO 48000 spectrofluorometer in the L-configuration equipped with a Neslab circulating water bath and magnetic stirrer. Polarizers were set to the magic angle. The sample was excited at 275 nm (4-nm excitation slit width), and emission was monitored at 302 nm (16-nm emission slit width). Each fluorescence intensity measurement is the average of at least 200 data points acquired over 2-5 s, and each data point was averaged five times. The titration was performed manually as described (26). The sample was allowed to equilibrate with stirring for at least 2 min before each measurement.
Titrations monitored by CD were performed in PIPES or phosphate buffer with similar results. Data from unfolding of full-length TraM were acquired on an Aviv 62A DS spectropolarimeter with a computercontrolled Hamilton Microlab 500 titrator with two 500-l syringes (27,28). The samples were stirred for 90 s, and then data were averaged for 30 s. This was sufficient to achieve equilibrium, as indicated by the stability of the CD signal. Manual titrations were performed on a Jasco J-710 spectropolarimeter equipped with a PTC-348WI thermostat. The procedure was as described for the fluorescence measurements, except that each measurement is the average of 600 data points acquired over 30 s. The S.D. of each measurement is indicated by error bars on the graphs.
Spectroscopic data were fit using the nonlinear least-squares analysis curve-fitting function of Kaleidagraph Version 3.51. A signalweighted two-state model was employed (26,29,30). A similar threestate model was derived to fit the observed spectroscopic signal (Y obs ) when appropriate. Y obs is the sum of the fraction of each species (native, f N ; intermediate, f I ; and unfolded, f U ) multiplied by its contribution to the signal (Y N , Y I , and Y U ) (Equation 1).
The spectroscopic signal associated with each species (Y N , Y I , and Y U ) varied linearly with denaturant concentration. K 1 and K 2 are equilibrium constants for the first and second steps of the three-state equilibrium, respectively. According to the linear extrapolation model (29), each K can be expressed as a function of the standard state equilibrium constant, K 0 , and the m-value. ⌬G 0 was calculated from the fit K 0 values. An unfolding curve was determined to exhibit hysteresis if the forward and reverse curves were reproducibly offset greater than the error in the measurement.
DNA Dissociation Titration-The dissociation of sbmA DNA (2.5 M in PIPES buffer) upon addition of GndHCl was monitored by absorbance at 260 nm with a Beckman DU-70 spectrophotometer equipped with a temperature-controlled cell holder. DNA was diluted into the buffer at 37°C and allowed to equilibrate for 5 min before each measurement.
Analytical Ultracentrifugation-Protein solutions in PIPES buffer with various concentrations of GndHCl and matching buffer solutions were centrifuged at 55,000 rpm in a Beckman Optima TL ultracentrifuge at 4°C for at least 1 h. The top half of each solution was removed for use in analytical ultracentrifugation experiments. Sedimentation velocity analytical ultracentrifugation (SV-AUC) experiments were performed in a Beckman XL-I analytical ultracentrifuge using absorbance optics and an An-60Ti rotor. All runs were performed at 35°C and 2 Primer sequences are available upon request. 60,000 rpm in cells with charcoal-filled Epon 12-mm double-sectored centerpieces and sapphire windows. The rotor and cells were prewarmed. Radial absorbance data were collected at 275 nm at a nominal point spacing of 0.003 cm with no averaging in continuous scan mode.
Weight average sedimentation coefficients (͗s* 20,w ͘), extrapolated to t ϭ 0, were calculated with the program DCDTϩ Version 1.14 (31, 32). Solvent density, viscosity, and the buoyant molecular weight of the protein species were calculated with SEDNTERP Version 1.06 (33). Data for CTD and NTP were analyzed by fitting to the modified Fujita-Machosham function using the dc/dt mode of SVEDBERG Version 6.39 (32,34,35), which is better suited for small molecules. In all analyses, 14 -20 scans were included in the fit. Goodness of fit was evaluated by randomness in residuals, reduced 2 , and visual examination of the fits overlaid onto the experimental data. Regions of the data that showed signs of meniscus effects or that indicated accumulation at the bottom of the cell were excluded from the fits, although, in the figures, the fit values are extended over the entire s* 20,w range.
