Four shape principal components describe tepal variation in M. haageana and its sister species
We characterized shape and color-pattern variation in a collection of 20 M. haageana accessions that belonged to the six subspecies as previously proposed [15], plus eight accessions from the sister species M. albilanata, giving a total of 717 tepals, from 272 flowers, from 28 accessions. We have recently used some of these accessions to characterize the natural variation in root architecture development [16]. To perform the tepal shape analysis, we flattened the perianth and fitted a 13-landmark template on each studied tepal (Fig. 2). First, to extract the tepal shape variation, we performed a Principal Component Analysis (PCA) with Procrustes for rotation, translation, but not size, because size is one of the features that display variation among the M. haageana accessions.
We found that 99% of the variation was covered by eight Principal Components (PCs), but the 7th and 8th capture less than 0.1% of the variation in each, and their variation was so subtle that could not be described. The first 6 PCs captured 98.59% of the variation in tepal shape (Fig. 3). The first shape Principal Component (PC1S) a large proportion of the variation (86.19%) confirming that M. haageana displays wide phenotypic variation in tepal size. Accordingly, PC1S seems to capture the tepal lamina length. PC2S is an orthogonal axis to PC1S, and therefore should capture the shape, regardless of the size; thus PC2S captured 5.76% of the variation and seems to quantify the tepal lamina width. PC3S captured 2.22% of the variation, and seems to capture whether the tepal is tilted towards the left or the right; however we believe this to be an artefact of the tepal flattening process, and therefore it was not taken into account. PC5S captured 2.16% of the variation, which seems to capture aspects of the tepal lamina shape, such as the width of the proximal end, or the sharpness of the distal end. PC5S captures 1.65% of the variation but similar to PC3S, we believe this to be an artefact of the teal flattening process, and thus it was not further considered. PC6S captures 0.61% of the variation and seems to describe the width of the middle section of the lamina. These four shape Principal Components (PC1S, PC2S, PC4S and PC6S) were then considered to evaluate shape variation in M. haageana accessions, plus its sister species M. albilanata.
We then plotted the biologically meaningful shape PCs according to their subspecies or their accession (Fig. 4). We noted that according to PC1S, M. albilanata has the smallest tepals, while M. haageana subsp. conspicua and M. haageana subsp. meissneri had the largest tepals (Fig. 4a). Concerning the lamina width regardless of size (PC2S), M. haageana subsp. acultzingensis had the widest tepals, while M. haageana subsp. meissneri had the narrowest tepals (Fig. 4c). On lamina shape (PC4S) M. haageana subsp. haageana had the widest distal end of the tepal, but there were no differences between M. albilanata as compared to three other M. haageana subspecies (Fig. 4e). On the lamina middle section (PC6S), the subspecies had broad variation and little distinction was observed between them (Fig. 4g). Despite the variations observed between subspecies, differences between accessions belonging to the same subspecies were far more striking. For instance, on PC1S we observed ample variation within the M. haageana subsp. conspicua accessions, or M. haageana subsp. haageana accessions; but a similar pattern was observed within the M. albilanata accessions (Fig. 4b). This pattern was also observed in the other shape traits PC2S, PC4S, and PC6S, which was less obvious in the later. In other words, each accession has a distinctive set of shape traits, which seemingly do not agree with their assigned subspecies.
Four color principal components describe tepal variation in M. haageana and its sister species
Using the geometric morphometric landmark data-points, we warped all the tepal images into the mean shape, using 10,000 pixels contained in the polygon (Fig. 2) in the RGB channels (30,000 pixels). With this new dataset, we performed a PCA to extract the main trends of variation in tepal color. 95% of the variation was captured by 47 color PCs. We only present the first 6 PCs, as they seem to capture trends of variation easily be visualized (Fig. 5). Other PCs have very subtle trends that could not be described (data not shown). The first color Principal Component (PC1C) seems to capture the color saturation; the PC2C captures the hue of the tepal; PC3C captures the proximal-distal hue of the tepal, and PC4C captures the mid-stripe hue. PC5C seems to capture left-right hue, which we believe to be an artefact because the right or left-handedness of the pigmentation is not observed in M. haageana natural accessions. PC6C was also not described and further considered, as the variation it captures is very subtle. Thus, we propose four color geometric morphometric traits (PC1C, PC2C, PC3C and PC4C) in M. haageana (and M. albilanata) to compare pigmentation patterns between and within subspecies.
