A Novel Strain of Stenotrophomonas Acidaminiphila Produces Super-active Alkaline Protease During Cassava E�uent Fermentation: Process Optimization, Kinetic Modeling and Scale-up

Purpose: Microbial fermentations for value-added metabolites production are exploited for efficient bioconversion of agro-industrial wastes for the dual purposes of pollution abatement and cost-effectiveness. Methods: In the present study, the regular 2-level factorial design was employed to screen fermentation parameters that enhance production of a novel alkaline protease by a strain of Stenotrophomonas acidaminiphila using cassava processing effluent as substrate. Data from randomized experiments of central composite rotatable design for improved enzyme activity, guided by path of steepest ascent experiments, were modeled and optimized by response surface methodology (RSM). Shake flask kinetics of production under optimized conditions was modeled by logistic and modified Gompertz models and determinations of maximum specific growth rate, µ max , maximum volumetric rate of substrate consumption, r sm, maximum volumetric rate of biomass formation, r xm and specific yield of product, Y p/x were made. Results : Logistic model poorly fitted RSM-generated product formation and substrate consumption data. However, biomass formation was accurately fitted (adjusted r 2 >99%), with µ max of 0.471 h -1 . The modified Gompertz model, on the contrary, more accurately fitted all three major response data with minimal mean squared error. Potential for scale-up of bioprocess evaluated in 5-L bioreactor satisfactorily revealed 8.5-fold more substrate consumption in bioreactor than in shake flask. The 86.76-fold aqueous two-phase system-purified protease had a specific activity of 1416.73 Umg -1 which improved with increasing surfactant concentration. Conclusion : These results suggest significant bioprocess potential for sustainable cassava effluent management and concomitant commercial production of alkaline protease for industrial detergent application.


