Gallic acid, catechin, 2,6-Di-tert-butyl-4-methylphenol (BHT), Folin-Ciocalteu′s phenol reagent, linoleic acid, potassium ferricyanide (III), iron (III) chloride, iron (II) chloride tetrahydrate, 2,2-diphenyl-1-picrylhydrazyl (DPPH), 3-(2-Pyridyl)-5,6-diphenyl-1,2,4-triazine-p,p′-disulfonic acid monosodium salt hydrate (ferrozine), pyrocatechol violet (PV) and sodium nitrite were purchased from Fluka-Sigma-Aldrich^® (St. Louis, MO, USA). Aluminium chloride anhydrous and 2,4-dinitrophenylhydracine were obtained from Panreac (Barcelona, Spain). The phenolic compound standards caffeic acid, genkwanin and apigenin were from Extrasynthese^® (Genay, France), and rosmarinic acid, carnosol and carnosic acid from Fluka-Sigma-Aldrich^® (St. Louis, MO, USA). The Juglone standard was from Merck (Darmstadt, Germany). All other chemicals were of analytical grade.

An extensive prospection of Rosmarinus officinalis L. wild populations was carried out in Spain in the years 2007, 2008 and 2009. The plant material was collected at the full-flowering time, hydrodistilled and chemically analyzed. Ten of these populations were selected on the basis of the yield and chemical variability of their essential oil. Table 1 provides a description of the collecting places based on geographical parameters. Individual plants from each selected population were vegetatively propagated and rooted under greenhouse conditions in 2010. The cultivation plots were finally settled during March and April 2011 in two different locations: Centro Agrario de Albaladejito (Cuenca, Spain) and Finca Casa Concha (Aranjuez, Spain). Seventy-five individual plants from selected populations were planted in three furrows (25 plants each) with 1 m between plants and amongst rows.

The aerial parts (inflorescences, leaves and stems) of R. officinalis were randomly collected from the experimental fields at the flowering period and dried at room temperature for 5 days prior to the hydrodistillation. Around 100 g of dry sample were hydrodistilled for 4 hours using a Clevenger type apparatus according to Boland et al. (1991). The solid residue from hydrodistillation was recovered and dried at room temperature.

Approximately 50 g of dried solid residue were thoroughly extracted with ethanol in a Soxhlet apparatus for 48 h. The crude extracts were subsequently filtered (Whatman grade 1) under vacuum and evaporated in a rotary evaporator to obtain the dried extracts that were kept in sealed tubes at 4 °C until analysis.

The content of total polyphenols was determined by the Folin-Ciocalteu method as described in Slinkard and Singleton (1977). Gallic acid was used as standard and the total polyphenol content was expressed as gram equivalents of gallic acid per 100 grams of dry extract. The total flavonoid content was determined using catechin as standard according to Zhishen et al. (1999) and expressed as gram equivalents of catechin per 100 grams of dry extract.

Solid residue extracts were analyzed using the LC-MS (ESI) technique in a Shimadzu LC/MS-2010A equipped with an LC-10ADvp binary pump, a DGU-14A degasser, a SIL-10ADvp auto sampler, and a SPD-M10Avp Photo Diode Array Detector. Dry extracts solubilized in methanol (10 mg/mL) were injected into a 250 mm ×4.6 mm, 5 μm particle size Discovery HS-C18 column (Supelco, Bellefonte, PA, USA) at a flow rate of 0.4 mL·min⁻¹. The general experimental procedures and conditions as described in Santana-Méridas et al. (2014) were followed.

The scavenging activity of the stable 1,1-diphenyl-2-picrylhydrazyl (DPPH) free radical was determined according to Braca et al. (2001). The method is based on the reduction of the stable free radical DPPH in the presence of a hydrogen-donating antioxidant, and the formation of the non-radical form DPPH-H as result of the reaction. This reduction can be monitored at 517 nm by measuring the bleaching of DPPH (violet) to DPPH-H (yellow). Distilled water and BHT were used as negative control and as standard antioxidant of the assay, respectively. The inhibition activity (%) of DPPH radicals was calculated as [(A₀–A₁)/A₀] ×100, where A₀ was the absorbance of the negative control, and A₁ was the absorbance of the extracts. The concentration of sample required to reduce 50% of DPPH radicals (IC₅₀) was calculated from a linear regression analysis.

