Is the labellum informative to distinguish species of Scaphyglottis(Orchidaceae)? Insights from geometric morphometrics

Abstract Scaphyglottis Poepp. & Endl. (Orchidaceae) is a neotropical genus with 78 species, 14 of which are found in Brazil. Some of these taxa have an intricate taxonomy, especially due to the morphological similarity among species and the existence of species complexes. Our study aims to test if the labellum shape, the main source of species diagnostic characters, can confirm the proposed circumscription of Brazilian species of Scaphyglottis. We analyzed 136 flowers from 12 species, to which we applied 12 landmarks, using a geometric morphometrics approach. Our results show that most species can be distinguished. Most of the variation is observed in the claw and apices of labellum lobes, confirming that the labellum is important tool in taxon distinction. Species with distinctly 3-lobed labellum showed some overlap but, for these taxa, we have found distinction specially in the lateral lobes. Our study showed new morphologically variable features in Scaphyglottis that can be used in future taxonomic studies. For widely distributed species we suggest a population-level approach, as local environmental factors can impact morphology, causing expected intraspecific variation.


Introduction
Standard external morphological features, traditionally used in plant taxonomy, are not always sufficient to delimit taxa (Galtier 2019;Lawley et al. 2022).In closely related or morphologically similar species, the use of qualitative morphological data alone will likely merge distinct taxa or segregate populations of a same taxon.Although molecular approaches can help solve some of these issues, a morphological circumscription is still needed to delimit species in a practical sense (Galtier 2019).
Distinguishing species that suffered allopatric speciation is usually simpler, but this is often not the case with young or cryptic lineages (Pinheiro et al. 2018;Pessoa et al. 2021).In the first case, these lineages can be included within one of the many species concepts (De Queiroz 2007;Pinheiro et al. 2018), but this is not always applicable to the second case, as other evolutionary and ecological processes, such as hybridization and introgression, or development homeostasis (Mayr 1963), can be at play and lead lineages to form species complexes (Pinheiro et al. 2018;Pessoa et al. 2022).
The labellum is considered a fundamental character to distinguish species in Orchidaceae (Dressler 1993;Carnevali & Ramírez-Morillo 2003;Szlachetko & Kolanowska 2014a;Araújo et al. 2022), and have been used as a major source of information in previous studies that applied geometric morphometrics (Menini-Neto et al. 2019;Lau et al. 2021;Pessoa et al. 2021;Camelo-Júnior et al. 2022).Thus, in this study we evaluate if geometric morphometrics of the labellum is able to distinguish the species of Scaphyglottis that occur in Brazil.The main goals are: i) to produce evidence of valuable diagnostic characters to delimit these species and ii) to evaluate delimitation of taxa with similar vegetative morphology.

Sampling
We sampled 136 flowers from 103 specimens from herbarium collections and field expeditions (Tab.S1).Our sampling is representative of the geographical distribution of Scaphyglotttis in Brazil (Fig. 2).Specimens from the following herbaria were analyzed: BHCB, EAC, EAN, FURB, PEUFR, HST, IAN, INPA, IPA, MG, RB, UEC, UFMT, UFP and UPCB (acronyms according to Thiers 2022, continuously updated).Three field expeditions were carried out at Reserva Florestal Adolpho Ducke (Manaus, AM), Parque Estadual Serra do Aracá (Barcelos, AM), Serra do Tepequém (Amajari, RR) and Serra Grande (Cantá, RR), where we collected several specimens of six species of Scaphyglottis (S. bidentata, S. fusiformis, S. imbricata, S. reflexa, S. sickii and S. stellata).All specimens collected on the field were kept in cultivation and until now, ten specimens have been deposited in herbarium collections.For each sampled specimen, we removed one to two flowers for analysis, with authorization of the herbarium curators.Some taxa had low sampling, either due to rarity and scarcity of specimens in collections or due to the available specimens being sterile or having only fruits.Specifically, S. graminifolia (one single flower), S. livida (three flowers), S. reflexa (three flowers) and S. brasiliensis (four flowers) had a low number of sampled flowers.We could not include S. emarginata and S. imbricata for the same reason.Therefore, 12 from the 14 Brazilian species were Acta Botanica Brasilica, 2023, 37: e20230017 analyzed.Specimens identifications were verified according to Dressler (1993) and Szlachetko and Kolanowska (2014a).