Introduction
Proteases constitute the largest enzyme market worldwide (Lazim et al. 2009) arising from the copious commercial applications and the range of proteinaeous substrates they catalyze. They are produced by all biological systems including plants, animals and microorganisms; however, commercial interest rests mostly on microbial sources by reason of their ease of production, recovery and amenability to improved productions in terms of cost and activity, in addition to demonstration of substrate diversity (Razzaq et al. 2019). Among microorganisms, bacterial proteases are the most encountered but fungal proteases are also available for commercial use. Bacterial and fungal proteases appear to have significant differences in terms of molecular weight and stability to environmental conditions of temperature, pH and ionic strength (Saggu and Mishra 2017). The conditions of temperature and pH are particularly significant in the classification of proteases and have therefore influenced their commercial applications in industries as varied as detergent, tannery, pharmaceutical, leather and environmental nitrogen cycling (Saxena and Singh 2010;Asitok et al. 2017;Ammasi et al. 2020). Mesophilic proteases are the most abundant but thermostable proteases are more suitable for specialized applications that involve high temperature processing like paper and textile processing (Abusham et al. 2009). In terms of pH discrepancy, proteases are categorized as acidic, neutral and alkaline. Alkaline and neutral proteases dominate literature while acid proteases, produced mostly by filamentous fungi, are less frequently encountered (Gupta et al. 2002;Razzaq et al. 2019;Ammasi et al. 2020).
Like every other value-added microbial product, the natural or baseline yield is usually low and costeffectiveness very poor (Ekpenyong et al. 2021a). To improve production economics, biosynthesis of one or more value-added metabolite during bioconversion of agro-industrial waste on a large scale has been suggested (Ekpenyong et al. 2017a). To improve yield and/or activity, a number of nutritional and environmental requirements of high-yielding microbial strains need to be optimized (Ekpenyong et al. 2017a, b;2021a). Response surface methodology (RSM) and artificial neural network (ANN) methods have become gold standards for bioprocess optimizations for the dual purposes of improving yield and cost-effectiveness (Kanno et al. 2020;Ekpenyong et al. 2021a;Dwibedi et al. 2021). Response surface methodology typically has a limitation in its handling of non-linear relationships, if at all. Artificial neural network methods, on the other hand, are naturally designed to handle complicated non-linear relationships, which in microbial fermentative productions are typically stochastic. There is no end to reports on the superiority of one approach over the other (Karri and Sahu 2018). However, with coefficient of determination, r 2 and mean squared error (MSE) or other performance metrics, one approach could be selected ahead of the other (Lau et al. 2020). Response surface methodology typically involves four basic steps which include selection of process influencing factors using one-factor-at-a-time (OFAT) approach (Ekpenyong et al. 2017a), screening of selected process factors for just significant factors using Placket-Burman design (PBD) or any other 2-level factorial design (Ekpenyong et al. 2017b;Long et al. 2018), moving the levels of significant variables close to the region of experimentation using the path of steepest ascent (Yingling and Zhengfang 2013) and finally designing response surface experiments using any of central composite or Box-Benkhen designs to locate the actual region of interest (Karri and Sahu 2018;Ekpenyong et al. 2021b).
Time-related changes in bioconversion and simultaneous production of value-added biotechnological products are of utmost consideration in deciding bioprocess economics. The need for speed from industry to market becomes necessary for improved turn-over of bioprocesses (Truppo 2017;Boshagh and Rostami 2021). Microbial process turn-over is inherently determined by the growth rate of the transforming microorganism which is substratespecific. Kinetic studies reveal latent parameters of bioprocesses that are amenable to optimization. Such parameters as volumetric substrate consumption or conversion rates are particularly significant in environmental engineering technologies that remove contaminants from test systems (Salahi et al. 2013;Ekpenyong et al. 2021c).
Bioconversion kinetics are usually confirmed by established or newly proposed models which are useful to predict future performances in similar systems. Very frequently in microbial bioconversions, the transforming microorganism uses the substrate as carbon or energy source, thus making determinations of such parameters as length of lag phase, maximum specific growth rate, volumetric and/or specific productivities sacrosanct. The Monod, Logistic, modified Gompertz and Luedeking-Piret models are a few of the models that have been developed to tackle environmental and industrial bioconversion kinetic problems (Mercier et al 1992;Tjørve and Tjørve 2017;Sulieman et al. 2018).
The bacterium Stenotrophomonas acidaminiphila UCCM 00065 was found to produce an alkaline caseinolytic protease with pH optimum of 9.0 at a temperature optimum of 40ºC (Edet et al. 2018). An activity of 153 U was reported under un-optimized conditions. In the present study, RSM was used to optimize bioprocess variables selected by statistical methods towards improved production of the alkaline protease during bioconversion of cassava processing effluent by the bacterium. Authors are not aware of any prior report(s) that describe(s) optimized alkaline protease production by this bacterium, or the promise it holds for environmental pollution abatement. A vial containing stocked bacterium was retrieved from the culture collection, reactivated and quality checked according to collection's guidelines (www.wfcc.info/ccinfo/collection/by_id/652). The phylogenetic tree of the bacterium showing closely related species is presented as Figure 1.

Regular two-level factorial screening of variables for improved protease production
One-factor-at-a-time (OFAT) screening of carbon, nitrogen and phosphorus sources for improved alkaline protease production by the bacterium selected the primary substrate cassava processing effluent (CPE), casein, corn steep liquor and K2HPO4/KH2PO4 (2:1) (Edet et al. 2018) as significant major influencing variables for alkaline protease production by the bacterium. Additional factors included in the screening were cations including Na + , K + , Mg 2+ , Ca 2+ , Zn 2+ , Co 2+ , Mn 2+ , Cu 2+ , Ni 2+ , Fe 2+ , Fe 3+ and Cr 3+ . The regular two-level factorial design (2-LFD) was chosen to screen these 16 factors for inclusion in a medium formulation to enhance alkaline protease production by the bacterium. The design matrix was developed in Design Expert software version 12 with each factor tested at either high (+1) or low (-1) levels. Results of the 2 k factorial experiments were analyzed by ANOVA model and significant adjusted and predicted r 2 values charted the course for future experimentation and conclusion. Protease activity was the only response variable and was determined as described in Edet et al. (2018). One unit of protease was defined as the amount of crude enzyme that was needed to digest 1 mg of azocasein in 1 min under assay conditions. Coefficients of significant main effects of the obtained regression equation were employed to explore the path of steepest ascent experiment. The regular 2-LFD equation for enzyme activity as response variable takes the form expressed as equation 1 where y is protease activity, b0, a constant term, bi, the coefficient of the linear term and and bij coefficient of interaction term, xi and xj the investigated variables and ɛ the error term.