The ferric reducing antioxidant power was determined according to the method of Oyaizu (1986) with modifications. This method determines the ability of a sample to reduce Fe³⁺ to Fe²⁺, which is recorded by measuring the formation of Perl′s Prussian blue at 700 nm. BHT was used as the standard antioxidant of the assay. The concentration of sample required to obtain 0.5 absorbance units (A.U.) was calculated from linear regression analysis.

The Fe²⁺-chelating activity was determined according to Carter (1971) with modifications. The method is based on the chelation of this metal ion with ferrozine to yield a red colored (562 nm) complex. The Cu²⁺-chelating activity was determined according to Saiga et al. (2003) based on the formation of the blue colored (632 nm) complex between Cu²⁺ and pyrocatechol violet (PV). In the presence of other chelating agents, the complex formation is disrupted and the color of the complexes decreases. The measurement of the rate of color reduction therefore allows for the estimation of the chelating activity. Distilled water and gallic acid were used as the negative control and as the standard metal chelator of the assay, respectively. The percentage of inhibition of complex formation was calculated as [(A₀–A₁)/A₀] ×100, where A₀ was the absorbance of the negative control, and A₁ was the absorbance of the extracts.

Extracts of R. officinalis were incubated with linoleic acid in order to evaluate their capacity to inhibit the oxidation of this unsaturated fatty acid. One milligram of dry extract was dissolved in 6 mL of distilled water (167 μg dry extract·mL⁻¹) in screwed tubes, thoroughly mixed with five microliters of linoleic acid, and incubated at 60 °C. Oxidation of linoleic acid was monitored at 24, 48 and 72 h of incubation by measuring the formation of oxidation products in the medium, including conjugated dienes, lipid hydroperoxides, and aldehydes and ketones (carbonyl groups). Distilled water and BHT were used as the negative control and as the antioxidant standard of the assay, respectively.

The formation of conjugated dienes was determined by measuring their absorbance at 233 nm and expressed as the increase in the initial absorbance. The determination of lipid hydroperoxides was based on their capacity of oxidation of Fe²⁺ to Fe³⁺, and the subsequent analysis of Fe²⁺ by ferrozine (Carter, 1971) as described in section 2.7. The results were expressed as percentage of initial absorbance at 562 nm, and calculated as (A_t/A_i) ×100, where A_i is the absorbance before incubation and A_t is the absorbance at 24, 48 or 72 h of incubation. Aldehydes and ketones were determined by the reaction of their carbonyl groups with 2,4-dinitrophenylhydracine (460 nm), and the results were expressed as the increase in the initial absorbance (Henick et al., 1954).

Spodoptera littoralis Boisd. (Lepidoptera: Noctuidae), Leptinotarsa decemlineata Say (Coleptera:Chrysomelidae) and the aphid Myzus persicae Sulzer (Hemiptera:Aphididae) were reared on an artificial diet (Poitut and Bues, 1970), of potato foliage (Solanum tuberosum) and bell pepper (Capsicum annum) plants, respectively; and maintained at 22±1 °C, >70% relative humidity, with a photoperiod of 16:8 h (L:D) in a growth chamber. The bioassays were conducted with newly emerged S. littoralis L6 larvae or L. decemlineata/M. persicae adults. Feeding or settling inhibition (%FI or %SI) were calculated as FI = [1−(T/C)] ×100, where T and C are the consumption of treated and control leaf disks, respectively, or as %SI = [1–10 (%T/%C)] ×100 where %C and %T are percent aphids settled on the control and treated leaf disks, respectively, as described in Burgueño-Tapia et al. (2008).

