Optimization by mixture design of the antimicrobial activities of five selected essential oils

The individual antimicrobial activities of essential oils have been reported by many authors. However, there is little information about the effects of their mixtures in order to maximize their effect and reduce the growing resistance of pathogens to existing medicines. So, the aim of this work is to optimize the antibacterial and antifungal activities of essential oils from Plectranthus glandulosus, Ocimum gratissimum, Cymbopogon citratus, Cymbopogon nardus and Eucalyptus PF1. The mixtures of these essential oils were tested on seven bacterial and one fungal strain by employing the Mueller Hinton disc diffusion method. The diameters of the inhibition zones were measured after 24 h of incubation at 37°C. The results showed significant effects and regressions due to pure and composite mixtures on the response. The highest diameters of 30 and 27 mm were observed respectively with the pure essence of C. citratus on Candida albicans and the composite mixture. The binary mixtures showed more significant effects than the pure ones with the highest positive coefficient of regression 17.20 due to the Plectranthus glandulosus and Eucalyptus PF1 mixture on Pseudomonas aeruginosa . The growth inhibition data fitted the quadratic models for all individual strains except those of Staphylococcus aureus that better fitted the special cubic model. Some regression models of individual and combined microorganism responses have been proposed, as well as optimizations to maximize the inhibition zone diameters.


INTRODUCTION
The use of plant derivatives to treat diseases is universal and dates back to immemorial times (Van Wyk and Wink, 2018). Plants contain active substances capable of inactivating pathogenic microorganisms and parasites and correcting physiological dysfunctions (Daniel, 2006). They treat cancers (Mukherjee et al., 2001;Cragg and Newman, 2005;Mohan et al., 2013), malaria (Krettli et al., 2001;Andrade-Neto et al., 2003, Negi et al., 2014, influenza (Phillipson and Wright, 1991;Wang et al., 2006;Mukhtar et al., 2008), diarrhea (Barbosa et al., 2007;Dubreuil, 2013) and many other diseases that rage around the world. All the underground and above ground parts of fresh or dried plants as well as sap are used in various forms, most often as a decoction or infusion in water or alcoholic beverage, sometimes in powder form. Plant products are ingested, inhaled or applied to the skin.
Research has allowed identifying numerous bioactive plants molecules, some of which are isolated and consumed directly as drugs and others serve as models for the synthesis of more active and inexpensive analogs (Dewick, 2002;Evidente andKornienko, 2009: Roleira et al., 2018).
The microorganisms' strains that are the subject of this study are responsible for toxiinfections which can cause death in individuals of all ages (Morrison and Wenzel, 1984;Cross, 1985;Griffin and Tauxe, 1991;Kotiranta et al., 2000;Mavor et al., 2005;Skippen et al., 2006). These microorganisms, like other pathogens, develop resistance to existing drugs. This constitutes a risk with sometimes fatal consequences for patients in addition to unnecessary financial expenses. One way to overcome this obstacle is to design remedies that integrate active principles with complementary mechanisms. To this end, mixture design can help optimize the efficacy of these products. The aim of this work was to check whether there are significant interactions between essential oils from Plecthrantus glandulosus, Ocimum gratissimum, Cymbopogon citratus, Cumbopogon nardus and Eucalyptus PF1 to improve their antimicrobial activity.

Essential oils extraction
The leaves of P. glandulosus, O. gratissimum, C. citratus, C. nardus were harvested in May 2019 in an experimental culture carried out in Douakani village. The specimens of this plant were identified at the National Herbarium, in Brazzaville. The Eucalyptus PF1 leaves were provided by the National Reforestation Service (SNR) based in Dolisie between January and February 2019. The specimens were identified by the engineers in this service. Eucalyptus PF1, is a natural hybrid between E. urophylla and two or three individuals of Eucalyptus alba (mother tree) and one group of poorly identified hybrid Eucalyptus (parent trees) that come from a Brazilian arboretum. Mikolo et al. 571 The essential oils were extracted by hydro-distillation using a manmade distiller. Three kilograms of dried leaves of P. glandulosus, O. gratissimum, C. citratus, C. nardus and Eucalyptus PF1 were placed in a 10 l boiler and water was added to 2/3 of the boiler that was then heated to boiling with firewood. Running water from a waterfall was used to cool water and oil vaporsl. Water was eliminated from the resulting two-phase liquid mixture by opening the tap of the separator funnel. The essential oil was then collected in a shade bottle, mixed with anhydrous sodium sulphate and stored at 4°C until required.

