Understanding the effect of interaction among aeration, agitation and impeller positions on mass transfer during pullulan fermentation by Aureobasidium pullulans

Abstract

Pullulan is a non-ionic, water-soluble homopolysaccharide produced via fermentation using Aureobasidium pullulans, a black yeast. The unique physicochemical properties of pullulan have made it a potential candidate for a diverse range of applications in various industrial sectors such as food, pharmaceutical, cosmetics, and biomedicine. Low yield and productivity are major challenges for the extensive commercialization of this biopolymer, and they are often found to be associated with poor mass transfer during the fermentative production of pullulan. The present study is an attempt to address this challenge by a unique multivariate approach. The interactive influence of air-flow rates, agitation speed and impeller spacing on volumetric mass transfer coefficient (k_La) was investigated using a statistical design. The design has demonstrated a unique correlation between mass transfer capabilities and process variables, whereas pullulan fermentation kinetics has revealed a dependence of microbial activity on k_La values in a bioprocess. Furthermore, the resultant mass transfer conditions were applied for pullulan production, which enhanced the productivity significantly (1.61 g L⁻¹ h⁻¹). These observations might help in obtaining improved mass transfer conditions for the economic production of pullulan.

---

1. Introduction

Pullulan is a neutral exopolysaccharide that is mostly produced by a yeast-like fungus, Aureobasidium pullulans. This biopolymer is composed of maltotriose subunits connected via α-1,4 and α-1,6-glycosidic bonds in a 2 : 1 ratio.¹ This unique pattern of linkages provide several interesting physicochemical properties to the polysaccharide and makes it an ideal candidate for a wide range of applications in food, pharmaceuticals, cosmetics and other industrial sectors. In the food industry, pullulan has been mostly used as a gelling agent, emulsifier, stabilizer, dietary fiber and thickening agent.² It has also been recognized as a GRAS (Generally Recognized As Safe) product for food industries by USFDA. Pullulan has become a potential candidate for targeted drug and gene deliveries due to its nontoxic and non-immunogenic properties.³ All these factors suggest the plausibility of pullulan for becoming the most widely used biopolymer in numerous high-value technological platforms, including biomedical and pharmaceutical industries. However, high production cost remains a major bottleneck for its acceptance and widespread application across industrial sectors.

The fermentative production of pullulan is associated with the problems of high viscosity and poor mass transfer conditions. Viscous fermentation broths exhibit pseudo-plasticity and viscoelasticity owing to their changing fluid rheology with the passage of fermentation time. The non-Newtonian behavior of viscous fermentation fluids impart detrimental effects on bioreactor performance through perturbed mixing conditions and oxygen mass transfer criteria.⁴ The inhomogeneity produced in the fermenter due to viscous fluid rheology may affect the yield and productivity of the bioprocess. Improved mass transfer conditions might help in resolving issues such as lower product yield and poor quality products during pullulan production via fermentation.⁵ Oxygen requirements of bioprocesses in a fermenter are met through the gas–liquid mass transfer phenomenon,^6,7 in which oxygen is transferred from ascending bubble streams into aqueous phase and is often characterized by a volumetric mass transfer coefficient (k_La).

Generally, the enhancement of k_La values is achieved by either increasing agitation or aeration values in a fermenter; however, their associated disadvantages of generating new challenges, such as increased process costs, excessive foaming, and high shear rate, limit their application to a certain extent.⁸ Increasing these variables beyond a critical value decreases both mass transfer capability as well as microbial activity. Earlier, Bandaiphet and Prasertsan (2006) reported reduced biopolymer production from Enterobacter cloacae when fermentation was executed beyond an agitation rate of 600 rpm.⁹ The higher level of agitation could maintain available dissolved oxygen in the fermenter; however, increasing the agitation speed beyond a particular level may inhibit cell growth and product formation by inducing cell deformation at high shear rates.¹⁰ Similarly, Roukas and Mantzouridou (2001) also reported lower pullulan yields with higher levels of aeration, suggesting that the univariate enhancement of either agitation or aeration might not resolve the mass transfer issues of viscous fermentations.¹¹ Moreover, flow pattern and oxygen transfer rates in stirred tank reactors are significantly affected by the spacing between the impellers.¹² It may be emphasized that increasing the distance between the impellers beyond a critical level would create an ineffective agitation region between adjacent impellers, resulting in reduced mass transfer.¹³

To address these challenges, the present study was attempted to create an improved mass transfer environment for carrying out viscous pullulan fermentation. A unique approach was followed for the first time to improve the process of mass transfer capability of fermentation systems. This approach employed a statistical model to understand the interaction of aeration, agitation and impeller spacing on the oxygen mass transfer capabilities of a fermenter using a simulated Newtonian system. Depending upon the multivariate effect of these parameters on oxygen mass transfer, an optimized fermentation operating condition was devised for resolving oxygen limitation during the fermentative production of pullulan. These optimized conditions were further applied in pullulan fermentation to overcome mass transfer issues related to viscous fermentations, resulting in an enhanced productivity of the biopolymer.

2. Experimental section

2.1 Determination of k_La values

A 5 L stirred tank reactor (STR) (BioFlo® 310, New Brunswick Scientific Co., Inc., Edison, NJ) containing two six-bladed Rushton turbine impellers was filled with 3.2 L distilled water to measure the k_La value in Newtonian system. The k_La value was measured using the static gassing-out method described by Garcia and Gomez (2009).¹⁴ Briefly, the dissolved oxygen concentration in water was removed by flushing nitrogen gas, and the deoxygenated water was then aerated. The increase in dissolved oxygen concentration in solution was then measured to calculate the k_La value.

