Industrial starch fermentation for the production of alcohols, amino acids, nucleotides, antibiotics, and high-glucose and high-maltose syrups (Ram and Venkatasubramanian, 1982; Malviya et al., 2010) relies on pullulanase to degrade α-1,6-glycosidic bonds to improve the efficiency of starch hydrolysis (Reddy et al., 2015). Pullulanase is also used to produce high-amylose starch, resistant starch, slow-digestion starch, maltooligosaccharides, and branched cyclodextrins (Shikaishi et al., 2014; Li et al., 2017).

Because pullulanase is used in these applications, a safe, low-cost, high-yield production method is needed. Although heterologous expression of pullulanase in Escherichia coli under various conditions has yielded as much as 580 U/mL (Nie et al., 2013) and 2523.5 U/mL (Zou et al., 2014), there are many restrictions for its use in foods, feeds, and pharmaceuticals because of the endotoxins and exotoxins produced by E. coli. Safer alternative species have been used for pullulanase production, including Bacillus subtilis (24.5 U/mL) (Song et al., 2016) and Pichia pastoris (350 U/mL) (Xu et al., 2006); however, the yields from these strains are relatively low. To solve these problems, researchers have isolated new types of pullulanase enzymes from various microorganisms, such as Bacillus sp. AV-7 (Kunamneni and Singh, 2006), Thermus thermophiles (Wu et al., 2014), Bacillus deramificans (Duan et al., 2013), Anoxybacillus sp. SK3-4 (Kahar et al., 2016), and Bacillus naganoensis. In addition, the yield of pullulanase has been increased through mutation breeding (2.82 U/ml) (Wang et al., 2014), protein engineering (46.9 U/ml) (Chen et al., 2016; Nisha and Satyanarayana, 2016), and the manipulation of culture condition (543 U/ml) (Zou et al., 2016). Despite these efforts, the yield of pullulanase from these strains is still too low to meet industrial demand.

B. subtilis is a viable species for improving pullulanase yield because it is a generally recognized as safe (GRAS) microbial-derived product (Ming et al., 2010). Therefore, we chose to use B. subtilis for our pullulanase production study. Protein yield is known to be closely related to the strength of the promoter; thus, a strong promoter is a necessary requirement for high protein yield (Blazeck et al., 2012). The most well-known promoter in B. subtilis is the cytidine deaminase (ccd) promoter P43 (Wu et al., 1991), which has been used to express GFP (Kong et al., 2009), β-galactosidase, staphylokinase (Kim et al., 2008) and alkaline protease (Kim et al., 2008). Yang et al. isolated a strong B. subtilis promoter (Plaps) that is 13 times stronger than the P43 promoter by using a promoter trapping system. Inducible promoters have also been widely used in B. subtilis; including promoters that are regulated by xylose, sucrose (Biedendieck et al., 2007), maltose (Biedendieck et al., 2007; Yue et al., 2017), starch, phosphates (Abdel-Fattah et al., 2005; Makarewicz et al., 2006), citric acid (Yamamoto et al., 2000), tetracycline (Geissendörfer and Hillen, 1990), and glycine (Phan and Schumann, 2007).

Moreover, promoters an also be combined to form a multiple-promoter complexes to further enhance the expression (Zhang et al., 2017), and have been shown to increase enzyme production 1.6- (Yang et al., 2013), and 12-fold (Kang et al., 2010). Zhang et al. (2017) designed a dual-promoter expression system, PhpaII-PamyQ. Using this system, they increased enzyme activity to 571.2 U/mL in a 3 L fermenter, which was 18.7 times the activity obtained in shake flasks. Guan C. R. et al. (2016) showed that amino peptidase could be expressed in B. subtilis by the synthetic dual promoter PgsiB-PHpaII. Using this system, the obtained enzyme activity was 88.86 U/mL in shake flasks and 205 U/mL in a 5 L fermenter. In addition, the core elements of promoters, including the −35 and −10 regions (Jiao et al., 2017) and ribosome recognition site (Wang and Doi, 1984), have been optimized to enhance the promoter strength.

One bioinformatic method for selecting candidate strong promoters to improve production efficiency is analyzing the amount of mRNA expressed in a transcriptome, which should represent the strength of the promoter (McCleary, 2009). This method has been used for other applications to increase production efficiency, thus saving time and reducing costs. For example, Liu et al. (2017) analyzed the top 10 most highly expressed genes and operons among 3,959 genes and 1,249 operons in transcriptome data from Bacillus licheniformis ATCC14580. Using this method, a novel high-efficiency promoter (PBL9) was identified, which showed 23% higher expression than P43 in B. subtilis. Geng et al. (2014) cloned a root-specific promoter and developed a high-yield screening system in peanut by establishing a simple digital expression profile based on Illumina sequencing data from peanut. However, no study has utilized transcriptome data to select highly active promoters in B. subtilis based on gene expression levels.

The major objective of this study was to improve the yield of pullulanase production by B. subtilis using different promoters. To this end, we first chose three B. subtilis transcriptome data sets to screen for strong promoters. Next, pullulanase expression driven by the selected promoters was evaluated in B. subtilis BS001. Then, these promoters were combined to generate dual- or triple-promoter expression systems to improve yield.

