Expression of dehydroshikimate dehydratase in poplar induces transcriptional and metabolic changes in the phenylpropanoid pathway

Abstract Modification of lignin in feedstocks via genetic engineering aims to reduce biomass recalcitrance to facilitate efficient conversion processes. These improvements can be achieved by expressing exogenous enzymes that interfere with native biosynthetic pathways responsible for the production of the lignin precursors. In planta expression of a bacterial 3-dehydroshikimate dehydratase in poplar trees reduced lignin content and altered the monomer composition, which enabled higher yields of sugars after cell wall polysaccharide hydrolysis. Understanding how plants respond to such genetic modifications at the transcriptional and metabolic levels is needed to facilitate further improvement and field deployment. In this work, we acquired fundamental knowledge on lignin-modified poplar expressing 3-dehydroshikimate dehydratase using RNA-seq and metabolomics. The data clearly demonstrate that changes in gene expression and metabolite abundance can occur in a strict spatiotemporal fashion, revealing tissue-specific responses in the xylem, phloem, or periderm. In the poplar line that exhibited the strongest reduction in lignin, we found that 3% of the transcripts had altered expression levels and ~19% of the detected metabolites had differential abundance in the xylem from older stems. The changes affected predominantly the shikimate and phenylpropanoid pathways as well as secondary cell wall metabolism, and resulted in significant accumulation of hydroxybenzoates derived from protocatechuate and salicylate.


Introduction
Lignocellulosic biomass represents an important renewable source of sugars and aromatics for the production of fuels and specialty chemicals.These sugars and aromatics are contained in polymers that constitute plant cell walls: cellulose is a polysaccharide made of glucose, while hemicelluloses are largely made of mixed sugars but are often dominated by xylose.In contrast, lignin is formed via the oxidative polymerization of 4-hydroxycinnamyl alcohols (or monolignols) that derive from the phenylpropanoid pathway.These polymers are particularly abundant in tissues that develop thick secondary cell walls, and lignin deposited in these tissues confers cohesiveness, mechanical strength, and hydrophobicity.A better understanding of the biological mechanisms that govern cell wall synthesis and regulation is needed to optimize plant biomass properties for efficient and sustainable conversion into fuels, biopolymers, and specialty chemicals (de Vries et al., 2021b).
Woody perennials are expected to represent a significant share of the dedicated biomass grown for bioenergy (Langholtz et al., 2016).Short-rotation woody crops, such as fast growing Populus, Salix, and Eucalyptus species produce large amounts of lignocellulose and can typically be harvested year-round and over several years, which ensures a stable and predictable supply of bioenergy feedstocks (Hinchee et al., 2009;Mizrachi et al., 2012).In particular, several species in the genus Populus display rapid growth and high productivity, do not require a large amount of inputs for cultivation, and can be grown on marginal lands to restore soil fertility (An et al., 2021).The availability of the genome sequence of Populus trichocarpa (black cottonwood) has facilitated the genetic analysis and improvement of several Populus species (Tuskan et al., 2006).
The efficient release of monosaccharides and simple aromatics from cell walls is an essential step for biological conversion of biomass into advanced fuels and chemicals via industrial fermentation.However, lignin impedes this process by interacting with polysaccharides and thereby protecting them from enzymatic deconstruction.Lignin itself is not easily deconstructed into simple aromatics due to its inherent complexity, but also due to molecular rearrangements occurring within the polymer during biomass pretreatment processes.These challenges have prompted the plant research community to genetically engineer plant cell walls to facilitate biomass utilization.In particular, several lignin bioengineering approaches have been designed to reduce cell wall recalcitrance in crops (Mahon and Mansfield, 2019;Liu and Eudes, 2022).One classic approach is the generation of plant lines affected in one or several genes involved in a specific step of the lignin biosynthetic pathway, resulting in plants with reduced lignin and enhanced biomass saccharification efficiency (de Vries et al., 2021a;Sulis et al., 2023).Our group has shown that heterologous expression of bacterial 3-dehydroshikimate dehydratase (QsuB) targeted to plastids is an effective approach to reroute the shikimate pool away from lignin biosynthesis, resulting in plants that accumulate the simple aromatic protocatechuate (3,4-dihydroxybenzoic acid, DHBA) at the expense of lignin (Eudes et al., 2015).This strategy was translated to poplar to design lines with reduced lignin and higher DHBA content in biomass (Unda et al., 2022).Engineered QsuB poplar lines show reduced recalcitrance to enzymatic saccharification, release higher amounts of simple aromatics, and enable higher titres of biofuels and bioproducts after deconstruction and microbial fermentation (C.-Y.Lin et al., 2022;Umana et al., 2022).
In this study, considering the important role of the shikimate and phenylpropanoid pathways in plant secondary metabolism (Maeda and Dudareva, 2012;Dong and Lin, 2021), we investigated the metabolic and transcriptional changes that occur in stems of transgenic QsuB poplar.In order to gain a comprehensive overview of these changes we analysed separately the xylem, phloem, and periderm tissues.The xylem consists of fibre cells and vessels with lignified thickened secondary cell walls that participate in the transport of water and provide mechanical strength for vertical growth; the phloem is composed of sieve elements for the transport of nutrients as well as fibres; and the periderm is rich in suberized cells and has protective functions.Moreover, the xylem tissue was collected from three different parts of the stem (i.e.older, intermediate, and younger) to obtain three distinct stages of growth and secondary cell wall formation.In addition to the wild-type control, three QsuB transgenic lines with varying levels of QsuB protein abundance, lignin reduction, and DHBA accumulation were analysed (Fig. 1A, B).

