Domains in the 1α Dynein Heavy Chain Required for Inner Arm Assembly and Flagellar Motility in Chlamydomonas

35 kb of genomic DNA in the region of the Dhc1 gene was recovered from a large insert, wild-type (21gr) Chlamydomonas library. The position of the Dhc1 transcription unit within this region was determined by probing Northern blots of wild-type RNA with selected subclones, and the Dhc1 transcription unit was thereby narrowed down to ∼22 kb of genomic DNA (see Fig. 1; Myster et al. 1997). Sequence information was obtained from both strands of subclones A–D, and G using a series of nested deletions (Erase-a-Base System; Promega Corp.) and Sequenase 2.0 (Amersham Life Science, Inc.) following the manufacturer's instructions. Subclones E and F were sequenced by the DNA Sequencing Facility (Iowa State University) on an ABI Prism sequencer (Perkin Elmer Corp.). The sequence data were assembled using the GCG software package versions 8 and 9 (Genetics Computer Group).

Potential open reading frames were identified using the GCG program CodonPreference and a codon usage table compiled from the coding regions of 73 different Chlamydomonas nuclear sequences (Nakamura et al. 1997; available at http://www.dna.affrc.go.jp/~nakamura/codon.html). Potential splice donor and acceptor sequences within the open reading frames were identified based on splice junction consensus sequences found in Chlamydomonas nuclear genes (Mitchell and Brown 1994; LeDizet and Piperno 1995; Zhang 1996; R. Schnell, personal communication).

In five regions of the Dhc1 gene, the presence of multiple potential splice donor or acceptor sequences did not allow a confident prediction of the putative exons. In those cases, the splice junctions were determined directly by sequence analysis of reverse transcriptase–PCR (RT-PCR) products generated from the Dhc1 transcript (see Fig. 1). Total RNA was isolated from wild-type cells 45 min after deflagellation, and then 5 μg of total RNA was reverse transcribed using either a random primer or a sequence-specific reverse primer and the Superscript Preamplification System (GIBCO BRL) according to manufacturer's instructions. 5 μl of the resulting 25-μl cDNA product was used in a 100-μl PCR reaction with sequence specific primers. PCR reactions were performed using 20 mM Tris-HCl, pH 8.4, 50 mM KCl, 1.5 mM MgCl₂, 2 mM deoxynucleotide triphosphates, 0.2 mM of each primer, and 2.5 U Taq polymerase (Life Technologies, Inc.). Some reactions also contained 3% DMSO. The PCR reactions were first denatured at 94°C for 3 min, followed by 30 cycles of 58°C for 1 min, 72°C for 3 min, and 94°C for 1 min, and then completed with a final cycle of 58°C for 1 min and 72°C for 5 min. The final reaction products were analyzed on agarose gels, and then purified using Wizard PCR preps (Promega Corp.) for direct sequencing with sequence-specific primers.

The proposed translation start site was determined by the recovery of an RT-PCR product using a forward primer downstream of the TATA box sequence and a reverse primer in exon 3. The resulting RT-PCR product contained stop codons in all three frames immediately preceding the proposed start codon.

The predicted amino acid sequence encoded by the Dhc1 gene was analyzed using the GCG program Motifs. The programs Bestfit, Compare, and Pileup were used to compare the 1α Dhc sequence to Chlamydomonas outer arm Dhc sequences α, β, and γ (Mitchell and Brown 1994, Mitchell and Brown 1997; Wilkerson et al. 1994) and the cytoplasmic Dhc from Dictyostelium (Koonce et al. 1992). Regions with the potential to form α-helical coiled coils were identified using the program COILS, version 2.2 (Lupus et al. 1991; Lupus 1996).

To identify clones that might contain a full-length Dhc1 gene, we screened two different Chlamydomonas cosmid libraries (Purton and Rochaix 1994; H. Zhang et al. 1994) that were generously provided by S. Purton (University College, London) and D. Weeks (University of Nebraska). Because the Purton library contains the ARG7 gene within the cloning vector, the cosmid clones can be used to directly transform arg7 strains. 10⁵ independent clones from each library were screened on Magnagraph (Micron Separations, Inc.) nylon membrane lifts in duplicate with probes from the 5′ and 3′ ends of the Dhc1 gene (see Fig. 5 A). Probes used for hybridization were purified in low melting point agarose (GIBCO BRL) and radiolabeled with [³²P]dCTP and random hexamer primers using the Prime It II kit (Stratagene). Conditions for prehybridization and hybridization were as described previously (Porter et al. 1996; Porter et al. 1999; Myster et al. 1997). After single colony isolation, cosmid DNA was purified using alkaline lysis procedures and CsCl gradient centrifugation (Sambrook et al. 1989).

A truncated version of the Dhc1 gene was constructed by fusing sequences from the 5′ end to sequences from the 3′ end. To recover the 5′ end, a 19-kb SalI fragment was subcloned to form the plasmid pSM8 (see Fig. 5 D). pSM8 contains ∼1.7 kb of genomic DNA located 5′ of the coding region, but ends in the middle of the Dhc1 transcription unit. pSM8 was digested with SalI and AscI to release the Dhc1 gene as an 11-kb fragment that is truncated before the region encoding the ATP hydrolytic site (P1). The 3′ end of the Dhc1 gene was subcloned as a 4.3-kb SalI, EcoRI fragment to form the plasmid p14SE, which was digested with SalI and AscI to release the region 5′ of the AscI site. The SalI-AscI fragment from pSM8 was ligated into the digested p14SE subclone to form the construct pD1SA (see Fig. 5 D). pD1SA joins sequences from the 5′ end of the Dhc1 gene to the 3′ end at the AscI site. It is predicted to encode the first 1,956 amino acids of the 1α Dhc, and then terminate translation after adding nine novel amino acids (QCHGCGPGV) to the COOH terminus of the polypeptide.

A modified pBELO BAC library containing Chlamydomonas genomic DNA was screened with selected subclones to identify large insert BAC clones containing the Dhc1 gene. This library was constructed by N. Haas and P. Lefebvre (University of Minnesota, St. Paul, MN) using genomic DNA from the cell-wall less strain cw92 and is currently available from Genome Systems, Inc. BAC DNA was isolated from positive clones using a modified version of the manufacturer's protocol available from C. Amundsen (University of Minnesota, St. Paul, MN) at the following URL: http://biosci.cbs.umn.edu/~amundsen/chlamy/methods/bac.html. The final pellet of BAC DNA was resuspended in 200 μl of TE and stored at −20°C. To identify clones containing full-length Dhc1 genes, 5 μl of BAC DNA was digested with the restriction enzyme SacI and analyzed on Southern blots using subclones from the 5′ and 3′ ends of the Dhc1 gene.