DNA Binding Interference Assay-TraM, preincubated with NTP or CTD, was combined with sbmA DNA. Electrophoretic mobility shift assays were performed essentially as described (22), except that equilibration was at 37°C.
Mating Assays-DY330 FЈ ER and DY330 FЈ ER M::kan were transformed with pET21aϩ (empty vector), pET-TraM, pET-NTP, or pET-CTD by electroporation. Overnight cultures from single colonies were diluted 1:100 into medium containing all appropriate antibiotics and regrown to mid-log phase. Donor cells (150 l) were spun down and resuspended in 100 l of warm LB medium, and then 900 l of mid-log phase TB1 recipient cells were added. Mating cultures were incubated for 2 min at 37°C without agitation. Conjugation was interrupted by vigorous vortexing, and then cells were incubated on ice for 10 min. Serial dilutions in cold sterile phosphate-buffered saline (18) were plated to select for donors and transconjugants. All mating experiments were repeated with at least three independent donor clones.
Limited Proteolysis-TraM or TraM-His 6 was mixed with trypsin at a molar ratio of 75:1 and incubated at room temperature. Aliquots were removed at hour intervals, and the reactions were stopped by addition of phenylmethylsulfonyl fluoride to 2 mM. The fragments generated were separated by electrophoresis on an SDS-polyacrylamide gel (15% acrylamide) and visualized by staining with Coomassie to monitor extent of digestion to ensure that both proteins generated the same proteolysis pattern. The C-terminal proteolytic fragments of TraM-His 6 were purified after 4 h. The reaction was stopped with phenylmethylsulfonyl fluoride and mixed for 10 min at 4°C with 25 l of Ni 2ϩ -NTAagarose equilibrated in phosphate buffer with 5 mM ␤-mercaptoethanol. The Ni 2ϩ -NTA-agarose was washed once, and bound protein was eluted in phosphate buffer with 5 mM ␤-mercaptoethanol and 600 mM imidazole. Eluted protein was run on an SDS-polyacrylamide gel; stained with Coomassie; and then transferred to Hybond-P membrane (Amersham Biosciences) in 20 mM CAPS (pH 10), 1 mM DTT, and 10% methanol at 300 V for 8 h at 4°C. Bands were excised from the membrane and subjected to N-terminal peptide sequencing.
Ni 2ϩ -NTA Coelution Assay-TraM-His 6 (5 M) was combined with 15 M TraM, NTP, or CTD in 50-l reactions in PIPES buffer with 10 mM DTT and incubated for at least 30 min at 37°C. Negative controls included all components except TraM-His 6 . Each reaction was mixed with 25 l of Ni 2ϩ -NTA-agarose at room temperature for 5 min; the agarose was washed once with 0.5 ml of PIPES buffer; and then bound protein was eluted in 50 l of PIPES buffer with 0.5 M imidazole. Protein that did not bind to Ni 2ϩ -NTA-agarose (flow-through) was compared with retained species by SDS-PAGE. The gels were scanned, and the band intensities were quantitated with the FluorChem imaging system (Version 2.00) controlled by AlphaEase FC software (Alpha Innotech Corp.). The background was subtracted from each band, and the coelution efficiency was calculated as the ratio of eluted test protein to TraM-His 6 in the same lane.

