The strains used here were purchased from the General Microorganisms Deposit Management Center of Jiangnan University (Supplementary Table S1). The tool plasmids used were previously deposited in our laboratory. The genes and primers were synthesized by Suzhou Jinweizhi Biotechnology Co., Ltd. Kits for genome extraction, plasmid extraction, DNA recovery and DNA purification were purchased from Shanghai Shenggong Co., Ltd. Restriction enzymes and DNA recombinases were obtained from Dalian TAKARA, Co., Ltd.

For gene mining, AS-A and AS-B from E. coli were selected as probes (Wang et al., 2005), and BLAST procedures were conducted in the NCBI database to search for AS-A and AS-B with different sequence identities. In order to obtain enzymes with different enzymatic properties, representative proteins with different sequence identities were selected for gene cloning, expression and analyses of enzyme activities.

Guided with the primers in Supplementary Table S2, different AS genes (asn) and polyphosphokinase 2 genes were amplified by polymerase chain reaction (PCR) using the genome of each strain as a template. Each asn was connected to pET28a and pMA5 and the recombinant plasmids were transformed into E. coli JM109 for gene cloning. Then the recombinant plasmids were extracted and transformed into E. coli BL21 (DE3) and Bacillus subtilis WB600, respectively. On the other hand, PPK2-Ⅲ genes from Deinococcus ficus and D. phoenicis were linked to pET21a by using DNA recombinase and transformed into E. coli strains.

The recombinant E. coli cells harboring expression plasmids were cultured in the TB liquid media supplementated with kanamycin (final concentration of 50 μg ml⁻¹) until the optical density (OD₆₀₀) reaching 0.4–0.6. Then isopropyl-β-d-Thiogalactopyranoside (IPTG) was added with a final concentration of 0.1 mmol L⁻¹ and incubated at 16°C for 24 h. The recombinant B. subtilis WB600 cells harboring expression plasmids were cultured in the SB liquid media with kanamycin (final concentration of 50 μg ml⁻¹) at 36°C for 24 h.

The collected cells were washed, resuspended in lysis buffer (20 mM Tris-HCl buffer, pH 8.0). Then cells were crushed by two rounds of ultrasound treatment and centrifuged at 12,000 rpm for 20 min at 4°C. Crude enzyme extract was prepared by filtering supernatant with a 0.22 μm microporous membrane for use. The crude extract was loaded onto a Ni-NTA column pre-equilibrated with buffer. The binding solution and eluent were used as the mobile phase for linear elution (eluent ratio: 0–100%), followed by SDS-PAGE analysis of the collected eluate.

The 5 ml reaction mixture contained 0.1 mol L⁻¹ Tris-HCl (pH 8.0), 0.1–0.5 mol L⁻¹ L-Asp, 0.1–0.8 mol L⁻¹ NH₄Cl, 1–10 mmol L⁻¹ ATP, 5–250 mmol L⁻¹ MgCl₂, 5–150 mmol L⁻¹ sodium hexametaphosphate and 1% Triton X-100. The reaction began with the addition of recombinant cells of AS and PPK2-Ⅲ. The reaction temperature is 25–45°C and the rotation speed is 180 rpm. Whole-cell catalytic reaction is terminated in a boiling water bath for 5 min.

The activity of AS was measured in 1 ml reaction system containing 100 mmol L⁻¹ Tris-HCl buffer (pH 8.0), 20 mmol L⁻¹ L-Asp, 10 mmol L⁻¹ MgCl₂, 10 mmol L⁻¹ NH₄Cl (or L-Gln), 10 mmol L⁻¹ ATP and 0.1 ml enzyme solution. After incubation at 37°C for 15 min, the sample was boiled for 5 min to terminate the reaction. Then the sample was centrifuged at 12,000 rpm and 4°C for 5 min, and the supernatant was filtered through a microporous (0.22 μm) membrane and poured into a sample bottle for content determination. The L-Asn content was determined by HPLC (High Performance Liquid Chromatography) with orthophthalaldehyde (OPA) on-line derivatization, and the amount of enzyme required to produce 1 μmol L-Asn from L-Asp per minute was defined as one enzyme activity unit.

