Tubulin perturbation leads to unexpected cell wall modifications and affects stomatal behaviour in Populus

Highlight Growth-compatible, post-translational modifications of tubulin in transgenic Populus alter cell wall pectin-xylan networks and guard cell movement.


Introduction
In plants, cortical microtubules (MTs) play essential roles in cell expansion, cell wall biogenesis, and stomatal behaviour via their influence over cellulose microfibril (MF) deposition (Ledbetter and Porter, 1963;Wasteneys, 2002;Paredez et al., 2006;Eisinger et al., 2012). It is now established that this classic function depends on MT-mediated membrane trafficking of the cellulose synthase complexes (Crowell et al., 2009;Gutierrez et al., 2009). MTs are also intimately linked to organelle trafficking (Crowell et al., 2009;Hamada et al., 2012), and have been recently implicated in intracellular movement of cell wall matrix components to the cell cortex (Kong et al., 2015;Zhu et al., 2015). Consistent with this role, MTs aggregate near sites of pectin mucilage secretion in the seed coat of Arabidopsis (McFarlane et al., 2008). Disruption of the MT-MF alignment by cobtorin in tobacco BY-2 cells altered pectin distribution, which implicated not only pectins in MT-MF alignment, but also MTs in pectin deposition (Yoneda et al., 2010). Similarly, depolymerization of MTs by oryzalin affected not only cellulose but also pectin deposition in Penium margaritaceum (Domozych et al., 2014).
Genetic manipulation of MTs for functional investigation is challenging, owing not only to the multiplicity of MT functions, but also to the complex transcriptional and post-translational regulation of the tubulin subunits. MT function depends on the polymerization and depolymerization of α-tubulin (TUA) and β-tubulin (TUB) heterodimers in response to developmental and environmental cues (Ehrhardt and Shaw, 2006). TUA and TUB are encoded by multi-gene families in all sequenced plant genomes, with the woody perennial Populus harbouring eight TUA and 20 TUB genes (Oakley et al., 2007). Biased expansion of the Populus TUB family coincides with greatly reduced transcript levels of individual genes when compared to the TUA family, implicating a complex TUA:TUB transcriptional regulation (Oakley et al., 2007). Interestingly, overexpression of TUB, but not TUA, led to rapid loss of viability in yeast (Saccharomyces cerevisiae), suggesting that TUB in excess of TUA is lethal (Burke et al., 1989;Weinstein and Solomon, 1990). In tobacco, co-transformation of TUA and TUB was deemed necessary in order to obtain viable transformants (Anthony et al., 1999). Despite the potential challenges, it has been possible to recover Arabidopsis with defective tubulin synthesis. However, many such plants display abnormal organ development, cell wall synthesis, and growth (Burk et al., 2006;Ishida et al., 2007), making it difficult to discern direct and indirect effects of MT perturbations.
In animals, tubulins undergo extensive post-translational modifications (PTMs) that affect MT dynamics (Janke and Bulinski, 2011). One such modification involves the evolutionarily conserved C-terminal Tyr of TUA, which can be enzymatically removed or re-attached via the tyrosination-detyrosination cycle (Janke and Bulinski, 2011). The detyrosinated (dY) TUA may be irreversibly converted to non-tyrosinatable (Δ2 or dEY) TUA by removal of the penultimate Glu (Paturle-Lafanechere et al., 1991). These PTM isoforms exhibit spatiotemporal distribution and are enriched in long-lived stable MTs during mammalian development (Gundersen et al., 1984;Webster et al., 1987;Peris et al., 2009). In contrast, tubulin tyrosination-detyrosination is poorly understood in plants. Putative dY and dEY TUA isoforms have been reported in plants using animal tubulin PTM-specific antibodies (Duckett and Lloyd, 1994;Smertenko et al., 1997;Wang et al., 2004), but evidence by direct proteomics analysis is lacking.
The current investigation exploited tubulin PTMs to perturb MT regulation and dynamics in Populus tremula x alba 717-1B4. Expression of dY or dEY PTM mimics of TUA1 was found to facilitate recovery of transgenic Populus that was otherwise unsuccessful by co-transformation with native TUA and TUB genes. This permitted an analysis of multiple MT-dependent processes in transgenic Populus. Perturbation of tubulin PTMs was found to affect cell wall pectin-xylan organization, alter expression of cell wall-modifying enzymes, and attenuate transcriptional response to gravitational stress in wood-forming tissues. In addition, stomatal responses to environmental cues were affected in transgenic leaves. These findings that non-cellulosic carbohydrate organization and guard cell behaviour are sensitive to tubulin PTMs support the emerging role of MTs in matrix polysaccharide network assembly.

