Tubulin in sea urchin embryonic cilia: post-translational modifications during regeneration

Tubulin is the major constituent of the detergent-solubilized ciliary membrane/periaxonemal matrix fraction from sea urchin embryos, accounting for >1/4 of the tubulin in the assembled organelle (Stephens, 1991). Synthetically, this tubulin initially behaves as if it were derived from a pool separate from that of the axoneme, as judged from labeling studies done during regeneration, but in successive regenerations the two fractions become indistinguishable. This behavior is easily rationalized if one assumes that the axoneme is initially constructed with mainly unlabeled (i.e. presynthesized) tubulin during early growth while the membrane/matrix reflects the total pool labeling after growth.

In mature cilia, both tubulin fractions are also indistinguishable with respect to their levels of a- tubulin acetylation and detyrosination (Stephens, 1990a, 1991). The former post-translational modification in Chlamydomonas, involving acetylation of lysine 40 (LeDizet and Piperno, 1987), distinguishes cytoplasmic or soluble α tubulin from axonemal tubulin (L’Hemault and Rosenbaum, 1983; Piperno and Fuller,1985) while the latter, involving removal of the C-terminal tyrosine, typically occurs on assembled micro tubules in a variety of systems (Gunderson et al., 1987,1989). In Chlamydomonas, the acetyltransferase enzyme is found tightly bound to the axoneme while both the deacetylase and an inhibitor of acetylation are cytoplasmic (Maruta et al., 1986). The localization of the detyrosinating enzyme has not been explored with respect to cilia or flagella.

To investigate the potential role of membrane/matrix tubulin as a precursor to axonemal tubulin, I undertook a study of detyrosination and acetylation during active ciliary regeneration. A preliminary report of this work has appeared in abstract form (Stephens, 1990b).

Gastrula stage Strongylocentrotus droebachiensis embryos, grown as mass cultures at 8°C in 500 ml tissue culture vessels, were deciliated by brief hypertonic salt treatment and returned to normal sea water for synchronous regeneration. The regrowing cilia were isolated by hypertonic salt treatment at specific time intervals from samples of the culture, each containing 1–2 ml of embryos. The cilia were harvested and purified by differential centrifugation (Stephens, 1986, 1991). Three-times-recycled cytoplasmic tubulin was prepared from mid-gastrula embryos by the method of Suprenant and Marsh (1987) with only minor modification (Stephens, 1991).

In order to perform membrane extractions or gel analyses with similar levels of protein, cilia were extracted for 2 min at 0°C with a volume of 0.25% NP-40 in 3 mM MgCh, 30 mM Tris-HCl (pH 8) proportional to their expected length and then sedimented at 10,000 g for 5 min. The resulting pellets of 9+2 axonemes were resuspended in an equivalent volume of the extraction buffer and concentrated SDS-sample buffer was added to both fractions. The membrane/matrix and axonemal fractions were run pairwise on 8%T/2.66%C polyacrylamide gels, using 0.1% Sigma L-5750 SDS in the Laemmli discontinuous system (Laemmli, 1970). Gels were stained by the equilibrium method of Fairbanks et al. (1971) using Coomassie Blue (Serva). Silver staining was by the method of Heukeshoven and Demick (1985).

For general radioactive detection of proteins, embryos were labeled for 4 h with [³H]leucine. The embryos were washed free of isotope, deciliated, and then allowed to partially regrow cilia. The cilia regenerated from pre-labeled protein pools were isolated, fractionated, resolved by SDS-PAGE, and the labeling patterns were evaluated by fluorography. For 2-dimensional analysis, cilia regenerating for 2 h were labeled for 4 h with [¹⁴C]leucine, chased with unlabeled leucine for 2 h, then isolated. Additional details of these methods have been published (Stephens, 1989, 1991).

Gels were analyzed by the Jandel Video Analysis (JAVA) system (Jandel Scientific, Corte Madera, CA), calibrated with neutral density standards to read optical density directly, using white light transillumination. Each a-tubulin band was placed perpendicularly within a constant, defined area on the monitor (Fig. 2B, below) and an averaged scan (software-defined‘vertical average’) was obtained for the band. This is equivalent to conventional densitométrie scanning along the lane axis, using the full width of the lane. These scans, in turn, were sequentially accumulated and printed as a data set (Fig. 2C). The individual peaks were integrated, either automatically as each scan was tabulated, or manually using a digitizing tablet and the‘area’ feature of the software when a relative baseline correction was required. This operation determines the total integrated optical density, correcting for any nonuniformity or skewing within the band.