Steepest ascent experimentation
The path of steepest ascent (PSA) was conducted as an intermediary experimentation to move variable conditions as close as possible to the experimental region within the design space. Coefficients of significant factor effects in the first-order model of the regular 2-LFD were employed to calculate step sizes for the PSA (Ekpenyong et al. 2021b). The slope was first calculated using the variable with the largest absolute coefficient, then the coded step sizes were converted into natural variables using the relationship between coded and natural units and the natural step change of the variable with the highest coefficient. A total of 8 PSA experiments were then conducted using the values obtained from the determined step sizes for each of the variables. The point of steepest ascent was scored as the experiment that recorded highest protease activity beyond which activity gradually decreased.
Experiments were compared using ordinary one-way ANOVA in GraphPad Prism 8.4.3 and significant means separated by Tukey's HSD multiple comparisons test at 95% confidence interval.

Central composite design of experiment (DoE) and response optimization
A central composite rotatable design (CCRD) matrix was adopted to investigate the significant linear effects obtained from the regular 2-LFD with levels of factors guided by the step sizes suggested by the PSA. Table 1 presents the factor codes as well as actual levels calculated from step sizes from PSA. A quadratic function of a response surface method (RSM) was employed to investigate the effects of the designed variables and their levels, with contour and surface plots as pictorial representations. Analysis of variance of the multiple regression data for the response variable was used to compare data variability. Coefficients of significant model parameters were employed to build regression model using the method of least squares. The second-order function of the CCRD-RSM took the form below: where y is protease activity, β0, βi, βii and βij are coefficients of the constant, linear (xi), quadratic (xi 2 ) and interaction terms (xixj) of the k factors respectively, with ɛ as error term of computation of protease activity. Model performance was evaluated with adjusted and predicted r 2 values, as well as lack-of-fit error probability.
The response optimizer in Design Expert 12 was employed to optimize variable conditions using the desirability function and parameter importance with regression model of the CCRD as fitness function. Except cassava processing effluent (CPE), all other variables were left at 'in range' mode while goal or response variable (protease activity) was set as 'maximize'. Maximum level of importance was also assigned to both CPE and protease activity.

Confirmation experiment
The optimum variable conditions suggested by CCRD-RSM optimization were employed to set up a fresh fermentation in 1-L Erlenmeyer flask. Experimental set up was in triplicates and enzyme activities determined as previously described (Edet et al. 2018). A difference of less than 5% in protease activities between optimized factor settings and real-life situation at the optimized variable conditions ratified the model as fitting for future predictions of protease production by the bacterium using cassava processing effluent as substrate.

Fermentation kinetics study
Experimental set up was as described in Ekpenyong et al. (2021c) Ekpenyong et al. (2021b). Ionic strength of medium was adjusted to optimized level using NaClO4. Medium pH was adjusted to 7.0 before autoclaving at 121ºC for 15 min. Initial reducing sugar concentration was determined after sterilization by Cooled medium was inoculated with 2% (vv -1 -1.58 x 10 8 cfumL -1 ) of 18-h old seed culture from section 2.6.1 and triplicate flasks incubated at optimized temperature on a rotary shaker adjusted to agitate at 200 rpm for 36 h.

