Novel LC8 Mutations Have Disparate Effects on the Assembly and Stability of Flagellar Complexes*

LC8 functions as a dimer crucial for a variety of molecular motors and non-motor complexes. Emerging models, founded on structural studies, suggest that the LC8 dimer promotes the stability and refolding of dimeric target proteins in molecular complexes, and its interactions with selective target proteins, including dynein subunits, is regulated by LC8 phosphorylation, which is proposed to prevent LC8 dimerization. To test these hypotheses in vivo, we determine the impacts of two new LC8 mutations on the assembly and stability of defined LC8-containing complexes in Chlamydomonas flagella. The three types of dyneins and the radial spoke are disparately affected by dimeric LC8 with a C-terminal extension. The defects include the absence of specific subunits, complex instability, and reduced incorporation into the axonemal super complex. Surprisingly, a phosphomimetic LC8 mutation, which is largely monomeric in vitro, is still dimeric in vivo and does not significantly change flagellar generation and motility. The differential defects in these flagellar complexes support the structural model and indicate that modulation of target proteins by LC8 leads to the proper assembly of complexes and ultimately higher level complexes. Furthermore, the ability of flagellar complexes to incorporate the phosphomimetic LC8 protein and the modest defects observed in the phosphomimetic LC8 mutant suggest that LC8 phosphorylation is not an effective mechanism for regulating molecular complexes.

LC8 is an important component of vital complexes such as cytoplasmic dynein, myosin V, neuronal nitric-oxide synthase (1), and the nuclear pore complex (2). In Chlamydomonas flagella, LC8 is found in three types of dynein motors (the outer dynein arm (ODA), 3 the inner dynein arm (IDA) called the I1 dynein, and the cytoplasmic dynein motor CD1b for intraflagellar transport (IFT)) and in the radial spoke (RS) (3)(4)(5)(6)(7). The LC8-null mutant, fla14, has short, paralyzed flagella in which these complexes are absent or reduced (6) ( Table 1). Drastic pleiotropic phenotypes also are noted in Drosophila LC8 mutants (8). In particular, the amorph mutant is embryonic lethal because of apoptotic death, consistent with the discovery that LC8 helps to sequester the pro-death apoptotic factor Bim (9).
Despite its significance, the function and regulation of LC8 are only partially understood. LC8 functions as a dimer with two identical binding sites for the target proteins (10,11). One model postulates that the dimer tethers dynein motors to various cargoes (12). However, in some systems, the role of LC8 is unrelated to dynein motors (1,13,14). Another emerging model, founded upon structural analyses, contends that the bivalent structure of the LC8 dimer increases the stability (15) and promotes the refolding of the scaffold proteins in the complex (16).
Additional data indicate that LC8 is regulated by phosphorylation (17)(18)(19). The C-terminal Ser 88 residue is phosphorylated in tumor cells, and the phosphorylation may promote cell survival by release and subsequent degradation of Bim (17). Consistent with this idea, a phosphomimetic LC8S88E mutant protein is predominantly monomeric in vitro and cannot bind to Bim or the target peptide from IC74 of cytoplasmic dynein (18,19). However, LC8S88E does bind to peptides of Swallow, a target protein in the Drosophila oocyte (18). It was postulated that Swallow peptides have a stronger affinity for LC8 and bind to the small dimeric pool of LC8S88E, reversing the monomeric shift, and that LC8 phosphorylation is a selective regulatory mechanism. These observations raise a critical question about LC8 phosphorylation as a regulatory mechanism. How can phosphorylated LC8 promote cell survival while failing to bind vital target proteins, such as intermediate chains in cytoplasmic dynein? Testing these models for functional and regulatory mechanisms ultimately requires in vivo analyses.
In this study, we characterize the Chlamydomonas mutant pf5 (20). The PF5 gene product was implicated in integrating the assembly and function of axonemes because in addition to an RS deficiency and the lack of two minor RS proteins, RSP13 and RSP15, pf5 axonemes also lack a small protein and an ϳ110-kDa polypeptide of the inner dynein arms. The RS is part of a system that controls the activation of axonemal dynein motors. We discovered that the intriguing phenotype of pf5 is actually caused by a mutation in the LC8 gene FLA14. This new mutant, expressing LC8 with a C-terminal extension (LC8CT), and an additional LC8 phosphomimetic mutant (LC8S90E) were used for analyses of the assembly and stability of the defined LC8-containing dyneins and the RS complexes in the Chlamydomonas flagellum (Table 1). Comparison of the phenotypes from these two new mutants and the LC8 null mutant (6) strongly support the predicted structural role of LC8 on target proteins to promote the stability and assembly of indi-vidual complexes at various levels. Furthermore, our results suggests that this function and molecular complexes are not significantly altered by LC8 phosphorylation.

