Novel Arabidopsis microtubule-associated proteins track growing microtubule plus ends

MT-binding regions in GPT1/2

Although GPT1/2 do not exhibit any significant amino acid homology with functionally characterized proteins [21], we noticed that the middle (M) and C-terminal (C) regions of these proteins are enriched in the basic amino acid residues Arg, Lys, and His (18.9% in GPT1 and 17.9% in GPT2; Fig. 1a and b). By contrast, the N-terminal (N) regions are abundant in acidic Asp and Glu (22.2% in both GPT1/2). Thus, GPTs are polar proteins with short, negatively charged regions at their N-termini (pI values of 4.4 for GPT1 and 4.5 for GPT2) and longer, positively charged regions in their middles and C-termini (pI values of 10.7 for GPT1 and 10.6 for GPT2).

To identify MT-binding regions in these proteins, we fused full-length and truncated versions of GPT1 and GPT2 to GFP, and transiently expressed these fusions in onion epidermal cells, together with the red fluorescent MT marker tagRFP-MAP4 (Fig. 1). Dual color visualization of GPT1-GFP and tagRFP-MAP4 proved difficult, possibly due to GPT1 having a weak MT-binding capacity, competition between GPT1 and MAP4 for overlapping MT binding sites, or both. Therefore, the MT-binding capacity of GPT1-GFP was determined based on the presence of fine filaments (presumably representing cortical MTs decorated by GPT1-GFP). Colocalization of the GFP and RFP signals confirmed that full-length GPT2 localized to cortical MTs (Fig. 1f). The N-terminal (N) regions of GPT1 (1–180) and GPT2 (1–207) did not localize to MTs, whereas the N-terminal-deleted (M + C) (181–530 for GPT1, and 208–553 for GPT2) and the middle (M) (181–300 for GPT1, and 208–328 for GPT2) fragments were clearly localized to cortical MTs (Fig. 1f). The C-terminal (C) fragments (301–530 for GPT1, and 329–553 for GPT2) were somewhat associated with MTs, but to a lesser extent than the fragments containing the M region. These results from transient expression assays indicate that the positively charged basic amino acid residues (especially in the M region) mediate the binding of GPT1/2 to MTs in vivo.

MT co-sedimentation assay

To test whether GPT1/2 directly bind to MTs, we fused full-length GPT1 and GPT2 to maltose-binding protein (MBP), expressed the recombinant proteins in bacteria, and purified the proteins by affinity purification using amylose resins. Some degradation occurred during the purification, particularly in the case of MBP-GPT1. When 1 μM of purified proteins was incubated with MTs assembled from bovine brain tubulin and pelleted by ultracentrifugation, intact MBP-GPT1 and MBP-GPT2 were identified in the microtubule pellet fraction (P) (arrowheads in Fig. 2a). The partial degradation products of MBP-GPT1 (asterisk in Fig. 2a) were also pelleted, whereas further degraded MBP-GPT1 and MBP-GPT2 fragments were not. In the absence of MTs, MBP-GPT1 and MBP-GPT2 remained in the supernatant (S). MBP alone was also evaluated to verify that it is not responsible for binding of GPT1/2 to MTs. As shown in Fig. 2, MBP remained in the supernatant fraction after ultracentrifugation. These results show that GPT1/2 bind MTs directly.

To determine the stoichiometry and affinity of GPT toward MTs, various amounts of MBP-GPT2 were centrifuged with a constant amount of MTs that were polymerized from 1 μM tubulin heterodimer, and a binding curve was obtained (Fig. 2b). MBP-GPT2 bound to MTs in a concentration-dependent manner and saturated at a stoichiometry of approximately 2.5 +/− 0.9 mol of GPT2 per mol of tubulin dimer. The dissociation constant K_d was calculated to be 1.9 +/− 1.0 μM. However, if GPT2 associates with MTs in more than one binding mode with different affinities (as indicated below), this simple regression analysis does not provide true binding values.

