DC3, the Smallest Subunit of the Chlamydomonas Flagellar Outer Dynein Arm-docking Complex, Is a Redox-sensitive Calcium-binding Protein*

The outer dynein arm-docking complex (ODA-DC) tar- gets the outer dynein arm to its correct binding site on the flagellar axoneme. The Chlamydomonas ODA-DC contains three proteins; loss of any one prevents normal assembly of the outer arm, leading to a slow, jerky swimming phenotype. We showed previously that the smallest ODA-DC subunit, DC3, has four EF-hands (Casey, D. M., Inaba, K., Pazour, G. J., Takada, S., Wakabayashi, K., Wilkerson, C. G., Kamiya, R., and Witman, G. B. (2003) Mol. Biol. Cell 14, 3650–3663). Two of the EF-hands fit the consensus pattern for calcium binding, and one of these contains two cysteine residues within its binding loop. To determine whether the predicted EF-hands are functional, we purified bacterially expressed wild-type DC3 and analyzed its calcium-binding potential in the presence and absence of dithiothreitol and Mg 2 (cid:1) . The protein bound one calcium ion with an affinity ( K d ) of (cid:1) 1 (cid:2) 10 (cid:3) 5 M . Calcium binding was observed only in the presence of dithiothreitol and thus is redox-sensitive. DC3 also bound (25). To determine the average flagellar beat frequency change induced by photoshock, cells were imaged using a (cid:4) 40 objective (instead of the usual (cid:4) 10 objective) and at low cell density to permit fast Fourier transform analysis of individual cells. Forward swimming cells were initially viewed and analyzed using red light; photoshock was then induced by removal of the red filter from the observation beam, and the beat frequency during the resulting tran-sient backward swimming was determined. About 20 cells were used to calculate the average flagellar beat frequency before and during photoshock for each strain. Assays— Cells that had been dark-adapted for 3 h were placed in a custom phototaxis chamber (26) that was mounted on the stage of a Zeiss universal microscope (Carl Zeiss, NJ). Ran-domly swimming cells were initially viewed with dim red illumination to prevent light-induced behaviors. Photoshock was assayed by moni-toring the behavior of a population of cells in response to an intense flash of light (a Vivitar 283 camera flash). The response of the cells was recorded on videotape by capturing images with a CCD camera mounted on the microscope. Cells were scored as having a normal photoshock response if, after a flash of intense light, they stopped forward swimming, briefly swam backwards, and then resumed forward swimming.

Intracellular calcium plays an important role in the control of eukaryotic cilia and flagella. For example, calcium signals control the beat frequency of mammalian airway cilia (1), the hyperactivation of mammalian sperm flagella (2), the reversal of ciliary beating in ciliates (3), and phototaxis and photoshock in the biflagellate alga Chlamydomonas reinhardtii (for review see Ref. 4). Despite the important role that calcium plays in these processes, little is known about the biochemical pathways or the calcium-binding proteins involved. A few Ca 2ϩbinding proteins, e.g. calmodulin (5,6), centrin/caltractin (7), and dynein light chain 4 (see Ref. 8 and see below), have been identified within the axoneme, but their precise functions in Ca 2ϩ -mediated flagellar behavior have not yet been clearly defined. Given the complexity of the axoneme and the likely need to modulate the activity of multiple axonemal structures to achieve a coordinated change in waveform in response to changes in Ca 2ϩ concentration, it is probable that the axoneme contains additional Ca 2ϩ -binding proteins. Identification and characterization of these proteins will be essential for understanding the molecular basis for the Ca 2ϩ control of ciliary and flagellar waveform.
Chlamydomonas normally swims forward by means of a "breaststroke"-like beat pattern in which the flagella propagate asymmetrical bends. The photoshock response, which is induced by a strong step-up light stimulus and mediated by an increase in intraflagellar Ca 2ϩ from Յ10 Ϫ6 to 10 Ϫ5 -10 Ϫ4 M, consists of an abrupt cessation of forward swimming, followed by a brief period of backward swimming during which the flagella propagate symmetrical bends, and finally a return to the normal asymmetrical bending pattern and forward swimming (4). The generation of symmetrical bends during photoshock requires the outer dynein arms (9,10), suggesting that the outer arms are regulated by Ca 2ϩ . One outer dynein arm light chain, LC4, 1 is a Ca 2ϩ -binding protein, and it has been hypothesized that this subunit may have a role in mediating outer arm function during photoshock (8); however, there currently is no experimental evidence in support of such a role for LC4.
In the axoneme, the outer arms are in direct contact with the outer dynein arm-docking complex (ODA-DC) (11), which targets the outer arm to its correct binding site on the doublet microtubule (12). The ODA-DC contains three subunits, termed DC1, DC2, and DC3, with masses of 83, 62, and 21 kDa, respectively. DC1 and DC2 are predicted to be long coiled-coil proteins that have a structural role in assembling the ODA-DC, and loss of either of these results in complete loss of the outer arms and a slow, jerky swimming phenotype (11,13). Loss of DC3 causes partial loss of the outer arms and a phenotype intermediate between those of DC1 or DC2 mutants and wild type (14). Interestingly, sequence analysis of DC3 indicates that it may be a Ca 2ϩ -binding protein (see Ref. 14 and see below). Moreover, mutants lacking DC3 are unable to swim backwards during photoshock, despite having a partial complement of outer arms (14). This raises the question of whether DC3 is a Ca 2ϩ -binding protein and, if so, if it is involved in this Ca 2ϩ -mediated flagellar response.
