Cultivation of Saccharomyces cerevisiae with Feedback Regulation of Glucose Concentration Controlled by Optical Fiber Glucose Sensor

Glucose belongs among the most important substances in both physiology and industry. Current food and biotechnology praxis emphasizes its on-line continuous monitoring and regulation. These provoke increasing demand for systems, which enable fast detection and regulation of deviations from desired glucose concentration. We demonstrated control of glucose concentration by feedback regulation equipped with in situ optical fiber glucose sensor. The sensitive layer of the sensor comprises oxygen-dependent ruthenium complex and preimmobilized glucose oxidase both entrapped in organic–inorganic polymer ORMOCER®. The sensor was placed in the laboratory bioreactor (volume 5 L) to demonstrate both regulations: the control of low levels of glucose concentrations (0.4 and 0.1 mM) and maintenance of the glucose concentration (between 2 and 3.5 mM) during stationary phase of cultivation of Saccharomyces cerevisiae. Response times did not exceed 6 min (average 4 min) with average deviation of 4%. Due to these regulation characteristics together with durable and long-lasting (≥2 month) sensitive layer, this feedback regulation system might find applications in various biotechnological processes such as production of low glucose content beverages.


Introduction
The number of people with diabetes has risen from 108 million in 1980 to 422 million in 2014, and between 2000 and 2016 there was a 5% increase in premature mortality from diabetes the World Health Organization (WHO) predicts that the diabetes will be the seventh leading cause of death in 2030 [1]. The accurate evaluation of the glucose content in foods is extremely important for the maintenance of its physiological level in blood of diabetic individuals. Information about glucose content of foods and beverages is essential for both producers and consumers. Glucose monitoring is crucial in tracing the fermentation processes in the wine, brewing, and dairy industries.
The first qualitative test of glucose was published in 1848 [2]. Since that time many methods of glucose quantification have been described [3]. Web of Science links to 25,000 references for key words glucose detection. Plenty of physical detection principles have been used which are often non-specific to glucose [4]. For example, microwave resonatorbased sensors might be advantageous for glucose detection in blood [5,6] or in some industrial applications as their linear range of measured glucose concentrations is from zero to more than ten weight percent [7]. In comparison with these microwave-based sensors, optical biosensors with glucose oxidase exhibit a high specificity to glucose. curable. Covalent attachment to a carrier (Sepabeads ® ) protects the enzyme against harsh conditions after mixing with Ormocer ® and during the UV curing of the sensitive layer on the acrylate lens.
Here, we present on-line feedback regulation of glucose concentration controlled by this glucose sensor placed in the bioreactor vessel. The regulation was demonstrated in both modes: in dilution and fed batch cultivation. Dilution mode was proposed for production of beverages with limited content of glucose and for diabetics, where glucose content should be close to zero. In fed-batch cultivation, microorganisms consume glucose, but to keep cells alive and producing, the concentration of glucose must be maintained at a level which allows cells to survive but limits their proliferation. In the fed batch process, glucose regulation is demonstrated with the most industrially used microorganism, Saccharomyces cerevisiae.

Microorganisms
Saccharomyces cerevisiae was obtained from Collection of microorganisms of the Institute of Biochemistry and Microbiology UCT Prague. Overnight culture (50 mL) was added into the bioreactor. Optical density (OD) of the overnight culture (3× diluted) was 0.5.

Preparation of Optical Sensitive Layers
Sensitive layers were prepared by procedure described in details by Kostejnova et al. [29]. Briefly, 200 mg of SEPA was activated by a stirring with glutaraldehyde (4 mL) for four hours. After centrifugation and washing, the enzyme solution was added to activated SEPA and the mixture was stirred for 18 h with the same velocity as was used during activation. The mixture was centrifuged, the supernatant was removed, and Sepabeads ® with immobilized glucose oxidase (SEPA-GOX) were washed twice with buffer. Ormocer ® was mixed with Ru complex, Irgacure 500 and sucrose to form ORM-RC. Components of sensitive layers were mixed on glass slide in the ratio 2:1, ORM-RC: SEPA-GOX. The mixtures were deposited on plastic lenses and cured by UV light for ten minutes. After UV polymerization the lenses were immersed in phosphate buffer (50 mM, pH 5.9) overnight to wash out sucrose. The thicknesses were measured with the microscope Tescan (Czech Republic) in the center and on the periphery of the layers (see Figure 1). Parameters of preparation and analytical characteristics (sensitivity (SN), linear dynamic range (LDR), and response time (RT) of sensitive layers used in the tests are presented in Table 1 and on Figure 1. during activation. The mixture was centrifuged, the supernatant was removed, and Sepabeads ® with immobilized glucose oxidase (SEPA-GOX) were washed twice with buffer. Ormocer ® was mixed with Ru complex, Irgacure 500 and sucrose to form ORM-RC. Components of sensitive layers were mixed on glass slide in the ratio 2:1, ORM-RC: SEPA-GOX. The mixtures were deposited on plastic lenses and cured by UV light for ten minutes. After UV polymerization the lenses were immersed in phosphate buffer (50 mM, pH 5.9) overnight to wash out sucrose. The thicknesses were measured with the microscope Tescan (Czech Republic) in the center and on the periphery of the layers (see Figure 1). Parameters of preparation and analytical characteristics (sensitivity (SN), linear dynamic range (LDR), and response time (RT) of sensitive layers used in the tests are presented in Table 1 and on Figure 1.