Determination of fermentation response variables
Periodically, at 4 h interval, 5 mL of sample was withdrawn from triplicate flasks and centrifuged at 10,000 x g for 5 min and determinations of total protein, protease activity, biomass concentration and amount of substrate consumed made from appropriate fractions. Protease activity was determined as described by Edet et al. (2018) using azocasein as substrate and protease activity defined as in section 2.2.
Biomass was quantified by the dry cell weight (DCW) technique (Rodrigues et al. 2006) from the pellet obtained from centrifugation at 10,000 x g for 5 min. Biomass concentration (gL -1 ) was given by the regression equation that established the relationship between dry cell weight (DCW) and optical density at 600 nm as given by the equation = ( 600 × 0.633) + 0.031; 2 = 0.9984.
Amount of substrate consumed was calculated by first determining residual reducing sugar at time t h using the DNS assay (Miller 1959) with glucose (Merck, USA) as standard carbohydrate and then subtracting the result from the initial amount at time 0 h. The regression equation for the relationship between glucose concentration and optical density was given by the expression 540 = 0.023 + 0.0282; 2 = 0.9991.
Triplicate time-related response data were subjected to descriptive statistics to obtain mean parameters +/standard error. Means were compared by one-way ANOVA and significant means separated by Tukey's HSD posthoc multiple comparisons test using 95% confidence interval.

Kinetic model-fitting of experimental data and model-performance evaluation
Mean data for the three major responses of biomass formation, protease activity and substrate consumption were fitted to the modified Gompertz and logistic models (Ekpenyong et al. 2021c) to ascertain which would better explain the data or whether a new model would have to be proposed. Equation 3 describes the logistic models for biomass formation; where X is biomass concentration (g/L) at any time t, X0 is the initial biomass concentration at time t0, Xmax is maximum biomass concentration (g/L), µmax is the maximum specific growth rate (h -1 ), S is the amount of substrate consumed, S0 is the substrate concentration at time t0.
The modified Gompertz model equations are presented as equations 4, 5 and 6 for biomass formation, protease activity and substrate consumption respectively.
where X, P and S are biomass, protease activity and amount of substrate consumed at any time of fermentation, t; Xmax and Pmax are maximum biomass concentrations (g/L) and maximum protease activity, Smax is the potential drop in substrate concentration; rxmax and rpmax (g/L.h) are maximum biomass and biosurfactant formation rates and rsmax is the maximum rate of substrate consumption (g/L.h); tlag is the lag time (h) defined as the time to reach exponential biomass formation, protease activity and exponential substrate consumption.
Kinetic parameters were obtained in triplicates and results reported as mean values +/-standard error.
Model performances were evaluated using coefficient of determination, r 2 , mean squared error (MSE) and mean absolute error (MAE) (Ekpenyong et al. 2021a), using the expressions in equations 7, 8 and 9 respectively.
where n is the number of samples, y and ŷ the actual value and predicted values respectively.

Scale-up of fermentative alkaline protease production in 5-L bioreactor
Batch mode fermentation of cassava processing effluent by Stenotrophomonas acidaminiphila UCCM 00065 under optimized conditions was also conducted in a 5-L bench-scale bioreactor (BioStat, Sartorius) with a working volume of 3.5 L. The medium had the same composition as that used in the shake flask experiment.
Temperature was maintained at the CCRD-RSM optimized level while pH, agitation speed and dissolved oxygen were maintained at 7.0 ± 0.2, 150 rpm and 50% respectively. Filter-sterilized air was allowed to flow into the headspace of the vessel at a rate of 1 L/min controlled by a mass flow controller. The bioreactor was equipped with a single in-place sterilization mechanism. The vessel, allowed to cool near optimized temperature after sterilization, was inoculated, through culture inlet, with 2% (v/v) of overnight seed bacterial culture as in the shake flask experiment and reactor operated batch-mode for 4 to 36 h. Each batch of fermentation was set up in triplicates.
Determinations of total protein, total enzyme activity, biomass concentration and residual total carbohydrate were made as earlier described in section 2.6.3. Once again, data was modeled by means of the logistic and modified Gompertz models to determine lag time, maximum specific growth rate, specific and volumetric substrate consumption rates, and volumetric rates of product and biomass formation under bioreactor conditions to compare with kinetic parameters obtained from shake flask studies.