Many calcium-binding proteins, including calmodulin (CaM), troponin C (TnC), and centrin, share a common calcium-binding motif called an EF-hand. EF-hand motifs consist of two ␣-helices (helix E and helix F, respectively) separated by a 12-residue loop. The loop of the helix-loop-helix motif forms the calcium-binding pocket. Loop residues 1, 3, 5, 7, 9, and 12 coordinate the calcium ion; their positions within the loop are designated X, Y, Z, ϪY, ϪX, and ϪZ, respectively (Fig. 1). The last coordinating residue (ϪZ) is highly conserved and provides two oxygens for liganding calcium.
Analysis of the predicted amino acid sequence and secondary structure of DC3 revealed four EF-hand motifs (EF1 to EF4, starting at the N terminus; Fig. 2), two of which (EF2 and EF4) fit the consensus pattern for calcium binding: , where X indicates any amino acid, parentheses indicate either/or, and braces indicate any amino acid except those shown (PROSITE accession number PS00018). DC3 is a novel member of the CTER (calmodulin, troponin C, essential and regulatory light chains of myosin) group of EF-hand proteins (14). DC3 is unique among CTER proteins in that one of its EF-hands contains two cysteines, which may be capable of forming a disulfide bond at the potential Ca 2ϩ -binding site.
To determine the in vitro properties of the EF-hands of DC3, we expressed DC3 in bacteria (fused to the C terminus of GST), and we assayed its Ca 2ϩ /Mg 2ϩ binding ability in the presence and absence of a dithiol reductant using an in vitro binding assay. The results indicate that DC3 binds one Ca 2ϩ with a K d of ϳ1 ϫ 10 Ϫ5 M and binds Mg 2ϩ with a much lower affinity; this Ca 2ϩ binding activity is redox-sensitive. To determine the in vivo function of the EF-hands of DC3, we used site-directed mutagenesis to inactivate them and then transformed the mutated gene into a DC3-null strain that assembles only some of the outer dynein arms, swims slowly, and does not generate symmetrical bends during photoshock. We then assayed the swimming speed, number of outer arms, and ability to undergo photoshock of the transformant and compared it to a DC3-null strain transformed with the wild-type DC3 gene. Both the mutant and wild-type genes restored all parameters tested to wild-type levels.
Growth Media-The following media were used: M medium I (17) modified to contain 0.0022 M KH 2 PO 4 and 0.00171 M K 2 HPO 4 and supplemented with 0.0075 M sodium acetate (referred to as R medium in this study); R ϩ Arg (R medium supplemented with 50 g/ml arginine); and TAP (18).
Site-directed Mutagenesis-Starting with the plasmid clone pDC3S-2 (14), which contains the intronless wild-type DC3 (ODA14) gene, two sequential rounds of mutagenesis (Chameleon™ Site-directed Mutagenesis Kit, Stratagene, La Jolla, CA) were used to create point mutations in the DC3 gene. One primer (E74Q, 5Ј-CTGCATCAGCCT-TCTACAATTCCAGACGCTATAC-3Ј) was designed to inactivate the first consensus EF-hand by changing the highly conserved glutamate residue in position 12 of the calcium-binding loop to a glutamine. After sequencing (DNA Sequencing Facility, Iowa State University, Ames) to confirm that the first consensus EF-hand had been mutated, a second primer (E152Q, 5Ј-CCTCACGCTCACGCAGTTCCTGCACTGCTTGC-3Ј) was used to make a similar Glu to Gln mutation in the second consensus EF-hand. Sequencing confirmed the presence of the second mutation. The construct was named pDC3Qx2-9.
Bacterial Expression of Wild-type and Mutant DC3-The DC3 coding region of either pDC3S-2 or pDC3Qx2-9 was amplified by the PCR using the primers DCFPF-1 (5Ј-CGCGGATCCATGGCGAGTGCC-3Ј) and DCFPR-1 (5Ј-CCGGATCCTCACTTGCGCTTCTG-3Ј). Amplification with these primers resulted in a 5Ј BamHI site directly upstream of the DC3 start codon and a 3Ј BamHI site directly downstream of the stop codon. PCR products were purified using the QIAquick PCR Purification Kit (Qiagen, Valencia, CA), digested with BamHI, and ligated into BamHI-digested and alkaline phosphatase-treated pGEX-6P-1 (Amersham Biosciences). This construct was designed to express DC3 fused to the C terminus of glutathione S-transferase (GST). The wild-type fusion protein is termed "GST-DC3" and the EF-hand mutant fusion protein is termed "GST-DC3-(E74Q,E152Q)." The constructs were transformed into Escherichia coli strain BL21 (Amersham Biosciences), and fusion protein expression was induced with either 0.1 mM (GST-DC3-(E74Q,E152Q)) or 0.3 mM (GST-DC3) isopropyl-␤ϪD-thiogalactopyranoside for 2 h at room temperature. A lower concentration of isopropyl-␤ϪD-thiogalactopyranoside was used to induce the mutant protein to aid in its solubility (according to The Recombinant Protein Handbook from Amersham Biosciences). After induction, the recombinant protein was released from the bacterial cells using a French® press (Thermo Spectronic, Rochester, NY). The bacterial lysate was centrifuged, and the soluble fraction was applied to a glutathione-Sepharose 4B column (Amersham Biosciences) at 4°C; the affinity-purified fusion protein was eluted with 50 mM Tris-HCl, pH 8.0, 10 mM reduced glutathione, 1 mM dithiothreitol (DTT).
