Modified agar diffusion bioassay for better quantification of Nisaplin®

Authors


Correspondence

Minal Lalpuria, FDA/CFSAN/OFS/FPET/LACF, 5100 Paint Branch Parkway, College Park, MD 20740, USA. E-mail: minallalpuria@gmail.com

Abstract

Aims

To investigate the effect of different well sizes and pre-diffusion times at 4°C, on the sensitivity, accuracy and precision of nisin quantification by agar diffusion bioassay.

Methods and results

Nisin solution (0·625–125 μg ml−1) was filled in wells (3·5 mm or 7 mm diameter) made on agar plates inoculated with Micrococcus luteus, followed by pre-diffusion (0, 24, 48 or 72 h), incubation and measurement of inhibition zone. Regression analysis indicated that wells with 3·5 mm diameter had smaller standard deviation and higher predictive accuracy, compared to wells with 7 mm diameter. Based on Tukey's test, pre-diffusion resulted in significantly different inhibition zones at different nisin concentrations. Pre-diffusion also improved sensitivity of the assay. Different regression models were considered to explore the relationship between inhibition zone and nisin concentration for different pre-diffusion times. A spline model was determined to be the best-fit model, and 48 h was the best pre-diffusion time.

Conclusions

Wells with 3·5 mm diameter demonstrated higher accuracy for nisin quantification compared to wells with 7 mm diameter. 48 h was the best pre-diffusion time for nisin concentration in the range 0·625–125 μg ml−1.

Significance and impact of the study

The findings from this study will be helpful in quantifying nisin and compounds with antimicrobial properties accurately over a wide range of concentrations using agar diffusion bioassay.

Introduction

There are increasing food safety concerns among the governmental agencies, food industry and consumers alike. As consumer demands for ‘natural’, ‘safe’ and ‘minimally processed’ foods are increasing, antimicrobial peptides from bacteria that target foodborne pathogens without adverse effects to human health have received considerable attention (Cleveland et al. 2001). Nisin is one such bacteriocin produced by Lactococcus lactis subsp lactis fermentation (except Nisin U, which is produced by L. uberis) and is approved for use in over 50 countries. Nisin is a polypeptide composed of 34 amino acids and is classified as ‘generally regarded as safe (GRAS)’ by the Food and Drug Administration (FDA) and the World Health Organization (WHO). It is used in a variety of dairy, plant-based and meat-based food products. Nisin shows antimicrobial activity against a wide range of Gram-positive bacteria such as Bacillus, Clostridium, Listeria and Staphylococcus species, but has little or no activity against Gram-negative bacteria or fungi (Hurst and Hoover 1993). Nisin attacks the target micro-organism by interacting with lipid molecules in the microbial cell and forming pores in the bacterial cytoplasmic membrane. This permeabilization of the cytoplasmic membrane results in rapid leakage of essential cellular constituents, causing cell death (Wiedemann et al. 2001).

Nisin is considered ‘natural’ only when used in concentrations found in foods naturally fermented with a nisin-producing starter culture (Cleveland et al. 2001). Various countries have set limits for maximum levels of nisin in foods. Therefore, it is necessary to have a precise and accurate method for the quantification of nisin in foods. Several techniques have been developed for the detection and quantification of nisin, such as methylene blue and dilution assay (Hirsch 1950), turbidity assays (Berridge and Barrett 1952), agar diffusion bioassay (Mocquot and Lefebvre 1956; Tramer and Fowler 1964), high-performance liquid chromatography (Liu and Hansen 1990; Rollema et al. 1995), photometry (Parente et al. 1995), capillary zone electrophoresis (Rossano et al. 1998), colorimetric methods such as bicinchoninic acid protein assay (Ripoche et al. 2006) and MTT [3-(4,5-dimethyl thiazol-2-yl)-2,5-diphenyltetrazolium bromide] colorimetric assay (Fang et al. 2007), chemiluminescence (Bouksaim et al. 1998), bioluminescence (Immonen and Karp 2007) and ELISA (Bouksaim et al. 1999; Falahee et al. 1990; Leung et al. 2002).

