Expression Patterns and Levels of All Tubulin Isotypes Analyzed in GFP Knock-In C. elegans Strains

Most organisms have multiple α- and β-tubulin isotypes that likely contribute to the diversity of microtubule (MT) functions. To understand the functional differences of tubulin isotypes in Caenorhabditis elegans, which has nine α-tubulin isotypes and six β-tubulin isotypes, we systematically constructed null mutants and GFP-fusion strains for all tubulin isotypes with the CRISPR/Cas9 system and analyzed their expression patterns and levels in adult hermaphrodites. Four isotypes—α-tubulins TBA-1 and TBA-2 and β-tubulins TBB-1 and TBB-2—were expressed in virtually all tissues, with a distinct tissue-specific spectrum. Other isotypes were expressed in specific tissues or cell types at significantly lower levels than the broadly expressed isotypes. Four isotypes (TBA-5, TBA-6, TBA-9, and TBB-4) were expressed in different subsets of ciliated sensory neurons, and TBB-4 was inefficiently incorporated into mitotic spindle MTs. Taken together, we propose that MTs in C. elegans are mainly composed of four broadly expressed tubulin isotypes and that incorporation of a small amount of tissue-specific isotypes may contribute to tissue-specific MT properties. These newly constructed strains will be useful for further elucidating the distinct roles of tubulin isotypes.

TBA-9 was detected in ciliated sensory neurons in the head (CEPV and CEPD) and in the ventral motor neuron PDE, as described previously (Hurd et al., 2010). TBA-9 was expressed in an additional 13 neuronal cells in the head region whose identity was unable to be determined due to their low GFP signals (Fig. 3B) but might include the ones reported in Hurd et al. (2010), i.e., ADF, AFD, ASE, ASI, AWA, AWC, and ADE. The average expression of TBA-9 was ~5% of that of TBA-1 in the cell bodies (Fig. 2F). TBB-4 was expressed in many ciliated sensory neurons, consistent with a previous report (Hurd et al., 2010). TBB-4 accumulated mainly in the cell body, and weak signals were detected in axons and cilia (Fig. 3B). In cell bodies, the average expression of TBB-4 was ~8% of that of TBB-2 (Fig. 2F).
Signal quantification in the cross-section that included the dendritic region (Fig. 2A,(a)) indicated that the levels of TBA-5, TBA-6, and TBA-9 were ~5% of those of the broadly expressed α-tubulin isotypes (TBA-1 and TBA-2) (Fig. 2C), and the level of TBB-4 was ~10% of those of the broadly expressed β-tubulin isotype TBB-2 (Fig. 2D). Within the ciliated neurons, TBA-5 accumulated in the region of the cilia, and TBA-9 was enriched in both cilia and cell bodies. TBA-6 was slightly enriched in the cilia and was detectable in cell bodies (Fig. 3C).
To determine at which stage these ciliated neuronspecific α-tubulin isotypes (TBA-5, TBA-6, and TBA-9) and β-tubulin isotype TBB-4 initiate co-expression, we examined their expression during embryogenesis. TBB-4 was detectable in the precursor cells of ciliated sensory neurons around the dorsal enclosure stage (~300 min after fertilization) (Fig. 4B). At this stage, the broadly expressed isotype TBB-2 was assembled into spindle MTs in mitotic cells, but TBB-4 was not detected in the spindle MTs and instead was diffusely present in the cytoplasm (Fig. 4A,B), suggesting that TBB-4 was not efficiently incorporated into spindle MTs. TBB-4 was strongly expressed in the cell body of ciliated neuron precursors in 1.5-fold embryos (~450 min after fertilization). In contrast, ciliated neuronspecific α-tubulin isotypes (TBA-5, TBA-6, and TBA-9) became detectable during the 2-fold stage (~500 min after fertilization), and their accumulation in the tips of neurons, which correspond to the region of the cilia, became prominent at the 3-fold stage (~550 min after fertilization) (Fig. 4C).
Based on these expression patterns, we speculate that TBB-4 may contribute to the general MT properties of ciliated sensory neurons, whereas expression of additional tubulins TBA-5, TBA-6, and TBA-9 may confer some MT features specific to the subtype of ciliated neurons.
In the gfp::mec-7 strain, we detected the MEC-7 signal in these six mechanosensory neurons as reported (Hamelin et al., 1992;Mitani et al., 1993), as well as in an additional 16 neurons (Fig. 5A,B). In all detectable cells, the expression of MEC-12 and of MEC-7 in cell bodies was comparable and was ~6 % of that of the corresponding broadly expressed -tubulin and -tubulin isotype, respectively (Fig. 2F).
BEN-1 was expressed in various types of neuronal cells, including mechanosensory neurons (Figs. 1 and 5C). The average expression level of BEN-1 in cell bodies was three to four times higher than that of the neuron-specific β-tubulin isotypes TBB-4 and MEC-7 but was ~20% of that of the broadly expressed β-tubulin isotype TBB-2 (Fig. 2F).
Thus, in neuronal cells, the broadly expressed isotypes (TBA-1, TBA-2, TBB-1, and TBB-2) are expressed at a high level, and neuronal-specific isotypes are additionally expressed at a low level and may confer cell-type-specific MT features.