Plant material
Hybrid poplar (Populus alba × grandidentata; P39) wild type (WT) and QsuB transgenic lines were previously described (Unda et al., 2022).WT and lines QsuB1, QsuB5, and QsuB15 were grown in a greenhouse as described in C.-Y. Lin et al., 2022.Samples for transcriptomic and metabolomic analyses were collected following 5 months' growth.Stems were cut 5 cm above the root collar and segments of 5 cm were collected above the first (bottom), 20th (middle), and 40th (top) internodes.The bark was peeled and developing xylem tissue was collected by scraping the surface of the debarked xylem segments using a sterile razor blade.This approach generated developing xylem samples from the older (bottom segment), intermediate (middle segment), and younger (top segment) parts of the stem (Supplementary Fig. S1A).The bark collected from the bottom stem segment was used to generate periderm (outer bark) and phloem (inner bark) tissue samples using a razor blade.Since the tissues were manually separated, some cork and vascular cambium tissues were possibly present in the phloem samples.All the samples were immediately placed in liquid nitrogen, and stored at −80 °C until time of use.Frozen samples were pulverized using a freezer mill (model 6875D, SPEX SamplePrep LLC, Metuchen, NJ, USA).Each sample was ground twice for 1 min at a grinding rate of 15 counts per second, including a 15 s break between the two cycles.

RNA extraction, library preparation, and sequencing
Four biological replicates from each transgenic poplar line and WT control were used for large-scale transcriptomic analysis.Total RNA was extracted using the RNeasy Plant Mini Kit (Qiagen) and treated with RNAse-free DNAse.Total RNA was checked for integrity using a BioAnalyzer with an Agilent RNA 6000 Nano Chip, following the manufacturer's instructions (Agilent).Plate-based RNA sample preparation was performed on the Sciclone NGS robotic liquid handling system (PerkinElmer), using the TruSeq Stranded mRNA HT sample prep kit with poly-A selection of mRNA following the manufacturer's instructions (Illumina).Total RNA starting material was 1 μg per sample and eight PCR cycles were used for library amplification.The prepared libraries were then quantified by qPCR using the Kapa SYBR Fast Illumina Library Quantification Kit (Kapa Biosystems), and run on a LightCycler 480 real-time PCR instrument (Roche).The quantified libraries were prepared for sequencing on the Illumina HiSeq sequencing platform using a TruSeq paired-end cluster kit v4 and Illumina's cBot instrument to generate a clustered flow cell for sequencing.Sequencing of the flow cell was performed on the Illumina HiSeq2500 sequencer using HiSeq TruSeq SBS sequencing kits v4 following a 2 × 150 indexed run procedure.

RNA-seq data processing
Raw fastq file reads were filtered and trimmed using the Joint Genome Institute QC pipeline and BBDuk (https://sourceforge.net/projects/ bbmap/).Raw reads were evaluated for artifact sequence by kmer matching (kmer=25), allowing one mismatch, and detected artifact trimmed from the 3ʹ end of the reads.RNA spike-in reads, PhiX reads, and reads containing any Ns were removed.Quality trimming was performed using the phred trimming method set at Q6. Finally, following trimming, reads under the length threshold were removed (minimum length 25 bases or a third of the original read length-whichever is longer).Filtered reads from each library were aligned to the Populus trichocarpa v4.1 reference genome (Tuskan et al., 2006; https://phytozomenext.jgi.doe.gov/info/Ptrichocarpa_v4_1) using HISAT2 version 2.2.1 (Kim et al., 2015).Strand-specific coverage bigWig files were generated using deepTools v3.1 (Ramírez et al., 2014).FeatureCounts was used to generate the raw gene counts file using gff3 annotations (Liao et al., 2014).Only primary hits assigned to the reverse strand were included in the raw gene counts.
UHPLC normal phase chromatography was performed using an Agilent 1290 LC stack, with MS and MS/MS data collected using a QExactive HF Orbitrap MS (Thermo Fisher Scientific).Full MS spectra were collected from m/z 70 to 1050 at 60k resolution in both positive and negative ionization mode, with MS/MS fragmentation data acquired using stepped then averaged 10, 20, and 40 eV collision energies at 15 000 resolution.Mass spectrometer source settings included a sheath gas flow rate of 55 (au), auxiliary gas flow of 20 (au), spray voltage of 3 kV (for both positive and negative ionization modes), and capillary temperature or 400 °C.For polar metabolites, normal phase chromatography was performed using a HILIC column (InfinityLab Poroshell 120 HILIC-Z, 2.1 × 150 mm, 2.7 µm, Agilent), at a flow rate of 0.45 ml min -1 with a 2 μl injection volume.The column was held at 40 °C equilibrated with 100% buffer B (99.8% 95:5 v/v ACN:H 2 O, 0.2% acetic acid, 5 mM ammonium acetate) for 1 min, diluting buffer B to 89% with buffer A (99.8% H 2 O, 0.2% acetic acid, 5 mM ammonium acetate, 5 µM methylenedi-phosphonic acid) over 10 min, down to 70% over 4.75 min, to 20% over 0.5 min, and finally isocratic elution for 2.25 min, followed by column re-equilibration by returning to 100% B over 0.1 min and isocratic elution for 3.9 min.For non-polar metabolites, reverse phase chromatography was performed using a C 18 column (Zorbax Eclipse Plus C 18 , Rapid Resolution HD, 2.1 × 50 mm, 1.8 μm, Agilent) at a flow rate of 0.4 ml min -1 with a 2 μl injection volume.To detect metabolites with C 18 chromatography, samples were run on the column at 60 °C equilibrated with 100% buffer A (100% H 2 O, 0.1% formic acid) for 1 min, diluting buffer A to 0% with buffer B (100% acetonitrile, 0.1% formic acid) over 7 min, and isocratic elution for 1.5 min, followed by column reequilibration by returning to 100% A over 1 min and isocratic elution for 1 min.Samples consisted of four biological replicates each and extraction controls, with sample injection order randomized and an injection blank of 100% methanol run between each sample, with the blank replaced by an injection of internal standard mix every third sample, as well as QC mix every 15 samples.The raw data from the metabolite LC-MS runs are provided in Supplementary Dataset S1.
For data analysis, features with 0.5 min<retention time (RT)<9.5 min (C 18 column) and 0.8 min<RT<18.5min (HILC column), and max peak height fold-change between sample and extraction control >3 were selected.LC-MS data were analysed using custom Python code (Yao et al., 2015), with each detected peak assigned a level of confidence indicated by a score from 0 to 3 in the compound identification.We performed a Feature-Based Molecular Networking (FBMN) workflow using MZmine 2 (Pluskal et al., 2010) and Global Natural Products Social Molecular Networking (GNPS; http://gnps.ucsd.edu;Wang et al., 2016).The MZmine workflow was used to generate a list of features obtained from extracted ion chromatograms containing chromatographic peaks within a narrow m/z range and filtered to remove isotopes.For each feature, the most intense fragmentation spectrum was uploaded to GNPS for putative identification by comparison with mass spectra deposited in the database.Compound classes were attributed to identified compounds using ClassyFire (Djoumbou Feunang et al., 2016) or NPClassifier (Kim et al., 2021).Putative metabolite structures were assigned based on RT and fragmentation patterns.For some metabolites, identification was based on exact mass and comparing RT and fragmentation spectra with those of standard compounds using the same LC-MS methods (Supplementary Dataset S2; Supplementary Table S1).For each compound, the measured masses agreed with the expected theoretical masses within less than 5 ppm mass error.

