The Dodecin from Thermus thermophilus, a Bifunctional Cofactor Storage Protein*

Dodecins are so far the smallest known flavoproteins (68-71 amino acids) and are most likely involved in prokaryotic flavin storage. The dodecin monomers adopt a simple βαββ-fold and assemble to hollow sphere-like dodecameric complexes. Flavin binding by the dodecin from Thermus thermophilus showed a 1:1 stoichiometry and apparent dissociation constants in the submicromolar to nanomolar range as characterized by isothermal titration calorimetry and fluorescence titrations. The x-ray structures of the flavin-prebound and FMN-reconstituted state of the T. thermophilus dodecin revealed binding of FMN dimers in a novel si-si-rather than the re-re-orientation of their isoalloxazine moieties as found before in an archaeal dodecin. Electron paramagnetic resonance studies demonstrated that upon reduction the excess electron is localized only on one flavin, thus making dodecin-bound flavins highly refractory to redox chemistry. Besides FMN dimers, trimers of coenzyme A are additionally bound to this eubacterial dodecin along the 3-fold symmetry face II of the dodecin complex. Therefore, dodecins can act as bifunctional cofactor storage proteins that sequester catalytic cofactors in prokaryotes very efficiently in an aggregated and unreactive state.

Flavoproteins are ubiquitous proteins using flavins as prosthetic groups. Because riboflavin (vitamin B 2 ) serves mostly as a biosynthetic intermediate, FMN and FAD are the principal flavin cofactors (1). They are involved in the catalysis of a wide range of biological redox reactions, such as the dehydrogenation of NAD(P)H, lipid esters, and D-amino acids, the oxidation of amines to imines, the formation and cleavage of disulfide bonds, the hydroxylation of aromatic substrates, or the activation of molecular oxygen. Furthermore, flavoproteins serve as electron transmitters in electron transfer processes like oxidative phosphorylation. Here, they act as mediators between typical 2-electron donors like NADH and 1-electron acceptors like the heme group. This versatility to engage in one-and two-electron transfer reactions is due to their isoalloxazine moiety that adopts either an oxidized, a semiquinoid, or fully reduced redox state (2). Flavins are not only involved in redox reactions but also in the sensing of blue or ultraviolet light (3). In cryptochromes, flavin chromophores mediate the flowering and daily light/dark cycles in plants (4), in phototropins they regulate phototropism (5), and in photolyases they are involved in DNA repair (6). Beside catalytic or sensoric actions, flavoproteins are also involved in the binding or transport of flavins. During pregnancy, a specific carrier system evolves in vertebrates including the riboflavin-binding proteins (RCPs or RfBPs), since adequate amounts of vitamin B 2 are essential for the normal fetal development (7).
Dodecins are a novel family of flavin-binding proteins of unknown function that were discovered in the Archaea Halobacterium salinarum during an inverse structural genomics project on halophilic proteins (8). So far it is the smallest known flavoprotein with only 68 amino acids. Apart from haloarchaea, dodecins are found in many eubacterial genomes, as 16% of all completely sequenced eubacteria possess dodecin encoding genes (84 of 520 total, as of July 2007). These organisms constitute mainly proteobacteria (64 of 273) but also actinobacteria (14 of 47), chlorobi (4 of 16), and deinococcus/thermus species (2 of 4), which thrive either in soil or aquatic habitats or as parasites in host organisms. Interestingly, dodecins occur in a large variety of pathogenic bacteria (e.g. Pseudomonas aeruginosa, Mycobacterium tuberculosis) as well as in organisms known for their biotechnological and environmental relevance (e.g. Chlorobium tepidum, Geobacter sulfurreducens). Based on sequence conservation, dodecins can be divided into two groups; whereas eubacterial dodecin sequences share an average sequence identity of 53%, the archaeal dodecin sequences cluster separately with only 31% sequence identity to the eubacterial dodecins. The crystal structure of the halophilic dodecin complexed to riboflavin shows a dodecameric hollow sphere-like arrangement of monomeric subunits with cubic 23 symmetry. Flavins are bound as dimers along the interfacial surfaces of the dodecameric complex. The main stabilizing interactions between the flavin dimer and the dodecin complex occur via an aromatic tetrade involving the isoalloxazine groups of the flavin dimer and the indole groups of two neighboring tryptophan residues. The dodecin monomer exhibits a new protein fold adopting a simple ␤␣␤␤ topology, where the amphipathic ␣-helix is partly enwrapped by the three-stranded antiparallel ␤-sheet. Structural studies and fluorescence-based binding experiments for the H. salinarum dodecin (9) showed the binding of several kinds of flavins and flavin-like molecules. Furthermore, lumiflavin and lumichrome bind like FMN and riboflavin as stacked dimers in the flavin binding pockets with 1:1 stoichiometries, whereas FAD is bound in a U-shaped locked conformation with an apododecin/ligand ratio of 2:1 (10). Binding constants obtained by fluorescence titrations indicate preferred binding of small flavins and flavin-like molecules such as lumiflavin and lumichrome compared with the bulkier flavins FMN and FAD.
We have characterized the dodecin from Thermus thermophilus, previously assigned as hypothetical protein TTHA1431, in terms of its structure and flavin binding characteristics. Besides a surprisingly different mode of flavin dimer binding compared with the H. salinarum dodecin, the binding of an additional cofactor, coenzyme A, was observed in T. thermophilus dodecin suggesting that dodecins have evolved to form highly efficient storage proteins for sequestering multiple cofactors in prokaryotic organisms.
