The substrate auto-delivery system is based on the controlled glucose release into the liquid cultivation medium through enzymatic degradation of starch. One major objective was to reach cell densities comparable to fed-batch bioreactor cultivations. However, the amount of starch necessary to achieve this is far above the concentration which remains soluble in a liquid. Even solubilised starch easily looses it's solubility due to the interactions between the amylose chains. In addition, a high concentration of starch in the cultivation medium would decrease the oxygen transfer rate due to the increased medium viscosity. Therefore the starch concentration in the medium should be low. This practical conflict was solved by generation of a two-phase system, consisting of a gel as the starch reservoir, and a liquid cultivation medium (Fig. 1b). The gel composition was optimized so that starch diffuses just fast enough from the gel to the medium without limiting the enzyme reaction and for a high mechanical gel stability to resist vigorous shaking.

A starch concentration of 10% was embedded in the gel to provide enough glucose for a cell density of about 30 g l^-1 (OD₆₀₀ ≈ 100) in the liquid phase assuming that (i) all starch is available and gets processed, and (ii) one gram of glucose yields about 0.5 g cell dry weight. In initial experiments it was found that the mechanical stability of the starch gel could be increased by addition of agar (data not shown). Separate optimization of the gel for use in shake flasks and in 96-well plates resulted in different agar concentrations, 5% for shake flasks and 1.5% for 96-well plates. This gel is referred to as the storage gel.

An additional intermediate agar layer between the storage gel and the cultivation medium (referred as regulating gel) prevents the release of insoluble starch particles to the liquid medium and provides a better control of the release rate of soluble starch. The results in Figure 2a clearly show for 96-well plates that the agar concentration in the regulating gel affects the diffusion of starch into the growth medium significantly. The regulating gel demonstrated sufficient mechanical resistance and was able to sustain intensive agitation when the concentration of agar was kept above 3.25% in microwell plates and above 5% in shake flasks. The thickness of the regulating gel slightly influenced the release of starch (data not shown). Its volume was kept as small as possible to maximize the cultivation volume. A 50 μL agar layer with 1.5% agar was considered to be optimal as regulating gel for 96-well plates. A 75 ml agar layer with 5% agar was ideal for 1 litre shake flasks.

The concept of controlling the glucose release by varying the concentration of glucoamylase in the cultivation medium is demonstrated in Figure 2b for the 96-well plate system. For establishing the system we have applied a glucoamylase from Aspergillus niger (AG 300L) which is commercially available and characterised by good stability.

The results show a clear dependency of glucose accumulation on the glucoamylase amount in the cultivation medium over the tested range of enzyme concentrations (Fig. 2b, c). For fed-batch type cultivation experiments with E. coli a glucoamylase amount of 6 to 12 AGU l^-1 (amyloglucosidase units per liter) may be proposed as a good starting point, assuming that a release of 5 to 10 g l^-1 glucose over a time period of 20 hours would be related to an overall specific growth rate of 0.2 ± 0.1 h^-1. Based on these calculations, and considering a yield coefficient of 0.5 g cell dry weight per one gram of glucose, a culture inoculated with an OD₆₀₀ of 0.1 should growth up to an OD₆₀₀ of 8 to 16 within 20 hours.

After optimization of the composition of the system, the principle of the enzymatic substrate auto-delivery system was first applied in shake flask cultures. The growth behavior with differing amounts of glucoamylase was tested with E. coli K-12 RV308 and E. coli BL21(DE3) as examples. Both strains are widely used in bioproduction of recombinant proteins. The physiological events in the shake flasks were monitored by the wireless Senbit^® system [15] with on-line sensors for pH and pO₂, and off-line analyses for glucose, organic acids, and the cell density (OD₆₀₀). As expected, the cultures at all three tested glucoamylase concentrations of 60, 30, and 12 AGU l^-1 showed an initial batch phase, during which the glucose level was clearly not growth limiting (Figs. 3a–c). The glucose concentration first increased, and then by the higher metabolic activity at increasing cell concentrations, decreased. During the batch phase, lasting approximately for 5.5 to 7 hours, depending on the enzyme concentration, the culture density increased exponentially and the pO₂ and pH levels decreased. The pO₂ and pH levels decreased more in the cultures with higher glucoamylase concentration. With both glucoamylase concentrations, 30 and 60 AGU l^-1 (Figs. 3a, b), the cultures reached a critical pO₂ level close to zero. Glucose limited growth after the batch phase was only established in the cultures with 12 and 30 AGU l^-1. In the culture with the highest enzyme concentration tested glucose accumulated constantly until the end of the cultivation. In this culture, despite the availability of glucose, cell growth was inhibited and did not rise above an OD₆₀₀ of 10. This was due to the accumulation of acetate and the connected drop of the pH value to 4.5 (Fig. 3a, d), which is non-optimal for growth of E. coli [35]. In contrast, the acetate level remained low and pH and pO₂ levels were rather stable in the cultures with lower glucoamylase concentrations (12 and 30 AGU l^-1), and also a high final cell density was obtained (OD₆₀₀ > 25). None of these cultures showed significant accumulation of either lactate or formate, being well in agreement with the observed pO₂ values. This indicates that the reason for the growth inhibition at the highest release rate of glucose is rather the acidification of the medium by overflow metabolism than anaerobic metabolism.

