Could leaf morphoanatomy characters help in the delimitation of Dyckia selloa complex?

ABSTRACT Among species of the genus Dyckia Schult. & Schult. f. there are 13 endemic species of the Brazilian states of Rio Grande do Sul and Santa Catarina informally treated as the Dyckia selloa complex. This study employed standard plant anatomy techniques to investigate variation in leaf morphology of species belonging to the genus Dyckia with a focus of establishing characters that help delimit the Dyckia selloa complex. The results allowed the survey of morphological and anatomical characters important for the characterization of species. Such characters include color of spines, the presence of water-storage parenchyma and mechanical hypodermis on both leaf surfaces, and the presence of tetracytic stomata on only the abaxial surface. Analyses support the current delimitation of the complex and recommend the investigation of reproductive and/or vegetative organs to better understand the relationships among these species of Dyckia.

The circumscription of Dyckia has been the subject of several morphological and taxonomic studies in recent years, with several authors using informal categories to group species with morphological similarities (e.g., Dyckia macedoi complex and Dyckia saxatilis complex by Guarçoni (2015), Dyckia ferruginea complex and Dyckia selloa complex by Büneker et al. (2021)).Among these species complexes, the Dyckia selloa complex is particularly interesting for being endemic to the Brazilian states of Rio Grande do Sul and Santa Catarina (Büneker et al. 2021) and for having been delimited based on in an unique combination of the morphological characteristics for the genus.According Büneker & Mariath (2022) the Dyckia selloa complex which was previously treated as the genus Prionophyllum by Koch (1873) and Mez (1896Mez ( , 1935)), Dyckia subg.Prionophyllum by Baker (1889), the Prionophyllum group by Leme et al. (2012) and Krapp & Eggli (2019), or the D. maritima complex by Strehl & Beheregaray (2006) and Büneker et al. (2015).According to these authors the Dyckia selloa complex currently consists of 13 species:  Mez. Büneker & Mariath (2022) emphasize that the group is clearly distinguished from other species of the genus based on morphology, they say the main distinguishing characters are: inflorescences usually composed of first to third order (vs.simple or compound to first order), with numerous first order branches (usually more than 10 vs. less than 10 branches), generally small and inconspicuous floral bracts, numerous flowers (more than 100 vs. less than 80) with non-unguiculate (vs.usually unguiculate) petals, stigma during anthesis usually with erect to suberect (vs.twisted) stigmatic lobes, pauciovulate ovary locules (less than 18 vs.more than 30 ovules) and oblanceoloid (vs.flattened) seeds with a little developed wing.Circumscription attempts (Krapp et al. 2014;Schütz et al. 2016;Pinangé et al. 2016;Gomes-da-Silva et al. 2019) and the existence of species complexes within Dyckia evidence the scarcity of morphological characters available for robust delimitations and, thus, the need for studies that complement available information to provide efficient data for taxonomic decisions and circumscriptions.
Morphological and anatomical studies within Bromeliaceae frequently address morphological comparisons between species and ecological adaptations to xeric environments (Tomlinson 1969;Smith & Downs 1974;Benzing 2000).In addition to these, some studies have involved morphoanatomical aspects with the aim of helping to delimit taxa (e.g.Santos-Silva 2015; Carvalho et al. 2016;2017;Guarçoni et al. 2014;2017;Büneker & Mariath 2022).Also of significance are studies that have sought to identify characters of taxonomic value and ecological significance, such as that of Silva & Scatena (2011), who compared leaf anatomy within the subfamily Tillandsioideae and concluded that, as they are epiphytic, some of the xeromorphic characteristics may represent ancestral adaptations during speciation.Recently, leaf anatomy data have revealed xeromorphic synapomorphies for the genera Deuterocohnia Mez, Dyckia and Encholirium Mart.ex Schult.& Schult.f., mainly the presence of mechanical hypodermis and water-storage parenchyma, which would be related to the occupation and diversification of these genera in the dry region of South America (Santos- Silva et al. 2013).As highlighted, leaf anatomical studies within Dyckia can provide data that help to solve taxonomic problems and reveal adaptations that have contributed to the diversification of the group.
Considering that most studies carried out within the Pitcairnioideae and, more specifically, within the genus Dyckia, have focused on the taxonomy and phylogeny of the group (Krapp et al.2014;Schütz et al. 2016;Pinangé et al. 2016;Gomes-da-Silva et al. 2019), the investigation of morphoanatomical characters is necessary.This is especially true regarding the analysis of vegetative characters related to leaf anatomy, which can provide data helpful in the delimitation of the Dyckia selloa complex and the species it comprises.Thus, the present study aimed to investigate the leaf morphology of species of Dyckia, with a focus on establishing characters that help delimit the Dyckia selloa complex.More specifically, the aims were to: (1) analyze leaf external and internal morphology; (2) to test which characters are important for species delimitation, and based on this, (3) to check whether the data obtained can be useful for delimiting the Dyckia selloa complex.