Protein extraction and proteomic liquid chromatographytandem mass spectrometry analysis
Tryptic peptides were prepared by following an established proteomic sample preparation protocol (Chen et al., 2023).Proteins were extracted from the xylem of intermediate stems by homogenizing frozen tissue powder (200 mg) with 500 µl of protein extraction buffer consisting of 50 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 5 mM DTT in a centrifuge tube and using a clean pestle.Extracts were cleared by centrifugation (20 000×g) and proteins were precipitated with addition of 1 mM NaCl and 4 volumetric equivalents of acetone, followed by two additional washes with 80% acetone in water.The recovered protein pellet was homogenized by pipette mixing with 100 mM ammonium bicarbonate in 20% methanol.Protein concentration was determined by the DC protein assay (Bio-Rad Laboratories, Hercules, CA, USA).Protein reduction (15-20 µg total protein) was accomplished using 5 mM tris 2-(carboxyethyl)phosphine for 30 min at room temperature, and alkylation was performed with 10 mM iodoacetamide for 30 min at room temperature in the dark.Overnight digestion with trypsin was accomplished at a 1:50 trypsin:total protein ratio.The resulting peptide samples were analysed on an Agilent 1290 UHPLC system coupled to a Thermo Fisher Scientific Orbitrap Exploris 480 mass spectrometer for discovery proteomics (Chen et al., 2022).Briefly, peptide samples were loaded onto an Ascentis ES-C18 Column (Sigma-Aldrich, St Louis, MO, USA) and separated over a 10 min gradient from 98% solvent A (0.1 % formic acid in H 2 O) and 2% solvent B (0.1% formic acid in acetonitrile) to 65% solvent A and 35% solvent B. Eluting peptides were introduced to the mass spectrometer operating in positive-ion mode and were measured in data-independent acquisition (DIA) mode with a duty cycle of three survey scans from m/z 380 to m/z 985, and 45 MS2 scans with precursor isolation width of 13.5 m/z to cover the mass range.DIA raw data files were analysed by an integrated software suite DIA-NN (Demichev et al., 2020).The database used in the DIA-NN search (library-free mode) is the latest Populus alba (white poplar) UniProt proteome fasta sequences (UP000309997) plus the protein fasta sequences of QsuB (UniProt accession number Q8NT86) and common proteomic contaminants.DIA-NN determines mass tolerances automatically based on first pass analysis of the samples with automated determination of optimal mass accuracies.The retention time extraction window was determined individually for all MS runs analysed via the automated optimization procedure implemented in DIA-NN.Protein inference was enabled, and the quantification strategy was set to Robust LC=High Accuracy.Output main DIA-NN reports were filtered with a global false discovery rate (FDR)=0.01 on both the precursor level and protein group level.The sum of peak areas of identified tryptic peptides that derive from the QsuB protein was used to quantify the QsuB protein abundance in the samples.Four biological replicates were used for each genotype.

Data analysis
Principal component analysis plots were obtained with MetaboAnalyst 5.0 available at https://www.metaboanalyst.ca/(Pang et al., 2021) and venn diagrams were generated with the webtool available at https:// bioinformatics.psb.ugent.be/webtools/Venn/.For the Gene Ontology (GO) enrichment analysis, the web-based g:Profiler tool was used to identify over-represented biological processes in the QsuB lines (Kolberg et al., 2023).The built-in algorithms of g:Profiler were used to correct for multiple testing and to identify the leading GO terms and remove the redundant terms in each function component.Leading GO terms that were significantly enriched (adjusted P-value≤0.05) in at least one QsuB line were visualized in a dot plot using the ggplot2 package v3.4.4 in R v4.3.1 (Wickham, 2009;R Core Team, 2018).For the Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, the enrichKEGG function with the clusterProfiler package v4.8.3 (Wu et al., 2021) in R was applied to identified over-represented KEGG categories in the QsuB lines based on the UniProt IDs of their differentially expressed genes.The P-value was adjusted using the FDR method with a cut-off of 0.1.The significantly enriched KEGG categories were visualized in a dot plot, as described above.

