The chromatographic analysis (GC retention index (RI) and GC/MS) of the essential oils allowed the identification of 144 components representing 87.40 to 99.37% of the total oil content [23–25]. Twenty five major compounds at an average concentration greater than 0.9 ± 0,2% have been retained for the statistical analysis (Table 1). The main components were 1,8-cineole (4.5 ± 1,61 -70.4 ± 2.5%) followed by cryptone (0.0 - 20.9 ± 1.3%), α-pinene (1.0 ± 0.7 - 17.6 ± 7.5%), p-cymene (0.8 ± 0.1 - 16.7 ± 5.2%), α-terpineol (0,6 ± 0,3 - 10,3 ± 1,1%), trans-pinaocarveol (0.8 ± 0.2 - 7.0 ± 2.5%), phellandral (0.0 - 6.6 ± 0.4%), cuminal (0.0 - 6.6 ± 0,6%), globulol (0.6 ± 0.2 - 6.2 ± 0.9%), limonene (0.4 ± 0.2 - 4.4 ± 0.3%), aromadendrene (0.1 ± 0.0 - 3.6 ± 1.2%), sapthulenol (0.1 ± 0.1 - 3.2 ± 0.9%) and terpinene-4-ol (0.3 ± 0.1 - 3.0 ± 0.8%).
Principal components analysis (PCA) and hierarchical cluster analysis (HCA)
The yield content of the 25 selected component was significantly different (p < 0.05) among species (Table 1). These 25 components were used for the PCA and the HCA analysis. The PCA horizontal axis explained 47.2% of the total variance while the vertical axis a further 23.80% (Figure 1). The HCA based on the Euclidean distance between groups indicated three specie groups (A, B and C) (with a dissimilarity of 11.0) (Figure 2), identified by their essential oil chemotypes. Group A clearly stood out forming a separate group in the PCA (Figure 1) and a deep dichotomy in the HCA (Figure 2). It was correlated positively with the axes 1 and 2. Groups A and B were negatively correlated, their separation was mainly due to axis 2. Group A species reduced to E. odorata, the essential oil of which was characterized by the highest mean percentage of cryptone (20.9 ± 1.3%), cuminal (6.6 ± 0.6%), phellandral (6.6 ± 0.4%), verbenone (0.9 ± 0.2%), p-cymen-8-ol (2.9 ± 0.6%), sapthulenol (3.2 ± 0.9%), carvacrol (1.7 ± 0.3%), p-cymene (16.7 ± 5.2%), terpinene-4-ol (3.0 ± 0.8%), caryophyllene oxyde (1.7 ± 0.2%), viridiflorol (4.5 ± 1.6%) and by the lowest level of 1,8-cineole (4.5 ± 1.6%). Group B, made up of E. maidenii, E. lehmannii, E. sideroxylon and E. cinerea, has essential oils characterized by the highest amount of limonene (3.1 ± 0.2 to 4.4 ± 0.2%), α-terpineol (2.2 ± 0.3% for E. maidenii to 10.3 ± 1.1% for E. cinerea). The PCA showed that the variation between these species is mainly due to the variation of 1,8-cineole content (57.8 ± 1.9% for E. maidenii to 70.4 ± 2.5 for E. cinerea) and of α-pinene (4.5 ± 0.7% for E. cinerea to 17.6 ± 7.5% for E. lehmannii). Group C, consisting of E. astringens, E. leucoxylon and E. bicostata, has essential oils distinguished by their highest mean percentage of epiglobulol (1.0 ± 0.2 - 1.2 ± 0.3%), globulol (5.4 ± 1.2 - 6.2 ± 0.9%), trans-pinocarveol (4.3 ± 1.0 to 7.0 ± 2.5%), aromadendrene (2.0 ± 0.8 - 3.6 ± 1.2%) and pinocarvone (1.2 ± 0.2 - 2.2 ± 0.5%). The E. bicostata and E. leucoxylon oils differed from E. astringens oil by their richness in 1,8-cineole (68.0 ± 5.3, 59.2 ± 10.1%, respectively) and their poverty in α-pinene (3.7 ± 1.2, 7.8 ± 2.3%, respectively) against 42.0 ± 5.9% and 22.0 ± 6.0%, respectively for the latter species.