For recombinant T. thermophilus dodecin production, Escherichia coli strain BL21(DE3)Gold was transformed with the appropriate plasmid and selected with 35 g/ml kanamycin. 2 liters of LB medium were inoculated with 50 ml of an overnight culture and grown at 150 rpm and 37°C for 20 h until an A 595 of 4.5 was reached to promote autoinduction (11). The cells were pelleted, resuspended in 20 mM Tris/HCl, pH 8.0, 200 mM NaCl, and stored at Ϫ80°C.
The frozen cells were thawed and disrupted by a French press. The cell lysate was centrifuged (15,000 rpm, 30 min), and the supernatant was incubated at 65°C for 10 min. Heat-denatured proteins were removed by centrifugation (15,000 rpm, 30 min), and the supernatant was dialyzed against 20 mM Tris/ HCl, pH 8.0, overnight. The protein was further purified by anion-exchange chromatography using a Q-Sepharose high performance (Amersham Biosciences) column equilibrated with 20 mM Tris/HCl, pH 8.0, and a linear gradient of 0 -1 M NaCl in 20 mM Tris/HCl, pH 8.0. Fractions containing the T. thermophilus dodecin were pooled, concentrated, and applied to a Superdex 200 (Amersham Biosciences) gel filtration column (120 ml) that was equilibrated with 20 mM Tris/ HCl, pH 8.0, 100 mM NaCl. Fractions of the dodecameric dode-cin with more than 90% purity were concentrated to about 10 mg/ml and stored at Ϫ20°C. Protein concentrations were determined by the Bradford protein assay (Bio-Rad).
Apododecin was prepared by an unfolding/refolding protocol. The T. thermophilus dodecin obtained from anion exchange chromatography was denatured at a concentration of 1 mg/ml with 6 M guanidinium chloride and dialyzed 3 times against 6 M guanidinium chloride, 20 mM Tris/HCl, pH 8.0, and refolded by dialyzing two times against 20 mM Tris/HCl, pH 8.0. Alternatively, the protein was precipitated with 6% trichloroacetic acid followed by dissolving the pellet in 6 M guanidinium chloride, 20 mM Tris/HCl, pH 8.0, and dialyzed stepwise against 20 mM Tris/HCl, pH 8.0, 100 mM NaCl. In both cases aggregated protein was removed by size exclusion chromatography.

Protein Characterization
Electrospray ionization mass spectra were taken with an Agilent 1100 Series mass spectrometer to obtain the molecular mass of the dodecin monomer. The apparent molecular mass of the oligomeric assembly was determined by gel filtration chromatography on a Superdex 200 HR10/30 (Amersham Biosciences) column equilibrated with 20 mM Tris/HCl, pH 8.0, 100 mM NaCl. Here, 200 l of T. thermophilus dodecin with a concentration of 1.1 mg/ml were loaded onto the column and eluted with a flow rate of 0.5 ml/min. To obtain a calibration line, mixtures of various size standards (Amersham Biosciences) were loaded.

Cofactor Identification
Bound flavin cofactors were identified by HPLC 2 /electrospray ionization-time-of-flight. After releasing the cofactor by denaturing a dodecin solution with 6% trichloroacetic acid, the cofactor extracts as well as FMN and riboflavin standards were loaded onto a Nucleodur 2 ϫ 125-mm C18 gravity column with a particle size of 3 M (Macherey & Nagel) and eluted with a water/methanol gradient at a flow rate of 0.2 ml/min. The elution was monitored by UV absorption at 215 nm, and masses were analyzed by electrospray ionization-time-of-flight using a Q-Star pulser (Applied Biosystems).

Cofactor Reconstitution
To obtain cofactor-reconstituted dodecin, a saturated flavin solution was added to the protein solution in a concentrator cell (Vivascience) until the flow-through became yellow. Reconstituted T. thermophilus dodecin was separated from the excess of flavin by gel filtration chromatography. For an estimation of the cofactor concentration, the protein was precipitated with 6% trichloroacetic acid, and the absorption of the supernatant at A 448 was measured. Using the extinction coefficient of free FMN and a modified extinction coefficient of FMN in 6% trichloroacetic acid (⑀ 448 ϭ 10,400 M Ϫ1 cm Ϫ1 ) the extinction coefficient of dodecin-bound FMN was determined to be 9,500 M Ϫ1 cm Ϫ1 . A calculated FMN:protein ratio of 1:1.2 was consistently obtained for different preparations indicating that FMN was bound in a 1:1 stoichiometry.
Orange-colored, hexagonally shaped crystals of the FMNreconstituted R65A mutant (P3 2 21, a ϭ b ϭ 67.59 Å, c ϭ 169.19 Å, half a dodecamer per asymmetric unit) appeared under sitting-drop conditions within a few weeks. The sitting drop consisted of 300 nl of 26 mg/ml dodecin and 300 nl of reservoir solution (0.4 M ammonium phosphate), and the reservoir volume was 100 l. Data sets of the wild type dodecins and the R45A mutant were collected at 100 K from flash-frozen crystals using a Bruker-Nonius FR591 (50 kV, 80 mA)-rotating CuK ␣ anode with a MAR imaging plate (MAR-Research). For the R65A mutant, a 1.4-Å dataset was collected at beamline BW7A at EMBL outstation, Hamburg. For the non-reconstituted dodecin, a 2.4-Å data set was recorded, and the crystals of the reconstituted dodecin diffracted up to 2.6 Å ( Table 1). The data sets were processed using MOSFLM and scaled with SCALA (13). All structures were solved by molecular replacement with MOLREP (14) using the H. salinarum dodecin structure as an initial search model and refined by REFMAC5 (15) and COOT (16) using restrained TLS refinement.

