Establishing a synthetic pathway for high-level production of 3-hydroxypropionic acid in Saccharomyces cerevisiae via β-alanine

for


Introduction
3-hydroxypropionic acid (3HP) is a platform chemical, which can be converted into acrylic acid, 1,3-propandiol, malonic acid, biodegradable polyesters, and other valuable chemicals.In 2011 the world annual production of acrylic acid was 5,000 kMT and the market size was USD 11.5 billion ("Acrylic Acid Market -Global Industry Analysis 2012 -2018", Research and Markets, July 2013).
Acrylic acid-derived products include superabsorbent polymers used in baby diapers and incontinence products, plastics, coatings, adhesives, elastomers, and paints.Currently, acrylic acid is primarily made by catalytic oxidation of propene.Establishment of 3HP production from glucose or other renewable carbon sources would provide a biosustainable alternative to acrylic acid production from fossil resources.
Several methods for biological production of 3HP have been described.Bacterial or fungal hosts can be used, where fungal hosts have the advantage of tolerance to low pH (3HP has a pKa value of 4.5), therefore enabling a more economical process without the need for neutralization during fermentation and acidification on product recovery.An attractive fungal host is baker's yeast, Saccharomyces cerevisiae, which is already used in several industrial processes for production of chemicals such as first and second generation bioethanol, succinate, isobutanol, resveratrol, amorphadiene and others [1][2][3] .Moreover S. cerevisiae can be engineered for tolerance to high concentrations of 3HP, thus making a robust production strain (KRK, Björn M. Hallström, THB, Nikolaus Sonnenschein, NBJ, SS, Scott J. Harrison, JM, MJH, Agnieszka S. Juncker, JF, JN, IB, accepted to Metabolic Engineering).
In biological systems, 3HP can be synthesized via at least four different intermediates: glycerol, lactate, malonyl-CoA or E-alanine 4 .Glycerol conversion to 3HP via glycerol dehydratase and aldehyde dehydrogenase by engineered Escherichia coli and native and engineered Klebsiella strains has been reported most extensively [5][6][7][8][9][10] .The application of the glycerol route in S. cervisiae is challenging because glycerol dehydratase is dependent on co-factor vitamin B12, which is not naturally synthesized by S. cerevisiae.Expression of active enzymes for conversion of lactate into 3HP has been described 11 .However, this route is thermodynamically unfavorable 12 and likely to result in a mixture of lactate and 3HP.Another route utilizes malonyl-CoA as precursor, which can be reduced to 3HP via combined action of malonyl-CoA reductase (malonate semialdehydeforming) and 3-hydroxypropionate dehydrogenase or 3-hydroxyisobutyrate dehydrogenase.
Alternatively a bi-functional malonyl-CoA reductase can be utilized, for example from Chloroflexus aurantiacus 11,13 .Furthermore the host can be engineered for improved supply of precursor and redox co-factor as the bi-functional malonyl-CoA reductase requires NADPH 13,14 .Finally, the route where E-alanine is converted into malonic semialdehyde either by the action of E-alanine-pyruvate aminotransferase (BAPAT) or J-aminobutyrate transaminase (GABT) has been reported in E. coli 15 and yeast 16 respectively (Fig. 1a).Malonic semialdehyde is further reduced into 3HP by the action of 3-hydroxypropionate dehydrogenase or 3-hydroxyisobutyrate dehydrogenase.E-alanine is a precursor for the biosynthesis of pantothenic acid (vitamin B5), which by itself is a component of coenzyme A and acyl carrier protein, co-factors required by a large number of enzymes.In bacteria, E-alanine is produced via decarboxylation of L-aspartate by aspartate-1-decarboxylase encoded by gene panD.No ortholog of panD was found in the genome of S. cerevisiae and it was shown that the only active route for E-alanine biosynthesis was via spermine 17 .Spermine can be naturally synthesized in yeast during ornithine metabolism.As it was our intention to engineer a high-3HPproducer, it was apparent that the biosynthesis of E-alanine precursor via spermine will most likely not be able to provide high flux and therefore it was preferred to engineer an alternative route for Ealanine biosynthesis from aspartate in yeast.
Hence, to engineer a de novo route for 3HP production in yeast via the E-alanine pathway it was essential to identify enzymes capable of converting E-alanine into 3HP and also to establish a synthetic pathway for efficient synthesis of E-alanine itself.Although a possibility of using BAPAT for 3HP production in yeast was mentioned, this has, to the best of our knowledge, not been experimentally demonstrated.Specifically, according to Liao et al. (2010), enzymes having BAPAT activity can be obtained from Pseudomonas putida or P. aeruginosa 15 .We, however, found that the BAPAT enzyme from P. putida was not active in S. cerevisiae.
The aim of our study was to establish a biosynthetic route for production of 3HP from glucose via E-alanine in yeast S. cerevisiae and to demonstrate that 3HP production at low pH was possible.

