**2.2.1 Spectrophotometric analysis**

Seven hundred absorbance values were utilised as quantitative variables for population analysis. Samples were collected in different provinces and seasons, mainly taking into account the quantitative variability of secondary metabolites during the year (Table 1). The principal component analysis showed that the first nine components explain almost 98.81% of the total variability. The second (PC2) and the third (PC3) principal components gathered relevant information for classifying species. Figure 4, shows a two dimension plot of PC2 vs. PC3 using all the variables. Samples could be classified in five large groups containing the species *B. crispa*, *B. microcephala*, *B. phyteumoides*, *B. triangularis* and *B. trimera*. PC2 clearly separates *B. microcephala* and *B. trimera* populations from *B. crispa* and *B. phyteumoides* populations. Moreover, samples corresponding to *B. triangularis* species were separated from those belonging to *B. microcephala*, *B. trimera*, *B. crispa* and *B. phyteumoides* populations by PC2. While PC3 separates *B. microcephala* samples from the rest of the species, it also separates *B. crispa* samples from *B. phyteumoides*, *B. trimera* and *B. triangularis* and between samples of the two latter species. However, PC3 did not completely distinguish *B. phyteumoides* from *B. trimera* and *B. triangularis* samples.

Fig. 4. Two dimensional model of PC2 *vs.* PC3 (15.63 and 7.80 %, respectively) derived from PCA of 700 quantitative variables of 53 *Baccharis* populations. Ba, *B. articulata*; Bc, *B. crispa*; Bg, *B. gaudichaudiana*; Bm, *B. microcephala*; Bp, *B. penningtonii*; Bphy, *B. phyteumoides*; Bs, *B. sagittalis*; Btr, *B. triangularis*; Bt, *B. trimera*; BA, Buenos Aires; CHU, Chubut; CO, Córdoba; COR, Corrientes; ER, Entre Ríos; FO, Formosa; LP, La Pampa; MI, Misiones; RN, Río Negro; SF, Santa Fe; SL, San Luis.

Quality Control of Herbal Medicines with Spectrophotometry and Chemometric Techniques

Fig. 6. Representative UV-Visible spectra of *B. articulata* (Ba), *B. gaudichaudiana* (Bg),

Fig. 7. A-B, PC loadings graph. **A**, PC2 (explains 15.63 % of the total data variance); **B**, PC3 (explains 7.80 % of the total data variance). C, Representative UV-Visible spectra of *B. crispa* 

Figure 7 B (PC3 eigenvalues) shows two regions, one with positive values (270 to 350 nm) and one with negative values (350-420 nm). From Figure 7 C it is possible to verify that in the first region the analytical signals for *B. microcephala* are more intense than those of *B. trimera*, *B. crispa*, *B. phyteumoides* and *B. triangularis*, and that the *B. microcephala* samples have positive PC3 scores. The rest of the species have negative PC3 scores. In Figure 7 B, the region with negative values (270 to 350 nm) matches the more intense analytical signals for

*(Bc)*, *B. microcephala (Bm)*, *B. phyteumoides (Bphy)*, *B. triangularis (Btr)*, *B. trimera (Bt)*.

*B. crispa*, *B. phyteumoides* and *B. triangularis* in Figure 7 C.

*B. sagittalis* (Bs).

– Application to *Baccharis* L. Species Belonging to Sect – Caulopterae DC. (Asteraceae) 409

Figure 5, shows a two dimensional plot of PC7 vs. PC9 using all the variables. Here, the seventh (PC7) and ninth (PC9) principal components enable the *B. penningtonii* samples to be separated from the rest of the species.

Fig. 5. Two dimensional model of PC7 *vs.* PC9 (0.62 and 0.36 %, respectively) derived from PCA of 700 quantitative variables of 53 *Baccharis* populations. Ba, *B. articulata*; Bc, *B. crispa*; Bg, *B. gaudichaudiana*; Bm, *B. microcephala*; Bp, *B. penningtonii*; Bphy, *B. phyteumoides*; Bs, *B. sagittalis*; Btr, *B. triangularis*; Bt, *B. trimera*; BA, Buenos Aires; CHU, Chubut; CO, Córdoba; COR, Corrientes; ER, Entre Ríos; FO, Formosa; LP, La Pampa; MI, Misiones; RN, Río Negro; SF, Santa Fe; SL, San Luis.

