Arogenate Dehydratase Isoenzymes Profoundly and Differentially Modulate Carbon Flux into Lignins*

Background: The plastid-localized arogenate dehydratase (ADT) gene family is hypothesized to differentially control carbon flux for lignin deposition, with lignin being the main contributor to lignocellulosic recalcitrance. Results: Single and multiple ADT knock-outs resulted in differential control over lignin content/composition. Conclusion: The first evidence for Phe upstream metabolism differentially controlling carbon flux into distinct secondary cell wall types was discovered. Significance: Upstream metabolic networks regulate secondary cell wall formation. How carbon flux differentially occurs in vascular plants following photosynthesis for protein formation, phenylpropanoid metabolism (i.e. lignins), and other metabolic processes is not well understood. Our previous discovery/deduction that a six-membered arogenate dehydratase (ADT1–6) gene family encodes the final step in Phe biosynthesis in Arabidopsis thaliana raised the fascinating question whether individual ADT isoenzymes (or combinations thereof) differentially modulated carbon flux to lignins, proteins, etc. If so, unlike all other lignin pathway manipulations that target cell wall/cytosolic processes, this would be the first example of a plastid (chloroplast)-associated metabolic process influencing cell wall formation. Homozygous T-DNA insertion lines were thus obtained for five of the six ADTs and used to generate double, triple, and quadruple knockouts (KOs) in different combinations. The various mutants so obtained gave phenotypes with profound but distinct reductions in lignin amounts, encompassing a range spanning from near wild type levels to reductions of up to ∼68%. In the various KOs, there were also marked changes in guaiacyl:syringyl ratios ranging from ∼3:1 to 1:1, respectively; these changes were attributed to differential carbon flux into vascular bundles versus that into fiber cells. Laser microscope dissection/pyrolysis GC/MS, histochemical staining/lignin analyses, and pADT::GUS localization indicated that ADT5 preferentially affects carbon flux into the vascular bundles, whereas the adt3456 knock-out additionally greatly reduced carbon flux into fiber cells. This plastid-localized metabolic step can thus profoundly differentially affect carbon flux into lignins in distinct anatomical regions and provides incisive new insight into different factors affecting guaiacyl:syringyl ratios and lignin primary structure.


How carbon flux differentially occurs in vascular plants following photosynthesis for protein formation, phenylpropanoid metabolism (i.e. lignins), and other metabolic processes is not well understood. Our previous discovery/deduction that a sixmembered arogenate dehydratase (ADT1-6) gene family encodes the final step in Phe biosynthesis in Arabidopsis thaliana raised the fascinating question whether individual ADT isoenzymes (or combinations thereof) differentially modulated carbon flux to lignins, proteins, etc. If so, unlike all other lignin pathway manipulations that target cell wall/cytosolic processes, this would be the first example of a plastid (chloroplast)-associated metabolic process influencing cell wall formation.
Homozygous T-DNA insertion lines were thus obtained for five of the six ADTs and used to generate double, triple, and quadruple knockouts (KOs) in different combinations. The various mutants so obtained gave phenotypes with profound but distinct reductions in lignin amounts, encompassing a range spanning from near wild type levels to reductions of up to ϳ68%. In the various KOs, there were also marked changes in guaiacyl: syringyl ratios ranging from ϳ3:1 to 1:1, respectively; these changes were attributed to differential carbon flux into vascular bundles versus that into fiber cells. Laser microscope dissection/ pyrolysis GC/MS, histochemical staining/lignin analyses, and pADT::GUS localization indicated that ADT5 preferentially affects carbon flux into the vascular bundles, whereas the adt3456 knock-out additionally greatly reduced carbon flux into fiber cells. This plastid-localized metabolic step can thus profoundly differentially affect carbon flux into lignins in distinct anatomical regions and provides incisive new insight into different factors affecting guaiacyl:syringyl ratios and lignin primary structure.
The final step of Phe biosynthesis, catalyzed by arogenate dehydratase (ADT) 3 in planta (1)(2)(3), is potentially a major regulatory point due to both its pivotal position at the branch point of Tyr and Phe biosynthesis (see Fig. 1) and as a linkage point between plastid/chloroplast-localized shikimate-chorismate and cytosolic/membrane/cell wall-associated phenylpropanoid metabolic networks. Together, these pathways comprise some of the most metabolically intensive networks in vascular plants. Indeed, depending upon the species, up to 50% of captured photosynthetic carbon can be in the form of Phe-derived phenylpropanoids (4,5). Furthermore, downstream phenylpropanoid-derived products can have important but distinct physiological functions in planta, including fragrances/flavors, defense molecules, UV protectants, pigments, and cell wall structural biopolymers, e.g. allyl/propenyl phenols, lignans (6), flavonoids (7), (proantho)cyanidins, stilbenes (8), phytoalexins (e.g. isoflavones) (9), lignins (10), and suberins (11,12). The broad physiological functions of phenylpropanoid-derived metabolites thus translate into a diverse and ever changing demand for the pathway intermediate Phe in different tissues and organs, i.e. in addition to Phe utilization for protein synthesis and other metabolic pathways.