Large-scale preparations of genomic DNA were isolated from wild-type and mutant transformant strains using CsCl gradients as described in Porter et al. 1996. A smaller scale mini-prep procedure (Newman et al. 1990) was used to isolate DNA samples from tetrad progeny and some of the transformants. Restriction enzyme digests, agarose gels, isolation of total RNA, and Southern and Northern blots were performed as previously described (Porter et al. 1996, Porter et al. 1999; Myster et al. 1997).

The strains used in this study are listed in Table. All cells were maintained as vegetatively growing cultures at 21°C as previously described (Myster et al. 1997). The arg2 (Eversole 1956) strains were grown on rich medium that contained reduced ammonium nitrate (one tenth the normal concentration), but was supplemented with l-arginine to 0.6 mg/ml.

The pf9-2 arg2 strain (Porter et al. 1992) was crossed to the outer arm mutant pf28 using standard genetic techniques (Levine and Ebersold 1960; Harris 1989) to obtain the triple mutant pf9-2 pf28 arg2. The pf9-2 pf28 arg2 strain assembles short, immotile flagella, and requires arginine for growth. After growth in tris-acetate phosphate (TAP) media supplemented with 0.6 mg/ml l-arginine, this strain was cotransformed using the glass bead method (Kindle 1990; Nelson et al. 1994) with various constructs of the Dhc1 gene (2–4 μg) and the plasmid pARG7.8, which contains a wild-type copy of the ARG7 (argininosuccinate lyase) gene (Debuchy et al. 1989). After transformation, cells were washed and plated on TAP media lacking arginine to select for arg+ transformants. After 10 d of growth, single arg+ colonies were picked into liquid media and tested for rescue of the flagellar assembly and motility defects.

Positive transformants were picked into 96-well plates and screened for motility on an inverted microscope (Olympus CK). Wells containing motile cells were streaked for single colonies and rescored on a phase-contrast microscope (Axioskop; Carl Zeiss, Inc.) using a 40× objective and a 10× eyepiece. The phenotypes of motile transformants were further analyzed by measuring forward swimming velocities and beat frequencies as previously described (Porter et al. 1992, Porter et al. 1994; Myster et al. 1997).

Transformants were tested for their ability to phototax using two different assays. In the first assay (King and Dutcher 1997), actively swimming cells were put in a dark box with a 3-cm-wide horizontal slit cut out along the bottom such that only the lower portion of a 10-ml suspension was illuminated. The box was placed ∼35 cm from a fluorescent light source for 40 min. Positively phototactic cells would concentrate in the lower, illuminated portion of the tube, whereas phototaxis defective cells would remain uniformly suspended throughout the tube. In the second assay, motile cells were transferred to a 96-well plate and placed on a dissecting microscope with substage illumination. Thick posterboard was placed under the plate and positioned to cover half of the well. The location of the cells within the well was followed over a 60-min time course. Positively phototactic cells would move quickly to the side of the well exposed to light, whereas cells unable to phototax would remain uniformly distributed throughout the well.

To verify that the rescued motility in the transformants was due to expression of the Dhc1 transgene and not a reversion event at the PF9 locus, the motile transformants were backcrossed to a pf28 allele (oda2), which lacks the outer arms but is wild-type at the PF9 locus (Kamiya 1988). Tetrad progeny were recovered following standard genetic methods (Harris 1989), and their motility phenotypes were scored using a phase-contrast microscope as described above.

Axonemes were prepared from large-scale (5–40 liters) liquid cultures of vegetative cells using procedures described by Witman 1986 and S.M. King et al. 1986, as modified by Gardner et al. (1994) and Myster et al. 1997. Crude dynein extracts were obtained by brief (∼30 min) high salt extraction of whole axonemes (Porter et al. 1992; Myster et al. 1997). To isolate I1 complexes, crude dynein extracts were fractionated by sucrose density gradient centrifugation as previously described (Porter et al. 1992; Myster et al. 1997). Aliquots of each fraction were analyzed by SDS-PAGE and Western blotting.

Protein samples from whole axonemes and sucrose gradient fractions were separated on 5% polyacrylamide gels using the Laemmli 1970 buffer system and either stained directly with Coomassie brilliant blue (R250; Sigma Chemical Co.) or transferred to either nitrocellulose (Schleicher and Schull, Keene) or Immobilon-P (Millipore) in 25 mM Tris, 192 mM glycine, and 12.5% methanol at 800 mA for 90 min at 4°C using a Genie electroblotter (Idea Scientific, Co.). Nitrocellulose blots were blocked in 1× PBS (0.58 M Na₂HPO₄, 0.017 M NaH₂PO₄·H₂O, 0.68 M NaCl), 5% normal goat serum (Sigma Chemical Co.), and 0.05% Tween 20 (polyethylenesorbitan monolaurate), whereas the Immobilon-P membrane was blocked in 0.2% I-Block (Tropix) in 1× PBS and 0.5% Tween 20.

Four different antibody preparations were used to probe the blots. The 1α Dhc antibody has been previously described in detail and is highly specific for the Dhc1 gene product (Myster et al. 1997). This antibody was raised against the peptide sequence DGTCVETPEQRGATD, which corresponds to amino acids 1,059–1,073 of the 1α Dhc polypeptide. The 1α Dhc antibody was affinity-purified on Western blots of dynein extracts, and then used at a 1:10 dilution. The IC140 antibody was provided by P. Yang and W. Sale (Emory University, Atlanta, GA). This antiserum was raised against a fusion protein containing a fragment of the 140-kD intermediate chain of the I1 complex (Yang and Sale, 1998), and it was typically used at a dilution of 1:3,000. The rabbit polyclonal antibody R5205, which was raised against a fusion protein of the human 14-kD dynein LC (S.M. King et al. 1996), was provided by S. King (University of Connecticut). The R5205 antibody cross-reacts with the 14-kD LC (Tctex1) of the I1 complex (Harrison et al. 1998) and was used at a dilution of 1:50. After incubation overnight at 4°C, the blots were washed in 1× PBS and 0.05% Tween 20. Immunoreactivity was detected using an alkaline phosphatase–conjugated secondary antibody, BCIP (5-bromo-4-chloro-3-indolyl phosphate), and NBT (nitro blue tetrazolium) following the manufacturer's instructions (Sigma Chemical Co.). An mAb to tubulin (T5168; Sigma Chemical Co.) was used at a dilution of 1:1,000, and then detected using an HRP-conjugated secondary antibody, 4-chloro-1-naphthol, and hydrogen peroxide following the manufacturer's protocol (Sigma Chemical Co.).