Results
Evaluation of E-alanine metabolic route using genome-scale modeling We used the iTO977 genome-scale metabolic model of S. cerevisiae to evaluate metabolic capabilities of the two promising routes towards 3HP, one via malonyl-CoA and another via Ealanine (Fig. 1b).We did not consider the glycerol route due to the difficulties associated with expression of B12-dependent glycerol dehydratase in S. cerevisiae or the lactate route because of the unfavorable thermodynamic constraints 12 .The predicted maximum yield of the malonyl-CoA route is strongly oxygen-dependent because of the requirement of ATP for acetyl-CoA synthesis.
The E-alanine route has a higher maximum theoretical yield, which is not significantly affected by the degree of aeration, and this makes the E-alanine route more attractive for an industrial process.
Identification of E-alanine-pyruvate aminotransferase (BAPAT) from B. cereus 3HP production in E. coli via a BAPAT enzyme from P. aeruginosa and P. putida has previously been described 15 .We expressed P. putida BAPAT in combination with 3-hydroxypropionate dehydrogenases (HPDH) or 3-hydroxyisobutyrate dehydrogenases (HIBADH) in S. cerevisiae, but we did not observe conversion of E-alanine into 3HP by the recombinant cells.Therefore, we looked for an alternative BAPAT ortholog.The presence of an enzyme with BAPAT activity in the cell extracts of B. cereus has been reported previously 18 .The enzyme was reported to have high substrate affinity with K m of 1.1 mM for both L-alanine and pyruvate as compared to K m 14 and 62 mM for L-alanine and pyruvate, respectively, in P. fluorescens BAPAT.The gene encoding the BAPAT enzyme in B. cereus was, however, not known.
We therefore performed a sequence search with the BAPAT gene aptA from Achromobacter dinitrificans (GenBank: AY330220) against the genome sequence of B. cereus to discover an uncharacterized aminotransferase yhxA (GenBank: WP_002076750) as the closest homologue (35% identity).We built a 3D protein model based on the x-ray structure of Bacillus anthracis aminotransferase (PDB: 3N5M ), which exhibits an overall sequence identity of 68% to YHXA and covers all but the first six residues (of 451).Eleven of the 14 amino acids of YHXA (S119, G120, S121, Y153, H154, E222, D255, V257, I258, K284, T320) located in the active site (defined as residues within 4Å from pyridoxal-5'-phosphate) are conserved among BAPAT proteins with experimentally confirmed enzymatic activity and YHXA (Fig. 2b, Supplementary Results).
Furthermore, W63 and R169 -both conserved and found in the active site, the latter being unique to BAPAT enzymes -are ideally positioned to form hydrogen bonds to the carboxyl group of the substrate (Fig. 2a).
The BAPAT-coding genes from P. putida and from B. cereus were synthesized in codon-optimized versions for S. cerevisiae and expressed in S. cerevisiae.The cell extracts were analyzed for BAPAT activity using an enzymatic assay.BAPAT activity was detected in the cells expressing YHXA from B. cereus, but not in the cells expressing BAPAT gene from P. putida or in a control strain carrying an empty plasmid (Supplementary Fig. S1).No activity was observed when pyruvate dehydrogenases or 3-hydroxypropanoate dehydrogenases.For the conversion of malonic semialdehyde to 3HP, the majority of the NADPH-dependent enzymes (HPDH) performed better than the NADH-dependent enzymes (HIBADH).This could be due to the predominance of the reduced (NADPH) versus oxidized (NADP) co-factor forms and the opposite relationship for NADH and NAD in the cytosol of S. cerevisiae 19 .
The identity of 3HP in the best sample was confirmed by NMR analysis (Supplementary Fig. S2).
The concentration measured by NMR was 1.75±0.15g•L -1 , which was slightly lower than the value found by the enzymatic assay.The enzymatic assay likely leads to overestimation of 3HP concentration due to interference of other metabolites from the fermentation broth with the enzymatic reaction.Therefore, in all the following experiments the concentration of 3HP was measured by HPLC instead of enzymatic assay.In the HPLC analysis 3HP was well separated from the other components of the fermentation broth and quantification was therefore more accurate.