Spectroscopic Measurements
Absorption spectra from 200 to 700 nm were recorded using an Ultrospec 3100 spectrophotometer (Amersham Biosciences) at room temperature with a scan rate of 350 nm/min and 0.5-nm steps. Buffer spectra were recorded under identical conditions and subtracted afterwards.
Fluorescence emission spectra were recorded with a FP-6500 spectrofluorometer (Jasco) equipped with a Jasco ETC-273T Peltier-type temperature control system at 20°C using a 200-l quartz cuvette with an excitation wavelength of 450 nm and an emission wavelength of 460 -700 nm. Spectra were taken with a scan rate of 1000 nm/min in 0.5-nm steps. For fluorometric measurements, filtered protein and cofactor solutions with a concentration of 5 M (related to flavin concentration) in 20 mM Tris/HCl, pH 8.0, 100 mM NaCl were used. Circular dichroism measurements were carried out on a J-810 (Jasco) equipped with a PTC-423S Peltier type temperature control system using quartz cuvettes with 0.1-cm path lengths. Protein samples had concentrations of 12.5 M in 10 mM Tris/HCl, pH 8.0. Data were recorded from 190 to 280 nm with 0.2-nm steps at a scan rate of 100 nm/min and averaged from 3 to 5 scans recorded at 20°C. The base line was corrected by subtracting the spectrum of the buffer collected under the same conditions. Thermal unfolding curves were measured by  3.1% (R45A), 5.1% (wild type/reconstituted), and 5.9% (wild type) of the data withheld from the refinement for cross-validation.

The Dodecin from T. thermophilus
following the ellipticity change at 198 nm during continuous heating from 20 to 95°C at a scan rate of 2°C/min.

Determination of FMN Binding
Flavins (Sigma) with a purity of Ͼ95% were used without further purification. Lumichrome (Sigma) required further purification by HPLC. Samples were separated on an Agilent 1100 HPLC system using a Nucleodur 125/2 C18ec column with a 3-m particle size at a flow rate of 0.2 ml/min and a column temperature of 40°C with a linear gradient of 10 -80% methanol, 0.1% trifluoroacetic acid in water, 0.1% trifluoroacetic acid within 25 min. Pooled fractions of interest were lyophilized.
Dissociation constants of flavin-dodecin complexes were determined by fluorescence titrations, where the fluorescence quench of the flavin ligand upon binding to the T. thermophilus dodecin was monitored with a FP-6500 spectrofluorometer (Jasco) equipped with a PTC-423S Peltier-type temperature control system. In this experiment 11.5 M dodecin containing endogenously bound FMN was used to avoid the kinetic effect observed before by ITC measurements and fluorescence titrations where apododecin was used. Ligand concentrations were 200 nM (FAD, lumichrome), 203 nM (riboflavin), 198 nM (FMN), and 232 nM (lumiflavin). All solutions were sterile-filtered and pre-equilibrated at 20°C in the dark. Measurements were made in 20 mM Tris/HCl, pH 8.0, 100 mM NaCl.
Fluorescence titrations were performed at 20°C in a 2-ml round cuvette. In a typical titration experiment, 1000 l of flavin solution was titrated with 4-l aliquots of dodecin every 2 min under stirring until an end point was reached. For riboflavin, FMN and lumiflavin, emission at 529 nm after excitation at 450 nm was measured with medium sensitivity and slit widths of 10 nm for emission and excitation. In the titration using FAD, the parameters were set to 450/532 with slid widths of 5 nm (excitation) and 10 nm (emission) and high sensitivity due to low FAD fluorescence at pH 8.0. In the case of lumichrome, the emission at 460 nm after excitation at 380 nm with slid widths of 5 nm (excitation) and 10 nm (emission) were measured. For the correction of the fluorescence of endogenously bound FMN, the protein was titrated into buffer under the same conditions as in the binding experiments.
The fluorescence was normalized to the fluorescence starting point of the ligand and corrected for the fluorescence of prebound FMN. With regard to dilution of ligand fluorescence, the data were plotted against the concentration of free dodecin. The apparent K D values were obtained by fitting these curves against the following one-site binding equation.
c total and r total are the ligand and total-protein concentrations after each titration point, respectively, F free is the fluorescence of the free cofactor, F meas is the corrected fluorescence readout, and F bound is the fluorescence of bound cofactor. K D is the dissociation constant. The listed K D values have been averaged from at least three titration experiments and were corrected for dilution effects as well as for FMN fluorescence of the first dimer. For a measurement of the time-dependent effect of FMN binding, 600 l of FMN (2 M) were titrated with 100 l of dodecin solution (100 M), which was equilibrated with 2 M FMN overnight at 20°C in the dark. Fluorescence emission spectra were taken before injection and every 30 min after injection with slid widths of 5 nm (excitation) and 5 nm (emission) with medium sensitivity and a scan speed of 1000 nm/min. For the determination of the kinetic effect, measurements were done at 20, 37, 50, and 65°C. Half-life periods were calculated by fitting the experimental data against the exponential decay formula.