Molecular Biology, LC8 Genomic Constructs
The paromomycin resistance cassette was amplified from the pSI103 vector (21) and blunt-end cloned into the pBluescript vector at the SmaI site to generate the pPMM plasmid. Subsequently, LC8 genomic fragments, containing 0.55 and 1 kb of 5Ј-and 3Ј-flanking sequences (23) and the LC8 coding region, were PCR-amplified from WT and CC1331 genomic DNA. The PCR products were cloned into the XhoI site in pPMM to generate pPMM-LC8 and pPMM-LC8CT plasmids.
For the LC8S90E phosphomimetic construct, PCR mutagenesis was performed to mutate the Ser 90 codon (TCG) to a glutamate codon (GAA) using a primer containing the mutated sequence and a nearby BssHII restriction sequence. The PCR product was used to replace the equivalent fragment in the pPMM-LC8 plasmid to generate the pPMM-LC8S90E plasmid. For the LC8S90A construct, PCR-based site-directed mutagenesis was carried out using the QuickChange mutagenesis kit (Stratagene, La Jolla, CA).

Expression of His-tagged Recombinant Proteins
LC8 coding sequences were PCR-amplified and cloned into the NdeI and EcoRI sites of pET28(a) vector. The constructs were transformed into BL21(DE3) for expression as described (21). All constructs were confirmed by restriction digest and sequencing.

Genetics
Transformation rescue and genetic crosses were performed as described (21,22). The experiments in which fla14-1 was transformed with WT and mutant LC8 constructs allow for a direct comparison of the impacts of the LC8 mutations in an identical genetic background.

Biochemistry
Axoneme Isolation-Axoneme isolation, high salt extractions at 5 mg/ml and velocity sedimentation were carried out as described (21).
Analytical Gel Filtration Chromatography-Purified, recombinant LC8 proteins were dialyzed in 50 mM phosphate/citrate buffer, pH 3 or pH 7, for 24 h. 100-l aliquots of 50 M protein were fractionated on a Superose 12 HR 10/30 column with the dialysis buffer at a flow rate of 0.1 ml/min powered by an AKTA FPLC system.
Chemical Cross-linking-Isolated axonemes were treated with 1,5-difluoro-2,4-dinitrobenzene (Pierce) at a series of concentrations as described (10). After treatment, the axonemes were fixed for SDS-PAGE and analyzed by Western blots.
Antibodies-Antibodies were used in Western blots at 1:5000 dilutions to detect the following axonemal structures: for LC8 and the ODA, LC8 and IC78 (Dr. S. M. King

Light Microscopy
Bright field light microscopy for motility assessment, imaging, and statistics was carried out as described (21). For flagellar length measurement, cells were fixed in 50% Lugol solution (Sigma). Images were magnified ϫ2 optically and further enlarged ϫ6 digitally. Flagella were traced on the monitor with a cotton thread, and the length was measured by a ruler. The measurement was converted to actual length by comparison with an image of a micrometer taken in parallel.

RESULTS
pf5 Mutant Displays Multiple Flagellar Defects Because of a Mutation in the LC8 Gene-The pf5 strain, CC1331, exhibits paralyzed or twitching flagella, typical for RS mutants with the occasional, slow swimming cell. We also found that these cells were defective in flagellar generation as well. Compared with the 10 -12-m WT flagella (Fig. 1A, upper panel), CC1331 flagella were about one-half length (Fig. 1A, lower panel; Table 1). Some cells have unequal length flagella (Fig. 1A, arrowhead) or flagella with swollen tips (Fig. 1A, arrow in insets), a phenotype typical of IFT mutants defective in the subunits of CD1b, the retrograde IFT motor, such as LC8 (6,25).
To characterize the protein deficiencies in pf5, we performed Western blots of axonemes, probing for the subunits of the flagellar dyneins and the RS, structures that are crucial for flagellar motility and generation (Fig. 1B). As reported previously (20), compared with different loads of the WT control, the RS in pf5 axonemes, represented by RSP16 and RSP3, was reduced to ϳ50%, and the lower band for RSP3, indicative of the dephosphorylated form (26), was more prominent. The lack of RSP13 and RSP15 (20) was not further confirmed because antibodies to these RS subunits are not available. In contrast, the light intermediate chain (DLIC) of the retrograde IFT motor CD1b (25,27), IC140 in I1 (28), p28 in single-headed IDAs (29), and IC78, LC6, and LC9 of the ODA (30) were not obviously affected. Interestingly, IC97 (also called IC110) in I1 (31) was not detectable, indicating IC97 is the missing 110-kDa IDA component in pf5 (20) and is not essential for the assembly of I1.