The experiments were conducted with Lactuca sativa cv Teresa (Fito, España) and Lolium perenne (Batlle) seeds, as described (Moitero et al., 2006). Briefly, a 2.5 cm diameter filter paper with 20 μL of the test compound (10 μg·μL⁻¹ for extracts and 5 μg·μL⁻¹ for pure compounds) were placed on 12-well plates (Falcon). Five hundred μL H₂O/well and 10/5 seeds (L. sativa/L. perenne presoaked in distilled water for 12 h) were added and the covered plates were placed in a plant growth chamber (25 °C, 70% RH, 16:8 L: D). Germination was monitored for 6 days and the root/leaf length was measured at the end of the experiment (25 plantlets randomly selected for each experiment and digitalized with the application ImageJ 1.43, http://imagej.nih.gov/ij/). Juglone (JU) (5 μg·μL⁻¹) was included as a positive control.

The antifeedant effects (FI/SI) were analyzed for significance by the non-parametric Wilcoxon signed-rank test. Extracts (100 μg·cm⁻²) and pure compounds (50 μg·cm⁻²) with an FI or SI >70% were tested in a dose–response experiment to calculate their relative potency (EC₅₀, the effective dose to give a 50% feeding reduction) which was determined from the linear regression analysis (%FI/SI on log dose). A Kruskall–Wallis test and an analysis of variance (ANOVA) were performed on germination and radicle/leaf length data, respectively. In all cases, the significance level was set at p<0.05. The relative concentration data (% values) was subjected to a principal component analysis. Relative concentration data, antioxidant and antifeedant effects were subjected to the Pearson correlation. All statistical analyses were carried out using the IBM^® SPSS^® Statistics ver. 22 (IBM corp.^©, 2013) package.

The extraction of the solid residues of R. officinalis by a Soxhlet system yielded a high recovery of ethanol soluble compounds that ranged from 20.4 to 26.2 g extract·100 g⁻¹ dry matter.

Tables 2 and 3 show the chemical profile and the composition data (area percent) of the residues extracted from ten R. officinalis populations cultivated in Cuenca (Castilla-La Mancha, Spain) and Aranjuez (Madrid, Spain). The chemical composition of the rosemary solid residue was similar to that described for other R. officinalis extracts (Herrero et al., 2010; Borras-Linares et al., 2011; Navarrete et al., 2011; Zhang et al., 2012; Kontogianni et al., 2013; Santana-Méridas et al., 2014). Rosmarinic acid (2), genkwanin (13), carnosol (16), epirosmanol ethyl ether/rosmarinic acid methyl ether (17) and carnosic acid (21) were the most abundant compounds tentatively identified and represented between 28% and 48% of the total area of peaks in all populations (Table 3). Significant differences in the area percent of compounds were found among populations and notably between cultivation places (Table 3, Figure 1).

The composition data (area percent) of the 20 extracts of R. officinalis populations from Table 3 were submitted to principal component analysis (Figure 2). As a result of the analysis, three main components explaining 80.11% of the original variability of the data pool were obtained. The first component explained 41.44% of the variability and was highly correlated both negatively with compounds 1, 2, 3, 7; and positively with compounds 8, 10, 13, 15, 17 (numbers according to Table 2). The second component was negatively correlated with carnosol (16) and carnosic acid (21); and explained 23.41% of the variability. The third component explained 15.56% of the variability and was correlated with the rest of compounds.

Plots using principal components from first to second as axes (64.85%) did not show groupings of populations according to their original collecting places, although clear trends were observed according the growing area (Figure 2). The first principal component showed the highest score values for P6-P9 cultivated in Cuenca which shared low relative contents of rosmarinic acid and high contents of genkwanin. The values for the second component were higher for populations P4, P7 and P10 from Aranjuez with a relatively high content in rosmarinic acid and low-medium content in carnosol. P3 from Cuenca and P8 and P9 cultivated in Aranjuez ranked the highest in carnosic acid contents and appeared separated from the rest of populations in the plot.

Our results support the hypothesis that the relative amounts of polyphenols depend on the specific environmental conditions in which the plant is developed (Josuttis et al., 2013), and could be considered a common adaptation of plants to adverse environmental conditions (Lattanzio et al., 2012). This information can be used to establish classifications according to the geographical conditions of the growing regions (Carrillo et al., 2014).