Curiously, the question of pivotal regulatory metabolic networks upstream of Phe and profoundly altering carbon flux/ allocation into phenylpropanoid/lignin metabolism versus protein synthesis etc. had essentially not been addressed before. Instead, previous biotechnological manipulations targeted the presumed entry point to the phenylpropanoid pathway, phenylalanine ammonia-lyase, as well as various downstream mono-lignol pathway steps (see Anterola and Lewis (13) and Davin et al. (10) for a discussion). Most of these approaches, however, did not take into consideration the potentially seamless integration of related upstream but differentially localized metabolic networks associated with carbon flux into phenylpropanoids and transcriptional regulation thereof. This was relevant because, in previous metabolic flux studies leading to monolignols in loblolly pine (Pinus taeda), it was established that factors apparently affecting Phe availability helped control/modulate carbon flux into phenylpropanoid metabolism (14,15), i.e. rather than phenylalanine ammonia-lyase having a central ratelimiting role as had often been reported due to its entry point position to phenylpropanoids.
The ADT family was thus considered a potentially promising candidate for involvement in regulating the previously documented changes in Phe availability in plants due to its branch point position in the shikimate-chorismate pathway and its sensitivity to feedback inhibition by Phe. Indeed, we had previously characterized all six ADT isoenzymes from Arabidopsis thaliana and provided molecular and biochemical evidence supporting the arogenate route as the major mode of Phe biosynthesis ( Fig. 1) (3). Specifically, three isoenzymes, ADT3, ADT4, and ADT5, demonstrated exclusive substrate preference for arogenate, whereas isoenzymes ADT1, ADT2, and ADT6 displayed instead a strong substrate preference for arogenate but also had a limited ability to utilize prephenate (3). Further confirmatory observations of a strong substrate preference for arogenate were subsequently made for one rice ADT isoenzyme (16) and three petunia ADT isoenzymes (17). Feedback inhibition of ADTs was also demonstrated to be an important factor influencing Phe biosynthesis, accumulation, and turnover as feedback-insensitive ADTs in both rice and Arabidopsis were found to accumulate approximately 55 and 160 times more Phe, respectively, compared with wild type (WT) (16). Thus, given the important role of ADT and considering the vast range of uses for Phe in planta, it was instructive to determine whether different isoenzymes potentially participate in distinct metabolic networks involving Phe.
To begin to delineate the potential individual physiological contributions of specific ADT isoenzymes, Arabidopsis lines containing knock-outs (KOs) of single and multiple ADT genes were generated and then initially analyzed for potential differential effects on phenylpropanoid metabolism (specifically lignification). It was thus established that lines with a combination of ADT4 and ADT5 KOs had profoundly altered lignin contents, including the various triple and quadruple KOs involving those isoenzymes, that led to even more pronounced effects. This is, therefore, the first demonstration that modulation of a network pathway step (ADT) upstream of phenylpropanoid metabolism, localized in plastids/chloroplasts, can differentially alter carbon allocation/flux into lignification (phenylpropanoid metabolism) versus formation of Phe for either protein synthesis or some other metabolic pathway.

EXPERIMENTAL PROCEDURES
All commercial kits were used according to the manufacturer's instructions with any minor deviations noted.
Generation and Confirmation of Single, Double, Triple, and Quadruple ADT Knock-out Lines-T-DNA insertion lines for all six ADT genes in Arabidopsis (supplemental Table S1) were obtained from either the Salk Institute Genomic Analysis Laboratory (18) or the Institut National de la Recherche Agronomique (19). For each T-DNA insertion line, DNA was extracted from leaves of individual plants using the RedExtract kit (Sigma), and these samples were then individually used as a template for two PCRs with different primer sets. For Salk Institute Genomic Analysis Laboratory lines, gene-specific left and right primers LP ϩ RP, respectively, were used to amplify WTspecific PCR products, and left border primer site "c1" (LBc1) ϩ RP were used to amplify T-DNA-specific PCR products (supplemental Table S1; LBc1, 5Ј-CACAATCCCACTATC-CTTCGC-3Ј). For the Institut National de la Recherche Agronomique line, the T-DNA-specific primer FLAG-LB (5Ј-GACGTAACATAAGGGACTGACC-3Ј) was substituted for LBc1. Homozygous T-DNA insertion lines were identified as those having T-DNA-specific PCR products only, and these were sequenced to confirm the presence and the specific site of each T-DNA insertion. ADT KO lines were confirmed using RT-PCR with primers designed to the 3Ј-end of each ADT mRNA transcript. Confirmed single KO lines were then crossed together to generate double heterozygous ADT KOs in all combinations, and double homozygous lines were identified in the subsequent generation using the same PCR screening approach described above. The same strategy was used to create triple and quadruple KO lines using double and triple KO parental lines, respectively. Each double, triple, and quadruple KO line was independently confirmed using the PCR strategy described above.
Complementation of adt5 Line-Complementation was carried out by expressing ADT5 under the control of its native promoter in the adt5 KO line. First, a 1888-bp fragment upstream of ADT5 was amplified using promoter-specific primers (see supplemental Table S2). The ADT5 coding gene was then cloned into the pENTR TM /D-TOPO vector to generate an entry construct. After full sequence verification, the entry vector construct was subcloned into a pK2GW7 binary vector (20). The cauliflower mosaic virus 35S promoter of the vector construct was next swapped for the ADT5 promoter following established procedures (21). The confirmed construct was transformed into Agrobacterium and used to transform the adt5 KO line using standard protocols (22).
Arabidopsis Growth and Harvest Conditions-All confirmed homozygous KO, complementation, and WT lines were grown in soil with four plants per pot in Washington State University greenhouses (16-h days, 27-28°C; 8-h nights, 24 -26°C; 200ppm nitrogen-based fertilizer added 5 days a week). For lignin analyses, the main stems of at least 48 plants were harvested weekly from after initial stem emergence up to maturity (ϳ3.5-10 weeks). The weights and lengths of 20 inflorescence stems from each line were measured, and stems were then subsequently cut into 0.5-1-cm-long pieces, lyophilized, and stored at room temperature prior to lignin analyses. For histochemical staining, two main stems for each ADT KO and WT line were harvested at 7 weeks.