RESULTS
GndHCl-induced Unfolding of F Factor TraM-The reversible unfolding reaction of TraM was followed by CD and intrinsic tyrosine fluorescence (Fig. 1). CD is a measure of secondary structure, whereas tyrosine fluorescence can report on close tertiary interactions, hydrogen bonding of the fluorophore, and secondary structure. The GndHCl-induced unfolding transition is biphasic, with an apparent intermediate state populated near 1 M GndHCl. The congruence of the curves when monitoring independent probes is consistent with each transition ob-served in the unfolding curve being two-state (36). This suggests that the full unfolding transition is three-state, involving only one intermediate. These experiments were performed at 37°C because the unfolding transition at 20°C exhibited pronounced hysteresis (data not shown).
The equation describing a three-state model includes too many parameters relative to the number of data points to obtain reliable fits. Therefore, the slope and intercept of the base line at high denaturant concentration were determined by fitting data from this region directly to a linear equation, and these values were fixed during subsequent fitting. It was not necessary to constrain the shorter native base line in this manner. By fitting the constrained model to the CD data, we calculated that the overall stability of 10 M TraM at 37°C is 4.7 kcal mol Ϫ1 . ⌻he first transition (⌬G 1 ϭ 2.7 kcal mol Ϫ1 , m 1 ϭ 3.7 kcal mol Ϫ1 M Ϫ1 GndHCl) results in a greater decrease in free energy than the second transition (⌬G 2 ϭ 2.0 kcal mol Ϫ1 , m 2 ϭ 1.1 kcal mol M Ϫ1 GndHCl). The fit of fluorescence data with the three-state model did not converge. This is likely due to a lower signal-to-noise ratio in these data. However, when the ⌬G and m-values derived from the CD fits were fixed, the fluorescence data were reasonably well fit by the three-state model (data not shown).
Although the unfolding data indicate two transitions in TraM unfolding, we do not know how each transition reflects events of the unfolding pathway. To determine which phase of TraM unfolding involves a change in oligomerization, we performed the unfolding titrations of TraM at protein concentrations of 1-45 M (Fig. 2). 10 M TraM had a greater molar ellipticity at 222 nm than 1 M, showing that TraM is stabilized by increasing the protein concentration. Although the offset between the two curves makes them difficult to compare, the first transition began at a slightly lower concentration of denaturant at the lower protein concentration, suggesting that the first transition has shifted. Neither transition of the fluorescence-monitored unfolding was dramatically shifted when the concentration of TraM was raised from 10 to 45 M. We also did not see a shift in either unfolding transition when sbmA, the highest affinity binding site, was included in the reaction. At this temperature and concentration, the midpoint of the GndHCl-induced dissociation of sbmA is located at ϳ0.5 M (data not shown).
SV-AUC Analysis of TraM Oligomerization-To directly evaluate how the oligomeric nature of TraM changes during unfolding, we performed SV-AUC at different concentrations of denaturant. The data from these experiments can be used to determine the molecular weight of the sedimenting species based on hydrodynamic models and can also be used to monitor changes in oligomerization (37-40). We were not able to perform complementary equilibrium analytical ultracentrifugation because TraM aggregation occurred before the sample reached equilibrium. The data for SV-AUC experiments were collected under conditions similar to the unfolding of 45 M TraM shown in Fig. 2. This was the lowest protein concentration that could be detected with the absorbance optics at 275 nm.
To determine whether the first transition involves a change in oligomerization, we measured the sedimentation of TraM at 0.5, 0.75, and 1 M GndHCl (Fig. 3), denaturant concentrations that span the first unfolding transition. Visual examination of the curves suggests that, even at the lowest concentration of GndHCl, the boundary shape is not symmetrical. This is consistent with the solution containing a heterogeneous mixture of species. As the concentration of GndHCl was increased, the distribution first became more asymmetric, and then the peak shifted to the left. This translation of the g(s*) distribution toward lower s* values is consistent with a change in oligomerization concurrent with the first phase in the unfolding. The g(s*) representation of the data shown is model-independent and provides a qualitative comparison of TraM sedimentation as a function of unfolding (34).