Purification of Stenotrophomonas acidaminiphila UCCM 00065 alkaline protease by aqueous two-phase system
Polyethylene glycol-sodium citrate (PEG-Na + Citrate) system was employed for aqueous two-phase system separation and purification of the enzyme. Five different sizes (molecular weights) of the PEG including 1500, 3000, 4500, 6000 and 7500 were investigated. The component phase systems were constituted by %w/w in a 15 mL graduated centrifuge tube. Total mixture of polymer, water, sodium citrate and 2% crude enzyme in the tube was 100%. Separation was conducted at different pH levels of sodium citrate; pH 7.5, 8.5 and 9.5 and holding temperatures namely 20, 35 and 50ºC. At each separation condition, the mixture was centrifuged at 3000 rpm for 20 min to facilitate phase separation and tube held at 20ºC for 24 h for equilibration (Abd Samad et al. 2017). Total protein and total enzyme activity were determined at each separation condition and results used to calculate specific activity of the enzyme, yield and extent (fold) of purification.

Stability evaluation of purified alkaline protease to surfactants
The stability of purified alkaline protease was evaluated by pre-incubating the enzyme with sodium dodecyl sulfate (SDS), triton X-100 and Tween 80 each at 1, 5 and 10 mM before repeating protease assay using azocasein as substrate. Stability to each denaturant was evaluated in triplicates by calculating the relative activity of the enzyme and data analyzed using two-way ANOVA in GraphPad Prism 8.4.3 software.

Phylogeny of Stenotrophomonas acidaminiphila UCCM 00065
The phylogenetic tree of the producing bacterium Stenotrophomonas acidaminiphila UCCM 00065 is presented as Figure 1.

Regular 2-level factorial design screening experiment
The results of the regular 2-level factorial design (2-LFD) experiments are presented in Table 2  The model was significant at p = 0.0002 < 0.001 with an adjusted r 2 of 98.86% and predicted r 2 of 90.62.

Central composite rotatable design and response surface methodology
The results of CCRD-RSM are illustrated in Figure 3 as contour or 2-D (a) and surface or 3-D (b) plots.
Only the best interactive plot, in terms of protease activity mediated, is reported here. The quadratic model of the

RSM optimization and model confirmation
Results of the optimization experiment are illustrated in Figure  to bring about the response. In the confirmation experiment, mean protease activity obtained using the optimum settings for the significant variables from the optimization experiment was 5784.83 ± 89.64 U as against the 5759.57U protease activity of the response optimizer; some 0.045% difference showing reliability of the suggested optimum levels.

Kinetics, model-fitting and performance evaluation
Results of the kinetic study are illustrated as Figures 7 and 8 Tables 3 and 4. While Table 3  In Table 4

Aqueous two-phase system purification of alkaline protease
Results of the purification of the alkaline protease using one step alkaline two-phase system at pH 10.5 is presented as Table 5. The table shows that 1500 molecular weight polyethylene glycol (PEG) purified the protein most with a fold purification of 85.11 and a specific activity of 1291.91 U/mg. Specific activities generally decreased with increase in size of the PEG as PEG-7500 showed lowest specific activity and fold purification of 156.69 U/mg and 10.32 respectively.

Stability of purified protein to surfactants
The demonstrate the potential suitability of the protease for laundry applications, the purified protease was

