Development of a Lipase Fermentation Process That Uses a Recombinant Pseudomonas alcaligenes Strain

Abstract

Pseudomonas alcaligenes M-1 secretes an alkaline lipase, which has excellent characteristics for the removal of fatty stains under modern washing conditions. A fed-batch fermentation process based on the secretion of the alkaline lipase from P. alcaligenes was developed. Due to the inability of P. alcaligenes to grow on glucose, citric acid and soybean oil were applied as substrates in the batch phase and feed phase, respectively. The gene encoding the high-alkaline lipase from P. alcaligenes was isolated and characterized. Amplification of lipase gene copies in P. alcaligenes with the aid of low- and high-copy-number plasmids resulted in an increase of lipase expression that was apparently colinear with the gene copy number. It was found that overexpression of the lipase helper gene, lipB, produced a stimulating effect in strains with high copy numbers (>20) of the lipase structural gene, lipA. In strains with lipA on a low-copy-number vector, the lipB gene did not show any effect, suggesting that LipB is required in a low ratio to LipA only. During scaling up of the fermentation process to 100 m³, severe losses in lipase productivity were observed. Simulations have identified an increased level of dissolved carbon dioxide as the most probable cause for the scale-up losses. A large-scale fermentation protocol with a reduced dissolved carbon dioxide concentration resulted in a substantial elimination of the scale-up loss.

One of the most persistent problems with laundry cleaning is the removal of fatty stains. A combination of high-temperature and high-alkalinity washing conditions can be used to emulsify and subsequently remove the fat-containing dirt. However, these conditions lead to damage to the fabric, and they require the input of large amounts of energy. At lower washing temperatures, e.g., 40°C, fatty-stain removal is very poor unless lipases are added.

The lipase from Pseudomonas alcaligenes, which is characterized by high alkalinity, has particularly advantageous characteristics under modern washing conditions (23). The gene for this P. alcaligenes lipase has been tested for heterologous expression in a variety of standard industrial production hosts, such as Bacillus licheniformis, Escherichia coli, Streptomyces lividans, Aspergillus niger, and Kluyveromyces lactis (unpublished observations). However, attempts to express this enzyme to commercially acceptable levels have been unsuccessful. Similar observations were made for lipases originating from other Pseudomonas species. The nature of this problem might reside in the complex process of folding and secretion of these lipase enzymes (LipA). These lipases require a specialized helper protein, LipB (9, 14, 17), and the involvement of proteins of the outer membrane secretion machinery (Xcp proteins) mediating the secretion (35). Although the precise role of LipB (also known as LipH, LimL, or Lif) is unclear at present, it has been suggested that it functions as a chaperone guiding the formation of correctly folded lipase. It has been claimed that a molar ratio between LipB and LipA of 1:1 (1) or even 4:1 (15) is needed for optimal functioning of the helper protein.

In order to develop a fermentation process for this lipase, we decided to optimize the expression of the gene in its original host, P. alcaligenes. Starting with very limited knowledge of the genetics and physiology of this organism, we have now made a large-scale production process. In this contribution the development of the fermentation process is described. Since P. alcaligenes cannot grow on glucose, a fermentation based on alternative substrates has been developed. Furthermore, the effects of amplification of lipase gene (lipA) and the lipase helper gene (lipB) on lipase production during fermentation have been tested. Finally, the prevention of scale-up losses by adaptations in the process conditions is described.

MATERIALS AND METHODS

Strains and plasmids.

For this study bacteria from the species P. alcaligenes and E. coli were used. The genetic modifications that have been introduced are described in Table 1. Plasmids used in this study are described in detail in Table 2.

TABLE 1.

Bacterial strains

Strain	Propertiesa	Reference
P. alcaligenes
M-1	Wild type; CBS473.85	This study
Ps93	Res− Mod+	This study
Ps224	Tetr (10 mg liter−1), pLAFlipA	This study
Ps225	Tetr (10 mg liter−1), pLAFlipAB	This study
Ps495	Neor (20 mg liter−1), pJRDlipAB	This study
Ps496	Neor (20 mg liter−1), pJRDlipA	This study
E. coli JM109	recA supE44 endA1 hsdR17 gyrA96 relA1 thi Δ(lac-proAB) F′ traD36 proAB+ lacIqlacZΔM15	39

^aRes⁻, restriction negative; Mod⁺, modification positive. 