As in the present study, E. cinerea, E. sideroxylon, E. bicostata, E. maidenii, E. leucoxylon, E. lehmannii and E. astringens have been reported to contain 1,8-cineole as a major component [26–30]. It was also reported that E. cinerea grown in Morocco contained a higher mean percentage of 1,8-cineole (87.8%) than that from Tunisia (70.4-2.5%) , whereas E. sideroxylon acclimated in Tunisia was richer in 1,8-cineole (69.2-0.6%) than that from Congo (59.9%) . The Tunisian E. astringens essential oil had the same chemotype as that extracted from leaves picked from Moroccan tree plantations with an important difference in their mean percentages. Both of them were different from the Australian E. astringens essential oil , which was represented by β-caryophyllene (14.75%), p-cymene (17.72%), and α-pinene (12.53%). The same principal components were also observed in E. leucoxylon essential oil [33, 34], whereas the essential oil obtained from Iran was significantly richer in 1,8-cineole (89.8%) than that from Tunisia .
The essential oils were tested for their putative antibacterial activity against seven bacterial isolates represented by 81 strains (Table 2). As shown in Table 2, E. odorata oil possessed the best activity against S. aureus (27.4 ± 10.7 mm, zdi), followed by S. agalactiae (19.4 ± 5.6 mm, zdi), H. influenzae (19.2 ± 9.6 mm, zdi), S. pyogenes (19.0 ± 0.0 mm, zdi) and S. pneumoniae (17.4 ± 4.1 mm, zdi). E. maidenii oil showed a relatively good activity against S. aureus (22.8 ± 6.8 mm, zdi). To evaluate the correlation between the antibacterial activities and the essential oils of the eight Eucalyptus species, all the mean values of the zone diameters inhibition were subject to the PCA and the HCA analysis. The statistical analysis of the antibacterial activities of the oils showed a significant difference among Eucalyptus species oils and the tested bacterial strains (p < 0.05). The PCA horizontal axis explained 59.39% of the total variance, while the vertical axis a further 13.84% (Figure 3). The HCA showed two species groups (I and II), identified by their bacteria growth inhibition with a dissimilarity ≥ 15.0 (Figure 4). When the dissimilarity was ≥ 5.0, group II was divided into three subgroups (IIa, IIb and IIc). The horizontal axis permitted the separation of group I from group II, however axis II separated all the species of the group II into three subgroups. Group I, limited to the Gram negative (Gram(−)) bacteria, P. aeruginosa and K. pneumoniae, forms a deep dichotomy within the HCA analysis and a clearly separated group in the PCA analysis. These two strains were the most resistant to the majority of Eucalyptus essential oils with zdi < 7.1 mm for P. aeruginosa and 10.7 ± 1.5 mm for K. pneumoniae. Compared to the growth inhibition zone produced by ciprofloxacin against P. aeruginosa (34.7 ± 5.0 mm, zdi) and K. pneumoniae (32.4 ± 2.9 mm, zdi). Subgroup Ia was limited to S. aureus which was characterized by the highest sensitivity to E. maidenii and E. odorata oils (22.8 ± 6.8 and 27.4 ± 10.7 mm, zdi, respectively). This high sensitivity could be due to the disposition of E. maidenii and E. odorata oils with a relatively high mean percentage of the monoterpene hydrocarbons p-cymene (7.4 ± 2.9, 16.7 ± 5.2%, respectively). Previous studies have reported the high sensitive character of S. aureus to essential oils with a high content of p-cymene . In addition, other researchers reported that this sensitivity of S. aureus was due to the single layer wall of the bacteria . Comparing these results with those obtained with antibiotics, E. odorata essential oil produced a similar inhibition to that produced by gentamicin, erythromycin, vancomycin and benzylpenicillin (29.6 ± 6.2, 29.9 ± 5.1, 25.3 ± 4.4 and 24.5 ± 7.5 mm, dzi, respectively). However, this activity remained lower than that produced by fosfomycin (34.3 ± 11.1 mm, dzi). Sub group IIb represented by S. pneumoniae, showed a particular sensitivity to E. odorata and E. bicostata essential oils (17.4 ± 4.1 and 17.0 ± 4.0 mm, zdi). This inhibition remained lower than that produced by its specific antibiotics with zone diameters inhibition ranging from 26.3 ± 12.0 mm (erythromycin) to 35.6 ± 5.5 mm (fosfomycin). E. lehmannii, E. sideroxylon and E. cinerea oils did not show significant antibacterial activities with inhibition zones of 9.8 ± 2.4, 10.7 ± 2.5 and 11.5 ± 2.8 mm, respectively. Subgroup IIc, consists of Streptococcus B, S. pyogenes and