Feedback Regulation System
A schema with photos of the feedback regulation system is on Figure 2.

Feedback Regulation System
A schema with photos of the feedback regulation system is on Figure 2.  without responding. In dilution mode, the pump administrated buffer in case that c GL was the same or higher than maximum allowed glucose concentration (c GL MAX ). In cultivation mode, the concentrated glucose solution was added into the bioreactor in case that c GL was the same or lower than minimum allowed glucose concentration (c GL MIN ). To reach desired concentration, c GL = c GL DES , the volume of a dose (buffer/conc. glucose solution) was calculated by the software with respect of volume of liquid in the bioreactor. The new t chec started after the pump finished dosing.

Reproducibility of Biosensor Response in Repetitive Measurements during 2 Months
The probe of the biosensor with sensitive layer (enzyme concentration 125 mg GOX X-S /g SEPA, thickness of the layer 300 nm) was immersed in non-sterile buffer (50 mM, pH 7), which was bubbled by sterile air with volume flow 16 mL/min, mixed 400 rpm, and tempered 25 • C. During two months, on working days, SN, LDR, and RT. were determined once a day. Measurements and calculations of analytical characteristics are described in details in previous paper [29].

Sterilization
The bioreactor filled with medium/buffer with inserted pH, dO2, T probe, together with storage bottles of base, acid, concentrated glucose, dilution buffer, and all connection pipes was sterilized in autoclave at 120 • C for 30 min. Before inserting into bioreactor, glucose probe, and acrylate lens with sensitive layer were sterilized by immersing in ethanol (70%) for 5 min and irradiation with UV for 10 min.

Off-Line Measurement of Glucose Concentration
Off-line glucose concentration was measured with Glucose oxidase Activity Assay kit from Sigma Aldrich s.r.o. (Prague, Czechia).  (Figures 3 and 4. green frames), which lead to increase of fluorescence lifetime (τ DES ). On user interface monitor were set τ 0 , τ DES and corresponding c GL 0 , c GL DES , times for averaging t chec = 10 min and c GL MAX = 0.5 mM resp. 0.125 mM (Figures 3 and 4, red frames) and corresponding τ MAX calculated from the calibration. After three t chec (Figures 3 and 4, position 1) concentration of glucose was increased from c GL DES = 0.4 mM resp. 0.1 mM to c GL MAX = 0.5 mM resp. 0.125 mM by hand pipetting of solution of concentrated glucose (0.2 mL, resp. 0.05 mL) into the bioreactor (Figures 3 and 4, position 2). After t chec , c GL MAX was detected and the pump of feedback loop dosed calculated buffer volume into the bioreactor to reach c GL = c GL DES (Figures 3 and 4, position 3). In reality, glucose concentration after regulation (c GL REG ) differs from c GL DES . The cycles of addition of glucose solution and regulation were repeated three times during both tests. The test for c GL DES = 0.4 mM was reproduced three times (Figure 3.I-III). decreased. Therefore, in case the desired glucose concentration was equal to limit of detection (LOD) of used biosensor (cGL DES = LOD = 0.1 mM, the test IV.), the average deviation increased to 3.7%.     Response time (RT 90 ) was calculated for each buffer dose according to equation where t 1 is time when actual measured glucose concentration exceeded glucose concentration after regulation for 10% c GL = c GLREG + 0.1 × c GLREG , and t 2 is time when buffer was added. Deviation (s) of glucose concentration after regulation from c GL DES was calculated for each buffer dose