2.2 Experimental design and statistical optimization

Response surface methodology was used to understand the interaction among the process variables and their cumulative effect on mass transfer. A central composite design (CCD) was developed using a full factorial matrix to understand the combined effects of agitation (A), aeration (B) and spacing between two impellers (C) on oxygen transfer rate and to identify the key variables responsible for high mass transfer. Each of the process variables were varied among five different levels (Table 1) to generate a second-order response surface.

Table 1 Experimental range of variables studied during designing of experiments^a

Factors	Symbols and units	Coded levels
		Min.	Low (−1)	Mid (0)	High (+1)	Max.	Std. dev.
a Note: * volume per volume per minute.
Agitation	A (rpm)	63.64	200	400.00	600.00	736	169.56
Aeration	B (vvm*)	0.15	0.75	1.63	2.50	3.10	0.74
Impeller position	C (inch)	0.11	1.10	2.55	4.00	4.99	1.23

---

For this purpose, agitation and aeration were controlled by providing set points in the fermenter, whereas the spacing between two impellers was changed by varying the position of top impeller. According to previous reports,^15,16 the distance between the lower impeller and the bottom of the vessel should be one-third of the vessel's diameter. In the present study, the diameter of the bioreactor vessel used was 6′′, and thus accordingly lower impeller position was fixed at a distance of 2′′ from the bottom of the vessel. A total of 19 experiments (Table 2) were designed using a statistical software package (Stat Ease Inc. Design Expert Ver. 8), which included five centre points, six axial points and eight factorial points with five replicates around the centre point. A multiple regression analysis of the data was carried out to determine the optimal oxygen mass transfer conditions in a fermenter. A second-order polynomial equation was devised to define the predicted response (Y) in terms of independent variables (A–C), expressed as

Y = x_o + x₁A + x₂B + x₃C + x₁₁A² + x₂₂B² + x₃₃C² + x₁₂AB + x₁₃AC + x₂₃BC

where Y is the response, x_o is the intercept coefficient, x₁, x₂, x₃ are the linear coefficients and x₁₁, x₂₂, and x₃₃ are the squared coefficients; and x₁₂, x₁₃, and x₂₃ are the interaction coefficients. The aptness and fit of the regressed empirical model were statistically ascertained through R² and F-values from the analysis of variance (ANOVA) of the model. Response surface graphs and contour plots were also generated to explain the interactive effect of each variable on the response.

Table 2 Experimental matrix designed through CCD for studying interactions among variables

Run no.	Factor A agitation (rpm)	Factor B aeration (vvm)	Factor C impeller spacing (inch)	Response: kLa (min^−1)
1	736.36	1.63	2.55	2.1
2	400.00	1.63	2.55	1.15
3	400.00	1.63	2.55	1.3
4	200.00	2.50	1.10	0.56
5	600.00	0.75	4.00	0.8
6	200.00	2.50	4.00	0.73
7	200.00	0.75	1.10	0.29
8	600.00	2.50	4.00	1.9
9	400.00	1.63	0.11	0.28
10	400.00	1.63	4.99	0.8
11	400.00	0.15	2.55	0.6
12	600.00	2.50	1.10	1.7
13	600.00	0.75	1.10	1.1
14	200.00	0.75	4.00	0.45
15	400.00	1.63	2.55	1.4
16	400.00	1.63	2.55	1.3
17	63.64	1.63	2.55	0.28
18	400.00	3.10	2.55	1.4
19	400.00	1.63	2.55	1.4

---

2.3 Application of the developed model in pullulan fermentation

The developed model was validated after performing experimental runs on pullulan fermentation, as suggested by the statistical optimization of the model.

2.3.1 Microorganism and inoculum development. The organism Aureobasidium pullulans RBF 4A3, isolated from the flower Caesulia axillaris,¹⁷ was used for pullulan production. Stock cultures of the isolate were maintained in 20% glycerol at −80 °C and subcultured monthly on YPD agar plates. The freshly grown cultures from the agar plates were then transferred to a 250 mL flask containing a seed culture medium (pH 6.0), comprising of glucose (2% w/v), yeast extract (1% w/v) and peptone (2% w/v) for inoculum development.¹⁸ The culture was incubated at 28 °C at 200 rpm in a rotary shaker for 24 h.

2.3.2 Pullulan fermentation. Five different pullulan batch fermentations were executed according to the process parameters laid out in Table 3 to understand the suitability of the developed model. The pullulan fermentation was carried out in a 5 L STR containing 3.2 L of production medium. The production medium (pH 6.2) consisted of 15.5% (w/v) glucose, 3.68% (w/v) yeast extract and 1.75% (w/v) peptone.¹⁸ The bioreactor was equipped with two conventional flat-bladed Rushton turbine impellers, and air was sparged using a perforated ring-shaped sparger. The temperature was maintained at 28 °C throughout the fermentation. The pH was continuously monitored but not controlled during the fermentation. The production medium for each fermentation batch was inoculated with 5% (v/v) of seed culture. Furthermore, all the fermentation batches were analyzed to obtain any marked influence of initial k_La values on the fermentation profile.