Spectrophotometric PCA data allowed the distinction of six out of the nine species examined in this study. However the three species *B. articulata*, *B. gaudichaudiana* and *B. sagittalis* could not be separated which is shown in the figure 6, where it is seen that the average spectra of the species are very similar.

As can be observed in Figure 4, the *B. microcephala* and *B. trimera* samples have PC2 scores with opposite signs to those of the *B. crispa*, *B. phyteumoides* and *B. triangularis* samples. This contrast can be explained with the help of Figure 7 A; C. Figure 7 A shows a graph of the loading values (eigenvalues) on PC2 vs. . Positive values are situated in a region between 200 and 230 nm. It is possible to verify from Figure 7 C that the analytical signals for *B. crispa*, *B. phyteumoides* and *B. triangularis* in this region are more intense than those of *B. microcephala* and *B. trimera*, and that the *B. crispa*, *B. phyteumoides* and *B. triangularis* samples have positive scores.

Figure 7 A also shows that three regions present negative values between: 225 and 275, 325 and 375 and 375 and 450 nm. In Figure 7 C, one can verify that in these intervals the most intense analytical signals belong to samples of the *B. microcephala* and *B. trimera* species. For this reason the *B. microcephala* and *B. trimera* species samples have negative PC2 scores.

Figure 5, shows a two dimensional plot of PC7 vs. PC9 using all the variables. Here, the seventh (PC7) and ninth (PC9) principal components enable the *B. penningtonii* samples to

Fig. 5. Two dimensional model of PC7 *vs.* PC9 (0.62 and 0.36 %, respectively) derived from PCA of 700 quantitative variables of 53 *Baccharis* populations. Ba, *B. articulata*; Bc, *B. crispa*; Bg, *B. gaudichaudiana*; Bm, *B. microcephala*; Bp, *B. penningtonii*; Bphy, *B. phyteumoides*; Bs, *B. sagittalis*; Btr, *B. triangularis*; Bt, *B. trimera*; BA, Buenos Aires; CHU, Chubut; CO, Córdoba; COR, Corrientes; ER, Entre Ríos; FO, Formosa; LP, La Pampa; MI, Misiones; RN, Río Negro;

Spectrophotometric PCA data allowed the distinction of six out of the nine species examined in this study. However the three species *B. articulata*, *B. gaudichaudiana* and *B. sagittalis* could not be separated which is shown in the figure 6, where it is seen that the average spectra of

As can be observed in Figure 4, the *B. microcephala* and *B. trimera* samples have PC2 scores with opposite signs to those of the *B. crispa*, *B. phyteumoides* and *B. triangularis* samples. This contrast can be explained with the help of Figure 7 A; C. Figure 7 A shows a graph of the loading values (eigenvalues) on PC2 vs. . Positive values are situated in a region between 200 and 230 nm. It is possible to verify from Figure 7 C that the analytical signals for *B. crispa*, *B. phyteumoides* and *B. triangularis* in this region are more intense than those of *B. microcephala* and *B. trimera*, and that the *B. crispa*, *B. phyteumoides* and *B. triangularis* samples

Figure 7 A also shows that three regions present negative values between: 225 and 275, 325 and 375 and 375 and 450 nm. In Figure 7 C, one can verify that in these intervals the most intense analytical signals belong to samples of the *B. microcephala* and *B. trimera* species. For this reason the *B. microcephala* and *B. trimera* species samples have negative PC2 scores.

be separated from the rest of the species.

SF, Santa Fe; SL, San Luis.

the species are very similar.

have positive scores.

Fig. 6. Representative UV-Visible spectra of *B. articulata* (Ba), *B. gaudichaudiana* (Bg), *B. sagittalis* (Bs).

Fig. 7. A-B, PC loadings graph. **A**, PC2 (explains 15.63 % of the total data variance); **B**, PC3 (explains 7.80 % of the total data variance). C, Representative UV-Visible spectra of *B. crispa (Bc)*, *B. microcephala (Bm)*, *B. phyteumoides (Bphy)*, *B. triangularis (Btr)*, *B. trimera (Bt)*.