Generation of transgenic poplars and quantitative RT-PCR
Coding sequences of TUA1 (or its 3′-truncated PTM mimics), TUA5, TUB9, or TUB15 were PCR-amplified from P. tremula × alba 717-1B4 xylem cDNA, cloned into pCR2.0-TOPO (Invitrogen), and sequence-confirmed. The Xba I-Sma I TUA fragments were subcloned into binary vector pCambia1302 at Spe I and Pml I, with the hygromycin selectable marker. TUB genes were similarly cloned into a modified pCambia2301 (with its Eco RI-Pml I fragment replaced with that of pCambia1302), with the kanamycin selectable marker. After mobilization into Agrobacterium tumefaciens C58/pMP90, TUA and TUB constructs were mixed in equal ratio for co-transformation into P. tremula × alba 717-1B4 (Meilan and Ma, 2006). Putative transformants were selected on media containing both kanamycin and hygromycin, and PCR-confirmed. Plant propagation and greenhouse maintenance were as described (Frost et al., 2012). Leaf plastochron index (LPI)-15 and developing xylem from internodes 60-80 were snap-frozen in liquid N for RNA extraction. Internodes 50-60 were debarked and air-dried for wood chemistry analysis. Total RNA extraction for real-time qRT-PCR analysis was performed as described (Oakley et al., 2007). A vector-specific primer was used in conjunction with TUA/TUB-specific primers to differentiate between transgenes and their endogenous counterparts (see Supplementary Table S1 for primers).

Histology
Paraffin stem sections were prepared as described (Tsai and Harding, 2013), stained with 0.05% (w/v) toluidine blue and imaged with a Zeiss Axioskop-50 microscope.

Tubulin proteomics and western blot analysis
Tubulin was purified from developing xylem (10 g) by DEAE-Sephadex chromatography as described (Smertenko et al., 1997), with a modified buffer-to-tissue ratio of two. Tubulin-enriched fractions were dialyzed, concentrated using a Nanosep centrifugal column (MWCO 10K, Pall Life Sciences), and an aliquot (~5 µg) separated by 10% SDS-PAGE. Bands corresponding to tubulins (~50 kDa) were excised and subjected to in-gel cyanogen bromide and trypsin digestion for proteomics analysis by LC-matrix-assisted laser desorption/ionization (MALDI) and MALDI time-of-flight tandem (TOF-TOF) mass spectrometry as described (Xiao et al., 2009) at the Proteomics core, Albert Einstein College of Medicine. Relative abundance of TUA1 and its PTM variants was estimated using the label-free method . For western blot analysis, crude proteins were extracted from leaves and xylem as described (Deng et al., 2007). Five micrograms of proteins were resolved by 10% SDS-PAGE and electro-transferred onto Immobilon-FL PVDF membrane (EMD Millipore). Replicate blots were hybridized with polyclonal antibodies raised in rabbits against affinity-purified recombinant TUA1 at 1:5000 dilution (anti-TUA1, Open Biosystems) or synthetic C-terminal peptides of TUA1 at 1:10 000 dilution (anti-dY: ESPDGEDGDEGDE or anti-dEY: ESPDGEDGDEGD, Sigma Genosys). Hybridization signals were detected using IRDye 680RD-conjugated goat anti-rabbit IgG secondary antibodies and an Odyssey infrared imaging system (Licor).

Wood chemical and physical properties, and tension wood induction
Stem wood was Wiley-milled (40-mesh), Soxhlet-extracted, and airdried. Lignin content was determined by the Klason method and an aliquot of the acid hydrolysate was used for structural carbohydrate analysis by anion exchange HPLC (Porth et al., 2013). Syringyl-toguaiacyl (S/G) lignin ratio was determined by pyrolysis GC-MS (del Rıó et al., 2001). Intact stem segments (~2 cm) were used for wood density, microfibril angle (MFA) and cellulose crystallinity analyses (Porth et al., 2013). A separate cohort of vegetatively propagated plants (~1 m height) was inclined at a 30° angle to induce tension wood (TW). After 3 weeks, the stem was debarked and developing xylem from the TW side of the basal half was collected for RNA extraction, while the distal half was air-dried and split lengthwise to obtain TW-enriched wood for chemical analysis as above. Stems from control, erect trees were processed identically. Intact stem segments (~1 cm in diameter) were used for X-ray micro computed tomography (μCT) analysis of wood density as described (Joshi et al., 2011).

