In silico comparison of the malonyl-CoA and the β-alanine pathways in P. pastoris
The v3.0 version of the iMT1026 P. pastoris genome-scale metabolic model [42] was used to evaluate the maximum 3-HP theoretical yield (Υ3−HP max) that could be achieved using methanol as sole carbon source through the malonyl-CoA and the β-alanine pathways. Assuming that all of the carbon source was used for product synthesis, that is, not used for biomass growth (µ = 0 h− 1), the maximum Υ3−HP through the malonyl-CoA pathway was calculated to be 0.718 g3 − HP gMetOH−1 (Fig. 1a), whereas a higher value was obtained when using the β-alanine route (0.795) (Fig. 1b). Methanol is a highly reduced carbon source. Its oxidation demands high amounts of oxygen to produce the required ATP for biomass and product formation. Therefore, high 3-HP yields are correlated to an elevated oxygen consumption, as shown in the color scale of Fig. 1a and Fig. 1b. Nonetheless, the Υ3−HP max value for the β-alanine route turned out to be higher (> 10%) than for the malonyl-CoA pathway, as less ATP is required to produce one molecule of 3-HP from 3 molecules of methanol in the former scenario (see Supplementary file 1). This also occurs when glucose is used as substrate. However, when glycerol is used, there is a net production of ATP in both pathways, although more ATP molecules are generated through the β-alanine route, which also results in a higher Υ3−HP max value.
<Suggested position for Fig. 1>
The expression of panD, yhxAandydfGgenes inP. pastorisgenerates a 3-HP-producing strain
To reconstruct the synthetic β-alanine pathway for 3-HP production, the three codon-optimized panD, yhxA and ydfG genes from T. castaneum, B. cereus and E. coli, respectively, were introduced into P. pastoris CBS7435, the parental strain. The strong and methanol inducible alcohol oxidase 1 (AOX1) and formate dehydrogenase (FDH1) promoters, were selected to conduct the expression of panD and yhxA genes, respectively. According to previous studies [37], the utilization of strong promoters such as the translation elongation factor EF-1 alpha promoter (PTEF1−α) for the expression of both panD and yhxA genes proved to be beneficial for 3-HP production with an engineered S. cerevisiae strain growing on glucose, since the reaction catalyzed by the PANDTc was determined to be the main flux control step of the β-alanine pathway. In these two first steps, an L-aspartate is transformed into β-alanine by the PANDTc, which in its turn is converted into MSA by the BAPATBc enzyme. 3-HP is finally obtained through the reduction of MSA by the action of the YDFGEc (Fig. 2). Due to the NADPH-dependence of the latter enzyme, we expected the potential generation of a redox imbalance in the last step of the pathway. For this reason, instead of using a strong and methanol inducible promoter to control the expression of the ydfG gene, two different constitutive promoters showing a significantly lower expression strength in comparison with the strong GAP promoter under methanol feed conditions (µmax up to 0.1 h− 1) were tested, namely the moderate strength mitochondrial porin (POR1) promoter, obtaining the PpCβ10 strain, and the weak pyruvate decarboxylase (PDC1) promoter, obtaining the PpCβ20 strain [5].
<Suggested position for Fig. 2>
The PpCβ10 and PpCβ20 strains were grown up to an optical density at 600 nm (OD600) of 2.4 (see Supplementary file 2) on 24 deep-well plates with minimal medium containing 11.9 g l− 1 (i.e., 1.5% v/v) of methanol, along with the reference strain (CBS7435). After 48 h of incubation, methanol was completely depleted. Since P. pastoris does not assimilate 3-HP as C-source [32], an endpoint sample was collected at the end of the culture to quantify the final 3-HP titers. The heterologous expression of panD, yhxA and ydfG genes resulted in 0.93 ± 0.03 g l− 1 of 3-HP after 48 h of cultivation when the PPOR1 was used to control the expression of the ydfG gene (PpCβ10 strain), whereas a slightly but significantly higher titer (p-value < 0.05), 1.04 ± 0.01 g l− 1 3-HP, was obtained with the weak PPDC1 driving expression of this gene (PpCβ20 strain). Notably, the 3-HP titers obtained herein were slightly higher than those achieved with a base S. cerevisiae strain expressing the same key enzymes under the control of strong constitutive promoters (TEF1 and PGK1) and growing on a 20 g l− 1 glucose medium (0.83 g l− 1 3-HP) [37]. Moreover, the expression of these three heterologous genes in P. pastoris resulted in similar 3-HP titers to those previously reported for a glucose-utilizing S. cerevisiae strain carrying overexpression cassettes for 5 genes, namely pyruvate carboxylase (PYC1 and PYC2), BAPATBc, PANDTc and YDFGEc. Only when xylose was used as substrate, a higher production of 3-HP was achieved (1.84 g l− 1) with the latter strain [28]. These initial screening experiments pointed at P. pastoris as a promising chassis cell for the bioproduction of 3-HP from methanol.