Expression of other tubulin isotypes in adult hermaphrodites
The isotypes TBA-4, TBA-7, TBA-8, and TBB-6 were expressed in distinct tissues. TBA-4 was expressed in a wide range of tissues, including the intestine and epidermis, and a small number of neurons but not in the germline (Figs. 1B and 6A). The general expression of TBA-4 was much lower than that of the broadly expressed isotypes TBA-1 and TBA-2 (~10% in section (a), (b), and (c) in Fig. 2). TBA-7 was expressed in the intestine, intestinal-rectal valve, rectal gland cells, and excretory pore cells (Fig. 6B), consistent with previous reporter gene assays (Hurd, 2018). Expression of TBA-7 was also significantly lower than TBA-1 and TBA-2 (~3% (a), ~2% (b), and ~2% (c) in Fig.   2). TBB-6 was expressed in unidentified cells in the head (Figs. 1 and 6C) and the expression level was low (~4% (a), ~5% (b), and ~3% (c) of TBB-2 in Fig. 2D). TBA-8 was undetectable in the adult stage, but in second larval stage (L2) and forth larval stage (L4) larvae, two linear signals along the anteroposterior axis were observed that correspond to seam cells in the lateral epithelium, which is consistent with a previous reporter assay (Portman and Emmons, 2004) (Figs. 1 and 6D). The expression level of TBA-8 was also significantly lower than that of TBA-1 and TBA-2 (~4% (a), ~2% (b), and ~6% (c) in Fig. 2).