ITC was performed at 25°C in a MCS MicroCal instrument with a reaction cell volume of 1.351 ml. Before use, solutions were degassed under vacuum to eliminate air bubbles. Titration experiments consisted of 25 injections of FMN with a concentration of 375 M into a 110 M dodecin solution. The volume of the first injection was 1.495 l, and the subsequent injection volumes were 10.021 l. Nonspecific heat effects were estimated from the magnitude of the peaks appearing after complete saturation. Raw data were integrated, corrected for nonspecific heats, and analyzed according to a 1:1 binding model using Origin TM 5.0 scientific plotting software.

EPR Spectroscopy
EPR/Electron-nuclear Double Resonance (ENDOR) Instrumentation-X-band-pulsed ENDOR spectra were recorded using a Bruker E580 pulse EPR spectrometer (Bruker BioSpin GmbH) in conjunction with a Bruker ER 4118X-MD5-EN dielectric-ring ENDOR resonator. For Davies-type ENDOR (17), a microwave pulse-sequence -t-/2-using 64 and 128-ns /2 and pulses, respectively, and a radio frequency pulse of 10-s duration starting 1 s after the first microwave pulse were used. The separation times t and between the microwave pulses were set to 13 s and 348 ns, respectively. To avoid saturation effects due to long relaxation times of flavin radicals, the entire pulse pattern was repeated with a frequency of only 400 Hz.
EPR Sample Preparation-For EPR studies, the dodecinbound FMN (246 M) was reduced with titanium(III) citrate (480 M) directly in an EPR tube (3 mm inner diameter) under anaerobic conditions. Results were checked photometrically.

RESULTS AND DISCUSSION
Expression and Purification-The gene coding for T. thermophilus dodecin was cloned by PCR from T. thermophilus strain HB8 and heterologously overexpressed in E. coli strain BL21(DE3)Gold. After purification by heat denaturation to remove most E. coli proteins, the T. thermophilus dodecin was purified to homogeneity as an orange-colored homo-oligomer by anion exchange and size exclusion chromatography (Fig.  1A). Overall yields were 17 mg of T. thermophilus dodecin per liter of culture with a purity greater than 95% (Fig. 1B). Mass spectrometric analysis of the recombinant T. thermophilus dodecin revealed a molecular mass of 7618.1 Da compared with 7616.6 Da as calculated for residues Gly-2 to Thr-69. Accordingly, the N-terminal formyl-methionine was removed by posttranslational modification.

JOURNAL OF BIOLOGICAL CHEMISTRY 33145
Spectroscopic Characterization of the Dodecin-Flavin Complex-UV/VIS spectroscopy of the orange-colored T. thermophilus dodecin complex showed the characteristic absorption spectrum of oxidized flavin species with absorption maxima at 270, 380, and 450 nm (Fig. 1C). An additional slightly red-shifted shoulder was observed in the range of 475-480 nm, indicating vibrational fine structure resulting from tight cofactor binding in the protein environment. Moreover, there is significant absorbance up to 600 nm representing charge-transfer interactions of the isoalloxazine rings with neighboring aromatic residues. According to fluorescence spectroscopy, the flavin fluorescence was quenched by Ͼ98% when compared with flavins in free solution (see Fig. 3A). Because the residual fluorescence was indistinguishable from free flavin, it is obvious that the binding of flavins to the dodecin oligomer almost completely quenches the endogenous flavin fluorescence. Likewise, attempts to induce a triplet state of the dodecin-bound flavin species in the frozen state failed when the dodecin complex was excited by a continuous 450-nm pump laser with triplet formation being followed by EPR (data not shown). Ultrafast UV-visible spectroscopy on defined flavin-dodecin complexes finally showed that a rapid electron transfer from neighboring tryptophan residues to the isoalloxazines takes place in the excited singlet state, thus avoiding consequent intersystem crossing and causing the efficient quench of flavin fluorescence. 3 Cofactor Binding and Oligomerization of the Dodecin Complex-The electrospray ionization-mass spectroscopy analysis of the flavin cofactor extracted under denaturing conditions from the dodecin complex showed by comparison with FMN, riboflavin, and FAD standards that the recombinant T. thermophilus dodecin predominantly binds the negatively charged FMN. This finding contrasts with the characterization of the previously discovered dodecin from H. salinarum, which was found to copurify exclusively with stoichiometrically bound riboflavin when being isolated from the cytosol of this haloarchaeon (8). The purified T. thermophilus dodecin com-plex exhibited a protein:cofactor ratio of 1:0.2, indicating substoichiometric flavin binding to the recombinant T. thermophilus dodecin, most likely due to the lack of sufficient flavin supply in the E. coli cytosol. However, the dodecin complex could be reconstituted afterward in vitro with either FMN, FAD, or riboflavin as shown by apparent cofactor:protein ratios of 1:1.2 after size exclusion chromatography, suggesting the formation of stoichiometric dodecin⅐flavin complexes.
To determine the oligomeric state of the dodecin complex, analytical size exclusion chromatography was performed as shown in Fig.  1A. Peak 1 indicated a molecular mass of 96.4 kDa, whereas peak 2 corresponded to a molecular mass of 39.8 kDa, suggesting that dodecameric T. thermophilus dodecin is in equilibrium with a hexameric state, if only substoichiometric amounts of FMN are bound. Reconstitution with FMN or riboflavin caused the disappearance of peak 2 indicating a stabilization of the dodecameric complex. SDS-PAGE analysis of the purified dodecin showed a high resistance against denaturation by SDS or heat. Even after boiling the protein samples for 5 min in gel-loading buffer comprising 4% (w/v) SDS, the monomer band was accompanied in the SDS-PAGE by an oligomeric band at the apparent molecular weight of a hexameric protein species (Fig. 1B). Whether this band corresponds to a partly denatured hexameric or a non-denatured dodecameric complex is unclear. As expected for a thermophilic protein the T. thermophilus dodecin oligomer exerts high thermal stability, since CD spectra of the holocomplex at 20 and 95°C as well as melting curves measured from 20 to 95°C showed no significant loss of secondary structure. Interestingly, flavin removal did not affect the thermal stability as no melting of the dodecameric apododecin was observed by CD spectroscopy.