Results
Transcriptional changes in stem tissues of engineered QsuB poplar The periderm and phloem tissues collected from the older part of the stem, as well as xylem tissues collected from the older, intermediate, and younger parts of the stem, were analysed by transcriptomics.Three selected genes known to be preferentially expressed in either the periderm (PtCER1, Rains et al., 2018), or the phloem (PtSEOR, Chen et al., 2021a), or the xylem (PtCESA8, Sundell et al., 2017) were used to validate sampling quality (Supplementary Fig. S1B).The QsuB transcript levels in the transgenic lines were higher in the xylem tissues compared with the phloem and periderm, and line QsuB1 showed higher QsuB expression compared with QsuB15.QsuB transcripts levels were comparable in lines QsuB1 and QsuB5, which does not correlate with QsuB protein abundance in these lines and suggests reduced translation of the QsuB transcripts or lower QsuB protein stability in QsuB5 (Supplementary Fig. S1C).In WT control trees, principal component analysis (PCA) demonstrated a clear distinction between the periderm, phloem, and xylem tissues, whereas the xylem samples from older, intermediate, and younger stem sections were more similar to each other (Fig. 1C).A total of 31 082 transcripts were identified across all tissue types in WT, representing 88% of the 34 699 predicted protein-coding genes found in the poplar genome.The expression level of these genes in the different tissues is shown in Supplementary Table S2.We found transcripts specific to the periderm (602), phloem (242), and xylem (615 including 88, 106, and 143 transcripts specific to the xylem from older, intermediate, and younger stem sections, respectively), while 26 430 genes were found to be expressed in all tissue types (Fig. 1D).A list of tissue-specific transcripts identified in stems of the WT is provided in Supplementary Table S3.In each tissue, PCA of the transcripts permitted the discrimination of different genotypes and showed a separation between the low-lignin QsuB lines (QsuB1 and QsuB15), which grouped together, and QsuB5, which was closer to the WT (Supplementary Fig. S2).These differences were more important in xylem tissues compared with the phloem and periderm.In more detail, the data showed that line QsuB1 had the largest number of differentially expressed genes (DEGs) compared with WT in all the tissues analysed, followed by QsuB15 and QsuB5, which correlates with the levels of lignin reduction and DHBA accumulation in these lines (C.-Y.Lin et al., 2022;Unda et al., 2022) (Fig. 2; Supplementary Dataset S3).In the xylem from older stem sections of lines QsuB1 and QsuB15, a total of 857 (329 down-regulated, 528 up-regulated) and 488 (279 down-regulated, 309 up-regulated) transcripts were differentially expressed, respectively.These values indicate that 3.1% and 1.8% of the genes expressed in this tissue had a different expression level in QsuB1 and QsuB15 compared with WT, respectively.The smallest number of DEGs was found in the periderm in QsuB1, accounting for 495 transcripts (185 down-regulated, 310 up-regulated), while the phloem showed the fewest DEGs in QsuB15 (209 transcripts, 65 downregulated, 144 upregulated) (Fig. 2).Line QsuB5 had a much smaller number of DEGs, with the xylem of intermediate stem sections displaying the highest count at 107 DEGs (Fig. 2).A list of the 10 most up-regulated and down-regulated transcripts in the different tissues of each QsuB line is provided in Supplementary Table S4.A majority of the DEGs were only observed in a specific tissue: In the case of QsuB1, 178, 291, and 856 transcripts were up-regulated specifically in the periderm, phloem, and xylem, respectively (Fig. 3A).Similarly, 92, 146, and 408 transcripts were down-regulated specifically in the aforementioned tissues (Fig. 3B).Moreover, new and silenced transcripts with distinct tissue-specific patterns were also observed in transgenics.In line QsuB1, 43, 77, and 262 new transcripts were detected in the periderm, phloem, and xylem, respectively (Fig. 3C), whereas 37, 29, and 143 silenced transcripts were identified in these tissues, respectively (Fig. 3D).Similar observations were made regarding the tissue specificity of up-regulated, down-regulated, new, and silenced transcripts identified in lines QsuB5 and QsuB15 (Supplementary Fig. S3; Supplementary Dataset S3).
We performed GO and KEGG analyses to determine whether the DEGs identified in the transgenic plants were significantly enriched in specific biological processes, molecular functions, and metabolic pathways.Among DEGs found in lines QsuB1 and QsuB15, we noticed an enrichment in genes associated with transmembrane transport, oxidoreductase activity, glutathione metabolism, secondary metabolism (phenylpropanoids, flavonoids, anthocyanins, and cytokinins), glycosylation, iron binding, and secondary cell wall synthesis in the xylem and phloem tissues.DEGs in line QsuB5 were only enriched in genes linked to transport and glycosylation activity in the xylem (Fig. 4; Supplementary Fig. S4).

Metabolic changes in stem tissues of engineered QsuB poplar
Separate aliquots of tissue samples used for transcriptome analysis were used for non-targeted metabolomics.We extracted metabolites with methanol and conducted metabolite profiling with a LC-MS platform that utilized gradient-based hydrophilic interaction liquid chromatography (HILIC) and reverse phase (C 18 ) chromatography, in both the positive and the negative ionization modes.A summary of the number of features detected in each stem tissue for each genotype using HILIC in the positive ionization mode is presented in Fig. 5A.In stem sections from WT trees, PCA showed a clear distinction between the periderm, phloem, and xylem tissues, whereas the xylem samples from older, intermediate, and younger stem sections were more similar to each other (Fig. 5B).A total of 16 095 features were identified in WT, including several specific to the periderm (168), phloem (323), and xylem (786 including 19, 16, and 67 features specific to the xylem from older, intermediate, and younger stems, respectively), whereas 10 059 features were detected in all tissue types (Fig. 5C).In every tissue examined, PCA grouped QsuB1 and QsuB15 together, separating them from the QsuB5/WT group (Supplementary Fig. S5).Most of the features detected in each tissue were present in all four genotypes, but some were only found in transgenics or the WT (Fig. 5D).In particular, 9880, 10 605, 11 574, 14 620, and 14 117 features were found in all genotypes in the xylem bottom, xylem middle, xylem top, phloem, and periderm stem tissues, respectively.A total of 10 774, 11 345, 12 130, 14 902, and 14 538 features were common to both WT and QsuB1 in the aforementioned tissues, respectively (Fig. 5D).We examined the relative abundance of these metabolites detected in both the WT control and transgenics (Fig. 6; Supplementary Dataset S4).In QsuB1, the line featuring the lowest lignin and highest DHBA content, we found that 19.4% (2092 out of 10 774) of the features detected in the xylem from older stem sections were differentially abundant compared with WT.This percentage was reduced to 16.7% and 13.2% in the xylem from intermediate and younger stems, respectively, but represented 23.0% in the phloem and 18.2% in the periderm (Fig. 6, top panels).In QsuB5 and QsuB15, the highest percentage of differentially abundant features was also found in the phloem (1.5% and 20.0%, respectively) (Fig. 6, middle and bottom panels).Overall, more features increased than decreased in the transgenics, across all tissues (Fig. 6; Supplementary Dataset S4).
We next focused on the distribution of the differentially abundant features across different tissues (Fig. 7; Supplementary Fig. S6).In QsuB1, a total of 4189 unique features were significantly more abundant compared with WT across all tissues combined, including 262 (6.3%) that were increased in all five tissue types, and 597, 478, and 1107 that were more abundant specifically in the periderm, phloem, and xylem, respectively (Fig. 7A).In contrast, 2316 unique features were significantly less abundant in QsuB1 in all tissues combined, including 23 (1.0%) that were reduced in all five tissue types, and others that were less abundant specifically in the periderm (175), phloem (1011), and xylem (589) (Fig. 7B).These results show that the highest number of more abundant features was found in the xylem, whereas the phloem contained the highest number of less abundant features when comparing QsuB1 to WT.The total number of unique, differentially abundant features was higher in QsuB1 (6505) compared with QsuB5 (492) and QsuB15 (5199).Several features were new or depleted (i.e.absent) across all tissue types in the different transgenic lines compared with WT (Fig. 7; Supplementary Fig. S6; Supplementary Dataset S4).In QsuB1, 223 new features were found in all tissue types (Fig. 7C), with only two features absent from all tissues analysed (Fig. 7D).The majority of these features were specific to the periderm (260 new/127 depleted), phloem (166 new/164 depleted), and xylem (1839 new/2293 depleted), with again some specificity within the xylem collected from old, intermediate, or young stem sections (Fig. 7C, D).Similarly, in line QsuB15, a large number of more abundant features (1033 out of 3158) was found specifically in the xylem, whereas the phloem contained the largest number of less abundant features (776 out of 2041), and the largest number of both new and depleted features was found in the xylem (Supplementary Fig. S6).QsuB15 showed fewer unique new features (2392) compared with QsuB1 (3024), but had a larger number of depleted features (942 vs. 843) (Fig. 7; Supplementary Fig. S6).Next, using GNPS, a subset of differentially abundant features observed in transgenics were classified and putatively identified.Consistent with the activity of QsuB within the shikimate pathway, a majority of the differentially abundant metabolites identified in transgenics were found to belong to the class of 'shikimates and phenylpropanoids' (Fig. 8; Supplementary Fig. S7).Finally, 10 962 unique features were detected across all stem tissues and genotypes using HILIC in the negative ionization mode.Their characteristics and the analysis of relative abundance and tissue others, as well as important percentages of differentially abundant features.For example, compared with WT and across all five tissues, line QsuB1 had 3180 and 1652 new and depleted features, respectively.In the xylem of older stems, 21.7% and 17.7% of the detected features were differentially abundant in QsuB1 and QsuB15, respectively.These metabolites mainly belong to the class of 'shikimates and phenylpropanoids', as previously observed in the metabolite analysis conducted with HILIC.