Geometric Morphometrics (GM)
The flowers were rehydrated, and the labella were mounted in white cardstock paper.The labella were later scanned using an Epson Expression 10000XL scanner in 1200 dpi resolution for geometric morphometrics analyses.The mounted flower images were digitized using TPSutil 1.78 (Rohlf 2015).We added 12 landmarks to the labellum, using TPSdig2 (Rohlf 2015).Landmarks 1 and 11 represent the labellum column foot junction, landmark 12 represents the median point between landmarks 1 and 11, landmarks 2 and 10 represent the claw and lateral lobes junction, landmarks 3 and 9 correspond to the lateral lobe apices, 4 and 8 represent the lateral lobe and median lobe junction, 5 and 7 represent the median lobe apices and 6 represents the apex or central point of the labellum invagination (Fig. 3).
The files with the added landmarks were transferred and analyzed in the software MorphoJ 1.06d (Klingenberg 2011), and then submitted to Procrustes analysis, where the size variation is removed from the analysis, preserving the variation of the points from the centroid and changes in shape (Goodall 1991;Bookstein 1996).An exploratory Principal Component Analysis (PCA) (Hotelling 1933) was performed, aiming to group the specimens according to the labellum variation and to detect potential outliers.We then conducted a Canonical Variable Analysis (CVA) (Campbell & Atchley 1981) to discriminate the variation between pre-determined groups (i.e., the species) and the sequence in which they separate according to the canonical variables (CVs).
Because of the high morphological variation of the 12 analyzed species, we conducted new PCA and CVA analyses for species with 3-lobed labellum separately (S. boliviensis, S. brasiliensis, S. prolifera and S. stellata), as their similarities in the labellum led to overlapping of the ellipses when the total dataset is analyzed.We applied the methods described above to this reduced dataset.

Results
The PCA exploratory analysis shows that the variation explained by the first two axes represents 52% and 25% of the variation respectively, and the accumulated variation of the three first axes is 86% (Tab.S2).For the first axis (PC1; variance = 52%), the larger landmark variation is found between landmarks 3 and 9, which represent the lateral lobe apices, 5 and 7 representing the median lobes, and 6 representing the labellum apex.For PC2 (variance = 25%), most of the variation is also found between landmarks 3 and 9, and between landmarks 4 and 8, which represent the junction between lateral and median lobes.For PC3 (variance = 8.6%), the larger variance is found in landmark 6, representing the apex of the labellum.Overall, the ellipses that represent the species are considerably overlapped (Fig. S1), except for S. bidentata and S. livida.
The CVA discriminant analysis showed an accumulated variation of 77.8% in the first three CVs (Tab.S2).On CV1 (variance = 50.5%)most of the variation is represented by landmarks 3 and 9, which represent the lateral lobe apices, 5 and 7 representing the median lobe apices, and 6, representing the apex position or the central point of the labellum invagination (Fig. 4A).In CV2 (variance = 27.2%) and CV3 (variance = 11.8%) the larger variation is seen in landmarks 5, 6 and 7, which represent the median lobe apices and the apex or central point of the labellum invagination (Fig. 4B-C).
The variation of the landmark position of CV1 allowed to separate S. punctulata, S. livida, S. graminifolia, S. fusiformis and S. bidentata in the negative pole, and S. prolifera and S. stellata in the positive pole, while S. boliviensis, S. sickii, S. brasiliensis, S. modesta, and S. reflexa were little impacted (Fig. 5A).The variation found in CV2 separated S. bidentata, S. modesta and S. reflexa in the positive pole and S. fusiformis, S. punctulata, S. livida, S. graminifolia, and S. sickii in the negative pole, while S. brasiliensis, S. boliviensis, S. stellata and S. prolifera were little impacted (Fig. 5A).CV3 allowed to separate S. livida, S. fusiformis, S. reflexa and S. brasiliensis in the positive pole and S. sickii in the negative pole, while the other species were little impacted.We highlight that S. boliviensis was not separated from the other species in any of the CVs that had higher variance, always being in the group of species that were not impacted.This is also clear from the general observations in Fig. 5A-B, where a strong overlap of this species with the remaining taxa is observed.
In the analysis conducted with the smaller species dataset (S. boliviensis, S. brasiliensis, S. prolifera and S. stellata), the larger variation in PC1 (variance = 54.425%) is found in landmarks 3 and 9, which represent the lateral lobe apices.For PC2 (variance = 30.589%),the largest variation is in landmarks 2 and 10, which represent the junction of the claw with the lateral lobes.The four species remained strongly overlapped in this PCA (Fig. S3-S4).In the CVA, the accumulated variation of the first two CVs is 99.99% (CV1 with 45.35%, CV2 with 40.99%, and CV3 with 13.65%, Tab.S2).In CV1, the largest variation is found between landmarks 3 and 9, which represent the lateral lobe apices, and 2 and 10, which represent the junction of the claw with the median lobes.The same landmarks showed the largest variation in CV2, with the addition of 4 and 8, which represent the lateral and median lobes junction (Fig. 6).CV1 separated S. prolifera in the positive pole and S. boliviensis in the negative pole, while CV2 separated S. stellata in the positive pole and S. brasiliensis in the negative.In this new analysis, we observed that S. brasiliensis overlapped with the remaining species, but S. boliviensis was strongly separated from S. stellata, which is the more morphologically similar species (Fig. 7).

