Optimization of the heme biosynthesis pathway for the production of 5-aminolevulinic acid in Escherichia coli

5-Aminolevulinic acid (ALA), the committed intermediate of the heme biosynthesis pathway, shows significant promise for cancer treatment. Here, we identified that in addition to hemA and hemL, hemB, hemD, hemF, hemG and hemH are also the major regulatory targets of the heme biosynthesis pathway. Interestingly, up-regulation of hemD and hemF benefited ALA accumulation whereas overexpression of hemB, hemG and hemH diminished ALA accumulation. Accordingly, by combinatorial overexpression of the hemA, hemL, hemD and hemF with different copy-number plasmids, the titer of ALA was improved to 3.25 g l−1. Furthermore, in combination with transcriptional and enzymatic analysis, we demonstrated that ALA dehydratase (HemB) encoded by hemB is feedback inhibited by the downstream intermediate protoporphyrinogen IX. This work has great potential to be scaled-up for microbial production of ALA and provides new important insights into the regulatory mechanism of the heme biosynthesis pathway.

important precursor, the biosynthesis of ALA is tightly regulated by the end product heme. In addition, the regulation mechanism of the heme biosynthesis pathway is more complex than expected 21 . Although the heme biosynthesis pathway enzymes have been well studied 3,22,23 , limited information on its regulation mechanism is available 3 .
In the present work, we systematically investigated the effect of overexpression of the downstream genes (hemB, hemC, hemD, hemE, hemF, hemG and hemH) on ALA accumulation. Interestingly, we discovered that in addition to the upstream genes hemA and hemL, hemB, hemD, hemF, hemG and hemH are also the major regulatory targets. Specifically, up-regulation of hemD and hemF was beneficial to ALA accumulation whereas overexpression of hemB, hemG and hemH was adverse to ALA accumulation. Through combinatorial overexpression of hemA, hemL, hemD and hemF, the titer of ALA was increased to 3.25 g l 21 . More importantly, in combination with transcriptional and enzymatic analysis, we demonstrated that ALA dehydratase (HemB) encoded by hemB is feedback inhibited by the downstream intermediate protoporphyrinogen IX.

Results
Up-regulation of hemD and hemF increased ALA accumulation. According to previous studies 24, 25 , it has been accepted that the biosynthesis of ALA is the rate-limiting step for heme biosynthesis and is tightly regulated in organisms including E. coli. In order to investigate the effects of downstream genes of the heme synthesis pathway on ALA accumulation, we individually overexpressed hemB, hemC, hemD, hemE, hemF, hemG and hemH with a lowcopy number vector pCDFDuet-1 in E. coli. Although the absolute production of ALA was low, distinct changes were observed with single overexpression of the above-mentioned genes (Fig. 2a). Up-

Glutamate
Glutamyl-tRNA GSA regulation of hemD or hemF increased ALA accumulation while reverse results were obtained when overexpressing hemB, hemG or hemH. In contrast, no significant differences were detected after overexpression of hemC or hemE. These findings indicated that up-regulation of the downstream genes hemD or hemF has a positive correlation on ALA accumulation whereas up-regulation of hemB, hemG or hemH has a negative correlation. More importantly, the results also suggested that in addition to hemA, hemL and hemF 3 , hemB, hemD, hemG and hemH are also the main regulatory points of the heme biosynthesis pathway.
To further validate the above results and in view of ALA biosynthesis is tightly regulated in E. coli 17 , the upstream metabolic flux towards ALA was increased by co-overexpression of the rate-limiting enzymes HemA s (a variant of HemA from S. arizona) and HemL (E. coli) in the above constructed recombinant strains. Compared to E. coli LA (hemL and hemA s ) (862.5 mg l 21 ), E. coli LAD (hemL, hemA s and hemD) and E. coli LAF (hemL, hemA s and hemF) produced more ALA (951.6 mg l 21 and 1,138.2 mg l 21 ), respectively ( Fig. 2b and Fig. 3a). In contrast, the recombinant strains E. coli LAG (hemL, hemA s and hemG), E. coli LAH (hemL, hemA s and hemH) and especially E. coli LAB (hemL, hemA s and hemB) exhibited remarkably reduced ALA production which were 242.3 mg l 21 , 580.6 mg l 21 and 39.2 mg l 21 , respectively ( Fig. 2b and Fig. 3a). Consistent with the above single expression results, no significant changes were observed when co-overexpressing hemC and hemE. The results confirmed that hemD and hemF are distinct from hemB, hemG and hemH in that they are beneficial to ALA production.