To view the I1 complex in strains with rescued motility, selected transformants were crossed to a pf9-3 strain to recover strains with rescued I1 complexes and the wild-type complement of outer dynein arms. Axonemes were prepared and processed for EM as previously described (Porter et al. 1992; Myster et al. 1997). Longitudinal images were selected, digitized, and averaged using the methods described in Mastronarde et al. 1992. Averages of individual axonemes were obtained by analyzing at least six 96-nm radial spoke repeats, and then averages from several axonemes were combined to obtain a grand average for each strain. The methods used to compute differences between two strains are described in detail in Mastronarde et al. 1992.

To identify the 3′ end of the Dhc1 transgene in the G3 transformant (which assembles the shortest 1α Dhc fragment), genomic DNA was isolated from wild-type and G3, digested with the restriction enzymes SacI and KpnI, and analyzed on Southern blots probed with Dhc1 subclones. A polymorphic 7.2-kb SacI-KpnI fragment was identified in G3 using subclone C. This polymorphic fragment was recovered from G3 genomic DNA by constructing a size-selected minilibrary, and then screening the library with subclone C. After single colony purification, the 3′ end of the truncated Dhc1 transgene was sequenced with Dhc1 specific primers to determine the predicted amino acid sequence at the COOH terminus of the 1α Dhc fragment.

In previous work, we identified a null mutation in the Dhc1 gene that resulted in the failure to assemble the I1 inner arm complex into the flagellar axoneme (Myster et al. 1997). To understand how the Dhc1 gene product might contribute to dynein complex formation, we have now sequenced the entire Dhc1 transcription unit (Fig. 1 a). A map of the deduced gene structure is shown in Fig. 1 b. Sequence analysis of subclone A identified the 3′ end of the neighboring gene (geranyl geranyl pyrophosphate synthase), ∼800 bp of intervening sequence, and a TATA box sequence 144 bp upstream of the proposed translation start site of the Dhc1 gene (see Materials and Methods). All of the 5′ sequence elements required for regulated Dhc1 expression should therefore be contained within an ∼1-kb region. The next 20 kb of the sequence contains the coding region located within 29 exons. The 3′ end of the gene is located in subclone G, which contains the last exon encoding the COOH-terminal 552 amino acids, a stop codon, and a consensus polyadenlyation signal sequence (TGTAA) 471 bp downstream.

The predicted amino acid sequence of the encoded 1α Dhc contains 4,625 amino acid residues and corresponds to a polypeptide of 522,806 D (Fig. 2). A search for potential nucleotide binding sites within the 1α Dhc sequence identified six consensus or near consensus phosphate-binding (P-loop) motifs with the sequence A/GXXXXGKT/S (Walker et al. 1982). Four of the P-loop motifs (P1–P4) are located within the central region of the Dhc, and both spacing and sequences of these P-loops are similar among all Dhc sequences reported thus far (reviewed in Gibbons 1995; Porter 1996). Two additional P-loop motifs were identified in the NH₂-terminal (Pn) and COOH-terminal (Pc) regions of the 1α Dhc respectively; these appear to be unique to the 1α Dhc (Fig. 2).

The predicted amino acid sequence of the 1α Dhc was compared with the three Dhc sequences (α, β, and γ) that form the outer dynein arm in Chlamydomonas (Mitchell and Brown 1994, Mitchell and Brown 1997; Wilkerson et al. 1994) and the cytoplasmic Dhc from Dictyostelium (Koonce et al. 1992). In each case, a high degree of sequence similarity was apparent over long stretches of the polypeptide, especially in the central and COOH-terminal thirds of the Dhc (28–38% identity, 58–67% similarity). However, the more variable NH₂-terminal third of the 1α Dhc also shares significant homology with the β and γ Dhcs of the outer arm (Fig. 3, ∼24% identity, ∼56% similarity). Alignment of the Dhc sequences using the GCG program PILEUP confirmed that the presence of conserved domains within the NH₂-terminal region, but also revealed several short stretches of unique peptide sequence in the 1α Dhc, including the region previously used to generate a monospecific 1α Dhc antibody (Fig. 2; Myster et al. 1997).

The 1α Dhc sequence was also analyzed using programs that predict secondary structure to identify regions with the potential to form α-helical coiled-coil domains (Lupus et al. 1991; Lupus 1996) (Fig. 4). One region located before P-loop 1 (residues 1,227–1,409) and a second after the P-loop 4 (residues 3,192–3,297, 3,400–3,494, and 3,701–3,789) show the highest probability of forming α-helical coiled coils. The presence of limited coiled-coil domains separating the central portion of the Dhc from the NH₂-terminal and COOH-terminal regions has been observed in other Dhc sequences (Mitchell and Brown 1994, Mitchell and Brown 1997; Porter 1996). These conserved structural domains are thought to play an important role in protein interactions within the dynein arms.

To better understand how the specific domains of the 1α Dhc polypeptide might be involved in the assembly and activity of the I1 inner arm dynein, we decided to analyze constructs of the Dhc1 gene in vivo in a pf9 mutant background. Because of the large size of the Dhc1 gene (∼21 kb), two cosmid libraries and one BAC library were screened with probes representing the 5′ and 3′ ends of the Dhc1 gene to improve the chances of recovering clones that contain the full-length gene (Fig. 5 A). The first cosmid library yielded a single clone, cA1, which was positive with both probes, but upon further analysis proved to be lacking a small portion at the 3′ end of the gene (Fig. 5 B). Screening the second cosmid library resulted in the recovery of a single clone, cW1, which contained the complete Dhc1 transcription unit as well as additional genomic sequences both 5′ and 3′ that might be required for proper expression in vivo (Fig. 5 C). Four larger clones (100–135 kb) containing the Dhc1 transcription unit were recovered from the BAC library; two of these clones were used in subsequent cotransformation experiments (Fig. 5 E). We also constructed a truncated version of the Dhc1 transgene known as pD1SA by fusing an 11-kb region encoding the NH₂-terminal 1,956 amino acids to a 1-kb region containing the 3′ end of the gene (Fig. 5 D). All of the Dhc1 transgenes were tested for their ability to rescue the pf9 mutant defects in vivo.

Mutations at the PF9/IDA1 locus typically result in strains that have a slow, smooth swimming behavior (Kamiya et al. 1991; Porter et al. 1992). To increase the sensitivity of the screen for rescue of the pf9 mutant phenotype, we introduced a second motility mutation into the pf9-2 background. pf28 is a mutation in the γ Dhc gene that results in the failure to assemble the outer dynein arms (Mitchell and Rosenbaum 1985; Wilkerson et al. 1994). Cells carrying the pf28 mutation swim with a jerky phenotype that can be easily distinguished from the slow, smooth swimming behavior of the pf9 mutant cells. In addition, pf9-2 pf28 double mutants assemble short, paralyzed flagella and sink in liquid medium (Porter et al. 1992). This short, paralyzed flagellar phenotype makes it very straightforward to identify transformants that have rescued the pf9 mutant defects by screening for cells that assemble full-length, motile flagella and swim with a pf28-like motility phenotype.