Glycome profiling and glycosyl composition analysis of cell wall extracts
Sequential cell wall extractions and glycome profiling of extractivefree wood meal were performed as described (DeMartini et al., 2011;Pattathil et al., 2012). Plant glycan-directed monoclonal antibodies (mAbs) were from CarboSource Services (CCRC, JIM and MAC series) at the Complex Carbohydrate Research Center, or BioSupplies (Australia) (BG1, LAMP). The data were filtered to remove weakly reacting mAbs (hybridization intensities <0.1 in all samples). The signal intensities from 127 mAbs were subjected to Z-score transformation (Cheadle et al., 2003) and self-organizing map clustering using MeV v4.5 (Saeed et al., 2003) and the bubble neighbourhood method (radius = 0.8). Plant samples within each group [wild type (WT), A1dYB9 or A1dEYB15+A1dEY] were pooled for statistical testing of transgenic effects using Limma (Smyth, 2005). The sequential cell wall extracts (200-500 μg per sample) were used for glycosyl composition analysis by GC-MS as described (York et al., 1986;Merkle and Poppe, 1994) at the Complex Carbohydrate Research Center's Analytical Services, University of Georgia.

RNA-Seq
Total RNA was extracted using the Direct-zol kit (Zymo Research) with Plant RNA Reagent (Life Technologies). Illumina TruSeq RNA library preparation and HiSeq-2000 sequencing were performed at the Genomics Core, University of Texas Southwestern Medical Center. Eight to eighteen million paired-end 100-bp reads were generated per sample for four plant lines (WT,A1dYB9,, each with two to eight biological replicates. After filtering to remove organellar contaminants, sequences were mapped to the Populus genome v3.0 (Phytozome) and processed using the Tuxedo tools (Trapnell et al., 2012). Differential expression was assessed for 9244 genes (FPKM ≥5 in all samples) to gauge transgenic or TW effects using Limma (Smyth, 2005) with multiple-testing corrections by SLIM (Wang et al., 2011). The RNA-Seq data are available under NCBI Sequence Read Archive accession number SRP042117.

Drought treatments and stomatal conductance
Vegetatively propagated plants of 1-1.5 m in height were used. Stomatal conductance was monitored on the same plants before and during the onset of drought response, or on plants maintained under different watering regimes. For drought treatment, water was withheld until the first visible sign of wilt in young leaves (typically 16-20 h), at which point gas exchange, stomatal conductance, and transpiration rates were measured on LPI-15 using a Licor LI-6400XS as described (Frost et al., 2012). Stomatal responses to light were measured at dawn. Data were analysed by paired two-sample t test or repeated measures ANOVA followed by the Student-Newman-Keuls post-hoc test using SigmaPlot 12 (Systat Software Inc.).