Subcellular localization

To determine the subcellular localization of GTP1/2 and to monitor their dynamics in vivo, GFP was fused to either the N-terminus or the C-terminus of GPT1 and expressed under the constitutive cauliflower mosaic virus 35S promoter, because the 5′-regulatory region of GPT1 was recalcitrant to cloning, whereas GFP was fused to the C-terminus of GPT2 and expressed under its native promoter. Both N-terminal and C-terminal GFP fusions of GPT1 localized to MTs in transient expression assays using onion epidermal cells (Additional file 1), indicating that the location of GFP does not affect the subcellular localization of fusion proteins. The Arabidopsis MT marker line that expresses mCherry fluorescent protein fused to β-tubulin 6 (mCherry-TUB6) was used as a transformation host.

GFP-GPT1 and GTP2-GFP co-localized with mCherry-TUB6 and decorated the nuclear envelope and preprophase bands, mitotic spindles, and phragmoplasts of the mitotic cells of the root meristem (Fig. 3). GFP-GPT1 labeling differed slightly from mCherry-TUB6 labeling, particularly in expanding phragmoplasts. This observation prompted us to carefully examine the localization and dynamics of GPT1 and GPT2 in cortical MT arrays, where the dynamics of single MTs can be clearly visualized.

GPT1 and GPT2 are novel + TIPs

Time-lapse imaging of root epidermal cells using a spinning disk confocal microscope revealed that GFP-GPT1 and GPT2-GFP not only decorated the MT lattice, but also accumulated as particles on MTs (Fig. 4 and Additional files 2, 3 and 4). Double intensity plots with MTs (mCherry-TUB6) and GFP-labeled GPT1/2 showed that GPT1/2 are localized to the MT ends in a comet-like pattern, with the highest signal intensity at the MT ends and a gradual decline in signal further from the MT ends. This MT-end localization was more pronounced for GFP-GPT1 than for GPT2-GFP. The MT ends labeled with GFP were highly dynamic, indicating that GPT1/2 both label the plus ends of cortical MTs. GFP-GPT1 was associated with polymerizing plus ends, and dissociated as soon as catastrophe occurred (Fig. 4c). However, when the MTs started to polymerize again, GFP-GPT1 was immediately recruited to the growing ends (Fig. 4d). GPT2-GFP showed a similar but weaker labeling pattern when compared with GFP-GPT1 (Additional file 2). These results demonstrate that GPT1 and GPT2 preferentially recognize the plus ends of growing MTs.

EB1 is a + TIP family member that recognizes the GTP-cap region of growing MTs [7]. To compare the labeling patterns of GPT1/2 with that of EB1, we stably co-expressed GFP-labeled GPT1 or 2 and mCherry-labeled Arabidopsis EB1b (EB1-mCherry) in Arabidopsis plants. EB1-mCherry showed strong plus-end labeling, as previously reported [11–15]. Many, but not all, EB1 particles co-localized with GFP-GPT1 (Fig. 5a; Additional file 5). When the relative fluorescent intensities of GFP and mCherry were plotted, the comet-shaped fluorescent signals of GFP-GPT1 and EB1-mCherry partially overlapped. In our confocal microscopy setup, the two emitted fluorescent signals from GFP and mCherry were sequentially detected after alternately exciting each fluorophore. When mCherry fluorescence was detected for 0.2 s, followed by the detection of GFP signal for 0.5 s, the highest GFP-GPT1 signal intensity was located approximately 0.2 μm closer to the MT end when compared with the highest EB1-mCherry signal (Fig. 5b, c). When the detection order was reversed, however, the highest EB1-mCherry signal intensity was located approximately 0.15 μm in front of the highest signal intensity of GFP-GPT1 (Fig. 5b, d). At the current spatial resolution, optical artifacts associated with our detection system make it difficult to conclusively determine whether EB1 and GPT1 completely co-localize at growing MT plus ends. In both observations, the fluorescent signal intensities of the EB1 comets decreased gradually and reached background levels at 1.5 μm behind the tips. By contrast, substantial levels of GFP-GPT1 signal (approximately 20% of the highest intensities) remained associated with the MT lattice behind the comet tails.

