Bulk samples of developing xylem were scraped into liquid nitrogen from the debarked trunks of 5-year-old, field-grown P. deltoides trees. TW xylem was obtained from the upper side of the trunk leaned at a 30–40° angle from the vertical axis for 4 weeks. Snap-frozen samples were stored at -80°C until use.

Approximately 5 g of xylem tissue was ground into a fine powder in liquid nitrogen for tubulin purification using a modified DEAE-Sephadex chromatography method (Morejohn and Fosket, 1982; Smertenko et al., 1997). The tissue powder was suspended in 10 ml of PEM buffer (50 mM PIPES, pH 6.9; 0.5 mM MgCl₂; 1 mM EGTA and 1 mM DTT) with protease inhibitors (1 mM benzamidine HCl; 2 mM leupeptin; 15 mM pepstatin A; 1 mM phenylmethylsulfonyl fluoride; 1 mM sodium fluoride and 50 μM N-tosyl-L-phenylalanine chloromethyl ketone) and 2 mM GTP, and vortexed vigorously. The mixture was first clarified at 50,000 g for 10 min, and the supernatant ultracentrifuged at 100,000 g for 45 min, both at 2°C. The resulting supernatant was mixed with 0.5 volumes of PEM-equilibrated DEAE-Sephadex A50 containing 0.5 mM GTP and incubated at 4°C for 1 h with gentle agitation. The mixture was loaded into a polyprep chromatography column (0.8 cm × 4 cm, BioRad), and washed with 3–5 volumes of 0.4 M KCl in PEM buffer containing 0.1 mM GTP. The bound tubulin proteins were then eluted with 0.8 M KCl in PEM buffer with 0.1 mM GTP. The protein-rich fractions were pooled and dialyzed against 1 L of 10 mM NH₄HCO₃ at 4°C overnight with one buffer change. The protein was concentrated using a Nanosep centrifugal column (MWCO 10K, Amicon) and the concentration was estimated using Bradford reagents (BioRad). Recombinant TUA1 in pET30a (Oakley et al., 2007) expressed in E. coli strain BL21 (DE3) was purified from inclusion body using BugBuster protein extraction reagent (Novagen), and quantified with BCA protein assay kit (Novagen).

Seventy-five nanograms of purified tubulins were resolved on a 10% SDS-PAGE gel, and transferred onto an Immobilon-FL polyvinylidenedifluoride membrane (EMD Millipore). The membranes were incubated with the Blocking Buffer for Fluorescent Western Blotting (Rockland Immunochemicals) in phosphate-buffered saline, and then with primary and secondary antibodies in the same blocker containing 0.1% Tween 20. The following primary antibodies were used. Polyclonal antibodies (Open Biosystems) raised against recombinant TUA1 in rabbits (1:5000) and recombinant TUB15 in chickens (1:5000) recognize TUA and TUB proteins, respectively. Polyclonal rabbit antibodies raised against synthetic C-terminal peptides of TUA1 (anti-dY, ESPDGEDGDEGDE at 1:1000; and anti-dEY, ESPDGEDGDEGD at 1:1000; Sigma Genosys) recognize detyrosinated and non-tyrosinated TUA isoforms as described previously (Swamy et al., 2015). Mouse-derived monoclonal antibodies for polyglutamylated (clone B3, 1:1500) and acetylated (clone 6-11B-1, 1:500) tubulins were obtained from Sigma. Hybridization signals were detected using IRDye 680RD-conjugated goat anti-rabbit IgG, 800CW-conjugated goat anti-mouse or 800CW-conjugated donkey anti-chicken secondary antibodies (1:15000, Li-cor) with an Odyssey infrared imaging system (Li-cor).