Impact of QsuB expression on the shikimate and lignin biosynthetic pathways
We analysed specifically the metabolites and genes involved in the different steps of the shikimate and lignin pathways (Fig. 9; Supplementary Dataset S9).Notably, these pathways are responsible for the synthesis of the monolignols p-coumaryl, coniferyl, and sinapyl alcohols that give rise to the p-hydroxyphenyl (H), guaiacyl (G), and syringyl (S) subunits of lignin.Consistent with the enzymatic reaction catalysed by QsuB, large increases of DHBA and significant decreases of 3-dehydroshikimate were measured in the different stem tissues from lines QsuB1 and QsuB15.These changes were more important in the xylem tissues where QsuB expression was higher compared with the phloem and periderm (Supplementary Dataset S9).The content of 3-dehydroquinate, the precursor of 3dehydroshikimate, was also reduced in the xylem and phloem, whereas shikimate was unchanged or even increased in the 'xylem top', despite the down-regulation of the two 3-dehydroquinate dehydratase/shikimate dehydrogenase genes (step 3, Potri.013G029800and Potri.010G019000).A conversion of 3-dehydroquinate into shikimate mediated by quinate dehydratase/dehydrogenase could explain these observations (step 4, Potri.013G029900 and Potri.005G043400)(Guo et al., 2014).Downstream of the shikimate pathway, the levels of the aromatic amino acids tryptophan, tyrosine, and phenylalanine were unchanged in the three xylem types.Nonetheless, a significant increase of tyrosine in the periderm and decrease of phenylalanine in the phloem and periderm was observed.
A major metabolic change in lines QsuB1 and QsuB15 was the increased levels of several metabolites derived from cinnamate and p-coumarate.4-Hydroxybenzoate, a metabolite known to be synthesized from p-coumarate via β-oxidative and non-β oxidative routes, was higher in both lines.Salicylate, which is preferentially made from cinnamate in poplar (Ullah et al., 2023), and its derivatives 2,3-dihydroxybenzoate and 2,5-dihydroxybenzoate were also increased in the transgenics.The closest homologs of the Arabidopsis genes encoding salicylate 3-hydroxylase (Potri.011G150100and Potri.011G150200) and salicylate 5-hydroxylase (Potri.015G002800and Potri.012G006300) were overexpressed and thus represent strong candidates for salicylate hydroxylation in QsuB1 and QsuB15.The higher abundance of hydroxybenzoates and dihydroxybenzoates led to a large accumulation of multiple hydroxybenzoate and dihydroxybenzoate glycosides in QsuB poplar: 59 distinct metabolites featuring m/z and fragmentation patterns that are characteristic of hydroxybenzoates conjugated to sugar moieties were found to be increased or only detectable in QsuB1 and QsuB15 (Supplementary Dataset S9).A significant increase in caffeate was also apparent in the xylem from these two lines, possibly due to the overexpression of phenylalanine ammonia lyase (Potri.006G126800)