Figure 7 B (PC3 eigenvalues) shows two regions, one with positive values (270 to 350 nm) and one with negative values (350-420 nm). From Figure 7 C it is possible to verify that in the first region the analytical signals for *B. microcephala* are more intense than those of *B. trimera*, *B. crispa*, *B. phyteumoides* and *B. triangularis*, and that the *B. microcephala* samples have positive PC3 scores. The rest of the species have negative PC3 scores. In Figure 7 B, the region with negative values (270 to 350 nm) matches the more intense analytical signals for *B. crispa*, *B. phyteumoides* and *B. triangularis* in Figure 7 C.

Quality Control of Herbal Medicines with Spectrophotometry and Chemometric Techniques

composition of flavonoids is different in the different species.

versa.

– Application to *Baccharis* L. Species Belonging to Sect – Caulopterae DC. (Asteraceae) 411

These results suggest that the substances responsible for the discrimination between species are those which have peaks around 225, 250, 300, 325, 350, 375 and 425 nm. Considering that the ethanol extracts used in this study might possess mainly flavonoids and that the region studied is where they present their main absorption bands (Greenham et al., 2003), it is likely that the components responsible for the spectra in this work would be flavonoids. It is worth taking into account that previous studies on the phytochemistry of *Baccharis* spp. have shown the presence of flavonoids, mainly flavanones and flavones (Coelho et al., 2004; Gene et al., 1996; Torres et al., 2000). From species of this genus 298 flavonoids were isolated. Among them 24 were flavanones and 85 were flavones and among them 48% are oxygenated in the C-3 (Gonzaga Verdi et al., 2005). They were described as good chemotaxonomic markers for the lower hierarchical levels of the Asteraceae family (Emerciano et al., 2001). In contrast to other characterization studies in which it is necessary to isolate and identify chemical substances, this study uses the complete ultraviolet-visible spectra measured between 200 nm to 900 nm, avoiding the isolation, purification and characterization of chemical compounds (Lonni et al., 2005). This methodology does not distinguish the spectra of individual components in the extract. Instead, the full spectrum ranges are analyzed as a whole. Absorption spectra of figure 7 C show that the qualitative

It has previously been observed that there is variability in the contents of several compounds in the different *Baccharis* species (Dresch et al., 2006; Gonzaga Verdi et al., 2005). With regards to the seasons, Borella et al. (2001) observed variations in the content of the total flavonoids, the largest number being found in a drug in the summer, which is an expected result due to the large number of functions attributed to them (Harborne & Williams, 2000). Our results are consistent with previous ones, showing different heights of peaks in the spectra in Figure 6, 7C and 8 B, even though these spectra were standardized. In a fertilization trial of *B. trimera,* in which the nutrient content of the soil was varied, no variation was observed in the contents of flavonoids (Borella et al., 2001). In our study we observed some changes in the content of flavonoids between collecting regions (Figure 4 and 5) but these did not prevent the grouping of various populations of the same species.

A routine step in multivariate data analysis is ordinarily to obtain a low-dimensional representation of the data. If two or three main components gives an accurate representation, a bi-or three-dimensional graph could be realized which mere observation is instructive. Clusters are usually easy to detect. After analysis of the eigenvalues (PC loading), more discriminant original variables are obtained. Then an ANOVA must be performing of each of these original variables between the OTUS (here species). The higher the eigenvalues, regardless of the sign will be more efficient in discriminating the OTUS. Variables that have negative eigenvalues (-) means that they are characterizing in the opposite direction in relation to the variables that have positive eigenvalues (+) and vice

Thus, absorbance values of the wavelengths (225, 250, 300, 325, 350, 375 and 425 nm) obtained from the analysis of the eigenvalues were submitted to ANOVA and we have established the wavelengths that differ between pairs of species (Table 2). Thus we confirm that there are differences in the UV / Visible spectra not observable to the naked eye but are

expressed as different clusters after a principal component analysis is applied.

*B. penningtonii* species samples have negative PC9 scores; PC9 separates *B. penningtonii* samples from the rest as seen in Figure 5. Figure 8 A shows a graph of the loading values (eigenvalues) on PC9 vs. ; it also shows that two regions present negative values between: 200 and 250 and 320 and 370 nm and it can be observed that the most intense analytical signals for *B. penningtonii* species samples are in these regions (Figure 8 B).