Agar diffusion bioassay is the most widely used method for quantifying nisin activity (Pongtharangkul and Demirci 2004). In this method, nisin is allowed to diffuse through agar gel seeded with nisin-sensitive indicator bacteria. The diameter of the inhibition zone produced by growth inhibition of nisin-sensitive indicator bacteria in the agar plate is correlated with the concentration of nisin. Greater nisin concentrations result in larger inhibition zones.

Factors such as nisin structure, concentration of agar, pH, detergent, number of indicator cells and temperature can affect diffusion of nisin through agar. Wolf and Gibbons (1996) improved the accuracy and precision of conventional agar diffusion bioassay for nisin by reducing the concentration of agar from 1·5 to 0·75% and then buffering the agar with phosphate salts. They also showed that addition of a detergent like Tween 20 to the agar greatly increased bioassay sensitivity by improving diffusion of nisin through the agar. Rogers and Montville (1991) showed that pre-diffusion of plates at 3°C improved the sensitivity and reproducibility of the assay by resulting in larger inhibition zones and lower variability on different days. Pongtharangkul and Demirci (2004) also reported that 24-h pre-diffusion of the agar plates containing nisin wells at 4°C produced larger inhibition zones and improved the accuracy and precision of nisin quantification as compared to the conventional bioassay. They concluded that nisin concentration should not exceed 300 IU ml−1 (7·5 μg ml−1) for accurate prediction of nisin concentration by agar diffusion bioassay. It seems that although pre-diffusion results in a more sensitive and accurate quantification of nisin, there is a need to improve the pre-diffusion step and achieve better correlation between inhibition zone and nisin concentration, at higher nisin concentrations.

Different researchers have used different well size, small (Arauz et al. 2008; Blom et al. 1997; Bonev et al. 2008) and large (Pongtharangkul and Demirci 2004; Tramer and Fowler 1964; Wolf and Gibbons 1996), for performing the agar bioassay with nisin. A smaller well allows for less need of sample and more wells can be made on a single agar plate. However, there is no information in the literature about the effect of well size on nisin quantification as yet.

The objectives of this study were to (i) investigate the effect of agar well size and pre-diffusion time at 4°C, on the sensitivity, accuracy and precision of nisin quantification, (ii) explore the relationship between inhibition zone and nisin concentration over a wide range of nisin concentrations and (iii) select the best pre-diffusion time for a wide range of nisin concentrations.

Materials and methods

Micro-organism and media

As in many previous studies, the nisin-sensitive micro-organism used in this study was Micrococcus luteus (ATCC 10240). Stock culture was maintained at −80°C in 20% glycerol. M. luteus was grown in Difco nutrient broth (Becton Dickinson and Co., Sparks, MD, USA) at 30°C on an orbital shaker at 300 rev min−1 for 24 h and refrigerated at 4°C before use. The media for agar bioassay consisted of 0·8% nutrient broth, 0·75% Bacto agar (Becton Dickinson and Co.) and 1% Tween 20 (BDH®, Solon, OH, USA).

Nisin standard

Stock solution of Nisaplin® (Danisco USA Inc., New Century, KS, USA) with nisin concentration of 125 μg ml−1 was prepared in sterile deionized water. Concentrations ranging from 0·625 to 125 μg ml−1 were prepared by dilutions in deionized water and utilized to construct the calibration plot. Nisaplin® contains about 2·5% nisin A, and one gram of Nisaplin® has an activity of 106 international units.