Tubulin manipulation in transgenic Populus was facilitated by PTM mimics
Two xylem-abundant TUA (TUA1, TUA5) and TUB (TUB9 and TUB15) genes (Oakley et al., 2007) were selected for co-transformation into Populus in eight construct combinations: four as native gene pairs (referred to as A1B9, A1B15, A5B9, and A5B15) and four pairs with dY/dEY PTM mimics of TUA1 in place of native TUA1 (denoted as A1dYB9, A1dYB15, A1dEYB9, A1dEYB15). (Note, TUA5 carries a C-terminal methionine and is not predicted to participate in the tyrosination-detyrosination cycle (Oakley et al., 2007)). The 35S promoter was used in all cases. Putative transgenic calli were obtained from all construct combinations, but the transformation efficiency was very low compared to the vector control in multiple trials. All of the callus lines derived from the native TUA-TUB construct pairs showed abnormal vascular development during organogenesis and eventually died (Supplementary Table S2, Supplementary Fig. S1). In contrast, the four pairs that contained PTM mimics had higher transformation efficiencies at the callus stage, and three produced viable transgenic plants (Supplemental Table S2). It appears that the TUA and TUB co-transformation strategy was not as effective in Populus as in tobacco (Anthony et al., 1999) for regeneration of viable plants, and that the PTM mimic-containing tubulin gene pairs facilitated whole plant regeneration in Populus.
PCR analysis of genomic DNA confirmed the presence of the TUA and TUB transgenes in the transgenic lines, except for the presumed A1dEYB15-5 and A1dEYB15-17 lines which lacked the TUB15 and NPTII PCR amplicons and were renamed A1dEY-5 and A1dEY-17 ( Supplementary  Fig. S2). The first cohort of transformants, including A1dYB9 (one line), A1dEYB15 (two lines), and A1dEY (one line) plants, was fully characterized. Endogenous transcript abundance was not changed in any of the transgenic lines ( Fig. 1). The transcript abundance of TUA1dY or TUA1dEY in mature leaves (LPI-15) was 6-17-fold higher than endogenous TUA1 in all transgenic lines (Fig. 1A). In contrast, expression of the TUB transgenes was low, and sometimes below endogene levels ( Fig. 1A). In developing xylem where endogenous tubulin transcripts were abundant, the transgenes were detected at very low levels in all viable transgenic lines we obtained. This strongly suggested that high levels of tubulin transgene expression were not tolerated in xylem during organogenesis. As in leaves, the relative transcript abundance of TUB transgenes was lower than that of TUA transgenes (Fig. 1B). The results suggest that the TUA and TUB transgenes were, much like their endogenous counterparts, differentially regulated in a gene family-and tissue-dependent manner.
The tissue-biased transgene expression was also detected at the protein level by western blot analysis. The total TUA (anti-TUA1) signal increased by as much as two-fold in transgenic leaves relative to the WT (Fig. 1C), but remained similar in xylem across genotypes (Fig. 1D). Anti-dY and anti-dEY signals were essentially absent in WT, but were readily detected in the respective transgenics. Specifically, anti-dEY signal was as strong as that of total TUA in leaves of dEYB15 and dEY plants, but accounted for only 17-27% of the total TUA signal in xylem of these plants (Fig. 1C,D). Proteomics analysis confirmed that the PTM mimics TUA1dY and/or TUA1dEY were present in transgenic xylem extracts (Fig. 2). Consistent with the western blot results, levels of the PTM isoforms were negligible in WT, but increased considerably in the transgenics, accounting for 3-6% of the TUA1 levels (Fig. 2D). In both western and proteomics analyses, the dY and dEY isoforms were only detected in the A1dYB9 and A1dEYB15-11 lines, respectively. This suggests that the enzymatic dY-to-dEY conversion reported in animal systems (Janke and Bulinski, 2011) may be absent in Populus. While the abundance estimate varied between detection platforms and cannot be directly compared, the results nevertheless provide protein-level evidence for active synthesis of TUA1 PTM mimics in the transgenics.