Fig. 8. A, PC loadings graph. PC9 (explains 0.36 % of the total data variance); B, Representative UV-Visible spectra of *B. articulata* (Ba), *B. crispa* (Bc), *B. gaudichaudiana* (Bg), *B. microcephala* (Bm), *B. phyteumoides* (Bphy), *B. penningtonii* (Bp), *B. sagittalis* (Bs), *B. triangularis* (Btr), *B. trimera* (Bt).

*B. penningtonii* species samples have negative PC9 scores; PC9 separates *B. penningtonii* samples from the rest as seen in Figure 5. Figure 8 A shows a graph of the loading values (eigenvalues) on PC9 vs. ; it also shows that two regions present negative values between: 200 and 250 and 320 and 370 nm and it can be observed that the most intense analytical

signals for *B. penningtonii* species samples are in these regions (Figure 8 B).

Fig. 8. A, PC loadings graph. PC9 (explains 0.36 % of the total data variance);

*B. triangularis* (Btr), *B. trimera* (Bt).

B, Representative UV-Visible spectra of *B. articulata* (Ba), *B. crispa* (Bc), *B. gaudichaudiana* (Bg), *B. microcephala* (Bm), *B. phyteumoides* (Bphy), *B. penningtonii* (Bp), *B. sagittalis* (Bs),

These results suggest that the substances responsible for the discrimination between species are those which have peaks around 225, 250, 300, 325, 350, 375 and 425 nm. Considering that the ethanol extracts used in this study might possess mainly flavonoids and that the region studied is where they present their main absorption bands (Greenham et al., 2003), it is likely that the components responsible for the spectra in this work would be flavonoids. It is worth taking into account that previous studies on the phytochemistry of *Baccharis* spp. have shown the presence of flavonoids, mainly flavanones and flavones (Coelho et al., 2004; Gene et al., 1996; Torres et al., 2000). From species of this genus 298 flavonoids were isolated. Among them 24 were flavanones and 85 were flavones and among them 48% are oxygenated in the C-3 (Gonzaga Verdi et al., 2005). They were described as good chemotaxonomic markers for the lower hierarchical levels of the Asteraceae family (Emerciano et al., 2001). In contrast to other characterization studies in which it is necessary to isolate and identify chemical substances, this study uses the complete ultraviolet-visible spectra measured between 200 nm to 900 nm, avoiding the isolation, purification and characterization of chemical compounds (Lonni et al., 2005). This methodology does not distinguish the spectra of individual components in the extract. Instead, the full spectrum ranges are analyzed as a whole. Absorption spectra of figure 7 C show that the qualitative composition of flavonoids is different in the different species.

It has previously been observed that there is variability in the contents of several compounds in the different *Baccharis* species (Dresch et al., 2006; Gonzaga Verdi et al., 2005). With regards to the seasons, Borella et al. (2001) observed variations in the content of the total flavonoids, the largest number being found in a drug in the summer, which is an expected result due to the large number of functions attributed to them (Harborne & Williams, 2000). Our results are consistent with previous ones, showing different heights of peaks in the spectra in Figure 6, 7C and 8 B, even though these spectra were standardized. In a fertilization trial of *B. trimera,* in which the nutrient content of the soil was varied, no variation was observed in the contents of flavonoids (Borella et al., 2001). In our study we observed some changes in the content of flavonoids between collecting regions (Figure 4 and 5) but these did not prevent the grouping of various populations of the same species.

A routine step in multivariate data analysis is ordinarily to obtain a low-dimensional representation of the data. If two or three main components gives an accurate representation, a bi-or three-dimensional graph could be realized which mere observation is instructive. Clusters are usually easy to detect. After analysis of the eigenvalues (PC loading), more discriminant original variables are obtained. Then an ANOVA must be performing of each of these original variables between the OTUS (here species). The higher the eigenvalues, regardless of the sign will be more efficient in discriminating the OTUS. Variables that have negative eigenvalues (-) means that they are characterizing in the opposite direction in relation to the variables that have positive eigenvalues (+) and vice versa.