Agar diffusion bioassay

The agar diffusion assay was performed according to the procedure described by Pongtharangkul and Demirci (2004). The agar medium was autoclaved, cooled to 40°C in a water bath and inoculated with 1% v/v of 24 h culture of M. luteus. To ensure that an equivalent number of M. luteus cells were inoculated into the agar medium each time, the optical density of the inoculum was maintained at 1·7, when measured at 600 nm. The final population of micro-organisms was approximately 106 CFU ml−1. Two different agar well sizes, 3·5 mm diameter with 2·6 mm depth and 7 mm diameter with 3·5 mm depth, were used. 15 and 20 ml of liquid agar seeded with M. luteus were poured into the Petri plates aseptically for small and large wells, respectively. This agar was allowed to solidify at 4°C for 30 min, and 3–4 wells were made on each plate using a sterilized steel cork borer. Twenty and 50 μl of nisin solution in the concentration range of 0·625–125 μg ml−1 were dispensed into the wells with 3·5 and 7 mm diameter, respectively. These agar plates containing nisin solution were stored at 4°C for 0, 24, 48 and 72 h (pre-diffusion) followed by incubation for 48 h in an environmental chamber maintained at 30°C and 75% r.h. The inhibition zone was measured from the edge of the well with a digital vernier caliper (VWR International Inc., Radnor, PA, USA). Four inhibition zone measurements were taken for each well and averaged. The assay was replicated three times.

Data analysis

The mean inhibition zone was plotted against the logarithm of nisin concentration. From this data, a regression equation for the assay was developed. The regression equation was of the type:

display math(1)

where Z = Inhibition zone (mm), CN = Concentration of nisin (μg ml−1), A = constant and B = constant.

The slopes of the regression equation indicated the sensitivity of the assay; the precision of the assay was inferred from the correlation coefficient (R2); and the accuracy of the assay was inferred from the standard error. The prediction accuracy of the model was inferred from the value of the predicted residual sum of squares (press) (Kutner et al. 2004a). Press values indicate how well a regression model predicts new observations. The smaller the press value, the better the model's predictive ability. Results were analysed using regression analysis and one-way anova (α = 0·05) in Minitab® 14 (Minitab Inc; State College, PA, USA). Tukey's multiple range test (α = 0·05) was used to determine any significant differences between mean inhibition zones.

To select the best pre-diffusion time, inverse prediction intervals of nisin were used (Kutner et al. 2004b). An inverse prediction interval is the 95% confidence interval for the predicted value of nisin concentration and is computed from the regression of inhibition zone vs nisin concentration. Smaller prediction intervals correspond to a more accurate prediction of an unknown nisin concentration from the observed inhibition zone. The first step in this analysis was to find the best-fit calibration model. In addition to the log-linear model given in Eqn 1, quadratic models and spline models with knot points (KN) at nisin concentrations 2·5, 6·25, 12·50, 18·75, 25 and 50 μg ml−1 were also fitted to the data to estimate the best-fit model for each pre-diffusion time.

The quadratic model used was of the following type:

display math(2)

while the spline model was as follows:

display math(3)

where,

display math

The goodness of fit was measured by the Akaike information criterion (AIC), R2 and lack-of-fit test. AIC is a goodness of fit measure that takes into account the value of the likelihood and the number of parameters in the model (Kutner et al. 2004a). The lower the value of AIC, the better the model. The lack-of-fit test measures how well the model represents the data and it should not be satisfactorily significant. This analysis was carried out in R (R Development Core Team 2011). After selecting the best model for each pre-diffusion time, the inverse prediction intervals of nisin as a function of the nisin concentration were computed (Kutner et al. 2004b).

Results

Effect of well size

The plot of inhibition zones vs nisin concentration for the wells with 3·5 and 7 mm diameter using Eqn 1 is shown in Fig. 1, and the results of regression analysis are presented in Table 1. The slopes for both well diameters were not significantly different (P > 0·05) from each other, and their R2 values were similar. Hence, their sensitivity and precision for nisin quantification are comparable. The standard error and press value were lower for wells with 3·5 mm diameter. Hence, the regression model for 3·5 mm diameter wells had smaller standard deviation of the residuals and can predict new observations more accurately than the model for 7 mm diameter wells.