Real Time RT-PCR Analysis of ADT KO Lines-Stem tissues for WT and selected ADT KO lines were harvested 5 weeks after planting, flash frozen in liquid N 2 , and stored at Ϫ80°C until use. Frozen tissue was ground using a mortar and pestle, and ϳ90 -110 mg was transferred to a 1.5-ml microcentrifuge tube. Total RNA was extracted using the Spectrum TM Plant Total RNA Extraction kit (Sigma-Aldrich). RNA quantity and quality were assessed using a Nanodrop 2000c spectrometer (Thermo Fisher Scientific Inc.), and mRNA (1 g) was reverse transcribed to cDNA using Superscript III (Invitrogen).
Gene-specific primers for each ADT isoform and housekeeping gene TIP41-like (AT4G34270; Ref. 23) were designed using Primer Premier 6.10 software (Premier Biosoft International) (see supplemental Table S2). The SYBR Green Real Time RT-PCR kit (Invitrogen) was used for real time RT-PCRs using 0.05 g of cDNA and 62.5 pmol of primers for each reaction. Triplicate reactions were run on an Mx 3505P Real Time Thermocycler (Stratagene), and data were analyzed with Mx Pro QPCR software (Stratagene).
Arogenate Dehydratase Assays-The following method from Jung et al. (1), modified by Maeda et al. (17), was applied for assaying ADT activity in Arabidopsis stems. Approximately 20 g of stem tissue was harvested, immediately ground in liquid N 2 , and extracted with 30 ml of buffer (20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 35 mg of leupeptin, and 35 ml of plant cell and tissue extract protease inhibitor mixture (Sigma-Aldrich)). The crude lysate was then subjected to an ammonium sulfate precipitation, and both the 20 -40 and 40 -80% fractions were collected. Each fraction was desalted with a PD-10 column (GE Healthcare) and then concentrated to ϳ500 l using an Amicon Ultra-4 Centrifugal Filter (Millipore). An aliquot (5 l) of each protein extract (containing 30 and 370 g for the 20 -40% and 40 -80% fractions, respectively) was added to the total volume of the 12-l reaction mixture containing 250 M arogenate and 20 mM Tris-HCl, pH 8.0. After incubation at 37°C for 15 min, the reaction was stopped by addition of 10 l of MeOH with 2 l of 10 mM alanine added as an internal standard. The assay mixtures were vortexed and centrifuged, half of the sample was derivatized with N-methyl-N-(tert-butyldimethylsilyl)trifluoroacetamide and analyzed by GC/MS, and the other half was derivatized using the Pico-tag system (Waters) and analyzed by HPLC as described previously (3,5). No enzymatic phenylalanine formation was detected under the conditions used.
Free Amino Acid Analysis-Total free amino acid pools were extracted from 5-week-old WT and ADT KO stems using methanol:chloroform:water (12:5:3) as described previously (24). Amino acids were derivatized using the AccQ⅐Tag TM Ultra Derivatization kit (Waters) and analyzed by ultraperformance liquid chromatography (Waters). Phe and Tyr levels in WT were ϳ18 and ϳ12 pmol/mg dry weight, respectively, whereas those of the ADT KO lines ranged from 12 to 21 and 5 to 11 pmol/mg dry weight, respectively, suggesting that no massive changes occurred in Phe/Tyr levels between WT and ADT KO lines.
Histochemical Staining-Histochemical staining and imaging of fresh hand-cut sections taken near the base of 7-week-old stems (Stage 3) (25) of both mutants and WT lines were carried out as described previously by Patten et al. (26). For detection of presumed guaiacyl (G) lignin, stem cross-sections were placed in a phloroglucinol-HCl (0.1%, w/v) solution for 30 min (27,28), transferred to a glass slide, and observed under the differential interference contrast setting using an Olympus System Microscope, Model BHT (Olympus Optical Co., Ltd., Tokyo, Japan). The Mäule reaction was used for presumed syringyl (S) lignin component detection. Hand-cut cross-sections of stems were treated for 10 min with KMnO 4 (0.5%, w/v; filtered through a 0.45-m MillexHV filter) and rinsed with distilled H 2 O (29). Samples were then treated with HCl (0.1%, v/v) for 5 min, rinsed, mounted in concentrated NH 4 OH, and observed as described above.
Laser Microscope Dissection of Vascular Bundles and Interfascicular Regions-Laser microscope dissection was used to separate the vascular bundle (VB) and interfascicular fiber (IF) regions from 7-week-old stems (Stage 3) for WT and adt1/4/5 lines as described previously (26). Briefly, sections were made using a Cryocut 1800 microtome (Leica Microsystems) and laser-microdissected using a P.A.L.M. Microbeam System (P.A.L.M. Microlaser Technologies) to isolate the VB and IF regions. Approximately 5000 samples from each region were individually collected for adt1/4/5 and WT, respectively. Each dissected sample was combined, then individually transferred to glass vials and extracted twice for 12 h with H 2 O:acetone (3:7, v/v) at RT, and dried in vacuo in preparation for pyrolysis GC/MS analyses.