Discussion
Stenotrophomonas acidaminiphila is a pale-yellow, strictly aerobic, Gram-negative rod-shaped bacterium belonging to a narrow genus of only about eight species, with S. maltophila as type species (Assih et al. 2002).
Although the bacterium, as well as the type species of its genus, has been described as having great metabolic versatility owing to the wide range of substrates utilized including aromatic hydrocarbons and fipronil (Ryan et al. 2009;Mangwani et al. 2014;Uniyal et al. 2016), only a few carbohydrates or sugars namely maltose, glucose, mannose and fructose are utilized (Assih et al. 2002;Mukherjee and Roy 2016). Although the bacterium does not utilize native starch as whole substrate, its efficient utilization of maltose and glucose prompted its investigation into the bioconversion of the pre-heated starch-based environmental effluent, namely cassava processing effluent (CPE). .0001 LGM = Logistic model; MGM = Modified Gompertz model; SF = Shake flask; BR = Bioreactor; Pmax = maximum protease activity (U); rpmax = maximum volumetric rate of protease activity (Uh -1 ); X0 = initial biomass concentration (gL -1 ); Xmax = maximum biomass concentration (gL -1 ); rxmax = maximum volumetric rate of biomass formation (gL -1 h -1 ); µmax = maximum specific growth rate (h -1 ); tLag = Lag time (h); Smax = Maximum predicted substrate consumption (gL -1 ); rsmax = maximum volumetric rate of substrate consumption (gL -1 h -1 ); Adj. r 2 = adjusted coefficient of determination; MSE = mean squared error; MAE = mean absolute error. The kinetic parameter values are means of triplicate determinations ± standard error. Product a concentration where a is total protein at time t mg - Protein concentration at onset of exponential protein production mg 48.04 73.334 -Pb0 Protease activity at onset of exponential protease activity   . 9 Relative activity versus surfactant concentration plot to determine stability of alkaline protease and potential for application in detergent industry For many years, cassava processing has been a major activity in Nigeria but disposal of waste arising therefrom has remained a major challenge and an environmental concern.
Considering the possibility of interaction between and among significant variables selected through onefactor-at-a-time (OFAT) approach (Edet et al. 2018), the regular 2-level factorial design (2-LFD) was adopted in preference to the popular Placket-Burman design (PBD) to screen for significant bioconversion process variables.
The percent contributions of such interaction terms as CPE*pH, CPE*agitation, CPE*Ca 2+ , CPE*Ni 2+ and CPE*Zn 2+ lend credence to the choice of the screening methodology. Very significantly, the linear effect of agitation was not significant in the study, nevertheless, the interaction term of CPE*agitation was significant suggesting that effect of agitation speed on fermentation could be a function of nature and/or amount of carbon source or that nature of medium determines the speed of agitation required to distribute oxygen in the system.
Selection of corn steep liquor as nitrogen source for synthesis of protease was not surprising because the complex medium contains almost all the amino acids in casamino acids which the bacterium utilizes readily as nitrogen source (Edet et al. 2018). Apart from amino acids, corn steep liquor also supplies the readily metabolizable invert sugars to enhance growth. The significance of divalent cations like Mg 2+ , Zn 2+ , Ni 2+ and Fe 2+ may indicate the extra requirement of these nutrients to compensate for their low levels in the ash content of corn steep liquor and suggest the enzyme as a metallo-protease (Li et al. 2016). Concentrations of the divalent cations also regulate the ionic strength of the medium which also significantly influenced the bioconversion process and enhanced protease activity. This complex medium has been reported as a reliable nitrogen source for microbial fermentations since the discovery of penicillin (Sherpa et al. 2021). Temperature is an important factor in the growth and metabolism of Stenotrophomonas acidaminiaphila as well as other related species. The study bacterium has a fairly wide mesophilic temperature of growth; 28-5-42.5 ºC (Ghosh and Saha 2013;Edet et al. 2018;Sherpa et al. 2021), nevertheless, some sort of synergy between growth and protease biosynthesis had to be achieved.
The path of steepest ascent was most successful in guiding the variable conditions towards optimum levels.
When the enzyme activity at the steepest ascent was compared with that obtained from RSM model and optimization, very little difference was found suggesting a remarkably successful path of steepest ascent experimentation. Frequently in literature, RSM is reported without regard to the path of steepest ascent experiment.