The size and purity of the recombinant protein was determined by SDS-PAGE. Protein concentration was determined with the bicinchoninic acid assay (Pierce) using bovine serum albumin as standard. Prior to determination of protein concentration, the fusion protein was dia- Asterisks indicate residues that precisely fit the consensus pattern for calcium binding (PROSITE accession number PS00018). The four cysteines are underlined. The single tryptophan is double underlined. The two glutamates that were mutagenized are dark and underlined. Numbers at the beginning and end of each EF-hand indicate amino acid numbers.
lyzed extensively at 4°C against 20 mM Tris-HCl, pH 7.5, 150 mM NaCl using a Slide-A-Lyzer® Dialysis Cassette (Pierce) with a molecular mass cut-off of 10,000 Da to remove reduced glutathione and DTT present in the column elution buffer (DTT interferes with the bicinchoninic acid assay). In some instances, the GST moiety was removed from the fusion protein by adding DTT to 1 mM and cleaving with PreScis-sion™ Protease (Amersham Biosciences) followed by glutathione-Sepharose 4B column chromatography. The protease is GST-tagged and was therefore retained by the column; DC3 was collected in the flow through. Before determining the protein concentration of DC3 alone, it was dialyzed extensively against 20 mM Tris-HCl, pH 7.5, 150 mM NaCl using a Slide-A-Lyzer® Dialysis Cassette with a molecular mass cut-off of 3,500 Da to remove DTT present in the cleavage buffer. Unless indicated otherwise, DTT was added to 1 mM before using the protein in Ca 2ϩ binding assays.
Electroporation Of Recombinant DC3 into Chlamydomonas Cells-Introduction of recombinant DC3 into DC3-null cells was done essentially as described by Hayashi et al. (19). Briefly, cell walls were removed from the DC3 deletion strain, V06, using Chlamydomonas autolysin (20). After cell wall removal, the cells were washed several times in HMDKCaS (30 mM HEPES, pH 7.4, 5 mM MgSO 4 , 1 mM DTT, 50 mM CH 3 COOK, 1 mM (CH 3 COO) 2 Ca, 60 mM sucrose) and resuspended to 1 ϫ 10 8 cells/ml in the same buffer. The cell suspension and a 2 M DC3 solution made in HMDKCaS were mixed in a 1:1 volume ratio. A 125-l aliquot of the mixture was transferred to an electroporation cuvette (model 620, BTX, San Diego) and incubated at 15°C for a total of 15 min after cell washing. An electric pulse of 350 V was then applied to the mixture using an ECM600 electroporation apparatus (BTX) with a resistance of 24 ohms and a conductance of 600 microfarads. The overall time constant used for each sample was around 12 ms. After electroporation, the cuvette was incubated at room temperature for 30 min with shaking every 5 min. The cells were then washed several times in TAP medium supplemented with 60 mM sucrose (TAP ϩ S), resuspended in 125 l TAP ϩ S, and left at room temperature for 2-3 h to allow the cells to recover.
Ca 2ϩ Binding Assay-Ca 2ϩ binding assays were performed at room temperature by the ultrafiltration method of Ladant (21), with minor modifications. Briefly, 1 ml of a 10 -20 M protein solution (either GST-DC3, GST-DC3-(E74Q,E152Q), or DC3 alone) in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, with or without 1 mM DTT was placed in the upper chamber of a Centricon Centrifugal Filter Device with a nominal molecular mass cut-off of 10,000 Da (YM-10; Millipore Corp., Bedford, MA). To reduce nonspecific protein binding, the filter unit was pretreated with 5% Tween 20 according to the manufacturer's instructions. Twenty l of a 500 M CaCl 2 solution containing ϳ2 Ci of 45 Ca 2ϩ (PerkinElmer Life Sciences) was added to the protein solution and mixed, and 20 l was removed for scintillation counting of total radioactivity (i.e. bound ϩ unbound 45 Ca 2ϩ ). Next, the filter assembly was centrifuged (IEC model HN-SII, 2500 rpm for ϳ1 min) until 35-40 l had filtered through. The filtrate was collected and added back to the top compartment, mixed, and centrifuged as before. Twenty l of the second filtrate was taken for scintillation counting of unbound 45 Ca 2ϩ . Samples were always collected from the second filtrate to ensure an accurate measure of the unbound Ca 2ϩ at each unique total Ca 2ϩ concentration (as opposed to a measure of the unbound Ca 2ϩ remaining in the filter from the previous data point). Thereafter, 40 l of 250 M non-radioactive CaCl 2 was added to the top compartment, and the above procedure was repeated until the desired range of calcium concentrations was covered. The ratio of radioactivity in the filtrate to radioactivity in the top compartment (i.e. unbound calcium/total calcium) was used to deduce the amount of calcium bound per molecule of protein at each total Ca 2ϩ concentration. The Ca 2ϩ affinity (reported as a dissociation constant, K d ) was derived from a Scatchard plot of [Ca 2ϩ ] bound /[Ca 2ϩ ] free versus [Ca 2ϩ ] bound . To determine whether the observed metal binding was Ca 2ϩ -specific, MgCl 2 (1 or 20 mM) was included in some assays.