However, screening for the optimal amount of glucoamylase is easy. Without measuring glucose, acetate, or the pO₂ in the culture, screening for highest cell density with different glucoamylase amounts will result in cultivation conditions which are still aerobic and where the acetic acid production is so low that the pH is not turned to very unfavourable conditions. As both, anaerobic metabolism [36] and overflow metabolism, can lead to a significant decrease of the biomass yield, the proper enzyme concentration can be selected straight forward by evaluation of the cell density.

Such screening can be readily done in 96-well plates due to the large number of possible parallel cultivations, as is demonstrated in Figure 4. In this case E. coli BL21(DE3) was cultivated in 150 μl liquid medium at 37°C with glucoamylase amounts ranging from 0.3 to 12 AGU l^-1 (Fig. 4a). As is concluded from the high cell densities and the linear growth profiles reached in the cultures with 3 to 6 AGU l^-1, these were the highest tested enzyme concentrations which supported glucose limited cultivation over the whole cultivation period. The data show that also lower enzyme concentrations can be applied if one would aim for slow controlled growth. An enzyme concentration of 12 AGU l^-1 resulted in a lower cell density; probably the higher glucose release had caused either accumulation of acetic acid and a steep decrease of the pH in the culture as described above, or even stronger oxygen limitation and anaerobic metabolism.

From the relative share of nitrogen in the biomass [37] we proposed that nitrogen limits the growth at a cell density of 20 to 25 in our medium. Therefore, it was then investigated whether the cell density can be further increased by addition of additional nitrogen. Addition of ammonia ions after 20 h of cultivation, at approximately the same as the initial concentration, resulted in continued growth up to a final cell density of about OD₆₀₀ = 40 (80 h of cultivation, Fig. 4b). The highest OD₆₀₀ was obtained if the extra nitrogen was added as a mixture of ammonium sulphate and ammonia to correct for the decreased pH value.

These experiments document how growth can be simply optimised with EnBase cultivations for the specific shaking conditions and the biological system. The optimal amount of enzyme is selected by testing different glucoamylase concentrations and by analysing the growth profile. Most favourable growth conditions resulted in the highest cell density in the defined time window.

As there is a big interest in a parallel cultivation system for recombinant protein production (e.g. [11,38]), the novel glucose auto-delivery system was applied for that application. As an example the production of recombinant Trypanosoma brucei triosephosphate isomerase (TbTIM) in the 96-well system was investigated. The gene had earlier been transferred into E. coli BL21(DE3) pET3a pLysS [39]. The test system was chosen for the reason that (i) the BL21 T7-RNA polymerase system is widely used and has even been proposed as the best consensus protocol for production of soluble recombinant proteins in E. coli [16], (ii) TbTIM shows a high expression level in this system [39], and (iii) TbTIM shows the non-desired tendency to aggregate irreversibly, which complicates the structural studies of the protein and it's variants with mutations in different amino acids [40]. We have earlier experienced that the cultivation conditions strongly affect aggregation of TbTIM. Therefore it was interesting to investigate whether the growth rate at the time of induction influence the amount of product and the relative share of soluble active TbTIM. A further challenge was that the optimal cultivation temperature for TbTIM production is 25°C, a temperature which was not tested for the novel auto-delivery system previously, but is generally used if aggregation is a problem [16]. Aside from analysis of the cell density, samples were taken to follow the TbTIM concentration by SDS-PAGE.

The accumulation of soluble TbTIM was increased in the EnBase cultures by a factor of 8 to 10 compared to the control on M9ZB medium without EnBase (Fig. 5, 364 to 512 μg l^-1 in the EnBase cultures dependent on the time of induction versus 51 μl l^-1 in the control culture). The production of TbTIM in M9ZB medium was earlier described [41] and had given the best results so far. That is why we used it as a production standard method for different TbTIM variants in shake flask cultures.

Despite the total production of TbTIM per cell was lower in the EnBase cultures (possibly due to the use of mineral salt medium), the total amount of soluble TbTIM was significantly higher compared to the control culture. The higher yield of total TbTIM in the EnBase cultures is due to the high cell density and, unexpectedly, a higher relative share of soluble TbTIM compared to the traditional batch culture system.