The pf9-2 pf28 arg2 strain was first cotransformed with the selectable marker pARG7.8 and the cW1 cosmid containing the complete Dhc1 transcription unit. Positive transformants were selected by growth on solid medium lacking arginine, and then single colonies were picked into liquid media and screened for motility. Fig. 5 C illustrates the combined results of several independent cotransformation experiments with the cW1 cosmid. Although arg+ transformants were recovered at expected frequencies (Kindle 1990), only 11 out of 2,880 transformants screened displayed any motility, for a frequency of rescue of <0.4%. Moreover, the motility of the cW1 rescued strains was not the same as pf28 (see below and Table).

Given these preliminary findings, we decided to test the other Dhc1 transgenes for their ability to rescue the mutant phenotypes. The cosmid cA1 contains an incomplete copy of the Dhc1 transcription unit but also contains the ARG7 gene cloned within the bacterial vector sequences, thereby physically linking the selectable marker and the Dhc1 gene. Therefore, all arg+ transformants might be expected to contain the Dhc1 sequence integrated along with the ARG7 gene. However, the frequency of rescue with the cA1 cosmid (0.61%; Fig. 5 B) was only slightly better than the cW1 cosmid. These observations indicated the Dhc1 transgenes were probably being fragmented during the transformation protocol, but the recovery of motile strains also suggested that a truncated version of the Dhc1 sequence was capable of restoring some function.

Previous study of an outer arm mutation, oda4-s7, had indicated that the NH₂-terminal third of the β Dhc polypeptide is sufficient for assembly of a dynein complex (Sakakibara et al. 1993). To test if this is also true for the 1α Dhc, we cotransformed the pf9 pf28 mutant with the smaller pD1SA construct, which encodes ∼40% of the 1α Dhc sequence. These experiments yielded seven motile strains (Fig. 5 D), which represented only a modest (0.87%) increase in the frequency of rescue, but these rescues confirmed that truncated Dhc1 transgenes could restore partial motility. To see if it was possible to completely rescue the motility defects, we also transformed the pf9 pf28 mutant with two BAC clones that contained the full-length Dhc1 gene located in the middle of ∼100–135-kb genomic inserts (Fig. 5 E). The frequency of rescue (∼0.1%) was still quite low, but the motility phenotypes of the rescued strains were very similar to pf28 (see below).

The recovery of motile isolates after cotransformation could also be due to an intragenic reversion event at the PF9 locus during the course of transformation and/or selection. To confirm that the motility of the transformants was due to the successful expression of the Dhc1 transgene and not a reversion event, two of the cW1 transformants (E2 and G4) were crossed with oda2, another mutant allele at the PF28/ODA2 locus, and the progeny from 11 complete tetrads were analyzed for each cross. If the rescued motility was due to reversion of the pf9-2 mutation, all the resulting tetrad progeny would be motile and swim with a pf28/oda2–like motility phenotype. However, if the rescued motility was due to the presence of the Dhc1 transgene, then the restored motility phenotype would be expected to segregate independently of the pf9-2 mutation, and a class of immotile progeny with the original pf9-2 pf28 genotype should be recovered. Surprisingly, we observed three different motility phenotypes in the tetrad progeny. The first class swam with a motility phenotype that was indistinguishable from either pf28 or oda2. The second class swam with a jerky motion like the pf28/oda2 strains but appeared slower. The third class was immotile with short, stumpy flagella. The recovery of aflagellate strains demonstrated that the original pf9-2 mutation was still present in the genetic background of the two transformants and that the rescued motility was due to the presence of the Dhc1 transgene.

Although the frequency of rescue was low, it was clear that the rescued motility was due to the presence of the different Dhc1 transgenes, and so the motility phenotypes of the Dhc1 transformants were analyzed in greater detail. More specifically, we measured the flagellar beat frequency, the forward swimming velocity, and the ability to phototax (Table). Transformants with complete rescue of the pf9 mutation would be expected to have a swimming phenotype nearly identical to that of pf28. The flagellar beat frequencies of the Dhc1 transformants were almost identical to the beat frequency of pf28, but measurements of forward swimming velocities clearly indicated that most of the transformants swam more slowly than pf28 (Table). In particular, the swimming velocities of the rescued strains obtained by transformation with the cosmid clones and pD1SA were slower than those obtained by transformation with the BAC clones. These results suggested that there were still some inner arm defects in most of the Dhc1 transformants.

We next tested if the Dhc1 transformants had recovered the ability to phototax. King and Dutcher 1997 have previously used a photoaccumulation assay to demonstrate that pf9 mutant cells do not phototax effectively. Using similar conditions, we have found that pf28 cells, which lack outer arms but have the full complement of inner dynein arms, are able to phototax, as assayed by their tendency to become concentrated in the illuminated portion of a tube within 40 min of exposure to a directional light source (Table). However, all of the Dhc1 transformants obtained with the cosmid clones remained equally distributed between the illuminated and darkened regions of the tube. To confirm these findings by direct observation of individual cells, the ability to phototax was also monitored in 96-well plates over a 60-min time course (see Materials and Methods). In the absence of outer arms, the Dhc1 transformants obtained with the cosmid clones remained equally distributed in both the illuminated and darkened portions of the microtiter well. Conversely, the majority of pf28 cells became concentrated on the illuminated side of the well within 15 min. These results indicated that this group of Dhc1 transformants does not phototax as effectively as pf28 control cells.

To examine the motility of the transformants in the presence of outer arms, two strains, G4 and E2, were crossed to pf9-3 and tetrad products containing the Dhc1 transgene in a wild-type outer arm background (G4+OA and E2+OA) were recovered. The two strains have beat frequencies almost identical to wild type, but their swimming velocities are intermediate in speed between pf9 and wild type (Table). Moreover, in the presence of the outer arms, the two strains could photoaccumulate as effectively as wild type. These results suggest that the outer arms can compensate in some way for the phototaxis defects in the Dhc1 transformants.