Mixtures design
The experiment was carried out in 3 replicates by applying a mixture design of experiment of extreme peaks of degree 1, with 5 components corresponding to the essential oils of P. glandulosus (X 1 ), O. gratissimum (X 2 ), C. nardus (X 3 ), C. citratus (X 4 ) and Eucalyptus PF1 (X 5 ). The inferior and superior levels of the oil proportions have been fixed at 0 and 2.5 respectively, with a unique total of 2.5 ml. The plan was increased with the central point and the points on the axes and designed using Minitab 3.17.1. The mixtures were prepared in 5 ml shaded flasks. The matrix of this experiment is shown in Table 1.

Strains identification
Different selective culture media were used for the isolation and identification of the strains: 1) Salmonella-Shigella Agar (Bio RAD) for isolation of bacteria of the genus Salmonella whose colonies are black-centered (H 2 S+) and lactose-positive and the genus Shigella whose colonies were colorless (H 2 S−) and lactose-negative; 2) Cetrimide Agar (BIO RAD) for Pseudomonas aeruginosa, 3) Mannitol Salt Agar (Bio RAD) for isolation of bacteria of the genus Staphylococcus including mannitol positive Staphylococcus (Staphylococcus aureus); 4) Mossel Agar medium for Bacillus; 5) Sabouraud chloramphenicol (Bio RAD) for Candida albicans and 6) Methylene Blue Eosin Agar (Bio RAD) for the isolation of enterobacteria in which the major species is Escherichia coli that is characterized by a metallic sheen. Strains of Klebsiella spp and E. coli were identified by the Enterobacteri Sytem gallery.

Essential oils antimicrobial activity evidence
Antimicrobial activity was assessed by the Mueller Hinton diffusion method (Prats et al., 2000;Matasyoh et al., 2007). An inoculum containing the strain to be tested was prepared from a pure and young colony. Its optical density was adjusted to 0.1 at a wavelength of 625 nm with a spectrophotometer which is equivalent to the cell density of 0.5 (Mac Farland No 0.5) (Boukhatem, 2013). The inhibition test was performed in Petri dish containing solid and sterile Mossel Agar (Bin RAD) medium for the growth of Bacillus, Mueller-Hinton agar (MHA) medium for the other bacteria, and Sabouraud dextrose agar (SDA) for fungi, previously inoculated with 0.1 ml of microbial suspension. Using forceps, the oil-soaked Wattman paper discs were placed on the medium. After 24 h of incubation at 37°C., the diameters of the observable zone of inhibition were measured.

Data analysis
Details of statistical methods used for data processing were found in Box and Draper (2007) and ReliaSoft (2015). The computing was done using the Minitab 3.17.1 software. The null hypotheses of equality of the means were tested by the analysis of variance (ANOVA) and the P-values at the significant level α of 0.05. The null hypothesis was rejected for a P-value<0.05. The P-value was used in order to know if the test statistic was just into the critical region or was far out into the region. The data shown in Tables 3 to 6 result from the ANOVA test of the inhibition zone diameters of the raw data from Table 2 followed by a two by two comparison test of Tukey. The aim was to know if there were significant differences from the diameter means due to the mixtures, on one side, and to the strains on the other side. Tables 7 and 8 show the results of ANOVA used in order to assess the main effects and interactions of the mixtures as well as the regressions of the inhibition diameters as a function of the essential oil proportions for each of the strains. The issue was to test whether the coefficients were equal to 0 or not. So, some statistics such as the coefficients of regression and determination were calculated to estimate and validate the models. For this purpose, the diameters were transformed into natural logarithms (ln) in order to obtain the highest coefficient of determination values (R 2 ), as well as knowing that inhibiting the growth of microorganisms involves biochemical processes that generally follow logarithmic distributions. These results allowed us to suggest the models showed below. The combined and individual optimizations were carried out by selecting the response variables to be optimized. The goal was to maximize the response; the lower level and the target being fixed respectively at ln(Y) = 2.7 (Y =14.88 mm) and ln(Y) = 3 (Y = 20.08 mm). Results showed in Table 9 were calculated automatically using the Minitab software.

Inhibition of the pathogens studied as a function of the mixtures
The inhibition zone diameters of the eight pathogen strains by essential oils are shown in Table 2. The maximum means from 25 to 29 mm were reached with the mixtures E4, E6, E8, E9 and E10 on P. aeruginosa_1 and C. albicans. According to the results of the analysis of variance, Table 3 (P = 0.000), there was at least one significant difference between the means at the significance level of alpha equal to 5%. The results of Tukey's two-by-two comparisons of the inhibition diameters (Table 4) classify the diameters obtained in three groups, A, B and C. The E6 and E10 mixtures yielded means of 14.46 and 14.83 mm, respectively higher than the others.