Redox Chemistry of the T. thermophilus Dodecin-To characterize the redox activity of dodecin, which might be governed by the unique stacking arrangement of the flavin chromophores, chemical reduction of the T. thermophilus dodecin was attempted using reductants such as EDTA, dithiothreitol, and sodium dithionite. Although these reagents did not yield detectable reduced flavin species, reduction with titanium(III) citrate finally produced minor amounts of the flavin semiquinone species that remained stable at liquid nitrogen conditions. By continuous-wave EPR, a radical signature centered at g ϭ 2.0035(5) was detected that is characteristic for a flavin radical (18 -20) (Fig. 2A). After the discussion outlined previously (18), neutral flavin radicals (FADH ⅐ ) show typically peak-to-peak EPR line widths of about 2.0 mT, whereas anionic radicals show a reduced line width down to about 1.3 mT as a result of deprotonation of N(5) (21,22). Experimentally, a peak-to-peak line width of 1.3 mT was observed, which is consistent with the latter. However, line-width reductions are also expected for dimers or N-mers of electronically interacting flavins, when the unpaired electron spin is either delocalized over the then-extended conjugated -electron system or when the spin is rapidly transferred between the nitrogen flavins. This limiting value of the line width in an N-mer is by a factor of N 1/2 smaller than the monomer line width, where N is the number of sites over which charge migration occurs. Such line narrowing phenomena due to delocalization or very fast electron transfer were first observed in EPR studies of the "special pair" of bacteriochlorophylls in bacterial photosynthetic reaction centers (23) and, more recently, in the circular arrays of bacteriochlorophylls of the light-harvesting complexes LH1 and LH2 (24). To distinguish between the various mechanisms for the EPR line narrowing of the dodecin-bound flavin radical species, pulsed proton ENDOR experiments were performed (Fig. 2B). Despite moderate spectral resolution due to the low concentration of the flavin radical in T. thermophilus dodecin, a pair of lines of axial symmetry was clearly observed and assigned to the hyperfine coupling of the methyl protons attached to C(8␣); A Ќ ϭ 9.8 MHz, A ʈ ϭ 11.7 MHz, yielding A iso ϭ 10.4 MHz. When compared with other flavoprotein radicals, either anionic or neutral, this value is rather large, thus excluding the dimeric nature of the flavin species for which all hyperfine couplings should be halved if delocalization between the flavin units is uniform (23). Accordingly, electronic coupling of the coplanar pair of flavins in the radical form appears to be rather small. Hence, the observed narrow line width in the continuous-wave-EPR spectra is due to the anionic nature of the flavin. This assignment is also corroborated by the experimental hyperfine coupling that is in a range typically expected for 8␣-methyl protons in flavin anion radicals (25). Taking together the findings that dodecin is only difficult to reduce both chemically and photochemically and that the semireduced state is not stabilized by delocalization, it is rather unlikely that dodecins are involved in biological redox processes.
Specificity and Time Dependence of Flavin Binding to the T. thermophilus Dodecin-To determine binding affinities between the apo form of T. thermophilus dodecin and the flavin cofactor FMN, fluorescence titrations as well as ITC measurements were carried out using apododecin prepared by a denaturation/ refolding protocol. For fluorescence titration, the emission at 530 nm after excitation at 450 nm was detected, exploiting the fluorescence quench occurring upon flavin binding (Fig. 3A).  Titration experiments, where the injections were done every 5 min, showed a binding curve with a sigmoidal shape, which could not be fitted by a 1:1 binding model (data not shown). Time-dependent measurements, where the decrease in emission after only one injection was monitored (Fig. 4A), showed that complete fluorescence quench by flavin binding required several hours. An increase in temperature from 20°C to 37, 50, and 65°C led to a faster decay of fluorescence, indicating that the binding equilibrium is reached more rapidly at higher temperatures (Fig. 4B). Half-life times of 166 min (20°C), 86 min (37°C), 41 min (50°C), and 25 min (65°C) were determined. Using the Arrhenius formula for the temperature dependence of reaction rates, an activation energy E A of 35.6 Ϯ 2.4 kJ/mol can be derived for the formation of the initial dodecin⅐flavin complex if starting from the flavin-free apododecin (Fig. 4B). For comparison, ITC experiments, where the FMN solution was injected to apododecin every 5 min at 25°C, could be fitted with a simple 1:1 binding model. However, the ITC experiments yielded a modest K D of 0.92 M and showed substoichiometric binding of only 15% flavin to apododecin (Fig. 4, C and  D). Overall, these experiments indicate that the binding of the first flavin dimer to the apo state of the T. thermophilus dodecin is kinetically impeded and might require significant conformational changes within the dodecameric complex.
