Improvement of Ficin-based Inhibitive Enzyme Assay for Mercury Using Response Surface Methodology and its Application for Near Real-Time Monitoring of Mercury in Marine Waters

Heavy metals pollution in the Straits of Malacca warrants the development of rapid, simple and sensitive assays. Enzyme-based assays are excellent preliminary screening tool with near real-time potential. The heavy-metal assay based on the protease ficin was optimized for mercury detection using Response Surface Methodology. The inhibitive assay is based on ficin action on the substrate casein and residual casein is determined using the Coomassie dye-binding assay. Heavy metals strongly inhibit the hydrolysis. A Central Composite Design (CCD) was utilized to optimize detection. The results show a marked improvement for the concentration causing 50% inhibition (IC50) for mercury, silver and copper. Compared to One-factor-at-a-time (OFAT) optimization, RSM gave an improvement of IC50 from 0.060 (95% CI, 0.0300.080) to 0.017 (95% CI, 0.0160.019), from 0.098 (95% CI, 0.0770.127) to 0.028 (95% CI, 0.0220.037) and from 0.040 (95% CI, 0.035.045) to 0.023 (95% CI, 0.0200.027), for mercury, silver and copper, respectively. A near real-time monitoring of mercury concentration in the Straits of Malacca at one location in Port Klang was carried out over a 4-h interval for a total of 24 h and validated by instrumental analysis with the result revealing an absence of mercury pollution in the sampling site.


Introduction
The Straits of Malacca is one of the busiest waterways in the world. It has become a collection point for pollutions, including heavy metals from shipping and terrestrial activities. It is also one of the largest sources of fishery and aquaculture products. The Maximum Permissible Limit for mercury stipulated by the Malaysian Department of Environment under the marine water quality standards [1] is 0.040 mg/L for fisheries (Class 2). The current contamination of heavy metals including mercury in the Straits of Malacca waters is under control. However, one study has found levels of the heavy metals Cr, Zn, Cu, As and Hg in the blood cockle (Tegillarca granosa) or also known as Anadara granosa exceeding the Malaysian standards [2]. In the West Coast of Peninsular Malaysia, Port Klang is one of the busiest ports with its surrounding areas heavily involved with aquaculture and fishery activities including the rearing of Anadara granosa [3]. Being able to monitor bioavailable mercury in near real-time (less than 1 h detecting time) allows the temporal detection of heavy metals that often elude monitoring authorities [3][4][5]. Previously, we have developed several heavy metal inhibition assays based on proteases including papain [6], bromelain [7], trypsin [8] which show promising potential in near real-time monitoring of heavy metals [9][10][11][12][13]. Previously, we had developed an inhibitive assay using ficin [14], which has the most sensitive response to mercury with an IC50 value of 0.085 mg/L. In order to detect mercury at the limit of the MPL (0.040 mg/L) the assay needs to be further optimized.
Customarily, analytical chemistry optimization has been done by changing important parameters of an experimental reaction by one component at a time or officially called one-factor-at-a-time (OFAT). The most significant drawback is that the interactive portion of the factors studied is not considered. As a result, the full effects of the parameter on the answer are not shown in this method. The increase in the number of experiments required to conduct the research is another drawback of the one-factor optimization which causes time and costs to be increased. In addition, OFAT uses more reagents and consumables. Response surface methodology (RSM) is a multivariate statistical methodology capable of resolving OFAT constraints. It is a series of mathematical and statistical methods that can explain the behavior of the dataset with the goal of making predictive forecasts, based upon the fits of a polynomial equation to experimental results. RSM is particularly useful when many factors affect a set of responses of interest. The goal is to maximize the rates of these variables simultaneously to produce the most optimum results. In analytical chemistry, RSM has often been used to optimize the detection of an analyte in a number of cases [15][16][17]. The use of RSM in improving the sensitivity of heavy metals detection in a protease-based inhibitive assay has never been attempted and this work, to the best of our knowledge, is the first of such an attempt.