Reconstruction of biosynthetic route for E-alanine production from glucose via L-aspartate
The S. cerevisiae strains described above produced 3HP only when E-alanine was added to the medium.Cultivating the strains on glucose as the only carbon source or on a glucose medium supplemented with L-aspartate or L-alanine did not result in 3HP formation (data not shown) as could be anticipated from the previous report on the lack of E-alanine synthesis in S. cerevisiae during growth on glucose 17 .In bacteria, E-alanine can be synthesized through decarboxylation of Laspartate by aspartate-1-decarboxylase. Some glutamate decarboxylases can also have an activity towards L-aspartate 20 .We expressed three different aspartate-1-decarboxylases, all with confirmed enzymatic activity in their native hosts, from Corynebacterium glutamicum, Tribolium castaneum, E. coli, and three glutamate decarboxylases from Rattus norvegicus, E. coli, and native S. cerevisiae GAD1 in combination with two enzymes converting E-alanine into 3HP (B.cereus BAPAT and E. coli YDFG).The resulting strains were evaluated for 3HP production on defined mineral medium containing glucose as the only carbon source or glucose and L-aspartate, and on feed-in-time medium for S. cerevisiae to simulate production in a fed-batch process.Only the S. cerevisiae strains expressing aspartate-1-decarboxylases from red flour beetle T. castaneum or from C. glutamicum were capable of producing 3HP (Fig. 3c).The 3HP titers for the strain overexpressing TcPAND, BcBAPAT and EcYDFG were 0.83±0.08g•L -1 and 3.66±0.72g•L -1 on mineral and feedin-time media, respectively.
Increasing the flux towards 3HP via optimization of gene expression and improving supply of Laspartate Once the biosynthesis of 3HP from glucose via E-alanine had been established, the next goal was to improve the expression of the biosynthetic genes and to increase the flux towards aspartate.As this would require stable simultaneous overexpression of several genes, we used the previously developed integrative vector set EasyClone 21 .We tested the effect of overexpressing genes encoding native cytoplasmic aspartate aminotransferase AAT2, pyruvate carboxylases PYC1 and PYC2, and combinations thereof.We also investigated the effect of multiple integration of the key biosynthetic genes, leading from aspartate to 3HP, using the vector designed for integration into retrotransposon TY4 elements (Supplementary Results).Increasing the copy number of integrated BcBAPAT/EcYDFG or of TcPAND genes led to improvement of 3HP titer in all the four background strains tested (reference, overexpressing AAT2, overexpressing PYC1&PYC2 and overexpressing AAT2&PYC1&PYC2) (Fig. 4).The effect of multiple integrations of TcPAND was larger than that of multiple copies of BcBAPAT/EcYDFG.We also cloned all three genes (BcBAPAT, EcYDFG and TcPAND) into TY-vector and attempted to integrate all the genes in multiple copies.The resulting transformants did not secrete 3HP and colony PCR revealed that the strains lacked one or several of the BcBAPAT, EcYDFG or TcPAND genes.While integration into chromosomes is usually rather stable, the loss of the integrated genes by homologous recombination still can occur, and in this case there must have been a strong selection pressure for these events.
The increased expression of AAT2, PYC1 and PYC2 had a positive effect on 3HP production mainly in the strain with multiple copies of TcPAND gene, thus indicating that the main flux control resided with TcPAND, when it was in single copy, however once PAND activity was increased, the flux control partially moved to the aspartate biosynthesis enzymes.The positive effect of overexpressing pyruvate carboxylase genes was only observed on feed-in-time medium, which simulates fed-batch conditions, but not in the batch cultivation.This may be because under batch conditions, i.e., during growth on glucose and then on ethanol, there is high pyruvate carboxylase activity in the cells 22 and therefore expression of additional copies of PYC1/PYC2 did not have an effect.The highest titers were obtained for the strain SCE-R2-200 overexpressing AAT2, PYC1, PYC2, BcBAPAT, EcYDFG, and multiple copies of TcPAND: 1.27±0.28g•L -1 and 8.51±1.05g•L -1 on mineral and feed-in-time media correspondingly.