After determining the relative amount of a-tubulin in each fraction from every time point pair, all samples were diluted to a common tubulin level, confirmed by densitometry after electrophoresis. From these initial standard dilutions, parallel dilution series were prepared and run together so that subsets of all experimental points would fall within the linear response range of the antibodies tested. Each set also contained either an internal standard of tubulin from mature cilia or one time point from another series for normalization. The sample sets were run in parallel on 2 or 4 small format 8% polyacryl-amide/SDS gels using the Laemmli (1970) formulation and transferred for 1 h at 60 volts to nitrocellulose by the‘Western blotting’ method of Burnette (1981), employing 0.1% Sigma SDS and 20% methanol in the transfer buffer. The protein blots were rinsed in Tris-buffered saline (pH 8) containing 0.05% Tween-20, blocked with 1% bovine serum albumin in the same buffer, and reacted with various primary antibodies (C-terminal Glu and Tyr α tubulin synthetic peptide polyclonal antibodies from Dr. Jeanette Chloe Bulinski; acetyl α tubulin monoclonal antibody 6-11B-1 from Dr. Gianni Pipemo; total a’-tubulin monoclonal antibody DM1A from Amersham). The primary antibodies were visualized with appropriate alkaline phosphatase-coupled secondary antibodies (Promega BioTec). Two-dimensional gel analysis and immunoblotting were performed by the methods of Linck et al. (1987).

The final immunoblots were quantitated with the JAVA system, using the basic method outlined above. In this case of reflectance densitometry, white light illumination was from above. A‘vertical average’ scan was taken at each band and an equivalent background reading, taken adjacent to the band, was subtracted from each. Both wet and dry blots were used, yielding essentially the same relative results. Even though the former showed higher sensitivity (contrast) the latter were generally preferable, due to a more uniform background.

Pellets of ciliary axonemes were fixed with 2.5% glutaraldehyde in 3 mM MgCb, 30 mM Tris-HCl (pH 8) for 1-2 h on ice, rinsed with the same buffer, and post-fixed for 1-2 h with osmium tetroxide/ferrocyanide (cf. Stephens, 1983). After overnight en bloc staining with 1% uranyl acetate, the specimens were dehydrated with ethanol and propylene oxide, embedded in Epon/Araldite, thin-sectioned with a diamond knife, stained with lead citrate, and observed and photographed with a Zeiss 10C electron microscope.

As described earlier (Stephens, 1977), the regenerating cilia of Strongylocentrotus droebachiensis (and other species) initially regrow in a near-linear fashion, then asymptotically approach a final length of about 20 μm. For reference, a typical regeneration is quantitated in Fig. 1, where cilia isolated at various times were measured directly. Regeneration can be repeated sequentially many times, with identical kinetics, while the embryos continue to develop normally. The growth curve extrapolates to a lag time of about 17 min, corresponding to the point when incipient cilia can be observed on the embryo (Stephens, 1977). Isolation of clean cilia shorter than 5 /rμm is difficult due to loss amongst debris of similar size; hence time points at and beyond 2 h were used in the systematic analyses that follow. For 1.5 h time points, an entire culture was processed through two partial regenerations to obtain sufficient mass for two duplicate analyses.

If the tubulin found in the membrane/matrix fraction is the eventual precursor to that of the axoneme, as metabolic studies would suggest, one might expect to see relatively large amounts of tubulin in this fraction during assembly. Consequently, regenerating cilia were analyzed during the near-linear growth phase. An example of one analysis of the detergent-solubilized membrane/matrix and axonemal fractions during growth is given in Fig. 2. Inspection of the gel and a comparison of the relative densitométrie peak heights demonstrate that the amount of tubulin found in the detergent-solubilized membrane/matrix fraction exceeds that of the axoneme in the earliest time point but that the proportion of tubulin in the axoneme fraction rapidly increases and becomes essentially constant beyond 7 h, when the cilia have nearly regrown. Since all samples were resuspended in a volume of buffer proportional to their expected length, the relatively smaller amount of total protein in the 2 h and 3 h time points reflects a progressively lower yield for shorter cilia. Normalizing by length, the amount of protein in the membrane/matrix fraction is, in fact, roughly constant, a point that will be addressed below.

One logical explanation for this basic observation is that newly formed cilia, although fully motile, might be labile and proteins resulting from axonemal breakdown could contaminate the earlier extracts. Three independent lines of evidence argue against this point.