Table 3 Operating parameters for executed viscous pullulan fermentation batches

Batch no.	Run no.	Agitation (rpm)	Aeration (vvm)	Impeller spacing (inch)	kLa value (min^−1)
1	Optimized	600.00	2.50	2.80	1.99
2	13	600.00	0.75	1.10	1.1
3	15	400.00	1.63	2.55	1.4
4	17	63.64	1.63	2.55	0.28
5	7	200.00	0.75	1.10	0.29

---

2.3.3 Analytical methods. Each of these batches was analyzed in terms of biomass, pullulan production and residual sugar concentration by harvesting fermentation broth at various time intervals. Collected fermentation broth was centrifuged at 10 000 rpm for 20 min (Sigma 4–16 K), and cell-free supernatant was used for residual sugar analysis as per the method described by Miller (1959).¹⁹ The cell pellet was dried overnight at 80 °C in an oven until a constant weight was obtained, and the dry-cell weight was calculated for biomass analysis. The exopolysaccharide was precipitated from cell-free fermentation broth by adding cold absolute ethanol in the ratio of 1 : 2 (v/v) and kept at 4 °C overnight for complete precipitation.¹⁸ The precipitate obtained was then dried overnight at 80 °C to remove residual ethanol. The crude exopolysaccharide was further purified by dialysis and re-precipitation as previously described by Choudhury et al. (2011).¹⁷ The influence of volumetric mass transfer coefficient on culture morphology was also investigated by the microscopic examination of harvested cells after a fixed interval using an Olympus CK X41 microscope.

3. Results and discussions

3.1 Statistical model development and optimization

Single-point optimization techniques do not consider the interaction amongst governing factors and may not include the appropriate optimizing conditions.²⁰ RSM investigates the univariate responses, as well as the interactive effects of all factors on the selective responses, through a regressed model governed by a suitable polynomial equation. Herein, RSM was used to analyze the effects of three fermentation parameters, (A) agitation, (B) aeration and (C) spacing between two impellers, on k_La value or mass transfer capability in a fermenter. A design evaluation indicated that the model has 9 degrees of freedom (df) with 5 df for ‘lack of fit’ and 4 df for pure error, suggesting that this design will be suitable for appropriate model development. Furthermore, in most of the cases, VIF (variance inflation factor) values were close to 1, describing the ability of the model to estimate the coefficients properly.

The model defining second-order polynomial in terms of coded factors is given as follows:

Y = 1.31 + 0.48A + 0.26B + 0.081C + 0.14AB − 0.054AC + 0.064BC − 0.031A² − 0.098B² − 0.26C²

The statistical significance of the model was confirmed by the Fisher's test for the analysis of variance (ANOVA), which revealed that the second-order polynomial function fitted well with the experimentation data (Table 4). The model F-value was observed to be 29.98, which implied that model was significant and that there is only 0.01% chance to obtain results not foreseen by the regression model. The large P-value for lack of fit (0.1874 > 0.05) showed that ‘lack of fit’ was insignificant relative to pure error.

Table 4 Analysis of variance (ANOVA) for all model terms

Source	ANOVA for kLa optimization
	Sum of squares	df	Mean square	F value	Prob > F
Model	5.36	9	0.60	29.98	<0.0001
A-agitation	3.12	1	3.12	157.24	<0.0001
B-aeration	0.95	1	0.95	47.66	<0.0001
C-impeller position	0.089	1	0.089	4.50	0.0630
AB	0.17	1	0.17	8.32	0.0180
AC	0.023	1	0.023	1.16	0.3088
BC	0.033	1	0.033	1.64	0.2327
A^2	0.013	1	0.013	0.64	0.4433
B^2	0.13	1	0.13	6.57	0.0305
C^2	0.93	1	0.93	46.60	<0.0001
Residual	0.18	9	0.020
Lack of fit	0.14	5	0.027	2.60	0.1874
Pure error	0.042	4	0.010
Cor total	5.54	18

---

Previously, Jiang (2010) employed the statistical optimization technique to study the combinatorial effects of temperature, time and initial pH on pullulan production.²¹ The F-value (14.71) and the P-value for lack of fit (0.259 > 0.05) of the model, as developed by Jiang, was found to be considerably similar with the model developed in the present study. Therefore, it could be suggested that the developed model is suitable for navigation within the design space. This statement was further strengthened by the fact that the R² (0.9677) value was in reasonable agreement with adjusted R² (0.9354) in the present report. These high R² values coupled with high model F-value and insignificant lack of fit indicated that the model predictions can appropriately explain the cumulative, as well as univariate, effects of all the selected variables on the k_La values of the fermenter.²² The model was further validated by executing a batch fermenter run with an aqueous system as per optimized conditions (600 rpm, 2.5 vvm and 2.8′′ impeller spacing) predicted by the model. The actual value of the response, i.e. k_La value (1.92 min⁻¹), was found to be in agreement with the predicted k_La value (1.99 min⁻¹); thus, the model was successfully validated. Parity plots (Fig. 1) also demonstrated good correlation between the actual and predicted values, indicating the aptness of the model.

Fig. 1 Parity plot: showing the relation between actual response and predicted values for k_La estimation.

3.2 Interactive effect of the selected variables on mass transfer

Three different response surface plots (Fig. 2a–c) were generated using the developed model to understand the effects of interactions among the selected variables on k_La. The response surface graph, presented in Fig. 2a, presents the interaction between impeller spacing and agitation speed at specific aeration rates.

Fig. 2 (a) Interactive effect of agitation and impeller position on k_La. Aeration was kept at a fixed value and the effect of agitation variation and impeller spacing on k_La values was investigated with the help of this response surface plot. (b) Dependence of k_La values on a possible combination of aeration and impeller position. Agitation was fixed at a constant value and the effect of aeration and impeller spacing on k_La value was investigated with the help of this response surface plot. (c) Dependence of agitation and aeration value interactions on the k_La value in a fermenter. Impeller spacing was fixed and the combined effect of aeration and agitation variations on k_La values was investigated with the help of this response surface plot.