Recycling of L-alanine
The BAPAT activity results in the formation of one L-alanine molecule per each E-alanine molecule formed (Fig. 1a).The generated L-alanine can either be recycled to pyruvate via various transamination reactions or via alanine aminotransferase, which transfers amino group to Dketoglutarate to form glutamate.It was also proposed that L-alanine can be converted to E-alanine via the action of a non-natural 2,3-alanine aminomutase (AAM) developed by directed evolution of L-lysine 2,3-aminomutase from B. subtilis 23 .The S. cerevisiae genome contains two genes encoding alanine aminotransferase enzymes, ALT1 and ALT2, of which the latter is catalytically inactive 24 .
We tested the effect of ALT1 or BsAAM overexpression on 3HP production.For the test a leucinenegative S. cerevisiae strain was constructed, which overexpressed PYC1, PYC2, BcBAPAT, EcYDFG, and multiple copies of TcPAND.Multiple transformants were screened for 3HP production and the best performing strain (SCE-R3-46) was selected.It was further transformed with either an empty integrative vector to complement the leucine auxotrophy or with vectors designed for overexpression of BsAAM or ALT1.Overexpression of either BsAAM or of ALT1 genes decreased the 3HP titer on feed-in-time medium by correspondingly 18% or 10% (Supplementary Fig. S3).

Production of 3-hydroxypropionic acid in controlled fed-batch fermentation
Strain SCE-R2-200 described above was cultivated in aerobic fed-batch reactor with glucoselimited feed at pH5 in triplicates.The medium contained mineral salts, vitamins and glucose as the sole carbon source.The strain produced 3-hydroxypropionic acid at 13.7±0.3g•L -1 titer, 14±0% Cmol•C-mol -1 yield on glucose and an overall process productivity of 0.17±0.0g•L -1 •h -1 .No significant amounts of by-products such as acetate, ethanol or glycerol were detected at the end of the fermentation (Fig. 5).Production of 3HP at low pH was also evaluated.Using a simplified constant feed protocol, 12.2 g•L -1 and 9.2 g•L -1 of 3HP were obtained when pH was maintained at correspondingly 5 and 3.5 throughout the fermentation (Supplementary Fig. S4).