Light Microscopy
Fresh leaves were collected from the leaf node right before the inflorescence for observation under light microscopy and stereomicroscopy.Leaves were sectioned transversally, at the median region, and immersed in a fixative solution containing 1% glutaraldehyde and 4% formaldehyde (McDowell & Trump 1976) in a 0.1M sodium phosphate buffer solution pH 7.2 and kept under vacuum for 24 hours.The material was subsequently dehydrated in an ascending ethanol series (Johansen 1940), and after subjected to an alcohol:chloroform series (3:1, 1:1, 3:1), and embedded in hydroxyethylmethacrylate resin (Gerrits & Smid 1983).Sections were made using a Leica 2265 rotary microtome, equipped with a high-profile disposable blade, to obtain 3-5 µm in thick sections.The material was stained with 0.1% Toluidine Blue O in 0.1M sodium phosphate buffer pH 4.4 (Feder & O'Brien 1968).Images were recorded using a Leica DMR HC microscope equipped with a Zeiss a Zeiss AxioCam digital camera and the free software Carl Zeiss ZEN LITE 2012.

Histochemistry
Fresh leaves were free hand transversally sectioned at the median region of the leaf and used for the histochemical characterization of the cell wall composition as well as the presence of the compounds being stored in the tissues.

Scanning Electron Microscopy (SEM)
For SEM analysis, fixed leaves were washed with 0.1M sodium phosphate buffer pH 7.2 and immersed in a 2.2-dimethoxypropane (DMP) solution.The samples were critical point dried (BAL-TEC CPD 030) (Gerstberger & Leins 1978), placed onto stubs and covered with a 10-15nm gold film using a BAL-TEC SPD 050 equipment.The material was analyzed with a JEOL JSM 6060 scanning electron microscope, under 10kV, at the Centro de Microscopia e Microanálise of UFRGS.
The resulting matrices were submitted to statistical tests using Pasw Statistics 18, SPSS (SPSS Inc., Chicago IL, USA).Simple descriptive statistics (mean, standard deviation, median, coefficient of variation and standard error) were calculated for each quantitative character.Quantitative character states, as well as mean and standard deviation values, are shown in Table 2. Characters were also evaluated for normality and homoscedasticity, while the significance of interspecific variability was estimated for each character using ANOVA (normalized morphometric characters) and the Kruskal-Wallis's test for the non-normalized morphological characteristics (Table S2).Characters were removed from the matrix when significant variability was not detected (p>0.05).
Two multivariate approaches were employed to identify morphological discontinuities among taxa and to detect characters that could contribute to the delimitation of the Dyckia selloa complex.
The first approach consisted of a Principal Coordinate Analysis (PCoA) of all analyzed characters, both quantitative and qualitative.In this analysis, the characters were discretized into categories and standardized by the Normatization Method -Object Principal.On a second approach, the quantitative characters were subsequently submitted to Discriminant Analysis (DA) to discriminate groups/species of Dyckia.For this analysis, species were defined by probabilities, estimated from the size of the established groups (case-wise test), through the covariance matrix and Fisher's linear predictive model.Thus, the functions that best discriminate the sampled groups were obtained (Table S3).