Feedback Regulation of Glucose Concentration of Fed Batch Cultivation of Saccharomyces cerevisiae in Stationary Phase (Cultivation Mode)
The bioreactor was filled with 2 L incomplete YPG medium. Throughout the experiment, the bioreactor was tempered to 30 • C and bubbled with sterile air, oxygen, or nitrogen to keep constant dO2 = 21%. Concentration of glucose in incomplete YPG medium was measured off-line. Double point glucose calibration was done in the first hour of the experiment. The first point was the concentration of glucose in incomplete YPG medium c GL MIN (2 mM, Figure 4, red frame) and the second point was c GL DES = 3.5 mM (Figure 4, green frame) acquired by hand pipetting of concentrated glucose (3 mL). Neofox corresponding fluorescence lifetimes, τ MIN and τ DES , were set on user software monitor together with t chec . Double the response time (RT 90 ), determined in calibration, was opted for checking time, t chec = 10 min.
After calibration, glucose (40 g) was added to complete YPG medium, thus c GL = 111 mM, which was out of the range (0-7 mM) of the biosensor ( Figure 5, position 1). The bioreactor was inoculated with night culture of Saccharomyces cerevisiae (50 mL, OD for 3x diluted culture was 0.5) and feedback regulation was switched on ( Figure 5, position 2). The growing cells consumed glucose. After 11.5 h of fermentation, c GL dropped below 7 mM ( Figure 5, position 3). Within 6 t chec , the measured glucose concentration c GL decreased from 6.8 to 1.9 mM and c GL < c GL MIN (Figure 6. position 1). At the end of 6th t chec the pump started to dose concentrated glucose solution into the reactor so that c GL = c GL DES , resp. c GL REG , after the seventh t chec (Figure 6, position 2). Culture of Saccharomyces cerevisiae consumed added glucose during the 8th t chec and c GL DES , and c GL REG dropped to (or under) c GL MIN , which activated dosing pump ( Figure 6, position 3). The cycle kept adding glucose to reach c GL DES followed by consumption with yeast culture to c GL MIN (cycle ↓↑) was repeated seven times. The experiment was twice reproduced.
Response times were calculated for each glucose dose according to where t 1 is time when measured glucose concentration reach 90% of concentration after regulation: c GL = 0.9. c GL REG and t 2 is time when concentrated glucose solution was added. Deviation (s * ) from c GL DES were calculated for each glucose dose according to shorter as the activity of enzyme increases and sensitive layer is thinner [26]. Nevertheless, these parameters are limited by technical feasibility of a preparing such layer. Selectivity, robustness, and long-term reliability are favored features of enzymatic glucose sensors with oxygen transducers for control of glucose concentrations in biotechnological processes but, if one minute or less response times are necessary, another type of glucose sensor should be used.  Activity of microorganisms resulted in cGL decreased from 6.8 to 1.9 mM (cGL < cGLMIN, position 1). After this point, the pump started to dose concentrated glucose solution into the reactor so that cGL = cGL DES (position 2). Culture of Saccharomyces cerevisiae continued to consume glucose and cGL dropped to cGLMIN, which activated dosing pump again (position 3). t1 * is time when measured glucose concentration reached 90% of concentration after regulation and t2 * is time when concentrated glucose solution was added.

Feedback Regulation of Glucose Concentration to Lower Level (Dilution Mode)
This regulation is demonstration of application of feedback system in production of beverages for diabetes, where the demand is to keep glucose concentration at a level close to zero. The sensitive layer, used in dilution mode, was chosen to meet the need of the lowest detection limit. Based on our previous study [29], such demand best fit sensitive layer comprising high content of enzyme, which is immobilized on undivided SEPA.
Monitoring and control of low glucose concentration levels are in Figures 3 and 4, and characteristics of regulation are in Table 2. In all experiments, response times were shorter than 5 min (RT 90 ≤ 5 min). Average deviation was 1.8% for glucose concentration hold at 0.4 mM. For lower glucose concentration, the relative precision of measurement decreased. Therefore, in case the desired glucose concentration was equal to limit of detection (LOD) of used biosensor (c GL DES = LOD = 0.1 mM, the test IV.), the average deviation increased to 3.7%. Table 2. Response times of the biosensor and deviations from desired glucose concentration in case of step increase of glucose concentration.