Table 1. Regression analysis of inhibition zone vs nisin concentration for different well size using Eqn 1
Well diameter (mm)SlopeaR2Standard errorpress value
  1. Press, predicted residual sum of squares.

  2. a

    Values with a different superscript are significantly different (P < 0·05).

71·61a0·980·180·98
3·51·48a0·990·100·61
Figure 1.

Inhibition zone vs log nisin concentration for well diameter (♢) 7 mm and (□) 3·5 mm. Data (Error bars: ±1 SD) are an average of three replications.

For wells with 7 mm diameter, the inhibition zone at a nisin concentration of 1·25 μg ml−1 was not significantly different (P > 0·05) from that of 2·50 μg ml−1 (Table 2). Similarly, no significant differences existed between the inhibition zones at nisin concentrations of 6·25 vs 12·50, 12·50 vs 18·75, 18·75 vs 25, and 50 vs 125 μg ml−1. For wells with 7 mm diameter, the inhibition zones were significantly different from each other (P < 0·05) for nisin concentration in the range 0·625–12·50 μg ml−1. However, no significant differences existed between the inhibition zones at nisin concentrations of 12·50 vs 18·75, 18·75 vs 25, 50 vs 75 and 75 vs 125 μg ml−1. These Tukey's test results indicated that wells with 3·5 mm diameter distinguish better between inhibition zones for neighbouring nisin concentrations in the range 0·625–12·50 μg ml−1.

Table 2. Tukey's multiple test for inhibition zone with different well size
CN (μg ml−1)Z(mm)a for well diameter
7 mm3·5 mm
  1. a

    In a given column, values with different superscripts are significantly different (P < 0·05).

0·6252·17 ± 0·16a1·98 ± 0·06a
1·252·73 ± 0·2b2·41 ± 0·13b
2·503·01 ± 0·25b2·80 ± 0·08c
6·253·94 ± 0·16c3·65 ± 0·09d
12·504·33 ± 0·17 cd4·11 ± 0·04e
18·754·65 ± 0·11d4·25 ± 0·11ef
254·83 ± 0·11d4·51 ± 0·09f
505·36 ± 0·22e4·84 ± 0·16 g
755·52 ± 0·15e5·05 ± 0·09gh
1255·78 ± 0·18e5·28 ± 0·10 h

Based on the results from Tables 1 and 2, wells with 3·5 mm diameter had smaller standard deviation of residuals, higher prediction accuracy and better distinction between inhibition zones at neighbouring nisin concentrations. Hence, this well size was used in further studies comparing the effect of different pre-diffusion times on the sensitivity, accuracy and precision of the agar diffusion bioassay.

Effect of pre-diffusion time

The plot of inhibition zone vs nisin concentration for all pre-diffusion times is shown in Fig. 2, and their regression analysis results are presented in Table 3. The slopes of the regression lines with pre-diffusion 0, 24, 48 and 72 h were all significantly different (P < 0·05) from each other. This indicates that an increase in pre-diffusion time would increase assay sensitivity. R2 values for all pre-diffusion times were high, while the standard errors and PRESS values increased with increasing pre-diffusion time.

Table 3. Regression analysis of inhibition zone vs nisin concentration for different pre-diffusion times using Eqn 1
Pre-diffusion time (h)Slopea R 2 Standard errorpress value
  1. Press, predicted residual sum of squares.

  2. a

    Values with a different superscript are significantly different (P < 0·05).

01·48a0·990·100·61
242·69b0·980·354·00
483·64c0·980·405·07
724·50d0·970·5811·13
Figure 2.

Inhibition zone vs nisin concentration for pre-diffusion times (○) 0 h, (♢) 24 h, (□) 48 h and (△) 72 h. Data (Error bars: ±1 SD) are an average of three replications.