To determine whether the multiple defects observed in CC1331 were caused by a single mutation, the CC1331 strain was backcrossed to WT. The motility and axonemes of tetrads derived from 10 zygotes were analyzed. Segregation of the CC1331 motility phenotype was 2:2 with only parental di-types observed (10:0:0). As shown in the representative Western blots of axonemes ( Fig. 2A), dephosphorylated RSP3 (dot), absence of IC97, and short twitching flagella (T) always co-segregated, whereas phosphorylated RSP3 and IC97 were only present in axonemes from 100% WT swimmers (S). The strict co-segregation of these biochemical markers and the motility anomaly suggested that the phenotypes are caused by a single mutation or tightly linked mutations.
To determine the genetic defect in pf5, transformation rescue experiments were carried out. CC1331 was co-transformed with BAC DNA that mapped near the pf5 locus on linkage group III and a plasmid that confers paromomycin (PMM) resistance for selection. Only BAC DNA containing the LC8encoding FLA14 gene rescued the pf5 mutant phenotype (data not shown). To determine whether pf5 is defective in the LC8 gene, FLA14 genomic DNA from CC1331 was amplified by PCR and sequenced. Two independent experiments showed that the TAA stop codon is replaced with the leucine-encoding TTA codon (asterisk, Fig. 2B) leading to an extended C terminus with an additional 23 amino acids, including 4 charged residues. The mutation is expected to result in a larger LC8 protein with a more basic pI than WT LC8. This result is consistent with the report that a small protein, presumably WT LC8, in the two-dimensional map of pf5 axonemes was absent (20). Secondary structure analysis with the PHD program predicts that the addition also changes the conformation of the last ␤-strand (␤5, Fig. 2B). As predicted, Western analysis showed that LC8 with the extended C terminus (LC8CT) was incorporated into CC1331 axonemes and is slightly larger than LC8 from WT axonemes (Fig. 2C).
To confirm that the CC1331 phenotypes were because of the read-through mutation in the LC8 gene, CC1331 cells were transformed with a single plasmid containing the WT LC8 gene (FLA14) and a paromomycin expression cassette for selection. The prediction is that the WT FLA14 gene will rescue the flagellar length and motility defects in CC1331. Among 200 transformants, 86 clones were randomly selected for further analyses. Based on the motility, they were categorized into three groups: CC1331-like, WT-like, and mixed populations. Axonemes from each group were analyzed by Western blots (Fig.  2D). The CC1331-like clones (H7, H8, and G1) were immotile, had difficulty generating flagella (see low amounts of tubulin), and their axonemes still had high levels of dephosphorylated RSP3 (Fig. 2D, dot). Interestingly, axonemes from the transformants with swimmers contained both forms of LC8 (LC8CT and WT LC8). Those with only ϳ30% swimmers (A1, B12, and C10 strains) had more LC8CT than WT LC8, whereas the strains with 100% swimmers contained largely WT LC8 (B11 strain) or a similar level of WT LC8 and LC8CT (B1 strain). Thus, LC8CT can compete with WT LC8 for assembly into the axoneme, and 50% WT LC8 is sufficient to restore WT-like motility. In the rescued strains, IC97 was restored, and RSP3 was largely phosphorylated. The varied ratio of WT LC8/ LC8CT is likely because of different expression levels from the randomly inserted plasmid. CC1331-like transformants are likely because of the failed integration of the WT LC8 gene from the transforming plasmid.
Based on backcross, sequencing, and transformation rescue, we conclude that the read-through mutation of the LC8 gene results in the combined deficiencies in flagellar length (Fig. 1A) On occasion, flagella of unequal lengths (arrowhead) and flagella with swollen tips (arrow in insets) are observed. B, Western blots of isolated axonemes reveal defects in I1 dynein and the RS in CC1331. Different loads of axonemes were included to better reveal the reduction level. The I1-dynein subunit, IC97, is undetectable, whereas the WD repeat subunit, IC140, appears normal. Two representative radial spoke proteins, RSP3 and RSP16, are reduced to less than 50% of the WT level, and the dephosphorylated form of RSP3 is more prominent in CC1331 (lower band in CC1331 RSP3 lane). IC78, LC6, and LC9 in the ODA, DLIC in CD1b, and p28 (a component of several single-headed IDAs) appear normal. and multiple axonemal complexes, including IC97 in I1 (Fig.  1B). Therefore, CC1331 (designated as pf5) and fla14 are defective in the same locus, and pf5 is renamed as fla14-3 after the two null mutants (6). Curiously, the FLA14 gene from strain CC1028, which is listed as a pf5 strain (Chlamydomonas Center), is normal and can rescue CC1331 phenotypes (data not shown). Thus, the paralysis of CC1028 is because of a mutation in a different locus.