Tryptophan Fluorescence Spectra-The effect of Ca 2ϩ and Mg 2ϩ on the tryptophan fluorescence of GST-DC3 was analyzed using a SPEX FluoroLog-3 spectrofluorometer (Jobin Yvon, Inc., Edison, NJ). Trp fluorescence spectra were recorded before and after the addition of either 20 mM Mg 2ϩ or 1 mM Ca 2ϩ to 1 M GST-DC3 in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM DTT. The spectra were recorded at 22°C with an excitation wavelength of 295 nm.
Transformation of Chlamydomonas-By using the silicon carbide fiber method of Dunahay (22), we co-transformed strain 026.2, an arg7 derivative of oda14-1, with the pARG7.8 plasmid containing the cloned argininosuccinate lyase (ARG7) gene (16), and plasmid clones containing the wild-type or modified ODA14 gene. ARG7 transformants were selected on solid R medium. Individual transformants were picked into liquid R medium and screened by light microscopy for restored wildtype forward motility.
Electron Microscopy-Flagellar axonemes were isolated from cells grown in TAP medium (18) by using the method of Witman (23) and prepared for electron microscopy as described by Kamiya (24).
Swimming Velocity and Flagellar Beat Frequency Measurements-Swimming velocity was measured as described by Kamiya (24). Briefly, motile cells were imaged using dark-field microscopy, and their swimming behavior was recorded with a CCD camera and VCR. Images from the videotape were "captured" using motion-capture software (Motion Capture, InterQuest, Osaka, Japan) at a rate of 10 frames/s. The video clips were then played on a computer monitor, and representative cells to be tracked were chosen by the operator, and their positions as a function of time were determined by cell-tracking software (Image Tracker PTV, InterQuest). The swimming velocity of ϳ30 cells was used to calculate the average swimming velocity for each strain. For photokinesis experiments, cells were dark-adapted for 2 h and then observed with or without 5 min of pre-illumination with white fluorescent light. Cells without pre-illumination were imaged using red light (630-nm cut-off filter), and those with pre-illumination were imaged using white light. Again, the average swimming velocity for each strain was determined from measurements of ϳ30 cells.
The average flagellar beat frequency of a population of forward swimming cells was determined by fast Fourier transform analysis, exactly as described by Kamiya (25). To determine the average flagellar beat frequency change induced by photoshock, cells were imaged using a ϫ40 objective (instead of the usual ϫ10 objective) and at low cell density to permit fast Fourier transform analysis of individual cells. Forward swimming cells were initially viewed and analyzed using red light; photoshock was then induced by removal of the red filter from the observation beam, and the beat frequency during the resulting transient backward swimming was determined. About 20 cells were used to calculate the average flagellar beat frequency before and during photoshock for each strain.
Photoshock Assays-Cells that had been dark-adapted for 3 h were placed in a custom phototaxis chamber (26) that was mounted on the stage of a Zeiss universal microscope (Carl Zeiss, Thornton, NJ). Randomly swimming cells were initially viewed with dim red illumination to prevent light-induced behaviors. Photoshock was assayed by monitoring the behavior of a population of cells in response to an intense flash of light (a Vivitar 283 camera flash). The response of the cells was recorded on videotape by capturing images with a CCD camera mounted on the microscope. Cells were scored as having a normal photoshock response if, after a flash of intense light, they stopped forward swimming, briefly swam backwards, and then resumed forward swimming.

Bacterially Expressed DC3 Is Biologically Functional-DC3
was expressed in E. coli as a C-terminal fusion with GST. A protein of the appropriate mass (ϳ47 kDa) was produced and purified; SDS-PAGE analysis of the purified protein is shown in Fig. 3A. To determine whether bacterially expressed DC3 was biologically functional, the GST tag was removed by digesting the fusion protein with PreScission TM protease (Amer- DC3 Is a Redox-sensitive Calcium-binding Protein sham Biosciences), and the free DC3 was purified on a glutathione-Sepharose 4B column (Amersham Biosciences) (Fig.  3B). The purified DC3 was then electroporated into the DC3null strain, V06. Because they lack DC3, V06 cells assemble a reduced number of outer arms and consequently have a lower flagellar beat frequency and slower swimming speed than wildtype cells (14).
During electroporation, most cells lose their flagella; this deflagellation is essential for efficient protein delivery (19). Flagellar regeneration after electroporation takes about 3 h (19). Therefore, the cells were viewed with dark-field microscopy 3 h after electroporation, and their movement was recorded using a microscope-mounted CCD camera and a VCR. By using motion-capture/cell-tracking software, we measured the average swimming velocity of wild-type cells, V06 cells, and As an additional test of rescue, we measured the average beat frequency of wild-type cells, V06 cells, and V06 cells electroporated in the presence of 1 M DC3. Wild-type cells had an average beat frequency of 61 Hz, whereas V06 had an average beat frequency of 37 Hz. V06 electroporated in the presence of 1 M DC3 had an average beat frequency of 59 Hz. Based on these data, we conclude that recombinant DC3 is biologically functional and supports the assembly of outer arms onto outer doublet microtubules.