Although the highest amount of soluble TbTIM could be harvested if the culture was induced at OD₆₀₀ = 5.0, it is remarkable that the process is very robust in view of the time of induction (cf. Fig. 5b). The amount of harvested product deviated from the average by less than 20% for the cultures which were induced between OD₆₀₀ 5.0 (16.4 h after inoculation) and 12.1 (32.8 h after inoculation).

E. coli RV308 (βlacX74 galISII::OP308 strA, ATCC 31608), E. coli BL21(DE3), and the plasmid containing strain E. coli BL21(DE3) pLysS pET3aTbTIM [41] were used in this study.

All cultivations with the novel substrate delivery system were performed in glucose-free mineral salt medium (MSM, [42]) with the following composition (per litre): 2.0 g Na₂SO₄, 2.7 g (NH₄)₂SO₄, 0.5 g NH₄Cl, 14.6 g K₂HPO₄, 3.6 g NaH₂PO₄ × H₂O, 1.0 g (NH₄)₂-H-citrate, 2 mM MgSO₄, 0.1 g l^-1 of thiamine hydrochloride, and 2 ml l^-1 trace element solution [43]. Additionally for precultures this medium contained 5 g l^-1 glucose (glucose-MSM). M9ZB medium contained (per litre): 10 g casein hydrolysate (Sigma-Aldrich, St. Louis, USA), 5.0 g NaCl, 1.0 g NH₄Cl, 3.0 g KH₂PO₄, 3.2 g Na₂HPO₄, 4 g glucose, and 1 mM MgSO₄. 100 μg ml^-1ampicillin and 35 μg ml^-1 chloramphenicol were added in order to maintain selective pressure for the pET3aTbTIM and pLysS plasmids, respectively.

The substrate auto-delivery system was based on a two-phase gel, the storage gel contained soluble potato starch (Sigma-Aldrich, St. Louis, USA) and agar (BD, Bacto Agar, Franklin Lakes, USA) as a stabilizing agent, the regulating gel only contained agar but no starch. In microwell plates (Perkin Elmer Spectra Plate™ – TC 96, Waltham, USA) the storage gel contained 10% starch and 1.5% agar. The gel was prepared by suspending the starch in a small volume of distilled water at room temperature. The suspension was poured into a larger volume of intensively mixed hot distilled water (≈95°C). After mixing the hot suspension for 20 to 30 min intensively by keeping it at about 90°C, the agar was added directly to the hot starch suspension, which was then filled with distilled water to the final volume. The upper regulating gel was prepared by resuspending the agar, 1.5 to 5% for microwell plates and 5% for shake flasks, in distilled water. The gels were autoclaved separately for 20 min at 121°C and afterwards stored at room temperature. The microwell plates were filled under aseptic conditions, with 100 μl of warm liquefied storage gel. After solidification 50 μl of warm regulating gel was added. The 1 l shake flasks contained 100 ml storage gels and 75 ml regulating gel layers.

Precultures were prepared from frozen glycerol-stocks in 100 ml Erlenmeyer flasks containing 10 ml of glucose-MSM by overnight cultivation of the respective strains at 37°C on an orbital shaker (180 rpm). For inoculation the preculture medium was removed by centrifugation (23°C, 3200 × g, 20 min) and the pellet was resuspended in glucose-free mineral salt medium. The preculture medium of E. coli BL21(DE3) pLysS pET3aTbTIM was removed by centrifugation (23°C, 3200 × g, 4 min) and the pellet was suspended either in glucose-free MSM or M9ZB.

Shake flask cultures with the glucose auto-delivery system were performed in 1 litre Erlenmeyer flasks with 4 baffles (Glasgerätebau Ochs, Bovenden-Lengelern, Germany) with side necks for the placement of the pH and oxygen sensors (cf. [15]) containing the two-phase gel and a total of 165 ml of glucose-free mineral salt medium. The indicated amounts of Aspergillus niger glucoamylase (Amylase AG 300L, Novozymes, Bagsværd, Denmark) were added after inocculation and the culture was immediately started (37°C, 180 rpm).

For cultivation in 96-well plates (Perkin Elmer Spectra Plate™ – 96 TC, Waltham, USA) the outer surrounding wells were not used for cultures but filled with water to limit evaporation. For cultivation the inner, gel-containing wells were inoculated with 150 μl of glucose-free MSM containing the cells and glucoamylase in the respective concentrations. The cultivations were performed with a Variomag Thermoshaker (with TEC-controller 485, Inheco, Munich, Germany) at 750 rpm. To minimise evaporation the plate was placed in a humid cabinet controlled at approximately 80% humidity. Under these conditions the evaporation was about 1 μl per well and hour at 37°C.

