Chlamydomonas fla mutants reveal a link between deflagellation and intraflagellar transport

Background Cilia and flagella are often lost in anticipation of mitosis or in response to stress. There are two ways that a cell can lose its flagella: resorption or deflagellation. Deflagellation involves active severing of the axoneme at the base of the flagellum; this process is defective in Chlamydomonas fa mutants. In contrast, resorption has been thought to occur as a consequence of constitutive disassembly at the tip in the absence of continued assembly, which requires intraflagellar transport (IFT). Chlamydomonas fla mutants are unable to build and maintain flagella due to defects in IFT. Results fla10 cells, which are defective in kinesin-II, the anterograde IFT motor, resorb their flagella at the restrictive temperature (33°C), as previously reported. We find that in standard media containing ~300 microM calcium, fla10 cells lose flagella by deflagellation at 33°C. This temperature-induced deflagellation of a fla mutant is not predicted by the IFT-based model for flagellar length control. Other fla mutants behave similarly, losing their flagella by deflagellation instead of resorption, if adequate calcium is available. These data suggest a new model whereby flagellar resorption involves active disassembly at the base of the flagellum via a mechanism with components in common with the severing machinery of deflagellation. As predicted by this model, we discovered that deflagellation stimuli induce resorption if deflagellation is blocked either by mutation in a FA gene or by lack of calcium. Further support for this model comes from our discovery that fla10-fa double mutants resorb their flagella more slowly than fla10 mutants. Conclusions Deflagellation of the fla10 mutant at the restrictive temperature is indicative of an active disassembly signal, which can manifest as either resorption or deflagellation. We propose that when IFT is halted by either an inactivating mutation or a cellular signal, active flagellar disassembly is initiated. This active disassembly is distinct from the constitutive disassembly which plays a role in flagellar length control.


Background
Intraflagellar transport (IFT) was first characterized in the unicellular green alga Chlamydomonas [1] and has since been shown to be required for flagellar assembly in a variety of systems [2,3]. IFT is the bidirectional movement of large protein complexes (IFT particles) along the flagellar axoneme, and has anterograde and retrograde components mediated by the plus and minus-end directed microtubule motors kinesin-II and cytoplasmic dynein, respectively [reviewed in [4,5]]. In Chlamydomonas, null mutations in genes necessary for activity of either kinesin-II or cytoplasmic dynein result in bald (flagella-less) cells or cells with very short, abnormal flagella [6,7]. Retrograde IFT is not proposed to be directly involved in disassembly, but rather is necessary to recycle IFT particles [7,8].
A model for flagellar length control has been proposed wherein anterograde IFT is required for transport of axonemal precursors to the distal tip of the flagellum; these precursors are necessary both for de novo flagellar assembly and to offset the constitutive disassembly that occurs at the tips of flagella [8]. This model suggests that the steady-state length of a flagellum is determined kinetically by the relative contributions of assembly, mediated by anterograde IFT, and disassembly at the tip, which is IFT-independent [9]. Thus, the phenotype of Chlamydomonas long flagella mutants could be a result of either an upregulation of anterograde IFT, or due to a decrease in the rate of disassembly at the tip [8][9][10].
A Chlamydomonas temperature-sensitive mutant for flagellar assembly, fla10, has been characterized as having a lesion in a subunit of kinesin-II [11]. fla10 cells have wildtype flagella at the permissive temperature (20°C), but are bald at the restrictive temperature (33°C) [12]. In agreement with the length control model, fla10 cells incubated at an intermediate temperature have intermediate-length flagella [8]. It has been accepted that the flagella of fla10 cells resorb at the restrictive temperature due to continued disassembly in the absence of anterograde IFT [e.g., [13]].
Several other temperature-sensitive flagellar assembly mutants (fla mutants) are available in Chlamydomonas [12,14,15]. Unlike fla10, the genes for these mutants have not yet been identified. However, like fla10, these fla mutants have been shown to have defects in IFT, even at the permissive temperature [15,16]. The majority of the fla mutants have been reported to undergo flagellar resorption at 33°C [12,15,16], presumably due to defects at different points in the IFT cycle disrupting flagellar assembly. Most fla mutants are unable to regenerate flagella at 33°C, with the exception of fla2 [12]. fla2 is also exceptional in that it has been observed to deflagellate, rather than resorb, at the restrictive temperature [2,12,14].