The SDS-PAGE bands containing tubulins were excised, destained and subjected to in-gel trypsin digestion as described (Shevchenko et al., 1996) at the proteomics core of Michigan State University. Peptides were fractionated by reverse phase HPLC using a Waters nanoAcquity UPLC. Eluted peptides were sprayed into a ThermoFisher LTQ Linear Ion trap mass spectrometer outfitted with a MICHROM Bioresources ADVANCE nano-spray source. The top five ions in each survey scan are then subjected to data-dependent zoom scans followed by low energy collision induced dissociation (CID) and the resulting MS/MS spectra are converted to peak lists in BioWorks Browser v3.2 using the default LTQ instrument parameters. Peak lists were searched against all Populus tubulin sequences using the Mascot searching algorithm v2.1. The Mascot output was then analyzed using Scaffold, v1.7.0 to probabilistically validate protein identifications.

An independent set of purified tubulin samples was subjected to SDS-PAGE and in-gel CNBr digestion according to Xiao et al. (2010) at the Proteomics core of the Albert Einstein College of Medicine. Peptides were fractionated by reverse phase HPLC (Ultimate 3000, Dionex), followed by MALDI-TOF analysis (ABI 4800, Applied Biosystems) in both positive and negative ion mode as detailed in (Xiao et al., 2010). Populus tubulin sequences were used as the database. Data were processed with Data Explorer software (v4.9). Label-free quantitation of tubulin isotypes was performed according to Miller et al. (2012).

The total RNA was extracted with the Direct-zol RNA Kit (Zymo Research) using Plant RNA Reagent (Life Technologies) and quantified with the Qubit RNA HS Assay Kit on a Qubit fluorometer (Life Technologies). RNA-Seq libraries were prepared using the Illumina TruSeq Stranded RNA LT Kit and sequenced on an Illumina NextSeq 500 at the Georgia Genomics Facility of University of Georgia. The data are available at NCBI Sequence Read Archive under accession number SRP076604. After filtering to remove rRNA sequences, data were processed by a local assembly pipeline (Gu and Tsai, unpublished) modified from Allen et al. (2015) for de novo assembly of TUA and TUB sequences. Briefly, TUA and TUB transcript sequences from P. trichocarpa were used as reference for read mapping by Bowtie 2, v2.2.3 (Langmead and Salzberg, 2012). Matched reads were pooled and subjected to de novo assembly by Trinity (Haas et al., 2013). The assembled contigs were Blasted against the reference, and relevant sequences were retained as new baits to repeat the process until the output was stable, usually within 10 iterations depending on transcript abundance. Following manual curation, the longest contig for each gene was retained for further analysis. This procedure was also performed for a previously published P. tremula ×alba xylem RNA-Seq dataset (SRP042117; Swamy et al., 2015). The translated peptide sequences were used to calculate the theoretical m/z for peptide mapping. For expression analysis, a region of up to 200 bp covering the hypervariable C-terminus and a portion of the 3′-UTR was retrieved for each gene from available transcript assemblies of P. deltoides and P. tremula ×alba above, guided by multiple sequence alignment. The P. trichocarpa sequences were used for those that were absent in our assemblies (not expressed). Reads were mapped to the reference sequences using Blat and filtered by >95% hit length coverage with a 2% mismatch allowance. Reads were only assigned to the best hit. Transcript abundance was estimated by fragments per million total reads per kilobase (FPKM).

Tubulin proteins were purified from developing xylem of P. deltoides using a modified DEAE-Sephadex anion exchange chromatography protocol (Swamy et al., 2015), originally developed for plant cell suspension cultures (Morejohn and Fosket, 1982). Based on SDS-PAGE and Coomassie Brilliant Blue staining, the purified proteins were enriched in a doublet of ∼50 kDa, close to the expected size for tubulins (Figure 1A). Western blotting confirmed that the lower band was TUA, and the upper one was TUB (Figure 1B). The yield of tubulins was 40–60 μg/g fresh-weight of developing xylem. The excised tubulin bands were subjected to in-gel cyanogen bromide (CNBr) digestion for proteomics analysis by MS.

