2.1. Claisen Rearrangement
The terpenoids are the largest and important group of fragrance compounds which may be obtained by Claisen rearrangements. Hence, great many literature references are available in this field. A key intermediate in the synthesis of many important synthetic fragrances (e.g. citral, linalool, geraniol, etc.) is 2-methyl-2-hepten-6-one (3). It can be obtained via a synthesis based on the Claisen rearrangement  (Scheme 4).
An analogous reaction, which involves dehydrolinalool, can yield pseudoionone (4), a key intermediate in the synthesis of farnesol and nerolidol  (Scheme 5).
The Claisen rearrangement can also lead to a number of other alcohols (5) and aldehydes (6-9) with isoprenoid or related skeletons [16,17,18,19,20] (Scheme 6).
Based on readily available myrcene, a complete synthesis was developed for β-sinesal (10), an important olfactory component of sweet orange and Amarylis oil [21, 22] (Scheme 7).
In an analogous reaction, geraniol can be converted into aldehyde (11), which has an odour reminis-cent of coriander  ( Scheme 8 ).
Condensation of linalool with isobutyraldehyde gives 2,2,5,9-tetramethyldeca-diene-4,8-al (12) with a floral, green odour [24,25] (Scheme 9).
The Claisen rearrangement can be also applied in case of alicyclic systems. Dehydrolinalool undergoes ring closure with ease in an “ene”-type reaction to yield the cyclopentane derivative (13), which can be utilised in the synthesis of ester (14), which was marketed as “Cyclopentenyl propionate musk” (RIFM) or Cyclomusk® (BASF), and which has a very interesting fruity, musky odor with a sandalwood note  (Scheme 10).
From alcohol (13) one can obtain spiro[4,5]decenone (15), which is one of the components of European Acorus Calamus L. oil  (Scheme 11).
Pinocarveol (16) is the parent substance for the production of “Pinolacetaldehyde” (IFF) (17), which possesses a floral odour  (Scheme 12), as well as its dimethyl derivative (18), with a floral, woody odor  (Scheme 13).
From DCP-ketone (19) it is possible to obtain 4-(tricyclo-[5,2,1,02,6]-decyl-8-ydiene)-butanal (20) (Dupical®, Quest Int.) with intensive floral, lily-of-the-valley odour  ( Scheme 14 ).
2.3. Essential Oil Composition
In leaves and stems 65 and 43 essential oil compounds could be identified, respectively. The detailed listing is available with the Supporting Information Tables S2 and S3. Bar charts showing the main oil compounds of the accessions are given in Figures S1 and S2 in the Supporting Information. To study the complex oil patterns in the leaves a multivariate approach has been attempted using hierarchical cluster analysis (HCA) and principal component analysis (PCA) with selected essential oil compounds as variables and the accessions as cases. In PCA also rosmarinic acid, total phenolics and antioxidant activity of both cuts have been included as variables.
The dendrogram from HCA (Figure 1) presents two distinct main clusters, the left containing all accessions of MOFF while the right grouped the samples MALT. Plants of MOFF had the monoterpene aldehydes geranial (=E-citral, citral a) and neral (=Z-citral, citral b) with citrus-like aroma as major oil compounds, while in MALT the sesquiterpenes β-caryophyllene, caryophyllene oxide and germacrene D were characteristic compounds. The oils of the MALT leaves contained also appreciable amounts of the monoterpenes α-pinene, β-pinene and sabinene. Starting with 21 variables, the PCA calculated four components having eigenvalues greater than one and representing together 83.9% of total variance. The first axis accounted for 43.5% and the second for 21.5% of the variance (Figure 2 and Figure 3). The scoring plot of the first two components could also clearly differentiate between the two subspecies (Figure 2). MOFF samples formed a group with negative factor 1 scores, while all MALT accessions had positive factor 1 scores.