Deflagellation, like IFT, is a conserved process in eukaryotic cilia and flagella [reviewed in [17]. Deflagellation/ deciliation is the regulated severing of the axoneme, and has been shown to occur in response to a number of stimuli, including pH shock, 42°C heat shock, and treatment with dibucaine or alcian blue. Calcium plays a central role in signalling deflagellation: calcium influx mediates acid shock-induced deflagellation [18], and axonemal severing can be induced in vitro in response to calcium [19]. Our lab has cloned two essential components of the deflagellation pathway, the flagellar autotomy genes fa1 and fa2 [20,21]. fa mutants do not undergo deflagellation in response to any known stimulus. We have also demonstrated that the microtubule-severing ATPase katanin is likely to mediate axonemal severing during deflagellation [19,22]. We have localized both katanin and Fa1p to the site of deflagellation, the flagellar transition zone between the basal body and the flagellum proper [ [20], M. Mahjoub and LMQ, unpublished observations].
In many ciliated cells, including vertebrate cells and Chlamydomonas, flagella/cilia are shed or resorbed prior to mitosis [23][24][25][26]. It has been proposed that mammalian primary cilia are important to maintain cells in a differentiated state; the best-characterized example is the role of kidney epithelial cilia in models of polycystic kidney disease [reviewed in [27]]. Therefore, as has already been demonstrated for flagellar assembly [3,27], genes required for flagellar resorption or deflagellation may be implicated in disease and/or development.
It is unknown what determines whether a given cell type deflagellates or resorbs prior to mitosis. Indeed, even in the same organism, flagellar loss may occur via different mechanisms at different stages of the life cycle [23]. Flagellar resorption and deflagellation have been thought to occur by dramatically different mechanisms and at different locations: resorption via disassembly at the tip [8,13] and deflagellation by severing at the base [28]. However, a role for severing activity at the base during resorption has been suggested by elegant EM studies performed on cells resorbing their flagella prior to mitosis [29]. We now provide further evidence that flagellar resorption involves severing activity at the base, and that active disassembly by resorption is distinct from the constitutive disassembly involved in flagellar length control.
Our examinations of the phenotypes of fla10 and fla2, reported here, lead us to the conclusion that flagellar resorption of fla mutants results from active disassembly, rather than constitutive disassembly in the absence of IFT. We show that in common culture medium for Chlamydomonas, the predominant mode of flagellar loss at 33°C for fla10 is deflagellation. Blocking the ability to deflagellate, either by lowering the extracellular calcium concentration or by genetically blocking the deflagellation pathway, causes fla10 cells to resorb their flagella at the restrictive temperature. Likewise, we demonstrate that fla2 cells undergo flagellar loss via deflagellation, and that resorption occurs in low calcium or by genetically introducing a fa mutation. Furthermore, fla10-fa1 and fla10-fa2 double mutant populations are defective for resorption, as many cells retain flagella past the time when fla10 mutants are bald. Finally, we find that flagellar resorption can be induced in cells unable to deflagellate in response to deflagellation stimuli. Our findings indicate that the deflagellation and resorption pathways are not separate, but that each can result from a disassembly signal which culminates at the flagellar transition zone. We propose that IFT and flagellar disassembly share common regulatory elements.

Results and Discussion
fla10 and fla2 deflagellate at the restrictive temperature Cultures of fla2 cells in TAP medium contain free flagella after incubation at the restrictive temperature of 33°C. Due to their small size, and tendency to stick to each other and to attached flagella, free flagella are difficult to accurately identify; however, as fla2 cells can regenerate and redeflagellate at 33°C, the appearance of abundant free flagella is quite obvious. Our initial observations of the fla10 temperature-sensitive phenotype led us directly to the conclusion that this mutant also deflagellates, as many free flagella are seen in the presence of calcium at 33°C in both fla10 and fla2 cultures ( Figure 1).
As fla10 has only been previously characterized as resorbing flagella at 33°C, and given our knowledge of the role played by calcium in the deflagellation pathway, we reasoned that deflagellation would be prevented if fla10 cells were incubated at 33°C in buffer lacking added calcium.
We found this indeed to be the case; there is no indication of deflagellation by fla10 cells at 33°C in the absence of calcium, yet cultures are still mostly bald after ~6 hours (Figures 1, 2). We extended our observations by demonstrating that fla2 cells also do not deflagellate at 33°C in the absence of calcium ( Figure 1). Adding calcium to minimal buffer to a concentration approximately that of TAP media (340 µM) [30] restored the deflagellation phenotype of both fla mutants. As a further control, blocking the deflagellation response by creating double mutants of either fla10 or fla2 with either of fa1 or fa2 also prevented the appearance of free flagella after six hours ( Figure 1 and data not shown). Wild-type cells do not deflagellate under these conditions.