We performed a more rigorous, quantitative analysis by determining the weight average sedimentation coefficient (͗s* 20,w ͘) as a function of denaturant concentration, shown in Fig. 3. The ͗s* 20,w ͘ represents the equilibrium distribution of all species, regardless of the rate of interconversion, as long as the system is at equilibrium at the beginning of the run (41). In Each g(s*) distribution was derived from 14 -16 radial absorbance scans at approximately the same 2 t and was area-normalized (37). As the concentration of GndHCl was increased, the peak in the g(s*) distribution became more asymmetric and shifted to lower s* values. The beginning and end points of the shift are marked with vertical lines. Lower panel, the weight average sedimentation coefficient is plotted as a function of GndHCl concentration (ࡗ). The two-state model was fit to these data to derive ⌬G tet(app) ϭ 2.9 Ϯ 0.8 kcal mol Ϫ1 and m tet(app) ϭ 3.1 Ϯ 2.3 kcal mol Ϫ1 M Ϫ1 GndHCl. AU, absorbance units. our experiments, the protein samples were incubated in the cell at 35°C for 20 -45 min before the experiment was initiated. When native TraM was diluted into 2 M GndHCl at this protein concentration, the fluorescence-monitored unfolding was complete within 3 min (data not shown). The ͗s* 20,w ͘ data indicate that the change in oligomerization that occurred upon unfolding of TraM was completed by 1.5 M GndHCl under these conditions. We fit the two-state model to the ͗s* 20,w ͘ data to derive the apparent tetramerization energy (⌬G tet(app) ). From this fit, we determined ⌬G tet(app) to be 2.9 kcal mol Ϫ1 and m tet(app) to be 3.1 kcal mol Ϫ1 M Ϫ1 GndHCl. These values agree well with ⌬G 1 (2.7 kcal mol Ϫ1 ) and m 1 (3.7 kcal mol Ϫ1 M Ϫ1 GndHCl) from the chemical unfolding, providing further evidence that the dissociation of the native tetramer is two-state. Together, these data indicate that the first phase in the unfolding of TraM involves dissociation of the native tetramer to a monomeric form.
Isolating Structural Domains of TraM-Using limited proteolysis, we generated a stable C-terminal fragment beginning at TraM residue 58 (data not shown). Expression vectors for this C-terminal domain (pET-CTD) and the remaining N-terminal peptide (pET-NTP) were constructed. The result from mass spectrometry of purified CTD matched the value predicted from the DNA sequence (8100; predicted molecular weight of 8103). However, mass spectrometry and N-terminal peptide sequencing of purified NTP indicated that it does not include the initial methionine and the final two residues coded for in the genetic construct.
To determine whether the two phases observed in the fulllength denaturation experiments correlate with sequential unfolding of these domains, we analyzed the unfolding of the individual domains (Fig. 4). Although the midpoint of the NTP transition is similar to the second transition in the full-length protein, both ⌬G NTP (3.0 kcal mol Ϫ1 ) and m NTP (2.0 kcal mol Ϫ1 M Ϫ1 GndHCl) are greater than the corresponding ⌬G 2 (2.0 kcal mol Ϫ1 ) and m 2 (1.1 kcal mol M Ϫ1 GndHCl). The unfolding of 45 M NTP overlays the data collected with 10 M protein, suggesting that this protein does not undergo a change in oligomerization upon unfolding. Furthermore, unlike TraM, the molar ellipticity of NTP did not change when the protein concentration was increased (data not shown).