DC3 Is a Ca 2ϩ -binding Protein-DC3 has two consensus EF-hands (Fig. 2), indicating that it may bind Ca 2ϩ . By using the ultrafiltration method of Ladant (21), we determined that GST-DC3 indeed bound ϳ1 mol of calcium/mol of protein in the presence of DTT (Fig. 4A). Similar results were obtained with DC3 after removal of the GST tag, demonstrating that calcium binding is an intrinsic property of DC3 rather than a nonspecific effect due to the GST moiety (data not shown). Scatchard analysis of the GST-DC3 Ca 2ϩ -binding data indicated that the calcium affinity of DC3 (expressed as a dissociation constant, K d ) is 1.2 ϫ 10 Ϫ5 M (Fig. 4B); similar results were obtained with DC3 alone (K d ϭ 1.3 ϫ 10 Ϫ5 M). This is in the range expected to be physiologically relevant for photoshock (see Introduction).
The Ca 2ϩ Binding Activity of DC3 Is Redox-sensitive-DC3 contains four cysteine residues (underlined in Fig. 2), two of which (Cys-66 and Cys-69) are within the first consensus calcium-binding loop. Cys-69 is one of the six loop residues (position ϪY) that coordinates the Ca 2ϩ ion (see Introduction). If Cys-66 and Cys-69 are in sufficiently close proximity, they might form a disulfide bond that could block Ca 2ϩ binding. To test this, we examined the Ca 2ϩ binding activity of recombinant GST-DC3 in dialysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl) without the addition of DTT. Under these conditions, the Ca 2ϩ binding activity of the protein was completely eliminated (Fig. 4A). Addition of 1 mM DTT restored Ca 2ϩ binding activity to the protein. Inasmuch as molecular modeling predicts that Cys-66 and Cys-69 are the only two cysteine residues in close proximity (14), these results strongly suggest that they can form a redox-sensitive disulfide bridge that prevents Ca 2ϩ binding. These data also suggest that the first consensus Ca 2ϩbinding loop is responsible for the in vitro Ca 2ϩ binding activity of DC3.
DC3 Binds Mg 2ϩ -In 1980, Reid and Hodges (27) proposed the acid pair hypothesis that correlated Ca 2ϩ affinity with the location and number of paired, negatively charged chelating residues. They predicted the highest Ca 2ϩ affinity will occur when an EF-hand has acidic residues in loop positions X, ϪX, Z, and ϪZ. Studies using synthetic EF-hand peptides have indicated that Mg 2ϩ will bind to those motifs containing 3 or 4 acidic residues in chelating positions, as long as two of the acidic residues are in Z and ϪZ positions (a "Z-acid pair") (28,29). Subsequently, Tikunova and colleagues (30) demonstrated that the presence of a Z-acid pair in the first EF-hand of CaM was essential for Mg 2ϩ binding with a physiologically relevant affinity.
Because EF2 has 4 acidic residues in chelating positions, two of which form a Z-acid pair (Fig. 2), DC3 could potentially be a Ca 2ϩ -and Mg 2ϩ -binding protein. To investigate this possibility, we performed the calcium binding assay in the presence of physiological (1 mM) concentrations of Mg 2ϩ . Under these conditions, the number of calcium ions bound per molecule of GST-DC3 decreased, but calcium binding was not eliminated (Fig. 5). Increasing the Mg 2ϩ concentration to 20 mM further lowered the number of calcium ions bound per GST-DC3 but  (21). Representative data sets for GST-DC3 Ca 2ϩ binding in the presence (squares) and absence (circles) of 1 mM DTT are shown. Under reducing conditions, GST-DC3 binds at least 1 mol of Ca 2ϩ /mol of protein between 10 Ϫ5 and 10 Ϫ4 M free Ca 2ϩ . In contrast, under non-reducing conditions, essentially no Ca 2ϩ is bound. B, the GST-DC3 Ca 2ϩ binding data were transformed to make a linear graph (Scatchard plot). The x axis is specific binding, and the y axis is the ratio of specific binding to concentration of free ligand. K d is the negative reciprocal of the slope.

DC3 Is a Redox-sensitive Calcium-binding Protein
still did not eliminate binding. These data indicate that the metal ion-binding site can bind either Ca 2ϩ or Mg 2ϩ . However, because Ca 2ϩ binding is not completely abolished even in the presence of ϳ200-fold excess Mg 2ϩ , the affinity of DC3 for Ca 2ϩ is much higher than its affinity for Mg 2ϩ .