The swimming behavior of the Dhc1 transformants obtained with the cosmid clones demonstrated that the introduction of these Dhc1 clones resulted in only a partial rescue of the pf9 motility defects. Given the large size of the Dhc1 transcription unit (>22 kb), we were initially concerned that these transgenes might not be expressing wild-type levels of the Dhc1 gene product. To address this question, we isolated axonemes from the Dhc1 transformants and analyzed the components of the I1 complex. Previous work has shown that the I1 complex is composed of eight polypeptides, two Dhcs (1α and 1β), three ICs (IC140, IC138, and IC110) (Smith and Sale 1991; Porter et al. 1992; Myster et al. 1997), and three LCs (LC8, LC12, and LC14) (Harrison et al. 1998). The Dhc1 gene encodes the 1α Dhc, which can be identified on Western blots using antibody directed against a peptide epitope in the NH₂-terminal region (Myster et al. 1997). Fig. 6 A shows three Coomassie blue–stained polyacrylamide gels containing whole axonemes isolated from several mutant strains: the 11 motile Dhc1 transformants obtained with the cW1 cosmid, the 4 motile Dhc1 transformants recovered with the cA1 construct, and the 5 transformants recovered with the Dhc1 BAC clones. Fig. 6 B shows the corresponding Western blots probed with the 1α Dhc antibody. The 1α Dhc antibody identified the ∼520 kD 1α Dhc in the pf28 control sample, but polypeptides significantly smaller than the 1α Dhc were identified in all of the motile strains obtained by transformation with the Dhc1 cosmids. In contrast, all of the axoneme samples prepared from rescued strains obtained by transformation with the Dhc1 BAC clones contained full-length 1α Dhc polypeptides. These results indicated that the partial rescue phenotype seen with the Dhc1 cosmid clones was not due to low levels of expression of a full-length 1α Dhc, but instead due to the expression of truncated 1α Dhcs, ranging in size from ∼165 to ∼300 kD.

To determine if other I1 subunits were associated with the truncated 1α Dhcs, Western blots of isolated axonemes were probed with an antiserum raised against the 140-kD intermediate chain (Yang and Sale, 1998). This antibody detects the IC140 in wild-type axonemes, but not in I1 mutant axonemes. As shown in Fig. 6 C, the IC140 antibody recognized a single polypeptide of ∼140 kD in pf28 and each rescued transformant, but did not detect the IC140 in any of the pf9 mutant strains. Similar results were seen using the antibody directed against the 14-kD Tctex1 light chain (data not shown).

To confirm that the other polypeptide subunits were assembled into an I1 complex, we isolated whole axonemes from large-scale cultures of two Dhc1 transformants, E2 and G4, as well as from control pf28 cells. Partially purified I1 complexes were obtained by high salt extraction of the isolated axonemes followed by sucrose density gradient centrifugation. The resulting fractions were analyzed by both SDS-PAGE and Western blotting. Fig. 7 A shows the 19S region of a sucrose gradient that was loaded with the pf28 dynein extract. The two Dhcs and three ICs of the I1 isoform cosediment as a complex that peaks in fraction number 4. Duplicate samples tested on Western blots probed with the 1α Dhc antibody confirmed the presence of the 1α Dhc in the 19S region (right). Fig. 7 B shows four fractions from the sucrose gradient that was loaded with the G4 dynein extract. The gel on the left reveals that the 1β Dhc and the three intermediate chains of the I1 complex have shifted and now cosediment at ∼16S, peaking in fraction 6. A novel polypeptide of ∼183 kD cosediments in the same region (see asterisks). Western blot analysis with the 1α Dhc antibody identified this novel band as the truncated 1α Dhc (Fig. 7 B, blot). Identical results were observed with dynein extracts isolated from the E2 strain (data not shown). The truncated 1α Dhcs in the Dhc1 transformants therefore form stable complexes with the other polypeptides of the I1 complex, but the resulting mutant complexes sediment more slowly than wild-type complexes.

To analyze the structure of the I1 complex in the Dhc1 transformants, we prepared purified axonemes from wild-type and mutant strains for thin section EM. To facilitate the analysis of the images, the transformants G4 and E2 were crossed to a pf9-3 strain to recover the Dhc1 transgene in a wild-type outer arm background (G4+OA and E2+OA, see Materials and Methods). Fig. 8 a shows the grand average of the 96-nm repeat from wild-type axonemes, and Fig. 8 b indicates the corresponding densities. Previous work has shown that the inner dynein arms repeat as a complex group of structures every 96 nm in register with the radial spokes (Muto et al. 1991; Mastronarde et al. 1992). The I1 complex is a trilobed structure located proximal to the first radial spoke (S1) in each 96-nm repeat (Fig. 8 b, lobes 1–3) (Piperno et al. 1990; Mastronarde et al. 1992; Myster et al. 1997). These three lobes are missing in pf9-3 axonemes (Fig. 8c and Fig. d), which lack the I1 complex (Myster et al. 1997). Fig. 8e and Fig. g, show the grand averages of the axonemes from the G4+OA and E2+OA strains, which contain I1 complexes with truncated 1α Dhcs. Lobes 1 and 3 of the I1 complex are present in the axonemes from these samples, but lobe 2 is still missing. Difference plots between these images and wild type confirms that the loss of lobe 2 is the only significant defect in the two Dhc1 transformants (Fig. 8f and Fig. h). These images demonstrate that the I1 complex is assembled and targeted to the appropriate axoneme location. In addition, these images suggest that the region of the 1α Dhc that is missing in the Dhc1 transformants corresponds to lobe 2 of the I1 structure.

Although the initial cotransformation experiments involved the use of a full-length or near full-length Dhc1 cosmid clones, all of the motile transformants recovered with these clones assemble partially functional I1 complexes with truncated 1α Dhcs (Fig. 6 and Table). To understand how the 1α Dhc fragments were related to the Dhc1 sequence, we analyzed the Dhc1 transcripts from several of the rescued transformants on Northern blots. Total RNA was first isolated from the G4+OA and E2+OA strains, which contain the Dhc1 transgene in the pf9-3 null mutant background. This background facilitated our analysis because the pf9-3 mutation is a large deletion (∼13 kb) in the Dhc1 gene and does not generate an endogenous Dhc1 transcript (Myster et al. 1997).

Fig. 9 A shows a partial restriction enzyme map of the Dhc1 gene and the subclones that were used as probes to analyze the Dhc1 transcripts. As shown in Fig. 9 B, probe A3′, which spans the Dhc1 transcription start site, identified a single, large (>13 kb) transcript in wild-type RNA. However, in G4+OA and E2+OA, the transcripts recognized by the A3′ probe were significantly smaller than the wild-type Dhc1 transcript, but these smaller transcripts were still upregulated in response to deflagellation (compare lanes 0 and 45). Identical results were observed with the next two subclones, probes B and C. Probe D, which includes the conserved region encoding the primary ATP hydrolytic site, hybridized to the truncated transcripts in E2+OA, but it did not recognize the truncated transcripts in G4+OA. Probes E–G did not hybridize with any transcripts in the transformants (Fig. 9 B, right panel, and data not shown). The Dhc1 transcripts in G4 and E2 are therefore truncated from the 3′ end of the Dhc1 gene, and G4 is truncated before E2.

