Production of muconic acid in plants

Muconic acid (MA) is a dicarboxylic acid used for the production of industrially relevant chemicals such as adipic acid, terephthalic acid, and caprolactam. Because the synthesis of these polymer precursors generates toxic intermediates by utilizing petroleum-derived chemicals and corrosive catalysts, the development of al- ternative strategies for the bio-based production of MA has garnered signi ﬁ cant interest. Plants produce organic carbon skeletons by harvesting carbon dioxide and energy from the sun, and therefore represent advantageous hosts for engineered metabolic pathways towards the manufacturing of chemicals. In this work, we engineered Arabidopsis to demonstrate that plants can serve as green factories for the bio-manufacturing of MA. In parti- cular, dual expression of plastid-targeted bacterial salicylate hydroxylase (NahG) and catechol 1,2-dioxygenase (CatA) resulted in the conversion of the endogenous salicylic acid (SA) pool into MA via catechol. Sequential increase of SA derived from the shikimate pathway was achieved by expressing plastid-targeted versions of bacterial salicylate synthase (Irp9) and feedback-resistant 3-deoxy-D - arabino-heptulosonate synthase (AroG). Introducing this SA over-producing strategy into engineered plants that co-express NahG and CatA resulted in a 50-fold increase in MA titers. Considering that MA was easily recovered from senesced plant biomass after harvest, we envision the phytoproduction of MA as a bene ﬁ cial option to add value to bioenergy crops.


Introduction
Muconic acid (MA) is a platform chemical that serves as a precursor for the synthesis of products such as adipic acid, terephthalic acid, and caprolactam which are widely used in the nylon and thermoplastic polymer industries. Current processes for the manufacturing of MA or its derivatives mainly rely on non-renewable petroleum-based chemicals. Such processes are not sustainable and eco-friendly since they require a high energy input and yield large quantities of toxic by-products (Xie et al., 2014).
As an alternative, the biological production of MA using engineered microorganisms and inexpensive carbohydrate feedstocks has received increasing attention over the past 20 years (Xie et al., 2014). Most of the established biological routes consist in the production catechol and its subsequent conversion into MA by ring-cleaving catechol 1,2-dioxygenase (Vaillancourt et al., 2006). All these routes exploit the intrinsic shikimate pathway for the biosynthesis of catechol precursors such as protocatechuate, anthranilate, salicylic acid (SA), and 2,3-dihydroxybenzoic acid (Kruyer and Peralta-Yahya, 2017). Recently, MA biosynthetic pathways have been implemented in various microbial strains capable of growing in the presence of aromatics derived from lignocellulosic biomass. These include engineered strains of Escherichia coli (Sonoki et al., 2014;Wu et al., 2017), Amycolatopsis sp. (Barton et al., 2017), Pseudomonas sp. (Vardon et al., 2015;Johnson et al., 2016Johnson et al., , 2017Sonoki et al., 2017), and Sphingobium sp. (Sonoki et al., 2017).
In addition to microbial synthesis, the metabolic engineering of photosynthetic organisms like plants also provides a sustainable approach for the production of valuable metabolites and materials (Börnke and Broer, 2010;Farré et al., 2014). These chemicals, when produced in engineered bioenergy and oilseed crops, represent valueadded renewable co-products on top of the lignocellulose and seed oil used to generate energy (Snell et al., 2015). Because plants are autotrophs able to capture solar energy, they represent an attractive chassis for implementing de novo metabolic pathways for cost-effective production of important chemicals (Yuan and Grotewold, 2015).
In plants, the shikimate pathway is confined to plastids and provides the precursors for the synthesis of aromatic amino acids and derived metabolites, vitamins K 1 and B 9 , and SA (Maeda and Dudareva, 2012). We therefore used Arabidopsis as a model system to investigate a novel bio-based approach for the phytoproduction of MA. In particular, the SA pool derived from chorismate via the shikimate pathway was converted to catechol and MA by dual expression and plastid-targeting of bacterial salicylate hydroxylase (NahG) and catechol 1,2-dioxygenase (CatA) (Fig. 1A). Additional supply of SA to the MA pathway was achieved via the expression of bacterial salicylate synthase Irp9 and feedback-resistant 3-deoxy-D-arabino-heptulosonate (DAHP) synthase (AroG*), which resulted in a~50-fold increase of MA content. Importantly, MA was recovered from the biomass of senesced mature plants which highlights its stability and suitability for storage when produced in target crops. Therefore, such engineered crops could represent high-potential feedstocks for existing MA microbial production platforms towards sustainable development of bio-based MA.