Preparation of casein and ficin solution
Casein (Sigma) was weighed (2 g) and mixed with 100 ml of deionised water. The pH of the mixture was adjusted to 8.0 using 5N NaOH and/or 5N HCl. The mixture was stirred at 60 C overnight to maximize dissolution. Several layers of cheesecloth were utilized to remove insoluble casein from the mixture. The slightly clear filtrate was further clarified by centrifugation at 10,000×g (4 o C). The protein content of the clear supernatant was measured using the Bradford assay with crystalline BSA (Sigma) as the standard. The solution is stored at 4 o C until further use or stored frozen at -20 C. Ficin (SIGMA, E.C. 3.4.22.3, lot no: F4165-1ku, crude dried fig tree latex. 0.5 Units/mg) was prepared at 4 C in 20 mM sodium phosphate pH 6.5 as a 10.0 mg ml −1 stock solution. Working solutions of ficin (2.0 mg ml −1 ) and casein (10 mg ml −1 ) were prepared from these stock solutions fresh daily.

Ficin optimization studies
Ficin activity and optimization studies based on OFAT were carried out according to previous work [14]. To match ambient temperature for field trial environment in Malaysia and to qualify for near-real-time measurement, the temperature was fixed at 30 C and the incubation duration was fixed for 30 min [10,11,18].
The optimum concentration of the enzyme was studied by varying the final concentration of ficin from the stock solution to the final concentrations ranging from 0.1 to 0.8 mg/L in 20 mM phosphate buffer pH 6.5. A 30 μL of casein was added to 50 mL of the ficin solution and the solution was mixed thoroughly. The final concentration of casein was 2 mg/mL. The volume was topped up to150 μL using 20 mM phosphate buffer pH 6.5 and the mixture was incubated for 30 min at 30 o C. After the incubation period has elapsed, a 10 μL aliquot was immediately withdrawn and mixed with 200 μL of Bradford dye-binding reagent. After 5 min of incubation at room temperature, the absorbance at 595 nm for time zero was taken. After 30 min, another 10 μL aliquot was again taken and the absorbance at 595 nm taken (5 min incubation at room temperature) after mixing with 200 μL of Bradford dye-binding reagent. A microplate reader (Bio-Rad Model 680 microplate reader, Bio-Rad Laboratories, Inc., 3110 Regatta Blvd, Richmond, CA 94804, United States) was utilized for absorbance measurement. For optimizing the concentration of the substrate casein, ficin was fixed at 0.5 mg/mL while casein concentrations were varied from 0.5 to 3 mg/mL in a final volume of 150 μL. To study the optimum pH for enzyme activity, ficin was set at 0.5 mg/mL and casein was set up at 2 mg/mL. A sodium phosphate buffer (20 mM) from pH 5.8 to 7.8 (±1 pKa of phosphate) was utilized and the assay was carried out in the same manner as before with the only difference is the pH of the assay [19].

Central composite design experiments:
The Central composite design (CCD) was applied for the optimization of three experimental factors namely enzyme-substrate incubation time, casein and ficin concentrations. A 2 3 factorial central composite experimental designs leading to a set of 20 experimental runs was used to optimize the detection of mercury at 0.040 mg/L. The response is the difference in the absorbance value difference of the Bradford dye-binding assay measured at 595 nm (after 20 min of incubation at 30 o C) with the greatest difference in absorbance as the most desired response.

Ficin mercury inhibition studies
The experiment was initiated by mixing 50 μL of ficin in 20 mM phosphate buffer pH 6.5 from the experimental runs stipulated by CCD with 50 μL of mercury (final concentration of 0.040 mg/L). The mixture was incubated for 10 min at 30 o C. In the control, mercury was replaced with 20 mM phosphate buffer pH 6.5. Then, 50 μL of casein was added and mixed thoroughly (stock solution of 2.0 mg/mL). Immediately, a 20 μL aliquot was mixed with the Bradford dye-binding reagent (200 μL). The mixture was incubated at room temperature for 5 min and the absorbance was taken at 595 nm as time zero absorbance. After a 30-min incubation, another 20 μL aliquot was again taken and mixed with the Bradford dye reagent and the absorbance at 595 nm taken after a 5-min incubation period as before.

Field trials
Marine water samples were sampled periodically into acid-washed HDPE bottles containing several drops of 1% (v/v) HNO3 every 4 h from a location at Port Klang Selangor with the GPS location 3°00'00.6"N 101°23'22.9"E ( Figure 1). Samples were first filtered using a 0.45 µ m syringe filter. Fifty microliter of the clear filtrate was immediately utilized for assaying mercury using the ficin assay at 30 o C using a portable egg temperature incubator (30 Watt) (Generic brand name) powered by DC12V to AC220V Car Inverter (ZTE Avid Plus, China) capable of maintaining a temperature of 30 ± 1 o C was utilized to incubate the reaction mixture at 30 o C. The absorbance was read using a portable mini-spectrophotometer (Model M6+, Axiom, Germany). A Perkin Elmer Flow Injection Mercury System (FIMS 400) was utilized to determine the concentration of mercury whilst silver and copper were quantified using Atomic Emission Spectrometry on a Perkin Elmer ICP OES (Optima 8300, PerkinElmer, Inc. 940 Winter Street. Waltham, MA 02451 USA).