Additionally, overexpression of the above genes generated distinct effects on cell growth and glucose consumption. In particular, upregulation of hemB resulted in reduced biomass (Fig. 3b) and glucose consumption (Fig. 3c), which is likely due to the accumulation of the harmful intermediate porphobilinogen (PBG) and its derivatives 12 (Fig. 3d). Interestingly, overexpression of hemG resulted in more biomass coupled with a long lag phase ( Fig. 3b and Supplementary  Fig. S1), which is likely due to its direct involvement in energy generation 26 and the toxic intermediate protoporphyrin IX 27 .
Modular optimization of the committed enzymes to improve ALA production. On the base of the above analysis, it could be speculated that simultaneous up-regulation of hemD and hemF would further enhance ALA production. To verify this speculation, the upstream genes (hemA s and hemL) and the downstream genes (hemD and hemF) of the heme biosynthesis pathway were simultaneously overexpressed in E. coli. As expected, the resulting strain E. coli  LADF harboring plasmids pACYCDuet-1-hemL-hemA s -hemF and pCDFDuet-1-hemD produced about 1,227.4 mg l 21 of ALA at 36 h ( Fig. 4b), which was higher than that of the recombinant strains E. coli LAD (hemL, hemA s and hemD) and E. coli LAF (hemL, hemA s and hemF). To further fine-tune these four committed enzymes and improve the production of ALA, the compatible plasmids pRSFDuet-1 (high-copy number) and pETDuet-1 (medium-copy number) were used to investigate the impact of plasmid copy number on ALA production. According to the above results, the committed genes hemA s , hemL, hemD and hemF were distributed in three modules (Fig. 1). Specifically, due to HemA and HemL forming a tight complex, with a 151 ratio, to quickly catalyze glutamyl-tRNA to ALA 28 , the upstream genes hemL and hemA s were contained within one operon and co-overexpressed with the downstream genes hemD and hemF as described in Fig. 4a. As we expected, the recombinant strains LADF-1, LADF-2 and LADF-3 with comparatively low-level overexpression of hemL and hemA s resulted in dramatically decreased ALA accumulation which were 11.3 mg l 21 , 211.7 mg l 21 and 12.1 mg l 21 , respectively. In comparison, the recombinants LADF-4, LADF-5 and LADF-6 with high-level overexpression of hemL and hemA s lead to much higher titer of ALA which were 1,202.8 mg l 21 , 1,617.0 mg l 21 and 2,048.0 mg l 21 , respectively. Furthermore, compared with the recombinant LADF-4 (RSF-hemL-hemA s , pBR322-hemF-hemD), LADF-6 (RSF-hemL-hemA s -hemF, pBR322-hemD) produced more ALA (2,048.0 mg l 21 , Fig. 4b). The results suggested that high-level overexpression of the upstream genes hemA s and hemL as well as the downstream gene hemF, and moderate overexpression of the downstream gene hemD are favorable to ALA accumulation.