A label-free method (Miller et al., 2012) was applied to determine the relative abundance of tubulin isotypes in NW and TW samples of P. deltoides using isotype-specific C-terminal reporter peptides (Table 1). RNA-Seq data from the same series of samples were used for de novo assembly of (expressed) P. deltoides tubulin transcripts in order to correct for sequence polymorphisms that deviate from the P. trichocarpa reference gene models (see “Materials and Methods” and Table 1). TUA1 and TUA4/5 were the most abundant isotypes in both NW and TW (Figure 2A). The predicted m/z for the TUA4 and TUA5 reporter peptides is indistinguishable by MS (Table 1). However, as TUA4 transcripts were barely detected in xylem based on RNA-Seq analysis (Figure 2B), the TUA4/5 reporter peptide signal probably represents (and hence was assigned to) TUA5 (Figure 2A). The TUB signals were much lower than the TUA signals, consistent with the immunoblot results shown in Figure 1. TUB15 and TUB 16 were the predominant TUB isotypes in P. deltoides xylem. The MS abundances of TUA and TUB isotypes were largely similar between NW and TW samples of P. deltoides (Figure 2A). We also analyzed the proteomics dataset of wild-type (WT) P. tremula ×alba xylem (NW) described in Swamy et al. (2015), with the same RNA-Seq-based sequence curation as above (Figure 2C). The TUA and TUB isotype profiles were generally consistent between species, with the exception of TUB9 which was among the top three most abundant C-terminal TUB peptides in P. tremula ×alba xylem (Figure 2C), but was undetected in P. deltoides (Figure 2A).

We searched the P. deltoides MS spectra of C-terminal polypeptides for evidence of tubulin detyrosination, non-tyrosination or glutamylation. The tyrosinated (unmodified) C-terminal peptide (residues 414–451) of TUA1 was detected with an m/z of 4180.5, very near the theoretical monoisotopic mass (MH+) of 4180.69 (Figures 3A,B). Its authenticity was previously confirmed by MALDI-TOF/TOF analysis using enriched tubulins purified from NW xylem of WT P. tremula ×alba (Swamy et al., 2015) (Figure 3C). In that same study, we also identified and MALDI-TOF/TOF-confirmed C-terminal peptides of detyrosinated (dY, m/z 4017.4) and non-tyrosinatable (dEY, m/z 3888.6) isoforms from transgenic P. tremula ×alba ectopically expressing the dY and dEY PTM mimics of TUA1, respectively (Swamy et al., 2015) (Figures 3I,N). However, the MS signal for either PTM peptide was near background in P. deltoides NW and TW samples or in WT P. tremula ×alba (Figure 3F–H,K–M) (Swamy et al., 2015). This was corroborated by Western blot analysis of purified xylem tubulins from both Populus species using PTM-specific antibodies raised against the dY (anti-dY) or dEY (anti-dEY) C-terminal peptide of TUA1. As reported previously (Swamy et al., 2015), the anti-dY and anti-dEY immunosignals were clearly detected with xylem tubulins purified from dY- and dEY-expressing transgenics, respectively (Figures 4A,B). The faint signals observed in the other samples likely represent background hybridization as seen for the recombinant TUA1 protein purified from Escherichia coli that was included as a negative control (Figures 4A,B). The results suggest that detyrosination and non-tyrosination of TUA were either absent or negligible in Populus xylem.

Polyglutamylation of Populus tubulins was examined by immunoblotting using a commercial monoclonal antibody raised against polyglutamylated TUAs of sea urchin (Lytechinus pictus) (Gagnon et al., 1996). A weak signal that overlapped with the anti-TUA signal was detected for all samples, including recombinant TUA1, indicative of background hybridization likely due to the TUA origin of the monoclonal antibody (Figures 4C,D). We then searched the MALDI mass spectra for peptide masses that were 129 Da (the mass of a Glu residue) larger than the C-terminal peptide masses of TUAs and TUBs in both Populus species. No such signals were found above background levels, suggesting that tubulin glutamylation is unlikely to occur in wood-forming tissues of Populus.