Data and Statistical Analysis
The per cent inhibition was calculated according to the following formula: % Inhibition = Test activity of sample -test activity of control x 100 Test activity of control Nonlinear regression using the one-phase exponential decay model was carried out using the software GraphPad Prism (Trial version 8.0.2). Means and standard deviations were determined according to at least three independent experimental replicates. RSM was carried out using the trial version of the Design-Expert version 7.0 software (Stat-ease Inc., USA).

Optimization using OFAT
The ficin assay mercury is an inhibitive assay. Under control (no mercury) situation, the ficin will degrade the substrate casein leaving oligopeptides (<2 kDa) that the Bradford assay cannot react, and the solution remains brown. In the presence of mercury, ficin is inhibited and the unreacted casein will be detected by the Bradford assay leading to an intense blue solution. The optimization studies via OFAT showed that ficin activity was optimum at 0.4 mg ml -1 ficin (Figure 2), 2.0 mg ml -1 of casein ( Figure 3) and at pH 6.5 (Figure 4). In the papain inhibitive assay for mercury, OFAT optimization gave the best combination of enzyme and casein both at 0.1 mg/mL [6]. In the bromelain assay, 0.11 mg/mL bromelain and 0.25 mg/mL casein are the best combinations [7].
Generally, protease assays including ficin can utilize a variety of substrates including natural such as casein and azocasein and artificial such as Nα-benzoyl-L-arginine-p-nitroanilide (BAPNA).
Similarly, other protein assays such as Biuret-Lowry, Folin Ciocalteu and Bichinconic acid can Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 31 August 2020 doi:10.20944/preprints202008.0689.v1 replace the Bradford assay. Despite this, the Bradford dye-binding assay continues to be a popular assay for a number of reasons including rapidity, sensitivity simplicity, and most importantly it is the most robust and can be used in the presence of many interfering compounds [20]. The mechanism behind this inhibition studies of this ficin enzyme is very much related to its active site.
Since this enzyme is a cysteine protease, it has SH group in its active site.

Optimization using Response Surface Methodology (RSM)
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 31 August 2020 doi:10.20944/preprints202008.0689.v1 The Central Composite Design (CCD) was employed to study the optimum concentration of the factors ( Table 1). Twenty experiments were designed by the Design expert software 6.0 with six replicates of midpoints, which are useful to determine the experimental error ( Table 2). was observed that the model had a low probability value (<0.0001) and a lack of the fit test F value of 1.435 (non-significant) implying that the model was fit. Insignificant lack of fit is important since it means that the observational errors are not significant, directed and systematic [21][22][23]. The "Pred R-Squared" of 0.9929 is within reasonable agreement with the "Adj R-Squared" of 0.9977. The reliability of the experiments can be further established with a low coefficient of variance (CV) value [24]. The CV of 4.3% for this study is an indication of the reliability of experiments performed.
The significance of regression for the coefficients was considered. In this case, A, B, C , A 2 , B 2 , C 2 , AB, AC & BC were the significant model terms (Table 3). Hence, statistical analysis of the experimental data revealed that all three factors had a significant effect on throughout the study.
The fitness of data into the selected model was examined through diagnostic model plots (Figures   5a-d). The plots are important particularly in the assessment of data error that differs from model predictions, which aids in assessing and improving model adequacy [24]. The plot of actual versus predicted values obtained from the experiment (Figure 5a) revealed a close relationship between the actual and predicted value as the data points assembled close to the line that divides the plot into equal halve (45 o ). The adequacy of the model was further verified by plotting the predicted values and studentized residuals (Figure 5b). Studentized residues are a variation between the predicted value and actual responses obtained from the model. The plot of normal probability demonstrates slight or no abnormality in the experimental data (Figure 5c). An outlier plot (Figure 5d) visualizes the distantly standout standard deviation of actual response from the rest of the data. No outlier was evident from the plot as all the data falls between 3.5 and -3.5. Visualization of the all the factors required for maximum growth is presented through 3-dimensional responses and contour plots (Figures 6 to 8). The plots are of utmost importance in determining the relationship at zero or intermediate levels of different combinations of independent factors before performing a real experiment [25,26]. The 3D response plots show the maximum response between each pair of a factor while the other factor is held constant. The curved contour lines indicate interaction with elliptical or saddle contour plots indicate significant interaction whilst circular contour plots indicate that interaction is not significant [27]. The perturbation plots, 2D plot for the combination of factors show the interaction between the factors (results not shown) especially between ficin and pH and casein and ficin concentrations. Normally, the interaction between two different factors occurs when there is a different response obtained when varying the outcome of one factor at different levels compared to the other factor [28]. As the perturbation lines amongst the ficin concentration and pH, and ficin concentration and casein concentration did not cross each other, this indicates that the interaction is synergistic instead of antagonistic [27].  +0.021667*pH*Ficin0.00032*pH*Casein+0.012222*Ficin*Casein