Having examined the culture media under various conditions, we turned our attention to the cells themselves. The average flagellar lengths of fla10 and fla2 cells, but not wild-type cells, decrease at 33°C (Figure 2). We have shown that the fla mutants deflagellate at 33°C in the presence of added calcium, yet the differences in the average flagellar length plots are difficult to discern whether or not calcium was added, especially for fla10 (diamonds in Figure 2A,2B). This similarity of the average flagellar To better illustrate the effects of calcium at the restrictive temperature on the fla mutants, we have chosen a novel means of presenting flagellar length data. By plotting the length of the longer flagellum versus the length of the shorter flagellum for fla10, fla2, and wild-type cells (Figures 3, 4, and 5, respectively), we can more clearly observe the states of individual cells during timecourses at the restrictive temperature. Points representing bald cells fall at the origin, while points representing uniflagellate cells are found on the x-axis. Additionally, we have plotted the percent of biflagellate, uniflagellate, and bald cells in the population for each timepoint as insets in the scatter plots. This presentation allows one to quickly assess the flagellation state of a population, and changes in that state over time.  Figure 2C).
After three hours at 33°C, cells in each population have shorter flagella, and there is an increase in the numbers of uniflagellate and bald cells in both populations. At this time in nominally calcium-free buffer, several cells with flagella shorter than 4 µm are observed. These short flagella indicate that resorption is occurring. Conversely, in the presence of calcium, no flagella <4 µm are seen, suggesting that cells deflagellate before flagellar resorption is complete. fla10 cells incubated at 20°C did not undergo significant flagellar loss or shortening, and after 6 hours at 20°C, populations resembled zero-hour timepoints (bottom panels in Figure 3). Figure 4 is a timecourse for the fla2 mutant, which was previously reported to deflagellate at the restrictive temperature [2,12,14]. After 6 hours in HEPES at 33°C, but not at 20°C, over one-quarter of the fla2 cells are bald. Fewer cells with flagella >9 µm are seen after 6 hours at 33°C compared with earlier timepoints and 20°C controls. At the three hour timepoint in the presence of calcium, there is a dramatic increase in bald cells in the fla2 population (Figure 4, inset). In contrast to fla10, fla2 cells are able to regenerate flagella at the restrictive temperature [ [12,14], and data not shown]. This regeneration accounts for the increase in the number of cells with short flagella in the fla2 population at later timepoints. After overnight incubation at 33°C, fla2 cultures are often entirely bald. Figure 5 shows the flagellar lengths of wild-type cells after a shift to 33°C. In contrast to the fla mutants, wild-type flagella do not undergo appreciable changes in length at 33°C. In principle, the resorption phenotype of fla10 at 33°C could be solely due to constitutive disassembly in the absence of anterograde IFT, and the deflagellation phenotype a secondary consequence due to the stress of elevated temperature. However, wild-type cells do not deflagellate at 33°C (Figures 1, 5), although they will deflagellate at 42°C (data not shown).

fla10-fa double mutants are slow to resorb flagella
The deflagellation phenotype of fla10 and fla2 at 33°C in calcium is blocked by either fa mutation (Figure 1 and data not shown). We wished to examine double mutant cell populations, and because fla2 is able to regenerate flagella at the restrictive temperature, we focused on fla10fa mutants. We find that both fla10-fa1 and fla10-fa2 double mutants resorb flagella more slowly than fla10 at 33°C ( Figure 6). This held true whether or not calcium was present in the medium, and, as for the fla10 single mutant, similar flagellar loss kinetics were observed regardless of calcium ( Figure 3 and data not shown). This result is reminiscent of the partial block of fla10 flagellar resorption by six different extragenic suppressors [31]; it is unknown whether any of these suppressors are fa genes. The prevalence of uniflagellate cells, as well as flagella of unequal lengths, is at odds with the classical picture of Chlamydomonas long-zero flagellar dynamics [8,32]. This model predicts that cells with flagella of unequal length should be rarely seen: the combination of length-control kinetics, and the fact that the two flagella share a common pool of precursors, means that any length imbalance between the flagella of a single cell will be corrected [8]. However, long-zero experiments are typically carried out in conditions permissive for flagellar regeneration, which (page number not for citation purposes) Scatter plots of fla10 timecourses in HEPES and HEPES+Ca is not the case for fla10 mutants at the restrictive temperature. In agreement with this, fla2 cultures contain fewer uniflagellate cells than fla10 cultures at the restrictive temperature, and this is likely due to the ability of fla2, but not fla10, to reassemble flagella at the restrictive temperature.