Tip-tracking of GPT does not require EB1 or SPR1

EB1 recognizes a structural feature of the MT GTP cap and directly binds to the plus ends of growing MTs both in vitro and in vivo [7]. Many animal + TIPs do not bind directly to the MT plus ends in vivo, but are recruited to this region by EB1. To test whether GPT accumulates at the MT plus end directly or through a hitchhiking mechanism involving interaction with EB1, we analyzed the tip-tracking behavior of GPT in the Arabidopsis eb1 null mutant. Because the Arabidopsis genome includes three EB1 genes that may function redundantly [14], a triple eb1 knockout mutant was used to study the expression of GFP-GPT1 and GPT2-GFP. GPT1/2 both localized to the plus ends of growing MTs, in the same pattern as observed in the wild-type background (Fig. 6a, b; Additional files 6, 7 and 8).

SPR1 and its homologues are plant-specific MT-localized proteins that bind to both the growing plus ends and the MT lattice [17, 18]. Arabidopsis contains seven SPR1 homologues that may be redundant, and SPR1 contributes predominantly to the anisotropic growth of seedling roots [19]. Therefore, we used a spr1 null mutant to test whether SPR1 is required for the plus-end tracking behavior of GPT. Cortical MTs were labeled with mCherry-TUB6. In the spr1 mutant, GPT1/2 labeled growing MT plus ends in interphase epidermal cells, forming a comet-like pattern (Fig. 6c, d; Additional files 6, 9 and 10). The tip-tracking behavior of the GFP-labeled GPTs in the spr1 null mutant was indistinguishable from that in wild-type epidermal cells.

These results demonstrate that EB1 and SPR1 are not required for the MT plus-end tracking function of GPT1 and GPT2.

GPT1/2 bind to MTs through its basic regions

With the exception of the N-terminal regions, GPT1/2 are relatively rich in Lys, Arg, and His residues, with GPT1/2 having average pI values of 10.7 and 10.6, respectively. These positively charged residues are distinct biochemical features of GPTs; however, these values are lower than those of the highly basic domains of some MAPs, such as BPP1, BPP2, and BPP3, which have pI values of 12.9–13.4 [21]. Transient expression of GFP-tagged full-length and truncated GPT1/2 proteins in onion epidermal cells showed that the basic regions of GPT1/2 are responsible for the in vivo localization to cortical MTs. The central and C-terminal basic regions were independently targeted to MTs in vivo, suggesting that several MT-binding regions exist in GPT1/2. Because recombinant GPT1/2 proteins bind to MTs in vitro, the basic regions appear to bind MTs directly.

Various MT-binding regions have been identified in MAPs, and a high pI value resulting from basic amino acid residues was found to be characteristic of the MT-binding regions of certain MAPs [16]. The positively charged regions are thought to interact electrostatically with the highly acidic Glu- and Asp-rich C-terminal tails of α- and β-tubulins. Because an acidic C-terminus is a common feature of all eukaryotic tubulins [23], this charge-based MT-binding mode is frequently used by MAPs in various organisms.

GPT preferentially accumulates at the plus ends of growing MTs

GFP-tagged GPT not only labels the MT lattice, but is considerably enriched at the growing plus ends of cortical MTs. This plus-end localization is stronger for GFP-GPT1 than for GPT2-GFP. We do not know whether the affinity for the plus end is inherently different between GPT1/2, or whether the location of the fused GFP molecule influenced the biochemical properties of GPT1/2. Although GPT1/2 have highly similar distributions of basic amino acids and considerable levels of amino acid sequence identity, we noticed that GPT1-GFP associates less efficiently with MTs than does GPT2-GFP when a MT marker is co-expressed in onion epidermal cells. Further biochemical analyses are required to identify possible differences between MT plus-end targeting by GPT1/2.