Generation and Gross Phenotype Comparisons of ADT Knock-out Lines-T-DNA insertion lines for each ADT isoenzyme used in this study (supplemental
The Arabidopsis growth/development time frame involves three distinct stages (25) as described previously in the comprehensive study of the lignin pathway-altered ref8, fah 1-2, and C4H::F5H Arabidopsis lines (26). During this time frame, phenotypic assessment of the above ADT mutant lines at each weekly harvesting point was carried out to identify single or double KO lines with potential reductions in lignin contents, alterations in stem lengths and weights, and/or the presence of prostrate phenotypes. Of the single KOs, only adt5 had a slight decrease in stem weights and lengths (to ϳ90% of WT levels; Fig. 2, A and B) and a partially prostrate (Fig. 3E) phenotype that was phenotypically distinct from WT (Fig. 3A). The remaining single KO lines, adt1, adt3, and adt4, each had stem lengths and weights relatively similar to WT over the course of growth and development (Fig. 2, A and B) and also did not appear visually to be phenotypically different from WT (Fig. 3, B-D). Additionally, one double KO, adt4/5, initially stood out as having stem weights and lengths reduced to ϳ82% of WT levels (Fig. 2, A and B) and a rather prostrate phenotype (Fig. 3F) relative to WT (Fig. 3A). Based on these observations, additional crosses were made to obtain two triple KOs that shared the same combination of adt4 and adt5, i.e. adt1/4/5 and adt3/4/5. Both triple KOs also had obvious prostrate phenotypes (Fig. 3, G and H), and adt1/4/5 had stem weights and lengths similar to adt4/5, corresponding to ϳ85% of WT levels, whereas adt3/4/5 displayed further reductions in both stem weight and length, corresponding to ϳ71% of WT levels (Fig. 2, A and B). Interestingly, the ADT isoenzymes ADT3, ADT4, and ADT5, which are in the same phylogenetic cluster, "subgroup III" (3), appeared to cause the greatest phenotypic effect, and therefore it was instructive to test whether the fourth ADT in that subgroup (ADT6) caused any further changes. Thus, we also obtained a quadruple KO, adt3/4/5/6, in which all members of subgroup III (3) were disrupted. This line had further reductions in stem weights and lengths (Fig. 2, A and B), corresponding to ϳ67% of WT levels, and displayed a prostrate phenotype (Fig. 3I).
ADT Expression Levels in Selected ADT Knock-out Lines-Real time RT-PCR was used to further verify that ADT transcripts were either not detectable or greatly reduced in each corresponding ADT KO line and to identify any potential increased expression in non-targeted ADT transcripts (supplemental Fig. S3). The relative mRNA expression levels of all six ADTs were also measured in WT and each of the eight selected KO lines described above. For each of these lines, ADT transcripts for each corresponding knocked out gene were absent except for ADT1 and ADT6, which were reduced to Ͻ25% of WT levels (supplemental Fig. S3). It is possible that the latter two genes are able to produce a small amount of transcript despite being homozygous for T-DNA insertions because the insertions in ADT1 and ADT6 are present in the first intron and the 5Ј-UTR region, respectively, and may allow for a small amount of correctly processed mRNA to be transcribed from these genes. Similar findings have been described elsewhere for T-DNA insertion mutants (40).
Consistent with the findings by Rippert et al. (41), ADT4 and ADT5 had the highest expression levels in WT stems, and the remaining ADT genes also were expressed in stems but at lower levels (supplemental Fig. S3). Expression of these two genes was increased slightly but perhaps significantly compared with WT in certain lines. ADT4 expression was increased in the adt3 and adt5 KO lines, whereas ADT5 was increased in the adt3 KO, and no change in expression was observed in the adt1 and adt4 KO lines. Small increases in expression were also noted for ADT1, ADT2, and ADT3 for certain ADT KO lines (supplemental Fig. S3), whereas there were no apparent increases observed for ADT6 in any lines.
As a decrease in expression levels does not (necessarily) translate directly to a decrease in overall ADT activity, we also attempted to measure ADT activity in Arabidopsis stems. Although ADT activity has been detected in certain species (1,2), it has never been described in Arabidopsis. Nevertheless, crude enzyme extracts were prepared using WT stem tissue using the modified protocol described by Maeda et al. (17).
Both arogenate and prephenate were tested as substrates; however, in both cases, no ADT or prephenate dehydratase activity could be observed (data not shown). Because no activity could be detected in WT stems, the KO lines were not assayed as they were expected to have even lower levels of ADT activity. A small decrease in the levels of free Phe was observed, however, in the stems of each KO line as compared with WT with the exception of adt1 and adt3. These changes were observable in 5-week-old stem tissue, but by 7 weeks (when lignification was complete) they had returned to WT levels (data not shown). 4 Histochemical Analyses of Single and Multiple ADT Knock-outs-Qualitative histochemical analyses of 7-week-old basal stem sections ( Fig. 4 and supplemental Fig. S4) for staining of both G and S lignin-containing phenotypes were also carried out using phloroglucinol-HCl (for G) (26 -28) and Mäule (for S) (26,29) reagents. G lignin component staining, used frequently for detection of coniferyl alcohol-derived moieties (see Fig. 5 for structures), was nearly identical throughout the VB and IF regions of WT, adt1, adt3, and adt4 lines (supplemental Fig. S4, A-D), whereas presumed G moieties were apparently less readily detectable in IF regions of adt5, adt4/5, adt1/4/5, adt3/4/5, and adt3/4/5/6 ( Fig. 4, B-F) as gauged by the decreased levels of red-pink staining in these cross-sections as compared with WT (Fig. 4A). Also apparent in the phloroglucinol-HCl-treated sections were irregularly shaped and partially collapsed cell walls in the metaxylem within the VBs of adt1/4/5 and adt3/4/5 lines (Fig. 4, H and I) in contrast to the WT line where this deformation was not evident (Fig. 4G). There was little if any visible difference in presumed S (sinapyl alcohol-derived) moieties for any of the KO lines using the Mäule reagent (Fig. 4, K-O and supplemental Fig. S4, F-H) as compared with WT ( Fig. 4J and supplemental Fig. S4E).