Morphological and anatomical characters
The analyzed species possess rosette phyllotaxis with succulent, rigid, lanceolate leaf blades ranging from erect to revolute.Aspects of leaf morphology can be seen Figures 1, 2 and Fig. 3, including details of external morphology, morphology in transverse section, and details of spines and epidermis trichomes (Table 2).
Leaf length exhibited great variation among the analyzed species, with D. rigida having the longest at 1002 mm ± 184.9 (mean ± SD) (Fig. 2M) and D. choristaminea the shortest at 99 mm ± 23.55 (Fig. 2Y).Analysis of the leaf in transversal section allowed the identification of different shapes, these shapes could be elliptical (Figs.1F, 2Z), narrow elliptical (Figs.1N, 1V, 1Z, 2B, 2F, 2J, 2R, 2V), short obovate (1J, 1R) or linear (2N, 2D').Leaf blade color was found to vary from green to grayish-green and whitish and the blades were covered with peltate scales on both sides (Fig. 1K).The leaf margins possess grayish-green (Fig. 1K, 2G, 2S, 2W), brown (Fig. 1G, 1S, 2K, 2O, 2A') or black spines (Fig. 1C, 1O, 1W, 1A', 2C, 2E', 3A), that are often rigid, with the exception of D. delicata, which develops flexible spines (Fig. 1K).Undulations are present on the abaxial leaf surface in transverse section, forming a costal zone and, consequently, establishing intercostal zones (Fig. 3B-F).Costal zone possesses only ordinary epidermal cells whereas the intercostal zone has also specialized cells: trichomes and stomata located together and aligned (Figs.3D-F and  3E).The leaves are hypostomatic with tetracytic stomatal complexes situated above the level of other epidermal cells and guard cells with equivalent periclinal thickening (Figs.3D-F).The adaxial leaf surface is generally flat or slightly undulating, causing the formation of depressions in which there is a reduced number of layers of mechanical hypodermis and the insertion of peltate trichomes (Fig. 3G).All analyzed species have unistratified epidermis on both surfaces, with thickening of the anticlinal and inner periclinal walls of the cells, reduced lumen and the presence of silica bodies (Fig. 3D, Table S1).
Water-storage, proved with apoplastic tracer (Fig. 3R-S), and armed parenchyma cells are present in the region of the mesophyll between the vascular bundles up to the abaxial side (e.g.Fig. 3Q).The greatest and least thickness of both tissues in this region are for D. polyclada (11.2 µm ± 2.4) (Fig. 2R) and B. antiacantha (3.6 µm ± 0.9) (Fig. 2D'), respectively.In these species, the cells of the waterstorage parenchyma are polygonal, which may or may not be anticlinally-elongatedand extending to the costal region of the abaxial side (Fig. 3Q).The armed cells of the armed parenchyma may have short or long cells extending to the intercostal area of the abaxial face of the epidermis.The types of the armed parenchyma cells are star-shaped armed parenchyma (Fig. 3V), lobe-shaped armed parenchyma (Fig. 3U), and intermediate armed parenchyma (Fig. 3T).
Among the analyzed species, D. rigida (Fig. 2N) was found to have the greatest number of vascular bundles in transverse section with 83 units while D. choristaminea (Fig. 2Z) was found to have the least with 23 units.These vascular bundles, inserted within the chlorenchyma of the mesophyll, are distributed in a linear plane through the mesophyll and are formed by collateral bundles (Fig. 3H) with one or two caps of fibers, depending on the caliber.