Improving 3-HP production by optimizing the flux through the β-alanine synthetic pathway
Once the functionality of the biosynthetic pathway was confirmed by the presence of 3-HP, we decided to improve the flux towards the production of this carboxylic acid by overexpressing the key biosynthetic enzymes. Previous studies with a S. cerevisiae strain expressing the β-alanine pathway, demonstrated that the effect of multiple integrations of PANDTc was larger than that of multiple copies of BAPATBc/YDFGEc, thus suggesting that the major control point of the flux through the biosynthetic pathway was the panD gene transcriptional levels [37]. Accordingly, we hypothesized the overexpression of this gene as the most straightforward strategy for 3-HP improvement.
The integration of a second copy of the PANDTc expression cassette into both PpCβ10 and PpCβ20 strains resulted in the PpCβ11 and PpCβ21 strains, respectively, which were cultivated on buffered minimal methanol medium (BMM) along with their parentals. After 48 h of cultivation, all strains reached a similar OD600 between 2.4–2.6 (see Supplementary file 2). The 3-HP production slightly increased by 3% (p-value = 0.04) in PpCβ11 strain (0.92 ± 0.01 g l− 1 3-HP), in comparison with its parental PpCβ10 strain (0.89 ± 0.02 g l− 1), while the strain PpCβ21 reached the highest 3-HP concentration (1.18 ± 0.03 g l− 1), i.e., a 12% increase respect to PpCβ20 (1.05 ± 0.01 g l− 1). Notably, the 3-HP yield on methanol was increased by 11% with the additional copy of the panD gene (strain PpCβ21) while this effect was much lower in the PpCβ11 strain (3%) (Fig. 3). These results indicate that lower expression levels of the ydfG gene (i.e., under the control of PPDC1) seem to be beneficial in terms of 3-HP production. This suggested the presence of a bottleneck at the final step of the pathway, since the YDFGEc requires NADPH to catalyze the reduction of the malonic semialdehyde to 3-HP, which is not being supplied. We hypothesized that this perturbation of the NADPH homeostasis could be further exacerbated when the β-alanine flux into the pathway is increased (i.e., with the insertion of an additional copy of the panD gene under the very strong AOX1 promoter), resulting in a potential limitation of NADPH equivalents for anabolism/cell growth.