Estimated Lignin Contents/Compositions of Single and Multiple ADT Knock-outs-
The potential effects on lignification in the various ADT KO lines relative to the WT line were next assessed with stems from each plant line harvested weekly. Extractive-free stem CWR were thus subjected to "AcBr lignin" (32-34) and thioacidolysis (34 -36) analyses to estimate gross lignin amounts and cleavable lignin monomeric contents/compositions. Note, however, that AcBr lignin contents were provisionally corrected for their H/G/S compositions using extinction coefficients for individual H, G, and S lignin-enriched preparations as described elsewhere (32). The thioacidolysis treatment by contrast releases monomeric H-, G-, and S-derived constituents via cleavage of presumed 8-O-4Ј interunit linkages in the lignin biopolymer(s) to afford the corresponding monomeric thioethylated derivatives, respectively (Fig. 5E).
Analysis of AcBr Lignin Deposition Patterns-Over the growth and development of the Arabidopsis lines being studied, the estimated AcBr lignin contents in the WT line plateaued at ϳ23.6% CWR at full maturation (i.e. average of 7-10 weeks) ( Fig. 5A and supplemental Table S3). The adt KO lines identified in our phenotype comparisons were also analyzed, and their estimated lignin contents were compared with WT. The adt1 and adt3 lines reached levels comparable with WT of ϳ23.8 and ϳ23.0% CWR, respectively, whereas the adt4 line putatively had a slightly higher lignin level of ϳ26.0% CWR. By contrast, the adt5 line had a marked reduction in the estimated AcBr lignin content to ϳ18.9% CWR at maturity, whereas an even more striking reduction (i.e. to ϳ14.4% CWR) was noted with the double mutant adt4/5. Further decreases in AcBr lignin levels were documented with the triple mutants; at maturity, adt1/4/5 and adt3/4/5 had roughly equivalent AcBr lignin levels of ϳ12.2 and 11.9% CWR, respectively, whereas adt3/4/ 5/6 was reduced to ϳ7.5% CWR. Additionally, complementation of adt5, however, resulted in a full restoration of lignin contents/compositions to WT levels; i.e. the adt5 complemented line had estimated lignin contents of ϳ24.2% CWR. (Note, however, that the AcBr lignin methods can overestimate lignin amounts due to the presence of non-lignin UV-absorbing components (see Anterola and Lewis (13).) Thioacidolysis Analyses and Lignin Monomer-derived Compositions and Contents-The thioacidolysis analysis results for the CWR of the WT and KO lines are depicted in Fig. 5, B-D (with actual values given in supplemental Table S3). In agreement with the estimated AcBr lignin analyses, thioacidolysis showed somewhat similar trends for the amounts of G ϩ S monomeric moieties (Fig. 5B) released from their lignins from the beginning of stem elongation until full maturity. In this regard, cleavage of the presumed 8-O-4Ј interunit linkages in WT released ϳ299 mol of G ϩ S monomers/g of CWR at maturation (average of 7-10 weeks), whereas adt4 was slightly higher at ϳ319 mol/g of CWR. These values correspond to approximately 25% by weight of the estimated AcBr lignin content. However, the other single KOs, adt1, adt3, and adt5, were slightly reduced compared with WT with ϳ257, 244, and 250 mol of G ϩ S monomers released/g of CWR, respectively. Larger reductions were also observed for the double KO adt4/5 (ϳ147 mol/g of CWR), triple KOs adt1/4/5 and adt3/4/5 (ϳ116 and ϳ88 mol/g of CWR, respectively), and quadruple KO adt3/4/5/6 (ϳ95 mol/g of CWR). Their amounts ranged from ϳ27 to 16% by weight of the putative lignin present with such reductions in releasable monomers frequently observed when overall lignin contents are reduced.
However, the adt5 line was significantly decreased in G cleavable monomers (Fig. 5C) but apparently had slightly higher levels of S monomers compared with WT (Fig. 5D). Similarly, the double adt4/5 KO had even more pronounced reduc- tions in G monomer levels released with little effect on the S amounts compared with WT. However, in the corresponding triple and quadruple KOs, adt1/4/5, adt3/4/5, and adt3/4/5/6, there were further reductions in both G and S releasable monomers compared with the adt4/5 KO. Again, the adt5 line complemented with the native ADT5 gene resulted in restoration of G ϩ S levels (ϳ310 mol/g of CWR) and G:S ratios restored to WT levels.