Acetylation of TUAs was studied using a commercial monoclonal antibody raised against acetylated TUA of sea urchin. As with the other immunoblot analyses described above, only background hybridization signals were detected regardless of tissue (NW or TW) or genotype (Figures 5A,B). A shot-gun LC-MS/MS analysis of trypsin-digested tubulin identified an N-terminal peptide shared by TUA1 and TUA5 that appeared to be acetylated at Lys-40 due to a mass shift of 42 Da (Figure 5C). Although this peptide was identified with a 94% probability, the acetylated Lys-40 signals (b5 and b6 ions) were very low. Unmodified Lys-40 is susceptible to trypsin cleavage, and accordingly, we identified a shorter N-terminal TUA1/TUA5 peptide with an exposed Lys-40 (Figure 5D). As TUA1 and TUA5 are abundant in Populus xylem, we interpret the immunoblot and MS data to suggest that both isotypes were subjected to very low levels of acetylation at the conserved Lys-40 residue.

The RNA-Seq data were also processed to estimate tubulin transcript abundance in xylem of both P. deltoides and P. tremula ×alba. Because TUA and TUB genes share high degrees of sequence identity, we extracted up to 200 bp sequences spanning the hypervariable C-termini and 3′-UTRs from the de novo transcript assemblies as references for read mapping (see “Materials and Methods” and Supplementary Data 1). This was intended to minimize ambiguous read alignment in the highly conserved coding region. The overall TUA and TUB transcript profiles were similar between P. deltoides and P. tremula ×alba, suggesting a conserved transcriptional regulation in this genus (Figures 2B,D). The transcript abundance estimates by RNA-Seq were largely congruent with the isotype profiles observed in both species (Figure 2), suggesting transcriptional regulation. Exceptions to this general trend were two TUBs with discordant patterns. Low MS signals were detected for TUB18 despite its very high transcript levels in both species, and TUB17 exhibited a similar though less pronounced pattern. The data suggested involvement of post-transcriptional or translational regulation. As mentioned above, the TUB9 reporter peptide was detected in hybrid aspen P. tremula ×alba, but not P. deltoides (Figures 2A,C). Accordingly, TUB9 transcripts were present at relatively higher levels in P. tremula ×alba (75th percentile of all TUBs) than in P. deltoides (50–55th percentile) xylem (Figures 2B,D). The results hint at a taxon-specific fine tuning of TUB regulation in Populus.

The proteomics data also help curate gene model annotation of the reference genome. We previously cloned all eight TUA cDNAs from P. tremuloides (Oakley et al., 2007). The experimentally identified TUA5 mRNA (GenBank accession EF583814) differed from the primary gene model (Potri.009G085100.1) predicted in the current P. trichocarpa genome v3 (Phytozome v11), but matched the secondary gene model (Potri.009G085100.2). The two models represent splice variants, with the primary gene model predicted to harbor an extra, 31-bp intron upstream of the stop codon that is not found in other Class I TUAs (Oakley et al., 2007). Proteomics data supported the presence of Potri.009G085100.2 in both P. deltoides (NW and TW) and P. tremula ×alba (NW) xylem, and no signal matching the predicted Potri.009G085100.1 C-terminal peptide was detected. Moreover, de novo assembly of xylem RNA-Seq data from both P. deltoides (this study) and P. tremula ×alba (Swamy et al., 2015) using a custom local assembly pipeline (Gu and Tsai, unpublished) recovered only the Potri.009G085100.2 transcript (Supplementary Data 2). Taken together, both transcriptomics and proteomics data suggest that Potri.009G085100.2 is the predominant TUA5 isotype in Populus, and the splice variant Potri.009G085100.1 is likely an annotation artifact.