H. influenzae. These strains were separated from all the others and correlated positively with the two axes and with E. cinerea and E. sideroxylon, the essential oils of which were characterized by a comparable activity against the previous bacterial strains, with inhibition zone diameters varying from 11.6 ± 1.4 mm to 13.0 ± 6.3 mm. Their activities were considered relatively as being lower than the tested antibiotics such as rifamicine and ampicilline. However E. odorata oil, which was removed from this group, showed the best activity against these bacterial strains with inhibition zone diameters varying from 17.4 ± 4.1 mm for S. pneumoniae to 19.4 ± 5.6 for Streptococcus B, but it remained much lower than that produced by their specific antibiotics. The MIC was performed for oils which have produced an inhibition ≥ 17 mm for clinical bacterial strains such as H. influenzae (reference 160), S. agalactiae (reference 3) S. pyogenes (reference 545) and S. aureus (reference 278). The result of their MIC was listed in Table 3. E. odorata and E. bicostata oils were characterized by the lowest MIC for Hemophylis influenzae (reference 160) (0.306 mg/mL), followed by S. agalactiae (reference 3) (10.4 mg/mL). These results were confirmed by the disc diffusion method. The highest MIC against S. aureus ( reference 278) was shown for the oils of E. bicostata (169 mg/mL), E. odorata (156.6 mg/mL) and E. maidenii (151.8 mg/mL). This finding was in contradiction to results obtained by the disc diffusion method. According to the classification of Schaechter et al. (1999)  and Soro et al. (2010) , E. odorata, E. bicosta and E. maidenii oils were considered bactericidal (MBC/MIC < 4) against the tested strains, however the two first oils showed a better bactericidal activity against H. influenzae (reference 160) and S. aureus (reference 278) than that obtained with S. pyogens (reference 545) and S. aglatctiae (reference 3).
Seven Eucalyptus oils were tested for their antifungal and anti-yeast activities. The result of their average percentages inhibition was listed in Table 4. Their variance analysis showed a highly significant effect for all oils (p < 0.05), except for E. odorata and E. cinerea oils (p > 0.05).
The PCA horizontal axis explained 73% of the total variance, while the vertical axis a further 14.15% (Figure 5). The HCA indicated three groups of species (A, B and C) with a dissimilarity ≥ 5.0 (Figure 6).
Group A, limited to Scopulariopsis brevicaulis, correlated negatively with axis 1 and with the group C. This species was distinguishable in the PCA and formed a distinct group. It was considered as being the most resistant to all the tested oils. Group B was formed by Trichophyton soudanense (correlating positively and negatively with axis 1 and 2, respectively) and by Trichophyton rubrum (correlating negatively with the two axes). These strains were distinguished by their highest sensitivity to E. odorata oil (100% of inhibition); however, the variability between these two fungi strains was mainly due to the higher sensitivity of Trichophyton soudanense to E. lehmannii and E. maidenii oils than that of Trichophyton rubrum and also to the resistance of Trichophyton soudanense to E. cinerea oil. Group C, which was formed by the yeast Candida albicans and by the fungi Microsporum canis, correlated positively with the two axes and was characterized by a relatively high sensitivity to E. bicostata, E. astringens and E. sideroxylon; however, these strains showed a higher resistance to the other oils.
According to the chemical composition of these essential oils, the antifungal activity was not related to the high content of one chemical compound, rather than to synergic effects between major and minor components. For example, 1,8-cineole, which was discussed above as the principal component of the most essential oils, did not correlate with the high antifungal activity because E. c inerea (70.4 ± 2.5%), E. leucoxylon (59.2 ± 10.1%), E. lehmanii (56.6 ± 4.3%), and E. maidenii (57.8 ± 1.9%) showed a lower antifungal activity than E. bicostata, E. astringens and E. sideroxylon. Additionally, the essential oils of E. bicostata, E. astringens, characterized by the highest content of pinocarvone (2.2 ± 0.5 and 1.8 ± 0.8%, respectively), tr-pinocarveol (4.6 ± 0.6 and 7.0 ± 2.5%, respectively) and globulol (5.4 ± 1.2 and 6.2 ± 0.9%, respectively), showed a lesser antifungal activity than the essential oil of E. sideroxylon.