This plot indicated that at lower values of impeller speed, the k_La value was small; however, upon increasing the impeller speed, k_La values were also enhanced. Conversely, in the case of impeller position, k_La was found to increase gradually as the spacing between the two impellers was increased; however, a further increase in spacing between the impellers beyond a critical value resulted in the reduction of volumetric mass transfer coefficient values. This inference obtained from the plot can further be justified by comparing the observed responses in Table 2. The data clearly indicated that increasing impeller separation from 0.11′′ to 2.55′′ (run no. 9 and 15) significantly enhanced the k_La in the fermenter (0.28 min⁻¹ to 1.4 min⁻¹), whereas a further separation of impellers from 2.55′′ to 4.99′′ (run no. 15 and 10) created a drop in the k_La value of the fermenter (1.4 min⁻¹ to 0.8 min⁻¹). This observation is in accordance with the previous report of Mishra and Joshi (1994), which suggested that the distance between two impellers has an enormous impact on mass transfer.²³ This group reported that if the impellers are placed too far apart, the chances of material being exchanged between the circulation loops is reduced, resulting in a poorer overall mixing and mass transfer. Conversely, if the impellers are moved closer together, the circulation loops from each impeller encroach on each other, allowing for improved fluid interchange.^12,13 However, placing the impellers close to each other may lead to inefficient mixing of large volumes of liquid in the upper or lower regions of the vessel. Therefore, although inter-impeller mixing is improved, overall mass transfer is reduced.

Similar observations could also be made while analyzing another response surface plot, which described the dependence of k_La on the distance between two impellers and air flow rate at specific agitation speeds (Fig. 2b). The response surface plot clearly indicated that although k_La increased with an increase in distance between the two impellers, it decreased when the distance between these two impellers were enhanced beyond a critical value, suggesting the importance of impeller positions in the creation of ideal mixing zones within the bioreactor.

Previously, Hudcova et al. (1989) also made similar observations¹⁵ and suggested that larger distances between the two impellers would result in the formation of parallel individual mixing zones in the fermenter for each impeller and would negatively affect the oxygen mass transfer rate. Conversely, the response surface plot also suggested that although the k_La values initially increased with the enhancement of aeration rates, enhancement was not that significant at very high aeration rates. These observations could be further confirmed by comparing values from run numbers 7 and 4, yielding k_La values of 0.29 min⁻¹ and 0.56 min⁻¹, respectively, wherein higher aeration rates facilitated improved mass transfer (Table 2). In run numbers 16 and 18, although the air flow rate was almost doubled, the k_La values (1.3 min⁻¹ and 1.4 min⁻¹) were not evidently increased. This could be attributed to the fact that at high aeration rates, the impeller is unable to disperse all the gas impinging on it and may become flooded. Moreover, higher aeration rates reduce the gas bubble retention time, and thus limits oxygen transfer.¹² As the oxygen flow rate is decreased, the impeller blades begin to process the gas and bubbles are dispersed towards the walls of the fermenter resulting in impeller loading. Thus, it may be concluded that neither low nor high air flow rates could solve the problems of inefficient mass transfer.

Fig. 2c depicts the effect of aeration and agitation values on k_La at specific impeller positions. The response surface plots have shown incremental gain in k_La values with the simultaneous enhancement of both of these variables. However, it showed a tendency to plateau at higher ranges of aeration and agitation values. Furthermore, a closer examination of the data (Table 2) showed that the increase in agitation value from 63 rpm to 400 rpm (run numbers 17 and 19) caused an increase in the k_La value from 0.28 min⁻¹ to 1.4 min⁻¹, whereas an increase in agitation from 400 rpm to 736 rpm (run numbers 1 and 19) increased the k_La value from 1.4 min⁻¹ to 2.1 min⁻¹, which was not so large as compared to previous runs (run numbers 17 and 19). This plateau trend can be attributed to the saturating effect of high agitation values. However, a slightly curved response plot for aeration values indicated a dominant agitation role on the response.

On comparing run numbers 16, 18 and 1, it was observed that aeration could not increase k_La values in run numbers 18 and 16, whereas high agitation resulted in enhanced k_La values, as shown in run numbers 16 and 1. This observation is in agreement with the previous report of Kim et al. (2002), which demonstrated the effect of agitation speed on pigment fermentation and reported that the k_La values increased gradually from 0.003 to 0.029 sec⁻¹ as the rotational speed was increased from 200 to 700 rpm.²⁴ In a similar study, Moscovici et al. (1996) reported that pullulan concentration increased with an enhancement in agitation rate.²⁵ This favorable effect was attributed to improved oxygen mass transfer and/or changes in the percentage of yeast-like, unicellular cells, which are often considered to be the major producers of pullulan. Thus, agitation plays a prominent role in mass transfer because it contributes both to air bubble distribution as well as the mixing of the system. However, aeration only affects the retention time of oxygen flow with a corresponding increase or decrease in gas velocity in the broth at high or low rates of aeration, respectively.⁹