Feedback Regulation of Concentration of Glucose of Fed Batch Cultivation of Saccharomyces Cerevisiae in Stationary Phase (Cultivation Mode)
In cultivation mode, the sensitive layer should possess fast response time and wide concentration range to measure glucose in sufficiently broad concentration range during stationary phase of cultivation. In our previous paper [29], it was shown that LDR of the layers increased with decreasing enzyme concentration. It was also shown that crushing of spherical SEPA with immobilized glucose oxidase resulted in higher LDR. Unfortunately, increasing LDR simultaneously increased RT, which is an undesirable effect for the feedback regulation system. Therefore, we must compromise between opposing demands on analytical features of sensitive layer for cultivation mode. We used the sensitive layer with RT ≤ 6 min and LDR = 0-7 mM.
A time record of complete cultivation of Saccharomyces cerevisiae is presented on Figure 5 and the detail of stationary phase, while glucose concentration was controlled with the feedback regulation system, is on Figure 6. Table 3 shows that in all seven cycles ↓↑, response times were below 6 min (RT 90 * ≤ 6 min) and deviation from regulation did not exceed 9%. The average RT 90 * was 4 min and the average deviation 3.9%. In situ monitoring and control glucose concentration during cultivation were described by Tric et al. [30]. They used also enzymatic sensor with optical glucose transducer; however, glucose oxidase was fixed on optically isolated oxygen sensor with glutaraldehyde and covered by perflorated hydrophilic membrane. In comparison with this report, where response times were 6 min for increasing and 10 min for decreasing of glucose concentrations, we reached response times shorter than 6 min in all tests. The shorter response times might be related to faster diffusion of oxygen and glucose in sensitive layer comprising both enzyme and fluorescent complex in one mixture. These results implicate that regulation response times less than few minutes are hard to reach with enzymatic glucose sensor with optical oxygen transducer. Response times become shorter as the activity of enzyme increases and sensitive layer is thinner [26]. Nevertheless, these parameters are limited by technical feasibility of a preparing such layer. Selectivity, robustness, and long-term reliability are favored features of enzymatic glucose sensors with oxygen transducers for control of glucose concentrations in biotechnological processes but, if one minute or less response times are necessary, another type of glucose sensor should be used.
Activity of microorganisms resulted in c GL decreased from 6.8 to 1.9 mM (c GL < c GLMIN , position 1). After this point, the pump started to dose concentrated glucose solution into the reactor so that c GL = c GL DES (position 2). Culture of Saccharomyces cerevisiae continued to consume glucose and c GL dropped to c GLMIN , which activated dosing pump again (position 3). t 1 * is time when measured glucose concentration reached 90% of concentration after regulation and t 2 * is time when concentrated glucose solution was added.

Reproducibility of the Biosensor Response during 2 Month.
During two months (42 measurements) the average SN was 0.306 µs L mmol −1 with relative deviation 10% ( Figure 7) and an average maximum of linear dynamic range (LDR MAX ) 1.6 mM with relative deviation 12% (Figure 8). At the first measurement, RT was 9 min. In the second measurement, RT increased to 14.7 min and this response time was preserved in following 40 measurements. An average RT (without the first day) was 15.1 min with relative determinative deviation 8% ( Figure 9) and it remained constant throughout the repetitions (p > 0.9971). After the first experiment, an increase of RT is probably a result of an adsorption of microorganisms from non-sterile buffer, which cause diffusion slowdown of both substrates glucose and oxygen, in the sensitive layer.

Wider Applicability of the Biosensor
The presented biosensor was developed with immobilized glucose oxidase aiming for the on-line monitoring of glucose concentration. Together with oxygen and pH, glucose concentration is one of the most often measured parameters in biotechnology. Nevertheless, the presented concept is general and replacing of glucose oxidase by other oxidases can result in various analogical biosensors, such as for biological amines [26] or cholesterol oxidase [27] for use on continuous systems. Of interest in near future might be sensors of various environmental pollutants. Biodegradation pathways of many organic pollutants often start with oxygenases enzymes [28,29] of different specificity, and these could serve as a biosensing elements for regulation of continuous water treatment processes. probably a result of an adsorption of microorganisms from non-sterile buffer, which cause diffusion slowdown of both substrates glucose and oxygen, in the sensitive layer.    probably a result of an adsorption of microorganisms from non-sterile buffer, which cause diffusion slowdown of both substrates glucose and oxygen, in the sensitive layer.

Conclusions
In this work, we presented the feedback system for regulation of glucose concentration based on the enzymatic sensor with optical oxygen transducer. The system was demonstrated for the case of maintaining low glucose concentration. An undesirable increase of glucose concentration was compensated below 0.125 mM by dilution in less than 5 min. In stationary phase of fed batch cultivation when glucose was continuously consumed by growing microorganisms, the feedback system adjusted glucose concentration to 3.5 mM in less than 6 min after detection of the concentration drop to 2 mM. The two-month stability and reproducibility of biosensor response was demonstrated by daily measurements, in which relative determinative deviations of analytical characteristics (sensitivities, linear dynamic ranges, and response times) were less than 12%. In comparison to known and commercially available glucose concentration regulations, the presented feedback system has advantage in use of in situ sensor, robust construction, and long-term stability.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to Pat. No. CZ30355 / PUV 2016-33183.

Conflicts of Interest:
The authors declare no conflict of interest.