Multivariate analyses Principal Coordinates Analysis (PCoA)
The results of the PCoA demonstrated that the morphological variation observed for the analyzed species could be adequately summarized using the first two dimensions (Fig. 4).The species formed a large grouping, except for the outgroup taxa D. choristaminea and B. antiacantha, which were discriminated from the species of the D. selloa complex.The first dimension (D1) represented 34.5% of the total variance (Cronbach's alpha=0.925)and the second dimension (D2) 20.1% (Cronbach's alpha=0.835),for a total of 54.6% of the total observed variation.
The values obtained for D1 and D2 showed that 21 of the 35 characters used in the analysis have high discrimination values (discrimination measures: DM > 0.7) in at least one of the first two dimensions, with the DM of seven characters being greater than 0.7 in both dimensions (Table  1. Acta Botanica Brasilica, 2023, 37: e20220079 S3).The qualitative characters that contributed the most to both D1 and D2 were spine color, non-vascular fibers, leaf shape in transverse section, chlorenchyma projections, shape of armed parenchyma cells in abaxial side and external leaf blade morphology.The analysis revealed ten morphometric characters with high discrimination values including leaf length and width, leaf height in transversal section, proportion of total chlorenchyma length x width, total length of water-storage parenchyma, length of 2nd stratum of water-storage parenchyma, total height of chlorenchyma, length of armed parenchyma cells, height of vascular bundle, and abaxial water-storage parenchyma and number of bundles in transverse section.The set of analyzed characters had high discrimination values capable of delimiting species from the outgroup taxa D. choristaminea and B. antiacantha.

Discriminant Analysis (DA)
Discriminant analysis performed with the set of 14 quantitative characters showed that 92.6% of the total variation can be explained by the first two discriminant functions, Discriminant Function 1 (DF1=75.0%)and Discriminant Function 2 (DF2=17.6%),as shown in the ordering Table S3.The DA scatter plot revealed the separation of the outgroup taxon B. antiacantha accompanied by D. rigida of the D. selloa complex (Fig. 4).Despite this subtle separation, the other analyzed species were grouped in the D. selloa complex, including the outgroup taxon D. choristaminea.
The two discriminant functions were strongly correlated (canonical correlation: DF1 = 0.951; DF2 = 0.889) and highly significant (DF1 to DF2: Wilks' Lambda = 4.55× 10 -54 ; DF2 to DF1: Wilks' Lambda = 6.87 × 10 -79 ).Based on the correlation coefficients obtained between each character and the discriminant functions, the highest absolute correlations were detected for four variables: DF1, leaf length and for DF2, leaf width, proportion of total chlorenchyma parenchyma length x width and armed parenchyma cells (Table S3).Although the set of characters has a high discriminant value, which enabled the separation of B. antiacantha and D. rigida from the other species of Dyckia sampled, the species from the D. selloa complex were grouped with the other outgroup taxon D. choristaminea.