The unfolding transition of CTD was completed by 0.5 M GndHCl. At this concentration of denaturant, there was no observed loss of TraM structure. The native base line in these unfolding curves is not well defined, which makes fitting these data difficult. The unfolding of 45 M CTD was shifted relative to 10 M, consistent with it being oligomeric. When the twostate model was fit to the 45 M data, ⌬G CTD(45) ϭ 2.0 kcal mol Ϫ1 and m CTD(45) ϭ 3.6 kcal mol Ϫ1 M Ϫ1 GndHCl. The m-value derived from these fits is very similar to m 1 (3.7 kcal mol Ϫ1 M Ϫ1 GndHCl) of the full-length protein. Because the m-value reflects the amount of surface area exposed upon unfolding, which should be the same at both protein concentrations, we fixed this parameter in fits of the two-state model to the data collected at 10 M. In these fits, ⌬G CTD (0.6 -0.9 kcal mol Ϫ1 ) is lower than ⌬G 1 (2.7 kcal mol Ϫ1 ). These experiments suggest that loss of CTD structure coincides with dissociation of the tetramer.
We used SV-AUC to determine whether CTD and NTP are oligomeric in solution (Fig. 5). The concentration of protein in these experiments (125 M) was higher than for the full-length protein (45 M) to achieve sufficient signal for data collection. These data indicate that CTD is a tetramer in solution, consistent with the concentration dependence of the unfolding curve. The SV-AUC data also suggest that NTP is dimeric in solution. These data are at odds with the observation that the NTP unfolding curve was not shifted when the initial protein concentration was changed. Together with the unfolding data, these experiments show that CTD structures are sufficient for tetramerization, although N-terminal residues may be involved in stabilizing the tetramer.
Functional Analysis of Domains-Our thermodynamic experiments indicate that CTD and NTP act cooperatively to stabilize the TraM protein. We proceeded next to see if both domains are required in vivo for the function of TraM. Indeed, only the full-length protein (and neither CTD nor NTP) was able to complement a traM deletion mutant (Table I). We also tested the effect of including each construct in trans to a wild-type FЈ episome. The transfer efficiency of DY330 FЈ ER was slightly, but reproducibly, decreased in the presence of CTD, but not TraM or NTP. This observed dominant-negative effect could indicate that CTD is included in tetramers with the full-length protein in vivo and that these tetramers are not active.
To determine whether the fragments interact with TraM in vitro, we monitored the coelution of the untagged species with TraM-His 6 from Ni 2ϩ -NTA-agarose (Fig. 6). The concentration of protein in these experiments was much lower than in the SV-AUC experiments. Both fragments coeluted with TraM-His 6 , albeit at lower efficiency than with TraM, although a genetic interaction was seen only for CTD. We used a gel-  retardation assay to address whether the observed interactions affect the DNA-binding activity of TraM (Fig. 7). Although no in vivo effect of NTP expression was observed, this fragment reduced the ability of TraM to bind DNA. Conversely, the interaction of CTD with TraM did not significantly affect its DNA-binding properties. DISCUSSION We have established conditions for the first reported reversible unfolding of the TraM protein, allowing us to describe its stability and oligomerization. Unfolding experiments were performed at 37°C, a temperature at which the unfolding transition is relatively free of hysteresis, to allow for application of thermodynamic models. Moreover, this is the physiological temperature at which TraM functions. Although calorimetry experiments with R1 TraM (17) indicate that, at 37°C, the protein shows some structural alteration, this thermal unfolding reaction is irreversible. In our equilibrium experiments, the GndHCl-induced unfolding curve of TraM at 37°C exhibits a well defined base line at low concentrations of denaturant, which suggests that the protein is in a native folded state under these conditions. Our characterization of TraM in these experiments defines a thermodynamic reference state that can be used in future studies to correlate observed phenotypes with defects in stability or oligomerization. The more precise characterization of TraM mutant phenotypes could help further elucidate the molecular role of TraM in mating signal transduction.