Pyrolysis GC/MS Analyses-Next, it was instructive to compare the lignin-derived pyrolysis products released from S-enriched IF and G-enriched VB tissues using both WT and adt1/ 4/5 lines as successfully carried out previously on Arabidopsis ref8, fah 1-2, and C4H::F5H lines (26). First, samples of WT and adt1/4/5 stem CWR were individually subjected to pyrolysis GC/MS and analyzed. Products were identified by either retention time, co-elution with authentic standards, or mass spectro-  A, D, and E is shown in G, H, and I, respectively, with metaxylem (mx), protoxylem (px), and xylem fibers (xf) labeled in G. Fainter staining of G moieties was detected using phloroglucinol-HCl in the IF regions of adt5, adt4/5, adt1/4/5, adt3/4/5, and adt3/4/5/6 (C-F, respectively), whereas there was no apparent decrease in the VBs in any KO lines. However, increased magnification (G-I) indicated that metaxylem cell wall integrity is affected in triple KO lines (H and I) with numerous irregularly shaped/partially collapsed vessels (denoted by *) present in these lines. Scale bars, 50 m. scopic fragmentation data. Simple inspection of releasable lignin-derived pyrolysis products in these two GC/MS chromatograms indicated substantial reductions in G and S components (relative to H-derived moieties) in the adt1/4/5 line (Fig.  6B) in comparison with WT ( Fig. 6A) with the largest reduction in G-derived components. (That is, in the WT line, the pyrolysis products identified were H (peaks 1-3, 5, and 8), G (peaks 4, 6,  7, 10, 11, 14, 15, 17, 19, 24, and 29), and S (peaks 13, 20, and  23) where the G components were most abundant (Fig. 6A, Table 1, and supplemental Fig. S1).) By contrast, the pyrolysis products obtained for the adt1/4/5 CWR detected the presence of H (peaks 1-3 and 5), G (peaks 4, 6, 10, 11, 14, 15, and 19), and S (peaks 13, 20, 23, and 26) with a significant reduction in both G/S-derived products relative to H moieties (Fig. 6B, Table 1, and supplemental Fig. S1). These data thus provisionally agreed with thioacidolysis results, which also indicated a greater reduction in G lignin-derived monomers as compared with S monomeric moieties in the adt1/4/5 line.
Next, laser microdissection was used to excise approximately 5000 individual VB and IF sections from the WT and adt1/4/5 lines, and these were subjected to pyrolysis GC/MS. Analysis of the VB regions of adt1/4/5 (Fig. 6D, Table 1, and supplemental Fig. S1) resulted in identification of H (peaks 1-3 and 5), G (peaks 4, 6, 10, 11, 15, 17, and 19), and S (peaks 13, 20, and 23) pyrolysis fragments where both G and S moieties were significantly reduced relative to the H-derived components (the S moieties were barely detectable). By contrast, the analysis of the IF regions gave a chromatogram with H (peaks 1-3 and 5), G (peaks 4, 11, and 17), and S (peaks 13, 20, and 21) pyrolysis fragments (Fig. 6C, Table 1, and supplemental Fig. S1). In this case, however, the relative amounts of S components still remained high, and the most notable effect was on G component reductions. These results are thus consistent with histochemical staining using phloroglucinol that indicated that the reduction was greatest in the G lignin constituents of the IFs (Fig. 4D).
GUS Expression Patterns in Stems-Expression of ADT4 and ADT5 was visualized using the GUS expression system. Putative promoter regions for ADT4 and ADT5 were fused in-frame with the GUS gene and stably transformed into WT Arabidopsis (see "Experimental Procedures"). Staining patterns were observed in 4-week stem cross-sections. ADT4 and ADT5 were shown to have overlapping expression patterns, which were localized to the vascular cambium regions in the VBs (Fig.  7, A and B, respectively). Staining was not observed in the interfascicular cambium or other regions of the stem.

DISCUSSION
ADT Manipulations and Phenotypic Effects-The possible differential contribution of distinct ADTs to carbon flux into the lignin pathway was investigated given its branch point position between shikimate-chorismate and phenylpropanoid metabolism. It was confirmed by real time RT-PCR that ADT mRNA transcripts from each corresponding ADT KO line examined, with the exception of ADT2, were barely detectable or below the level of detection (supplemental Fig. S3). ADT2 was not successfully knocked out (see "Experimental Procedures"). Vastly reduced levels of ADT1 transcript were observed in adt1 and adt1/4/5, whereas a small amount of ADT6 transcript was found in adt3/4/5/6; in each case, however, their levels were significantly reduced compared with WT.
Although slight increases in non-targeted ADT transcripts were observed in certain ADT KO lines, there was no apparent trend for increased "compensation" in the double, triple, and quadruple KOs relative to the single KOs. This was clearly demonstrated in the adt3/4/5/6 KO line for which ADT3, ADT4, ADT5, and ADT6 transcript levels were decreased/abolished, but the remaining ADT1 and ADT2 transcript levels were apparently identical to those of WT. These results suggest that the observed phenotypes are a direct result of diminished ADT gene expression levels in the knocked out ADTs. Although we were unable to confirm that the decrease in ADT expression led to a reduction of ADT activity in planta, the decrease in ADT expression is fully consistent with the phenotypes observed for these plants.
Phe pool sizes were also slightly affected in adt4, adt5, adt4/5, adt1/4/5, adt3/4/5, and adt3/4/5/6 but only during the time of growth and development in which lignification was occurring. Later in development when lignification slowed, there was no detectable difference compared with WT. 4 Depending upon the plant line generated, there were relatively significant but distinct reductions in stem lengths (to a range of approximately 67-85% of WT; see Fig. 2B) when adt5 was "knocked out" in combination with adt4, adt1/4, adt3/4, and adt3/4/6, respectively. This was also noted in dry weight stem tissue determinations, which showed reductions to approximately 86 and 76% of WT levels for adt1/4/5 and adt3/ 4/5, respectively, relative to WT. Such relatively small effects on biomass production, however, contrast with numerous other studies on monolignol pathway step modulations, which frequently result in, for example, extremely dwarfed plant lines (with greatly reduced biomass) and significantly compromised vasculature (for a discussion see Davin et al. (10)).