As seen earlier for the pre-diffusion time of 0 h, no significant differences existed between neighbouring nisin concentrations, for concentrations greater than 12·50 μg ml−1(Table 4). For 24-h pre-diffusion, inhibition zones were significantly different from each other (P < 0·05) for nisin concentration in the range 0·625–25 μg ml−1. However, there was no significant difference between inhibition zones at nisin concentrations of 25 vs 50, 50 vs 75 and 75 vs 125 μg ml−1. For 48-h pre-diffusion, inhibition zones were significantly different from each other (P < 0·05) for all nisin concentrations. For 72-h pre-diffusion, inhibition zones were significantly different from each other (P < 0·05) for all nisin concentration, except for 25 vs 50 μg ml−1. Thus, pre-diffusion at 4°C improved the distinction of inhibition zones at neighbouring nisin concentrations up to 48 h, with 48 h giving the best distinction.

Table 4. Tukey's multiple test for inhibition zones with different pre-diffusion times
CN (μg ml−1)Z(mm)a for pre-diffusion time
0 h24 h48 h72 h
  1. a

    In a given column, values with a different superscript are significantly different (P < 0·05).

0·6251·98 ± 0·06a2·98 ± 0·13a3·38 ± 0·18a3·38 ± 0·19a
1·252·41 ± 0·13b3·76 ± 0·34b4·68 ± 0·23b5·54 ± 0·19b
2·502·80 ± 0·08c4·53 ± 0·34c5·93 ± 0·11c7·10 ± 0·10c
6·253·65 ± 0·09d6·06 ± 0·14d8·02 ± 0·16d9·47 ± 0·18d
12·504·11 ± 0·04e6·86 ± 0·03e9·02 ± 0·07e10·67 ± 0·24e
18·754·25 ± 0·11ef7·43 ± 0·16f9·520·06f11·41 ± 0·09f
254·51 ± 0·09f7·95 ± 0·01 g9·97 ± 0·13 g12·09 ± 0·11 g
504·84 ± 0·16 g8·18 ± 0·11gh10·49 ± 0·04 h12·52 ± 0·14 g
755·05 ± 0·09gh8·6 ± 0·09hi11·05 ± 0·15i13·04 ± 0·05 h
1255·28 ± 0·10 h8·91 ± 0·06i11·80 ± 0·24j14·05 ± 0·18i

Based on the results from Tables 3 and 4, it was found that longer pre-diffusion times resulted in higher sensitivity and better distinction between different inhibition zones at neighbouring nisin concentrations, but lower accuracy of the assay. Hence, this model was not conclusive in selecting a best pre-diffusion time. Quadratic (Eqn 2) and spline (Eqn 3) models were fitted to the data and compared with the log-linear model. The AIC, R2 and lack-of-fit test values for these models at all pre-diffusion times are presented in Table 5. The R2 values for all models were greater than or equal to 0·97. For the 0 h pre-diffusion time, the models that passed the lack-of-fit test (P > 0·05) were the quadratic and spline models with knot points at nisin concentrations of 6·25, 12·50, 18·75, 25 and 50 μg ml−1. For the 24 h pre-diffusion time, only spline models with knot points at nisin concentrations of 18·75 and 25 μg ml−1 passed the lack-of-fit test (P > 0·05). Similarly, for the 48 h pre-diffusion time, only the spline model with knot point at nisin concentration of 12·50 μg ml−1 passed the lack-of-fit test (P > 0·05). None of the models passed the lack-of-fit test (P < 0·05) at 72 h pre-diffusion time. Hence, 72h pre-diffusion time was not considered for any further comparison. Based on the smallest AIC value for models that passed the lack-of-fit test (P > 0·05), spline models with knot points at nisin concentration 12·50, 25 and 12·50 μg ml−1 were the best for pre-diffusion times 0, 24 and 48 h, respectively. The plot of inhibition zone vs nisin concentration using a spline model is shown in Fig. 3.