<Suggested position for Fig. 3>
Engineering the redox balance by overexpressing the PseFDH(V9) enzyme variant from Pseudomonas sp. 101, a NADP + -dependent formate dehydrogenase
To increase NADPH supply, a mutated formate dehydrogenase from Pseudomonas sp. 101, namely PseFDH(V9), showing an improved efficiency and specificity towards NADP+ compared to all previously explored FDH [43], was expressed in both PpCβ20 and PpCβ21 strains, obtaining the PpCβ20-P and PpCβ21-P strains, respectively. The gene encoding for the PseFDH(V9) enzyme was expressed under the control of the strong and methanol inducible FDH1 promoter. The PpCβ20-P and PpCβ21-P strains were cultivated along with their corresponding parental strains on 24-deep well plates containing BMM. A comparable endpoint OD600 between 2.5–2.6 was reached after 48 h of cultivation by the strains PpCβ20, PpCβ21 and PpCβ20-P. Conversely, cell growth was slightly lower for the PpCβ21-P strain (DO600 = 2.2) (see Supplementary file 2). The final 3-HP titer was significantly reduced by about 11% in strain PpCβ20-P (1.12 ± 0.01 g l− 1) compared to its parental strain PpCβ20 (1.24 ± 0.03 g l− 1) (p-value < 0.05). The introduction of the PseFDH(V9) enzyme in strain PpCβ21 resulted in a minor but also significant decrease (p-value < 0.003) of about 6% on the final product titer, from 1.29 ± 0.01 g l− 1 to 1.22 ± 0.01 g l− 1 (PpCβ21-P). However, there were no significant differences (p-value = 0.07) in the 3-HP produced per OD600 unit between these strains. In summary, the expression of this NADP+-dependent formate dehydrogenase did not improve neither 3-HP production nor product yields using methanol as sole carbon and energy source in small-scale experiments.
Evaluating the impact of the β-alanine pathway on cell growth kinetics
To determine the maximum specific growth rate (µmax) of the different strains obtained in this study, series of growth kinetics experiments were performed by cultivating a representative clone from each strain in shake flask cultures using BMM medium. The introduction of the β-alanine route led to a reduction of the µmax from 0.14 h− 1 (reference CBS7435 strain) to 0.12 h− 1 and 0.10 h− 1 for PpCβ20 and PpCβ10 strains, respectively (see Supplementary file 2). The β-alanine pathway consumes 1 NADPH per 3-HP, thereby reducing NADPH availability for anabolic purposes, which has been previously correlated to decreased biomass yields on methanol [44]. Coherently, lower expression levels of the NADPH-consuming YDFGEc enzyme (i.e., under the control of PPDC1, strain PpCβ20) of the β-alanine pathway proved to be beneficial, both in terms of 3-HP production and cell growth. Moreover, this effect was more pronounced with the overexpression of the PANDTc expression cassette, as inferred from the µmax comparison between strains PpCβ21 and PpCβ11 (see Supplementary file 2). Introducing a second panD gene copy into the PpCβ10 strain substantially decreased the µmax value from 0.10 h− 1 to 0.06 h− 1 (PpCβ11), whereas this effect was minimal in the PpCβ20 strain, leading to a slight (not statistically significant) reduction from 0.12 h− 1 to 0.10 h− 1 (PpCβ21). According to previous studies [15], fine-tuning the panD gene dosage aiming at increasing the flux towards the β-alanine intermediate through aspartate decarboxylation may cause an impaired cell growth of P. pastoris, since aspartate takes part in many biological processes. The results obtained with both PpCβ21 and PpCβ11 strains were consistent with this phenomenon. Further metabolic flux analyses should help elucidating the cause of the differences between the strains PpCβ11 and PpCβ21 growth. In any case, PpCβ11 was discarded for further bioreactor-scale cultivation experiments.
The introduction of the mutated formate dehydrogenase FDH(V9) from Pseudomonas in the PpCβ20 strain seemed to cause a slight decrease in the specific growth rate, from 0.12 h− 1 to 0.10 h− 1 (PpCβ20-P) but was statistically not significant. A lower µmax value was determined for the PpCβ21-P strain (0.08 h− 1), in comparison with its parental strain PpCβ21. Since both strains showed similar 3-HP production titers, further characterization in bioreactor-scale experiments was required.