To investigate the binding of the remaining dimers, the T. thermophilus dodecin in its flavin-prebound state was used for fluorescence titrations, which has endogenously bound 20% FMN after purification. Here, binding of additional flavin to the T. thermophilus dodecin showed no measurable time dependence. For the binding affinity between FMN and T. thermophilus dodecin, an apparent K D of 0.311 M was determined with a 1:1 stoichiometry (Fig. 3, B  and C). Besides FMN, also FAD, riboflavin, lumiflavin, and lumichrome were tested for their binding affinities to the T. thermophilus dodecin. All flavins showed similar binding affinities, which are 589 Ϯ 26 nM for FAD, 311 Ϯ 18 nM for FMN, 233 Ϯ 37 nM for riboflavin, 141 Ϯ 11 nM for lumiflavin, and 80 Ϯ 3 nM for the lumichrome-dodecin complex (Fig. 3C). This is in contrast to the binding affinities reported for the H. salinarum dodecin (9). Here, the small ligands lumichrome, lumiflavin, and riboflavin showed significantly lower dissociation constants (11.6, 17.6, 35.8 nM) than the bulkier flavins FMN and FAD (11.3 and 0.44 M). Furthermore, FAD was found to bind in a 1:2 stoichiometry to the H. salinarum dodecin (9), which could be attributed to the binding of a closed conformer of FAD within the flavin dimer binding site of H. salinarum dodecin (10), whereas in the T. thermophilus dodecin FAD is bound like the other flavins with 1:1 stoichiometry.
Overall Structure of the T. thermophilus Dodecin-The T. thermophilus dodecin and its R45A and R65A mutants crystallized in four different crystal forms. Tetragonal crystals (space group P4 1 2 1 2) were obtained for the complex between wild type T. thermophilus dodecin and substoichiometrically bound amounts of FMN. Orthorhombic crystal forms (space group P2 1 2 1 2 1 ) were generated for the stoichiometric 1:1 complexes with FMN, which differed between the wild type and the R45A dodecin in the dimensions of the b and c axes, whereas the R65A mutant crystallized in the trigonal space group P3 2 21.
The T. thermophilus dodecin structures were solved by molecular replacement starting from the structure of the H. salinarum dodecin (34% sequence identity) as a search model (8). All crystal forms of the T. thermophilus dodecin showed the characteristic dodecameric assembly of monomeric subunits (Fig. 5, C and D), as anticipated before from analytical size exclusion chromatography. The T. thermophilus dodecin dodecamer with its cubic 23 point symmetry adopts a hollow sphere-like shape with an inner diameter of 23 Å and an outer diameter of 60 Å. The monomeric subunits have overall dimensions of 44 ϫ 20 ϫ 17 Å and adopt a simple ␤1␣1␤2␤3 topology, that is characteristic of the dodecin-fold (SCOP classification), where the ␣-helix is partly enwrapped by the threestranded antiparallel ␤-sheets (Fig. 5A). Although the ␣-helices face the outer surface of the dodecameric complex, the ␤ sheets line the interior of the dodecamer. Structural comparison between the T. thermophilus and H. salinarum dodecins shows the high conservation of the dodecin fold with a root mean square deviation of 0.67 Å for 58 C a atoms. Major structural differences are restricted to the variable ␤2-␤3 loop and the N and C termini (Fig. 5B), which reside on the outer surface of the dodecin complex and are not involved in flavin binding. Likewise, the quaternary structure is similarly conserved as a superposition of the chains A of both dodecin complexes on each other leads to largest divergences in the superposition of the other chains of below 1.5 Å.
The Flavin Dimer Binding Site of the T. thermophilus Dodecin-The various crystal structures of the T. thermophilus dodecin show the binding of flavin dimers along the 2-fold axes of the cubic-symmetric oligomer. The binding site is made up along the interface of four different dodecin subunits and corresponds to a 10 ϫ 15-Å large cleft (Fig. 5D) connecting the inner space of the hollow sphere-like particle with the exterior bulk solvent. The free access to the flavin binding sites is consistent with the reversible flavin binding and exchange as observed for the T. thermophilus dodecin in its flavin-prebound state. In the crystal structure of the T. thermophilus dodecin, where only endogenously bound flavin is present, the electron density observed at the end of the ribityl side chains clearly indicated the presence of the FMN phosphate group and, thus, corroborated the previous electrospray ionization-mass spectroscopy results concerning the cofactor content of the T. thermophilus dodecin (Fig. 6A). Due to substoichiometric FMN binding, only two of the six putative binding sites within the dodecin dodecamer show electron density for flavin molecules with refined occupancies of 70%. The other four binding pockets appear to be almost devoid of bound flavin (occupancies of Ͻ50%) as they only show residual density at the expected positions of the flavin dimers. However, the binding mode of the flavin dimer is not affected by the substoichiometric binding, because in the fully reconstituted dodecin⅐FMN complex the orientation of the FMN molecules is unchanged.
Important interactions stabilizing the flavin dimers within the T. thermophilus dodecins are shown in Fig. 6. Interestingly, only few highly conserved residues are involved in binding of the flavin dimer. As is typical for flavoproteins, the dimethylbenzene ring is primarily surrounded by hydrophobic side chains, whereas the nucleotide part is stabilized by hydrogen bonds to adjacent polar residues. Val-11 and Val-59 surround the hydrophobic side of the flavin forming an unpolar binding pocket. Val-59 is moderately conserved and can be mostly replaced by Ser, Thr, or Gly. Likewise, Val-11 can be replaced in several other dodecin homologues by either Ile or Thr (Fig. 7). The side chains of Gln-57 and Arg-45 contribute to the flavin binding by the formation of three hydrogen bonds. Gln-57 is a highly conserved residue present in all dodecin homologues, forming a hydrogen bond to C2AO und N3-H. Such hydrogen bonds between the apoprotein and the N3-H of the flavin cofactor are ubiquitously observed in other flavoprotein structures (26). Arg-45, stabilizing the flavin by hydrogen bonds to C4AO and N5, is conserved among eubacterial dodecins. In the archaeal dodecins, Arg-45 is substituted by Gly-45. The most important interaction is made by the indole groups of two highly conserved tryptophan residues (Trp-38) from two symmetry-related subunits. These residues stabilize the isoalloxazine moiety bystacking, forming a sandwich-like aromatic tetrade. The binding of the isoalloxazine ring by --stacking is also known from other flavin binding proteins; in flavodoxins and riboflavin-binding proteins the flavin is stabilized by forming an aromatic triad where the flavin is sandwiched between a tryptophan and a tyrosine residue (7,27).