A great variability among populations and locations was observed in the total polyphenol content in the residues of R. officinalis, which ranged from 4.4 g·100 g⁻¹ of dry extract in P6 from Cuenca to 12.6 g·100 g⁻¹ in P8 from Aranjuez (Table 4). These data were consistent with the recovery of polyphenols from other aromatic plants (Sánchez-Vioque et al., 2013) although an overall minor content in total polyphenols was observed in the residues from the populations cultivated in Cuenca in comparison with those from Aranjuez. The total content in polyphenols and flavonoids were well correlated (r=0.78, p<0.05) (Table 5) but flavonoids showed a variability even higher than that observed for the content in polyphenols, from 0.6 (P6, Cuenca) to 6.2 g·100 g⁻¹ of dry extract (P7, Aranjuez) (Table 4). The content of total flavonoids was superior in populations cultivated in Aranjuez in comparison with those from Cuenca, up to more than 7.5 times higher in the case of P7 and P10.

The large differences in the contents in polyphenols and flavonoids resulted in very different DPPH radical-scavenging activities among populations, whose IC₅₀ values ranged from of 21.6 (P8, Aranjuez) to 80.5 μg·mL⁻¹ (P6, Cuenca) (Table 4). The IC₅₀ of BHT was 4.9 μg·mL⁻¹. The correlation between total polyphenols and DPPH activities was very high (r=−0.85, p<0.05) (Table 5). The agreement between the results of DPPH radical-scavenging activity and total polyphenols was expected, and it is consistent with the fact that both determinations are based on a chemical reduction: either Folin-Ciocalteu′s reagent in the case of determination of the total polyphenols or DPPH radicals. Unlike total polyphenols, the method used for the determination of total flavonoids is not based on a reduction but on a complexation of Al (III) by flavonoids. Consequently, DPPH activities were less correlated with flavonoids (r=−0.53, p<0.05) than with total polyphenols (Table 5). Obviously, the DPPH activities depend on the total content of flavonoids but also on the flavonoid composition of the extract since each individual flavonoid has a specific antioxidant activity against DPPH radicals. For example, the extracts of P3 from Cuenca and of P6 from Aranjuez had similar flavonoid contents, 1.5 and 1.7 g·100 g⁻¹ dry extract, respectively, but P3 (IC₅₀=24.5 μg·mL⁻¹) was a much better reducer of DPPH radicals than P6 (IC₅₀=56.3 μg·mL⁻¹) (Table 4). Similarly, the extracts of P1 and P7 from Cuenca (both with 0.8 g flavonoids/100 g dry extract) showed DPPH scavenging activities (IC₅₀) of 37.0 and 56.6 μg·mL⁻¹, respectively. The polyphenol compositions of these populations showed large differences: P1 and P3 from Cuenca were mainly characterized by a high content of carnosol, 17.9 and 24.4% of the total peak area, respectively, whereas the content was much lower in the P6 from Aranjuez (8.6%) and marginal in the P7 from Cuenca (0.6%) (Table 3). Moreover, the compounds other than flavonoids endowed with reducing capacity may be present in the extracts and increase the DPPH activity.

The data of the reduction in Fe (III) by R. officinalis extracts were comparable to those of DPPH activities, showing a very high correlation with total polyphenols (r=−0.86, p<0.05), and a moderate correlation with total flavonoids (r=−0.70, p<0.05) (Table 5). The concentration of extract to yield 0.5 A.U. ranged from 21.0 (P8, Aranjuez) to 81.3 μg·mL⁻¹ (P6, Cuenca) (Table 4). The standard BHT attained such absorbance at 8.0 μg·mL⁻¹ of concentration. As in the previous antioxidant test, the FRAP of extracts was very dependent on the location and on the population; those cultivated in Aranjuez were more antioxidant than their equivalent populations from Cuenca.