The activity of PPK2-Ⅲ was measured in 1 ml reaction system containing 100 mmol L⁻¹ Tris-HCl (pH 8.0), 10 mmol L⁻¹ MgCl₂, 50 mmol L⁻¹ polyphosphate and 1 mmol L⁻¹ AMP. The reaction was conducted at 37°C for 10 min, and the sample was heated at 80°C for 10 min to terminate the reaction. After centrifuge at 12,000 rpm for 10 min, the supernatant was collected to determine the ATP content by ATP detection kit. As for the definition of activity of PPK2-Ⅲ, the amount of enzyme required to produce 1 μmol of ATP per minute is defined as a unit of enzyme activity.

Concentrations of L-Asn, L-Asp and glutamine were determined by HPLC on-line o-phthalaldehyde (OPA)-derivation method with an Agilent Zorbax Eclipse AAA column (4.6 × 150 mm, 5 μm). The control conditions were set as follows, column temperature 40°C, detection wavelength 338 nm, flow rate 1.0 ml min⁻¹, sample volume 10 μL. The mobile phase contained A (40 mmol L⁻¹ Na₂HPO₄, pH 7.8) and B (acetonitrile: methanol: water = 45: 45: 10). The linear elution programme represented as the proportion of mobile phase B was as follows, 5% at 0–2 min, 5%–57% at 2–12 min, 100% at 13–17 min, 5% at 18–24 min.

Through sequence alignment, the enzyme molecules exhibiting 20–100% identity with the amino acid sequence of the probe were selected in order to obtain a novel AS with different enzymatic property from the probe. Thereafter, eight genes of asnB and four genes of asnA were selected for clone and heterologous expression (Supplementary Table S1). Heterologous expression of these enzyme proteins in E. coli and B. subtilis is indicated by SDS-PAGE (Supplementary Figure S1 and Supplementary Figure S2). Theoretically, the molecular weights of AS-B and AS-A are ∼60 and ∼40 kDa, respectively. EcoAS-B, SplAS-B, HalAS-B and EcoAS-A were achieved high soluble protein expression and consistent with the theoretical value, while VruAS-B, SmaAS-B had less soluble protein expression but consistent with the theoretical value in E. coli. B. subtilis is considered to be a good host of protein soluble expression and was selected as host for these enzyme proteins expression. SDS-PAGE shows that HalAS-A, LsaAS-A, EcoAS-B, SplAS-B, HalAS-B, VruAS-B and SmaAS-B are expressed in soluble state in B. subtilis WB600 (Supplementary Figure S2). Enzyme activity assys in Table 1 shows B. subtilis is more suitable as a host for heterologous expression of these enzyme proteins, except for EcoAS-A. LsaAS-A expressed in B. subtilis produced highest enzyme activity (1356.7 mU·mL⁻¹) and this value increased to as high as 3850.0 mU·mL⁻¹ under optimal culture condition in triangular flask.

Figure 1A shows that the activity of LsaAS-A was temperature-dependent with an optimal temperature around 40°C. When temperature exceeded 50°C, the enzyme activity decreased rapidly. In order to obtain the half-life data of LsaAS-A, the first-order exponential decay formula was used to determine the inactivation rate constant by plot of the decay formula (Figure 1B). Through calculation by inactivation rate constant, the half-lives of LsaAS-A at 30, 40 and 50°C were 857.9, 94.6 and 51.0 min, respectively. In previous reports, the AS-A derived from E. coli had a half-life of less than 1 min at 45°C and was completely inactivated after being stored at 50°C for 5 min (Humbert and Simoni, 1980), suggesting that LsaAS-A was significantly more stable than E. coli-derivived AS-A. Moreover, effect of pH on LsaAS-A activity was also investigated by incubating the reaction solution under buffers with different pH. Figure 2A shows that LsaAS-A is pH-dependent with an optimal pH of 8.0. The suitable reaction buffer seems to be acetate buffer and Tris-HCl buffer, rather than boric acid buffer. In addition, pH stability of LsaAS-A was also investigated by storing LsaAS-A at 4°C for 24 h under different pH conditions and its residual activity was measured. LsaAS-A was stable in the acetate buffer at pH 5, while storage of LsaAS-A in buffer with pH 4 or 6–9 resulted in partial loss of enzyme activity. When storage pH was up to 10, only 29.8% of the residual enzyme activity was retained (Figure 2B). On the other hand, metal ions were added to the reaction system to investigate the effect of them on enzyme activity and Table 2 shows that Mg²⁺, Mn²⁺, Fe²⁺, and Zn²⁺ have an activation effect on the enzyme activity. Among them, Fe²⁺ has the highest activation performance (136%), while Zn²⁺ gave a relatively low activation effect (23.61%). The lost of LsaAS-A activity without addition of metal indicates that LsaAS-A is a metal-dependent enzyme. The main reason is that metal ion including Mg²⁺ binds to the C-terminal active site of AS and participates in the reaction (Richards and Schuster, 1998).