Discussion
To our knowledge, this is the first report to systematically analyze the expression levels of all tubulin isotypes in a single metazoan organism. We found that the broadly expressed tubulin isotypes were expressed at significantly higher levels than tissue-specific isotypes, suggesting that the incorporation of even a small amount of tissue-specific tubulin isotypes into the MTs, which comprise mainly the broadly expressed isotypes, may confer the specific MT features of that particular tissue (Fig. 7).
Although the expression patterns of tubulin isotypes detected in these strains were generally consistent with previous studies that involved extrachromosomal arrays, GFP::TBA-6 in our study was not detected in HSN motor neurons as previously reported (Hurd et al., 2010). This discrepancy could have several explanations. First, transgenes used in previous studies might contain incomplete promoter regions that are not sufficient to reproduce endogenous expression patterns. Second, some endogenous-level GFP signals in the knock-in strains may be too weak to detect, whereas GFP signals in the multicopy-transgenic strains (e.g., Hurd et al., 2010, Hao et al., 2011 might be expressed at a higher, and thus detectable, level. qPCR analyses have also been used to determine transcript levels of tubulin isotypes in C. elegans PLM neurons (Lockhead et al., 2016). In that study, transcripts of the mechanosensory neuronspecific α-tubulin isotype MEC-12 and β-tubulin isotype MEC-7 were detected at higher levels than those of the broadly expressed isotypes TBA-1 and TBB-2, whereas we found that GFPtagged endogenous TBA-1 and TBB-2 were detected at higher levels than those of MEC-7 and MEC-12. This discrepancy might reflect differences in translational efficiency of the mRNAs of each tubulin isotype. Post-transcriptional regulation of tubulin isotype genes needs to be explored in the future.
Our collection of GFP knock-in strains also revealed the preference and efficiency with which each tubulin isotype was incorporated into different types of MTs (e.g., spindle MTs, ciliary MTs, or axonal MTs). It is unclear whether the presence of some isotypes in neuronal cell bodies was due to an excess of tubulin proteins that were not incorporated into MTs in axons or dendrites or was due to their being required in the cell bodies.
The C terminus of tubulins represents the most divergent region of these proteins and is subjected to PTMs, which can modulate MT properties. For example, detyrosination of tubulins affects mechanotransduction in skeletal and heart muscles in mice, and this disruption causes muscular dystrophy (Kerr et al., 2015). Knockdown of the tubulin glycine ligase TTLL-3 results in shortened cilia in zebrafish and Tetrahymena thermophila (Wloga et al., 2009). However, how isotype-specific PTMs affect MT properties in vivo is not well understood. Both ciliated neuronspecific α-tubulin isotype TBA-6 and the relative levels of tubulin glutamylase TTLL-11 and deglutamylase CCPP-1 are crucial for the structure and function of cilia in C. elegans neurons, although TBA-6 does not have polyglutamylation sites (Silva et al., 2017, O'Hagan et al., 2017. Thus, other tubulin isotypes expressed in the ciliary neurons, possibly TBA-1, TBA-2, TBB-1, TBB-2, and/or TBB-4, might be regulated by polyglutamylation. How PTMs in each isotype contribute to the properties of tissue-specific MTs is an important topic to be analyzed. In other organisms, some tubulin isotypes are not replaceable by other isotypes. In Drosophila melanogaster, the somatic β-tubulin isotype β3 does not complement the function of the testis-specific β-tubulin isotype β2 (Fackenthal et al., 1993;Hoyle and Raff, 1990). In mice, platelets require βtubulin isotype β1 and α-tubulin isotype α4 for the assembly of a MT structure called the marginal band that maintains platelet structure (Schwer et al., 2001;Strassel et al., 2019). Our analysis demonstrated that TBA-5, TBA-6, and TBA-9 are expressed in distinct subsets of ciliated neurons, and these isotypes affect ciliary structures (Hao et al., 2011;O'Hagan et al., 2017). It will be of interest to determine whether the loss-of-function phenotype of these isotypes in C. elegans can be rescued by other ciliary neuronspecific isotypes, which will help us to further understand the functional specificity of tubulin isotypes.
The combination of tubulin isotypes and PTMs is proposed to generate "tubulin codes," which can be read out by the interaction between MTs and MAPs including motor proteins (Gadadhar et al., 2017). In vitro, MT dynamics and stability are modulated according to the composition of different human β-tubulin isotypes (Ti et al., 2018;Vemu et al., 2017). Thus, the combination of tubulin isotypes can fine-tune MT functions via isotype-specific PTMs and interactions with MAPs and motors (Sirajuddin et al., 2014). Our analysis of the tissue-specific composition of tubulin isotypes will be a starting point for decoding tubulin codes in vivo, and this comprehensive collection of GFPknock-in strains will be useful for further studies.

Worm strains and maintenance
Bristol N2 strain was used as wild type. Strains used and constructed in this study are listed in Table SI. All worms were grown in standard nematode growth medium (NGM), fed OP50, and kept at 20°C or 24.5°C as indicated (Table SI) (Brenner, 1974).