The Fe²⁺-chelating activities (IC₅₀) of R. officinalis extracts also depended on the population and location, and ranged from 238 (P7, Aranjuez) to 410 μg·mL⁻¹ (P9, Aranjuez) (Table 4). In terms of comparison, the IC₅₀ of the standard gallic acid was 52.6 μg·mL⁻¹. Nevertheless, the Fe²⁺-chelating activities of R. officinalis extracts were not significantly correlated with the total contents of either polyphenols or flavonoids (Table 5). A similar low correlation has been observed in several aromatic plant residues and attributed to the fact that only polyphenols having neighboring hydroxyl and/or keto groups in their chemical structures are able to chelate divalent metals, as well as to the possible presence of other compounds in the extracts that could act as chelators (Sánchez-Vioque et al., 2013; Santana-Méridas et al., 2014). None of the polyphenols showed a significant and negative correlation with Fe²⁺-chelating activities (IC₅₀), and only carnosol showed a certain significant but positive correlation (r=0.46, p<0.05). Nevertheless, some observations with respect to the polyphenol composition and the chelation of Fe²⁺ can be made. Thus, the Fe²⁺-chelating activities (IC₅₀) of the extracts of P6 from Cuenca and of P6 from Aranjuez were identical (313 μg·mL⁻¹) despite the fact that the P6 from Cuenca had a flavonoid content of around 3 times lower than that of P6 from Aranjuez (Table 4). Both populations showed a very different polyphenol composition: the P6 from Cuenca was characterized by a high content in genkwanin (27.5% of total peak area) while the major polyphenols in the P6 from Aranjuez were carnosol (8.6%), rosmarinic acid (7.4%), epirosmanol ethyl ether/rosmarinic acid methyl ether (6.6%) and rosmanol/epirosmanol (5.7%) (Table 3). Likewise, the extract of P9 from Cuenca (1.3 g flavonoids·100 g⁻¹ dry extract) was a better Fe²⁺-chelator (IC₅₀=281 μg·mL⁻¹) than P9 from Aranjuez (3.8 g flavonoids·100 g⁻¹ dry extract; IC₅₀=410 μg·mL⁻¹). The polyphenol profile of P9 from Cuenca showed a predominance of genkwanin (18.4%), whereas rosmarinic acid (15.2%), carnosic acid (8.3%) and carnosol (8.2%) were the major polyphenols tentatively identified in P9 from Aranjuez. Extracts from P7 (19.3% genkwanin) and P9 (22.9% genkwanin) from Cuenca also showed good chelating activities in relation to their modest contents in total flavonoids. This suggests that a high content of genkwanin improves the Fe²⁺-chelating activity, which is consistent with the chelating activity of the 4-keto and 5-hydroxyl groups present in the structure of this flavonoid (Van Acker et al., 1996).

The Cu²⁺-chelating activities (IC₅₀) of R. officinalis extracts ranged from 271 (P8, Aranjuez) to 695 μg/mL (P10, Cuenca), whereas the IC₅₀ of gallic acid was 24.7 μg·mL⁻¹. In contrast to the chelation of iron, the Cu²⁺-chelating activities were highly correlated with the total content in polyphenols (r=−0.86, p<0.05), and moderately correlated with the total content in flavonoids (r=−0.70, p<0.05) (Table 5). In addition to these good correlations, some differences among populations in the Cu²⁺-chelating activities could be in part attributed to their polyphenol composition. Carnosic acid (r=−0.52, p<0.05) showed the best correlation with Cu²⁺-chelating activity, and its predominance in some extracts seemed to promote this activity. Thus, the extracts from populations P2 and P3 from Cuenca had the same flavonoid content, 1.5 g·100 g⁻¹ dry extract, but P3 (IC₅₀=371 μg·mL⁻¹) was a better copper chelator than P2 (IC₅₀=529 μg·m⁻¹) probably due to the higher content in carnosic acid in P3 (10.3%) in comparison with P2 (0.5%) (Tables 3 and 4). P8 and P9 from Aranjuez also showed similar flavonoid contents but P8 (IC₅₀=271 μg·mL⁻¹) was a much better chelator than P9 (IC₅₀=478 μg·mL⁻¹). The polyphenol composition of P8 was characterized by a higher content in carnosic acid as compared with P9 (17.0% versus 8.3%) These comparisons suggest that for equal contents of flavonoids, carnosic acid could enhance the Cu²⁺-chelating activity in comparison with other flavonoids of R. officinalis extracts.