Discussion
In production of organic acids, it is advantageous to carry out the fermentation process at low pH: it saves the costs of neutralization and subsequent acidification to obtain the free acid and also reduces the likelihood of bacterial contamination.Therefore fungal hosts are often preferred (e.g, A. niger for citric acid production and yeast for lactic acid and succinic acid production).
We engineered a heterologous route for de novo production of 3-hydroxypropionic acid via Ealanine in yeast S. cerevisiae.To the best of our knowledge, this is the first report on utilization of E-alanine-pyruvate aminotransferase for 3HP production in yeast.Overexpression of BAPAT from B. cereus resulted in higher 3HP titers than overexpression of the native J-aminobutyrate transaminase, presumably due to higher intracellular concentration of pyruvate than Dketoglutarate, which are the substrates for BAPAT and GABT, respectively.Interestingly, while the uncharacterized aminotransferase YHXA from B. cereus has a very low sequence identity to the other known BAPAT enzymes, the residues in the active site within 4Å of the pyridoxal-5'phosphate moiety are strongly conserved between the known BAPAT enzymes.Furthermore, R169 is ideally positioned to hold the carboxyl group of the E-alanine substrate in collaboration with W63 on the opposite side of the substrate-binding pocket.This arginine residue not only appears to be conserved among all characterized BAPAT enzymes, it is also unique to them and is possibly involved in defining the specificity of BAPAT enzymes.
3HP is a source of commodity chemicals and needs to be produced at a price competitive to the petrochemical acrylic acid.Despite the promising production parameters (~14 g•L -1 titer and 14% C-mol/C-mol glucose yield), the titer and the yield probably need to be improved at least 4-5-fold to reach commercial feasibility for production of 3HP from glucose while volumetric productivity must probably be increased more than 10-fold 25 .With the maximal theoretical yield of nearly one gram 3HP per gram of glucose via E-alanine pathway (Fig. 1b), it should, however, be possible to achieve the commercial process requirements by further strain and process optimization, and we are therefore confident that this work forms the basis for a commercial 3HP production using S. cerevisiae.

Conclusions
In this study we engineered a synthetic pathway for de novo biosynthesis of E-alanine and its subsequent conversion into 3HP in yeast S. cerevisiae.Furthermore, we optimized the expression of the critical enzymes in the pathway and increased aspartate biosynthesis to obtain a high-3HPproducing strain.We demonstrated 3HP production at low pH 3.5.
M2P labs GmbH (Germany).The medium was supplemented with the supplied vitamins solution (final 1% v/v) and the enzyme mix (final concentration 0.5% v/v) immediately prior to use.

Metabolites analysis by HPLC
The extracellular metabolites concentrations were analyzed on HPLC.Therefore 25 μl of the sample were analyzed for 45 min using Aminex HPX-87H ion exclusion column with a 1 mM H 2 SO 4 flow of 0.6 ml/min.The temperature of the column was 60°C.The refractive index at 45°C and the UV absorption at 210 nm were measured.Glucose, glycerol and ethanol were detected using RI-101 Refractive Index Detector (Dionex).3HP, pyruvate, succinate and acetate were detected with DAD-3000 Diode Array Detector at 210 nm (Dionex).The UV spectrum was recorded with diode array and the identity of the 3-hydroxypropionic acid was additionally verified by comparison with the spectrum of the standard.

NMR spectroscopy
The spectra were acquired on a Bruker AVANCE III 800 spectrometer (Bruker, Rheinstetten, Germany) working at a proton operating frequency of 800.33 MHz, equipped with a four channel 5 mm inverse detection probe head with pulse-field gradients along the Z axis.
Spectra were run at 25 o C with presaturation of the water signal and a recycle delay of 6 seconds.
For quantification purposes, formate was added as an internal concentration standard and spectra run with a repetition delay of 66 s.