Comparison of OFAT and RSM in mercury detection using ficin
In order to compare the efficacy of RSM in optimizing detection of mercury, the ficin Coomassie dye-binding assay for mercury is compared to OFAT by constructing the inhibition curve.  IC50 value refers to the concentration of heavy metals that inhibits the enzyme activity by 50% while LOD is the limits of detection, which is three times the standard deviation of the blank. The non-overlapping confidence interval of parameter estimates in nonlinear regression generally indicates a significant difference at the alpha value utilized (0.05) whilst an overlapping CI does not, in general, indicate significance or non-significance with more data is needed to reach a conclusion. In this work we only consider non-overlapping CI as a benchmark for indicating process optimization (RSM) success in improving the sensitivity of heavy metals' detection using the ficin assay and is this case, RSM managed to improve the LOD and IC50 values for all heavy metals tested. Notably, the use of RSM increased the D595nm values by 0.1 absorbance value on average, which is a marked improvement. The use of RSM in analytical works to improve the sensitivity of detection has been documented in many cases [15][16][17][29][30][31], and the improving of  inhibition (<10%) to the ficin assays and instrumental analysis shows mercury level lower than the designated MPL in marine waters (0.040 mg/L). Other near real-time works using enzymes in rivers shows temporal levels of heavy metals [9,11,13,14,35] and this is the first study using marine water as samples. Marine waters are large bodies of water where heavy metals originating from terrestrial areas are rapidly diluted. Elevated levels of heavy metals have been found in this region but most are found in the sedimental fractions [36]. This could explain the lack of inhibition to the ficin and a lower concentration of mercury in marine waters within this area. Variation in heavy metals levels in running water especially in rivers and marine water bodies is common [9][10][11][12][13]37]. Even sedimentary samples have been found to be variable in their spatial and temporal concentrations of heavy metals [38]. This variability requires a fast detection system to capture the temporal variation in heavy metals' concentrations as this is important in environmental forensic applications. The current detection system can best be described as a batch system with samples needed to be transported to the laboratory prior to the determination of their heavy metals content [6,39,40]. One of the solutions to this problem is real-time or near real-time monitoring of heavy metals. The use of bioassays involving plants, microorganisms and enzyme assays can address this issue [41][42][43]. In enzyme assays, the sampling to detection period can be carried out in less than one hour using a portable spectrophotometer and thus is a perfect candidate for near real-time analysis. We have demonstrated the application of enzyme-based system in capturing in near real-time the temporal variation of heavy metals concentrations in rivers running through heavily industrialized areas [9][10][11][12][13]. The application of the ficin assay in monitoring mercury in marine water bodies in this study is a novel exercise meant as a proof of concept. More sampling locations need to be identified and more field trials will be carried out in the future. Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 31 August 2020 doi:10.20944/preprints202008.0689.v1

Conclusion
The use of RSM based on CCD was successful in optimizing the protease ficin dye-binding assay for mercury, silver and copper resulting in more sensitive determination of these metal ions. The resultant LOD and IC50 values were better than the values obtained with the OFAT approach. The sensitivity of mercury, in particular, was good enough to detect mercury at the Maximum Permissible Limit allowed for marine waters. The developed assay was then tested as a near real-time assay for the detection of mercury from a marine site in Port Klang. Mercury was not detected in this site at all of the sampling periods indicating the absence of pollution due to mercury in this water. More samples from diverse sampling points are currently being tested to monitor the presence of heavy metals especially mercury in coastal areas and rivers in Malaysia.
The assay is rapid, sensitive, easy to be carried out and has the ability to monitor in near real-time as a preliminary screening tool for heavy metals pollution.

Author Contributions:
Conceptualization