The oxidation of unsaturated fatty acids like linoleic acid can be monitored by the determination of the numerous oxidation compounds generated. The presence of antioxidant compounds in the medium delays the generation of oxidation compounds whose determination therefore allows for an indirect evaluation of their antioxidant capacity. Conjugated dienes are compounds formed at the early stages of the oxidation (primary oxidation products) and their determination constitutes an usual method of detecting the initiation of lipid oxidation. After 24 h of incubation, all the R. officinalis extracts reduced the generation of conjugated dienes with respect to the control, and especially effective were the extracts from populations P2 and P5 from Cuenca, and P5, P8 and P10 from Aranjuez. (Figure 3). The formation of conjugated dienes logically increased with the incubation time, but even after 72 h of incubation, the extracts from populations P5, P8 and P9 from Cuenca, and P5, P8 and P10 from Aranjuez still showed a remarkable reduction in the content of conjugated dienes as compared with the control.

Lipid hydroperoxides are also primary oxidation products whose generation was clearly inhibited by R. officinalis extracts (Figure 4). After 24 h of incubation, most of the rosemary extracts reduced the generation of hydroperoxides with respect to the control. The extracts from populations P2, P3, and P5 cultivated in Cuenca, and P8 and P10 from Aranjuez were the most effective in delaying the formation of lipid hydroperoxides throughout incubation. All these populations, except for P10 from Aranjuez, were characterized by a high content in carnosic acid and carnosol (16–35% total peak area). In fact, carnosic acid was the only polyphenol significantly correlated (r=0.61, p<0.05) with the inhibition of lipid hydroperoxide generation. As both lipid hydroperoxides and conjugated dienes are primary oxidation products, their generation was well correlated (r=−0.70, p<0.05) (Table 5).

Aldehydes and ketones are compounds generated when oxidation has reached an advanced stage (secondary oxidation products) and the development of rancidity is clearly perceptible. Their formation was stopped or drastically reduced by all the R. officinalis extracts after 24 h of incubation, and after 72 h the detection of carbonyl groups was only significant in incubations with extracts from populations P1, P3, P4 and P6 from Aranjuez (Figure 5).

The effectiveness of R. officinalis extracts to delay the oxidation of linoleic acid cannot be exclusively related with their total content of polyphenols and other compounds presumably present in the extracts like tocopherols, which could help to prevent such oxidation as deduced from the data on the formation of oxidation products. Thus, whereas some extracts with a high contents in polyphenols, like P2 and P3 from Cuenca, or P5, P8 and P10 from Aranjuez, extensively reduced the generation of oxidation products, others with similar contents in polyphenols, like P4 and P7 from Aranjuez, did not. Conversely, the extract from P5 from Cuenca was very effective as an antioxidant despite the fact that its content in polyphenols was lower.

Table 6 shows the antifeedant effect of the extracts from 10 populations of R. officinalis cultivated in Cuenca and Aranjuez. According to their feeding ecologies, L. decemlineata (olyphagous) was more sensitive to the extracts of R. officinalis populations when compared with S. littoralis and M. persicae (polyphagous). Based on the EC₅₀ values, the extracts from most of the populations cultivated in Aranjuez (P2-P5 and P7-P10) were strong antifeedants against L. decemlineata, whereas only the extracts from populations P2, P3 and P9, P10, grown in Cuenca, showed strong activity on this pest. The extracts from P1 and P9 cultivated in Cuenca showed antifeedant effects on S. littoralis, while the rest of the samples were inactive. When tested against M. persicae, only the extracts from P2 (Cuenca and Aranjuez), P4 (Cuenca) and P7 (Cuenca) had antifeedant effects at the maximum concentration tested (100 μg·cm⁻²). Among the major pure compounds present in the R. officinalis residue, rosmarinic and carnosic acids acted as strong antifeedants against L. decemlineata, while genkwanin was inactive. None of these compounds had activity against S. littoralis (Table 6).