Most of the earlier studies aimed at enhancing the mass transfer capability of a fermenter by individually increasing agitation or aeration within a prefixed range.^11,26–28 The effects of varying agitation or aeration values are observed on microbial performance, and the best optimal value is selected. However, these observations are generally aliased due to the univariate evaluation of the effects of agitation, aeration, and impeller type on k_La values. Roukas and Mantzouridou (2001) studied the effect of aeration rates only on pullulan production and observed that polysaccharide concentration increased significantly as aeration rates increased to 2 vvm, which then decreased at values higher than 2 vvm.¹¹ Similarly, Göksungur et al. (2005) investigated the effects of aeration, agitation and sugar concentration on pullulan production using response surface methodology.²⁷ However, the polysaccharide yield (17.2 g L⁻¹) and productivity (0.18 g L⁻¹ h⁻¹) were very low even at the optimum levels of process variables (aeration rate, 2.36 vvm; agitation speed, 345.3 rpm; and initial sugar concentration, 51.4 g L⁻¹). In contrast to these findings, Gibbs and Seviour (1996) studied the effects of agitation rate and oxygen saturation on polysaccharide production and found that pullulan yield was drastically reduced at higher agitation rates (≥750 rpm).²⁸ Cell deformation and damage suffered as a result of high shear rates, could be the main reason for the reduction of metabolite production at high agitation rates.¹⁰ Gibbs's group further reported that polysaccharide production at high agitation rates could be enhanced by maintaining the partial pressure of oxygen at low levels during the initial phase of fermentation.²⁸ All these reports suggested that increasing either agitation or aeration might not solve mass transfer issues. Moreover, to the best of our knowledge, none of the earlier reports have studied the impact of impeller spacing on mass transfer and pullulan production.

Unlike previous studies, the present report attempted to understand the cumulative effects of fermenter operating variables (such as aeration, agitation, and impeller spacing) on k_La in a fermenter with the help of a model developed in a Newtonian system through the employed statistical design (Table 2). Interestingly, in run numbers 13 and 15 (Table 3), k_La values obtained were almost comparable, although operating variables were far different from each other in both of the runs. In this case, the effect on k_La values, which was produced by a one-third reduction in agitation speed (600 rpm to 400 rpm), almost subsided or was compensated with a 46% increase in aeration values (0.75 vvm to 1.63 vvm) and a 44% increase in impeller separation (1.1′′ to 2.55′′). Similarly, in run numbers 17 and 7, a 70% decrease in agitation along with a nearly 50% elevation of aeration, as well as impeller separation, had achieved similar k_La values in the Newtonian system. The previously described correlation between operating variables and k_La values was not linear, suggesting that k_La values could not be enhanced in a fermenter by increasing only single variables such as agitation, aeration or impeller spacing; rather, a cumulative effect of all possible variables needs to be studied for proper evaluation of the effect on k_La values.

3.3 Effect of mass transfer on pullulan fermentation

The efficiency of the developed model in case of viscous fermentation was tested by applying selected operating conditions in pullulan fermentation. k_La values were selected at three different levels (high-1.99 min⁻¹; medium-1.4 min⁻¹ and 1.1 min⁻¹ and; low-0.28 min⁻¹ and 0.29 min⁻¹) from the statistical design. Accordingly, five pullulan fermentation batches were executed with operating conditions governed through these k_La conditions (Table 3). Fermentation profiles of all pullulan batches revealed that the activity of A. pullulans in various aspects of growth and product formation, as well as sugar assimilation, is closely correlated to the initial k_La values of the fermentation system (Table 5).

Table 5 Fermentation kinetics of A. pullulans cultured at five different levels of k_La

Run no.	Initial kLa values (min^−1)	Final biomass concentration (g L^−1)	Final EPS concentration (g L^−1)	Total substrate consumption (%)
Optimized	1.99	61.9	74	99
13	1.4	50	52	97
15	1.1	46	40	96.1
17	0.28	14	16.2	93.5
7	0.29	14.7	15.8	93.9

---

Fermentation processes with equal k_La values (0.28 min⁻¹ and 0.29 min⁻¹) demonstrated similar profiles for biomass, pullulan production and sugar consumption, irrespective of their operating conditions (Fig. 3a to d). Thus, a cumulative contribution of all variables was clearly reflected in the batch kinetics of pullulan fermentation.

Fig. 3 Graphical representation of batch kinetics observed in the fermentation of A. pullulans carried out under different k_La conditions: (a) biomass accumulation rate; (b) sugar assimilation trend; (c) rate of pullulan production; and (d) dissolved oxygen concentration profiles during fermentation.

A maximum biomass concentration of 61.92 g L⁻¹ was obtained within 46 h of fermentation run when the cultivation was performed at a higher initial k_La value (1.99 min⁻¹), whereas the cellular growth was reduced by nearly 18% and 77% when fermentation was executed at lower initial k_La values (1.4 min⁻¹ and 0.28 min⁻¹, respectively). A comparative growth curve plot for all batches (Fig. 3a) revealed that operation at higher initial k_La values supported the exponential growth rate of the microbe. A similar observation was reported by Himabindu and Gummadi (2015) during the fermentative production of xylitol by Debaryomyces nepalensis.²⁹ The growth curve of cultures grown at an initial k_La value of 1.99 min⁻¹ exhibited a plateau after 30 h of the fermentation run, whereas this plateau was previously observed at the 21st h in batches with initial k_La values of 0.29 min⁻¹ and 0.28 min⁻¹, suggesting an earlier onset of stationary phases at lower initial k_La values. Low oxygen availability cease the culture growth, and thus result in a remarkably low pullulan production.