The formation of subclusters in HCA (Figure 1) and the division of the scores in PCA (Figure 2) show for MOFF samples a lower variability than for MALT samples. In addition, in both subspecies, the oil compositions changed noticeably from the first to the second cut. In PCA first cut samples were mostly associated with a positives component 2 score while the second cut samples had rather a respective negative score (Figure 2).
In the dendrogram the first and the second cut samples of MOFF were classed into two distinct subclusters. The plants from the first cut had geranial as main compound, varying amounts of neral and citronellal and around 9% caryophyllene oxide in their oils. To compare, an oil from cultivated blooming plants of the Balkans with 23.4% geranial, 16.5% neral and 13.7% citronellal showed a similar composition as the present first cut oils  and similar oils were also reported from blooming Slovakian plants . However, the oils from the present study second cut appeared more homogeneous: in each of 15 accessions geranial and neral accounted together for 80–90% of the oil and the variability between the accessions of these two compounds with coefficients of variation (CV%) of 5.1% and 4.2%, respectively, was remarkably low. A similar result was obtained by Adzet et al. (1992)  where in 25 biotypes of lemon balm cultivated in the Spanish Ebro region the sum geranial + neral made up 93–96% of the respective oils with CV% less than 7%. In contrast to our results, in this latter study the ratio of these two compounds varied little between July/August and November . Accession M10 had already in the first cut oil very high geranial and neral levels like the oils from the second cut and was therefore in the subcluster of the second cut samples. In comparison with other MOFF samples the first cut from M8 had the lowest geranial and neral and the highest β-caryophyllene and germacrene D percentages and had in consequence its distinct position in the dendrogram and on the score plot.
With exception of the accessions M17, M21 and M22, also MALT samples of the first and the second cut could well be separated in distinct sublclusters (Figure 1).
To summarize, the individual oil compounds loadings on the principal components are represented in Figure 3. The loading on component 1 shows as mentioned above a strong differentiation of the citrus-like aroma monoterpenes from both cuts (citronellal, neral, geranial) typical for MOFF oils from the sesquiterpenes that are characteristic for MALT. Here these monoterpenes were strongly associated with negative values while the mentioned sesquiterpenes loaded with positive values. Further monoterpenes α-pinene, β-pinene and sabinene which were conspicuous in MALT leaves of the second cut had positive loadings on component 1 and negative loadings on component 2. These three monoterpenes, almost absent in MOFF, were present in low amounts in several first cut MALT samples. Of the latter, only accession M21 had at this time 12.6% sabinene, 10.8% β-pinene and 4.5% α-pinene. However, in the second cut the accessions M17, M21 and M22 were rich in these three compounds that together made up 55–60% of the respective oils. In consequence, they formed a subgroup with negative factor 2 scores in PCA and a distinct subcluster in HCA.
Hexadecanoic acid, present mainly in MALT first cut samples (up to 5.9%) and having positive component 2 scores, loaded accordingly positively on principal component 2. Furthermore epi-caryophyllene could be detected in MALT plants of the second cut, while hexadecanoic acid was not found in these samples. A subcluster in HCA with five accessions (M19, M28, M18, M12 and M20) had the highest caryophyllene oxide contents in the oils of the first cut.
In the present MOFF leaf oils geranial and neral were positively correlated (R = 0.976, p < 0.001) but negatively correlated with citronellal (R = −0.920 and −0.935, p < 0.001). Additionally, geranial and neral were negatively correlated with caryophyllene oxide, germacrene D and β-caryophyllene while caryophyllene oxide showed a positive correlation with citronellal and germacrene D. In MALT leaf oils α-pinene, β-pinene and sabinene were strongly correlated (R = 0.993–0.998, p < 0.001) while caryophylene oxide was negatively correlated with these three monoterpenes (R = −0.728 to −0.756, p < 0.001) and with germacrene D (R = −0.521, p = 0.008) and β-caryophyllenene (R = −0.439, p = 0.028). Small amounts of the sesquiterpenes α-copaene, β-cubebene and cadinol occured in samples from both subspecies and were usually higher in MALT.