First, the qualitative composition of the membrane/matrix fraction is essentially constant throughout regeneration, as can be demonstrated by silver staining the membrane fractions (Fig. 3A). Although there is some band-shape variability and staining differences characteristic of gel samples containing relatively high amounts of NP-40, the earliest and the latest membrane/matrix time points contain the same major bands, in the same relative ratios, in a pattern distinct from that of the axoneme.

Second, most newly synthesized axonemal proteins are not detected in the membrane/matrix fraction of early regenerates, as illustrated by an autoradiogram of membrane/matrix and axoneme fractions from a 4 h time point derived from pre-labeled embryos (Fig. 3B). The asterisk denotes tektin-A, a heavily labeled, integral component of outer doublet microtubules (Stephens, 1989). The absence of this and most other heavily labeled axonemal proteins in the detergent extract replicates a result obtained earlier with fully regrown cilia (Stephens, 1991). Since the addition of newly synthesized architectural proteins is most likely to be tip-wise (Johnson and Rosenbaum, 1991), this result would argue against a greater lability of ciliary tips.

Third, electron microscopy of randomly chosen survey sections shows no obvious morphological differences, apart from length, between axonemes of cilia regenerating for 3 h and those of fully grown cilia (Fig. 3D vs. 3C). Analyzing cross-sections quantitatively, the control axonemes show 6.9% splitting along the long axis, 2.2% loss of 1 central pair member, and <1% B-subfiber breakdown (n=187). The regenerating axonemes show 11.3% longitudinal splitting, 9.5% loss of 1 central pair member, and <3% B-subfiber breakdown (n=231). In both cases, more than half of the centralpair loss was found in split axonemes, suggesting mechanical disruption rather than solubilization. Total solubilization of both central pair members would account for only one-half of the tubulin found in the fully mature control cilia while total solubilization of the central pair and all B-subfibers would be required to explain the amount of membrane/matrix tubulin found in early regenerates. Furthermore, the tubulin content of early-regeneration axonemes (about 2/3 of the total axonemal protein) is essentially the same as that of controls (Fig. 2). Therefore, the relatively higher proportion of membrane/matrix tubulin in the earlier time points is unlikely to be due to greater breakdown of axonemes in younger cilia.

Using data derived from Fig. 2 (Series I data set, 1 h intervals), an independent early-point data set (Series II, 0.5 h intervals), and 2 single point determinations at 1.5 h, the increase in the axoneme to membrane tubulin ratio with time may be quantitated (Fig. 4). This ratio increases approximately linearly from values of about 0.7–0.8 at the earliest time points to >3 when regeneration is complete. The latter is essentially the same value obtained from mature cilia prior to isolation (Stephens, 1991). The cilia increase in length over this same time period by an equivalent factor. Dividing the axoneme to membrane tubulin ratio by length over the range of 2–7 h produces a value that remains constant during the growth phase (0.127 ± 0.010 s.d., n=11 successive points), indicating that the amount of tubulin per cilium in the membrane/matrix fraction remains relatively constant as the cilium grows by axoneme elongation.

Antibodies prepared against peptides having the sequence of tyrosinated (Tyr) and detyrosinated (Glu) a- tubulin (Gunderson and Bulinski, 1986) allow accurate estimates of detyrosination. Three-times-recycled embryonic cytoplasmic tubulin is almost fully tyrosinated (<3% Glu), as determined by serial dilution and relative cross-reactivity (Glu=Tyr at >1/32 dilution, not shown). Tyr antibody may be used in a dilution series to estimate the relative level of C-terminal tyrosine in axonemal tubulin from virgin cilia. Cytoplasmic tubulin at half the concentration of axonemal tubulin gives the same response as axonemal tubulin in a dilution series on an immunoblot (Fig. 5). This relationship can be shown quantitatively by video analysis of the blot, where the relative (secondary) dilution is plotted against the relative integrated density under reflected light. This type of analysis is the basis for the quantitations that follow.

Since detyrosination is generally considered a characteristic of assembled microtubules, at early time points one might expect axonemal tubulin to undergo this post-translational modification while membrane/matrix tubulin, like cytoplasic tubulin, should be mainly tyrosinated. Membrane/matrix and axonemal samples from all time points, diluted to the same a-tubulin level, were analyzed for the degree of detyrosination by immunoblotting with the Glu antibody. The blots in Fig. 6 illustrate the increasing degree of detyrosination of both membrane/matrix and axonemal tubulin with time for representative early time points from one data set.