TABLE 2.

Plasmids

Plasmid	Properties	Reference
pLAFR3	Tetr; IncP replicon	32
pJRD215	Neor Strr; IncQ replicon	5
pTZ18R	Ampr; multiple cloning site	22
pTZlipA	Ampr; 2.0-kb PvuII chromosomal fragment of P. alcaligenes containing lipase gene in SmaI-opened pTZ18R	This study
pTZlipAB	Ampr; 2.4-kb EcoRI-HindIII fragment containing lipase gene and lipase helper gene of P. alcaligenes in EcoRI-HindIII-opened pTZ18R	This study
pLAFlipA	Tetr; IncP replicon; 2.0-kb EcoRI-HindIII fragment derived from pTZlipA inserted into pLAFR3	This study
pLAFlipAB	Tetr; IncP replicon; 2.4-kb EcoRI-HindIII fragment derived from pTZlipAB inserted into pLAFR3	This study
pJRDlipA	Neor; IncQ replicon; 2.0-kb SstI-HindIII fragment derived from pTZlipA inserted into pJRD215	This study
pJRDlipAB	Neor; IncQ replicon; 2.4-kb SstI-HindIII fragment derived from pTZlipAB inserted into pJRD215	This study

Media.

For small- and large-scale DNA isolations P. alcaligenes and E. coli strains were propagated in 2× TY medium (16 g of Bacto tryptone liter⁻¹, 10 g of Bacto yeast extract liter⁻¹, 5 g of NaCl liter⁻¹, pH 7.0) supplemented with the appropriate antibiotic (Table 1) (tetracycline, 10 mg liter⁻¹; neomycin, 20 mg liter⁻¹). Colonies were plated on 2× TY agar plates or on tributyrin agar plates. Tributyrin agar plates were composed of buffered (0.2 M HEPES [Sigma]) minimal medium containing, per liter, 0.6 g of citrate, 0.6 g of K₂HPO₄, 1.0 g of (NH₄)₂SO₄, 0.3 g of MgSO₄ · 7H₂O, 0.08 g of CaCl₂ · 2H₂O, and 10 ml of tributyrin oil (Merck), supplemented with the appropriate antibiotics for selection (Table 1) and solidified with 1.5% agar (Difco).

Recombinant DNA techniques.

Chromosomal DNA from P. alcaligenes was isolated according to the method of Andreoli (2). Plasmid DNAs of E. coli and P. alcaligenes were obtained by the method of Birnboim and Doly (3) unless otherwise stated.

Isolation of mRNA from P. alcaligenes was done as follows. An aliquot of 1 ml of fermentation broth was diluted in 9 ml of milliQ water. The cells were spun down in a centrifuge (4,500 × g). The pellet was homogenized in 10 ml of a solution containing 3 M LiCl and 6 M urea with an Ultra-turrax blender (20,000 rpm) for 1 min on ice. The suspension was centrifuged (4,500 × g), and the homogenization procedure was repeated for 45 s on ice. Subsequently, the suspension was centrifuged in an ultracentrifuge at 140,000 × g for 30 min at 4°C. The pellet was dissolved in 3 ml of 1× FEB, containing 0.1 M NaCl, 10 mM Tris-HCl (pH 9.0), 0.5% sodium dodecyl sulfate, and 5 mM EDTA. To the suspension was added 3 ml of hot phenol solution (phenol–chloroform-isoamyl alcohol–3× FEB; 50:48:1:2.5, vol/vol). The solution was mixed thoroughly and centrifuged (4,500 × g) at room temperature. The upper layer was collected, and 1 volume of chloroform solution (chloroform-isoamyl alcohol; 24:1, vol/vol) was added, mixed thoroughly, and centrifuged (4,500 × g) at room temperature. The last two steps were repeated twice. Furthermore, an ethanol precipitation was performed on the water layer, and an appropriate volume of RNase-free water was added to the sample.

Restriction endonucleases and other enzymes were purchased from GIBCO-BRL and used according to the manufacturer’s instructions. For Southern hybridizations, transfers of chromosomal DNA fragments and colonies to nitrocellulose filters were done by the method of Southern (31). For Northern hybridizations, agarose gel electrophoresis of RNA was performed with addition of formaldehyde to the gel matrix, and transfer to nitrocellulose filters was done by the method described by Maniatis et al. (20). Detection of both DNA- and RNA-containing filters was executed with DNA probes labelled with [α-³²P]dATP (Amersham), using the random-primer-labelling kit from Pharmacia according to the instructions of the supplier.