Effect of Ca 2ϩ or Mg 2ϩ on the Fluorescence Spectra of DC3-Some EF-hand proteins that bind calcium, such as CaM, change their shape and thereby regulate cellular processes. Others bind calcium but do not change shape; these proteins may be involved in Ca 2ϩ buffering or anchoring protein complexes (31). Single tryptophan residues in Ca 2ϩ -binding proteins are often used as intrinsic fluorescent probes to determine protein structural changes and substrate binding. Among the properties used are changes in the fluorescence intensity and wavelength maximum ( max ). In order for the information they provide to be meaningful, Trp residues must be located close to the metal ion-binding site. Because DC3 contains only one Trp residue (double underlined in Fig. 2), which is located in the C-terminal part of EF2 helix E, we measured the Trp fluorescence of GST-DC3 in the absence and in the presence of 1 mM CaCl 2 or 20 mM MgCl 2 . The maximal Trp fluorescence intensity decreased in the presence of either metal (Fig. 6), supporting the above finding that DC3 binds Ca 2ϩ and providing further evidence that it also binds Mg 2ϩ . DC3 underwent an ϳ1.1-fold decrease in its maximal Trp fluorescence intensity upon Mg 2ϩ binding; binding of Ca 2ϩ caused an ϳ1.2-fold decrease. The magnitude of the intensity change upon binding of either metal suggests local conformational changes within the Ca 2ϩ /Mg 2ϩbinding loop (32). Binding of either metal caused a small 1-nm red shift in the max of fluorescence emission from 345 to 346 nm (Fig. 6). The shift to a slightly longer wavelength suggests that Trp-60 is in a more polar environment when DC3 is in its metal-bound state.
Site-directed Mutagenesis Abolishes DC3 Ca 2ϩ Binding-The results of our calcium binding assays indicated that DC3 contains at least one functional Ca 2ϩ -binding site. To determine whether DC3 was binding calcium through either of its consensus sites, we introduced a point mutation into both EF2 and EF4 to change the highly conserved glutamate residue at position ϪZ (Fig. 2, dark and underlined) to a glutamine. This Glu to Gln mutation has been shown to abolish or greatly reduce the calcium binding of other EF-hand proteins (for review see Ref. 33). We expressed the mutant protein as a C-terminal fusion with GST; SDS-PAGE analysis of the bacterially expressed, affinity-purified mutant protein, which we named GST-DC3-(E74Q,E152Q), is shown in Fig. 7. The purified protein was of the expected mass (ϳ47 kDa). When the calcium binding potential of the modified protein was tested using the same ultrafiltration method as was used for the wild-type protein, we found that the Glu to Gln mutations completely abolished DC3 calcium binding (Fig. 8). Therefore, either EF2 or EF4 must mediate the in vitro calcium binding ability of wild-type DC3.
The EF-hand Mutant Gene Restores Outer Arms to a DC3-Null Strain-Because they are missing ϳ50% of their outer dynein arms, Chlamydomonas cells lacking DC3 have a reduced beat frequency and swim in a jerky manner at about one-half the speed of wild-type cells (14). We showed previously that nuclear transformation with the wild-type DC3 gene restores outer dynein arms, wild-type motility, and a wild-type photoshock response (14). To determine the role of the calcium binding activity of DC3 in vivo, we transformed a DC3-null strain with a DC3 gene containing the same Glu to Gln mutations that abolished the calcium binding activity of DC3 in vitro. If the calcium affinity in vivo of DC3 is similar to its calcium affinity in vitro (ϳ10 Ϫ5 M), the EF-hand mutant protein should not affect either ODA-DC or outer arm assembly, because flagellar assembly in Chlamydomonas presumably occurs at a steady-state calcium concentration of ϳ10 Ϫ8 M (4).
Indeed, transformation of the DC3-null strain with the modified gene resulted in the complete restoration of outer dynein arms (Fig. 9). Moreover, the EF-hand mutant gene restored wild-type swimming in 29% of the transformants screened (45/154), a percentage comparable to that obtained by transformation with the wild-type gene (24/96 or ϳ25%). (Because transformed cells were selected on the basis of co-transformation with a selectable marker (ARG7; see "Experimental Procedures"), it is expected that only some of the selected strains would express DC3.) To verify our qualitative conclusion that the mutant gene had restored wild-type motility to the DC3-null strain, we quantitatively measured its swimming speed and beat frequency and compared them to those of the original DC3-null strain and a DC3-null strain transformed with the wild-type gene. As reported above, DC3-null cells swim with an average velocity of ϳ130 m/s and have an average beat frequency of ϳ37 Hz. In contrast, DC3-null cells transformed with the wildtype gene swam with an average velocity of 203 m/s and had an average beat frequency of 55 Hz. Similarly, DC3-null cells transformed with the EF-hand mutant gene swam with an average velocity of 206 m/s and had an average beat frequency of 52 Hz. Taken together, these data indicate that Glu to Gln mutations in the consensus EF-hands of DC3 do not affect the assembly of the ODA-DC or the outer dynein arm or the generation of normal forward motility.
Photoshock Is Restored to Normal in the EF-hand Mutant Strain-As briefly reviewed in the Introduction, the switch from an asymmetrical to a symmetrical waveform during photoshock is dependent on an increase in intraflagellar Ca 2ϩ from ϳpCa 6 to 4 (4). Strains lacking DC3 have an altered photoshock response in that they stop but are unable to generate effective backward swimming (14). Transformation of the DC3null strain with the wild-type DC3 gene restores not only outer dynein arms but also the photoshock response (14). Therefore, the altered photoshock response of the DC3-null mutant could be due to the incomplete number of outer arms, or it could be because DC3 is missing and unable to signal an increase in intraflagellar Ca 2ϩ to the outer arm. The in vitro calcium affinity of DC3 (ϳ10 Ϫ5 M) is within the appropriate range for detecting the rise from pCa 6 to 4 that induces the change in flagellar waveform. Thus, inasmuch as the ODA-DC is in direct contact with dynein (11), DC3 could be the calcium sensor for photoshock. To investigate this possibility, we tested the photoshock response of the DC3-null strain transformed with the EF-hand mutant gene. Photoshock appeared to be restored to normal, indicating that DC3 is unlikely to be the sole flagellar calcium sensor for photoshock and furthermore that the inability of the DC3-null strain to generate backward swimming is probably due to the incomplete number of outer dynein arms.