The cultivations of E. coli RV308 and E. coli BL21(DE3) were performed at 37°C and the cultivation of E. coli BL21(DE3) pLysS pET3aTbTIM was performed at 25°C. Liquid culture controls were placed into the same plate with 150 μl M9ZB but without gels. Product synthesis was induced by addition of 0.5 mM IPTG. Additional nitrogen (at time point 26.2 h) and magnesium (at time point 36.85 h) were added to the cultivation of E. coli BL21(DE3) pLysS pET3aTbTIM by adding NH₃ (as 25% solution) and (NH₄)₂SO₄, and MgSO₄ (as 0.5 M solution) to final concentrations at 0.033%, 2.6 g l^-1, and 3.3 mM, respectively.

For medium analysis cells were removed by centrifugation in a microfuge (16 000 × g, 3 min, 4°C) and the clear supernatants were used for analysis.

The starch concentration was analysed by the amount of glycosyl residues after complete acid hydrolysis and following glucose analysis, according to the following procedure (modified from Ritte et al. [44]): 100 μl of sample was hydrolyzed by addition of an equal volume of 2 N HCl solution at 100°C for 2 hours. The solution was then neutralized with 100 μl of 2 N NaOH and 100 μl of 0.5 M potassium phosphate buffer.

Glucose was analysed with a YSI 2700 Select Biochemical Analyzer (YSI Inc., Yellow Springs, USA).

Acetate was analysed by high-performance liquid chromatography (HPLC). Medium samples were stored at -20°C until HPLC analysis. Samples were thawn on ice, incubated at 80°C for 5 min to solubilise possible precipitates, centrifuged at 16 000 × g for 5 min at +4°C, and finally filtered with 0.2 μm cellulose filters. The HPLC analysis was done with a Merck-Hitachi HPLC-system (Model D-6000) and ICSep COREGEL 87H3 organic acids column (Transgenomic Inc., Omaha, U.S.A.). Running buffer was 0.01 N H₂SO₄. Detection of organic acids was performed with a UV-VIS detector (L-4250, Merck Hitachi) at 210 nm. Data analysis was performed by the D-7000 HPLC System Manager (version 3.1.1, Hitachi) software.

Cell growth was followed with a spectrophotometer after the dilution of culture samples in growth medium at either 600 nm (OD₆₀₀) for samples from shake flask cultures with an UltroSpec Pro 2100 UV/Visible Spectrophotometer (GE Healthcare, Buckinghamshire, UK), or at 490 nm (OD₄₉₀) from the 96-well plate cultures with a Victor³ Multiwell Plate Reader (Perkin Elmer, Waltham, USA) and a sample volume of 5 μl. 1 unit of OD₄₉₀ corresponds to 0.73 of OD₆₀₀. All cell density values are presented as OD₆₀₀ values. One unit of OD₆₀₀ corresponds to a dry cell weight of 0.3 g l^-1 [36].

The SENBIT^® wireless system (teleBITcom GmbH, Teltow, Germany) was used in the shake flask cultures to follow the pO₂ and pH continuously in 12 s intervals as described earlier [15].

Overexpression and purification of TbTIM, to provide a standard of pure protein for the SDS-PAGE analysis, was done as previously described [41]. For samples the culture media was removed by centrifugation (16 000 × g, 3 min, 4°C) and the cell pellets were frozen at -20°C. For analysis the cells were disrupted simply by the action of the internal T7-lysozyme by suspending them in 50 μL lysis buffer (20 mM Tris, 100 mM NaCl, pH 8) with addition of MgCl₂, DNAse and RNAse as described earlier [41]. Insoluble cell debris (IN) was separated from the soluble cell debris (SP) by centrifugation at 16 000 × g (30 min). The amounts of cellular debris suspensions were normalized by dilution with the mentioned buffer, without the additives, on the basis of lowest optical density at OD₆₀₀. The IN was washed twice with the same buffer by re-suspending and centrifugation at 16 000 × g (15 min). The IN was finally re-suspended in 50 μl of 6 M urea. Both samples, IN and SP, were applied to SDS-PAGE. As a reference pure TbTIM (dissolved in the same buffer for the SP samples and dissolved in 6 M urea for the IN samples) was applied in different concentration on the same SDS-PAGE gel. The bands were visualized by coomassie brilliant blue staining and by scanning with Image scanner III and band intensities were quantified with the ImageQuant 5.2 software (both GE Healthcare, Buckinghamshire, UK). Known TbTIM quantities were used to calculate the amount of TbTIM on the sample gels (SP and IN).