Sequencing was carried out on an Applied Biosystems 373A DNA sequencer with the ABI PRISM Dye Termination Cycle Sequencing Ready Reaction Kit and AmpliTaq DNA polymerase.

Electroporation and isolation of restriction-negative, modification-positive strain Ps93.

Electroporation of P. alcaligenes M-1 with DNA isolated from E. coli gave only 1 to 10 transformants per μg, indicating a very low transformation frequency, whereas electroporation with plasmid DNA isolated from P. alcaligenes resulted in a higher transformation frequency. Because of this observation, we reasoned that there could be a selection for strains that are restriction negative after electroporation. Therefore, 24 independent transformants obtained after electroporation of P. alcaligenes M-1 with plasmids isolated from E. coli were cured from their plasmid. In these strains an equal amount of plasmid pJRD215 (5) isolated from E. coli was introduced by electroporation. Two strains showed a transformation frequency of 10⁵ colonies per μg of DNA, and one of the strains was named strain Ps93. DNA isolated from strain Ps93 and reintroduced in P. alcaligenes M-1 demonstrated the same high transformation frequency, indicating that strain Ps93 is a restriction-negative, modification-positive mutant of P. alcaligenes M-1.

Electroporation of plasmid DNA to E. coli was carried out as described by Calvin and Hanawalt (4). For electroporation of P. alcaligenes strains, the method of Wirth et al. (37) was used, with the modification that all treatments were performed at room temperature. Transformants of E. coli or P. alcaligenes with plasmid pLAFR3 were selected on 2× TY agar plates with 10 mg of tetracycline liter⁻¹. Cells of E. coli or P. alcaligenes transformed with pJRD215 were selected on 2× TY agar plates with 20 mg of neomycin liter⁻¹.

Cloning of the lipase gene.

Purification of the lipase of P. alcaligenes M-1 from the growth medium was performed as described by Stuer et al. (33). After gel electrophoresis and blotting on Immobilon transfer membrane, the N-terminal amino acid sequence was determined (21). The N-terminal sequence was found to be GLFGSTGYTKTKYPIVLTHGMLGF. …

The following oligonucleotide probe, matching residues 6 to 16 (according to the preferential codon usage in Pseudomonas aeruginosa [36]) was prepared: 5′-ACC GGC TAC ACC AAG ACC AAG TAC CCG/C ATC GT-3′. This oligonucleotide was used to hybridize chromosomal DNA of P. alcaligenes M-1 digested with several restriction endonucleases. Chromosomal PvuII and BclI fragments (between 1 and 4 kb) were cloned in the multiple cloning site of vector pTZ18R. Positive clones were detected from the library with PvuII fragments by colony hybridization with the oligonucleotide matching the N-terminal sequence of M-1 lipase.

Lipase assay.

Lipase activity in growth media was assayed titrimetrically as described by Gilbert et al. (12) with some modifications. A pH-Stat unit (Radiometer type ETS 822) was used at pH 9.0 and at 30°C with a standard olive oil emulsion (Sigma) diluted to 10% (vol/vol) and supplemented with 20 mM NaCl and 10 mM CaCl₂. One lipase unit (LU) was defined as the amount of lipase capable of releasing 1 μmol of titratable fatty acid per min under the assay conditions used.

Lipase detection on plates was done with the use of minimal medium (see “Media” above) agar with 1% tributyrin oil (Merck). Plates were incubated at 37°C during 48 to 72 h, and clearing zones indicated lipolytic activity.

Controlled fed-batch fermentation process.