RNA extraction and qRT-PCR analysis
Total RNA was extracted from stems of 5-week-old wild type and T3 homozygous transgenic lines (pools of three plants per line,~100 mg) using the Plant RNeasy extraction kit (Qiagen, Valencia, CA), and treated with DNase (Qiagen, Valencia, CA) to remove genomic DNA contamination. First-strand cDNAs were synthesized from 2 µg of total RNA using the SuperScript III First-Strand Synthesis SuperMix (Thermo Fisher Scientific, Waltham, MA) followed by qPCR analysis using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Hercules, CA) on a CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA) following the manufacturer's instruction. Oligonucleotide primers (Supplementary Table S1) were tested in annealing temperature gradients and 58°C was chosen as the annealing temperature. Melting curve analyses were performed after each run to ensure single amplicons were produced. The data were analyzed using the 2^(-ΔΔCt) method (Livak and Schmittgen, 2011).

Metabolite extraction
SA was extracted from developing stems using 80% (v/v) methanolwater at 70°C as previously described (Eudes et al., 2015). MA was extracted from stems of mature senesced plants ball-milled with a mixer mill MM 400 (Retsch, Newtown, PA). Ball-milled stem material (50 mg) was mixed with 1 mL of 80% (v/v) methanol-water and mixed (1400 rpm) for 15 min at 70°C. This extraction step was repeated twice. Extracts were pooled and cleared by centrifugation (5 min, 20,000×g), mixed with 1.5 mL of analytical grade water and filtered using Amicon Ultra centrifugal filters (3000 Da MW cutoff regenerated cellulose membrane; EMD Millipore, Billerica, MA) prior to LC-MS analysis.
Alternatively, MA was released from stems of mature senesced plants (line nahG-catA-1.2 × irp9-aroG 2.2) using dilute alkaline and dilute acid treatments. For dilute alkaline treatments, 10 mg of ballmilled biomass was soaked with 90 µL of either 1.2% or 0.62% (w/v) NaOH and heated at 100°C or 130°C for 30 min in an autoclave, respectively. For dilute acid treatment, biomass (10 mg) was soaked with 90 µL of 1.2% (w/v) H 2 SO 4 and heated at 120°C for 30 min in an autoclave. After cooling down and centrifugation, an aliquot of the hydrolyzates was mixed with 4 volumes of 80% (v/v) methanol-water and filtered using Amicon Ultra centrifugal filters prior to LC-MS analysis.

LC-MS metabolite analysis
SA and catechol were analyzed using liquid chromatography (LC), electrospray ionization (ESI), and time-of-flight (TOF) mass spectrometry (MS) as previously described (Haushalter et al., 2017). LC-ESI-TOF-MS analysis of muconic acid was carried out with a similar method except that the LC gradient elution was conducted as follows: linearly increased from 5% solvent B (0.1% formic acid in methanol) to 60.9% B in 4.3 min, increased from 60.9% B to 97.1% B in 1.3 min, decreased from 97.1% B to 5% B in 0.4 min, and held at 5% B for 2 min. The flow rate was held at 0.42 mL/min for 5.6 min, increased from 0.42 mL/min to 0.65 mL/min in 0.4 min, and held at 0.65 mL/min for 2 min. The total LC run time was 8 min. All metabolites were quantified via calibration curves of standard compounds (Sigma-Aldrich, St Louis, MO) for which the R 2 coefficients were ≥ 0.99. Cis,trans-MA was prepared from cis,cis-MA as previously described (Matthiesen et al., 2016).