Wood physicochemical properties were not quantitatively changed but pectin and xylan extractability was altered in the transgenics
Young stem cross sections of WT and transgenic plants were anatomically similar ( Supplementary Fig. S3). Wood density, MFA, and cellulose crystallinity of the woody stems did not show transgenic effects (Table 1). Cell wall carbohydrate and lignin contents also did not differ between genotypes; however, the S/G monolignol ratio was lower in the transgenics ( Table 1). The results suggested that tubulin perturbation in this system led to subtle compositional modifications, but not quantitative changes in major constituents of the cell wall.
To further investigate possible cell wall modifications in the transgenics, extractive-free wood meal was subjected to sequential fractionation of cell wall polysaccharides for glycome profiling (Pattathil et al., 2010;Pattathil et al., 2012). In general, mild solvent extractions by oxalate (fraction I) and carbonate (II) preferentially release arabinogalactans (AGs) and pectic components. Subsequent alkaline extractions with 1 M (III) and 4 M KOH (IV) remove hemicellulosic components (xylans and xyloglucans). Chlorite extraction (V) degrades lignin and associated carbohydrates, with the post-chlorite 4 M KOH treatment (VI) releasing additional tightly bound polysaccharides. Total sugars of each cell wall fraction estimated by the phenol-sulfuric acid method differed little between genotypes ( Supplementary Fig. S4A). Self-organizing map clustering of ELISA data from 127 cell wall glycan-directed mAbs revealed six major epitope clusters across the cell wall fractions and genotypes (Fig. 3, Supplementary Table S3). Cluster 1 consisted mainly of xyloglucan epitopes, with similar extractability between genotypes (Fig. 3). The other clusters were dominated by pectin-(clusters 2-5) and/or xylan-derived (clusters 5 and 6) epitopes, and showed a general trend of increased extractability in the transgenics (Fig. 3). Statistical analysis confirmed that the vast majority of epitopes with significant changes (P ≤0.05 Data are mean ± SD of n = 3-7 biological replicates, except for A1dEY-5 (mean ± range of n = 2) and A1dEYB15-12 (n = 1). (C, D) Western blot analysis of TUA levels in leaves (C) and xylem (D) using polyclonal antibodies raised against recombinant TUA1 (anti-TUA1) or synthetic C-terminal peptides of TUA1 (anti-dY and anti-dEY). Relative (%) signal abundance normalized to marker bands was estimated within the anti-TUA1 blot for genotypic comparison, or across the three blots (anti-TUA1, anti-dY, and anti-dEY) for isoform comparison. and fold-change ≥1.5) in both transgenic groups (39 out of 44) were derived from pectins (Supplementary Table S3).
The five pectin/xylan clusters revealed changes in the epitope extractability across cell wall fractions due to tubulin perturbation. Cluster 2 consisted of more easily-extracted pectic epitopes, such as AGs [classified under mAb groups AG-1 and AG-3 (Pattathil et al., 2010)], arabinans (AG-2), and methyl-esterified homogalacturonans (HG-2). They were abundant in fractions I and II and showed modest transgenic effects (Fig. 3). In contrast, pectic epitopes in cluster 3 [AG-3 and rhamnogalacturonan I (RG-I/AG and AG-4)] and cluster 4 (RG-I backbone, RG-I/AG, arabinan and galactan) were present in all cell wall fractions and showed large transgenic effects (Fig. 3, Supplementary Table S3). Their extractability was most strongly enhanced in the transgenics after removal of lignin (fractions V and VI). The small cluster 5 contained both pectic (RG-I) and xylan (Xylan-3) epitopes with increased extractability in fractions II-V (Fig. 3). Cluster 6 included mostly xylan (Xylan-6 and Xylan-7) and a few pectic (RG-I/AG and de-esterified HG-1) epitopes that were abundant in all cell wall extracts, especially the most tightly bound fraction VI. Their extractability was more strongly affected in the A1dEYB15/A1dEY plants than the A1dYB9 line.
while GalA was enriched in fraction V (Supplementary Fig.  S4C-F). Xyl and Rha were recovered at significantly higher levels in fractions III, IV, and VI of A1dEYB15/A1dEY plants than in WT ( Supplementary Fig. S4G,H,J). Lignin removal by chlorite (fraction V) released proportionately more non-Xyl monosaccharides, with GalA, Ara, Rha, and Gal higher in A1dEYB15/A1dEY than in WT plants ( Supplementary Fig. S4I). The response of A1dYB9 was weaker. These results are interpreted to suggest that pectins and xylans deposited early during cell wall biogenesis are tightly associated with lignin, and most sensitive to tubulin perturbation. In contrast, subsequently deposited pectins that were more readily extracted by mild solvents were not as significantly affected in the transgenics. Thus, while cellulose and hemicelluloses were not quantitatively changed by the subtle tubulin manipulation reported here, the pectin-xylan polysaccharide organization, and their interactions with lignin polymers, appear to be modified in the transgenics.

Elevated expression of cell wall-modifying enzymes and attenuated TW response in transgenic xylem
Vegetatively propagated plants were subjected to TW induction, a treatment known to increase both MT and MF abundance (Pilate et al., 2004). X-ray µCT imaging analysis confirmed TW formation on the upper side of the inclined stem (Fig. 4A), which exhibited significantly higher density than the normal wood (NW; Fig. 4B). The expected lignocellulosic compositional shifts in TW were observed (Foston et al., 2011), including significantly increased levels of glucose, arabinose, rhamnose, and galactose, and decreased xylose, mannose, and lignin ( Fig. 4C-I), but there were no genotypic differences in either NW or TW. In agreement with the wood trait response, RNA-Seq analysis revealed large TW effects, but small genotypic differences in either NW or TW (Fig. 5). The small number of genes significantly up-regulated by tubulin perturbation overlapped substantially with those induced by TW ( Fig. 5A-C), and, in effect, their response to TW was attenuated in the transgenics compared to WT (Fig. 5B,D). Many of these genes encode cell wall-modifying enzymes, such as pectin lyases (e.g. Biswal et al., 2014), pectin acetylesterase (Gou et al., 2012), cellulase (Lewis et al., 2013), β-xylosidases (Goujon et al., 2003), and α-expansins (Gray-Mitsumune et al., 2008) (Fig. 5D, Supplementary Table S4). In fact, overexpression of pectin lyase Potri.003G175900 in transgenic Populus has been shown to alter cell wall pectin and xylan extractability without affecting the wood composition, similar to what was observed in this study. The RNA-Seq data thus provide molecular support for cell wall remodelling in the transgenics due to tubulin perturbation.