Stem essential oils were analyzed from five selected accessions of each subspecies (Supporting Information, Figures S3 and S4). Like in the leaves, the stem oils of the first cut MOFF plants had geranial, neral and citronellal in varying proportions but not more than 35% together, so these oils had a high proportion of sesquiterpenes such as β-caryophyllene, caryophyllene oxide and α-copaene. The latter reached 18% of the oil in accession M10. In the second cut, the stems had as the leaves geranial and neral as main oil compounds, ranging together between 40% and 85% of the respective oils. In contrast, germacrene D that played only a marginal role in MOFF stems was a major compound in MALT stem oils. In the oil from the first cut of accession M21 it was the only detectable compound. Other sesquiterpenes present were caryophyllene oxide and β-caryophyllene and in accessions M12 and M20 also α-copaene. Stems of the first cut had more caryophyllene oxide and less β-caryophyllene than the respective stems of the second cut. The accessions M17, M21 and M22, having high proportions of β-pinene, sabinene and α-pinene in the leaves, had these compounds also in their stem oils from the second cut. By this way the stem oils reported here differed clearly from a stem oil of Iranian plants where the main constituents were: n-hexadecanoic acid (47.4%), (Z,Z)-9,12-octadecadienoic acid (14.9%), dodecanoic acid (4.6%), β-caryophyllene (4.2%) and geraniol (2.2%) .
In sum, lemon balm essential oils show a great variability and plasticity. Literature references in most cases do not differentiate into the two subspecies when referring to essential oil composition but a great variation is documented. There are reports of oils having geranial and neral as main compounds. Leaf oil from plants grown in Algeria, being composed of 44.2% geranial, 30.2% neral, 6.3% citronellal and less than 4% sesquiterpenes, was similar to the present leaf oils from the second cut . Further citrus-like aroma monoterpenes may also play a major role in the oils. Several Turkish lemon balm oils had citronellol (37–44%) as main compound . An Iranian lemon balm flower essential oil displayed trans-carveol (28.9%), citronellol (25.2%), δ-3-carene (5.3%), citronellal (4.9%) and geranial (2.2%) as main compounds . Popova et al.  described a Bulgarian Melissa oil containing 18.5% citronellal, 15.2%, geraniol, 9.5% citronellol, 7.2% geranyl acetate and 5.9% geranial.
Greek lemon balm leaf oils with α-pinene, β-pinene, sabinene, β-caryophyllene, caryophyllene oxide and germacrene D as reported by Basta et al.  were therefore presumed to derive from subsp. altissima. A leaf oil from Jordan having as main compound caryophyllene oxide (43.6%) reportedly also contained considerable amounts of γ-muurolene (28.8%) . In this case a confusion of this latter compound with germacrene D appears probable, as both components have similar retention behavior and fragmentation patterns in GC/MS. In some oils these citrus aroma monoterpenes occur along with comparable amounts of sesquiterpenes as in the case of a Moroccon leaf oil with 14.4% citronellal, 10.2% geranyl acetate, 5.2% nerylacetate, 11.0% caryophyllene oxide and 8.2% β-caryophyllene  but also in various lemon balm strains from Poland .
Besides genetic factors such as the existence of the two subspecies, the basis of this observed high variability remains complex: There is the experience of the present study that second MOFF leaf oils had a highly uniform composition with little varying neral and geranial contents in contrast to the oils from the first cut of the same plants. Similarly, plants cultivated in Poland had in two consecutive years nearly the same geranial (45%) and neral (33%) contents in their oils while the citronellal content varied (0.4–8.7%) . A further experiment from Poland reported higher neral and geranial levels under higher insolation . In pot experiments, soil water content hardly influenced essential oil composition . On the other hand, various accessions cultivated on two different sites in Turkey clearly differed in their oil composition .