Thus, absorbance values of the wavelengths (225, 250, 300, 325, 350, 375 and 425 nm) obtained from the analysis of the eigenvalues were submitted to ANOVA and we have established the wavelengths that differ between pairs of species (Table 2). Thus we confirm that there are differences in the UV / Visible spectra not observable to the naked eye but are expressed as different clusters after a principal component analysis is applied.

Quality Control of Herbal Medicines with Spectrophotometry and Chemometric Techniques

Apigenin Genkwanin Luteolin Band

and this retention time corresponds to the luteolin marker (Figure 10).

0,83

*B. articulata* X X - X X - - -

*B. gaudichaudiana* X X X - - X - - *B. sagittalis* X X X - - X X X

Table 3. Summary of the bands obtained by TLC for *B. articulata, B. gaudichaudiana* and *B. sagittalis.* Mobile phase: DCM: Hexane: MeOH (4:2:1). x indicates presence of the band, -

HPLC analysis was carried out on the same extracts as used for the studies with UV-Visible spectrophotometry and for TLC in *B. articulata*, *B. gaudichaudiana* and *B. sagittalis.* The chromatophotographic profiles showed the main peaks with the following retention times for *B. articulata*: 2.32, 3.00, 3.22 and 3.30 min; for *B. gaudichaudiana*: 2.26, 3.00, 3.12 and 3.30 min and for *B. sagittalis*: 2.26, 2.40, 3.00, 3.12 and 3.30 min. The retention time for the apigenin marker was 3.00 min and for genkwanin was 3.30 min. These peaks appear in all three species studied. In *B. gaudichaudiana* and *B. sagittalis* there is also a peak at 3.12 min,

Band 0,75

Band 0,58

Band 0,50

Band 0,33

Rf 0,33, 0,5 and 0,58 (Figure 9 A).

indicates absence of the band.

**2.2.3 HPLC analysis** 

– Application to *Baccharis* L. Species Belonging to Sect – Caulopterae DC. (Asteraceae) 413

characterized by the previous method. NP-PEG reagent under UV 365 nm and UV 254 nm were used to detect the polyphenolic compounds present in the extracts (Wagner & Bladt, 1996), in accordance with that published for the *Baccharis* genus on account of the high occurrence of these compounds in the genus (Bohm & Stuessy, 2001; Gonzaga Verdi et al., 2005). Different color bands will be detected with the NP-PEG reagent under UV 365 nm or quenching bands will be detected under UV 254 nm if these compounds are present in the extract. On the other hand, given that one of our objectives was the identification of the species in the state of a crude drug, it was very important to select appropriate components of easy access for chemical quality control. So apigenin, genkwanin and luteolin, which are compounds present in several *Baccharis* species, were selected as markers. Genkwanin has been reported in *B. articulata* (Gianello & Giordano, 1984) and apigenin in *B. gaudichaudiana*  (Fullas et al., 1994). Luteolin has not yet been reported in any of the three species analyzed by TLC, but it has been found in the following species from the same section: *B. microcephala*  (Bohlmann et al., 1985), *B. trimera* (Soicke & Leng-Peschlow, 1987) and *B. triangularis*  (Pettenati et al., 2007). Our results are shown in Table 3. The TLC chromatograms showed differences for the three species studied (Figures 9A and 9B). The flavonoids apigenin and genkwanin were found in all three species, although the band that corresponds to genkwanin somewhat overlaps in *B. gaudichaudiana* and *B. sagittalis.* In the case of the other marker, we observed that the band corresponding to luteolin appeared in *B. gaudichaudiana*  and *B. sagittalis*. There are at least two more bands at Rf 0,83 and 0,75 for *B. articulata* and there is another band for *B. gaudichaudiana* at Rf 0,58 and three more bands for *B. sagittalis* at


Table 2. Wavelengths (nm) with statistically significant differences (p <0.05) among the species. *B. articulata* (Ba), *B. crispa* (Bc), *B. gaudichaudiana* (Bg), *B. microcephala* (Bm), *B. penningtonii* (Bp), *B. phyteumoides* (Bphy), *B. sagittalis* (Bs), *B. triangularis* (Btr) and *B. trimera* (Bt)