Discussion
In this study we tested if geometric morphometrics of the labellum is useful to distinguish species of Scaphyglottis that occur in Brazil, as it has been used to separate taxa in Orchidaceae in many recent studies (e.g.: Pinheiro & Barros 2007;Pinheiro & Barros 2009;Menini-Neto et al. 2019;Lau et al. 2021;Pessoa et al. 2021;Camelo-Júnior et al. 2022).Previous taxonomic studies of Scaphyglottis used a combination of vegetative characters (shape and size of pseudobulbs and leaves), besides flower color, size, and overall shape of the labellum to distinguish species (Dressler 1993;Carnevali & Ramírez-Morillo 2003;Szlachetko & Kolanowska 2014a;Araújo et al. 2022).We found evidence that the lateral lobe apices, the junction of the claw with the lateral lobes, the median lobe apices, and the apex or central point of the labellum have enough morphological variation to delimit these species, allowing their identification solely based on this structure.
The species with higher morphological variation in Brazil, S. bidentata and S. livida, were easily separated when compared to the other analyzed species in the first PCA (Fig. S1-S2), which could not separate most of the species.This pattern was also found in previous studies with the labellum in Orchidaceae, in which the PCA showed little power to  Acta Botanica Brasilica, 2023, 37: e20230017 distinguish among morphologically similar species, being mostly used as a preliminary approach (Goldman et al. 2004;Pinheiro & Barros 2009;Menini-Neto et al. 2019;Lau et al. 2021;Pessoa et al. 2021;Camelo-Júnior et al. 2022).
The CVA allows better separation of the species, besides indicating the regions of the labellum that show higher variation, such as the median lobe apex (landmarks 5, 6 and 7), and the wider point of the labellum (landmarks 3 and 9) (Fig 3).Similar results indicating these regions as more variable were also found by Pinheiro and Barros (2007;2009) in Maxillariinae, Pessoa et al. (2021) in Laeliinae and Camelo-Júnior et al. (2022) in Oncidiinae.However, among the species of Scaphyglottis with a more distinctly 3-lobed labellum, S. boliviensis, S. brasiliensis, S. prolifera and S. stellata, these regions showed low variation when compared to the other species, remaining overlapped in the analyses (Fig. 4).
The second CVA analysis, containing only these four species, allowed their separation with higher variation in the lateral lobe apex and claw junction (landmarks 3 and 9, 2 and 10, respectively) (Fig. 6), separating S. boliviensis and S. stellata, which are frequently misidentified in herbarium collections (CRIA 2022).Previous studies used similar strategies to try and separate species that remain overlapped in preliminary analyses (Camelo-Júnior et al. 2022).In the case of these four species, landmarks 2 and 10, referring to the junction of the claw with the median lobes, had an essential role in separating them.These landmarks contain the length and width of the claw (Fig. 7).
Among these four taxa, S. stellata and S. boliviensis can be vegetatively identified only by leaf width (> 0.4 cm vs. < 0.3 cm, respectively), while regarding floral characters, S. boliviensis has a more prominent callus in the labellum than S. stellata (Dressler 1993;Koch et al. 2018).Our results indicate that the claw is smaller in S. stellata than in S. boliviensis.Both species have synonyms that were described based on specimens with cream-colored/whitish or lilac flowers (Reichenbach 1869;Schweinfurth 1955;Mansfeld 1928).We believe that further investigation at population level is needed to decide either to maintain those as synonyms or reinstate them as separate species; nonetheless our results indicate that the claw size is an important feature that might help in this task.