Production of 3-HP in fed-batch cultivations
Initial fed-batch cultivations of the 3-HP-producing strains using a growth defined mineral medium typically used for recombinant protein production in P. pastoris [45], which is only supplemented with biotin, revealed that such medium did not support neither growth nor 3-HP production during the feed phase on methanol. We hypothesized this was probably due to the special co-factor requirements described for some of the enzymes of the β-alanine pathway, specifically for the PANDTc (UniProtKB: A7U8C7) and BAPATBc (UniProtKB: C2VE79) enzymes, which use at least one pyridoxal phosphate as cofactor for its synthesis according to the UniProtKB database [46]. This coenzyme is derived from pyridoxine (vitamin B6) and plays an important role in amino acid metabolism. Moreover, β-alanine is an intermediate in pantothenic acid (vitamin B5) biosynthesis, which, in turn, is an integral part of the coenzyme A and acyl carrier protein, co-factors required for several enzymes [37]. The redirection of this β-amino acid flux towards the production of 3-HP instead of vitamin B5, could be an explanation for the observed growth impairment in the bioreactor-scale cultivations. Indeed, the presence of both vitamins B6 and B5 in the yeast nitrogen based (YNB)-containing medium used in the small-scale cultivations is likely to be the reason why the 3-HP-producing strains grow normally in those conditions. Subsequently, bioreactor cultivations were carried out using a growth medium supplemented with a vitamin solution containing, among others: calcium pantothenate (B5), pyridoxine hydrochloride (B6), and niacin (B3), which is an important precursor for the essential redox cofactors NAD+ and NADP+. This vitamin solution is usually added to Delft medium, commonly employed for bioreactor cultivations of S. cerevisiae [47], allowing for sustained growth on methanol and 3-HP production.
The best performing strains PpCβ20 and PpCβ21, along with the strain expressing the mutated FDH(V9) enzyme from Pseudomonas, PpCβ21-P, were further evaluated in aerobic fed-batch reactors. After the initial glycerol batch and methanol adaptation (transition) phases, the fed-batch phase was started, where methanol was fed following a pre-programmed exponential feeding strategy for controlled growth rate at µ = 0.03 h− 1. The feeding phase was finalized before reaching a dry cell weight (DCW) of 50 g l− 1 of biomass, accounting for 39.3 h of the total process duration, with a total methanol concentration added into the culture of almost 125 g l− 1. Neither methanol nor by-products were detected by HPLC analysis over the course of fermentation. Nonetheless, exometabolome profiling analyses by NMR of the supernatant samples from the early, mid, and late methanol-feeding phase revealed the presence of small amounts (mM range, below HPLC detection limit) of ethanol as well as some metabolites potentially derived from leucine (Leu) and isoleucine (Ile) biosynthesis and degradation pathways (or pyruvate fermentation) such as 2-oxoisocaproate and 3-methyl-2-oxopentanoate (see Supplementary file 3 for NMR spectra of the culture supernatants). Both compounds are involved in transamination reactions with 2-oxoglutarate and L-glutamate, which, in turn, are also present in the oxaloacetate conversion into L-aspartate catalyzed by an aspartate aminotransferase (AAT2), reaction that takes places upstream of the β-alanine pathway (see Fig. 2). 3-hydroxyisobutyrate, pyruvate and acetate were also detected in much lower amounts as well as traces of methanol. In contrast to 3-HP-producing P. pastoris strains engineered with the malonyl-CoA pathway growing in fed-batch cultures using glycerol as C-source [32], no arabitol was accumulated during the fermentation process.
At the end of the culture, the PpCβ21 strain produced up to 21.4 ± 0.7 g l− 1 of 3-HP, whereas 20.3 ± 0.6 g l− 1 were achieved by the strain PpCβ20. In both cases the 3-HP generated during the batch phase of the cultures was slightly above 2 g l− 1 (Fig. 4). The addition of a second copy of the panD gene in PpCβ21 resulted in a 12% improvement of the final 3-HP titer in the small-scale screenings. This increase was reduced to 6% when both strains were tested in bioreactor fed-batch cultivations. Nevertheless, the 3-HP yield on biomass (YP/X) calculated for the fed-batch phase was increased by 18% in PpCβ21 with respect to PpCβ20. Consequently, this resulted in a 10% improvement of the specific productivity (qp) (Table 1), and thus demonstrating that an enhanced supply of β-alanine intermediate in the metabolic pathway had a positive effect on 3-HP production, coherent with previous studies [37]. The expression of the PseFDH(V9) enzyme in PpCβ21-P strain allowed for a final 3-HP titer of 19.0 ± 0.5 g l− 1, along with lower biomass concentration levels (38.8 g l− 1 DCW) (Fig. 4), in comparison with the above-mentioned strains. This yielded a further 14% and 10% increase in YP/X and qp, respectively, compared to PpCβ21, reflecting an enhanced cell NADPH availability. The PpCβ21-P strain also showed a remarkably higher specific methanol consumption rate (qs), since the same amount of substrate was consumed, and less biomass was generated in comparison with PpCβ20 and PpCβ21 strains (Table 1).