Studied strains sensitivity
In the ANOVA table (Table 5), the p-value (0,000) for strain indicates that at the 5% significance level, not all means are equal. Tukey's test provides grouping information and two sets of confidence intervals with multiple comparisons. The grouping table (Table 6) shows that group A includes the most sensitive strains overall, namely: P. aeruginosa_1, C. albicans and Bacillus subtilis while group B contains the less susceptible strains: E. coli, Salmonella, P. aeruginosa S. aureus and Klebsiella spp.

Modeling inhibition zone diameters
The estimates of the regression coefficients are shown in Table 7. They explain how well some mixtures have significant effects on the microorganism growth inhibition. The positive coefficients obtained from binary mixtures mean that the two components act in synergy or are complementary (Table 8). In other words, the inhibition zone diameter is greater than that obtained by calculating the mean of each pure mixture diameters. The synergism   of the mixture X 1 *X 5 is the most important on 4 out of 8 strains (B. subtilis, E. coli P. aeruginosa_1 and C. albicans). Negative coefficients mean that the two components are antagonistic. Thus, the mean of the inhibition zone diameter is less than the one obtained by calculating the mean of each pure mixture diameters. This was observed particularly for interactions X 1 *X 2 and X 1 *X 3 on 5 out of 8 strains followed by X 1 *X 4 . The mixtures X 1 *X 2 , X 1 *X 3 , X 1 *X 4 , X 1 *X 5 and, X 2 X 3 are the only mixtures with two components and X 1 *X 2 *X 3 with three components giving significant results (p <0.05). The following terms could not be estimated; they were deleted: X 2 *X 4 , X 2 A 5 , X 3 *X 4 , X 3 *X 5 and X 4 *X 5 . Values of p <0.05 were obtained for the special cubic regression model for S. aureus and quadratic for the other seven strains of pathogens, indicating that the models estimated by the regression procedure are significant at the 0.05 alpha level. This means that at least one coefficient is different from zero. The importance of the coefficients in the five pure mixtures indicates that the essential oils of P. glandulosus (1.810-2.970), O. gratissimum (1.785-2.633), C. nardus (1.900-3.030), C. citratus (1.842-3.385 ) and of Eucalyptus PF1 (1.842-3.385) give significant inhibition zone diameters ( Table 7).
The R 2 values in Table 8 indicate that the predictors explain 79.51 to 95.10% of the variance in the diameter of the inhibition zone; those of R 2 (prev) from 64.64 to 89.09% indicate that the quadratic regression models predict correctly the responses of the new observations for all the strains, except for the special cubic regression of the data obtained with S. aureus that gave a slightly lower value of R 2 (prev) equal to 53.8 %. The R 2 (fitted) values varying between 70.19 and 92.51%, showing that the theoretical data actually fit the models. All the quadratic models are written as: β 1 X 1 + β 2 X 2 + β 3 X 3 + β 4 X 4 + β 5 X 5 + β 12 X 1 X 2 + β 13 X 1 X 3 + β 14 X 1 X 4 + β 15 X 1 X 5 + β 23 X 2 X 3 For example, using the coefficients of the terms shown in Table 8, the quadratic regression model equation for the diameter of the Klebsiella inhibition zone is: Y = ln (d) = 0.92X 1 + 0.86X 2 + 0.91X 3 + 0.74X 4 + 0.97A 5 -1.21X 1 X 2 -1.055X 1 X 3 + 0.53X 1 X 4 + 0.48X 1 X 5 + 0.19X 2 X 3, where d is the diameter in mm.

Inhibition zone diameters optimization
The combined optimizations allowed examining the combinations of essential oil proportions that simultaneously optimize multiple responses to meet the requirements for all responses in the set, with the individual and composite desirabilities (d) of the response variables equal to 1.0. To simultaneously inhibit P. aeruginosa _1, B. subtibilis and C. albicans, a combination of X 1 (1.22 ml), X 2 (0 ml), X 3 (0 ml), X 4 (0 ml) and X 5 (1.24 ml), yields the following responses: 6.7115 (821.80 mm) for P. aeruginosa _1, 4.9116 (135.86 mm) for B. subtibilis and 5.9191 (372.08 mm). A combination of X 1 (1.6162 ml), X 2 (0.2946 ml), X 3 (0 ml), X 4 (0.2503 ml) and X 5 (0.33389 ml) provides the following responses for Klebsiella spp: 2, 1601 (8.7 mm), Salmonella spp: 2.4531 (11.6 mm), P. aeruginosa: 2.0250 (7.6 mm), P. aeruginosa _1: 3.6168 (37.2 mm), B. subtibilis: 3.0176 (20.4 mm), C. albicans: 3.4926 (32.87 mm) and S. aureus: 2.1819 (8.86 mm). The optimizations of the individual responses are shown in Table 9. It is observed that the binary mixtures make it possible to obtain inhibition diameters of the germs studied from 19 to 410 mm. That is, the growth of certain strains such as P. aeruginosa _1, B. subtilis and C. albicans should be completely inhibited.