To evaluate the capacity of the recombinant strain LADF-6 for the production of ALA, a scaled-up fermentation experiment was carried out in a 3 l bioreactor with E. coli LA as the control (Fig. 4c). Obviously, compared with E. coli LA, LADF-6 showed much bigger capacity for ALA accumulation (Fig. 4d). At 32 h, the titer of ALA was increased to 3.25 g l 21 with a final productivity of 0.102 g l 21 h 21 which was 1.79-fold of that of E. coli LA (1.82 g l 21 ). The results further demonstrated that up-regulation hemD and hemF is beneficial to ALA accumulation and rational up-regulation of the upstream genes (hemA and hemL) and the downstream genes (hemD and hemF) is a good strategy for efficient synthesis of ALA. Moreover, the results also confirmed that the regulatory mechanism of the heme biosynthesis pathway is complicated and needs to be uncovered.  T7   T7  T7  T7   T7  T7  T7   T7  T7  T7   T7  T7  T7   T7  T7  T7   T7 T7 T7 The heme biosynthesis pathway is tightly regulated at the transcriptional level. To understand the underlying biology behind the different effects of overexpressing the above mentioned five key genes hemB, hemD, hemF, hemG and hemH on the heme biosynthesis pathway, we quantified the mRNA expression of these genes. As expected, single overexpression of these genes substantially increased the level of each respective mRNA transcript. Moreover, we observed that overexpression of hemB, hemG or hemH caused reduction of most of the pathway genes while overexpression of hemD resulted in significant increase of the all pathway genes (especially gltX and hemA) (Fig. 5), which are in agreement with ALA production (Fig. 2) and the results from the C4 pathway 29 . These findings suggested that hemB, hemG, hemH and hemD have central regulatory roles in the heme biosynthesis pathway and their transcriptional levels are tightly regulated. Although ALA accumulation was significantly increased with overexpression of hemF, no obvious transcriptional alterations of the pathway genes were detected (Fig. 5), suggesting that up-regulation of hemF resulted in other effects but not transcriptional changes in pathway genes. Compared to E. coli LA (hemL, hemA s ) and LAB (hemL, hemA s and hemB), E. coli LAF (hemL, hemA s and hemF) showed lighter color even with accumulation of more ALA (Fig. 3d). In addition, although up-regulation of hemG significantly decreased the transcriptional level of the genomic hemA gene, the recombinant strain E. coli LAG (hemL, hemA s and hemG) harboring the plasmid-borne genes hemL and hemA s still accumulated much less ALA compared with E. coli LA (hemL, hemA s ). Taken together, we concluded that HemB is a key control node and might be feedback inhibited by the intermediate protoporphyrinogen IX which catalyzed from coproporphyrinogen III by coproporphyrinogen III oxidase (HemF, encoded by hemF) (Fig. 1).
HemB is a feedback inhibitory node of the heme biosynthesis pathway. To further validate our speculation, the HemB activity of the recombinant strains E. coli LA (hemL, hemA s ), E. coli LAF (hemL, hemA s and hemF), E. coli LAG (hemL, hemA s and hemG) and E. coli LAB (hemL, hemA s and hemB) were directly examined (Table 1). Compared with E. coli LAB (52.04 U mg 21 ), E. coli LA exhibited much lower activity (6.97 U mg 21 ) which further confirmed that expression of HemB is tightly regulated to maintain a relatively low level in E. coli. Notably, E. coli LAF with overexpression of hemF showed an obvious decline of the HemB activity (3.34 U mg 21 ), which supported our speculation that HemB is feedback inhibited by the intermediate protoporphyrinogen IX (Fig. 1). Furthermore, E. coli LAG with overexpression of hemG gave rise to a substantial increase in HemB activity (22.21 U mg 21 ), which might be attributed to the direct activation by protoporphyrin IX or the weakened inhibition that caused by decreased protoporphyrinogen IX (Fig. 1).
To further confirm the above speculation and due to the unavailability of the intermediate protoporphyrinogen IX, we quantified HemB activity of the E. coli LAB strain extracts with addition of protoporphyrin IX in vitro to indirectly determine which of the two possibilities is correct. As shown in Table 1, activation of HemB following the addition of protoporphyrin IX was not detected. In combination with the above results from E. coli LAF, our speculation that HemB is a key regulatory node of the heme biosynthesis pathway and feedback inhibited by the intermediate protoporphyrinogen IX (Fig. 1) was demonstrated, which in turn provided the reason why high-level overexpression of hemF enhanced the production of ALA.