The FMN phosphate group forms multiple salt bridges and hydrogen bonds to residues Lys-3, Tyr-5, Arg-65, and Asp-37. Whereas Tyr-5 is present in most dodecin homologues (53 of 56 sequences), only moderate conservation is observed for the residues Lys-3 (only 13 sequences), Arg-65 (35/55), and Asp-37 (19/57) (Fig. 7). Additional interactions by the ribityl group of FMN are only formed between the 2Ј-OH group and the mainchain carbonyl of Asp-37 and the 3Ј-OH group and the side chain of Tyr-5.
Comparison with H. salinarum Dodecin-Although the overall structure and the binding of flavin dimers appears to be similar in the dodecins from T. thermophilus and H. salinarum, the two dodecins differ significantly in their mode of flavin binding (Fig. 6B). Whereas the 1.7-Å structure of the H. salinarum dodecin clearly shows the presence of riboflavin dimers,  The Dodecin from T. thermophilus NOVEMBER 9, 2007 • VOLUME 282 • NUMBER 45 the T. thermophilus dodecin binds FMN as a cofactor. For this discrepancy the positively charged Arg-65 residue appears to be responsible because this residue is substituted in H. salinarum dodecin by a negatively charged glutamate residue.
Surprisingly, the stereochemistry and orientation of the aromatic tetrades differ between the dodecins from the two organisms. In contrast to H. salinarum dodecin, where the isoalloxazine moieties contact each other via their prochiral re sides, the isoalloxazine rings in T. thermophilus contact each other via their si sides. Interestingly, this inverse binding mode is accompanied by only slight changes in the conformations of the residues mostly responsible for flavin binding. Despite contacting non-equivalent flavins in both dodecins, Gln-57 is responsible for hydrogen bonding with the pyrimidine part of the isoalloxazine. A unique interaction that is only found in the H. salinarum dodecin is concerned with the recognition of the isoalloxazines from the inner compartment. Here, a magnesium ion was found to bridge between Gly-45 and the C4 carbonyl and N5 group of the isoalloxazine moiety. This interaction is missing in all eubacterial dodecins, where Gly-45 is replaced by the highly conserved Arg-45 that forms hydrogen bonds with the same groups. However, the structure of the R45A mutant (Fig. 8) shows that this interaction does not explain per se why the binding mode differs from the H. salinarum dodecin because in the R45A mutant the flavins adopt the same si-si association as in the wild type dodecin.
Whereas in the H. salinarum dodecin a large overlap of all four -systems occurs in the binding site of the flavin dimer, this overlap is diminished in the T. thermophilus dodecin complex. Here, the contact surfaces between the isoalloxazine moieties are by 21.2 Å 2 (20%) smaller (total, 85.0 Å 2 ) than in the H. salinarum dodecin (106.2 Å 2 ). The larger overlap in the H. salinarum dodecin also coincides with an almost parallel orientation of the ribityl side chains which allows additional stabilization of the flavin dimer by intermolecular hydrogen bonding between symmetry-related 3Ј-and 5Ј-hydroxy groups. Due to the increase of the tilt between the two isoalloxazines raising from Ϫ27°(H. salinarum) to 32°(T. thermophilus) as calculated for the relative orientation of the axes going through the N5 and N10 atoms, the two ribityl side chains of FMN in the T. thermophilus dodecin point to opposite directions so that less intermolecular interactions occur within the flavin dimer. The different binding modes might be determined by the recognition of the 5Ј-phosphate of FMN by charged residues at the surface of the T. thermophilus dodecin (e.g. Arg-65) which are missing in the halophilic protein. The reason for their absence might reside in the requirement to present mostly acidic residues on the outer surface of the halophilic protein (28), which is stable under conditions of Ͼ4 M NaCl. Whereas the dodecin from T. thermophilus comprises almost equal amounts of acidic (17%) and basic (17%) residues and exhibits a nearly neutral pI of 5.8, the H. salinarum dodecin exhibits a large excess of aspartates and glutamates (24%) over basic residues (6%) and possesses a remarkably low pI of 3.8. Accordingly, 13 of these acidic residues are localized on the outer surface of the H. salinarum dodecin complex. This does not only provide the required stabilization of the hydrate shell to avoid aggregation by a salting out effect but would also weaken the recognition of negatively charged flavin species by the H. salinarum dodecin (8).
The Binding Site for Trimeric Coenzyme A-Due to the 23 point symmetry of the dodecin complex, 2 kinds of 3-fold symmetric faces can be distinguished; in face I the 3-stranded ␤-sheet of each monomer extends to a 5-stranded antiparallel ␤-sheet (Phe-39 -Val-41) by local pairing with the ␤ 2 strand (Ile-44 -Gly-46) of the other monomer. In the 3-fold symmetric face II the ␤-sheets are continued throughout the whole dodecamer by main chain-main chain pairing between Gly-2-Lys-7 and Glu-9 -Thr-13.