In order to obtain K_m and V_max, double reciprocal plot was used to determine these kinetic parameters and the results were shown in Table 3. The K _m of LsaAS-A for Asp is 0.78 mmol L⁻¹, which is similar to AS from Klebsiella aerogenes (Richards and Schuster, 1998), indicating that LsaAS-A has a high affinity for Asp. The K _m of LsaAS-A for NH₄ ⁺ is 6.5 mmol L⁻¹, which is similar to AS from Streptococcus bovis (Burchall et al., 1964). The K _m of LsaAS-A for ATP is 2.24 mmol L⁻¹, indicating that the affinity to ATP is high.

PPK2-III from Deinococcus ficus and D. phoenicis was cloned and expressed in RosettagamiB (DE3), Rosetta (DE3) and BL21 (DE3) of E. coli, respectively. Electrophoretic analysis shows that they were expressed effectively (Supplementary Figure S3) and their enzyme activities were shown in Table 4. Each host can express ppk2-III, while E. coli Rosetta (DE3) exhibits the best expression performance among three hosts. The reason may be that E. coli Rosetta (DE3) can enhance the soluble expression of genes containing rare codons when compared with E. coli BL21 (DE3). Although E. coli RosettagamiB (DE3) also has this function, its slow growth restricts the improvement of enzyme activity. Considering that the polymerization degree has a great influence on the activity of most PPK (Ishige et al., 2002; Motomura et al., 2014; Meng et al., 2016), polyphosphates with different polymerization degrees were used to evaluate the activity of DfiPPK2-Ⅲ. Among the measured polyphosphates, sodium hexametaphosphate and sodium polyphosphate could be used as substrates for DfiPPK2-Ⅲ to synthesize ATP, while sodium pyrophosphate and sodium tripolyphosphates was not suitable as phosphate donors (Table 5).

A coupling reaction system for L-Asn preparation was constructed using LsaAS-A and DfiPPK2-Ⅲ ATP regeneration system (Supplementary Figure S4) either including free enzymes or whole cells. From the point of view of application, free enzyme and whole cell have their own advantages and disadvantages. Table 6 shows that the catalysis efficiency of whole cell as a catalyst is higher than that of crude enzyme solution whether ATP or AMP was used as energy donor. The reason may be that endogenous PPase in living cell can eliminate the inhibitory effect of pyrophosphate (Suzuki et al., 2018; Lelièvre et al., 2020) and LsaAS-A and DfiPPK2-III remained in orininal cells were more stable than the free enzymes. Table 5 also indicates that the addition of surfactants to the whole-cell catalytic system can improve L-Asn titer, which may be ascribed to the fact that surfactant could enhance the permeability of the cell membrane and thereby improve the effect of mass transfer.