P. alcaligenes was cultured either in laboratory fermentors with a 10 liter capacity or in 100-liter, 4-m³, or 100-m³ production fermentation vessels. The fermentors were inoculated with full-grown seed cultures that were grown at 35°C for 16 h on a yeast extract medium (10 g of yeast extract liter⁻¹, pH 7.0). The inoculation percentage used on all scales was 5%. For the main fermentation a defined minimal salt medium was used. The medium contained (per liter) K₂HPO₄, 6.4 g; (NH₄)₂SO₄, 3.33 g; MgSO₄ · 7H₂O, 3.2 g; CaCl₂ · 2H₂O, 0.8 g; CoCl₂ · 6H₂O, 40 mg; MnSO₄ · 1H₂O, 32 mg; FeSO₄ · 7H₂O, 20 mg; ZnSO₄ · 7H₂O, 12 mg; CuSO₄ · 5H₂O, 3 mg; H₃BO₃, 3 mg; Na₂MoO₄ · 2H₂O, 3 mg; and KI, 1 mg. The fermentation was run at 35°C. The pH of the culture was maintained automatically at 7.0 ± 0.1 by using sulfuric acid and ammonia as titrants. Foam was controlled automatically by use of a polyalkylene-silicone mixture (SAG 5693 silicone antifoam; Union Carbide). As a carbon source, citric acid (6 g liter⁻¹) was added. When the citric acid was fully consumed, 10 to 16 h after inoculation, a feed of soybean oil (Cargill) was applied at the rate of 1 g of soybean oil per liter of broth per h. This feed rate was kept constant for the rest of the fermentation. The fermentation was run for 48 to 100 h. Fully aerobic conditions (dissolved oxygen tension, >20% air saturation) were maintained throughout by injecting air, at a rate of 1 standard liter of air per liter of broth per min (= 1 vvm), into the region of the impeller that was rotating at about 400 rpm.

For scaling up the small-scale process to 100-m³ production fermentors, 100-liter and 4-m³ fermenters were used as intermediates. The medium, the temperature, the pH, and the soybean oil feed rate did not vary with scale. Only the aeration changed with varying scale, being 1 vvm at the 10-liter scale, 2 vvm at 100 liters, 0.5 vvm at 4 m³, and 0.4 vvm at 100 m³ to prevent excessive foaming and a high holdup. Except for the 10-liter scale, a back pressure of approximately 50 kPa was applied.

RESULTS

Lipase fermentation process.

In order to develop an economical production process for the microbial lipase produced by P. alcaligenes, a fermentation process under fed-batch control was developed. The inability of P. alcaligenes to metabolize sugars like glucose restricted the choice of suitable (industrial) carbon sources. Citrate was selected as the preferred initial carbon source during the first phase of the fermentation when the cells were growing batchwise. This was based partly on the fact that citrate enabled the cells to grow at high specific growth rates (μ ≥ 0.9 h⁻¹), which resulted in a relatively short batch phase (10 h). Furthermore, citrate kept the metal ions in solution during sterilization and the initial stages of the fermentation. When the citrate was fully consumed, the oxygen uptake rate sharply decreased (Fig. 1). At that moment a carbon-limited feed of soybean oil was applied to the culture to construct a fed-batch process, where the temperature, pH, and dissolved oxygen concentration were controlled.

FIG. 1.

Production of lipase by P. alcaligenes M-1 in a fed-batch fermentation process. After an initial exponential growth period, the process was controlled by a constant feed rate of soybean oil, which resulted in the oxygen uptake rate (OUR) (no symbol), biomass profile (▴), lipase activity (•), and yield profile of lipase on soybean oil (Y_P/S) (□) shown. The feed was started at 10 h after the initial carbon source (citrate) was fully consumed and the OUR was down to almost zero.

The characteristic parameters of this standard fermentation of the wild-type P. alcaligenes M-1 are exemplified in Fig. 1. Hardly any synthesis of lipase took place during batch phase growth on citrate. Only after the start of the soybean oil feed was lipase expression induced. It should also be stressed that a significant development of biomass occurred during the soybean oil feed phase. The highest yield of lipase on soybean oil (Y_P/S) was detected halfway through the fermentation. Until that time, the soybean oil was primarily converted into biomass, lipase, and carbon dioxide. Thereafter, the lipase synthesis leveled off rapidly and the major part of the soybean oil was used for maintenance purposes besides some growth. As a result, the Y_P/S decreased dramatically.

Characterization of the lipase gene.

As a result of the screening with the oligonucleotide corresponding to the N-terminal sequence of M-1 lipase, a positively reacting clone containing a 2.0-kb PvuII fragment (Fig. 2A) was detected. The 2.0-kb PvuII fragment from the P. alcaligenes M-1 chromosome was cloned in SmaI-opened E. coli vector pTZ18R (pTZlipA) (Fig. 2A). This fragment was found to contain an open reading frame that encodes the same NH₂-terminal peptide as determined for the lipase from P. alcaligenes M-1. Therefore, we have named the gene lipA. The encoded protein shows homology to other Pseudomonas lipase gene products (Fig. 3), with the highest homology (77% amino acid identity) to the lipase from P. aeruginosa (14, 38).