In Vitro Analysis of Recombinant LC8 Variants-Structural analyses indicate the two C termini, including Ser 88 , of the rat LC8 dimer are positioned in close proximity to the dimeric interface but not at the target binding sites (Fig. 3A). A phosphomimetic mutation in Ser 88 (Fig. 3A, S88E; yellow residues with side chains) near the C terminus prevents dimerization in vitro because of charge repulsion, suggesting that phosphorylation is a mechanism for regulating LC8 by controlling dimer formation (18).
We postulate that the residues and charges in the additional C-terminal tail of Chlamydomonas LC8CT perturb dimerization leading to the severe phenotypes of the LC8CT strain (fla14-3). To test this, recombinant LC8, LC8CT, and LC8S90E (equivalent to S88E) were analyzed by gel filtration chromatography. As expected, at pH 7, WT LC8 was dimeric, and LC8S90E was largely monomeric (Fig. 3B). However, contrary to our hypothesis, recombinant LC8CT migrates as a slightly larger particle than WT LC8, indicating that LC8CT actually forms stable dimers. As expected, all LC8 constructs were monomeric at pH 3 because of protonation (16). Therefore, the deficiencies in LC8CT flagella are not because of a defect in its dimerization.
Effect of Phosphomimetic LC8S90E on Flagellar Complexes in Vivo-Because Drosophila phosphomimetic LC8S88E cannot bind IC74 of cytoplasmic dynein in vitro, we predict that the equivalent monomeric LC8S90E will result in severe deficiencies in the ODA, I1, and retrograde IFT. To test this, fla14-1 was transformed with a plasmid expressing LC8S90E, WT LC8, LC8CT, or LC8S90A, a variant that cannot be phosphorylated (19).
More than 50 transformants for each construct were screened microscopically. As expected, the WT LC8 gene completely rescued the fla14-1 phenotype (fla14-LC8), whereas the LC8CT transformants (fla14-LC8CT) resembled fla14-3 (CC1331), with short, paralyzed flagella. Interestingly, fla14-LC8CT flagella were slightly longer than those of fla14-1, and both gradually shorten as time progressed (compare CT and null in Fig. 4A), even in the presence of WT levels of the retro-grade motor (see DLIC in Fig. 1B). To our surprise, LC8S90E transformants (fla14-S 90 E) appeared indistinguishable from WT LC8 strain in flagellar length (compare LC8 and S 90 E, Fig.  4D) and overall motility. A detailed analysis of three LC8S90E strains by high speed video microscopy (21) showed that their flagellar waveform was similar to WT LC8 strains, but the beat frequency was slightly slower (Table 1), a common trait of mutants with mild defects in the ODA (32).
To determine whether LC8S90E proteins were incorporated into the axoneme as monomers, isolated axonemes were treated with increasing concentrations of the short length cross-linker 1,5-difluoro-2,4-dinitrobenzene, which has been shown to cross-link axonemal LC8 dimers (10). Western blots showed that, in axonemes, LC8S90E was cross-linked into dimers as does WT LC8 (Fig. 4B), indicating that although LC8S90E is monomeric in vitro, it is incorporated into axonemes as a dimer in vivo.
To determine whether LC8S90E transformants exhibit any axonemal defects, axonemes with LC8S90E and WT LC8 were compared by Western blots. Interestingly, IC97 failed to assemble in LC8S90E axonemes (Fig. 4C), a result similar to that observed in LC8CT axonemes (Fig. 1B). In addition, LC6 and LC9 from the ODA were reduced, a result that differed from LC8CT axonemes. The slight defect in ODA assembly is consistent with the slightly reduced beat frequency observed in these transformants (Table 1). However, in some preparations, these deficiencies were not as severe, suggesting variability in the assembly of these complexes in the LC8S90E strains. LC8S90A strains were indistinguishable from WT controls (data not shown).