Muconic acid (MA) production in plants expressing nahG and catA
The plastidial SA pool derived from the shikimate pathway was used as precursor for the biosynthesis of MA in Arabidopsis stems. To this end, we co-expressed plastid-targeted versions of the salicylate hydroxylase NahG and catechol 1,2-dioxygenase CatA from Pseudomonas putida for the sequential conversion of SA into catechol and MA. Although NahG has been shown previously to be functional in plants (Friedrich et al., 1995), the use of CatA for the synthesis of MA in plants has never been described. Since mature senesced Arabidopsis plants mainly consist of stem biomass, we selected two Arabidopsis promoters (pIRX8 and pC4H) which are both strongly active in stem tissues that develop secondary cell walls for synchronized expression of nahG and catA (Fig. 1B). Specifically, both promoters are known to be active in xylem vessels and interfascicular fibers: pIRX8 is the promoter of a glycosyltransferase family 8 (GT8) involved in the synthesis of secondary cell wall xylan (Persson et al., 2007), and pC4H is the promoter of the cytochrome P450 cinnamate 4-hydroxylase involved in the general phenylpropanoid pathway and lignin biosynthesis (Bell-Lelong et al., 1997). Five independent lines were selected, and expression of nahG and catA was confirmed by RT-qPCR using mRNA extracted from stems of 5-week-old homozygous plants at the T3 generation ( Fig. 2A). For these five lines, the content of MA extracted from stem biomass of senesced plants varied between 8.3 and 13.8 µg/g DW (Fig. 2B), which validates the dual nahG-catA expression strategy for the production of MA in plants. Although CatA converts catechol into cis,cis-MA, we detected a mixture of cis,cis-MA and cis,trans-MA in our plant extracts ( Supplementary Fig. S1), presumably due to the partial conversion of cis,cis-MA acid during the extraction procedure performed at 70°C. Therefore, MA titers reported in this work are the sum of cis,cis-MA acid and cis,trans-MA. Moreover, SA content measured in 5-week-old stems from the transgenic lines that co-express nahG and catA was reduced 5-to 10-fold compared to wild-type plants (Fig. 2C), suggesting that SA could be limiting for MA synthesis in transgenics.

Enhancement of SA content in stems by expressing bacterial salicylate synthase (Irp9) and feedback-insensitive DAHP synthase (AroG*)
In order to increase the content of SA in Arabidopsis stems, a plastid-targeted version of the salicylate synthase Irp9 from Yersinia enterocolitica was expressed using the promoter of the Arabidopsis secondary cell wall cellulose synthase gene IRX5 (CESA4) which is specifically active in stem vascular tissues (Eudes et al., 2012). Expression of irp9 has previously been shown to be effective to increase SA content without negative growth consequences in poplar (Xue et al., 2013). Seven independent lines were selected and expression of irp9 was confirmed by RT-qPCR using mRNA extracted from stems of 5week-old homozygous plants at the T3 generation (Fig. 3A). For six of these seven lines, the content of SA extracted from 5-week-old stems was increased significantly 1.8-to 4.4-fold compared to wild-type plants (Fig. 3B), which validates the important role of Irp9 to increase SA content in Arabidopsis stems. None of these transgenic lines showed any visible growth defects. Next, in order to further increase the carbon flux through the SA biosynthesis pathway in stem, we generated a new set of transgenic lines for co-expression of Irp9 with a mutant feedbackinsensitive DAHP synthase (AroG*) from E. coli. Expression of plastidtargeted AroG* in Arabidopsis was previously shown to increase the content of metabolites derived from the shikimate pathway such as aromatic amino acids and hydroxycinnamates (Tzin et al., 2012). The promoter pCCoAOMT of the caffeoyl-CoA O-methyltransferase gene involved in the monolignol pathway and lignin biosynthesis in Arabidopsis was chosen to express aroG* in stem vascular tissues that produce secondary cell walls (Do et al., 2007). Expression of irp9 and aroG* was verified in six independent lines using mRNA extracted from 15-cm  Wild type irp9-aroG-7.2 irp9-aroG-5.2 irp9-aroG-11.5 irp9-aroG-6.9 irp9-aroG-1.3 irp9-aroG-3.11 Salicylic acid (ng/g FW) Wild type irp9-aroG-7.2 irp9-aroG-5.2 irp9-aroG-11.5 irp9-aroG-6.9 irp9-aroG-1.3 irp9-aroG-3.11 Relative expression nd nd Fig. 3. Overproduction of SA in plants expressing Irp9 and feedback-resistant DAHP synthase (AroG*). (A) Detection by qRT-PCR of mRNA expression levels in stems from 5-week-old wild type and pA6-pIRX5::schl1-irp9 (irp9) transgenic lines. ACT2 was used as an internal control. Irp9 expression level was normalized to 1 in line irp9-6.1 and was calculated relative to this value in the other lines. Error bars represent the SD from technical duplicates. (B) SA content in stems of 5-week-old wild type and irp9 transgenic lines. Error bars represent the SE from four biological replicates (n = 4). Asterisks indicate significant differences from the wild type using the unpaired Student's t-test (*P < 0.005). (C) Detection by qRT-PCR of mRNA expression levels in 15-cm stems from wild type and pA6-pIRX5::schl1-irp9-pCCoAOMT::schl2-aroG (irp9-aroG) transgenic lines. ACT2 was used as an internal control. Irp9 and aroG expression levels were normalized to 1 in line irp9-aroG-6.9 and were calculated relative to these values in the other lines. Error bars represent the SD from technical duplicates. (D) SA content in 15-cm stems of wild type and irp9-aroG transgenic lines. Error bars represent the SE from four biological replicates (n = 4). Asterisks indicate significant differences from the wild type using the unpaired Student's t-test (*P < 0.001). stems of homozygous plants at the T3 generation (Fig. 3C). For these six lines, the content of SA extracted from 15-cm stems was increased significantly 11-to 65-fold compared to wild-type plants (Fig. 3D). The two lines showing the highest SA increase (irp9-aroG-1.3 and irp9-aroG-3.11) had an obvious dwarf phenotype compared to the other lines and wild-type plants (data not shown), a phenomenon previously observed in Arabidopsis mutants that overproduce SA (Sha, 2003). Therefore, co-expression of Irp9 and AroG* genes under secondary cell wall promoters represents an efficient strategy for over-producing SA in Arabidopsis stems.