Comparison of Histochemical and Pyrolysis GC/MS Analyses of Vascular Bundles and Interfascicular
Fibers-Guaiacyl and syringyl entities are often qualitatively detected using either the histochemical staining reagent phloroglucinol-HCl for G components in lignified tissues or the Mäule reagent for S-derived lignin moieties. Histochemical staining of the various lines generated herein (adt1, adt3, adt4, adt5, adt4/5, adt1/4/5, adt3/ 4/5, and adt3/4/5/6) upon comparison with WT thus provided some useful insights into the limitations of these staining protocols. Specifically, for the double, triple, and quadruple  Table 1, and for structures, see supplemental Fig. S1.) mutants containing the adt5 knock-out, there was essentially no indication that G moieties were present in IF regions, whereas the VBs all stained positively even though there was considerable distortion/weakening of the metaxylem cell walls (Fig. 4, B-F). By contrast, for each of the lines examined using the Mäule reagent, S staining was notably detected in IF cell walls as well as in fiber-containing cell walls within the VBs (Fig.  4, K-O). These qualitative data thus suggested the absence of G moieties in the IF regions.
Pyrolysis GC/MS of stem cross-sections (CWR) of both WT and adt1/4/5 was also carried out, resulting in facile detection of various H-, G-, and S-derived monomers (Fig. 6, A and B, and Table 1). By comparison with the H-derived constituents, however, the levels of both G-and S-derived moieties were much reduced overall in the adt1/4/5 line with the highest reduction in G-derived moieties (Fig. 6B). Next, when subjected to pyrolysis GC/MS, laser microscope-dissected IF and VB sections of both WT and adt1/4/5 lines provided considerable insight into the type of lignins present in these distinct anatomical regions. For the IF region of adt1/4/5, it was evident (relative to H-derived monomers) that the amounts of G-derived pyrolysis products were substantially diminished relative to S-derived moieties, which were comparatively more readily detectable (Fig. 6C). Thus, the IF regions still contained an S-enriched lignin even though overall lignin amounts were substantially reduced. Failure to detect G moieties in IF regions of some of these mutants by histochemical staining thus demonstrates a serious limitation in this qualitative staining protocol when G levels are low (relative to S-and H-derived moieties).
Laser microscope dissection/pyrolysis GC/MS of WT IF regions resulted in facile detection of H (peaks 1 and 3), G (peaks 4, 6, 7, 10, 11, 14, 15, 17-19, and 24) and S (13, 20-23, 27, and 28) lignin-derived moieties with the G and S moieties predominating relative to the H-derived component (data not shown; see Patten et al. (26)). By contrast, analysis of the IF regions of the adt1/4/5 indicated substantial reductions in G component release with both S and H component detection relatively comparable (Fig. 6C). Pyrolysis GC/MS of WT VB regions indicated the presence of presumed H (peaks 1 and 3), G (peaks 4, 6, 11, 14, 15, 17-19, and 24), and S (peaks 13, 20-23, and 27) lignin-derived moieties (data not shown; see Patten et al. (26)). In the adt1/4/5 VBs (Fig. 6D), it was evident that in comparison with the H pyrolysis products the amounts of both G and S moieties were also significantly reduced with the S components barely detectable. However, the G moieties still predominated. The pyrolysis data are thus in general agreement with the bulk studies of the AcBr lignin contents and ADT Manipulations and Effect on Lignification-Total lignin monomeric compositions and contents were also systematically studied over a period of 3-10 weeks, reflecting the three phases of Arabidopsis growth/development until maturation and senescence (25,26). As indicated above, massive yet differential reductions in lignin contents and compositions were observed via manipulation of this plastid-localized enzyme family.
We next plotted correlations between estimated AcBr lignin contents versus thioacidolysis-released G ϩ S monomer levels in the different lines generated at different growth stages (sampled weekly) until the plant stems matured and ultimately senesced (Fig. 8A). Although there was considerable experimental variation in the samples tested, all of the lines examined essentially showed linear increases overall in releasable G ϩ S monomer amounts relative to estimated lignin contents as observed previously (10,26). At maturation, however, the adt4/5 (green OE), adt1/4/5 (orange •), adt3/4/5 (blue •), and adt3/4/5/6 (purple •) lines had G ϩ S monomer release/AcBr lignin levels that did not surpass the 4-week-old levels for the WT line (pink f). As documented elsewhere (26), there was also an initial ϳ5% AcBr lignin deposition in which essentially no G ϩ S moieties are released; this early stage AcBr lignin, presumably H-derived (from p-coumaryl alcohol), can also include other non-lignin components as discussed in Jourdes et al. (32) and Davin et al. (10). Most importantly, however, the essentially linear correlations of G ϩ S monomer release versus estimated lignin contents are considered indicative of a biochemical process in place leading to (near) conservation of the 8-O-4Ј interunit linkage frequency in these various lignins. This again reflects control exercised over lignin macromolecular assembly (10). FIGURE 8. Comparison of thioacidolysis-determined G and/or S lignin-derived monomer contents versus total AcBr lignin contents for single KO lines adt1, adt3, adt4, and adt5 as well as multiple KO lines adt4/5, adt1/4/5, adt3/4/5, and adt3/4/5/6. Total G ϩ S-derived thioacidolysis monomeric degradation products are compared with total AcBr lignin for all single and multiple KO lines (A). G-derived thioacidolysis degradation products are compared with total AcBr lignin for single (B) and multiple (C) KO lines. S-derived thioacidolysis degradation products are compared with total AcBr lignin for single (D) and multiple (E) KO lines. Error bars represent means Ϯ S.D. (n ϭ3).