Discussion
Because of its numerous potential applications in medicine and agriculture, market demand of the valuable compound ALA is rapidly rising. Over the past several years, much research has been dedicated to optimizing whole-cell catalytic synthesis of ALA 1 by overexpression of ALAS. Although metabolic engineering has been widely used as a powerful tool for construction and optimization of the target pathways towards various valuable compounds [30][31][32][33][34]   ies have made significant contributions towards elucidating the regulatory mechanism of the heme biosynthesis pathway 1,2,35,36 (Fig. 1), while the structure-function relationship for all the heme biosynthesis enzymes have been well understood 23 . Therefore, exploration of the heme synthesis pathway would not only be beneficial to rational engineering towards ALA production but also invaluable for uncovering regulatory mechanism of this highly conserved pathway. Although the transcriptional model of hemA and hemF has been previously examined under different conditions 3 , the overall regulatory mechanism of heme biosynthesis pathway genes is still vague. In this study, we discovered that in addition to HemA, HemL and HemF 3,35 , HemB, HemD, HemG and HemH are also the key regulatory targets of the heme biosynthesis pathway. Overexpression of these genes showed significant effects on cell growth and metabolism ( Fig. 3b and 3c) suggesting that they are tightly controlled under normal conditions. Recent studies have reported that hemB and hemH are negatively regulated by RyhB (an iron-associated small non-coding RNA) at the post-transcriptional level 37 . Moreover, by investigation at transcriptional and protein levels, we demonstrated that HemB is feedback inhibited by the intermediate protoporphyrinogen IX (Fig. 1), which suggest that regulation of heme biosynthesis pathway genes is far more complicated than imagined. In addition, Kang et al. previously discovered that overexpression of the native gltX gene (glutamyl-tRNA synthetase, GluRS) resulted in dramatic decrease of ALA 17 . Here, we also confirmed that overexpression of gltX and artificial addition of glutamate lead to reduced ALA accumulation and cell growth (Supplementary Fig. S2). The reason might be attributed to the synthesis of glutamyl-tRNA Gln (harmful to cell growth) that catalyzed by the nonspecific GluRS 35,38 .
In the field of metabolic engineering, it was generally accepted that increase and balance of the objective synthetic pathways are the most critical parameters for high-yield production of the endproduct 34,39,40 . For this purpose, many combinatorial optimization strategies have been developed and applied to assemble and regulate expression of the key genes [41][42][43][44][45] . In this study, after investigation and identification of the four positive key genes towards ALA, the heme biosynthesis pathway was further optimized with two compatible plasmids pRSFDuet-1 (high-copy number) and pETDuet-1 (medium-copy number) according to previous results. As we expected, the production of ALA was significantly increased (Fig. 4d) with an optimized combination of the increased flux towards ALA. One side, the results demonstrated that moderate expression of HemD is crucial to ALA accumulation since overexpression of HemD not only up-regulates the upstream genes ( Fig. 5) but also draws more flux to downstream reactions from ALA (Fig. 1). Moreover, the results were consistent with our finding that HemB is a key regulatory point of the heme biosynthesis pathway and feedback inhibited by protoporphyrinogen IX. Previously, it has been reported that a hemB mutant E. coli strain failed to increase the production of ALA 46 indicating that low activity of HemB is essential for cell growth and ALA accumulation. In comparison, the strategy of down-regulating HemB with overexpression of HemF was an alternative to improve ALA production.
In conclusion, through investigation of the heme biosynthesis pathway genes we discovered that hemB, hemD, hemF, hemG and hemH are also the main regulatory targets of the heme biosynthesis pathway. More importantly, overexpression of hemD and hemF increased the accumulation of the upstream intermediate ALA. By combinatorial overexpression of hemA, hemL, hemD and hemF with different copy-number plasmids, the titer of ALA was increased from 862.5 mg l 21 to 3.25 g l 21 . Furthermore, in combination with transcriptional and enzymatic analysis, we demonstrated that HemB is feedback inhibited by the downstream intermediate protoporphyrinogen IX, which provides new important insights into the regulatory mechanism of the heme biosynthesis pathway. To further increase the production of ALA, comparative investigation of different E. coli hosts and overexpression of the ALA transporter RhtA 17 should be available. In addition, dynamic regulation of HemB with synthetic regulatory elements or circuits 40,47,48 should also be desirable for efficient production of ALA.