Because of the varied severities in the LC8S90E phenotype, we tested if the deficiency could be enhanced by culturing cells in adverse conditions, on agar plates in a dim-lighted room at 32°C for 14 days followed by resuspension in water at 32°C. All strains were assessed as percentage swimmers at three time points after resuspension. Under these conditions, the fla14-1 and fla14-LC8CT strains were entirely immotile, and most cells lacked flagella (Fig. 4D, fla14-1 and LC8CT). Although fast swimmers were noted in the fla14-LC8 and fla14-S90A strains after 30 min, only ϳ10 -50% cells swam in 3.5 h (Fig. 4D, WT  and S 90 A), indicating the culture condition was not favorable for flagellar generation. Importantly, the LC8S90E swimmers did not appear until 3.5 h (Fig. 4D, S 90 E), and the movement was sluggish. Together, these results indicate that LC8S90E affects the ODA more severely than other flagellar complexes. The deficiency becomes more severe and extends to other flagellar complexes when cells are grown in challenging conditions. As additional controls, LC8S90E cells grown in 32 or 24°C growth chambers with bright light and controlled humidity appear normal (data not shown), indicating the enhanced phenotype of LC8S90E is not merely because of temperature sensitivity.

Dissociation of Axonemal Dyneins and RS Complexes in LC8
Mutants-To test if the mutant LC8 molecules affect the stability of flagellar complexes, axonemal dyneins were extracted by 0.6 M NaCl followed by velocity sedimentation in 5-20% sucrose gradients and Western blot analyses (Fig. 5A). As anticipated for WT, the ODA and I1 sedimented as intact ϳ20 S particles (represented by IC78 and IC140, respectively), and LC8 primarily co-sedimented with the abundant ODA. In the FIGURE 2. CC1331 phenotype is because of a read-through mutation in the LC8 gene. A, motility and Western blots of isolated axonemes from representative progeny derived from a cross between CC1331 and WT reveal only parental di-type tetrads, twitching flagella (T) characteristic of the CC1331 phenotype and wild-type swimming cells (S). The phenotypes, including twitching short flagella (T), dephosphorylated RSP3 (dot), and the absence of IC97, strictly co-segregate. B, comparison of the C-terminal sequences of LC8 in CC1331 and WT. The stop codon in the LC8 gene is mutated into a leucine codon (asterisk), resulting in the addition of 23 amino acids at the C terminus. This C-terminal extension may alter the original 5th ␤-strand into a helix (h). C, Western blot shows that the LC8 with a C-terminal extension (LC8CT) is present in CC1331 axoneme and slightly larger than WT LC8. The middle lane contains both WT and CC1331 axonemes. D, representative Western blots show that the WT LC8 gene rescues the motility and axonemal deficiencies of CC1331. In the axonemes of the transformants, WT LC8 co-assembles with the larger endogenous LC8CT. The ratio of WT:LC8CT correlates with the percentage of swimmers and the amount of IC97 and phosphorylated RSP3. Interestingly, 50% assembly of WT LC8 rescues the flagellar phenotypes (B1). Like the CC1331 parental strain, the CC1331-like strains (H8, G1, and H7) yield less axonemes (see tubulin protein stain), but dephosphorylated RSP3 (dot), typical of CC1331 axonemes, is still observed.

FIGURE 3. Two C-terminal mutations affect LC8 dimerization differently.
A, tertiary structure modeling illustrates the location of Ser 88 (yellow residues with side chains) and the C terminus near the dimeric interface. The diagram is based on the NMR study of the rat LC8 dimer (blue and green) and Bim peptide (pink) (12) and was generated by the web-based program, Protein Data Bank Workshop 1.50. B, gel filtration shows that at pH 7 LC8CT migrates as a larger particle than dimeric WT LC8 and the monomeric phosphomimetic LC8S90E (upper panel). In contrast, at pH 3, all LC8 forms migrate primarily as monomers (bottom panel).
fla14-LC8CT strain, the ODA remained at ϳ20 S. In contrast, IC140 sedimented as a smaller particle, at a position similar to p28 (Fig. 5A, arrow) in single-headed dyneins, and IC138 was found in multiple peaks, indicating the I1 complex is less stable and dissociates upon extraction. Interestingly, in the LC8S90E gradient, even though IC97 was absent, I1 did not dissociate, whereas the ODA did (Fig. 5A, IC78 and IC69). In this case, IC78 sedimented closer to the top of the gradient, although IC69 sedimented at a position similar to IC140 in the LC8CT gradient. Thus, I1 is more sensitive to LC8CT, although the ODA is more sensitive to LC8S90E. Because CD1b dissociates readily under these conditions (27), the stability of this complex was not assessed.