We analyzed three transcriptome data sets from B. subtilis to select the genes with the highest expression levels. The promoters of the top 10 genes were determined through predictive bioinformatic analyses and were used to express pullulanase in B. subtilis. Four promoters (PsodA, Phag, PtufA, and PfusA) were stronger than PamyE. Transcriptome mining, with the goal of engineering promoter-induced modifications to increase gene expression, has been previously reported. For example, Liu et al. (2017) selected a promoter, PBL9, from 3,595 genes and 1,249 operons in a B. licheniformis transcriptome, that was 23% stronger than P43. Liao et al. (2015) screened the candidate promoter Pr2 (the promoter of sigW) from 288 genes with higher expression levels (RPKM values) than the control gene P43, and observed the strongest β-galactosidase activity in post-log phase. Park et al. (2007) identified a cadmium-inducible promoter via transcriptome analysis of Hansenula polymorpha SEO1 that had broad specificity for heavy metals and was also responsive to arsenic and mercury. This study of pullulanase expression in B. subtilis reinforces the idea that selecting promoters from transcriptome data is a good approach for identifying strong promoters and can be used to optimize the expression of industrially important microbial products, saving time, reducing costs, and improving safety. This study is also the first to select promoters based on B. subtilis transcriptome data, showing that it is a viable option to modify expression of pullulanase.

The yield of pullulanase from a single promoter was unsatisfactory because of insufficient strength. Some researchers proved that optimizing the −35 and −10 regions of promoter could enhance promoter strength. Jiao et al. (2017) constructed a super-strong promoter, Pg3, by −35 and −10 regions mutations, which was 1.63 times higher than that before mutation in B. subtilis. In addition, Feng et al. (2017) generated P43 promoter variants, which was 1.77 times higher than P43 promoter. Research has also shown that artificial dual-promoters are typically stronger than single promoters. For example, the dual promoter PgsiB-PhpaII was shown to be stronger than PhpaII, PyxiE, P43, PgsiB, Pluxs, or PaprE alone (Guan C. et al., 2016). In addition, the strength of the dual promoters PhpaII-PamyR and PhpaII-Pblma was 11- to 12-fold higher than the single promoter PhpaII in B. subtilis (Kang et al., 2010). Sinah et al. (2012) also constructed a set of two promoters for high protein expression in both E. coli and S. cerevisiae. Therefore, we combined strong single promoters to generate artificial multiple-promoter systems to increase the yield of the target protein. Pullulanase mRNA transcript levels and enzyme activity were significantly increased with the number of promoters (Figure 5).

PsodA+fusA+amyE was a semi-constitutive promoter constructed from PsodA and PfusA, which are constitutive promoters, and PamyE, which is a starch- and maltose-inducible promoter. Therefore, this triple promoter system could be induced in CSA medium and did not require an inducer. Constitutive promoters are advantageous in large-scale industrial production because they do not require an inducer. This simplifies the composition of the medium and the fermentation conditions, thus reducing production costs. Although the activity of pullulanase heterologously expressed under various conditions was as high as 580 U/mL (Nie et al., 2013) and 2523.5 U/mL (Zou et al., 2014) in E. coli, it was comparatively low in B. subtilis, at 5.7 U/mL (Chen et al., 2001), 2.82 U/mL (Wang et al., 2014), and 24.5 U/mL (Song et al., 2016). In this study, the yield of pullulanase was as high as 1,555 U/mL, which is the highest yield reported to date. In addition, we recently improved pullulanase activity to 2,180 U/mL by optimizing the medium composition and controlling the fermentation conditions.

Interestingly, the strain containing PsodA+hag+tufA had the highest mRNA expression, but not the highest enzyme activity (Figures 3, 5). In fact, the enzyme activity in the strain containing PsodA+hag+tufA was only 84.8% of that in the strain containing PdosA+fusA+amyE (Figure 3). This suggests that post-transcriptional modifications may modulate enzyme levels or activity. Alternatively, the overexpressed mRNA might not be used as a template for translation due to limited amounts of tRNA or ribosomes (Yuan and Wong, 1995; Wu et al., 1998). Further, the concentration of pullulanase protein was 14.3 g/L at 24 h, while its specific activity was only 10.2 U/mg. This indicates that some of the pullulanase might be misfolded and have no enzymatic activity (Li et al., 2004; Yan and Wu, 2017). In the future studies, we would like to determine the structure of pullulanase by NMR or X-ray crystallography to confirm protein misfolding and explore how changes in the culture conditions or chaperones can be employed to improve the folding rate.

In this study, four strong promoters from B. subtilis were identified by analyzing transcriptome data in GenBank, and these promoters were used to express pullulanase in B. subtilis BS001. Both gene expression and protein production increased significantly with increasing tandem combinations of promoters. The enzyme activity of the strain with the triple promoter complex PsodA+fusA+amyE reached 336.4 U/mL in a shake flask and 1,555 U/mL in a 50-L fermenter, which was 4.73 times higher than that of the strain with PamyE. The strain with PsodA+hag+tufA showed the highest mRNA levels, which were 53.5 and 37 times higher than that of the strain with PamyE at 24 and 48 h, respectively. Taken together, these results demonstrate that bioinformatic analysis in combination with genetic recombination technology can be used to develop microbial bioproduct advancements that can quickly and safely benefit industrial production in a cost-effective way. We will furtherly upgrade the expression strength by promoter mutation and optimize fermentation conditions in order to furtherly enhance the pullulanase production.

The datasets used and/or analyzed during the current study are available from the corresponding authors on reasonable request.

FM designed and performed the experiments, analyzed the data, and wrote the manuscript. XZ and TN constructed plasmids and transformed into host strain B.subtilis B001. FL and XB conceived the project, designed the experiments. YL and FT wrote a part of the Discussion section and helped with language editing. ZL designed the research content and analyzed the data. All authors read and approved the final manuscript.