Methods
Strains and plasmids. Bacterial strains and plasmids used in this study are described in Supplementary Table S1 and Table S2, respectively. Specifically, E. coli BL21 (DE3) was used as host for gene expression due to the T7 RNA polymerase. E. coli JM109 was used for DNA manipulations and plasmids construction.
Media and culture conditions. Luria-Bertani (LB) medium (g l 21 ) composed of tryptone 10.0, yeast extract 5.0 and NaCl 10.0 was used for the DNA manipulation process and seed cultures. Agar (2.0%) was added when a solid medium was required. The modified minimal medium (g l 21 ), which contains glucose 20.0, yeast extract 2.0, (NH 4 ) 2 SO 4  The batch culture was performed in a 3 l fermentor (BilFlo 115, New Brunswick Scientific Co., Edison, NJ, USA). 2.0% inoculation of the seed culture was transferred into the fermentor with approximate 1.5 l medium after being cultured at 37uC for about 12 h. Chloramphenicol (20 mg ml 21 ) and streptomycin (50 mg ml 21 ) or ampicillin (100 mg ml 21 ) and kanamycin (50 mg ml 21 ), glucose (35.0 g l 21 ) and IPTG with a final concentration of 0.1 mM to induce the expression of genes were initially added to the medium. Agitation speed was 400 r min 21 and aeration rate was 1.0 vvm. The cultures were incubated at 37uC and pH was maintained at approximate 6.5 by adding 4.0 mol l 21 NaOH.
Analytical procedures. All engineered E. coli strains were cultured in the modified minimal medium containing appropriate antibiotics. For growth studies, optical densities (OD) of the cell were measured at a wavelength of 600 nm with a UV-1,700 spectrophotometer (Shimadzu, Kyoto, Japan). Glucose concentration in the supernatant was detected using a glucose-glutamate analyzer SBA-40C (Biology Institute of Shandong Academy of Sciences, Jinan, China). The production of ALA was analyzed using the Modified Ehrlich's Reagent after the cultures were centrifuged 49 .
Enzyme assays of HemB. For detecting HemB activity, cells were harvested after addition of IPTG following 8 h of culture by centrifugation for 10 min (10,000 r min 21 , 4uC). Cells were washed twice with disodium hydrogen phosphate-citric acid buffer (pH 6.7), then they were resuspended in the above buffer and disrupted for 5 min using an ultrasonic oscillation (Sonics VCX750, amplitude 25%). After removing the cellular debris by centrifugation (10,000 r min 21 , 4uC), the supernatant was analyzed for enzyme activity. Enzyme activity was quantified using the method described previously without the addition of dithiothreitol (DTT) to the reaction mixture 50 . HemB enzyme activity of one unit was defined as the amount of enzyme required to consume 1 pmol of ALA per min.
Quantitative real-time PCR (qRT-PCR) analysis. Cultures used for RNA extraction were cultivated for approximately 5 h after the addition of 0.1 mM IPTG. Cell quantities corresponding to approximately 1.5-2.0 OD600 nm were harvested and frozen immediately at 280uC. Total RNA of all E. coli strains was extracted using the RNA Extraction Kit (Takara, Dalian, China) according to the manufacturer's instructions. The quantity and purity of the RNA was determined using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA) by optical density measurements at 260 and 280 nm.
The RNA level was measured by qRT-PCR. Genes and their respective primer sequences that were used for qRT-PCR studies are listed in Supplementary Table S3. gapA encoding D-glyceraldehyde-3-phosphate dehydrogenase was selected as internal standard. The cDNA templates used for qRT-PCR were obtained by reverse transcribing mRNA transcripts using PrimeScript TM RT-PCR Kit (Takara, Dalian, China). Gene expression analysis via qRT-PCR was carried out in a 96-well plate with a total reaction volume of approximately 20 ml using a SYBRH Premix Ex Taq TM (Takara, Dalian, China) according to manufacturer's specifications. Reactions were performed with a LightCycler 480 II Real-time PCR instrument (Roche Applied Science, Mannheim, Germany).