Altered stomatal behaviour of transgenic leaves
Next, whether another MT-dependent process-the dynamic opening and closing of stomata-might also be  affected in the transgenic plants was investigated. Putative changes in MT stability of the guard cells due to tubulin perturbation would be expected to alter stomatal conductance in response to external stimuli. Drought stress is known to induce stomatal closure, as shown by the significant decrease of stomatal conductance in WT (Fig. 6). However, transgenic plants showed delayed stomatal closure in response to drought (Fig. 6A, Supplementary  Fig. S5). This slowed response led to the prediction that stomatal opening would also be delayed in the transgenics. Indeed, the rate of stomatal opening in response to light was slower in the transgenics than in the controls, whereas the steady-state stomatal conductance remained similar between genotypes under dark-acclimated or lightsaturated conditions (Fig. 6B). These results imply that tubulin PTM perturbation in transgenic poplar altered some MT-dependent aspect of the stomatal response to environmental cues.

Tubulin manipulation in transgenic Populus was facilitated by PTM mimics
In this study, transgenic Populus have been obtained with perturbed tubulin homeostasis. The overall transformation efficiency was very low, consistent with developmental defects and/or lethality frequently associated with tubulin manipulation (Weinstein and Solomon, 1990;Abe et al., 2004). While TUA-TUB co-transformation to obtain balanced expression was deemed essential for viable transgenic tobacco (Anthony et al., 1999), that strategy did not produce viable transgenic Populus in the present study. Instead, expressing a PTM mimic (dY or dEY) of TUA1, with or without TUB, was more effective for recovery of Populus transformants. Interestingly, the transcript levels of the TUA transgenes (TUA1dY or TUA1dEY) were several-fold higher than those of TUB transgenes in all transgenic lines examined, despite the use of a constitutive CaMV 35S promoter in all cases. This is reminiscent of previous findings that endogenous TUA transcripts are more abundant than TUB transcripts across a range of Populus tissues (Oakley et al., 2007). The results suggest that both endogenous and introduced TUA and TUB genes are subject to post-transcriptional or co-translational regulation in Populus, presumably via auto-regulation or auto-feedback regulation as reported in animal systems, to maintain a proper ratio of the two subunits (Pachter et al., Heatmaps of response ratios (TW/NW or transgenic/WT) of cell wall remodelling genes from the intersections of C. Significant differences (P ≤ 0.05) are underlined. Data were from n = 3 for WT, n = 8 for A1dYB9, n = 5 for A1dEY(B15), or n = 2-3 for transgenic lines A1dEYB15-11 and A1dEY-5. AGP, arabinogalactan protein; FC, fold-change; PMEI, pectin methylesterase inhibitor; XTH, xyloglucan endotransglucosylase/hydrolase. 1987; Gonzalez-Garay and Cabral, 1996). The inherent TUAbiased transcript abundance in Populus also suggests that high levels of TUB transcripts, either endogenous or transgenic, may be detrimental for development, as reported in yeast (Weinstein and Solomon, 1990). This may explain the limited success of the tubulin co-transformation strategy in Populus.
Besides the TUA bias, expression of tubulin transgenes also exhibited a tissue-level bias, where higher transcript abundance was observed in the leaves than in xylem, opposite to the endogenous transcripts (Fig. 1). Although the CaMV 35S promoter exhibits slightly lower activity in xylem than leaves of Populus (Eriksson et al., 2000;Babst et al., 2014), it has been successfully used to overexpress genes in Populus xylem (Zhong et al., 2011;Lin et al., 2013). In support of its efficacy in xylem, transcripts of 35S-driven hygromycin phosphotransferase marker gene were detected at high levels (average FPKM 1011), on par with the xylem-abundant TUB15 (FPKM 1023) and several lignin biosynthetic pathway genes in A1dYB9 (NW) xylem, based on RNA-Seq analysis (Supplementary Table S4). Thus, the unusually low levels of TUA and especially TUB transgene expression in xylem of all viable transgenic lines strongly suggests that higher levels might constrain recovery of developmentally normal Populus transformants. Further supporting this argument was the observation that most of the callus lines failed to advance during shoot regeneration, characteristically exhibiting defective vasculatures (Fig. S1). Together, these results underscore the complexity of tubulin regulation at multiple levels, which likely contributed to the difficulties in obtaining transgenic alteration of tubulin expression.