Shikimates and Phenylpropanoids Terpenoids
Metabolite classes: Fig. 8. Classification of differentially abundant metabolites identified in each tissue of QsuB1 using hydrophilic interaction liquid chromatography (positive ionization mode).For each class, the number of metabolites is indicated inside the corresponding slice of the pie chart.A breakdown of new and depleted metabolites (upper panels) and of increased and decreased metabolites (lower panels) is indicated next to each colour class symbol.and of the two putative p-coumarate 3-hydroxylase genes (Supplementary Dataset S9) (Barros et al., 2019), whereas ferulate and sinapate were reduced in this tissue, and, counterintuitively, 5-hydroxyferulate could only be detected in QsuB1 and QsuB15.In addition, metabolites that are derived from p-coumaroyl-CoA such as naringenin and p-coumaraldehyde were increased in the transgenic lines.Naringenin was only detected in lines QsuB1 and QsuB15 in the xylem, and showed a 32-fold increase in the phloem from these two lines.The increase of p-coumaraldehyde observed in all tissues from QsuB poplar is consistent with the higher content of H lignin units measured in transgenic trees, since they are derived from the polymerization of p-coumaryl alcohol (Unda et al., 2022).This change in lignin monomer composition is presumably the consequence of reduced hydroxycinnamoyl-CoA: shikimate hydroxycinnamoyl transferase (HCT) expression and/ or activity.To support this hypothesis, we observed a downregulation of the two HCT genes involved in lignin synthesis (step 19, Potri.003G183900/PtHCT1 and Potri.001G042900/PtHCT6) and a reduction of the HCT product p-coumaroyl shikimate in the different tissues from QsuB1 and QsuB15 (Fig. 9).More generally, all the genes from the phenylpropanoid and monolignol pathways appeared down-regulated in most tissues analysed, except for one cinnamyl alcohol dehydrogenase gene (Potri.016G078300)that was up-regulated in the periderm and for the gene encoding p-hydroxybenzoyl-CoA monolignol transferase (Potri.001G448000/pHBMT;Zhao et al., 2021;de Vries et al., 2022) that was up-regulated in both the phloem and periderm.We examined the expression of peroxidase and laccase genes predicted to be involved in lignin polymerization and observed mixed expression profiles showing either up-regulation or down-regulation across the different tissues from QsuB1.For example, focusing on three peroxidase and three laccase genes that have been implicated genetically in lignification, we found inconsistent fold changes in expression when comparing laccases and peroxidases, and also when comparing the xylem tissues with phloem and periderm (steps 26/27 on Fig. 9; Supplementary Dataset S9).

