Functional expression and evaluation of heterologous phosphoketolases in Saccharomyces cerevisiae
This study evaluated yeast expression of phosphoketolases from bacterial origin due to their potential application in metabolic engineering. The phosphoketolase from A. nidulans was included as a reference as it had been used beneficially in yeast metabolic strategies previously (de Jong et al. 2014a; Kocharin et al. 2013). From the phylogenetic tree in Fig. 2, it can be seen the selected enzymes in this study fall into two primary groups—enzymes from the Bifidobacterium genus and the Firmicutes phylum, both distinct from the fungal phosphoketolase. This is expected, in part because of the evolutionary distance between these groups, but also because phosphoketolase function differs between the two major clusters. The bifidobacterial phosphoketolase has a crucial role in breaking down hexose sugars via the characteristic “bifid shunt” (Pokusaeva et al. 2011), while in lactic acid bacteria, phosphoketolases primarily function to break down pentose sugars (Ganzle 2015). This was also displayed in the in vitro assay as a difference in substrate preference towards F6P and X5P respectively. Indeed, enzymes from the Bifidobacterium genus showed higher relative activity towards substrate F6P, compared with the enzymes from the Firmicutes phylum.
The in vitro phosphoketolase results were based on crude cell free extract of strains expressing the heterologous gene candidates. Several of the selected enzymes showed significant activity in the assay, while the two candidates from A. nidulans and L. paraplantarum did not. As phosphoketolase quantities were not directly measured in this study, it cannot be excluded that low protein production is the cause for this inactivity. Previously reported activities of the candidates upon X5P ranged between 2 and 29 U/mg, whereas in our study this range is between 0.27 and 1.05 U/mg total protein (see comparisons of obtained and reported values in Additional file 1: Table S3). The corresponding ranges for reported F6P specific activity values are 0.1–147.3 U/mg, while we in our study obtained values ranging from 0.11 to 0.48 U/mg. The characterization studies were conducted with purified proteins, with some of the studies reporting purification factors of about 11–24 (Jeong et al. 2007; Meile et al. 2001; Yevenes and Frey 2008), which could explain these large differences. Also, expression differences between yeast and bacterial systems could be a factor explaining the dissimilarities in specific activity measurements.
A general similarity between our obtained data and previously reported values, is that the specific activity towards X5P was higher than towards F6P for all enzymes tested. Previously reported X5P:F6P activity ratios were between 5:1 and 2:1. Also in the present study, the X5P activity was in all instances higher than the F6P activity. In the specific cases, however, differences can be seen, for example for the Bifidobacterium species where the obtained ratio between X5P:F6P activity was closer to 3:2, thus showing a higher specificity towards F6P than previously reported. This is in agreement with the fact that the enzyme is solely responsible for F6P degradation in the Bifidobacterium species, as discussed above. For all enzymes from the Firmicute phylum the X5P:F6P ratio were approximately 5:1, supporting that their role is to break down pentose sugars in their native hosts.
The candidate that differed most in sequence similarity from the other candidates was Xfpk(AN) from A. nidulans (see a percent identity matrix for all protein sequences in Additional file 1: Table S4). The strain expressing xfpk(AN) failed to show any significant difference in activity level compared to the negative control in the defined assay conditions. In a previous study, the expression of this gene in combination with a phosphotransacetylase from Bacillus subtilis induced a 3.7-fold increase in the production of FAEEs (de Jong et al. 2014a). This could be an indication that the enzymatic activity of Xfpk(AN) is too low to be detected in vitro, possibly due to low sensitivity of the assay. It also shows promise that utilization of the more active phosphoketolase candidates found in this study has potential to increase titers of acetyl-CoA derived products further.
Our study confirmed that S. cerevisiae has a strong endogenous capability to break down acetyl-phosphate. We deleted the genes GPP1 and GPP2, encoding two endogenous phosphatases involved in glycerol biosynthesis, as suggested by a patent filed by the company Amyris (Hawkins et al. 2014), which allowed us to determine their individual contribution to the degradation of AcP. Deletions of these two genes may pose a requirement for efficient recombinant pathway flux if a phosphoketolase would be used in combination with a phosphotransacetylase or an acetate kinase, as considerable amounts of carbon otherwise would be processed by Gpp1 and Gpp2, which would reduce the potential energy benefits of the heterologous pathway. However, increased levels of the non-native metabolite AcP might have unexpected impact on yeast physiology. AcP has been suggested to be a key contributor to protein acetylation in E. coli, and the compound can acetylate proteins non-enzymatically in vitro (Weinert et al. 2013). As protein acetylation affects gene expression and metabolism in yeast on a global level (Henriksen et al. 2012), it would be beneficial to keep the AcP concentrations in the cell as low as possible.