Morphological and anatomical characters
The analyses carried out here found morphological characteristics that are shared between species of the D. selloa complex and other species of Bromeliaceae, such as: spine color, presence of trichomes on both sides of the leaf, lignified mechanical hypodermis, layers of water storage parenchyma, cells morphology in armed parenchyma cells, and chlorenchyma disposition in the mesophyll (Tomlinson 1969;Smith & Downs 1974;Reitz 1983;Benzing 2000;Proença & Sajo 2007;Voltolini et al. 2009;Santos-Silva et al. 2013;Krapp et al. 2014;Pinangé et al. 2016;Schütz et al. 2016).Dyckia possesses leaf anatomical characteristics common to other genera of the xeric clade of Pitcairnioideae, such as the presence of mechanical hypodermis and waterstorage parenchyma on both sides (Givnish et al. 2007;2011;Gomes-da-Silva et al. 2017).
Species of the D. selloa complex exhibit great plasticity in external leaf morphology, ranging from lanceolate to revolute with intermediate morphologies.In Dyckia, rosettes do not form cisterns, which makes for a larger contact surface to acquire water and nutrients from the xeric environment of these species (Smith & Downs 1974;Reitz 1983;Benzing 2000;Proença & Sajo 2007;Dettke & Milanez-Gutierre 2008;Voltolini et al. 2009;Aoyama et al. 2012).The absence of cistern in Dyckia indicates a variation related to environmental conditions compared to other xeric members of Bromeliaceae.
Analysis of leaf morphology and anatomy showed the occurrence of spines in Dyckia through visualization of the continuous vascular system (Tomlinson 1969;Reitz 1983;Benzing 2000).Aculeae and spines are similar, pointed elements on the surface of plant organs.However, an aculea is an exclusively epidermal structure, while a spine can result from the modification of a branch, leaf, stipule, or root.Such structures are vascularized and firmly attached to the plant body (Ferri et al. 1978).Spines are normally found on leaf margins within Bromelioideae and Pitcairnioideae and are rigid and organized along the entire leaf margin (Tomlinson 1969).Although rare, aculeus have been recorded in Bromeliaceae, on the leaves of Aechmea calyculata Baker (Favretto & Geuster 2017).Species of the D. selloa complex have extraordinarily strong spines with a horny, black-brown constitution and the apex of the leaf blade can end in a spine, in agreement with Reitz (1983).As for their color, spines in Dyckia are usually white or the same color as the leaf (Smith & Downs 1974;Reitz 1983).The extensive sampling in this study showed variation in the color of these structures, enabling the separation of the D. selloa complex into two groups: one containing D. agudensis, D. aff.maritima, D. myriostachya, D. nigrospinulata, and D. rigida, which have black spines; and one containing D. alba, D. delicata, D. domfelicianensis, D. hebdingii, D. polyclada, D. retroflexa, D. selloa and D. tomentosa, which have brown spines.Many terrestrial bromeliad species invest in spines as a mechanical defense against herbivores (Benzing 2000).This is an important feature to be included not only in analyses for the circumscription of the D. selloa complex itself, but in analyses of other taxonomic studies as well.
Another interesting character is the position of stomata.The present results show that stomata in Dyckia are located above the level of the epidermis.This character has already Acta Botanica Brasilica, 2023, 37: e20220079 plants less palatable to predators and thus prevent herbivory (Mauseth 1988).Despite the presence of raphids being used in these taxonomic studies, our analysis did not show that it is an important character to discriminate species.
The vascular bundles of the D. selloa complex species are collateral and with one or two caps of fibers, as reported for other Dyckia species (Santos- Silva et al. 2013).This character may also be useful in characterizing groups of species, especially considering the number of layers of these caps.All Dyckia species do not present non-vascular fiber bundles when compared to B. antiacantha which has non-vascular fiber bundles, constituting a possible character for future taxonomic analysis.