Worm strain construction
The gfp::tba-1, gfp::tba-2, gfp::tbb-1, and gfp::tbb-2 strains were constructed in our previous study (Honda et al., 2017). The other strains were constructed using the method developed by Dickinson et al. (2015). For the loss-of-function strains, the GFP coding sequence and self-excision cassette (SEC) were inserted just before the start codon of each tubulin coding region; these strains show the Rol phenotype because of the mutated sqt-1 gene in the SEC. For the GFP-fusion strains, SEC was excised by heat-shock so that the GFP coding region became adjacent to the tubulin coding region (the resulting strains become non-Rol).
Each repair template fragment was cloned into the pDD282 vector (Addgene #66823) by Gibson assembly method (Gibson et al., 2009). Primers used in this study are listed in Table SII. As left homology arms, 500-to 800-base pair (bp) DNA fragments upstream of the start codon of each tubulin gene were PCR-amplified from N2 genomic DNA or fosmids. As right homology arms, 500to 3000-bp DNA fragments downstream of each stop codon were amplified. To prevent the repair template from being cleaved by Cas9, silent mutations were incorporated in the single-guide RNA (sgRNA) binding site or protospacer adjacent motif (PAM) sequence. All Cas9 target sites were chosen using the online design tool CRISPRdirect (http://crispr.dbcls.jp/) (Naito et al., 2015).
All constructed alleles were confirmed by PCR and sequencing of the corresponding genomic regions.

Microscopy
For live imaging of the GFP signals in the whole body of adult hermaphrodites, worms were immobilized with 0.5% phenoxypropanol on 2% agarose pads. Images were taken with a CSU-X1 spinning-disk confocal system (Yokogawa Electric, Musashino, Japan) mounted on an IX71 inverted microscope (Olympus, Tokyo, Japan) with a UPlanSApo 60×/1.30 NA silicone objective lens (Olympus) under the control of MetaMorph software (Molecular Devices, Sunnyvale, CA). All images were taken by an Orca-R2 12-bit/16-bit cooled CCD camera (Hamamatsu Photonics, Hamamatsu, Japan). Images were acquired for 40 z-sections with 1-µm steps at every field of view by using a 2000ms exposure time with camera gain set to 0 and without binning. Images were processed and analyzed by ImageJ/Fiji software (National Institutes of Health, Bethesda, MD). For Fig. 1, z-Sectioned image stacks were projected using the Max intensity algorithm, connected to generate the image of the whole body with the MosaicJ plug-in, and then corrected to create the image of a straightened worm with the Straighten plug-in in ImageJ. Some images in Fig. 1 were enhanced according to the signal intensity using the Brightness/Contrast function in ImageJ.
Live imaging of embryos was performed as described (Toya et al., 2010). In brief, hermaphrodite adults were dissected transversely with an injection needle, and embryos were collected in egg buffer (NaCl 94.4 mM, KCl 32 mM, MgCl 2 2.72 mM, CaCl 2 2.72 mM, HEPES [pH 7.4] 4 mM) (Edgar, 1995) on a cover glass. Embryos were mounted on 2% agarose pads with egg buffer and sealed with Vaseline to fill the gap between the cover glass and glass slide. GFP and mCherry images were acquired every minute at each of 21 z-sections with 1-µm steps and a 500-ms exposure time with camera gain of 0 and without binning by confocal microscopy as described above.

Quantification of expression levels of each tubulin isotype
Expression levels of GFP-tagged tubulin isotypes in Fig. 2

Immunostaining
Embryos were collected from hermaphrodite adults of N2 and of the GFP-expressing mutant strains as described above and were incubated in egg buffer for 68 hours until they reached the 1.5to 3-fold stage. These embryos were transferred to poly-L-lysinecoated slides and were fixed by conventional freeze-cracking and methanol-acetone treatment (-20°C methanol for 10 min, -20°C acetone for 5 min) (Albertson, 1984). The samples were rehydrated by passing the slide through an acetone series (90%, 70%, 50%, and 30%), followed by transfer into PBS + 0.5% (w/v) Tween 20 (PBST). After incubation for 1 hour at 4°C in a humid chamber with PBST containing 0.  Scale bar: 100 µm. All images were arranged such that the head is at the left and the ventral side is at the bottom, except for the images of tba-9 and mec-12, which were acquired from the ventral side.
Images without enhancement are shown in Fig. S1.