Concomitant with cell growth, the glucose concentration of fermentation broth decreased prominently with increasing k_La values (Fig. 3b). Almost complete sugar depletion was observed at an initial k_La value of 1.99 min⁻¹, whereas 3.0%, 3.9% and 6.5% of residual sugar were present in the other batches of 1.4 min⁻¹, 1.1 min⁻¹ and 0.28 min⁻¹ k_La, respectively. This trend indicated that the glucose was efficiently utilized in cultures operating at optimum k_La conditions (1.99 min⁻¹), compared with its lower counterparts (1.1 min⁻¹, 1.4 min⁻¹ and 0.28 min⁻¹). Lazaridou et al. (2002) observed a similar trend and reported that the assimilation of sugars increased with oxygen availability during pullulan fermentation,³⁰ which was directly related with oxygen transfer and it simultaneously increased with an increase in biomass and pullulan concentration.

Pullulan production followed a similar pattern of microbial growth, suggesting that pullulan was a growth-associated metabolite³⁰ (Fig. 3c). The maximum pullulan concentration (74 g L⁻¹) was obtained at the 46th h of fermentation executed at a k_La value of 1.99 min⁻¹, wherein the maximum dry cell weight, as well as the lowest residual concentration (1%), was observed at a similar age. The higher initial k_La value (1.99 min⁻¹) facilitated improved microbial growth and pullulan production with a maximum polymer yield and productivity of 0.47 g g⁻¹ of substrate consumed and 1.61 g L⁻¹ h⁻¹, respectively, whereas the batches operated at the initial k_La values of 1.4 min⁻¹, 1.1 min⁻¹ and 0.28 min⁻¹ exhibited lower pullulan concentrations of 52 g L⁻¹, 40 g L⁻¹ and 16.21 g L⁻¹, respectively, along with lower sugar assimilation rates, and in turn led to lower polymer yields (0.33 g g⁻¹, 0.26 g g⁻¹ and 0.11 g g⁻¹, respectively) (Table 5). This observation is in accordance with the previous reports on pullulan fermentation.^30,31 Lazaridou et al. (2002) stated that biomass level and polysaccharide concentration increased significantly with oxygen availability and that the highest concentration of polysaccharide (49 g L⁻¹) was obtained in the culture grown at 700 rpm, whereas in cultures grown at 300 and 500 rpm, the maximum concentration of pullulan was 34 g L⁻¹ and 43 g L⁻¹, respectively.³⁰

Similarly, Rho et al. (1988) observed that a high oxygen transfer rate or high k_La favored both cell growth and pullulan synthesis,³¹ which was indicated by a continuous increase in biomass concentration under the highest simulated oxygen transfer condition, e.g. in the Erlenmeyer flask containing the smallest medium volume. Moreover, this group demonstrated that the pullulan synthesis rate and pullulan yield were directly proportional to the oxygen availability. Broth viscosity also increased as k_La values were increased, which was due to the continuous enhancement in biomass and pullulan concentrations till 46 h at high k_La values (ESI Table S1†). A high initial k_La value (1.99 min⁻¹) facilitated improved mass transfer rate, which ultimately resulted in significantly higher viscosity of 25.95 cP after 46 h due to enhanced pullulan production. However, at lower k_La values (1.1 min⁻¹ and 0.29 min⁻¹), viscosity was reported to be comparatively low, i.e. 16.41 cP and 5.76 cP, respectively, at the 46th h.

Fig. 3d shows the dissolved oxygen (DO) concentration (as the percentage of air saturation) profiles in the broth during the fermentative production of pullulan under different k_La conditions. It was observed that in the case of lower initial k_La values, namely, 0.29 min⁻¹ and 0.28 min⁻¹, the DO concentration falls rapidly to a low level (below 20%) within 5 to 8 h, and then remains constant until the end of fermentation. In case of 1.1 min⁻¹ and 1.4 min⁻¹ k_La, DO concentration dropped to a very low level (9% and 4%, respectively) within 12 h and 10 h, respectively, and then remained constant. Conversely, at a high initial k_La value of 1.99 min⁻¹, the decrease in DO level was more gradual, and it was possible to maintain DO within 20–40% of the range because of improved mass transfer in this case. This observation suggested that oxygen availability throughout the entire fermentation process resulted in improved pullulan production.

Interestingly, the maximum specific growth rates (μ_max) of the microbe were also increased from 0.046 h⁻¹ to 0.084 h⁻¹ when initial k_La values were enhanced from 0.28 min⁻¹ to 1.99 min⁻¹, respectively (Fig. 4). Previously, Gibbs and Seviour (1996) reported that the organism's growth rate (μ) was slower at 125 rpm, which may be a reflection of a reduced initial k_La values at low agitation rates.²⁸ The actual oxygen transfer to the cells was reduced at a low agitation rate that caused the large air-bubble size and poor bubble breakup, leading to the formation of viscous, non-stirred zones. Reduced mass transfer to cells could affect the organism's growth and biopolymer yield by blocking nutrient transfer and its consumption by the cells.¹¹ Batch productivity data also supported the importance of high k_La value implication in enhancing pullulan production. Fig. 4 illustrates the comparative account of batch productivity, product yield and μ_max at various k_La values. All these findings suggest that metabolite production by A. pullulans becomes enhanced under high oxygen transfer conditions, which is in accordance with previous reports.^27,31

Fig. 4 Effect of different k_La on batch productivity and pullulan yield at the 46th h of fermentation. Maximum specific growth rate (μ_max) values have also been demonstrated graphically to show the effect of k_La value on A. pullulans fermentation.

Therefore, it may be suggested that enhanced k_La conditions in the fermenter are responsible for the improvement in pullulan fermentation, which is attributed to the improved oxygen mass transfer.