To test the effect of the LC8 mutations on the stability of the RS complexes, axonemes were extracted with 0.5 M KI and then fractionated similarly. As shown by RSP16 Western blot, the RSs sedimented as intact 20 S particles (Fig.  5B). In contrast, in LC8CT extracts, all RSPs tested, including RSP16 and four other RSPs that are located at different subdomains of the RS complex (24), sedimented near the top of the gradient, indicating a drastic disintegration of the RS. Therefore, LC8CT interferes with the assembly of individual RS complexes, phosphorylation of RSP3, and the incorporation of the RS into axonemes (Fig. 1B and Fig. 5B).

DISCUSSION
The major challenges in understanding LC8 function are the severe pleiotropic phenotypes of the previously described LC8 mutants and the lack of in vivo context in the simplified in vitro systems. These constraints are overcome by the two new LC8 mutants and the Chlamydomonas experimental system with defined LC8-containing complexes.
The key results are that phosphomimetic LC8S90E, although monomeric in vitro, is a dimer in vivo and only causes minor defects. In contrast, dimeric LC8CT results in disparate assembly deficiencies and the dissociation of complexes that are otherwise rather stable. These results support the proposed structural role of LC8 in forming a core scaffold (15,16) and shed new light on the regulation and ultimate function of LC8.
LC8 Promotes the Formation of the Retrograde IFT Motor for the Generation of Full-length Flagella-Because the transient short flagella of fla14-1 lack retrograde IFT and the responsible motor CD1b, it was postulated that retrograde IFT recycled the IFT machinery for flagellar generation and maintenance, and thus fla14 generated short flagella that resorb prematurely because of depleted IFT rafts (6). Yet defects in multiple axonemal complexes in fla14-1 could result in short flagella (29). The IFT deficiency model was confirmed by fla14-3 (LC8CT For these experiments, fla14-1 was transformed with constructs to express LC8, LC8CT, LC8S90E, and LC8S90A. A, flagellar lengths of the transformants expressing LC8 variants. Each data point was obtained from measuring 100 flagellated cells from each transformant at different time points after the beginning of the light cycle. As expected, the WT LC8 transformant has full-length flagella (LC8) and the LC8CT transformant has short flagella that are about onehalf the length of wild type and slightly longer than that of fla14-1. The transformants expressing the LC8S90E protein are full-length (S90E). CT, C terminus. B, Western blot of axonemes treated with increasing concentrations of 1,5-difluoro-2,4-dinitrobenzene shows that LC8S90E is cross-linked into a dimer as effectively as WT LC8 (arrowhead). C, Western blots of isolated axonemes demonstrate IC97 from I1 is absent, and LC6 and LC9 from ODA are reduced in axonemes from LC8S90E transformants (lanes #17, #8, and #2). RS assembly appears normal (RSP3 and RSP16). IC140 and p28 are unaffected and thus serve as loading controls. D, differences between LC8S90E transformants and the control, LC8S90A, and WT become obvious when cultured in a dimly lighted 32°C room. The histogram shows the percentage of swimmers after resuspension of cells from agar plate. Note under this condition, 10 -50% of the cells in the control groups can swim after 0.5-3 h, compared with ϳ100% in the standard laboratory condition. However, only ϳ10% LC8S90E becomes motile after 3.5 h. Cell numbers counted from five randomly chosen fields were averaged and rounded off to reflect the approximate assessment of moving cells. NOVEMBER 6, 2009 • VOLUME 284 • NUMBER 45 strain) that also displayed a similar length phenotype but milder axonemal defects that are insufficient to affect flagellar length. Interestingly, the amount of DLIC appears normal in LC8CT flagella (Fig. 1B) indicating that the retrograde IFT motor can enter flagella, but its function is still impaired. Importantly, the defects in the IFT motor and axonemal complexes appear independent.

Function and Regulation of LC8
LC8 Directly Promotes the Proper Assembly of Individual Complexes at Multiple Levels-One theory proposes that numerous LC8 target proteins are the cargoes of dynein motors and that the bivalent LC8 dimers adhere the motors and the cargoes together. This idea is consistent with the deficiencies of the LC8 null flagella in which the retrograde IFT motor, CD1b, is absent and its putative axonemal cargoes are reduced or absent. However, defective motorcargo coupling cannot explain the defects in LC8CT-containing flagella in which retrograde IFT is dysfunctional, although the ODA appears largely normal; the RS is unstable (Fig. 5B) and poorly incorporated into axonemes, and I1 is only missing a single subunit (20) (Fig. 1), and it dissociates upon extraction (Fig. 5A). Such disparate structural deficiencies in the four flagellar complexes are indicative of individual assembly defects and strongly suggest that LC8, normal or mutated, is in a critical position in each flagellar complex to directly affect the assembly unique to each complex.