The cytotoxicity effect of the eight Eucalyptus essential oils on Vero cell lines varied significantly within species (Table 5). Vero cells were resistant to the essential oils of E. maidenii, E. sideroxylon and E. cinerea with CC50 values of 253.5, 247.3 and 204.5 mg/mL, respectively. The essential oils of E. odorata, E. leucoxylon, E. lehmannii, E. astringens and E. bicostata demonstrated a different behavior and their cytotoxicity increased considerably with CC50 varying from 6.2 to 16 mg/mL. We did not notice any clear correlation between the chemical composition of the tested oils with the results of the cytotoxic effect and further investigation needs to be undertaken. However, the lowest cytotoxicity was observed with oils having a high content of 1,8-cineole but with a moderate amount of α-pinene and limonene such as those of E. maidenii, E. sideroxylon and E. cinerea. The high cytotoxic effect was shown with E. lehmannii, E. astringens oils, which were characterized by a higher mean percentage of the monoterpene α-pinene and by a moderate mean percentage of 1,8-cineole. The present result was confirmed by Setzer et al. (2006) , who demonstrated that the monoterpene hydrocarbons α-pinene had a stronger cytotoxicity activity against Hs 578 T and Hep-G2 cell lines than 1,8-cineole. It was also observed that a synergic effect among the oxygenated sesquiterpens globulol and viridiflorol and the monterpene hydrocarbons α-pinene of E. astringens, E. lehmannii and E. leucoxylon oils could make the cell lines more sensitive. On the other hand, the significant high cytotoxicity of E. odorata oil could be explained by the latter’s lack of 1,8-cineole and its richness in the ketone cryptone, the monoterpene hydrocarbons p-cymene and in theses aldehydes: phellandral and cuminal . This cytotoxicity effect could be due to the synergetic effect of the previous main constituents of this essential oil. Compared to the previous studies, the cytotoxicity of our studied essential oils was very low CC50 > 20 μg/mL . Therefore they could be considered as being safe for use at non cytotoxic concentrations.
In order to elucidate the mode of antiviral action and to identify the target site, cells were pre-treated with essential oils before viral infection (pre-treatment of cells) and the virus was incubated with essential oils before cell inoculation (pre-treatment of virus). All samples tested were used at their maximum non-cytotoxic concentrations (Table 5). The essential oils of E. sideroxylon, E. lehmannii, E. leucoxylon and E. odorata showed no inhibition of viral infection, whereas the most significant antiviral activity was shown with the essential oils of E. bicostata (IC50 = 0.7 - 4.8 mg/mL) and E. astringens (IC50 = 8.4 mg/mL), followed by essential oils of E. cinerea (IC50 = 102–131 mg/ml) and E. maidennii (IC50 = 136.5 - 233.5 mg/mL). The selectivity index describes the ratio between the cytotoxic and the antiviral activity of a tested sample. The virus pretreatment with E. bicostata essential oil showed a better antiviral activity (IC50 = 0.7 mg/mL, SI = 22.8) than cell-pretreatment (IC50 = 4.8 mg/mL, SI = 3.33). The essential oil of E. astringens showed an antiviral activity only when incubated with a virus prior to cell infection. This activity was dose-dependent and the antiviral activity decreased with the diminishing essential oil concentration. E. cinerea and E. maidenii essential oils showed an antiviral activity at a concentration of 150 mg/mL when incubated with cells. This activity increased significantly at the same concentration when the sample was incubated with a virus prior to infection.