Tubulin perturbation affected pectin-xylan networks in transgenic Populus wood
A subtle perturbation of the tubulin subunit pool was achieved by overexpressing PTM mimics of TUA in Populus. The perturbations were found to affect cell wall properties, but not in a manner predicted by the classic function of MTs in directing cellulose MF deposition. Results showed that pectin-xylan polysaccharide extractability and lignin S/G ratio, but not cellulose characteristics (content, MFA, or crystallinity), were sensitive to tubulin manipulation in the xylem. The absence of a clear effect on cellulose in the viable plants might be due to weak transgene expression in xylem as discussed above. It is plausible that strong transgenic effects might have compromised structural integrity as reported previously (Burk et al., 2006;Ishida et al., 2007), and contributed to regeneration failure in the present study. Altered matrix polysaccharide extractability could signify changes in structure, composition, and/or linkage. Because vascular structural integrity was maintained and cell wall carbohydrate levels did not change quantitatively in the transgenics, the differential extractability of cell wall glycans can be best explained by qualitative changes that alter the pectin-xylan networks.
Pectins are abundant in the primary cell wall and middle lamella, and their deposition precedes that of cellulose, hemicelluloses, and lignin (Keegstra, 2010). At the onset of secondary cell wall formation, the primary cell wall is pushed outwards into the middle lamella and cell corners between adjoining cells (Jarvis et al., 2003) where lignin (typically G lignin) deposition first occurs (Boerjan et al., 2003;Tobimatsu et al., 2013). The preferential increase of G lignin, and the enhanced extractability of multiple pectic (arabinan, AG, RG-I, and de-esterified HG) epitopes in the lignin-bound cell wall fractions of transgenic wood are both consistent with a close association between early-deposited pectin and lignin polymers in the middle lamella and cell corners. This association likely involved xylan, because several xylan epitopes also exhibited increased extractability in the more tightly bound cell wall extracts of the transgenics. In support of this view, a recent study showed that the Arabidopsis irregular xylem 8 (irx8) mutant deficient in xylan and HG owing to a mutation in galacturonosyltransferase is compromised in lignin biosynthesis (Hao et al., 2014). Specifically, defective matrix polysaccharide deposition led to reduced accrual and altered extractability of G lignin in the irx8 mutant (Hao et al., 2014). These and observations from the current study strongly support a close association of pectin, xylan, and lignin during early stage of cell wall biogenesis.
Pectins and other matrix polysaccharides are synthesized in the Golgi and transported to the plasma membrane for deposition into the cell wall (Keegstra, 2010), a process that implicates MT-mediated vesicle trafficking (Crowell et al., 2009;Hamada et al., 2012). A role for MTs in pectin deposition has been inferred based on immunofluorescence microscopy and pharmacological studies. Pectin vesicles were observed in close association with cortical MTs near the mucilage secretion pockets during Arabidopsis seed coat development (McFarlane et al., 2008). Oryzalin treatment was shown to affect not only cellulose but also pectin deposition in unicellular green alga (Domozych et al., 2014). Two recent studies showed that Arabidopsis mutants defective in kinesin motor proteins exhibited abnormal Golgi morphology, with altered pectin secretions and accumulation of arabinose-containing carbohydrates in the cell wall (Kong et al., 2015;Zhu et al., 2015). It is worth noting that tubulin C-termini are exposed on the outer surface of MTs where their interactions with a host of MT-associated proteins (MAPs) modulate MT dynamics (Janke and Bulinski, 2011). It is conceivable that tubulin PTM perturbation interfered with MT-mediated vesicle trafficking and affected pectin polysaccharide deposition and its integration with xylan and lignin polymers. The involvement of spatiotemporally regulated MAPs (Sedbrook and Kaloriti, 2008;Gardiner, 2013) may underscore the conditional nature of the phenotype affecting mostly early cell wall deposits.

Stomatal behaviour was sensitive to tubulin perturbation in transgenic Populus
Tubulin perturbation delayed drought-induced stomatal closure as well as light-induced stomatal opening without changing steady-state stomatal behaviour in transgenic Populus. A dynamic relationship between MTs and stomatal function has been unequivocally demonstrated using immunofluorescence microscopy or live cell imaging coupled with pharmacological treatments (Fukuda et al., 1998;Eisinger et al., 2012). The MT-depolymerizing drug oryzalin prevents stomatal opening, whereas the MT-stabilizing drug taxol impedes stomatal closure (Fukuda et al., 1998;Eisinger et al., 2012). Thus, the observed delay in stomatal closure in drought-stressed leaves may be attributed to more stable MT arrays due to ectopic expression of tubulin PTM mimics. This is consistent with the frequent association of dY-and dEY-TUA isoforms with stable MTs in animal systems (Peris et al., 2009;Janke and Kneussel, 2010). On the other hand, lightinduced stomata opening requires polymerization of tubulins (Eisinger et al., 2012). The delayed response observed in transgenic Populus raises the possibility that MT polymerization was slowed owing to altered tubulin pool composition. In support of this idea, dY-TUA has been shown to assemble more slowly than the tyrosinated counterpart in taxolpromoted in vitro polymerization (Banerjee and Kasmala, 1998). In addition, the delayed stomatal responses may hint at altered guard cell wall properties as a result of tubulin perturbation. Guard cell wall chemistry is not well understood, but studies based on immunolocalization and Fourier transform infrared spectroscopy all point to pectins as the major component imparting strength and flexibility necessary for dynamic change of cell shape during stomatal opening and closure (Majewska-Sawka et al., 2002;Jones et al., 2003;Jones et al., 2005). Based on this, and on the glycome profiling results showing that the pectin-xylem networks were sensitive to tubulin perturbation, one may speculate that the transgenic effect on pectins also occurred in guard cells to impact stomatal movement.