It was instructive to also compare G and S monomer release from the different lines generated. As discussed earlier above ("Results"), the adt5 KO (burgundy ᭜) had the largest (of any single adt mutant) reductions in lignin contents and thioacidolytic monomer release (Fig. 8, B and D) relative to WT. However, the effect on reductions in monomer release essentially only impacted G monomer levels (approximately 74% of WT; Fig. 8B) but not the levels of the S constituents, which (if anything) were slightly increased (Fig. 8D). These data thus provisionally suggested that at a minimum essentially the same amount of cleavable S lignin-derived monomers was generated in the adt5 line as in WT. The overall amounts of G and S moieties released from adt1 (green ᭜), adt3 (light brown ᭜), and adt4 (cyan ᭜) lines were also similar to that of WT at maturity albeit with a slight increase in levels of G moieties for adt4 (Fig. 8, B and D). In the latter case, there also appeared to be a small increase in total AcBr lignin contents (approximately 10% more relative to WT). Given the large reduction in G monomer levels but not in S levels in adt5, these data provisionally suggested that the primary effect was on the lignin-forming biochemical machinery in the G-enriched VBs rather than in the various regions containing fiber cells. This, however, provides no insight into where these monomers are spatially located.
Examination of the double, triple, and quadruple KOs (Fig. 8,  C and E), all of which contained the adt5 KO, was also very informative as this added and extended to the observations made above. As indicated earlier, the first of these, adt4/5 (green OE), displayed a quite pronounced prostrate phenotype (Fig. 3F) and at maturity had AcBr lignin contents estimated to be ϳ61% of WT (Fig. 5A). However, at maturity, there was again essentially no difference in the amount of releasable S monomers relative to WT (Fig. 8E), whereas by contrast, the G monomers released were reduced to approximately 35% of WT levels (Fig. 8C). These data were thus again consistent with the primary target for lignin and G monomer reduction being within the VB region.
For the triple (adt1/4/5 and adt3/4/5) and quadruple (adt3/ 4/5/6) knock-outs, the releasable G monomer levels were reduced further down to approximately 29, 20, and 21% relative to WT at maturity (Fig. 8C), whereas the releasable S monomers were now also reduced to approximately 72, 58, and 67% of WT amounts, respectively (Fig. 8E). Thus, with the triple and quadruple knock-outs, there was now clearly also an effect on S monomer amounts and hence on S lignin deposition in the fiber cells. Overall, the lignin contents in these lines were also significantly reduced down to ϳ52, 50, and 32% of WT levels, respectively; these represent some of the largest reductions in lignin levels ever reported in genetic manipulations.
Taken together, these data again point to different modulation of ADT-regulated carbon flux into distinct anatomical regions (VBs versus fiber cells). As a result, it was not unexpected that the G:S ratios in the lignins so obtained also changed markedly with different KO lines from ϳ3:1 (WT) to ϳ1:1 (adt3/4/5/6). This was because these cell types still contained the biochemical machinery to predominantly generate G-and S-derived monomers, respectively, i.e. thereby generating different amounts of G and S lignins.
In this respect, the GUS expression patterns of both ADT5 and ADT4 indicated that they were localized to the vascular cambium adjacent to the VBs (Fig. 7), whereas the others were localized as relatively faint bands throughout the cambial regions (data not shown). Thus, these expression data are also consistent with the analyses of the VBs and the substantial decrease in G levels in the VBs.
Such differential effects on lignin deposition are striking, however, given the different subcellular localizations of ADTs in chloroplasts/plastids for Phe formation versus monolignol pathway enzymes that are cytosol/membrane-localized. This is the first time, however, that a plastid/chloroplast-localized upstream step involving the shikimate-chorismate pathway has been shown to have specific isoenzymes dedicated to profoundly and differentially directing carbon flux into Phe for lignin biosynthesis (42,43).
Importantly, the effect of these manipulations was that carbon allocation to the pathway was simply significantly reduced. The resulting phenotypes were, however, substantially weakened from a structural vasculature viewpoint due to their decreased lignin levels relative to WT levels. These data thus again suggest that only relatively modest lignin reductions can occur to maintain physiological integrity and that loss of physiological function occurs when lignin levels are significantly decreased. This in turn demonstrates the limitations in plasticity of physiological responses/structural properties when lignins are not fully deposited.
Toward Identifying Physiological Roles of ADT Isoenzymes and ADT Phylogeny-Our earlier phylogenetic tree comparisons (3) also indicated that ADT3 and ADT6 are in the same cluster (subgroup III) as ADT4 and ADT5, whereas ADT1 and ADT2 are in subgroups I and II, respectively. All six Arabidopsis ADT genes, however, were found to be expressed in stems, leaves, roots, flowers, siliques, and seeds (3,41). Of these, ADT2 was the most highly expressed in leaves and seeds, and it was also suggested earlier to have a "housekeeping" role in Phe biosynthesis (41). By contrast, ADT4 and ADT5 were more highly expressed in stems and roots, whereas the remaining isoenzymes had generally much lower levels of expression (supplemental Fig. S3) (41). Co-expression of ADT genes with other Arabidopsis genes was also explored using the Botany Array Resource Expression Angler (44). Each of the six ADT genes was used as bait, and this analysis showed evidence for differential transcriptional regulation of the ADT genes. ADT3, ADT4, ADT5, and ADT6 were provisionally co-regulated with numerous shikimate, phenylpropanoid, and aromatic amino acid biosynthesis genes (supplemental Table S4A). Of particular significance in light of the findings herein, ADT4 and ADT5 are provisionally co-expressed with at least one gene of each step in the monolignol pathway between Phe and the three monolignols, p-coumaryl, coniferyl, and sinapyl alcohols with the exception of F5H (supplemental Table S4A). By contrast, ADT1 and ADT2 are mainly associated with genes involved in basic cellular functions, such as transcription, translation, cell division, and nucleic or amino acid biosynthesis (supplemental Table S4B).
Together, transcriptional expression patterns and co-expression data strongly support the finding that ADT4 and