3.3. MA production in plants expressing NahG, CatA, Irp9, and AroG* To assess the effect of SA over-accumulation on MA production, the line nahG-catA-1.2 was transformed with the constructs used for expression of irp9 and for co-expression of irp9 and aroG* (Fig. 1B). First, we selected several transgenic lines that show expression of irp9 in the nahG-catA-1.2 background (Fig. 4A). Muconic acid content in stems of senesced nahG-catA-1.2 × irp9 lines was increased 2-to 24-fold compared to the parental line (Fig. 4B). Second, several transgenic lines that show co-expression of irp9 and aroG* in the nahG-catA-1.2 genetic background were selected (Fig. 4C). In these lines, muconic acid content range between 483 and 637 µg/g DW, which represents a 39-to 51-fold increase compared to the values measured in the nahG-catA-1.2 parental line (Fig. 4D). These results demonstrate the positive effect of increasing the carbon flux through the SA route to enhance the production of MA in the proposed biosynthetic pathway. Furthermore, measurement of SA and catechol in two MA-producing lines revealed contents far below those of MA, suggesting that these two intermediates could limit MA biosynthesis (Supplementary Fig. S2).

Discussion and conclusions
We demonstrate in this work that bacterial catechol 1,2-dioxygenase (CatA) is functional in Arabidopsis plastids and thus can be exploited for the production of MA in plants. A biosynthetic route for catechol has been elegantly demonstrated in white campion flowers: it originates from phenylalanine and uses cinnamic acid, benzoic acid, and SA as intermediates (Akhtar and Pichersky, 2013). In this pathway, the biosynthetic genes involved in the steps for sequential conversion of benzoic acid into SA and catechol remain to be identified. Therefore, a plastid-targeted bacterial salicylate hydroxylase (NahG) was used in this work for the conversion of chorismate-derived SA into catechol. Gratifyingly, co-expression of NahG and CatA in plastids resulted in the   4). Asterisks indicate significant differences from the nahG-catA-1.2 line using the unpaired Student's t-test (*P < 0.05). (C) Detection by qRT-PCR of mRNA expression levels in stems from 5-week-old nahG-catA-1.2 and nahG-catA-1.2 × irp9-aroG transgenic lines. ACT2 was used as an internal control. Irp9 and aroG expression levels were normalized to 1 in line nahG-catA-1.2 × irp9-aroG 4.9 and were calculated relative to these values in the other lines. Error bars represent the SD from technical duplicates. (D) MA content in mature senesced dried stems of nahG-catA-1.2 and nahG-catA-1.2 × irp9-aroG transgenic lines. Error bars represent the SE from four biological replicates (n = 4). Asterisks indicate significant differences from the nahG-catA-1.2 line using the unpaired Student's t-test (*P < 0.001).