FIG. 2.

Genetic organization of the lipase operon and plasmid construction. (A) Construction of plasmid pTZlipAB and map of the chromosomal fragments of P. alcaligenes that are present on plasmids pTZlipA, pTZlipB, and pTZlipAB. Restriction endonuclease sites used for cloning: B, BclI; E, EcoRI; H, HindIII; P, PvuII; S, SstI; V, EcoRV. lipA and ′lipA, lipase gene and part of the lipase gene; lipB and lipB′, lipase helper gene and part of the lipase helper gene; bla, β-lactamase gene. (B) Genetic organization of the lipase operon. lipA, lipase gene; lipB, lipase helper gene; PR, lipase promoter sequence; ATT, attenuator sequence; TER, terminator sequence. The arrows represent the two lipase mRNAs, one starting from the transcription start and ending at the attenuator signal (1.2 kb in length) and the other starting from the transcription start and ending at the terminator (2.1 kb in length).

FIG. 3.

Sequence alignment of lipases from Pseudomonas rRNA group I and group II. All residues identical to those in the P. alcaligenes protein sequence are shown as white letters. The alignment was carried out by using the Clustal method of the WINSTAR software package. LIP-PALC.PRO, P. alcaligenes (this study); LIP-PAER.PRO, P. aeruginosa (38); LIP-BGLU.PRO, B. glumae (8); LIP-BCEP.PRO, B. cepacia (17).

The second incomplete open reading frame shows homology to a gene, lipB, found to be involved in the expression of lipase in other pseudomonads. In order to determine the complete sequence for lipB, the 2.0-kb PvuII fragment was used as a probe to hybridize with a library of BclI fragments. A colony containing a 1.7-kb BclI fragment could be identified as positive in this experiment. This 1.7-kb BclI fragment (pTZlipB) (Fig. 2A) harboring the complete second open reading frame was isolated and fused to the overlapping 2.0-kb PvuII fragment by using a common EcoRV site (Fig. 2A). Subsequently, a 2.4-kb fragment with both lipA and lipB could be subcloned in pTZ18R (pTZlipAB) (Fig. 2A), and its complete sequence was determined. The amino acid sequence of the putative LipB homolog was derived and was compared with those of other lipase helper gene products of Pseudomonas origin (Fig. 4). The N-terminal 25 residues show high hydrophobicity, suggesting that LipB is anchored to the membrane.

FIG. 4.

Sequence alignment of lipase helper proteins from Pseudomonas rRNA group I and group II. All residues identical to those in the P. alcaligenes protein sequence are shown as white letters. The alignment was carried out by using the Clustal method of the WINSTAR software package. LIPBPALC.PRO, P. alcaligenes (this study); LIPBPAER.PRO, P. aeruginosa (16); LIPBBGLU.PRO, B. glumae (9); LIPBBCEP.PRO, B. cepacia (17).

Effect of gene copy number on lipase production.

The 2.0-kb lipA fragment from pTZlipA was used to construct the high-copy-number plasmid pJRDlipA (Table 2) and the low-copy-number plasmid pLAFlipA (Table 2) in order to study the effect of gene copy number on lipase expression.

The transformation of the wild-type P. alcaligenes strain by the electroporation method with DNA isolated from the restriction-negative P. alcaligenes strain Ps93 (see Materials and Methods) gave a transformation frequency of 10⁵ colonies per μg of DNA for both plasmids pLAFlipA and pJRDlipA. In this way, P. alcaligenes Ps224 and Ps496, containing pLAFlipA and pJRDlipA, respectively, were made. On tributyrin plates these strains exhibited significantly larger halos than the wild-type strain. The P. alcaligenes strains were grown in a fermentor at a 10-liter scale by using the standard protocol. The lipase expression levels are shown in Fig. 5. It can be seen that there is a clear gene dose effect, with the high-copy-number plasmid (pJRDlipA)-containing strain exhibiting the highest lipase expression.

FIG. 5.