Previous reports showed the strong antifeedant effects afainst L. decemlineata and moderate against S. littoralis of residues from rosemary (Santana-Méridas et al., 2014) and other economically important aromatic plants (Santana-Méridas et al., 2012). However, to our knowledge, this is the first report on a population-based survey including the comparative chemistry and biological properties of the solid residues from the distillation of R. officinalis.

The antifeedant effects observed for the extracts of R. officinalis populations were not accounted for by their chemical composition. A Pearson correlation analysis was performed to ascertain the relationship between the antifeedant activity and the major individual compounds detected in the extracts and no significant correlation was observed (data not shown). However, we found a significant positive correlation between the antifeedant activity against L. decemlineata (%FI) and the total content in flavonoids (r=0.55, p=0.01) or the total content in polyphenols (r=0.52, p=0.03), suggesting that the antifeedant activity of rosemary extracts could be explained by the additive or synergistic effects of their individual phenolic components. Similar results have been reported by Julio et al. (2014) when comparing the conventional and supercritical extracts from two Lavandula luisieri populations. In the same way, Rojht et al. (2012) reported the potential repellent effect of a non-toxic ethanol extract from the aerial part of R. officinalis against adults of Acanthoscelides obtectus and L. decemlineata, and propose a synergistic effect among the volatile compounds as responsible for the activity. Interestingly, copper chelating activity was correlated with the antifeedant activity of the extracts on L. decemlineata (r=0.49, p=0.03), suggesting a possible impairment of the metal-based enzymes of insects by metal chelators (Dittmer and Kanost, 2012).

The involvement of plant phenolics in the plant mechanism against insects and its role on the feeding behavior of herbivores is well known through the ages, and has been recently reviewed (Lattanzio et al., 2012). Specific polyphenols, either individually or in combination, have been associated with defense against insect herbivores in Quercus oleoides (Moctezuma et al., 2014). Fornoff and Gross (2014) reported the increase in polyphenols as one of the induced resistance and defensive traits of Myriophyllum spicatum to prevent herbivore damage. In the same way, the feeding activity of some chewing insects has been correlated with the concentration of polyphenols and flavonoids in plant tissues (Lattanzio et al., 2012).

None of the samples tested had a phytotoxic effect on the germination and radicle growth of L. sativa. On the contrary, most of the extracts and especially those from populations cultivated in Cuenca had a strong stimulating effect on L. sativa radicle length (Table 7). The effects observed in the case of L. perenne were more controversial. The rosemary residues from P1 cultivated in Cuenca and Aranjuez showed the strongest inhibition of L. perenne germination, effective up to 144 h. The root and leaf growth of L. perenne were strongly affected by P2 grown in Cuenca (60 and 43%, respectively) and Aranjuez (88 and 52%, respectively) while the rest of the samples had stimulating effects, these being more evident in populations cultivated in Aranjuez (Table 8).

The allelopathic potential of phenolic compounds is reviewed by Li et al. (2010), and some phenolic acids present in rosemary, such as rosmarinic and caffeic acids have shown allelopathic effects on Arabidopsis seeds (Jian et al., 2010). More recently, López-Iglesias et al. (2014) proposed that enhanced plant growth or, on the contrary, plant growth inhibition, could be the result of a positive or negative balance between nutrient and polyphenol concentration. In a previous work, Santana-Méridas et al. (2014) reported a stimulating effect of rosemary residues on L. sativa and a moderate phytotoxic effect of L. perenne. Rojht et al. (2012) did not find significant differences in germination between the control and treated beans with rosemary ethanol extract, whereas Guimarães-Araújo et al. (2013) reported heterogeneous effects on the growth of hypocotyls and radicle of the seeds of lettuce and onion after the application of ethanol extract and fractions obtained from R. officinalis.