The tip-tracking behavior of GPT1/2 and the canonical + TIP, Arabidopsis EB1, was compared by coexpressing these proteins tagged with different fluorescent markers in Arabidopsis plants. EB1 specifically recognized the growing MT plus end and dissociated almost completely at 1.5 μm from the MT tip, where there are no binding sites for EB1 [8–10]. By contrast, the GPT1/2 tip-tracking fluorescent signal decreased with increasing distance from the MT plus end, but a substantial proportion of GPT1/2 molecules remained bound to the GDP-tubulin MT lattice. This propensity of GPT1/2 to bind to the MT lattice may obstruct observations of tip-localized GPT1/2 in dense mitotic MT arrays. Although further analysis at higher resolution is necessary, our present localization data suggest that EB1 and GPT1 have distinct binding sites at growing MT plus ends. EB1 recognizes the terminal GTP-cap and an extended region of several hundreds of tubulin molecules at growing MT ends [8, 9]. Recent imaging of EB1 localization using cryo-electron tomography showed that EB1 interacts with outwardly curved and straight regions of the MT lattice that probably represent GTP- and GDP-Pi-tubulin dimers, respectively, at the MT ends [10]. EBs bind at the intersection of four tubulin dimers in two adjacent protofilaments [8, 9]. This binding site is ideally suited for sensing the nucleotide states of the surrounding tubulin dimers. To autonomously track the growing MT ends, a + TIP should recognize subtle structural changes of tubulins that are associated with the nucleotide status and the protofilament closure into a cylinder. The predicted GPT-binding sites at the acidic tubulin C-terminus would be insufficient for such a recognition mechanism, and may require other components that associate with MTs, such as the N-terminal non-basic regions.

The efficient dissociation of EB1 from the MT lattice is partly caused by the negatively charged C-terminus of EB1. When this C-terminus was replaced with a neutrally charged motif, the EB1 mutant decorated the MT lattice without affecting its interaction with the growing MT ends [24]. The biochemical properties of GPT1/2 may render these molecules less able than EB1 to discriminate between MT ends and the lattice.

EB1 is not the only + TIP that autonomously interacts with the MT ends. Xenopus laevis XMAP215 (or ch-TOG in human) recognizes MT plus ends, but does not distinguish between polymerizing and depolymerizing ends [6]. XMAP215/ch-TOG binds to the extreme MT tip, whereas EB1 localizes to the tip region several tens of nanometers behind XMAP215/ch-TOG in vivo and in vitro [25, 26]. +TIPs that recognize distinct structural features at the plus ends would be expected to decorate different sub-regions at the MT ends.

GPT does not require EB1 or SPR1 for tip tracking

EB1 recruits non-autonomous + TIPs via their C-terminal EB homology domain to growing MT ends [7]. CAP-Gly domains [27] and SxIP motifs [28] have been identified in EB1-interacting proteins that accumulate at growing MT ends. Interestingly, GPT1/2 still tracked the MT ends in the Arabidopsis eb1 triple mutant background, which lacked EB1 function. The absence of CAP-Gly domains and SxIP motifs in GPTs suggests that their subcellular localization is independent of EB1. GPT1/2 localization to MT ends also does not require SPR1, another plant-specific + TIP [17, 18]. Although it is not known whether SPR1 autonomously tracks growing MT ends, these results suggest that if GPT does indeed use a hitchhiking mechanism to bind to a growing MT tip, it must couple with a + TIP protein other than EB1 or SPR1.