Beat Frequency Is Normal in Backward Swimming EF-hand Mutant Cells-During the photoshock response, the switch from forward to backward swimming is accompanied by an increase in flagellar beat frequency (34,35). Because this subtle change might have been missed in our qualitative analysis of photoshock (see above), we measured the flagellar beat frequency in the wild-type and mutant strains before and during the photoshock response. The beat frequency of wild-type cells was 62 Hz during forward swimming and increased to 84 Hz when the cells were induced to swim backward during photoshock. The beat frequency of DC3-null cells rescued with the wild-type DC3 gene was 59 Hz during forward swimming and increased to 83 Hz during backward swimming. The beat frequency of EF-hand mutant cells was 58 Hz during forward swimming and increased to 80 Hz during backward swimming. Therefore, the EF-hand mutations in DC3 do not prevent the increase in flagellar beat frequency that occurs as calcium rises into the 10 Ϫ4 M range during photoshock.
Photokinesis Is Normal in the EF-hand Mutant Strain-Photokinesis is a phenomenon in which the swimming speed of a microorganism varies as a consequence of the light stimulus intensity. Although this type of photoresponse has not been studied in detail in Chlamydomonas, Moss and Morgan (36) report that light-adapted wild-type cells swim faster than darkadapted cells, and Pazour et al. (15) have shown that the speed of wild-type cells increases as the intensity of the light stimulus increases. Interestingly, Moss and Morgan (36) report that photokinesis does not occur in strains that lack outer dynein arms. These observations suggest that a light-activated signal transduction cascade is acting at the level of the outer dynein arms to increase cell motility. A light-activated signal transduction cascade might use cytoplasmic reducing equivalents (i.e. electrons) to transmit signals because their levels change when Chlamydomonas is in the light (37).
Because the ODA-DC is in direct contact with dynein (11), and DC3 calcium binding is redox-sensitive, the Ca 2ϩ binding activity of DC3 may play a role in photokinesis. To investigate this possibility, we measured the average swimming velocity of dark-adapted versus light-adapted wild-type cells, DC3-null cells, DC3-null cells rescued with the wild-type DC3 gene, and EF-hand mutant cells. As DC3 calcium binding in vitro is abolished by mutation of the consensus EF-hands, DC3 in the EF-hand mutant strain presumably is constitutively "off," i.e. cannot bind Ca 2ϩ regardless of the redox state of the cell, and will not be able to mediate photokinesis if the latter depends on

DISCUSSION
In the present study, we have demonstrated that DC3, the 21-kDa subunit of the ODA-DC, has calcium binding activity. DC3 is the fourth CTER calcium-binding protein to be identified in the flagellum of Chlamydomonas. The other three CTER proteins are CaM, centrin, and LC4. CaM is an integral axonemal component (5,6) and is associated with the radial spokes (38). Centrin is a subunit of some inner dynein arms (7). LC4 is a subunit of outer dynein arms (8). The presence of multiple calcium-binding proteins each associated with a discrete axonemal structure suggests that there are multiple sites of calcium control in the flagellum. This is consistent with the complex architecture of the axoneme, which contains multiple structures that undoubtedly must respond in a coordinated manner to changes in intraflagellar Ca 2ϩ .
Although DC3 has four EF-hand motifs, only EF2 and EF4 fit the consensus pattern for calcium binding. EF1 has a lysine residue in position Y of the calcium-binding loop. According to the PROSITE consensus pattern for calcium binding (see Introduction), only aspartate, asparagine, and serine are tolerated at this position, so EF1 is not expected to bind calcium. EF3 has two residues that deviate from the consensus, one of which is in a calcium-chelating position (position Z). Furthermore, molecular modeling of DC3 (14) predicts that EF3 forms a ␤-hairpin that would most likely disrupt the coordination geometry of the loop and prevent calcium binding. Our in vitro calcium binding assays indicate DC3 has at least one functional calcium-binding site. Site-directed mutagenesis of EF2 and EF4 completely abolished DC3 calcium binding. These data indicate that either EF2 or EF4 is responsible for the in vitro binding activity of DC3. Based on the redox sensitivity of DC3 calcium binding, EF2 is the more likely candidate because it contains three cysteine residues, two of which are located within the calcium-binding loop.