Interestingly, the microscopic analysis of microbial samples grown at the same initial k_La values revealed that both the sets of samples displayed similar cellular morphology despite differences in other operating conditions. Furthermore, cell morphology analysis in pullulan fermentation batches revealed that the optimized pullulan batch (k_La 1.99 min⁻¹) supports the formation of the swollen cell morphology of microbe (Fig. 5a), whereas in the case of pullulan batches with the lowest initial k_La (0.28 min⁻¹), swollen cells were scarce (Fig. 5b). Previously, Campbell et al. (2004) reported that the swollen cellular morphology of A. pullulans were mostly responsible for pullulan production.³² Similar observations have also been reported by other researchers.^11,33 These reports are in line with the observations made in the present study, in which conditions yielding to higher k_La values resulted in a greater number of swollen cells with regard to higher pullulan production.

Fig. 5 Comparative morphological analysis of Aureobasidium pullulans grown under various k_La conditions: (a) higher k_La (1.99 min⁻¹); (b) lower k_La (0.28 min⁻¹).

3.4 A comparative account of pullulan fermentation

Essentially, the commercial success of any fermentation technology mostly depends on its economic viability. Therefore, it is important to employ a fermentation process that would result in optimum product formation with high productivity and yield for the economical production of pullulan. In literature, several investigators have attempted various strategies, including the statistical optimization of media components and process variables, to obtain high pullulan production.^11,27,30 However, low product concentration and productivity is still one of the major bottlenecks towards the commercialization of pullulan fermentation. In a previous report, Lazaridou et al. (2002) reported the production of 49 g L⁻¹ of pullulan with very low pullulan productivity (0.34 g L⁻¹ h⁻¹).³⁰ This low productivity would enhance the operative cost and ultimately increase the cost of biopolymer production. Bandaiphet and Prasertsan (2006) studied the effect of k_La values on cell growth and biopolymer yield at high agitation rates and reported a very low pullulan concentration and productivity of 4.76 g L⁻¹ and 0.07 g L⁻¹ h⁻¹, respectively, in 72 h of fermentation.⁹ In our previous report, we studied the effect of complex medium components, particularly glucose, yeast extract and peptone, on pullulan production using response surface methodology. Although a maximum pullulan concentration of 70.43 g L⁻¹ was obtained using the optimized concentration of variables, but pullulan productivity was still low (0.73 g L⁻¹ h⁻¹).¹⁸ In another study, Xia et al. (2011) demonstrated that A. pullulans AP239 could accumulate 39 g L⁻¹ pullulan in 60 h of fermentation, which ultimately resulted in a total productivity of 0.65 g L⁻¹ h⁻¹.³⁴ Recently, Özcan et al. (2014) compared the pullulan production in airlift and bubble-column bioreactor using statistical design to understand the effect of three factors, particularly initial sugar concentration, aeration rates and incubation time, on pullulan fermentation.³³ This resulted in pullulan production with a yield of 0.402 g pullulan g⁻¹ sugar consumed and productivity of 0.30 g L⁻¹ h⁻¹ (Table 6).

Table 6 Comparative literature report of pullulan production carried out under various k_La conditions

Reference	Operating conditions	Carbon source	Fermentation time (h)	Pullulan concentration (g L^−1)	Product yield (Yp/s)	Productivity (g L^−1 h^−1)
Present study	Agitation-600 rpm, aeration-2.5 vvm	Glucose	46	74	0.47	1.61
9	Agitation-200 rpm, aeration-1.25 vvm	Sucrose	72	4.76	0.16	0.07
28	Agitation-250 rpm, aeration-0.4 vvm	Glucose	140	11.27	0.37	0.08
30	Agitation-700 rpm, aeration-0.23 vvm	Beet molasses	144	49	0.49	0.34
11	Agitation-200 rpm, aeration-2 vvm	Sucrose	144	30	0.6	0.20
27	Agitation-345.3 rpm, aeration-2.36 vvm	Sucrose	96	17.2	—	0.18
35	Agitation-150 rpm, aeration not mentioned	Sucrose	168	22.6	0.32	0.13
36	Agitation-200 rpm, aeration-1.5 vvm	Sucrose	168	25.8	0.35	0.15
34	Agitation-800 rpm, aeration-1.3 vvm	Sucrose	60	39	—	0.65
18	—	Glucose	96	70.43	0.45	0.73
33	Aeration-1.93 vvm (airlift bioreactor)	Sucrose	128	38.77	0.402	0.30

---

Earlier reports suggested that, neither high agitation nor high aeration rates have supported high yield and productivity during pullulan production.^11,27,28 Poor mass transfer is one of the major problems that often leads to lower yield and productivity in the viscous fermentation system. However, none of the previous reports investigated the cumulative effect of all three operating variables, particularly agitation, aeration and impeller spacing, on mass transfer with regard to pullulan production. In the present study, it was possible to obtain a remarkably high pullulan concentration, 74 g L⁻¹, within 46 h of fermentation time at 600 rpm. The batch productivity (1.61 g L⁻¹ h⁻¹) obtained in the present study was also significantly higher than that in previously published reports, and the product yield obtained was reasonably good. The aforementioned comparisons suggested that the improvement in pullulan production could be achieved by applying the model developed using a conjoint and multivariate optimization of selected variables.

4. Conclusion

The present study has demonstrated that a conjoint multivariate effect of operating variables is more effective for obtaining successful oxygen mass transfer in a system, rather than its univariate responses. The comparative data analysis for pullulan production from the current study, as well as previous reports, has suggested that an improvement in pullulan production process could be achieved by applying a conjoint and multivariate optimization of fermentation variables such as aeration, agitation, and impeller position. This conjoint effect has been reflected in the sustainable enhancement of bioreactor performance, which has been suggested through higher productivity and economic viability of the fermentation process in the present study. The obtained batch productivity for pullulan by applying the model suggested that the optimized variables had not only proven the model validation but also have given this process an economic advantage and a scope for its commercial exploitation. Thus, the present study suggested the application possibilities of such models in resolving the mass transfer challenges of viscous fermentation systems.