Lipase production by various plasmid-containing strains at the 10-liter fermentation scale, showing the effects of low- and high-copy-number plasmids on lipase production levels. The lipase production of the wild-type P. alcaligenes strain M-1 (1.5 MLU/liter) is used as the standard and fixed at 100%. Bars: 1, no plasmid (strain M-1); 2, pLAFlipA (strain Ps224); 3, pLAFlipAB (strain Ps225); 4, pJRDlipA (strain Ps496); 5, pJRDlipAB (strain Ps495).

Effect of the helper gene.

In order to investigate the effect of the helper gene (lipB) on the lipase production, two new constructs were made. As the 2.0-kb PvuII fragment does not harbor the complete open reading frame for the putative helper protein, the 2.4-kb fragment harboring both open reading frames was made and used to construct pLAFlipAB (Table 2) and pJRDlipAB (Table 2). This results in P. alcaligenes Ps225 and Ps495, respectively. Fermentation results for these strains at the 10-liter scale are shown in Fig. 5.

In order to verify the expression of the second open reading frame, we have analyzed mRNAs on Northern blots with the 2.4-kb fragment as a probe. In Fig. 6 strains Ps496 and Ps495 are compared. It can be seen that the strain without additional lipB gene copies (Ps496) shows only a band of 1.2 kb, whereas strain Ps495 shows, besides the strong 1.2-kb band, an additional longer mRNA of 2.1 kb, which corresponds to the full length of the lipAB operon. The 1.2-kb band corresponds in size exactly to a transcript running from the promoter to the hairpin structure present between the lipA and lipB open reading frames (Fig. 2B). The 2.1-kb fragment would correspond to a transcript from the promoter to the terminator structure downstream of the lipB gene (Fig. 2B).

FIG. 6.

mRNA isolations (see Materials and Methods) from fermentation broth analyzed by using the Northern blot procedure with the 2.4-kb insert from plasmid pTZlipAB as a probe. The autoradiograph was overexposed in order to visualize the 2.1-kb transcript. Lane A, mRNA of P. alcaligenes Ps496 isolated from a sample taken at 48 h of fermentation; lane B, mRNA of P. alcaligenes Ps495 isolated from a sample taken at 48 h of fermentation; lane C, mRNA of P. alcaligenes Ps495 isolated from a sample taken at 72 h of fermentation.

Scale-up losses.

The small-scale reproducible fed-batch process, which resulted in a lipase activity of 55.3 ± 4.1 MLU/liter (n = 5), was scaled up from the 10-liter scale to a 100-m³ production fermentor. During the scaling up, important parameters did not vary with scale. However, a significant loss of lipase production was observed, without any obvious reason. In Fig. 7A it can be seen that the lipase concentration and thus the production decreased with increasing scale, accumulating to a loss of production of 65% at the 100-m³ scale. Besides the lipase production, no major differences between the fermentations at the different scales were detected.

FIG. 7.

(A) Lipase production loss during scaling up, showing fitted lipase production profiles based on five fermentations at a 10-liter scale (——), four fermentations at a 100-liter scale (·····), and two fermentations at a 4-m³ scale (–––) and a lipase production profile at a 100-m³ scale (▪). (B) Effect of CO₂ addition to the inlet air at a 10-liter scale, showing lipase production without addition of CO₂ (•) (control), with the addition of 5% CO₂ (○), and with the addition of 10% CO₂ (▪). (C) Lipase production at 10-liter (•) and 100-m³ (▪) scales after successful scaling up with the following modifications: increased ventilation rate (from 0.4 to 0.8 vvm), decreased back pressure (from 80 to 20 kPa), and lowered pH (from 7.0 to 6.7).

A systematic search at the 10-liter scale for possible causes of the negative scale-up effect revealed the dissolved carbon dioxide as a likely candidate. During fermentation, the pH was regulated at 7.0. At this pH, CO₂ dissolved in the broth at fairly high levels, predominantly as HCO₃⁻. Due to the relatively high pressure (both hydrostatic and back pressure) and the low aeration rate of 0.4 to 0.5 vvm, the dissolved carbon dioxide concentration increased with increasing scale. The total dissolved CO₂ concentration could be estimated from the fermentation parameters and the CO₂ equilibrium constants and was substantially higher (three- to fivefold) at the larger scales of 4 and 100 m³. In order to obtain dissolved CO₂ concentrations at the 10-liter scale representative of those for large scales, CO₂ gas was supplied to the inlet air to a final concentration of 5 to 10% (vol/vol). By doing these experiments, it could be demonstrated that increasing dissolved CO₂ concentrations had a negative effect on the lipase production (Fig. 7B).