Many EF-hand proteins such as CaM undergo a global conformational change when they bind Ca 2ϩ but only a local conformational change when they bind Mg 2ϩ (32,39). The difference between the Ca 2ϩ -and Mg 2ϩ -bound conformations can be determined by comparing the Trp emission spectra of the protein. For instance, the Trp fluorescence of CaM increases ϳ2.8fold in the presence of Ca 2ϩ yet only ϳ1.4-fold in the presence of Mg 2ϩ (30). Although the maximal Trp fluorescence intensity of Ca 2ϩ -bound DC3 differed from that of Mg 2ϩ -bound DC3, the intensity change was small (ϳ1.2-fold versus ϳ1.1-fold, see Fig.  6). Furthermore, Ca 2ϩ binding to CaM induces a small blue shift in max , whereas Mg 2ϩ binding induces a small red shift (30). With DC3, the position of max was the same in the presence of either metal. Taken together, these data indicate that although there are clearly two conformational states of DC3 depending on which metal is bound, the structural difference is minor when compared with that of CaM (30).
As mentioned above, a large shift in the intensity of Trp fluorescence occurs when some EF-hand proteins bind calcium and is indicative of extensive conformational changes (32). The slight fluorescence intensity shift of Ca 2ϩ -bound DC3 versus Ca 2ϩ -free DC3 (Fig. 6) suggests that calcium binding to DC3 does not cause major structural effects (32). Other examples of EF-hand proteins that bind calcium yet change shape only slightly include the regulatory domain of cardiac TnC (40,41) and the calcium signal modulator, calbindin D 9k (for review see Ref. 42). These proteins have non-polar amino acids in key regulatory positions, specifically a cysteine (Cys-84) in cardiac TnC and an isoleucine (Ile-73) in calbindin D 9k (43). Proteins that are CaM-like have buried polar residues in the structurally analogous positions. It is believed that these buried polar residues are critical for generating a large conformational change (reviewed in Ref. 42 and see Ref. 43). In fact, replacing the polar lysine residue with a non-polar isoleucine residue (K75I) in the N terminus of CaM prevents the protein from undergoing a large conformational change (43). DC3 has a non-polar cysteine residue (Cys-82) in the structurally analogous position, lending support to our interpretation that metalion binding to DC3 induces only a small conformational change.
In vivo, DC3 is in a complex with DC1 and DC2 (11); therefore, it is conceivable that DC1 and/or DC2 exert a substantial effect on the structure of DC3. For instance, DC3 may significantly change conformation upon DC1 and/or DC2 binding. This mode of structural change would be analogous to that of cardiac TnC, which does not significantly change conformation upon Ca 2ϩ binding yet does significantly change conformation upon TnI binding (44). Binding of TnI to TnC also increases the Ca 2ϩ sensitivity of TnC ϳ10-fold (31). Likewise, the presence of CaM-dependent enzymes or short peptides corresponding to CaM-dependent enzyme-binding sites increases the Ca 2ϩ affinity of CaM (45,46). These data indicate the calcium binding activity of DC3 in complex with DC1 and DC2 may be higher than the calcium binding activity of the purified protein.
The above studies highlight the important role that protein interactions have in determining the conformation and Ca 2ϩbinding properties of EF-hand proteins. This was clearly demonstrated by experiments in which target binding restored Ca 2ϩ binding to Ca 2ϩ -binding site mutants of CaM (47,48). In these studies, the binding of mutant CaMs to peptides representing CaM-binding sites changed the conformation of the mutant CaMs back to that of the wild-type protein so that they could still bind calcium and activate target enzymes. Therefore, although the site-directed mutagenesis of DC3 effectively eliminated calcium binding in vitro (Fig. 8), it is possible that the mutations are suppressed in vivo when DC3 binds to either DC1, DC2, or some other flagellar protein. Moreover, there are several examples of non-canonical EF-hands binding calcium when their structures are stabilized by interacting proteins (reviewed in Ref. 49); this may also be the case with DC3. Regardless of the mechanism, if mutant DC3 binds calcium in vivo, this may explain why no deleterious effect on photoshock was observed in the DC3-null strain that was rescued with the EF-hand mutant gene.
Alternatively, another calcium-binding protein may be the flagellar calcium sensor for photoshock. As mentioned above, CaM, centrin, and LC4 have all been localized to the axoneme. One or more of these proteins (or a yet unidentified calciumbinding protein) acting alone or in concert may initiate photoshock by relaying a message to the outer dynein arms when the intraflagellar calcium concentration rises above pCa 6. The presence of proteins with redundant functions would ensure that the cell could respond to bright light even if one of the other sensor proteins were unable to function.
Our finding that the in vitro calcium binding activity of DC3 is redox-sensitive suggests that DC3 calcium binding in vivo may be subject to regulation. If the cysteine residues in the calcium-binding loop of EF2 are vicinal when DC3 is attached to the axoneme, they could form a disulfide bridge that keeps the calcium-binding loop in a conformation that prohibits accommodation of a calcium ion. For DC3 to bind calcium, the disulfides would have to be reduced. A class of proteins known as thioredoxins can catalyze the formation as well as the reduction of protein disulfide bonds and thereby regulate the functional state of disulfide-containing proteins. The outer dynein arm contains two thioredoxin-related light chains, which are associated with the stem domains of the ␣and ␤-heavy chains, respectively (37,50). Because the ODA-DC directly associates with the outer arm (11), these light chains may be appropriately positioned to interact with and regulate the redox state of DC3. Intriguingly, Harrison et al. (37) recently demonstrated that the ATPase activity of the ␥-heavy chain of the Chlamydomonas outer dynein arm is activated by sulfhydryl oxidation. Thus, DC3 may be one of several flagellar proteins whose function is regulated by redox state.