Acknowledgements

The authors gratefully acknowledge the Council of Scientific and Industrial Research (CSIR), Government of India for financial support.

Notes and references

1. R. S. Singh, G. K. Saini and J. F. Kennedy, Carbohydr. Polym., 2008, 73, 515–531.
2. V. D. Prajapati, G. K. Jani and S. M. Khanda, Carbohydr. Polym., 2013, 95, 540–549.
3. R. S. Singh, N. Kaur and J. F. Kennedy, Carbohydr. Polym., 2015, 123, 190–207.
4. S.-O. Enfors, M. Jahic, A. Rozkov, B. Xu, M. Hecker, B. Jürgen, E. Krüger, T. Schweder, G. Hamer and D. O'beirne, J. Biotechnol., 2001, 85, 175–185.
5. R. J. Seviour, B. McNeil, M. L. Fazenda and L. M. Harvey, Crit. Rev. Biotechnol., 2011, 31, 170–185.
6. Y. Kawase, B. Halard and M. Moo-Young, Biotechnol. Bioeng., 1992, 39, 1133–1140.
7. Z. Zhang and C. R. Thomas, Chem. Eng. Commun., 1995, 140, 207–217.
8. H. G. lawford and J. D. Rousseau, Appl. Biochem. Biotechnol., 1991, 28, 667–684.
9. C. Bandaiphet and P. Prasertsan, Carbohydr. Polym., 2006, 66, 216–228.
10. P. Çalık, G. Çalık and T. H. Özdamar, Biotechnol. Bioeng., 2000, 69, 301–311.
11. T. Roukas and F. Mantzouridou, J. Chem. Technol. Biotechnol., 2001, 76, 371–376.
12. P. M. Doran, Bioprocess Engineering Principles, Elsevier, Amsterdam, 2nd edn, 2013, ch. 8, pp. 255–332.
13. D. Cronin, Trans. Inst. Chem. Eng., Part C, 1992, 72, 35–40.
14. F. Garcia-Ochoa and E. Gomez, Biotechnol. Adv., 2009, 27, 153–176.
15. V. Hudcova, V. Machon and A. Nienow, Biotechnol. Bioeng., 1989, 34, 617–628.
16. A. Kiełbus-Rąpała and J. Karcz, Mass Transfer in Multiphase Systems and its Applications, ed. E.-A. Mohamed, InTech Open Access, 2011, ISBN: 978-953-307-215-9, ch. 5, pp. 93–116.
17. A. R. Choudhury, P. Saluja and G. Prasad, Carbohydr. Polym., 2011, 83, 1547–1552.
18. A. R. Choudhury, M. Bhattacharyya and G. Prasad, Biocatal. Agric. Biotechnol., 2012, 1, 232–237.
19. G. L. Miller, Anal. Chem., 1959, 31, 426–428.
20. F. Francis, A. Sabu, K. M. Nampoothiri, S. Ramachandran, S. Ghosh, G. Szakacs and A. Pandey, Biochem. Eng. J., 2003, 15, 107–115.
21. L. Jiang, Carbohydr. Polym., 2010, 79, 414–417.
22. A. Mehta, G. Prasad and A. R. Choudhury, Int. J. Biol. Macromol., 2014, 64, 252–256.
23. V. Mishra and J. Joshi, Chem. Eng. Res. Des., 1993, 71, 563–573.
24. H. J. Kim, J. H. Kim, H. J. Oh and C. S. Shin, Process Biochem., 2002, 38, 649–655.
25. M. Moscovici, C. Ionescu, C. Oniscu, O. Fotea, P. Protopopescu and L. D. Hanganu, Biotechnol. Lett., 1996, 18, 787–790.
26. A. Wecker and U. Onken, Biotechnol. Lett., 1991, 13, 155–160.
27. Y. Göksungur, S. Dağbağlı, A. Uçan and U. Güvenç, J. Chem. Technol. Biotechnol., 2005, 80, 819–827.
28. P. Gibbs and R. Seviour, Appl. Microbiol. Biotechnol., 1996, 46, 503–510.
29. K. Himabindu and S. N. Gummadi, Br. Biotechnol. J., 2015, 5, 24–36.
30. A. Lazaridou, T. Roukas, C. Biliaderis and H. Vaikousi, Enzyme Microb. Technol., 2002, 31, 122–132.
31. D. Rho, A. Mulchandani, J. H. Luong and A. LeDuy, Appl. Microbiol. Biotechnol., 1988, 28, 361–366.
32. B. S. Campbell, A.-B. M. Siddique, B. M. McDougall and R. J. Seviour, FEMS Microbiol. Lett., 2004, 232, 225–228.
33. E. Özcan, S. Sargın and Y. Göksungur, Biochem. Eng. J., 2014, 92, 9–15.
34. Z. Xia, S. Wu and S. Pan, Carbohydr. Polym., 2011, 86, 1814–1816.
35. K.-C. Cheng, A. Demirci, J. M. Catchmark and V. M. Puri, J. Food Eng., 2010, 98, 353–359.
36. K.-C. Cheng, A. Demirci and J. M. Catchmark, Food Sci. Technol. Int., 2011, 17(2), 99–109.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra03715h

‡ Authors have contributed equally.

---