To characterize the Dhc1 transcripts present in other transformants, RNA was isolated from three additional strains: G3, G9, and A2. G3 and G9 are the cW1 transformants that assemble the smallest and largest 1α Dhc fragments respectively, whereas A2 is a cA1 transformant that assembles the largest (>300 kD) 1α Dhc fragment obtained thus far (Fig. 6). As shown in Fig. 9 C, in each strain, probe C hybridized to a truncated transcript that is significantly smaller than the endogenous Dhc1 transcript derived from the pf9-2 mutant background. However, probe D, which corresponds to the region encoding the ATP hydrolytic site, failed to hybridize with the truncated transcripts present in the G3, A2, and G9 samples. Similar results were seen with probes E and F (data not shown). Therefore, all three strains encode 1α Dhc fragments that are truncated before the proposed motor domain.

To identify the sites where the Dhc1 cosmid clones were being modified during transformation, we isolated genomic DNA from the transformants and analyzed the structure of the integrated Dhc1 transgenes on Southern blots. Fig. 10 A shows a blot of SacI digested genomic DNA that was hybridized with a probe for subclone C. As expected, this probe hybridized to the endogenous Dhc1 gene present in the pf9-2 mutant background of the transformants. However, for each transformant, a second polymorphic band could be detected using either probe C (Fig. 10 A) or probe D (data not shown). From these and other blots, we concluded that the Dhc1 cosmids were being rearranged during integration into genomic DNA. In addition, the region of the Dhc1 gene encoding the conserved motor domain appeared to be the most common target for disruption during these integration events.

Because the G3 transformant assembles the smallest 1α Dhc fragment identified thus far (Fig. 6), we recovered the modified Dhc1 transgene using probe C to screen a mini-library made from genomic DNA of the G3 transformant (see Materials and Methods). The Dhc1 transgene was then sequenced with Dhc1 specific primers to identify the junction between the Dhc1 sequence and the site of integration in G3 genomic DNA. The Dhc1 sequence in G3 is fused to an unidentified DNA sequence, and the resulting hybrid gene is predicted to encode up to amino acid residue 1,249 of the 1α Dhc, followed by the addition of 17 novel amino acids before encountering a stop codon (Fig. 10 B). Sequence analysis of an RT-PCR product derived from G3 RNA has confirmed the presence of this hybrid transcript. The polypeptide encoded by the modified transgene would, therefore, correspond to a 1α Dhc fragment of ∼143 kD that is truncated just COOH-terminal to the epitope recognized by the 1α Dhc antibody (Fig. 10 C). The recovery of this fragment in G3 axonemes (Fig. 6) reveals that the NH₂-terminal coiled-coil domains of the 1α Dhc are not required for I1 complex assembly.

16 different Dhc genes have been identified in Chlamydomonas: 2 cytoplasmic Dhc sequences (Pazour et al. 1999; Porter et al. 1999), 3 genes that encode the Dhcs of the outer dynein arm (Mitchell and Brown 1994, Mitchell and Brown 1997; Wilkerson et al. 1994), and 11 other Dhc genes whose expression patterns are consistent with gene products that are involved in flagellar function (Porter et al. 1996, Porter et al. 1999; Knott, J., and M. Porter, unpublished results). The analysis of specific Dhc mutations has revealed the role of each outer arm Dhc in flagellar motility (Sakakibara et al. 1991, Sakakibara et al. 1993; Porter et al. 1994; Rupp et al. 1996), but the specific functions of the multiple inner arm Dhcs are largely unknown. We have previously demonstrated that the Dhc1 gene product plays an essential role in the assembly and function of the I1 inner arm complex (Porter et al. 1996; Myster et al. 1997). To better understand how this inner arm Dhc contributes to flagellar motility, we have now analyzed the structure and function of the Dhc1 gene and the encoded 1α Dhc in detail.

As described in Fig. 1, we have determined the nucleotide sequence for the complete Dhc1 transcription unit and used this information to obtain the predicted amino acid sequence of the 1α Dhc (Fig. 2). To our knowledge, this is the first full-length, inner arm Dhc sequence to be reported in any organism. Comparisons to other full-length Dhc sequences indicate that the 1α Dhc is most similar to the β and γ Dhcs of the outer arm, and that these sequence similarities extend into the NH₂-terminal region (Fig. 3). As the NH₂-terminal region of the Dhc is thought to be involved in the association with isoform specific IC and LC subunits, these observations suggest that the 1α Dhc and the β and γ Dhc contain conserved sites for the binding of accessory subunits. Indeed, recent studies have revealed that the I1 complex and the outer arm do contain similar IC and LC components, such as the 8-kD LC, the Tctex1 and Tctex2 LCs, and a family of WD-repeat containing ICs (Patel-King et al. 1997; Harrison et al. 1998; Yang and Sale, 1998). However, a novel feature of the 1α Dhc sequence is the presence of a P-loop motif (Pn) at amino acid residues 960–967 (Fig. 2). A weakly conserved P-loop motif has also been identified within the first 200 residues of the β Dhc (Mitchell and Brown 1994; Kandl et al. 1995). Whether the Pn sequence in the 1α Dhc is a bona fide nucleotide binding site is unknown, but the future sequence analyses of 1α Dhc homologues in other organisms should indicate whether the Pn motif is a conserved feature of this class of Dhc.

The central and COOH-terminal thirds are the most highly conserved regions of the 1α Dhc, and our transformation experiments are consistent with previous proposals that this region corresponds to the dynein motor domain (Sakakibara et al. 1993; Koonce and Samso 1996, Gee et al. 1997). The amino acid residues around P1 (the primary ATP hydrolytic site) and P4 are highly conserved with other axonemal Dhcs, whereas the sequences around P2 and P3 are less well conserved (Gibbons et al. 1994; Gibbons 1995). For example, P3 does not strictly conform to the P-loop consensus sequence GXXXXGKS/T (Walker et al. 1982), as the glycine in position 6 is substituted by an alanine in the 1α Dhc sequence, but the same amino acid substitution was found in the γ Dhc sequence (Wilkerson et al. 1994).