Cultivation in microtiter plates
At least six single colonies originating from independent transformants were inoculated in 0.5 ml drop-out liquid medium without uracil, histidine, and leucine in a 96-deep well microtiter plate with air-penetrable lid (EnzyScreen, NL).The plates were incubated at 30°C with 250 rpm agitation at 5 cm orbit cast overnight.50 μl of the overnight cultures were used to inoculate 0.5 ml mineral medium or synthetic fed-batch medium in a 96-deep well plate.Fermentation was carried out for 72 hours at the same conditions as above.For some experiments the mineral medium was supplemented with 10 g•L -1 E-alanine.
At the end of the cultivation OD 600 was measured as following: 10 μl of the sample was mixed with 190 μl water and absorbance was measured at 600 nm wave length in microplate reader BioTek Synergy MX (BioTek).The culture broth was spun down and the supernatant analyzed for 3hydroxypropionic acid concentration using enzymatic assay or HPLC.
Controlled fermentations SCE-R2-200 glycerol stock (0.3 ml) was inoculated in 150 ml mineral medium in a 500-ml baffled shake flask and propagated at 30°C with 250 rpm agitation for about 24 hours.The culture was concentrated down to 50 ml by centrifugation at 4,000 x g for 2 min and used to inoculate 0.5 L medium in 1L-Sartorius reactor.The final medium in the reactors contained per liter: 15 g (NH 4 ) 2 SO 4 , 6 g KH 2 PO 4 , 1 g MgSO 4 •7H 2 O, 4 ml trace metals solution, 2 ml vitamins solution, 0.4 ml antifoam A (Sigma-Aldrich), and 44 g dextrose.The initial cultivation volume was 0.5 L.
Dextrose was autoclaved separately.Vitamins solution was sterile filtered and added to the medium after autoclavation.The trace metal and vitamins solutions were the same as described previously 21 .
The agitation rate was 800 rpm, the temperature was 30°C, aeration was 1 L min -1 air and pH was maintained at 5.0 by automatic addition of 2N NaOH.Carbon dioxide concentration in the off-gas was monitored by acoustic gas analyzer (model number 1311, Bruël & Kjaer).Once the glucose was exhausted, which was observed from decline in CO 2 production and was also confirmed by residual glucose detection using glucose strips Glucose MQuant™ (Merck Millipore), the feed was started at 5 g•h -1 .The feed contained per liter: 45 g (NH 4 ) 2 SO 4 , 18 g KH 2 PO 4 , 3 g MgSO 4 •7H 2 O, 12 ml trace metals solution, 6 ml vitamins solution, 0.6 ml antifoam A, and 176 g dextrose.Dextrose was autoclaved separately, vitamins solution was sterile filtered and added to the feed after autoclavation.24 hours after the feed start the feed rate was ramped up to 10 g•h -1 and 48 hours after the feed start it was further increased to 15 g•h -1 .The total feed volume used per reactor was 0.5 L. The measured glucose concentration did not exceed 0.3 g•L -1 during the fed-batch phase.The reactors were sampled twice a day to measure biomass dry weight and metabolites.For metabolites analysis the sample was immediately centrifuged and the supernatant was stored at -20°C until HPLC analysis.
For estimating the effect of pH on 3HP production, the fermentations were carried out with constant feed rate of 10 g•min -1 .

Modelling of YHXA
A homology model of YHXA was built using the HHpred server (http://toolkit.lmb.unimuenchen.de/hhpred;PMID 15980461).The model was based on the x-ray structure of Bacillus anthracis aminotransferase (3N5M), which exhibits 68% sequence identity to YHXA and covers residues 7-451 (of 451 residues; 98.4%).Following homology modeling, a pyridoxal-5'-phosphate (PLP) moiety was manually added from the structure of Chromobacterium violaceum Ztransaminase (4A6T; PMID 22268978) and linked to the side chain amino group of the universally conserved K284.The monomer YHXA model with the PLP co-factor was duplicated and arranged as a dimer by alignment to the template structure.Finally, the substrate E-alanine was manually docked into the binding site.x We report high-level production of 3HP from a sugar feedstock by yeast x A de novo route for 3HP biosynthesis via E-alanine was engineered x Using yeast as the cell factory enables a more economical low pH process

Figure 2 :
Figure 2: BAPAT structure and alignment.(a) Predicted 3D structure of YHXA protein from B.

Figure 3 :
Figure 3: Production of 3HP by recombinant S. cerevisiae strains.(a) Overview of the genes

Figure 4 :
Figure 4: Strain optimization.Influence of integrating multiple copies of genes and of

Figure 5 :
Figure 5: Fed-batch cultivation of 3HP producing strain.Growth and metabolite production