The high resolution data for the R65A mutant displayed additional electron density in the 3-fold symmetry face II which could be identified as trimerized coenzyme A molecules (Fig.  9A). As becomes obvious from inspection of the electrostatic surface potential (Fig. 9B), the CoA binding pocket is formed primarily by positively charged (Arg-28, Thr-32, Arg-34, Glu-67) and hydrophobic (Leu-10, Ala-21, Leu-33, Phe-64, and Leu-66) residues. Overall, 3 of the 12 positively charged amino acids per monomer participate in CoA binding.
The thiol group of CoA, which is often acylated in other proteins, is buried in a mainly hydrophobic pocket built by residues Leu-10, Thr-32, Ala-29, and Arg-28 (Fig. 9C). The adenine moiety of the CoA is stabilized by hydrophobic and polar interactions. Besides the stabilization by the perpendicularly arranged residue Phe-64, Leu-33 and Leu-66 form a hydrophobic binding pocket. Aside from hydrophobic interactions, the main chain amide group of Glu-67 and Arg-34 form hydrogen bonds to the adenine atoms N1 and N7, respectively, and the main chain carbonyl groups of Arg-34 and Ala-65 stabilize the adenine N6 atom.
The ribose 3Ј-phosphate, the 5Ј-phosphate ester, and the pantetheine primarily form hydrogen bonds and salt bridges. For example, Lys-6 located on the bottom of the binding pocket forms a hydrogen bond with the carbonyl group of the pantetheine. The diphosphate of CoA is bound by salt bridges to residues Thr-32 and Arg-28 and the ribose 2Ј-hydroxy group H-bonds to residue Arg-34. Besides the stabilization by the protein environment, the formation of an intermolecular hydrogen bond between the phosphate ester and the hydroxyl group of the pantetheine of the neighboring CoA monomer takes place. Induced by the form of the binding pocket, intramolecular hydrogen bonds between the phosphate ester oxygen atoms (PO10-AO2), the carbonyl group of the pantetheine with the phosphate ester (AO6-AO2), and the 2Ј-hydroxy group (AO2*) of the ribose with the 3Ј-phosphate group (AO8, AO9) of the ribose occur, leading to the V-shaped form of the CoA molecules. In contrast to the adenine moiety, the phosphopantetheine, and the thiol group, which are stabilized by a number of amino acids, the 3Ј-phosphate group of the ribose projects out of the binding pocket, thus making no interactions with the CoA binding site.
From the protein sequence it might not be excluded that the halophilic dodecins are also capable of binding coenzyme A. The archaeal dodecin sequences contain some of the residues of the CoA binding site (Lys-6, Arg-28, Thr-32, Lys-33, Phe-64, Leu-66). However, Arg-34, one of the major constituents of the CoA binding site, is substituted by the negatively charged Asp-34, possibly due to the halophilic characteristics of enhancing the solubility and preventing aggregation in hypersaline solution.
Concluding Remarks-Although the T. thermophilus and the H. salinarum dodecins share the same protein fold and quaternary structure, they differ in their flavin binding characteristics. For the si-si orientation of the FMN dimers in the T. thermophilus dodecin versus re-re orientation in the H. salinarum dodecin, neither the Arg-45 nor the Arg-65 residue alone is responsible as could be expected from the alignment of various dodecin orthologs (Fig. 7). Because most of the residues involved in FMN binding are present in the majority of eubacteria (e.g. Tyr-5, Asp-37, Arg-45, Arg-65), the flavin binding mode realized in the T. thermophilus dodecin seems to be generally widespread. The equivalent residues Phe-5, Gly-45, Glu-65 are highly conserved in all archaeal dodecins known so far, indicating that the binding mode exhibited by the H. salinarum dodecin is the prevalent one in Archaea.
The high binding affinity for lumiflavin and lumichrome in the halophilic dodecin led to the hypothesis that lumichrome is the cognate ligand for dodecins in vivo. Accordingly, it was postulated that the dodecins might serve as a waste-trapping device, protecting the cellular environment from high amounts of phototoxic lumichromes, which are generated by the photoinduced degradation of riboflavin (9). In contrast to the H. salinarum dodecin, the T. thermophilus dodecin binds all flavins with similar binding constants. However, this finding is not necessarily contradictory to a supposed function of dodecins as protection against phototoxic compounds because free riboflavin, FMN, or FAD are known to act as photosensitizers with detrimental effects in vivo (29,30). A scenario for the biological function of dodecins might be that of a flavin trap that gets into action when the cytosolic concentration of free flavin increases, e.g. after heat shock and flavin release from denatured flavoproteins. Notably, dodecins from the mesophiles M. tuberculosis and Streptomyces coelicolor exhibit a similar thermal stability as the T. thermophilus dodecin. 4 In this case, the observed retarded flavin binding by apododecin may allow maintaining low resting concentrations of free flavins under nonstressed conditions before switching the dodecin to an efficient flavin scavenger.
Besides the flavin dimers, the T. thermophilus crystal structure also shows the binding of coenzyme A in a trimeric form. Accordingly, dodecins comprise cofactor binding motifs for two pathway-related cofactors, e.g. both cofactors are involved in the synthesis of fatty acids, which are bound in an inert manner either for cofactor storage or to overcome cellular stress caused by transient high cofactor concentrations.