The observation of unequal shortening of flagella in a resorbing population (Figure 6) parallels the observation that one flagellum is prone to deflagellate before the other (Figures 3, 4, 6 and data not shown). Loss of one flagellum at a time is not observed in response to a strong deflagellation signal, but is observed after treatment with weaker stimuli such as a mild acid shock (data not shown) or threshold concentrations of alcian blue [33]. The two flagella of Chlamydomonas are not equivalent; the cis flagellum is defined as the one closest to the eyespot. Rather than the random loss of either flagellum, we hypothesize that the cis and trans flagella are differentially sensitive to disassembly signals, as they are to motility-related calcium signals [34]. This hypothesis predicts that disassembly will be initiated earlier and/or proceed more rapidly in one flagellum, which could explain the abundance of unequal length flagella observed in Figure 6.

Resorption is mediated at the transition zone
On the basis of a survey of the modes of flagellar loss in unicellular algae and fungi, Bloodgood divided flagellar loss into several categories [23]. Examples were cited of the same organism (or even the same cell) undergoing flagellar loss sometimes by resorption and sometimes by deflagellation [23]. Rather than supposing that organisms undergo flagellar loss by different mechanisms at different stages of their life cycles, we contend that the mechanism of flagellar loss is conserved, and that subtle variations in signaling leads to the appearance of either deflagellation or various forms of resorption. This conserved mechanism would involve severing of the axoneme at the base of the flagellum.
We now summarize specific experimental data suggesting that resorption is mediated at the transition zone, the site of deflagellation, rather than by constitutive disassembly at flagellar tips. First, fla10 and fla2 mutant cells disassemble their flagella at the restrictive temperature via deflagellation in the presence of calcium, and via resorption in the absence of added calcium. Second, fa mutants, which are unable to deflagellate, are slow to resorb; this effect would not be predicted if resorption were entirely due to dynamics at flagellar tips. The slow resorption phenotype of fa1 is especially informative, as this mutant is slow to assemble flagella (JDKP, Ben Montpetit, LMQ, unpublished observations), in which case the length-control theory predicts that fa1 should be fast to resorb. Third, the slow resorption of fla10-fa double mutant cells pro-duces uniflagellate cells in the absence of deflagellation, a consequence not predicted by the dynamic length control model but in accordance with the results of mild deflagellation treatments. Fourth, while Chlamydomonas cells resorb their flagella prior to mitosis [24], EM studies have provided evidence that flagella are detached from basal bodies, but not lost from cells, during pre-mitotic resorption [29]. This seeming contradiction may be explained by our proposal that resorption requires microtubule severing activity at the flagellar transition zone.
There are also some theoretical considerations. Not only the tips of the outer doublet microtubules, but rather the whole axoneme, has been shown to undergo turnover in sea urchin embryos [35]. If this turnover occurs in all eukaryotic flagella, then flagellar tip dynamics alone [8] or lattice translocation models [36] could potentially account for this turnover. However, we believe a regulated ratcheting of the axoneme at the transition zone, coupled with dynamic activity at the tip, provides the most satisfying explanation for the occurrence of axonemal turnover even while a flagellum remains functional. Additionally, there is the consideration that flagellar length in Chlamydomonas is cell-cycle dependent, and flagella undergo a period of slow resorption prior to a period of fast resorption preceding cell division [8]. We hypothesize that the slow phase of resorption is mediated by the length-control mechanism, while the fast phase of resorption is mediated by severing of axonemal microtubules in the transition zone.
Acid shock causes resorption of mutants unable to deflagellate Figure 7 Acid shock causes resorption of mutants unable to deflagellate. At the indicated times post-acid shock, flagellar length distributions were determined and percents of cells with long flagella (flagella longer than 7 µm) were plotted.