Considering that polyphosphate is the phosphate donor for regeneration of ATP via the catalysis of PPK2-Ⅲ, the amount of polyphosphate added in the reaction system were investigated. Figure 3A shows that L-Asn production is hexametaphosphate-dependent. When the concentration of sodium hexametaphosphate added in the reaction solution was lower than 60 mmol L⁻¹, the production of L-Asn increases with the enhancement of sodium hexametaphosphate concentration. While the concentration of sodium hexametaphosphate is greater than 60 mmol L⁻¹, the production of L-Asn gave an opposite trend. The main reason may be that high concentration of sodium hexametaphosphate will chelate with Mg²⁺ to reduce enzyme activity (Iwamoto et al., 2007). Furthermore, sodium hexametaphosphate is loaded with high negative charge and is able to combine with enzyme, thereby reducing enzyme activity (Kameda et al., 2001).

As mentioned in enzymatic properties of LsaAS-A, metal ions are necessary for LsaAS synthesis of L-Asn, so DfiPPK2_S also requires Mn²⁺ or Mg²⁺ as cofactors (Sato et al., 2007; Motomura et al., 2014). However, the present study showed that Mn²⁺, Fe²⁺ and Zn²⁺ had no stimulating effect on L-Asn production (data not shown). The reason may be that their strong binding force with sodium hexametaphosphate leads to the formation of precipitation, so these ions can not bind to LsaAS to play the role of cofactor (Larsen et al., 1999). As for Mg²⁺, Figure 3B shows that the situation is different. When the concentration of Mg²⁺ is no more than 200 mmol L⁻¹, the production of L-Asn is stimulated by the enhancement of the concentration of Mg²⁺. However, the titer of L-Asn would be reduced if Mg²⁺ exceeds 200 mmol L⁻¹ since chelation derived from high concentration Mg²⁺ is easy to genarate precipitation (Kameda et al., 2001).

In order to reduce costs, effect of initial ATP concentration (0–10 mmol L⁻¹) on whole-cell catalytic synthesis of L-Asn was investigated. Without the addition of ATP, only 3.56 mmol L⁻¹ of L-Asn was produced. When 6 mmol L⁻¹ of ATP was added, the L-Asn output reached 50.10 mmol L⁻¹ (Figure 3C). Theoretically, 1 mol ATP is consumed for every 1 mol L-Asn produced, so the ATP regeneration system has recycled ATP for 7.35 times. The final concentraion of L-Asn was not restricted by ATP when its intial concentraion was no less than 6 mmol L⁻¹, so addition amount of 6 mmol L⁻¹ is a suitable value in this ATP regeneration system.

Under the aforementioned optimal conditions, effects of L-Asp concentration (40–300 mmol L⁻¹) and NH₄ ⁺ concentration (0.1–0.7 mol L⁻¹) on the whole-cell catalytic synthesis of L-Asn were investigated. When the concentration of L-Asp is 100 mmol L⁻¹, the yield of L-Asn reaches the maximum (55.87 mmol L⁻¹). L-Asp concentration high than 100 mmol L⁻¹ would lead to reduction in L-Asn titer, which may be ascribed to substrate inhibition (Figure 3D). As for another substrate NH₄ ⁺, Figure 3E shows that enhanced concentration of NH₄ ⁺ improves the production of L-Asn, but 0.4 mol L⁻¹ may be an optimal value since more addition of NH₄ ⁺ is not able to bring about more production of L-Asn.

Above results indicate that addition of high concentration of sodium hexametaphosphate will cause inhibition in L-Asn production due to its chelation of Mg²⁺ and inhibition of enzyme activity. In order to release the inhibitory effect of sodium hexametaphosphate, a fed-batch strategy for addition of sodium hexametaphosphate was used. After 12 h, the conversion rate reached 90.15% (Figure 4A), which was nearly 80% higher than that of adding 180 mmol L⁻¹ sodium hexametaphosphate directly, indicating that feeding sodium hexametaphosphate in batches can suppress its negative effect (Kameda et al., 2001). Furthermore, the strategy of feeding L-Asp and sodium hexametaphosphate simultaneously was adopted in order to obtain higher yield of L-Asn. At 4, 8, 12, 16, and 20 h after the start of the reaction, L-Asp and sodium hexametaphosphate were fed at a final concentration of 60 mmol L⁻¹. At the end of the catalytic reaction, 218.26 mmol L⁻¹ L-Asn was produced with a yield of 64.19% (Figure 4B).