DISCUSSION
It appears from this work that some pure and composite mixtures of tested essential oils are capable of inhibiting the growth of the microbial strains studied. This is justified by the average inhibition zone diameters observed ranging from 10 to 19 mm. The maximums diameters up to 30 mm were reached with the essential oils of C. nardus and C. citratus on Candida albicans. There were significant differences between the means of the inhibition zone diameters for the mixtures and the strains. The mean highest diameter of inhibition zone was given by the mixtures and E10.
There was also significant difference in the sensitivity of the strains to essential oils. Three strains were more sensitive namely: P. aeruginosa_1, B. subtilis and C. albicans; the others being less sensitive. These diameters of inhibition are close to those reported by Leopold et al. (2002) and Naik et al. (2010) and Leopold et al. (2002). However, Kon and Rai (2012) obtained the inhibition zone diameter reaching 64 mm with Cinnamomum zeylonicum essential oil.
The results of the regression analyses showed that the effects of certain essential oil mixtures were significant. Some synergistic and antagonistic effects were observed. The inhibition was greater with binary mixtures which gave the highest regression coefficients. The X 1 *X 5 combination of P. glandulosus and Eucalyptus PF1 gave coefficients of 17.20 and 16.43 mm on P. aeruginosa _1 and C. albicans, respectively, showing the existence of a synergistic or additive effect. This mixture is therefore better indicated for treating pathologies caused by these two germs. The results reveal that the most efficient mixtures of essential oils to use for designing medicines or antiseptics should be those with two ingredients, as the active ingredient will probably be diluted with the increasing of the mixture components number. The 576 J. Med. Plants Res. The most antagonistic effect was observed on the X1*X3 mixture composed of P. glandulosus and C. nardus on S. aureus, giving a regression coefficient of -16.01. The combinations X 1 *X 2, X 3 *X 3 and X 1 *X 4 that showed antagonistic effects on 4-5 strains out of 8 should be avoided for inhibiting most of the microorganisms studied. What is important in these results is not the size of the inhibition zone diameter, but the regression coefficient and the degree of correlation between the input variables and the responses allowed to explain the variability of the response and to predict responses of untested factor levels. In fact, the size of the inhibition diameter depends both on the concentration of the active principle in the ingredient and on its efficacy. However, the degree of correlation between the variable and the response depends on the effectiveness of the growth inhibition. Thus, an essential oil that will give higher coefficients and values of R 2 will probably contain the most effective molecule to inhibit the growth of the tested germ even if the diameters values are low.
The tests of the regression validation led to select the quadratic model for all the strains except one, S. aureus, the data of which better fitted the special cubic model. Optimization was done to maximize the diameter of individual or group of germs. Individual or combined optimizations yielded combinations of essential oils that would significantly inhibit the growth of the tested strains. The combination of essential oils of P. glandulosus and Eucalyptus PF1 was the one that gave the best results on P. aeruginosa_1, B. subtilis and C. albicans. Forecast diameters of the order of 400 mm simply mean that microorganisms can be completely inhibited, depending on the proportion and diffusivity of the ingredient.

Conclusion
The blending design allowed 88 combinations of essential oils and microbial strains to be tested in one experiment. Statistical analyses of the obtained results made it possible to identify both the most effective combinations of essential oils and the most sensitive microbial strains. It should be noted, as revealed by other authors, that it is the binary mixtures that indicate the most important positive and negative interactions. The most significant positive and negative interactions were observed respectively with the combinations, X1*X5 (P. glandulosus + Eucalyptus PF1) and X1*X2 (P. glandulosus and O. gratissimum).
Analyses of the obtained data allowed the selection of the mathematical models best adapted to the variation of the diameter of the inhibition zones of the microbial strains according to the proportions of essential oils in the mixtures. Data from all strains were more suitable for quadratic regression except for Staphylococcus that was better adapted to the special cubic model. This led to optimizations of the individual and combined responses of the essential oils to obtain the proportions of essential oils that would maximize the growth inhibition diameters of the microbial strains. Finally, this is the first time that the antimicrobial activity of the essential oil of the PF1 clone of Eucalyptus has been demonstrated.