Scaphyglottis brasiliensis, one of the taxa included in this second analysis, is morphologically related to S. reflexa and S. emarginata in Brazil, with these two latter species presenting cylindrical leaves (Lindley 1839;Schlechter 1922;Pessoa & Alves 2012) and being relatively rare in herbarium collections.Our analysis, however, did not show similarities between the labellum of S. brasiliensis and S. reflexa (Fig. 4A).Among the group of species with 3-lobed labellum, S. brasiliensis has the longest and narrowest claw (Hoehne 1949).However, this species overlaps with other three taxa, presenting the wider ellipse in the analysis plot, with datapoints further separated from each other, indicating either insufficient sampling or large morphological variation.This group of species with cylindrical leaves, already treated as a separate genus, Reichenbachanthus Barb.Rodr.(Brade 1935;Garay 1967), deserves a dedicated taxonomic study, although our results show these two taxa are distinguishable.
Lastly, we could easily distinguish S. prolifera from S. sickii based on the first CVA (Fig. 4).These two species are considered more morphologically related to each other than to other species that occur in Brazil, due to the smaller leaves and pseudobulbs, and flower color and size.S. prolifera has the least prominent 3-lobed labellum among the species included in the second analysis, and according to Dressler (1993), it is not a typically 3-lobed labellum as seen in other species, but a labellum where the median portion extends to the sides.However, this character (position of the landmarks) probably caused the overlapping with S. boliviensis, S. brasiliensis and S. stellata in the first analysis, due to being different than the remaining species with unlobed labellum.
Studies that highlight the morphological variation in Orchidaceae species with wide geographical ranges and varied habit (epiphytic, rupicolous or facultative terrestrial) were carried out in Epidendroideae and Maxillariinae (Pinheiro & Barros 2007;2009) and Laeliinae (Menini-Neto et al. 2019).Taking into account that some species of Scaphyglottis have similar attributes, further studies to evaluate the geographic organization of morphological variation and its relation to environmental gradients should be considered, especially in widely distributed species like S. stellata, that is known from Nicaragua to central Brazil.The labellum has an essential role in pollinator attraction in Orchidaceae, mainly due to color, size, odor, and floral resources (nectar, pollen, oils) (van der Pijl & Dodson 1966).In Scaphyglottis, nectar usually accumulates in the labellum base (Dressler 2001), therefore, selective pressures associated with pollination might influence the variation found in this structure, especially in the size and width of the claw.It might explain differences in the labellum morphology found in similar species studied here.However, reproductive biology studies need to be carried out in Scaphyglottis species to elucidate evolutionary pressures related to labellum shape variation and avoid speculative conclusions.
Our study showed new morphologically variable features in Scaphyglottis that can be used in future taxonomic studies, or even for identification of specimens based on labellum fragments (Fig. 8).We highlight that six of the more morphologically close species did not overlap in their labellum morphology (i.e., S. sickii and S. prolifera; S. reflexa and S. brasiliensis; S. punctulata and S. fusiformis), possibly indicating different pollination strategies.For widely distributed species, such as S. prolifera, S. stellata and S. reflexa, we suggest a population-level approach, as local environmental factors can impact morphology, causing expected intraspecific variation (Jonas & Geber 1999;Blinova 2012;Menini-Neto et al. 2019).

Figure 2 .
Figure 2. Distribution map of species sampled in the morphometric analysis, indicating the number of flowers analyzed by specimen.The symbol (+) indicates that specimens are from the same place, circles represent the specimens.Map prepared by Falen, L.

Figure 3 .
Figure 3. Location of landmarks applied to the labellum of Scaphyglottis for geometric morphometric analysis.A. S. prolifera, B. S. bidentata.Photos by Araújo, A. M. Plate prepared by Miranda, G.