Table 1
Averaged value of key process parameters obtained for the methanol fed-batch phase using a preprogrammed µ of 0.03 h-1. Volumetric productivity (QP), biomass yield on methanol (YX/S), 3-HP yield on methanol (YP/S), 3-HP yield on biomass (YP/X), specific substrate consumption rate (qS), specific 3-HP production rate (qP), and experimentally measured mean specific growth rate (µ). Cultivations were performed in duplicate and biomass concentration analyses were performed in triplicate. ± indicates SD of the biological replicates.
| PpCβ20 | PpCβ21 | PpCβ21-P |
QP (g3 − HP l− 1h− 1) | 0.46 ± 0.01 | 0.48 ± 0.02 | 0.46 ± 0.01 |
YX/S (gDCW gMetOH−1) | 0.28 ± 0.01 | 0.25 ± 0.01 | 0.21 ± 0.01 |
YP/S (g3 − HP gMetOH−1) | 0.14 ± 0.01 | 0.15 ± 0.01 | 0.15 ± 0.01 |
YP/X (g3 − HP gDCW−1) | 0.52 ± 0.03 | 0.61 ± 0.03 | 0.69 ± 0.03 |
qS (gMetOH gDCW−1 h− 1) | 0.103 ± 0.002 | 0.107 ± 0.002 | 0.124 ± 0.002 |
qP (mmol3 − HP gDCW−1 h− 1) | 0.166 ± 0.001 | 0.183 ± 0.001 | 0.201 ± 0.001 |
µ (h− 1) | 0.029 ± 0.001 | 0.027 ± 0.001 | 0.026 ± 0.001 |
The final 3-HP titers and volumetric productivities obtained herein are substantially higher to the ones reported for a S. cerevisiae strain expressing the β-alanine pathway growing in a glucose fed-batch bioreactor (13.7 g l− 1, 0.17 g l− 1 h− 1) [37]. The difference of electrons per carbon atom between methanol (6) and glucose (4) might be an explanation for this observation since excess electrons can be channeled to 3-HP production [2]. Moreover, S. cerevisiae produces ethanol during growth phase under aerobic glucose excess conditions (i.e., in the batch phase of a fed-batch cultivation) due to overflow metabolism (Crabtree effect). This can have a significant impact on the global process 3-HP yield, since there is a competition for carbon between the pyruvate carboxylase (PYC), catalyzing the conversion of pyruvate into oxaloacetate (which is a precursor of the β-alanine pathway), and the pyruvate decarboxylase (PDC), channeling the pyruvate into ethanol. This fact may explain why higher 3-HP yields are achieved using xylose as C-source (0.29 g g− 1) [28], as this substrate does not induce the Crabtree effect on S. cerevisiae. Similarly, arabitol accumulation was detected in the late stages of the batch and fed-batch phases of P. pastoris cultivations on glycerol for 3-HP production [32], causing a significant decrease in the global product yield. Conversely, when P. pastoris grows on methanol, exometabolite levels are lower than when growing on other substrates such as glucose as sole carbon source [48].
<Suggested position for Fig. 4>
Production of 3-HP from methanol has also been reported by an engineered Methylobacterium extorquens AM1 expressing the reductive malonyl-CoA pathway [49], although the yields obtained were an order of magnitude smaller than the ones reported in this study. Furthermore, the volumetric productivities achieved herein are significantly higher (about 3.2-fold) to the productivities observed in the recently published P. pastoris strain producing 3-HP from sole methanol (0.15 g l− 1 h− 1, calculated from data reported by Wu et al. [10]. To our knowledge, the values obtained in this study are the highest volumetric productivities reported so far for a 3-HP production process based on the β-alanine pathway.