Fig. 6 also illustrates the quantitative data for the full range of time points from all data sets, normalized with an internal standard of virgin axonemal tubulin taken as 50% Glu. With the possible exception of the earliest points, tubulin in the membrane/matrix fraction is detyrosinated to essentially the same extent as that incorporated into the axoneme. At 2 h, both fractions are not discernably different from embryonic tubulin (about 2.5% Glu). Note that these data resemble the tubulin ratio data in Fig. 4 in the sense that both axonemal tubulin increase and tubulin detyrosination take place with similar kinetics after an apparent 2 h lag.

It was shown earlier (Stephens, 1990a, 1991) that the membrane/matrix and axonemal tubulin fractions from mature cilia are acetylated to the same extent while the embryonic tubulin from which they must be derived is not detectably acetylated. If it is a precursor to axonemal tubulin, one might expect membrane/matrix tubulin to be less acetylated early in assembly. Using samples adjusted to identical tubulin levels, crossreactivity with the acetyl a-tubulin monoclonal antibody 6-11B-1 (Pipemo and Fuller, 1985) demonstrates that membrane/matrix tubulin from early time points is apparently as acetylated as that from virgin cilia (Fig. 7). On parallel blots, cross-reactivity with the Glu antibody shows the characteristic increasing detyrosination, while a monoclonal antibody to a-tubulin (DM1A; Blose et al., 1984) confirms that the relative amount of tubulin is the same in all samples after serial dilution and immunoblotting.

Expressed quantitatively, the degree of acetylation of membrane/matrix and axonemal tubulin is the same in both fractions and remains relatively constant throughout ciliary regeneration, as can be illustrated by plotting acetylation against time after deciHation (Fig. 7). The two multi-point data sets were normalized by taking the maximum integrated density obtained within each set as 100% relative acetylation. Recycled cytoplasmic tubulin from gastrula stage embryos showed no detectable acetylation and was taken as 0%; the total embryonic tubulin from which this was derived was estimated to contain <1% acetylated tubulin. The axoneme data average 90.3 ± 7.3% (s.d., n=14) relative acetylation while the membrane/matrix data average 91.8 ± 5.7% (s.d., n=14). The 1.5 h single data points yield a value for membrane/matrix tubulin acetylation that is lower than these averages (69.5 ± 12%, s.d., n=4) but this difference is not statistically significant. However, there does appear to be a trend of increasing acetylation in the membrane/matrix fraction during the early time points that is not apparent in the axonemal tubulin data.

Two-dimensional gel and immunoblot analysis may be used to evaluate the relative proportion of a-tubulin actually present in the acetylated form. Using ciliary axonemes from regenerated cilia, protein staining reveals one major a-tubulin isoform (>80%), accompanied by a closely migrating, less-acidic minor isoform (Fig. 8A). Autoradiography of immunoblots from cilia labeled during regeneration indicates that this minor isoform is disproportionately more labeled (Fig. 8B), consistent with it being the more recently synthesized form. Comparative cross-reactivity with a general a-tubulin antibody (DM1A) and the acetyl a- tubulin antibody 6-11B-1 demonstrates that this minor isoform is non-acetylated (Fig. 8C vs. 8D). As would be expected from the quantitative immunoblotting results, the same isoform ratio results are obtained for membrane/matrix tubulin, although with somewhat poorer resolution (not shown).

A prominent characteristic of regenerating sea urchin embryonic cilia is the relative abundance of tubulin and other membrane/matrix proteins early in the growth process. On a per cilium basis, however, the total amount of protein in the membrane/matrix fraction remains relatively constant during regeneration, while the amount of tubulin (and associated proteins) in the axoneme necesssarily increases coincident with length. Since the membrane/matrix also increases in proportion to axonemal length, its protein concentration must decline and the ciliary membrane must grow in area mainly by incorporation of lipids and not by the addition of more protein.

Previous results indicate that ciliary‘membrane’ tubulin exists as a complex with lipids and other proteins, in both molluscan gill (Stephens, 1983, 1985) and sea urchin embryonic cilia (Stephens, 1991). It has been suggested that a lipid-tubulin complex could be involved in the delivery of both lipids and tubulin to the growing cilium (Stephens, 1990a), and protein synthetic studies of multiple regenerations likewise suggest a precursor-product relationship (Stephens, 1991). Operationally, one can define at least three distinct tubulin compartments: the soluble cytoplasmic pool, the ciliary membrane/matrix fraction, and the assembled 9+2 axoneme. A potential fourth compartment of soluble periaxonemal tubulin is probably relatively small, since most tubulin solubilized by detergent extraction is reconstituted into membranes (Stephens, 1991).