These deficiencies can be partially explained by the model that the LC8 dimer stabilizes or refolds dimeric scaffold molecules in molecular complexes (15,16). It is conceivable that the LC8 dimer associates with putative scaffold proteins, such as WDrepeat ICs in dyneins (30) (Fig. 6, left) and unknown proteins in the RS. This binding can enhance the stability of dimeric scaffold proteins and transform the flanking disordered regions into a conformation that favors the association with additional molecules, such as IC97 in I1, the LCs in the ODA (Fig. 4), RSP13/15 in the RS (20), IFT particle proteins in the case of CD1b, or docking proteins responsible for anchoring axonemal complexes. Thus, through modulation of target proteins, LC8 promotes the assembly of various complexes and higher level complexes. When LC8 is mutated, such as LC8CT, the additional sequence (Fig. 6, right, red triangles) at the dimeric interface may act as a wedge that alters the dimeric conformation or creates steric hindrance, consequently reducing the stability of the complex and interfering with further interactions.
As instrumental and important as LC8 is, it is critical to attribute its contribution objectively to understand the grander scheme of molecular assembly. Contrary to the prediction (33), LC8 is not necessary for initiating assembly and may not be involved in the initiation process at all. Studies of fla14-1 showed that without LC8, flagellar LC8 complexes actually assemble, although the process is defective at different steps for each complex. For example, the subunits of CD1b are dispersed in the cell body of a DLIC mutant (25,27,34), indicating that this subunit is essential for assembly of the retrograde IFT motor. In contrast, in WT and the LC8 null mutant (fla14-1), the other subunits of CD1b are concentrated around the peribasal body region. Thus, in the absence of LC8, the retrograde IFT motor is assembled into a defective complex that cannot enter flagella. In the case of the RS, the precursor RS complex enters flagella (26) but cannot be incorporated into axonemes, resulting in the absence of RS structures in fla14-1 axonemes. I1 is assembled in fla14-1 axonemes, although it is less abundant (6). Thus, the modulation of target proteins by LC8 is to correctly promote further assembly rather than to initiate the assembly of complexes.
The unique subunits in I1 also caution the strict interpretation that LC8 target proteins only function as molecular scaffolds (16). The actual target proteins of LC8 in I1 have not been determined. However, neither IC138, a candidate scaffold molecule with WD repeats, nor IC97, which is most sensitive to the LC8 mutations, is required for I1 assembly (31,35). Instead, both are implicated in phosphoregulation of I1 (31,35,36). These unique subunits explain the less severe defects of I1 in fla14-1 and demonstrate the varied roles of LC8 target proteins.
Differential Accommodation of Mutated LC8-The scaffold model portrays LC8 actively refolding target proteins (16). This study shows the individual ability of each flagellar complex to accommodate the modified LC8 molecules. For example, the ODA accommodates LC8CT better than LC8S90E, whereas the other complexes accommodate LC8S90E better than LC8CT.
Two possibilities can explain these differences. Despite a common LC8-binding motif (37,38), the actual LC8-binding sequences are longer and rather diverged (11) and may have different affinity for mutated LC8 proteins. Furthermore, the molecular context of each LC8 complex, which cannot be tested in vitro, may influence how severe a particular complex will be affected by each LC8 mutation. For example, it was postulated that five tightly packed LC8 dimers promote the folding of the dimeric scaffold proteins in the yeast nuclear pore complex (2). Similarly, two LC8-binding sites are immediately adjacent to each other in the WD repeat subunit of CD1b (34) that is particularly sensitive to LC8CT. It is conceivable that multiple tightly packed LC8CT dimers with additional tails will amplify the impact on the target proteins and complexes.
In contrast, the target proteins and neighboring subunits of each complex, which are also involved in assembly of the scaffold (15,30,39), may contribute to the ability of each complex in accommodating a particular LC8 mutation, particularly the phosphomimetic LC8 that is inherently monomeric in vitro (Fig. 3), thus reducing or eliminating the impact of modified LC8.