Table 5. R2, Akaike information criterion (AIC) and lack-of-fit P-value of various regression models for all pre-diffusion times (computed using R)
Pre-diffusion time (h)Model R 2 AICLack-of-fit P-value
0Log-linear0·99−29·770·005
Quadratic0·99−40·840·101
Spline knot log (2·50)0·99−30·210·006
Spline knot log (6·25)0·99−41·010·105
Spline knot log (12·50)0·99−45·460·285
Spline knot log (18·75)0·99−43·990·209
Spline knot log (25)0·99−44·310·224
Spline knot log (50)0·99−38·960·063
24Log-linear0·9820·470·000
Quadratic0·994·040·009
Spline knot log (2·50)0·9819·140·000
Spline knot log (6·25)0·997·550·003
Spline knot log (12·50)0·99−0·400·030
Spline knot log (18·75)0·99−5·520·112
Spline knot log (25)0·99−9·640·283
Spline knot log (50)0·995·010·006
48Log-linear0·9833·610·000
Quadratic0·99−4·600·004
Spline knot log (2·50)0·9919·250·000
Spline knot log (6·25)0·99−11·870·030
Spline knot log (12·50)0·99−14·070·053
Spline knot log (18·75)0·99−4·370·003
Spline knot log (25)0·992·450·000
Spline knot log (50)0·9924·000·000
72Log-linear0·9756·560·000
Quadratic0·992·730·001
Spline knot log (2·50)0·9930·860·000
Spline knot log (6·25)0·995·560·001
Spline knot log (12·50)0·997·840·000
Spline knot log (18·75)0·9914·170·000
Spline knot log (25)0·9919·520·000
Spline knot log (50)0·9845·100·000
Figure 3.

Inhibition zone vs log nisin concentration using Eqn (3) for pre-diffusion times (a) 0 h, (b) 24 h and (c) 48 h generated using R. The dashed vertical line shows knot point. Dotted lines show 95% confidence interval.

The spline model equations for inhibition zone vs nisin concentration for pre-diffusion times 0, 24 and 48 h are as follows:

For 0-h pre-diffusion:

display math(4)

where,

display math

For 24-h pre-diffusion:

display math(5)

where,

display math

For 48-h pre-diffusion:

display math(6)

where,

display math

Selecting the best pre-diffusion time

After finding the best-fit model for each pre-diffusion time, the next step was to determine the best pre-diffusion time. Nisin concentration and its 95% inverse prediction interval were calculated using spline models for each pre-diffusion time. Pre-diffusion time with smaller prediction intervals was considered better.

The plot of inhibition zone vs log nisin concentration with the observed and predicted values of nisin concentration and its 95% inverse prediction interval is shown in Fig. 4. The inverse prediction interval for various nisin concentrations and pre-diffusion times is shown in Fig. 5. It appears that longer pre-diffusion time produced shorter prediction intervals with 48 h pre-diffusion time having the shortest prediction interval. Hence, our results suggest that 48 h pre-diffusion time should be used to estimate nisin concentration in the range 0·625–125 μg ml−1. When nisin concentration was less than 18·75 μg ml−1(log1·3), the prediction interval for 24 h pre-diffusion time was in the same range as that of 48 h pre-diffusion time. Therefore, 24 h pre-diffusion time can be used to predict nisin concentrations up to 18·75 μg ml−1.

Figure 4.

Inhibition zone vs log nisin concentration with observed (○) and predicted (△) values (using Eqn 3) of nisin concentration generated using R and showing 95% inverse prediction intervals for different pre-diffusion times.

Figure 5.

Ninety-five percent inverse prediction interval of nisin concentration vs log nisin concentration for spline models with pre-diffusion times (○) 0 h, (♢) 24 h and (□) 48 h.

