LINEARIZATION OF THE BRADFORD PROTEIN ASSAY TO APPLICATION IN COW MILK PROTEINS QUANTIFICATION BY UV-Vis SPECTROPHOTOMETRY METHOD Linearização da metodologia de Bradford para aplicação na quantificação das proteínas totais do leite bovino por espectrofotometria no UV-Vis

Reliable methods for determination and quantification of total protein in food are essential information to ensure quality and safety of food trade. The objective of this study was to evaluate the linearity of calibration curves obtained from different proteins (blood serum albumin-BSA, α-LA, β-LG, caseins (CN): αs, β and κ-CAS) with the reagent of Bradford. Comercial UHT skimmed bovine milk was analyzed for the determination of total protein using the Bradford method by reading at 595 nm. The determination of the concentrations of total milk protein was achieved by linear regression. The Bradford method showed a high sensitivity for the determination of total proteins in bovine milk dilution 1:25 to values closer to those obtained by the Kjeldahl method. The results showed that the calibration curve of standard proteins β-CN and BSA obtained better linearity with less variation in the absorbance measurements for the determination of total protein of milk.

The development of methods for determination and quantification of total protein in food is an important consi deration in dairy research (DZIUBA et al, 2009). Kjeldahl is the official method for quantification of the total milk protein (HELRICH, 1990), however, this technique has been questioned because it is based on spectroscopic techniques to the quantification of samples total nitrogen which includes protein and non-protein nitrogen, besides being time-consuming and using corrosive reagents (DE KRUIF;HOLT, 2003). There are different ways to measure the total milk protein and methods based on spectrophotometry are commonly employed (LOWRY et al., 1951;GROVES et al., 1968). The ultraviolet spectrophotometry method has the advantage of not destroying the samples and to be very fast (PACE et al., 1995). One major limitation is that the results with complex samples may be imprecise when they are constituted by a mixture of substances (STOSCHECK, 1990).
The Coomassie Brilliant Blue (CBB), known as the Bradford method, is commonly used for protein determination in food (BRADFORD, 1976). This method is based on the electrostatic interaction between the dye, basic amino acids (arginine, lysine, histidine amino acid residues, and NH 2terminal amino group) and aromatic amino acids residues. Van der Waals forces and hydrophobic interactions are also supposed to be involved in the color formation. This technique is very reproducible, relatively fast, sensitive and specific (ZOR; SELINGER, 1996). When the protein binds to the red form of the reagent, it is converted to the blue form which is stable and non-protonated. Thus the method can provide a slight non-linearity in the Beer-Lambert law due to variation of the pH of the solution and the pattern of response due to overlapping of the spectra of two different types of reagent color (OWUSU-APENTEN, 2002;COMPTON;JONES, 1985;SPLITTGERBER;SOHL, 1989;TAL et al., 1985).
There are many studies for determination of protein comparing sensitivity, cost, robustness, and linearity of the different interference on the spectroscopic analysis. However it was not found studies comparing the linearity of the Bradford method standard curves with different proteins. Considering that milk is a complex solution and its composition has many different types of proteins with different conformations and structures, the aim of this study was to evaluate the linearity model using calibration curves obtained from different individuals proteins (α-, β-and κ-casein; α-LA, β-LG and BSA) to determine the concentration of total milk protein. The hypothesis was that these curves would present better reliability for quantification of total protein from bovine milk using the Bradford method with readings in the UV-Vis region of the spectra.

Preparation of curves of individual proteins
The Bradford reagent solution (Bio-Rad Protein Assay, Hercules, California) was prepared according to manufacturer's instruction. The standard proteins: BSA, α-LA, β-LG, α-casein, β-casein and κ-casein (Sigma-Aldrich, St. Louis, MO, USA) were prepared in the following concentrations: 0.2, 0.4, 0.6, 0.8, 1.0 and 1.5 mg / mL using ultrapure water (Ultrapure Milli-Q water, Millipore Corp., Bedford, MA, USA). For standard procedure, 4 μL of each of them were added to 200 μL of Bradford reagent. After incubated for 5 minutes at room temperature, absorbance measurements were performed using "Protein Bradford" module at 595 nm in a spectrophotometer (ND-1000 NanoDrop Technologies, Wilmington, USA). The Bradford reagent was used as a blank. Standard curves were prepared for each standard protein and used to quantification of protein in milk samples.

Spectrophotometry determination of the total milk protein
Sample of commercial skimmed bovine milk was diluted (1:25, 1:50, 1:100) in ultrapure water (Ultrapure Milli-Q water, Millipore Corp., Bedford, MA, USA) and 4 μL of each dilution were added to 200 μL of Bradford reagent and subsequent vortex mixing. After 5 minutes at room temperature, absorbances were measured at 595 nm in a spectrophotometer ND-1000 NanoDrop as above.