Limited Effects of LC8 Phosphoregulation-It has been proposed that LC8 phosphorylation is a selective regulatory mechanism (18). The subtle phenotype in the LC8S90E phosphomi- FIGURE 5. Differential instability of axonemal complexes in each LC8 mutant. Axonemal complexes were extracted by NaCl (A) or KI (B) buffer and separated by velocity sedimentation through a 5-20% sucrose gradient. The gradient fractions were analyzed by Western blots. A, differential dissociation of ODA and I1. The ODA and I1 dynein from WT LC8 axonemes represented by IC78 and IC140, respectively, sediment at ϳ20 S (arrowhead). LC8 primarily co-sediments with the ODA. In an LC8CT transformant (CC1331 as well, not shown), I1 dissociates; both IC138 and IC140 sediment as distinct smaller particles. In contrast, the ODA remain intact at ϳ20 S. Arrow indicates the p28 peak for single-headed dyneins. In the LC8S90E gradient, the ODA dissociates; both IC78 and IC69 sediment as distinct particles (arrowheads in IC78 and IC69 panels), whereas I1 remains stable (arrowhead in IC140 panel). LC8 largely co-sediments with IC69 of the ODA (arrowhead in LC8 panel). The LC8S90A gradient is similar to that of WT. B, RS from LC8CT axonemes dissociate upon extraction. Represented by RSP16, the RS in fla14-LC8, fla14-LC8S90E, and fla14-LC8S90A extracts sediments as an ϳ20 S particle. In contrast, RSPs extracted from LC8CT strains sediment near the top of the gradient indicating a drastic dissociation of the RS complex. FIGURE 6. Model explaining the influence of dimeric LC8 and LC8CT on molecular complexes. Binding of the WT LC8 dimer (left) to the disordered regions (squiggly lines) of target proteins (such as the dynein ICs) induces refolding and thus promotes the association of additional subunits, e.g. IC97 in I1 dynein and the LC6 dimer in ODA. Mutant LC8 (LC8CT, right) binds to its target proteins (the ICs), but the additional C-terminal tails at the interface (triangles) interfere with the proper structural changes required for recruiting IC97. Binding of the LC6 dimer is not affected by this mutation and thus no impact on the ODA is observed. The inherent differences among scaffold proteins and flanking molecules may determine the differential effects of LC8 modifications, including phosphorylation. metic mutants suggests that the phospho-regulation is more restricted than anticipated. One idea is that LC8 phosphorylation induces a monomeric shift that preferentially interferes with the binding of lower affinitive target proteins, such as the intermediate chain target peptides from Drosophila dynein (18). Consistently, the ODA and I1 in Chlamydomonas flagellar are abnormal (this study). However, the important differences are that the flagellar dyneins incorporate LC8S90E as dimers (Fig. 4B), and the defects are insufficient to cause obvious functional deficiencies in Chlamydomonas flagella in standard laboratory growth conditions (Fig. 4A). Although the phosphomimetic mutation is not identical to phosphorylation, these modest effects from LC8S90E raise doubts on the effectiveness of LC8 phosphorylation in regulating flagellar functions and molecular complexes in general. In fact, to date, we have not been able to identify phosphorylated LC8 in Chlamydomonas flagella (data not shown). The strains with LC8S90A that cannot be phosphorylated appear normal as well. This tolerance of phosphorylated LC8 by most complexes could be advantageous such that phosphorylated LC8 could primarily affect a small number of target proteins, like apoptotic factor, to promote cell survival (17,19) without causing deleterious side effects such as the disruption of dynein motors.
The enhanced phenotypes of the phosphomimetic mutants in adverse conditions suggest that the conserved Ser 90 is important but not for phosphorylation. The deficiency of S90E flagella became more severe under dimmer light and warmer growth condition that even challenges WT (Fig. 4D) but is not unusual in natural habitats. Yet it is under this condition that the difference between WT and the phosphomimetic mutant becomes evident. Perhaps the phosphomimetic LC8 is less effective in promoting assembly. The inefficiency is masked under ideal laboratory conditions but becomes apparent in challenging situations in which myriads of cellular reactions, such as protein synthesis, are critically reduced. As shown in a structural study, the hydroxyl group of Ser 88 at the interface may contribute a hydrogen bond (18) that favors the dimerization to effectively promote the proper assembly of molecular complexes under diverse conditions.
As small as LC8 is, it has been assigned an accolade of titles. By taking advantage of Chlamydomonas mutants and the well characterized LC8-containing flagellar complexes, this study has provided the critical in vivo evidence and the molecular context to substantiate the structural model for LC8 function, further revealing the ultimate function for LC8 in fostering the proper assembly of flagellar complexes and the limit of postulated phosphoregulation. These conclusions are likely applicable to diverse complexes as well as higher level complexes that contain LC8 in other systems (40,41).