The COOH-terminal third of the 1α Dhc also contains a small region ∼340 amino acids downstream from P4 that is predicted to form a limited coiled-coil domain (Fig. 4). A similar region in cytoplasmic Dhc sequences has been identified recently as the stalk structure that extends from the globular head domain and forms the microtubule binding site (Koonce 1997; Gee et al. 1997; Gee and Vallee 1998). Mutations in this region in the outer arm β Dhc can have dramatic effects on flagellar motility (Porter et al. 1994). Interestingly, a sixth P-loop motif (Pc) has been identified in the 1α Dhc sequence, ∼280 amino acids downstream from the proposed microtubule binding site. The function of this sixth P-loop in the 1α Dhc is unknown, but its position downstream from the microtubule binding site is intriguing. Recent sequence analysis has suggested that all dyneins may contain six ATPase-like repeat regions: the four central P-loops previously identified and two additional, less well conserved repeats after the COOH-terminal coiled-coil domains (Neuwald et al. 1999).

Using the Dhc1 sequence information, we recovered two BAC clones and two cosmid clones containing full-length or near full-length Dhc1 genes, and then used these constructs to rescue the pf9 motility defects (Fig. 5). 20 independent transformants with rescued motility were recovered. Backcrossing the transformants confirmed that the rescued motility was due to the presence of the Dhc1 transgene and not to a reversion event at the PF9 locus. However, analysis of the Dhc1 transformants produced two unexpected results. First, the frequency of rescue (<1%) was much lower than previously observed with other flagellar genes (5–10%) (Diener et al. 1990; Kindle 1990; Tam and Lefebvre 1993), and second, the 1α Dhcs were truncated in most of the motile transformants recovered thus far.

One reason that the cotransformation frequencies were so low might be due to the large size of the Dhc1 transcription unit, which could make the Dhc1 transgenes more susceptible to damage during the transformation protocol. Exogenous DNA sequences often undergo deletions as they integrate into the Chlamydomonas genome (Tam and Lefebvre 1993; Smith and Lefebvre 1996, Smith and Lefebvre 1997; Koutoulis et al. 1997). The recovery of truncated 1α Dhc fragments (Fig. 6) indicated that the Dhc1 cosmids were being disrupted during the transformation protocol, and both Southern and Northern blot analyses of the rescued transformants (Fig. 9 and Fig. 10) have confirmed that deletions from the 3′ end of the Dhc1 cosmids did occur. The presence of additional genomic DNA flanking the Dhc1 transcription unit in the BAC clones may have served to protect the Dhc1 transgenes and thereby permitted the full-length rescues observed with these clones (Fig. 5 and Fig. 6).

Another reason for the low frequency of rescue might be the relatively small amount of genomic DNA present on the 5′ end of the Dhc1 gene in certain constructs (Fig. 5). If this region was randomly deleted, the resulting cotransformants would not retain the sequences necessary for expression of the Dhc1 transcript and rescue of the mutant phenotype. Recent experiments with constructs encoding the IC140 subunit have indicated that sufficient DNA upstream from the 5′ end of the IC140 gene is essential for efficient rescue of the ida7 mutation (Perrone et al. 1998).

The second unexpected result was the frequency with which we recovered motile transformants expressing only NH₂-terminal fragments of the 1α Dhc (Fig. 6 B). Northern and Southern blot analyses have shown that the rescued strains retained the 5′ sequence elements required for regulated expression of the Dhc1 transcripts, but several of these strains lacked the 3′ end of the transgene (Fig. 9 and Fig. 10). Therefore, the truncated 1α Dhcs represent those NH₂-terminal fragments that were competent to assemble with other subunits into the I1 complex (Fig. 6B and Fig. C, and Fig. 7).

The observation that none of the 1α Dhc fragments is smaller than ∼143 kD may indicate that this is the shortest NH₂-terminal fragment capable of complex assembly. Studies of outer arm mutants have identified a novel β Dhc mutation with similar properties. The oda4-s7 mutant expresses a 160-kD fragment of the β Dhc that is capable of coassembly with other outer arm subunits at the correct axoneme location (Sakakibara et al. 1993). Likewise, low level expression of cytoplasmic Dhc constructs in Dictyostelium indicates that a 158-kD NH₂-terminal fragment is also capable of complex assembly (Koonce and Knecht 1998).

Although the NH₂-terminal third of the 1α Dhc is the most variable region, secondary structure programs have identified a region just before P1 that is predicted to form a limited coiled-coil domain (Fig. 4). This domain, which has been identified in nearly all Dhcs sequenced to date (Mitchell and Brown 1994, Mitchell and Brown 1997), has been proposed as a potential region that might mediate interactions between the Dhcs and their associated ICs and LCs. To determine if this coiled-coil domain is required for assembly of the I1 complex, we recovered the Dhc1 transgene from the G3 transformant, which expresses the shortest 1α Dhc fragment (Fig. 6 B). Sequence analysis of the truncated Dhc1 transgene demonstrated that the 1α Dhc sequence terminates before the region predicted to form the NH₂-terminal coiled-coil domain (Fig. 10). Given that this 1α Dhc fragment still assembles with other I1 components into the flagellar axoneme (Fig. 6), other sites within the NH₂-terminal region must be required for complex formation. We plan to analyze additional Dhc1 constructs to further delineate the domains required for specific subunit interactions and complex assembly.

If the Dhc1 transgenes were deleted randomly from the 3′ end, we would expect to recover a broad distribution of Dhc fragments ranging in size from the minimum required to assemble the I1 complex to nearly full-length. Therefore, why are almost all of the 1α Dhc fragments smaller than ∼217 kD (Fig. 6 B)? One possibility may be that larger 1α Dhc fragments are unstable and prevent assembly of the I1 complex into the axoneme. Studies in Dictyostelium have shown that constructs of cytoplasmic Dhc lacking significant portions of the COOH terminus are expressed poorly as compared with other constructs that contain the entire motor domain (Koonce 1997). Alternatively, the presence of larger 1α Dhc fragments with partial motor domains may inhibit flagellar motility. If so, we would not recover such transformants in our screen, which was based on the rescue of a motility defect. Indeed, Northern blot analysis of the two transformants (A2 and G9) that assemble the largest 1α Dhc fragments demonstrates that the sequences encoding the dynein motor domain are not present in the associated Dhc1 transcripts (Fig. 9). The NH₂-terminal regions of these larger 1α Dhc fragments must, therefore, be fused to other protein sequences. Interestingly, the absence of motile transformants with partial motor domains is consistent with previous reports that nucleotide binding by the cytoplasmic Dhc is inhibited by deletion of the COOH terminus, leading to the formation of rigor complexes (Gee et al. 1997).