This generates misleading optimum bioprocess conditions that may result in either underestimated and/or overestimated yield of target metabolite, with the consequent difficulty in scale-up and/or pilot studies. Response surface methodology is a sequential statistical procedure that typically involves four stages which must be followed painstakingly. It is impossible therefore to move from a factor selection process straight into RSM, the results of which would only come by guess-setting the levels. The method of steepest ascent has been successfully exploited by various researchers towards locating optimum conditions of significant bioprocess variables (Yingling and Zhengfang 2013;Long et al. 2018). In recent times, artificial neural network (ANN) has emerged as an alternative method to RSM for predicting optimum levels of bioprocess factors and has frequently been demonstrated as being superior. However, if the fitted model performance metric; adjusted r 2 from an RSM experiment; is able to explain up to 98% of the variations about the data, then the alternative ANN would be of little or no relevance (Karri and Sahu 2018;Lau et al. 2020). In this report, the second-order model equation for protease activity yielded an adjusted r 2 of 99.99% and could therefore satisfactorily serve as fitness function for the optimization experiment. The optimum conditions of the significant variables were shown to lead to high protease activity by the study bacterium which was confirmed in triplicate shake flask experiments. Successful application of RSM for modeling and optimization of process parameters for metabolite production and environmental remediation has been copiously documented (Olubunmi et al. 2020;Karray et al. 2021).
Production kinetic studies typically attempt to uncover some key parameters of presumed optimized experiments. In a study where waste bioconversion and enzyme activity were the desired goals, maximum volumetric rate of substrate consumption and specific product formation are viewed as key kinetic parameters to be maximized. A comparison of shake flask kinetics and that in 5-L bioreactor as scale-up attempt for the bioprocess, Successful exploitation of bacteria in waste treatment with simultaneous production of value-added metabolite has been recommended for reduction in microbial bioprocess economics (Ekpenyong et al. 2021c). Kumar et al. (2015) reported the simultaneous production of lipid during dairy wastewater treatment by Rhodococcus opacus in a batch bioreactor with applications in biodiesel production and Olubunmi et al. (2020) produced biodiesel during restaurant waste oil treatment. Our results suggest that the study bacterium could be exploited for concomitant alkaline protease production and treatment of cassava processing waste.
Attempts at modeling the kinetics of protease production and substrate consumption using the modified Gompertz and logistic models (Mercier et al. 1992) revealed that the logistic model ( is also recommended in this report. The study enzyme was purified by the aqueous two-phase system using polyethylene glycol with molecular weight of 1500 in association with sodium citrate. When results of specific enzyme activity, yield and fold purification were compared to those obtained from a combination of precipitation and column chromatography (data not shown), the later results were set aside and the former upheld for this study. This biphasic system purification had earlier been applied to the extraction of alkaline proteases from cell-free Bacillus subtilis TISTR 25 fermentation broth (Chouyyok et al. 2005) and Bacillus amyloliquefaciens B7 (Abd Samad et al. 2017). With the 87fold purification of the protein, more reliable results of characterization and kinetic studies on the protein could be obtained.
To demonstrate potential for application of the alkaline protease in the detergent industry, results of preexposure of the purified enzyme to different detergents revealed stability of the enzyme to sodium dodecyl sulfate (SDS) with increased activity occurring with increased concentration of SDS. These results are in concord with a number of findings for other bacteria (Ozcelik et al. 2014), the only difference being in the extent of stability which Stenotrophomonas acidaminiphila strain UCCM 00065 possesses great advantage. This suggests that the protease extracted from cassava processing waste bioconversion could be further employed in a gainful venture like formulation of detergents for household and industrial dishwashing purposes.

Conclusions
Stenotrophomonas acidaminiphila UCCM 00065 produced alkaline protease during bioconversion of pregelatinized cassava processing effluent. A regular two-level factorial designed (2-LFD) experiment selected additional eight factors which linear and interaction terms significantly influenced substrate consumption and subsequent protease activity. A central composite rotatable designed experiment of a surface methodology using factor levels that produced maximum enzyme activity in the path of steepest ascent improved total enzyme activity by 35.82-fold in shake flasks and 163.41-fold under bioreactor conditions. Cassava waste bioconversion in the bioreactor therefore improved significantly by a factor of 8.89 after shake flask optimization with reduced fermentation time. We conclude that statistical optimizations are useful to improve microbial productions through identification and setting of levels of significant factors for efficient bioconversion of substrates and that production in bioreactors are dependable indices for scale-up of the bioprocess.