The improved large-scale process.

Additional proof that the dissolved CO₂ was responsible for a significant part of the decrease in lipase production at the large scales was obtained by using the 100-m³ production fermentor. Decreasing the broth volume from 100 to 50 m³, increasing the aeration rate to 1.2 vvm, and decreasing the back pressure to 20 kPa resulted in a reduction of the CO₂ concentration, which had a positive effect on the production of lipase. The maximal lipase activity detected was increased by 70%. However, the total output (lipase concentration × total amount of broth) per fermentor was lowered as a result of a lower occupation of the volume.

These experiments indicated that the prevention of scale-up losses ought to be done by keeping the dissolved carbon dioxide concentration as low as possible. At a large scale, the ventilation rate is usually limited. With a total volume of 100 m³, the maximal aeration that could be achieved was 0.8 vvm. Together with the reduced back pressure of 20 kPa, a reduction in the CO₂ concentration by almost a factor of 3 could be obtained. This indeed gave a beneficial effect on production. Still, it was not sufficient to eliminate all of the scale-up losses.

At the lab scale, additional methods of lowering the dissolved CO₂ concentration during the fermentation were investigated. By decreasing the soybean oil feed rate from 1 to 0.5 g of soybean oil per liter of broth per h, a reduction of the carbon dioxide production rate was obtained. However, this had a negative effect on lipase production (38.9 compared to 55.3 MLU/liter), suggesting that the specific lipase production rate was linked to the specific growth rate, which was also affected by changing feed rate. Further reduction of the dissolved CO₂ concentration was achieved by decreasing the pH. From studies at the lab scale, however, it was known that decreasing the pH to 6.5 caused a 10 to 15% reduction in lipase production. An optimal balance for production at large scales between the decrease of production caused by lowering the pH and the increase of production caused by lowering the dissolved CO₂ concentration was found at pH 6.7. Subsequently, a combination of increased ventilation rate (from 0.4 to 0.8 vvm) and decreased back pressure (from 80 to 20 kPa) together with the lowered pH (from 7.0 to 6.7) was used to minimize the scale-up losses at 100 m³ considerably (Fig. 7C).

DISCUSSION

The finding that the P. alcaligenes lipase shows the highest homology with the P. aeruginosa lipase is in line with the taxonomic data that both species are classified as being in rRNA group I of the pseudomonads (29). The homology with the lipases from rRNA group II Pseudomonas species (now called Burkholderia species) is lower but still very significant. The residues in the vicinity of the catalytic-triad residues Ser111, Asp257, and His279 (Fig. 3) are especially strongly conserved. Interestingly, the two cysteine residues forming a disulfide bond in Pseudomonas lipases are also fully conserved (positions 213 and 263 in Fig. 3), stressing the importance of this single disulfide bond for lipase folding and stability. Both the P. alcaligenes and P. aeruginosa lipases lack a stretch of 20 amino acids (around position 232) as compared to lipases from representatives of rRNA group II, Burkholderia cepacia and Burkholderia glumae. From a comparison with the experimentally derived three-dimensional structure of B. glumae, it can be seen that this deletion overlaps precisely with two antiparallel β-strands and a β-turn in the protein, which seem to be replaceable in the lipase molecule (25). Another major difference is in the signal sequence, which is unusually long in the rRNA group II lipases but not in the rRNA group I lipases.