production of MA, and MA titers could be further enhanced by increasing the carbon flux through SA biosynthetic pathway. Although the synthesis of SA from chorismate involves known plastidial isochorismate synthases, the plant enzyme(s) involved in the conversion of isochorismate into SA remain to be identified (Widhalm and Dudareva, 2015). Similarly, the identity and sub-cellular localization(s) of the enzymes that contribute to SA biosynthesis from cinnamic acidan alternative SA pathway described in several plant specieshave not all been elucidated (Dempsey et al., 2011). Therefore, we targeted to plastids a characterized bacterial bi-functional SA synthase (isochorismate synthase / isochorismate pyruvate lyase, Irp9) that has been previously validated in planta to enhance SA synthesis in Arabidopsis. Moreover, expression of plastid-targeted bacterial feedback-insensitive AroG enhanced SA production when co-expressed with Irp9, which confirms previous observations in Arabidopsis about the positive effect of AroG expression on the accumulation of metabolites derived from the shikimate pathway (Tzin et al., 2012). Considering low SA titers measured in our transgenic Arabidopsis lines that produce MA ( Supplementary Fig. S2), additional engineering to enhance carbon flux through SA in these lines could improve MA titers. Whether the overexpression of other enzymes from the shikimate pathway (Fig. 1A, steps 2-7) would further increase the SA pool remains to be investigated. Similarly, the enolase responsible for the synthesis of phosphoenolpyruvate (Fig. 1A, step 1) could be targeted to increase SA content via the shikimate pathway since its overexpression was shown to drive carbon flux towards aromatic amino acid biosynthesis in tomato (Zhang et al., 2015). Specifically, the lack of correlation observed between aroG expression levels and SA (Fig. 3C-D) or MA (Fig. 4C-D) titers in our engineered plants suggests that one of the two aroG substrates could become limiting (Fig. 1A). Ultimately, since our MA biosynthetic route is confined to plastids, a trapping of SA inside plastids should be considered to avoid leak of this precursor, which could be achieved by downregulating known SA plastid exporters (Serrano et al., 2013). More generally, certain crops engineered for reduced biomass recalcitrance and enhanced digestibility overproduce SA (Gallego-Giraldo et al., 2011), making them ideal genetic backgrounds for the production of both fermentable sugars and value-added MA. Likewise, bioenergy Populus species (e.g., Salicaceae family) known to accumulate extremely high amounts of endogenous SA and SA-derived metabolites (up to 10% leaf dry weight) would represent adequate plant chassis for MA bioproduction (Lindroth and Hwang, 1996;Morse et al., 2007). In addition, we anticipate that bioenergy crops engineered for MA accumulation could serve as compatible feedstock for MA-producing microbial strains able to grow on lignin-enriched streams derived from lignocellulosic biomass (Vardon et al., 2015;Rodriguez et al., 2017). Because such streams are generated with high solids loadings (> 10% w/v), their enrichment with MA could be achieved using biomass containing MA. As an illustration, biomass containing 5% MA DW could potentially generate streams with 5 g/L MA (at 10% w/v biomass loading), a value similar to the best titers accomplished using engineered microbes and glucose as carbon source (Johnson et al., 2016). In this scenario, the MA titer we are reporting in Arabidopsis (0.64 mg/ g DW) would need to be improved by less than two orders of magnitude. More research will be needed to determine the optimal biomass pretreatment conditions for the release of MA. Our preliminary study indicates that optimal dilute alkaline biomass pretreatments used to generate lignin-rich fractions for downstream biological upgrading (Karp et al., 2014) can efficiently release MA from biomass of our engineered Arabidopsis plants ( Supplementary Fig. S3). On the other hand, lower amount of MA was recovered when biomass was treated with dilute acid, possibly due to the cyclization of MA into muconolactone under these conditions (Carraher et al., 2017).
As complementary approaches to the strategy presented in this work, the synthesis of catechol towards MA production could be achieved from alternate precursors such as anthranilate, protocatechuate, or 4-hydroxybenzoate as previously achieved in microorganisms (Kruyer and Peralta-Yahya, 2017). For this purpose, our preliminary work conducted in tobacco validated that both anthranilate 1,2-dioxygenase and protocatechuate decarboxylase can be functionally expressed in plastids for the synthesis of catechol from anthranilate and protocatechuate, respectively (Shih et al., 2016a). In addition, previous engineering strategies in Arabidopsis have demonstrated the overproduction of anthranilate, protocatechuate, and 4-hydroxybenzoate from chorismate (Eudes et al., 2016;Ishihara et al., 2006;Last and Fink, 1988). Since protocatechuate and 4-hydroxybenzoate synthesis can also be accomplished from the precursors 3dehydroshikimate and p-coumaroyl-CoA, respectively (Eudes et al., 2012(Eudes et al., , 2016Wu et al., 2017), production of high MA titers in plants could be envisioned by stacking branched biosynthetic routes that use diverse intermediates and products of the shikimate pathway ( Supplementary Fig. S4), and assisted by the use of synthetic promoters to optimize and synchronize the expression of multiple biosynthetic genes (Shih et al., 2016b).