The axoneme to membrane tubulin ratio and the level of detyrosination of both the membrane/matrix and axonemal a-tubulin, after remaining constant for about 2 hours, closely correlate with ciliary length during regeneration, although any functional correlation remains obscure. Detyrosination is generally considered to be a post-assembly event (Gunderson et al., 1987, 1989) and, while this appears also to be the case with axonemal tubulin, the fact that the tubulin in the readily solubilized membrane/matrix fraction is as detyrosinated as axonemal tubulin suggests that the former is processed as if it were in an assembled form. These results are consistent with earlier observations that neuronal membrane-bound tubulin is detyrosinated relative to the cytoplasmic form (Nath and Flavin, 1978) and also with later conclusions that tyrosinated and detyrosinated tubulin coexist in the neuronal membrane-associated tubulin fraction (Nath and Flavin, 1980).

The results presented here for the acetylation of α tubulin in regenerating sea urchin embryonic cilia confirm and extend the original observations of L’Hernault and Rosenbaum (1983) with regard to Chlamydomonas flagellar assembly. Although the 2-dimensional analyses in both studies are virtually identical with respect to the relative proportion of major and minor axonemal a-tubulin isoforms, the present antibody results would indicate a much more rapid or complete acetylation of membrane/matrix α-tubulin in sea urchin cilia than in Chlamydomonas flagella where, in contrast to sea urchin cilia, the non-acetylated isoform dominates the membrane/matrix fraction in assembled flagella. The possibility remains, however, that a small amount of soluble periaxonemal tubulin in sea urchin cilia, representing the fourth compartment noted above, is primarily non-acetylated, particularly at very early stages of regeneration.

Acetylation does not correlate with growth but, rather, appears to be a characteristic of tubulin localized within the cilium, whether in the membrane/matrix or the axoneme. Tubulin in the cytoplasm, from which such tubulin must be derived, shows no detectable acetylation. To some extent, the same holds true for detyrosination, where cytoplasmic tubulin is >97% tyrosinated while the membrane/matrix and axonemal tubulins are each about 50% detyrosinated when ciliary assembly is complete. Acetylation and tyrosination, therefore, do not distinguish membrane/matrix tubulin from axonemal tubulin in the growing cilium, although both post-translational modifications distinguish ciliary tubulin from the bulk tubulin of the embryonic cytoplasm. Consistent with compartmentalization of modifying enzymes within the ciliary lumen, a recent study has shown that both acetylation and an undefined post-translational modification that is characteristic of assembled cilia do not occur when the cilia are experimentally forced to assemble within the cytoplasm (Adoutte et al., 1991). In addition, Bowser and coworkers (1991) present evidence for a free diffusion barrier between the primary cihum and cytoplasm of tissue culture cells, which is able to exclude materials with M_r values in excess of 10,000.

These observations lead to a speculative model for partitioning tubulin between cytoplasmic and ciliary compartments. Tubulin contained in the membrane/matrix fraction may be determined or defined by acetylation. This fraction remains relatively constant during elongation and it may thus provide a constant store of tubulin destined for the axoneme. Detyrosination may take place in this membrane/matrix compartment prior to the incorporation of tubulin into the growing axoneme, only to be replenished with tubulin freshly acetylated from the cytoplasm. It may be significant that both the initiation of detyrosination and the marked increase in the axoneme to membrane tubulin ratio occur at about the same 2 h time point. Detyrosination may proceed to a point where an equilibrium is established between the membrane/matrix and axonemal tubulin compartments, resulting in the characteristic gradual cessation of growth.

One seeming complication with this model is that newly synthesized axonemal proteins, apart from tubulin, do not appear to any measurable extent in the membrane/matrix fraction, either at maturity (Stephens, 1991) or during assembly (Fig. 2B, above). This would restrict the model to the delivery of tubulin alone, via a membrane pathway, and would suggest that most other axonemal components co-assemble with tubulin without first passing through the membrane/matrix compartment, perhaps transported via the axoneme itself or else present in the periaxonemal matrix but at too low a level to be detected. Since the quantal synthesis of the integral outer doublet protein, tektin-A, limits the length of the axoneme (Stephens, 1989), this compartmentation model could provide an additional point for kinetic control by regulating tubulin availability.

This work was supported by USPHS grant 20,644 from the National Institute of General Medical Sciences. I am indebted to Drs. Jeannette Chloe Bulinski and Gianni Pipemo for their generous gifts of antibodies, to Gwen Prior Szent-Gyorgyi and Robert Crowther for expert technical assistance, and to Dr. Steven L’Hemault for encouraging me to resurrect and reprobe our earlier 2-D analyses in terms of acetylated tubulin isoforms.