Empirical evidence against an active tubulin tyrosination-detyrosination cycle in Populus
The transgenic phenotypes of plants expressing either TUA1dY or TUA1dEY PTM mimics were similar, differing only in degree. The comparatively strong phenotypes of A1dEYB15/A1dEY relative to A1dYB9 plants were consistent with higher TUA1dEY than TUA1dY signals at both RNA and protein levels in the respective lines (Figs. 1 and 2). The possibility was considered that the phenotypic variation was due to differing capacities of the two PTM mimics to participate in the tyrosination-detyrosination cycle (i.e. TUA1dY could be retyrosinated back to TUA1 or converted to TUA1dEY, but TUA1dEY could not be modified further), but several lines of evidence argue against this. First, despite the wealth of knowledge about tubulin tyrosination/ detyrosination in animal systems, support for such a cycle in plants has only been inferred from antibody-based detection (Smertenko et al., 1997;Wang et al., 2004). In fact, the dY-and dEY-TUA1 isoforms that might be indicative of a tyrosination-detyrosination cycle were essentially undetected in WT xylem by western blot and proteomics analyses in the present study. Second, no plant orthologues of the mammalian tubulin tyrosine ligase (Ersfeld et al., 1993) involved in dY-to-Y retyrosination have been identified (Oakley et al., 2007). The mouse cytosolic carboxypeptidase 1 reported to catalyse the irreversible dY-to-dEY conversion (Berezniuk et al., 2012) also lacks any apparent orthologue in plants (Phytozome v10.2), and indeed, no evidence of dYto-dEY conversion in the A1dYB9 line was found. Together, these data seem to suggest that the tyrosination-detyrosination cycle, evolutionarily conserved from animals to parasitic protozoa (Xiao et al., 2009), is absent in plants. How then did the dY/dEY PTM mimics of TUA elicit an effect in transgenic Populus? Tubulin C-terminal PTMs have been shown to affect MT assembly or interactions with MAPs (Konishi and Setou, 2009;Peris et al., 2009;Sirajuddin et al., 2014). Given that TUA1dY/dEY isoforms are normally absent in Populus, their ectopic expression in the transgenics may represent 'foreign' tubulin codes that can interfere with MAP interactions. Interestingly, among the suite of annotated plant MAPs (Hamada et al., 2013), only two putative Populus orthologues showed significantly altered (reduced) expression in all transgenic plants (Supplementary Table S4). The two genes (Potri.006G018000 and Potri.016G006900) are genome duplicates and orthologous to Arabidopsis AtMAP70-5, which is known to regulate secondary cell wall polymer organization (Pesquet et al., 2010). Although further research is needed to elucidate the interplay between MAP70-5 and cell wall polysaccharide organization in Populus, the data support the idea that tubulin PTM perturbation can modulate MAPs to affect MT-dependent processes.
A remaining question is whether the observed phenotypes were due to tubulin PTM perturbation specifically, or to altered tubulin abundance/composition in general, because transgenic lines expressing native tubulin genes were not obtained in the present investigation. Given the inherently disproportionate TUA and TUB transcript abundance in Populus, manipulating single tubulin expression should be feasible to address this question.

Supplementary data
Supplementary data are available at JXB online. Supplementary Fig. S1. Abnormal vascular development was observed in nonviable transgenic lines. Supplementary Fig. S2. PCR analysis of wild-type and transgenic plants using gene-specific primers. Supplementary Fig. S3. Stem cross sections of wild-type and transgenic plants. Supplementary Fig. S4. Total sugars and glycosyl composition of sequential cell wall extracts from wild-type and transgenic plants. Supplementary Fig. S5. Photosynthetic responses of wildtype and transgenic leaves from replicate drought experiments. Supplementary