Plant materials, growth conditions, and plant transformation

Arabidopsis thaliana ecotype Columbia was obtained from the ABRC stock center and used throughout the experiments. Arabidopsis seeds were sterilized with a solution of 0.1% Tween-20 and 10% sodium hypochlorite for 15 min, washed three times with sterilized water, and sown on Arabidopsis agar medium (2.5 mM KNO₃, 1.25 mM KPO₄, 1 mM Ca(NO₃)₂, 1 mM MgSO₄, 35 μM Fe-EDTA, 7 μM MnCl₂, 5 μM NaCl, 0.5 μM ZnSO₄, 0.25 μM CuSO₄, 0.1 μM NaMoO₄, 0.005 μM CoCl₂, 1.5% [w/v] agar, and 1% [w/v] sucrose). After stratification at 4 °C in darkness for 4 days, the seeds were germinated and grown vertically at 23 °C under a long photoperiod (16 h light/8 h darkness). Transgenic seedlings were identified by screening on ½ MS medium (Nihon Pharmaceutical Co.: 50 μg/mL myo-inositol, 0.2 μg/mL thiamine, 0.05% [w/v] MES-KOH pH5.7, 0.7% [w/v] agar, and 1% [w/v] sucrose) containing appropriate antibiotics. The seeds were germinated on the antibiotic-containing media and then scored after 14 days.

Arabidopsis was transformed using the floral dip method [29]. Agrobacterium tumefaciens GV3101 (pMP90) harboring the destination vector was selected on LB plates supplemented with the appropriate antibiotics. The bacteria were collected by centrifugation [29] and suspended in transformation solution containing 5% [w/v] sucrose and 0.05% [v/v] Silwet L-77. After the floral parts of the plants were dipped into the transformation solution, the plants were allowed to grow normally until the seeds were harvested.

For dual color visualization, a transgenic Arabidopsis plant expressing a GFP-tagged GPT protein was crossed with a plant harboring mCherry. To observe mitotic MT structures, mCherry-TUB6 was expressed under the control of the epidermis-specific promoter, a 3.4-kb 5′-upstream region from the putative transcription start site of AtML1 [30]. To analyze cortical MT arrays in root epidermal cells, mCherry-TUB6 was expressed under the control of an epidermis-specific promoter (i.e., the 2.0-kb fragment immediately upstream of the translation start site of WRKY72) [31]. The GFP marker in the genomic EB1b-GFP construct [15] was removed by inverse PCR, followed by ligation of mCherry before the stop codon of EB1 to generate EB1b-mCherry. Cortical MT arrays in cotyledon epidermal cells were labeled with mCherry-TUB6 expressed under the control of the UBIQUITIN 10 promoter [32].

The GPT-GFP constructs (described below) were directly transformed into the eb1a-2 eb1b-3 eb1c-2 triple mutant, which was generated in our lab [15]. Transformants were first screened for hygromycin resistance, and then for GFP fluorescence. The spr1-3 mutant, which had been screened in our lab [17], was transformed with pWRKY72::mCherry-TUB6, and subsequently crossed to the GPT-GFP expressing lines. Homozygous spr1 mutant lines expressing mCherry-TUB6 and GPT-GFP were screened for the root skewing phenotypes and for simultaneous fluorescence of mCherry and GFP.

Vector construction

For expression of GFP-tagged proteins in Arabidopsis plants, a full-length GPT1 cDNA and a genomic fragment of GPT2 that contained the 2.5 kb 5′-upstream region and the genomic region to just before the stop codon in the 8th exon, were cloned into pGWB5 and pGWB4 [33], respectively, using the Gateway Cloning System (Invitrogen). Transient expression vectors were constructed by fusing the appropriate cDNA fragment to GFP at the C-terminus of the expressed protein using the Gateway Cloning System in pUGW5 [33]. Truncated cDNA fragments were generated by PCR, and the ATG start codon was included before the N-terminus of a truncated cDNA.

Recombinant protein purification and MT co-sedimentation assay

Full-length cDNAs of GPT1/2 were cloned into the pCold-MBP vector [34] using the appropriate restriction enzyme cleavage sites. The recombinant MBP-GPT proteins were expressed in the Rosetta (DE3) strain of E. coli and harvested and purified using an amylose resin column according to the manufacturer’s instructions (New England BioLabs). Purified fusion proteins were concentrated by centrifugal filters (Amicon Ultra filters; Merck Millipore), and the buffer was changed to BRB80 (80 mM PIPES, 1 mM MgCl₂, 1 mM EGTA, pH 6.8). The final protein preparations were frozen in liquid nitrogen and stored at -80 °C.