Similarly to what we did on the shape, we plotted the biologically meaningful color PCC according to their subspecies (Fig. 6). Regarding saturation (PC1C), we found little variation between most M. haageana subspecies, except for M. haageana subsp. haageana which has lighter saturation tones; on the other hand, M. haageana subsp. san-angelensis has darker saturation tones (Fig. 6a). Concerning hue (PC2C), M. haageana subsp. acultzingensis had more reddish tepal color as compared to the rest of the M. haageana subspecies, but M. albilanata had the rather magenta tepal colors (Fig. 6c). On the proximal-distal tepal hue (PC4C), and similar to the distinctive patterns observed in PC1C, M. haageana subsp. haageana has darker magenta pigmentation towards the apex, whereas M. albilanata has a darker magenta color towards the proximal end of the tepal (Fig. 6e). In other words, M. haageana subsp. haageana and M. albilanata share pigmentation patterns, but not with the complement of subspecies recognized in M. haageana, based on PC1C and PC4C. Finally, the mid-stripe tepal hue (PC6C) is more prominent and noticeable in M. haageana subsp. haageana, further supporting the distinction of this subspecies from the rest (Fig. 6g).
While differences in color-pattern variations were remarkable between subspecies, when comparing accessions only PC1C and PC4C showed broad differences between accessions, even from the same subspecies. This was the case of M. haageana subsp. meissneri on PC1C, in which CC031 has significantly lower PC1C value (saturation) than CC030. Another example is the A. albilanata accession CC044 in which there is ample variation between individuals of the same accession, that is noticeable by the amplitude of the boxplot (Fig. 6b). Remarkably, M. haageana subsp. haageana accessions are similar to one another in PC1C and PC4C.
Phenetic analyzes suggests possible natural groups
Both shape and pigmentation analyses provided quantitative parameters of novel axes to compare the accessions of M. haageana and its sister species M. albilanata. As shown in Fig. 4 and Fig. 6, some traits partially recapitulate subspecies classifications, but others do not. To unravel the possibility that these traits could recapitulate the subspecies classification, a series of multivariate analyses were performed. First, a Linear Discriminant Analysis (LDA) with shape traits was implemented to test whether the four PCS parameters separated the five subspecies plus the sister species (Fig. 7a). This model gave a precision of 65.45% in the correct assignment of accessions. In other words, in an LDA, the four shape components in M. haageana failed to create a distinction between the five subspecies and the sister species. Subsequently, the four parameters corresponding to pigmentation were tested (PCc). The model gave a 80.67% accuracy in assigning the accessions to the subspecies. Nevertheless, in this analysis it was possible to see the separation of two groups of accessions corresponding to M. haageana subsp. conspicua and M. haageana subsp. san-angelensis (Fig. 7b). Finally, we performed a third LDA combining the eight components corresponding to shape (PCS) and color (PCC). This model has an accuracy of 71.32% in assigning subspecies. In this case, the two groups of accessions that separated from the rest were the species M. albilanata and the subspecies M. haageana subsp. haageana (Fig. 7c). In summary the LDA of shape and pigmentation parameters only partially recapitulated some of the previously proposed taxonomic units [15].
We also used another multivariate method based on Euclidean distances to test whether the taxonomic units can be recapitulated. This gave groupings between the accessions based on shape and color. It was observed that the accessions CC020, CC021 and CC022, corresponding to the subspecies M. haageana subsp. haageana, are grouped in a single branch as compared to the rest of the accessions (Fig. 7d). This result might indicate that the combination of shape and color parameters in M. haageana subsp. haageana could indeed define this group as a subspecies within M. haageana.