Discussion
The transcriptome and metabolome are commonly remodelled in response to transgene expression in plants.Our current results show that heterologous expression of a plastid-targeted bacterial 3-dehydroshikimate dehydratase induces noticeable transcriptional and metabolic changes in poplar (Figs 2,  6).In the developing xylem of the older stem section of a line displaying high QsuB expression, we found that 3% of the transcripts had altered expression levels and that ~19% of the detected metabolites had differential abundance.Similar observations were made in lignin-modified poplar lines that express a bacterial shikimate kinase and feature up to 2464 DEGs and 2025 differentially abundant metabolites (or ~27% of the total detected features) in the xylem compared with WT (Hu et al., 2022).These results are consistent with the modification of secondary cell wall synthesis via overexpression of serine hydroxymethyltransferase, which resulted in the deregulation of up to 1159 genes in poplar leaves (Zhang et al., 2019).
In connection with the shikimate pathway, the heterologous expression of bacterial salicylate synthase targeted to plastids led to 1716 DEGs (~8% of the total transcripts) in poplar leaves (Xue et al., 2013).Interestingly, field-grown engineered salttolerant poplar expressing a hormone-responsive transcription factor from tomato had 1.3% of DEGs in apical buds compared with WT, but this percentage was actually greater (~5% of the transcripts) when comparing identical poplar genotypes grown at two different field locations, indicating a large effect of environmental factors on gene expression (Zhang et al., 2022).This observation is perhaps unsurprising considering that a single transient bending of the stem in poplar can affect the expression of ~6% of the transcripts (Pomiès et al., 2017).Alteration in the expression of multiple monolignol biosynthetic genes is often detected in poplar RNAi lines silenced for one particular gene of the lignin pathway, suggesting genetic interactions between these genes such as epistasis or other unknown higher-order regulation (Wang et al., 2018;Matthews et al., 2020).In the case of QsuB poplar, we observed a down-regulation of genes from the core phenylpropanoid and monolignol pathways, indicating that modifying the pools of 3-dehydroshikimate and DHBA (i.e. the substrate and product of QsuB, respectively) indirectly changes the expression of these genes.Such observations on gene downregulation were also made in engineered poplar trees that express a bacterial shikimate kinase (Hu et al., 2022).Shikimate and quinate are synthesized inside plastids, while the actual pools of shikimate and quinate available in the cytosol for enzymes such as HCT and hydroxycinnamoyl-CoA:quinate hydroxycinnamoyl transferase (HQT) remain undetermined in QsuB poplar.The reduction of caffeoyl quinate (chlorogenic acid) and feruloyl quinate in the xylem and phloem cells in transgenics could be the result of low quinate levels in the cytosol since HQT expression is unchanged (Fig. 9; Supplementary Dataset S9) (Zhang et al., 2018).Previous work showed that silencing PtHCT1 and PtHCT6 reduces lignin and increases the relative amount of H units, similar to the lignin characteristics observed in QsuB poplar (Wang et al., 2018;Unda et al., 2022).Besides its canonical substrate shikimate, PtHCT6 was shown to accept DHBA, 2,3-dihydroxybenzoate, and 2,5dihydroxybenzoate (Eudes et al., 2016), suggesting that higher concentrations of these dihydroxybenzoates in the QsuB poplar lines could partially reduce HCT activity via competitive inhibition.The identification of features with mass ions matching those of p-coumaroyl dihydroxybenzoates and caffeoyl dihydroxybenzoates that are found to be more abundant or only detected in the QsuB lines supports this hypothesis (Fig. 9; Supplementary Dataset S9).
The occurrence of high amount of DHBA glycosides in QsuB poplar implies the export of DHBA from chloroplasts to the cytosol, the activity of UDP-glycosyltransferases (UGTs), and presumably the storage of DHBA conjugates in the vacuole, as previously described for other benzoates synthesized in plastids (Eudes et al., 2008).Both ATP-and proton gradientdependent transporters have been implicated in the uptake of benzoate glucose conjugates by plant vacuoles, but their identification is still pending (Bartholomew et al., 2002;Vaca et al., 2017).Very few plastidic transporters involved in the export of metabolites derived from the shikimate pathway have been characterized (Serrano et al., 2013;Widhalm et al., 2015), and whether DHBA produced in chloroplasts in QsuB poplar is exported via a transport protein remains unknown.Our transcriptomic data showed that DEGs in the QsuB lines are largely enriched in genes involved in transmembrane transport (Fig. 4; Supplementary Fig. S4).A gene encoding a protondependent transporter from the major facilitator superfamily (Potri.018G041700)and predicted to localize to plastids is among the 10 most overexpressed genes in all tissue types in QsuB1 and QsuB15 (Supplementary Table S4), making it a possible candidate involved in the non-specific export of DHBA.Moreover, we cannot exclude that DHBA export competes with the export of shikimate from plastids, hence affecting the cytosolic shikimate pool.Interestingly, results from previous studies suggest that the cytosolic shikimate pool could act as a sensor of the flux into lignin, and its reduction may favour the production of other phenylpropanoids (e.g.flavonoids) at the expense of lignin (Adams et al., 2019;Dixon and Barros, 2019).Such a redirection of the metabolic flux might also contribute to the overproduction of salicylates observed in QsuB poplar.DEGs were found to be significantly enriched in genes related to glycosyltransferase activity, including several encoding UGTs from the CAZy GT1 family (Fig. 4, Supplementary Fig. S4; Supplementary Table S4).UGT-7 (Potri.007G132400),UGT-37 (Potri.016G016800),and UGT-40 (Potri.018G096000),which are found among the 10 most overexpressed genes in QsuB1 and QsuB15, represent strong candidates for synthesizing DHBA and other benzoate glycosides in transgenics considering their known activity towards hydroxybenzoates (Saint-Vincent et al., 2023).
We previously observed an increase of xylan content and changes in the meso-scale structure and organization of cellulose microfibrils in the cell walls of QsuB poplar, but the molecular basis for these observations is unclear (C-Y.Lin et al., 2022;Unda et al., 2022;Senanayake et al., 2024).GO analysis retrieved 22 genes associated with 'secondary cell wall biogenesis' among the DEGs identified in the xylem from older stem sections of QsuB poplar (Fig. 4; yellow highlights in Supplementary Dataset S3).These include 10 fasciclin-like arabinogalactan protein (FLA) genes that are down-regulated in both QsuB1 (up to 60-fold) and QsuB15.The same FLA genes (PtFLA1-PtFLA10) were shown to be up-regulated in the tension wood produced in response to bending and characterized by a 'G-layer' of crystalline cellulose in fibre cells (Lafarguette et al., 2004;Andersson-Gunnerås et al., 2006).FLA genes have been implicated in cellulose synthesis and deposition in woody tissues, but their exact mechanisms of action remain unclear (MacMillan et al., 2010;S. Lin et al., 2022).Noteworthily, a recent study in Arabidopsis proposed that FLAs could act as cell surface sensors to fine-tune the balance of lignin and cellulose synthesis in secondary cell walls during development (Ma et al., 2022).In QsuB poplar, PtFLA1-PtFLA10 are specifically down-regulated in the xylem from the bottom part of the stem, suggesting that changes induced by QsuB affect PtFLA expression only at the later stages of secondary cell wall deposition.Whether the modification and reduced crystallinity of cellulose observed in QsuB poplar results, in part, from the down-regulation of PtFLA genes warrants further investigation (Senanayake et al., 2024).Moreover, a gene encoding a COBRA-Like (COBL) protein (Potri.004G117200-COBL4) is highly down-regulated in the 'bottom xylem' of transgenics QsuB1 and QsuB15 (Supplementary Dataset S3).COBL proteins are putative glycosylphosphatidylinositol-anchored proteins that control cellulose deposition and microfibril orientation in the cell wall, and COBL4 is also known to be upregulated during tension wood formation (Andersson-Gunnerås et al., 2006;Liu et al., 2021).Although glucan content is unchanged in QsuB poplar stems (C-Y.Lin et al., 2022;Unda et al., 2022), one cellulose synthase-like (GT2) gene was up-regulated in the 'bottom xylem' in both QsuB1 and QsuB15, and one α-expansin gene was found overexpressed in this tissue in QsuB1 (Supplementary Dataset S3).Xyloglucan is another component of the G-layer in poplar tension wood fibre cells, and xyloglucan endo-transglycosylases (XETs) are generally up-regulated during G-layer formation (Nishikubo et al., 2007).We found that one XET (GH16) and one putative xyloglucan fucosyltransferase (GT37) are strongly downregulated (up to 74-fold) and up-regulated (up to 18-fold), respectively, in both QsuB1 and QsuB15 (Supplementary Dataset S3).Further, the up-regulation in the xylem of genes involved in cytokinin production may be part of a response to cell wall modifications considering the role of these hormones in vascular cambium development and secondary growth (Fig. 4B) (Nieminen et al., 2008).Collectively, these results also suggest that tension wood synthesis upon mechanical stress could be affected in QsuB poplar, in contrast with the observations made in low-lignin antisense 4-coumarate:CoA ligase (4CL) poplar trees that produce more tension wood (Voelker et al., 2011).Three genes encoding GH3, CE13, and CE8 enzymes related to pectin synthesis were differentially expressed in the xylem of QsuB1 but not in QsuB15 (Supplementary Dataset S3), suggesting some indirect effects on pectin metabolism with higher QsuB expression levels.
The higher content of 4-hydroxybenzoate and its glycosides measured in methanolic extracts of QsuB poplar somehow coincides with the observed reductions in 4-hydroxybenzoate ester linked to lignin previously shown in the transgenics (Unda et al., 2022) (Fig. 9; Supplementary Dataset S9).One plausible explanation is the down-regulation in the xylem of the gene encoding the transferase responsible for these 4-hydroxybenzoate groups to be conjugated to monolignols, which can then be transported and participate in lignification (pHBMT, step 25 in Fig. 9).The preferential expression of pHBMT in the xylem as determined from our RNA-seq data is consistent with the occurrence of 4-hydroxybenzoate groups specifically in the cell walls of xylem fibers (Supplementary Dataset S3) (Goacher et al., 2021).pHBMT uses 4-hydroxybenzoyl-CoA and monolignols as substrates, but its activity towards dihydroxybenzoyl-CoA donors such as protocatechuyl-CoA has not been explored, and the transferase responsible for the formation of DHBA ester groups on lignin in QsuB poplar remains to be identified (Zhao et al., 2021;de Vries et al., 2022;Unda et al., 2022).
An increase of the stress hormone salicylate and salicinoids (i.e.grandidentatin, salicyloylsalicin, and tremulacin) were also observed in QsuB1 and QsuB15.Our RNA-seq data agree with other studies indicating a feedback inhibition of the shikimate pathway in modified poplar that features enhanced salicylate levels (Gordon et al., 2022).The higher levels of naringenin, taxifolin, myricetin, isorhamnetin, eriodictyol, and quercetin measured in QsuB transgenics are also consistent with previous observations that linked increases of flavonoids with higher salicylate levels (Supplementary Dataset S9) (Ullah et al., 2022).Notably, a few genes encoding transcriptional activators of flavonoid synthesis that are known to be induced by salicylate such as MYB115 (Potri.002G173900),MYB134 (Potri.006G221800),bHLH131 (Potri.005G208600),and WD40-1 (Potri.016G075800)are significantly overexpressed in the xylem of QsuB1 and/or QsuB15 (Ullah et al., 2019) (Supplementary Dataset S3).Moreover, an important enrichment of genes related to oxidoreductase, glutathione transferase, and chitinase activities among the DEGs is indicative of a stress state prevailing in QsuB poplar (Fig. 4; Supplementary Fig. S4).Iron is abundant in chloroplasts and essential for photosynthetic electron transport.DHBA produced in chloroplasts has the capacity to chelate ferric iron via its catechol moiety, which may disturb iron homeostasis, generate oxidative stress, and could explain the observed enrichment of genes involved in iron binding among the identified DEGs (Fig. 4; Supplementary Fig. S4).In this regard, specific secondary metabolites containing catechol moieties (e.g.coumarins) are typically secreted via root exudates to facilitate iron acquisition (Tsai and Schmidt, 2017), and possible changes in iron nutrition capacity remain to be explored in QsuB poplar.Considering that DHBA is a central intermediate in bacterial lignin catabolism and that lignin modifications can influence the endosphere microbiome and ectomycorrhizal colonization in poplar, it would be informative to study the microbiome in QsuB poplar (Beckers et al., 2016;Kamimura et al., 2017;Behr et al., 2020).
The differential abundance in transgenics of select metabolites from the phenylpropanoid pathway, such as the accumulation of 5-hydroxyferulate and concomitant decrease of ferulate and sinapate, cannot be explained from the observed changes in gene expression of the biosynthetic enzymes.These differences may be due to post-transcriptional regulation mechanisms and various degrees of inhibition of the enzymes by several pathway intermediates, as shown previously for phenylalanine ammonia-lyase, 4CL, caffeate 3-O-methyltransferase, ferulate 5-hydroxylase, and cinnamyl alcohol dehydrogenase (Wang et al., 2014;Zhang et al., 2020).Thus, metabolic flux analyses assisted by mathematical modelling that integrates these complex regulatory effects is needed to obtain a more comprehensive overview of the modified shikimate and lignin pathways in QsuB poplar (Wang et al., 2019;Rao and Barros, 2023).
Finally, the identification in this study of poplar transcripts that are specific to certain tissues (i.e.developing xylem at different growth stages, phloem, or periderm) provides concomitantly a comprehensive list of gene promoter candidates with wide range of activities that can be leveraged to achieve more precise spatiotemporal expression of transgenes in poplar stems (Supplementary Table S3).Fine-tuning transgene expression, notably with the design of synthetic promoters (Belcher et al., 2020;De Meester et al., 2021), should enable adequate expression of the desired engineered traits without compromising plant growth.