Determination of Kjeldahl
The total protein content in milk was determined by the Kjeldahl method (HELRICH, 1990).

Statistical Analysis
The method of least squares (OLS) was used to calculate the linear regression curve for each protein concentration correlating to absorbance signal. The statistical significance of linear regressions obtained was evaluated by conducting an analysis of variance (ANOVA) for each regression (Microsoft Excel) and F test was applied. The samples were analyzed in triplicate.

RESULTS AND DISCUSSION
Figures 1A to 1F present the absorbance as a function of each protein concentra tions and these graphs were used as standard curves. The determination of individual protein was done at 595 nm. The absorbance obtained for the proteins concentrations was correlated with the results of protein patterns as shown in standard curves of Figure 1 (A to F).
Results of variance analysis and the regression value were obtained for each parameter ( Table 1). The significance of parameter F, given by the quotient of the mean square of the model and by the mean square residual indicates the existence of a significant linear relationship between the two variables. Some authors recommend that the significance of F value should be at least > 5 for the regression to have practical utility (AGTERDENBOS et al., 1981). The higher the F value the greater is the significance of the linear relationship between the two variables. Thus, the analysis of Table 1 provides the assertion that all the linear regression obtained in this study was significant, due to the high F values. The tabulated F (F (0,05, 4, 24) ) is 2.78, so the analysis of values of lack of fit of F contained in Table 1 leads to the conclusion that all of the standard curves, except for the α-LA and β-LG, were suitable for analysis because there was no evidence of lack of fit at 95% confidence. Table 2 shows the coefficients of the linear regression equations, developed from the six concentrations tested with five replicates each. Each protein was analysed by standard statistical parameters of the equation and the corresponding R 2 were obtained. Table 3 shows the quantitative results using the Bradford method and the total protein concentration estimated from bovine milk in three dilutions (1:25, 1:50, 1:100) through each calibration standard curve prepared from the following proteins: BSA, α-LA, β-LG, α s -casein, β-casein and κ-casein. The quantification of total milk protein using the standard curve of α-LA in the protein dilutions 1:100, 1:50 and 1:25 underestimate the concentration found by the Kjeldahl method. When compared with the measurements obtained from the curves in dilutions 1:100, 1:50 and 1:25 of β-LG. α s -casein and κ-casein concentrations were higher than those found by the Kjeldahl method. Quantification of total protein in cow's milk 1:25 dilution obtained with the standard curves of the β-casein and BSA showed a result closer to that obtained from the Kjeldahl method.
Even though the Bradford method can provide different estimates of the concentration of total protein in milk for each standard curve, a better linearity of the curves was found. Kamizake et al. (2003) revealed that BSA and casein can be used as standard proteins and that the Bradford method for protein determination by spectrophotometry on total milk showed a higher sensitivity as compared with others standard curves. A possible explanation for these observations is the fact that the milk protein fractions have a differentiated structure presenting each physical and chemical property due to different amino acid composition and the same provision in the formation of the protein (DE KRUIF;HOLT, 2003). Although the Bradford method is rapid, sensitive and subject to a much smaller number of interference than other methods (BRADFORD, 1976), it presents the same drawbacks as the change in absorptivity to different specific proteins due to difference in hydrophobicity or molecular weight, and susceptibility to suffer interference in the presence of lipids and sodium and potassium chlorides (SHIBA et al., 1985;GOREN;LI, 1986;WIMSATT;LOTT, 1987). To circumvent the problems in specific absorptivity of the different proteins, some authors  BERLET, 1989) or heating the sample with urea or 2-mercaptoethanol (GOTHAN et al., 1988). However, the change of these variables depends on the molecular weight of each protein.
Considering the hydrophobicity of the proteins tested, BSA is a protein that has the lowest value followed by the α s -casein = α-LA < β-LG = κ-casein < β-casein. With regard to molecular weight and number of amino acid residues, the α-LA is the smallest protein with less residues available for interaction followed by β-LG < κ-casein < α s -casein < β-casein < BSA (WALSTRA; JENNESS, 1984). According to the hydrophobicity there was not a correlation since the best results for the quantification of total protein were obtained by BSA and β-casein. However there may be a correlation with the size of the protein and the amount of amino acid residues, since the less satisfactory results were obtained for measurement of the proteins of lower molecular weights (α-LA. β-LG and κ-casein) whereas, better results were obtained with higher molecular weight proteins (BSA and β-casein). These results may be explained due to the number of amino acid residues that interact with the colorant CBB. The higher the molecular weight observed in proteins tested the greater the amount of amino acid residues available for interaction with the Bradford reagent.  All curves consisted of six points with five replicates each. SE A = Standard error of the coefficient of the straight and SE b = standard error of the coefficient b of the line.