Pigmentation patterns are known to have impacts on color reflectance, and therefore color perception by pollinators, as demonstrated in Antirrhinum [17]. In order to test whether the environment might play a role in determining tepal pigmentation patterns in M. haageana, we performed an association analysis of solar radiation with our shape and color PCs (Fig. 8a). Interestingly, we found several correlations between solar radiation, with tepal shape and color. For simplicity, we show the plots of the six highest correlations between the environment and tepal attributes (Fig. 8b-g). Regarding tepal shape, we found that global horizontal radiation (GHI) and air temperature (TEMP) were negatively correlated with PC2S (lamina width) (Fig. 8b-c), implying that narrower tepal accession originates from cooler locations with lower global horizontal irradiation, while wider tepal accessions originated from warmer and higher global horizontal irradiation locations. We also found a positive correlation between the organ shape parameter PC6S (lamina middle section) and global horizontal radiation (GHI) (Fig. 8d), which interpretation requires further examination. Regarding tepal color, we found that PC1C (saturation) has a negative correlation with diffuse horizontal radiation (DIF) (Fig. 8e); this means that darker tepals are present in accessions originally from locations with lower DIF, while lighter tepals are present in accessions originally from locations with higher DIF. Another interesting negative correlation was found between PC1C (saturation) and air temperature (TEMP) (Fig. 8f), in which paler tepals (low PC1C) were from accessions originally from lower TEMP locations, as opposed to darker tepals (high PC1C) which were from accessions from higher TEMP locations. On the other hand, PC4C (mid-stripe tepal hue) was positively correlated with direct normal radiation (DNI) (Fig. 8g), in which the more noticeable mid-stripe (low PC4C), the lower DNI; meanwhile the less noticeable mid-stripe (high PC4C) the higher DNI. Overall, these results allowed us to establish a link between tepal shape and pigmentation patterns, and the environment, which might definitely have adaptive bases.
In order to provide a possible anatomical explanation on the origins of shape and color variation in M. haageana tepals, we examined the tepal anatomy of M. haageana subsp. san-angelensis (Fig. 9). To do this we performed histological sections of the proximal, middle, and distal sections of the tepals, and characterized the tepals qualitatively and quantitatively. We obtained that epidermis of the abaxial and adaxial sides are fairly similar to each other, and also similar cell size at the distal or proximal end of the tepal. These sections suggest that color variation along the transversal and longitudinal sections of the tepal lamina does not originates within the epidermis cell shape, and possibly the variation between accessions either. Thus, the variation in tepal color that we characterized between accessions must be given by the variation in betalain (purple color) and betaxanthin (yellow color) pigments, which concentrations remain to be quantified. Last, within the mesophyll we found large mucilage bodies, which sometimes seemed to be large single cells, but in some other cases, these mucilage bodies were surrounded by small cells. These mucilage bodies had a variety of sizes and occupied most of the mesophyll tissue towards the distal part of the tepal and were rather small at the proximal end of the tepal. On the other hand, the tepal was highly vascularized towards the proximal end, but fewer or smaller vascular bundles were observed towards the distal end. In leaves, vascular bundles are more prominent towards the leaf base, and at the apex they become mere vascular endings with tracheal elements. In the context of cacti tepals, this inverse relationship of vasculature versus mucilage bodies in the mesophyll might be part of the tepal adaptation for the flower to avoid desiccation towards the distal end, but maintain water flow towards the proximal end. Finally, to contextualize what we observed in M. haageana subsp. san-angelensis, we found that tepal anatomy is fairly similar to what has been observed in tepals of Epiphyllum phyllanthus [18], another distantly related cacti species, suggesting that tepal anatomy is fairly constant in the Subfamily Cactoideae.