Fig. 1 .
Fig. 1. 3-Dehydroshikimate dehydratase (QsuB) poplar lines used in this study and analysis of transcripts isolated from stems of wild type (WT).(A) QsuB protein abundance in the xylem from intermediate stems of QsuB poplar lines.Values are means and SD of four biological replicates.Means with different letters represent statistically significant differences using Tukey's pairwise comparison (P<0.05).ND, not detected.(B) Average lignin and protocatechuate (3,4-dihydroxybenzoic acid, DHBA) contents in stems from WT and QsuB1, QsuB5, and QsuB15 transgenic lines.Debarked stems are shown.Scale bar is 1 cm.Values are from C-Y. Lin et al. (2022).(C, D) Principal component analysis plot (C) and Venn diagram (D) of the 31 082 unique transcripts identified in stem tissues from WT.

Fig. 2 .
Fig. 2. Volcano plots of transcripts identified in different stem tissues from wild type (WT) and QsuB transgenic lines.The number of down-regulated (in blue) and up-regulated (in red) transcripts in transgenics compared with WT control is indicated on each plot (log 2 fold change +2/−2 and P-value<0.05).Grey dots represent transcripts not differentially expressed.

Fig. 3 .
Fig. 3. Venn diagrams of transcripts up-regulated (A), down-regulated (B), new (C), and silenced (D) in different tissues of line QsuB1.The number of unique transcripts is indicated in brackets.

Fig. 4 .Fig. 5 .
Fig. 4. Dot plots of Gene Ontology (GO) (A) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (B) enrichment analyses of differentially expressed genes identified in line QsuB1.The size of the dots represents the number of genes associated with each ontology term and pathway.

Fig. 6 .
Fig.6.Volcano plots of features detected in different stem tissues from wild type (WT) and QsuB transgenic lines using hydrophilic interaction liquid chromatography (positive ionization mode).The number of decreased (in blue) and increased (in red) features in transgenic lines compared with WT control is indicated on each plot (log 2 fold change +2/−2 and P-value<0.05).Grey dots represent features not differentially abundant.

Fig. 9 .
Fig. 9. Transcriptional and metabolic changes in the shikimate and lignin pathways in QsuB poplar.Changes (log 2 fold change) in transcripts and metabolites in different tissues of line QsuB1 are shown.Values in bold are significantly different from the WT (P<0.05).Transcript up-regulation is indicated as positive values (red) and down-regulation as negative values (blue).Increased and decreased metabolite abundances are shown in red and green, respectively.XB, XM, and XT, xylem tissue from older, intermediate, and younger sections of the stem, respectively; Ph, phloem; Pr, periderm.Black circled numbers denote enzymatic steps.Dashed arrows indicate multiple enzymatic steps or uncharacterized enzymes.'new' indicates metabolites that are detected in QsuB1 but are below the detection limit in WT.See Supplementary Dataset S9 for detailed information on genes and metabolites, as well as the data obtained from lines QsuB5 and QsuB15.