Structural studies of the isolated I1 complex by negative staining or rotary shadowing have shown that it is a two-headed isoform (Goodenough et al. 1987; Smith and Sale 1991) but the physical relationship between the globular heads seen in vitro and the structural domains identified in situ has been unknown. Because some of the Dhc1 transformants assemble I1 complexes that lack central and COOH-terminal regions of the 1α Dhc (Fig. 7), we can now identify the position of the 1α motor domain within the I1 structure. EM analysis of axonemes isolated from the E2 and G4 transformants has revealed that lobe 2 of the I1 structure corresponds to the missing 1α motor domain (Fig. 8). Lobe 2 is close to the first radial spoke, in a position that may permit direct signaling between the radial spoke and the 1α Dhc motor domain. We predict that the remaining two lobes of the I1 structure represent the positions of the 1β Dhc motor domain and stem region containing the I1 ICs and LCs respectively. We are currently transforming other I1 mutants with the genes for other I1 subunits to identify the polypeptide components that are located within these structural domains (Perrone et al. 1998; Perrone, C.A., E. O'Toole, and M.E. Porter, work in progress).

The recovery of transformants missing only the 1α Dhc motor domain but containing the other components of the I1 complex has allowed us to analyze the specific role of the 1α Dhc motor domain in flagellar motility. I1 mutants lacking both the 1α and 1β Dhcs have a slow, smooth swimming phenotype with an altered flagellar waveform (Kamiya et al. 1991; Porter et al. 1992). Measurements of swimming velocities reveal that the Dhc1 transformants with truncated 1α Dhcs swim faster than I1 mutants but slower than control strains containing both Dhcs (Table). The 1α Dhc motor domain, therefore, contributes directly to force production during motility.

The I1 complex is also an essential component of the phototaxis response in Chlamydomonas. Strains that have defects in outer dynein arms, the dynein regulatory complex, or other inner arm isoforms can phototax, but mutants lacking the I1 complex cannot (King and Dutcher 1997). Analysis of other phototaxis mutants that retain the I1 complex reveals that the phosphorylation state of IC138 is altered (King and Dutcher 1997). In vitro sliding assays have shown that the phosphorylation state of the IC138 affects microtubule sliding velocities (Habermacher and Sale 1997). These results suggest a model in which the phosphorylation state of the IC138 modulates the activity of the I1 Dhc motor domains.

To assess the specific role of the 1α Dhc motor domain in phototaxis, we compared the swimming behavior of the Dhc1 transformants to that of control cells in response to a directional light source. pf28 cells were clearly phototactic, but the Dhc1 transformants with truncated 1α Dhcs remained uniformly dispersed during the time course of our assays. These observations indicate that the motor activity of the 1α Dhc contributes to phototaxis, at least in the absence of the outer arms. However, if the outer arms were present, the Dhc1 transformant strains could undergo phototaxis, whereas I1 mutant strains could not (Table). This difference in behavior in the presence or absence of outer arms suggests that there are cooperative interactions between the I1 complex and the outer arms during the phototaxis response.

Previous studies have demonstrated that differences in the activity of the cis and trans flagellum are the basis of the phototaxis response (Kamiya and Witman 1984; Rüffer and Nultsch 1991, Rüffer and Nultsch 1997; Horst and Witman 1993; Witman 1993). This differential activity includes both differences in beat frequency between the two flagella as well as differences in flagellar waveforms (Sakakibara and Kamiya 1989; Rüffer and Nultsch 1991). The outer dynein arms are responsible for generating the differences in beat frequency observed between cis and trans flagella, as the cis-trans frequency differential is lost in mutants that lack the outer arm or the outer arm α Dhc (Sakakibara and Kamiya 1989; Sakakibara et al. 1991). More recent work has demonstrated that this differential beat frequency depends on the presence of the docking structure that facilitates the attachment of the outer arm to its specific binding site on the A-tubule (Takada and Kamiya 1997). The differential in beat frequency is not essential for the phototaxis response (Rüffer and Nultsch 1991), as outer arm mutant cells such as pf28 are capable of phototaxis (Table). However, it is clear that the cis-trans differences in beat frequency and flagellar waveform must be coupled in some way, because phototaxis mutants such as ptx1 are defective in both (Rüffer and Nultsch 1997; Takada and Kamiya 1997). In this context, it appears that regulation of the I1 complex is important for generating the asymmetries in flagellar waveform between the cis and trans flagella that contribute to phototaxis (King and Dutcher 1997), whereas regulation of the outer arm contributes to the differential in flagellar beat frequency (Takada and Kamiya 1997). Phototaxis can occur in the absence of the outer arms, but not the I1 complex (King and Dutcher 1997; Table). However, if both the outer arms and part of the I1 complex are present (such as in G4+OA or E2+OA), then their combined activity can apparently compensate for the absence of the 1α Dhc motor domain.

These observations raise several interesting questions about the mechanism by which changes in the phosphorylation state of IC138 might contribute to phototaxis. For example, where is IC138 located in the axoneme relative to the motor domains of the two I1 Dhcs and the three outer arm Dhcs? Is it in lobe 3 of the I1 structure, which is also in close proximity to at least one outer arm per axoneme repeat? Does IC138 interact directly with either the 1α or 1β Dhc? We are planning to address these questions by analyzing subunit interactions within the I1 complex. Other important questions concern the identity and location of the axonemal kinases and phosphatases that modulate the phosphorylation state of IC138. Work from other laboratories has indicated that many of these regulatory components are tightly bound to the flagellar axoneme (Habermacher and Sale 1997; King and Dutcher 1997; Yang, P., and W.S. Sale. 1998. The M_r 140,000 intermediate chain of Chlamydomonas flagellar inner arm dynein is a WD-repeat containing protein implicated in dynein arm anchoring. Mol. Biol. Cell. 9:3335–3349), but the specific enzymes that act on either the outer arm or the I1 IC138 in situ have not yet been identified. The future identification and localization of these regulatory components will provide new insights into the pathway that governs the activity of the multiple dynein motors during flagellar motility and phototaxis.

We wish to thank other members of the Porter laboratory for their advice and support during the course of this project. We are also grateful to the members of the laboratories of C. Silflow, P. Lefebvre, and R. Linck (University of Minnesota) for their helpful suggestions at our weekly group meetings. Parts of this work were completed by S.H. Myster in partial fulfillment of the requirements for a Ph.D. degree at the University of Minnesota.

This work was supported by a grant from the National Institutes of General Medical Sciences (GM 55667) to M.E. Porter and S.H. Myster was supported in part by a research training grant from the National Science Foundation for Interdisciplinary Studies on the Cytoskeleton (DIR91134444). E. O'Toole was supported by a National Institutes of Health Biotechnology Resource (RR00592) grant to J.R. McIntosh.