We have shown that monomeric and tetrameric forms of TraM are in equilibrium in solution. The existence of a folded monomeric species suggests that multiple forms of TraM may coexist inside a cell without being degraded. The three-state unfolding of F factor TraM we report is consistent with the observation that the irreversible thermal unfolding of TraM from F-like plasmid R1 is non-two-state (17). We calculate that, at 10 M TraM, the energy difference between the monomeric and tetrameric states is on the order of 2.5 kcal mol Ϫ1 at 37°C. This is only four times RT, the thermal energy at this temperature. The concentration of TraM in the cell is 2-3 orders of magnitude lower than in our experiments (12), which would decrease the energy difference between the monomer and tetramer. However, there are also factors in vivo that could stabilize the tetramer, such as specific interaction with DNA or other proteins or the cytoplasmic environment. The interplay of these factors will ultimately determine which oligomeric state is favored in the cell, although our experiments demonstrate it is feasible that both states are populated in vivo. TraM was not apparently stabilized by the presence of its DNA ligand in vitro (Fig. 2). However, in the absence of DNA, the first unfolding transition did not begin until 0.5 M GndHCl, which is near the midpoint of the sbmA DNA dissociation transition. Furthermore, it is likely that GndHCl can interfere with DNA binding by TraM. The effects of GndHCl on either DNA dissociation or TraM-DNA interactions could prevent observation of TraM stabilization by a TraM⅐DNA complex.
The equilibrium between tetramer and monomer is twostate, indicating that no other folding intermediates exist. These results are consistent with studies of R1 TraM that show DNA binding only as a tetramer (17), but argue against a model in which TraM initially binds to DNA as a dimer (9). Interestingly, SV-AUC indicates that NTP is dimeric in solution, although unfolding experiments suggest that it is not oligomeric. The discrepancy between the unfolding and SV-AUC experiments may indicate that dissociation of the NTP dimer does not perturb the structure in a manner that can be detected by either CD or intrinsic tyrosine fluorescence. An analogous fragment of R1 TraM, TraMM26, was also shown to be dimeric by gel filtration (17), although it was monomeric under conditions used to solve the NMR structure (42). The observation that TraM does not form dimers suggests that NTP structure may contribute to the stability of the TraM tetramer, but is not sufficient to support oligomerization alone.
Our domain studies indicate that structures important for oligomerization and DNA binding overlap. Although CTD is able to act as a tetramerization domain, the TraM tetramer is more stable than the isolated CTD tetramer, consistent with the assertion that N-terminal structures influence the stability of the tetramer. Furthermore, it has been shown that NTP contains residues that are important for DNA binding (11,43,44), although neither this fragment nor CTD can reproduce the DNA-binding activity of TraM. This suggests that the DNAbinding domain of TraM includes structures that are involved in, or stabilized upon, oligomerization. Consistent with this hypothesis, the R1 TraM homolog is stabilized upon the addition of DNA (17).
We have shown that, in vitro, both NTP and CTD can interact with full-length TraM. Although we have not ruled out the possibility of other types of interactions, it is possible that the constructs are incorporated into tetramers with the full-length protein. Our experiments demonstrate that, in vitro, CTD does not interfere with TraM-DNA binding, although it does show a dominant-negative interaction in mating experiments. We suggest that, in vivo, CTD interferes with an essential function of TraM that is separate from DNA binding. For example, it is possible that the TraM-TraD interaction is disrupted in these cells. In contrast, NTP interferes with the ability of TraM to bind DNA, although no genetic interaction was observed. This apparent contradiction can be explained if the interaction relieves the negative autoregulation of traM. The increased expression of TraM would interfere with any dominant-negative effect on conjugation.
We propose that the mating signal could modulate the oligomerization equilibrium of TraM. In this model, the negative autoregulation of TraM keeps the intracellular protein concentration at a level such that only sbmA and sbmB, the high affinity binding sites, are occupied by TraM tetramers. After successful mating pairs have been formed, the mating signal shifts the oligomerization equilibrium, and the concentration of TraM tetramers is increased. This equilibrium shift results in occupation of the sbmC binding site, thereby allowing for DNA transfer to begin. Our thermodynamic analysis of TraM oligomerization indicates that this model is consistent with the physical properties of TraM. Future studies will determine whether the activity of TraM is regulated by its oligomerization in vivo.