By using Southern blotting, copy numbers have been determined for plasmids in P. alcaligenes, resulting in values of 2 to 4 and 20 to 40 copies for pLAFR3 and pJRD215, respectively (data not shown). These experimental data show that the gene copy number for pLAFR3 is around 3 and that for pJRD215 is around 30. For pLAFR3 this is in line with literature data (11). For plasmid RSF1010, which has the same replicon as pJRD215, a copy number of 12 in E. coli has been reported (13). However, it is known that alterations in the expression of rep functions can lead to higher copy numbers in plasmid variants or in other host cells (10). Overexpression of the lipase from plasmids pLAFlipA and pJRDlipA in P. alcaligenes leads to an increase in the lipase production at a ratio that seems to correlate with the copy numbers of these plasmids. Conceivably, LipB is not limiting. However, when lipA and lipB of P. alcaligenes were combined on both expression vectors (pLAFR3 and pJRD215), an additional increase in lipase levels was observed only with the high-copy-number plasmid, pJRDlipAB. Evidently a single lipB gene copy, which is still present in the host chromosome, is sufficient to fully support the efficient expression of lipA from a low-copy-number plasmid but is insufficient to support the full expression of lipase from a high-copy-number plasmid, as judged from the fermentation results for strains Ps496 and Ps495. The accuracy of plasmid copy number determinations is insufficient to derive quantitative conclusions, but from the comparison of strains Ps496 and Ps495 it seems that the helper gene becomes limiting at lipA gene copy numbers of above 10.

Further support for a model in which the LipB gene product is required only in catalytic amounts comes from the fact that the amount of lipB transcript is very low. A hairpin structure situated between the lipA and lipB genes has all the characteristics to act as a transcriptional attenuator (for a review, see reference 18). Alternatively, it could represent a pausing position for mRNA degradation. The observed low lipAB/lipA transcript ratio seems to be in contrast to the report that for B. cepacia (1) the optimal ratio of LipB to LipA has been measured at 1:1. Also, for P. aeruginosa a maximal activation in denaturation-renaturation experiments was found at a LipB/LipA ratio of 1:1 (28). It should be kept in mind, however, that those measurements were done in a refolding experiment under in vitro conditions quite distinct from physiological concentrations. Interestingly, both P. aeruginosa and the two representatives of rRNA group II, B. cepacia and B. glumae, have similar hairpin structures, in front of and within the lipB gene, respectively. Indeed, Frenken et al. (9) showed that in B. glumae the dominant transcript is a 1,400-nucleotide-long mRNA that encodes only lipase. Although the precise mode of action of LipB is yet unknown, the observations for P. alcaligenes and P. aeruginosa (16) best fit with a chaperoning type of action, which requires only catalytic amounts of LipB. Inactivation of the helper gene in the chromosome of P. alcaligenes (unpublished results) leads to a lipase-negative phenotype, which is in line with the proposed function as a lipase-specific chaperone.

Although for economic reasons complex substances often are used for industrial fermentations, for reasons of consistency and quality control there is a strong tendency nowadays to develop a process with a medium as defined as possible. This is due mainly to the disadvantages of the complex raw materials, such as the unknown composition, the batch-to-batch variability, and the forced combination in a fixed ratio. Since P. alcaligenes does not require any special nutrients for the production of lipase, it was possible to develop a balanced mineral medium for the fermentation process. Subsequently, the use of this defined medium significantly facilitated the identification of CO₂ as the major cause for the scale-up problem. It is generally known that CO₂ can be responsible for problems during scaling up. However, despite numerous studies on the effects of carbon dioxide on microbial growth and metabolism of microorganisms, knowledge of the mechanism of CO₂ inhibition still remains inconclusive (for reviews, see references 6 and 26). In this case also, no explanation for the molecular mechanism by which CO₂ affects the lipase production by P. alcaligenes can be offered. Other than the lipase production, no major differences in metabolism that were caused by the increased partial CO₂ pressure could be detected. There were no indications that the increased CO₂ concentrations had any effect on the energetics of cell synthesis (34) or on the lipase enzyme itself by affecting its physicochemical properties (24). The results here could indicate that there is a CO₂-controlled repression of lipase synthesis. Induction and repression of enzyme synthesis by CO₂ is well known in the process of autotrophic CO₂ fixation, e.g., with photolithotrophically grown cells of Rhodospirillum rubrum (30).

By a combination of adaptations to the process, the scale-up losses at 100 m³ could be minimized considerably. There is still a remaining loss of 10 to 15%. Some of this can be attributed to gradients of soybean oil, ammonia, oxygen, and pH that are generally known to exist in large-scale fermentors. These gradients are caused by the lack of ideal mixing that is a characteristic of these large-scale fermentors (7, 27). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis and Western blotting (data not shown) confirmed that the remaining loss in lipase production was not caused by inactivation of the lipase by shear, as has been reported for a lipase produced by Candida cylindracea (19).

In a separate paper the purification process leading to the complete removal of the endotoxins of P. alcaligenes will be reported.