The flagellar disassembly pathway
The deflagellation phenotype of fla10 is significant, as it implies an active signalling event is at work. We predicted that if the deflagellation severing complex is responsible for resorption, then providing a deflagellation stimulus to a cell unable to deflagellate should induce resorption. Indeed, this is consistent with temperature-sensitive resorption of fla10 in the absence of added calcium, and with temperature-induced resorption of fla10-fa double mutants with or without added calcium. To investigate whether this effect is more general, or only due to some special nature of the fla phenotype, we examined the effect of pH shock on mutants unable to respond by deflagellation. As previously shown, acid treatment of fa1, fa2, or adf1 mutants does not lead to deflagellation as it does for wild-type cells [37,38]. However, as predicted by our model, these mutants do indeed resorb their flagella within one hour after a 30 s acid shock (Figure 7). This resorption is not necessarily complete, and is complicated by the tendency of the acid-treated fa cells to coil their flagella starting from the distal ends. These coiled flagella get smaller over time, and before disappearing entirely they become dense, stumpy structures. Sanders and Salisbury noted that the Chlamydomonas centrin mutant, vfl-2, will sometimes resorb rather than deflagellate in response to the deflagellation agent, dibucaine [39]. We have further extended these results by performing deflagellation experiments on wild-type cells in low-calcium buffer. Acid shock, dibucaine treatment, and 42°C heat shock all induce flagellar resorption of wild-type cells in low calcium buffer (data not shown).
We conclude that resorption and deflagellation both may result from a "flagellar disassembly" signal. If the fast phase of flagellar resorption results from disassembly mediated by the severing complex at the transition zone, as we have argued, then a new picture of deflagellation emerges: calcium-mediated hyperactivation of the severing complex following a disassembly signal leads to deflagellation. This model exposes our lack of understanding of the endogenous signaling pathways that lead to flagellar disassembly. Flagellar disassembly occurring prior to mitosis or meiosis in Chlamydomonas must occur via resorption, even in medium containing calcium, as cultures typically contain few free flagella. If agents such as intracellular acidification, dibucaine, and heat shock induce deflagellation by hyperactivation of an endogenous disassembly pathway by causing an increase in intracellular calcium, then why should similar treatments in low calcium cause resorption? Two nonexclusive possibilities present themselves: either the severing apparatus has progressively greater activity at higher calcium thresholds, and has partial activity sufficient to mediate resorption without full calcium activation; or there is a calcium-independent signaling pathway which is induced by treatments such as intracellular acidification.
A candidate for providing such a signal is the IFT pathway. Recent work has suggested that IFT has signaling functions [40], and it has been suggested that these signalling functions are perhaps more important for flagellar assembly than the physical transport of flagellar components [5].
Halting IFT in the presence of calcium induces deflagellation, as we have shown with the IFT mutant fla10, in which IFT halts quickly after shift to the restrictive temperature [41]. fla10 and fla2 are not the only IFT mutants which deflagellate; we have observed deflagellation at the restrictive temperature for several other fla mutants (data not shown). As most fla mutants have defects at some point in the IFT cycle at the permissive temperature [16], it is likely that IFT is disrupted in these mutants at the restrictive temperature. As first mentioned anecdotally by Kozminski et al. [41], halting IFT seems to lead to deflagellation.
As IFT is so intimately involved in flagellar assembly, it should not come as a surprise that it may play a role in flagellar disassembly. It is intuitive that a cell should not attempt to continue building a flagellum at a time when it is actively disassembling a flagellum, and this logic suggests that signalling events that regulate IFT should be related to flagellar disassembly signals. Deflagellation of fla10 at the restrictive temperature, when anterograde IFT is halted, raises the possibility that functional IFT counteracts a constitutive disassembly signal. This could be adaptive. For example, for cells with damaged flagella, deflagellation would be the favored disassembly response, due to its speed and ability of Chlamydomonas to upregulate genes required for flagellar assembly postdeflagellation. If deflagellation is blocked, flagellar disassembly will instead occur via resorption.

Conclusions
We find that temperature-sensitive flagellar loss in the Chlamydomonas IFT mutants fla10 and fla2 is not due to constitutive resorption at flagellar tips, which plays a crucial role in flagellar length control, but rather due to activation of a disassembly pathway (Figure 8). We also find that flagellar disassembly can occur either by deflagellation or resorption in response to the same stimuli, and that deflagellation is preferred if sufficient calcium is available. Mutants unable to deflagellate are also slow to resorb flagella. We propose that resorption is actively mediated in the transition zone at the base of the flagellum, and that cellular signals which regulate IFT may also regulate flagellar disassembly. Under conditions appropriate for flagellar assembly (Yes), the decision is made to activate IFT and block disassembly at the transition zone. Active IFT could provide a signal to inhibit active disassembly. When flagella are to be disassembled (No), IFT is inactivated. In the absence of IFT, an inhibitory signal would be removed, allowing disassembly to proceed. There may also be direct activation of disassembly.