Discussion

Based on our agar diffusion assay, wells with 3·5 mm diameter demonstrated higher accuracy for the quantification of nisin. Pre-diffusion at 4°C improved the distinction between inhibition zones at different nisin concentrations. Pre-diffusion at 4°C also improved the sensitivity of the assay. This observation is in agreement with results from other investigators (Pongtharangkul and Demirci 2004; Rogers and Montville 1991) who showed that pre-diffusion increased the inhibition zone and the slope of the regression line, enabling a more sensitive assay. Inhibition zone is the result of suppression of bacterial growth at or above the minimum inhibitory concentration (MIC) of the nisin. For a given agar formulation, agar thickness, temperature, well size and volume in the well, the inhibition zone is a function of the diffusivity of the antimicrobial within the agar, susceptibility of the bacteria and its growth rate (Cavenaghi et al. 1992). During pre-diffusion, there is no microbial growth, while nisin diffuses through the agar. As a result, the sensitive micro-organisms encounter higher concentration of nisin per bacterial cell. When there is no pre-diffusion, the sensitive micro-organisms encounter relatively lower nisin concentration per bacterial cell due to microbial growth. Therefore, with pre-diffusion, a lower concentration of nisin is sufficient to kill all micro-organisms within the inhibition zone. A longer pre-diffusion allows a larger volume of nisin to diffuse into the agar before the onset of microbial growth, shifting the MIC inhibition zone of nisin farther away from the well centre and giving larger inhibition zones for the same nisin concentration. Hence, longer pre-diffusion times result in higher sensitivity of the assay.

Based on AIC values, R2 and lack-of-fit test, spline models produced excellent correlation between nisin concentration and inhibition zone for 0-, 24- and 48 h pre-diffusion times. The statistical models tested in this study had poor correlation between inhibition zone and nisin concentration at 72 h pre-diffusion time. Spline model as the best-fit model in this study indicated (a) a change in the driving force for diffusion of nisin through agar in the vicinity of the knot point concentration and/or (b) different mechanism of bacterial inhibition at low and high concentrations of nisin.

The agar well has a limited volume of nisin solution. The progressive depletion of nisin solution from this well during pre-diffusion and incubation can reduce the driving force for diffusion of nisin through the agar. In addition, there was microbial growth during incubation at 30°C. An increase in microbial cell density in the agar can reduce the free volume, slowing down nisin diffusion further. Several researchers have reported a significant decrease in the diffusivity of substrate in gels containing immobilized cells due to increase in microbial cell density (De Backer et al. 1992; Mignot and Junter 1990a,b; Westrin and Axelsson 1991). The knot point may also be a result of different nisin killing mechanisms. Huang et al. (2002) showed that pore formation on a model bacterial membrane was dependent on the concentration of nisin. They found that nisin interacted with negatively charged lipid membrane and could form pores only above a certain concentration. At high concentrations, nisin behaved like a surfactant, breaking and removing the lipid membrane. This concentration-dependent behaviour of nisin may be the reason for different linear relationship between the inhibition zone and nisin concentration before and after the knot point. It should be pointed out that spline models are simplifications of a real situation where there is no real discontinuity. Although we have simplified the standard plot by using a spline model with one single knot point, the actual transition in the diffusivity of nisin or its kill mechanism would happen over a wide range of concentrations.

This modified agar diffusion assay can be used for the quantification of nisin in solutions and monitoring the rate of nisin production during fermentation. Study has shown that nisin A and nisin Z have significantly different diffusion behaviour in agar (De Vos et al. 1993). Our study is specific to the quantification of nisin A, and its application to other nisin variants (nisin Z, F, Q and U) needs to be tested. Our approach of using spline models can be helpful in accurately quantifying antimicrobials such as vancomycin (Walker and Kopp 1978) over a wide range of concentrations using agar diffusion bioassay. The proposed use of inverse prediction intervals to assess the accuracy of microbiological assays can be applied for the quantification of other antibiotics as well.

Ancillary