The native AcP-degrading power of yeast allowed us to estimate phosphoketolase activity in vivo, simply by quantifying acetate production and consumption in small scale shake flask cultivations of the xfpk expressing strains. The higher specific activity of Xfpk(BB) towards F6P made it an interesting enzyme to evaluate in vivo, since the intracellular substrate distribution of F6P and X5P differs, and F6P is present at a higher concentration during growth on glucose (Wasylenko and Stephanopoulos 2015). As shown in Fig. 5a, the greatest increase in acetate was indeed seen for the strain expressing xfpk(BB), which correlates well with the enzyme’s stronger ability to convert the high-abundant F6P to AcP and erythrose-4-phosphate. Expression of xfpk(LM) and xfpk(CA) also led to increased acetate accumulation, albeit lower compared to xfpk(BB) expression.
Xfpk expression furthermore correlated with decreased cellular fitness, such as a reduced growth rate and decreased biomass formation, indicating that expression of these genes leads to some form of cellular stress. The increased acetate formation following xfpk expression could explain this behavior, as production of acetic acid will increase ATP consumption when the cell attempts to maintain its intracellular pH levels caused by the proton decoupling effect of the organic acid (Verduyn et al. 1990). A net consumption of ATP will also be the result when acetate produced via the phosphoketolase pathway is metabolized, as its production, contrary to the native pathway, does not generate any ATP. Furthermore, the fact that maximal acetate levels are reached at a later time point in strains expressing the high-activity phosphoketolases compared to the control strain, indicates that the AcP levels in these strains are high after glucose is depleted, which potentially could affect yeast physiology negatively, as mentioned above. It will be important to understand the physiological response in detail in order to counter-act the negative effects of phosphoketolase expression in future studies, for example by studying the effect of an augmented degradation of acetate and AcP.
The results obtained in this study suggest that if a highly efficient phosphoketolase is used in metabolic engineering strategies for improved yields of acetyl-CoA derived products, its expression should be coupled with an efficient channeling of the produced AcP to acetyl-CoA. This can for example be achieved by co-expressing the phosphoketolase with a phophotransacetylase (de Jong et al. 2014a; Sonderegger et al. 2004). Yet, expression of a Pta alone might not be sufficient to relieve problems following phosphoketolase expression in yeast. In a study by Sonderegger et al. (2004), the xfpk(BL) was expressed in combination with pta from B. subtilis in an attempt to increase xylose fermentation and ethanol formation rates. However, xfpk expression led to high acetate levels and reduced xylose consumption and ethanol formation rates even in the presence of Pta, which is why their final strategy for increased xylose fermentation was to rely on a weak endogenous phosphoketolase activity in combination with the B. subtilis Pta. The deletion of yeast genes GPP1 and/or GPP2 in combination with utilization of a more efficient Pta candidate might provide a solution to the problem, as indicated in a patent filed by Amyris, which proposes utilization of Xfpk from Leuconostoc mesenteroides in combination with a Pta from Clostridium kluyveri for improved production of acetyl-CoA derived chemicals (Hawkins et al. 2014). During the time this manuscript was under review, results were published showing that yeast comprising a deletion of GPP1, expression of the above stated heterologous genes together with expression of an acetaldehyde dehydrogenase and a NADH-consuming HMG-CoA reductase produced 25% more of the isoprenoid compound farnesene compared to the reference strain (Meadows et al. 2016), indicating the strong potential of a phosphoketolase based strategy for improved production of acetyl-CoA derived compounds. In general, an efficient system to channel the formed acetyl-CoA to product formation would likely stimulate flux through the heterologous pathway, for example expression of a de-repressed version of acetyl-CoA carboxylase, active under glucose conditions and efficiently forming the fatty acid precursor malonyl-CoA (Shi et al. 2014).
In conclusion, this study has shown that several bacterial heterologous phosphoketolase candidates can be functionally expressed in S. cerevisiae and that this can efficiently shift the natural central carbon metabolism, channeling C5 and C6 carbon directly towards C2 synthesis without passing a CO2 emitting step. We confirmed that the endogenous metabolism of yeast interferes with the phosphoketolase pathway through a strong, promiscuous ability to degrade AcP, and that deletion of the two native genes GPP1 and GPP2 can reduce this conversion. Furthermore, we showed that the expression of phosphoketolases with high catalytic activity led to increased acetate formation and negatively affects growth properties of S. cerevisiae. Our analysis provides a valuable resource for identification of suitable phosphoketolase enzymes for metabolic engineering of yeast with the objective to improve acetyl-CoA supply.