According to these results, no correlation was found between the chemical composition and the antiviral assay. Therefore the activity of the tested essential oils could be due to a synergism between the major and the minor components. Altogether, the essential oils used in the present study exhibited the best antiviral effect when incubated with a virus. From the presented results, E. odorata oil was associated with E. maidenii, E. bic ostata, E. lehamnnii, E. astringens and E. leucoxylon oils in both the antibacterial PCA and HCA. However, it remained associated with E. maidenii and E. lehmannii in the antifungal PCA analysis. Therefore, the biological activity has allowed the association of different chemotypes in the same group producing a similar biological activity. This allowed us to deduce that the global biological activity of these oils was mainly due to an addition or a synergism effect between the major and the minor components. This was confirmed by Paster et al. (1995) . The best activity was recorded with E. odorata oil against the majority of the microbial strains with inhibition zone diameters almost equal to those produced by erythromycin against clinical strains H. inflenzae, S. aureus, and to the Pevaryl against the three dermatophytes fungi and the Candida albicans. This property could be explained by the richness of E. odorata oil in p-cymene (16.7 ± 5.2%), terpinene-4-ol (3.0 ± 0.8%), spathulenol (3.2 ± 0.9%), carvacrol (1.7 ± 0.3%), caryophyllene oxide (1.7 ± 0.2%), p-cymene-8-ol (2.9 ± 0.6%), verbenone (0.9 ± 0.2%), viridiflorol (1.0 ± 0.3%), phellandral (6.6 ± 0.4%), cuminal (6.6 ± 0.6%) and cryptone (20.9 ± 1.3%), and by its poverty in 1,8-cineole (4.5 ± 1.6%). It appeared that all the strains were resistant to oils rich in the latter component which varied from 42.0 ± 5.9% for E. astringens to 70.4 ± 2.5% for E. cinerea. On the other hand, oils characterized by a small quantity of 1,8-cineole and by a medium mean percentage of the monoterpene, p-cymene, the ketone, cryptone, the aldehydes, phellandral and cuminal were more active. These results were confirmed by Mulyaningsih et al. (2010) , who demonstrated that the antimicrobial activity of pure 1,8-cineole was inferior to the totality of E. globulus fruit oil, considered poor in the latter component and rich in aromadendrene and globulol. Dorman and Deans (2000), also confirmed that the antimicrobial activity increased with oils rich in aldehydes, ketones and phenols . E. maidenii oil was characterized by a relatively high mean percentage of 1,8-cineole (57.8 ± 1.9%), p-cymene (7.4 ± 2.9%), α-pinene (7.3 ± 0.7%), limonene (3.1 ± 0.2%), α-terpineol (2.2 ± 0.2%), the activity of which occupied the second position after E. odorata oil against all the bacterial strains (7.1 ± 1.5 mm, zdi, for P. aeruginosa to 22.8 ± 6.8 mm, zdi, for S. aureus) and also against the fungal strains with an inhibition percentage varying from 22.8 ± 2.1% for Scopulariopsis brevicaulis to 66.8 ± 17.1% for Trichophyton soudanense. However, oils of E. sideroxylon and E. cinerea, which were distinguished by the highest levels in 1,8-cineole, were less active against the tested microorganisms than the majority of the remaining oils. The present finding was in contradiction with previous studies reporting that 1,8-cineole had strong antimicrobial properties against many important pathogens [44, 45]. It seems that the activity of this chemical compound was inhibited by other minor components. Further investigations need to be carried out to better understand the present issue.
According to the study of Claudio et al. (2008), the essential oil of E. globulus has shown a higher antibacterial activity against Haemophilus influenzae (28 mm, zdi) than the tested essential oils in the present study, whereas it possessed comparative inhibition activities against S. pneumoniae (15 mm vs 9.8-17.4 mm, zdi). However, our essential oils exhibited a better activity against K. pneumoniae (6.6-10.7 mm vs 0 mm, zdi), S. aureus (12.2-27.4 mm vs 2 mm, zdi), S. agalactiae (11.1-19.4 mm vs 3 mm, zdi) and S. pyogenes (9–19 mm vs 5 mm, zdi) . Martin et al. (2010) have reported similar findings concerning the antibacterial activity of the essential oils of E. dives and E. staigeriana against Pseudomonas aeruginosa (7.7-9.1 mm vs 6–7.1 mm, zdi) .
We also noticed that Microsporum canis is more sensitive to E. sideroxylon oil than that of E. cinerea. This sensitivity could be attributed to the presence of a higher content of α-pinene and limonene and a lower percentage of α-terpineol in E. sideroxylon oil. The comparative study of the antibacterial activity result of the tested oils obtained by the disc diffusion method with the results obtained by micro-well dilution showed for some species a concordance of the two methods and discordance for others, especially against the Gram positive (Gram (+)), S. aureus (278). Compared to the activities produced by the tested antibiotics, all the essential oils were less active. An association of the most active oils together may moderate the activity against the most resistant strains.