For the MT co-sedimentation assay, thawed proteins were centrifuged at 100,000 x g for 15 min at 4 °C to remove protein aggregates. MTs were polymerized from bovine brain tubulin [35] in 1x BRB80 buffer supplemented with 1 mM GTP and 10 μM taxol (paclitaxel; Wako) at 37 °C for 35 min. Polymerized MTs were separated from unpolymerized tubulin dimers by centrifugation at 100,000 x g for 30 min at 30 °C. Precipitated MTs were suspended in the 1x BRB80 buffer supplemented with 10 μM taxol. MBP-GPT1 or MBP-GPT2 were mixed at 1 μM with MTs (equivalent to 1 μM tubulin) in 50 μL of 1x BRB80 buffer containing 10 μM taxol and incubated at room temperature for 1 h. MBP was purified as above and used as a negative control. After a 1-h incubation, the samples were centrifuged at 100,000 x g for 30 min at 30 °C. The pellet and supernatant fractions were mixed with SDS-PAGE sample buffer and analyzed by SDS-PAGE (using 10% polyacrylamide gels), followed by staining with Coomassie Brilliant Blue G-250 (Nacalai Tesque), as described previously [35]. Binding values were calculated from the equation q = (q _max x c)/(K _d + c) where q is the amount of protein bound to the tubulin dimer, c is the concentration of free MBP-GPT2 in solution, K _d is the dissociation constant, and q _max is the amount of bound MBP-GPT2 at the saturated level, by using a Prism 7 software (GraphPad Software Inc.).

Particle bombardment-mediated transient expression

Plasmid (2 μg) harboring a full-length or truncated form of GPT1-GFP or GPT2-GFP and the tagRFP-MAP4 MT marker plasmid (2 μg) [21] were mixed for 30 min with 1.5 mg of 1.0 μm gold particles (Bio-Rad) that had been suspended in 19.2% glycerol, 962 mM CaCl₂, and 1.5% spermidine, and were washed with 70% ethanol, followed by 100% ethanol. The plasmid-coated gold particles were bombarded into onion (Allium cepa) epidermal peels using the PDS-1000/He Biolistic Particle Delivery System (Bio-Rad) equipped with 1,100-p.s.i. rupture disks. After a 16-h dark incubation in a moist petri dish, the GFP-expressing cells were observed with a D-ECLIPSE C2 confocal microscope (Nikon). GFP was excited at 488 nm and observed using the ET514/30 band-pass emission filter, whereas tagRFP was excited at 543 nm and observed using the ET585/65 filter.

Visualization of MTs using confocal spinning disc microscopy

Time-lapse imaging of fluorescent protein localization was performed on an inverted microscope (ECLIPSE Ti; Nikon) with a spinning disk confocal unit (CSU-XI; Yokogawa) connected to an EM-CCD camera (iXon3 DU897; Andor). Images were acquired using a numerical aperture oil-immersion objective (Apo TIRF 60x/1.49; Nikon). Excitation of fluorophores was accomplished using lasers (Andor) at 488 nm with a 520/35 filter for GFP or at 561 nm with a 617/73 filter for mCherry. Images were acquired at 2-s intervals for interphase cells or at 30-s intervals for mitotic cells. All images were collected by sequential acquisition using single channels. For dual-color visualization, mCherry was excited and its emission collected with an exposure time of 0.2 s. GFP was subsequently excited and its emission signals collected with an exposure time of 0.5 s. In another set of experiments, GFP was visualized first, followed by mCherry. The observation time lags for laser swapping caused temporal separation between the two fluorophores. Trajectories of individual MTs were traced on images and converted into kymographs using ImageJ. See Additional file 11 for schematic illustration of the procedures.