Multivariate analysis and delimitation of the Dyckia selloa complex
The scatter plot obtained from the first two PCoA axes, incorporating all analyzed species, clearly shows that the D. selloa complex is morphologically distinct from the other taxa.The selected outgroup taxa, B. antiacantha and D. choristaminea, were discriminated from the other species by the first two axes of PCoA, with the characters most contributing to this delimitation being leaf length and spine color.The scatter plot also shows the proximity of D. choristaminea to the sampled species of the D. selloa complex, especially D. hebdingi and D. delicata, resulting from shared characteristics related to spine color (such as brown spines), upturned leaves and trichomes on both leaf sides.Dyckia aff.maritima, D. nigrospinulata, D. myriostachya and D. rigida group closely on the scatter plot, possibly because they share characters related to black spine color and leaf length.
The scatter plot of the first two discriminant functions clearly shows that the D. selloa complex groups with the outgroup taxon D. choristaminea.On the other hand, B. antiacantha and D. rigida appear discriminated from the other species in the right portion of the scatter plot, mainly by DF1, which is based mainly on characters such as leaf length and proportion of water-storage parenchyma.Despite the high values of discrimination and correlations obtained, the discriminant function analysis was not able to delimit the D. selloa complex using morphometric data.
From data observed in this work, combined with the incongruence between the multivariate analyses in the delimitation of species belonging to the D. selloa complex, in addition to the distinct seminal rudiment data obtained in Breitsameter (2017), we suggest maintaining the D. selloa complex based on morphological qualitative data, with the possibility of dividing the species into two groups.The first group would consist basically of the same species group treated like D. maritima complex sensu stricto (except for D. agudensis and D. retroflexa) of Büneker et al. (2015), consisting of species morphologically similar to D. maritima that have as main feature the black leaf spines.This first group we observe is formed by the species D. agudensis, D. aff.maritima, D. myriostachya, D. nigrospinulata and D. rigida.
The second group would be composed of the species D. alba, D. domfelicianensis, D. selloa, D. retroflexa, D. polyclada and D. tomentosa, for which a new name could be created to contrast to with D. maritima complex sensu stricto, as established before, showing that there are at least two species groups within the D. selloa complex.The fact that the species D. delicata and D. hebdingii share very similar leaf morphoanatomical characteristics with the outgroup taxon D. choristaminea, show their distinctions between the other species of the complex.
Choosing the optimal methodological approach is essential for identifying the boundaries between sets of species and inferring the number of species in a complex (Rieseberg & Burke 2001;Sites & Marshall 2003;De Queiroz 2007).Morphological characters are the original tools used by taxonomists to identify and discriminate species (Cronquist 1981;Dahlgren et al. 1985).However, many studies focus on few characteristics, with no attention to morphological variation in a morphometric and multivariate context.The past few decades have seen researchers develop several methods for recognizing new species or testing species hypotheses (Wiens 2007;Naciri & Linder 2015).However, these tools for species delimitation often involve expensive molecular and computational methods that are demanding of specialized work.In this sense, the morphological and anatomical approaches applied in the present study, mainly based on qualitative characters, proved to be robust for delimiting the D. selloa complex.
The analyses carried out here indicate that characters related to leaf morphology and anatomy can be useful for delimiting species or groups of species, but they should be applied with caution in taxonomic decisions involving the genus Dyckia.Characters related to the armed parenchyma cells shape and spine color proved to be important for the D. selloa complex, and so we indicate their importance for future studies.There also remains a need to include different technical approaches -including reproductive structures and palynology -that make analyses more sensitive and enable the delimitation of species complexes in Dyckia.

Figure 1 .
Figure 1.External and internal morphology of species of Dyckia: A-D -D.agudensis; E-H -D.alba; I-L -D.delicata; M-P -D.domfelicianensis; Q-T -D.hebdingii; U-X -D.aff.maritima; Y-B' -D.myriostachya.Evidencing in A, E, I, M, Q, U and Y -the general appearance of the plants; in B, F, J, N, R, V and Z -leaf shape in transverse section under stereomicroscopy without staining; in C, K, O, S, W and A´ -detail of adaxial leaf side showing spines and trichomes; in G -detail of abaxial leaf side showing spines and trichomes; and in D, H, L, P, T, X and B´ -detail of the adaxial leaf side and trichome arrangement.(cl = chlorenchyma; sp = spine; wp = water-storage parenchyma; ap = armed parenchyma; t = trichome), ** mechanical hypodermis in adaxial side, * mechanical hypodermis in abaxial side.Scale bar: C, F, G, J, K, N, O, P, S, V, W, Z and A' = 1 mm; B, D, H, L, T, R, X and B' = 0.5 mm; E, I = 2 cm; M, Q, U = 4 cm; and Y= 1 cm.

Figure 2 .Figure 3 .
Figure 2. External and internal morphology of species of Dyckia and Bromelia A-D -D.nigrospinulata; E-H -D.polyclada; I-L -D.retroflexa; M-P -D.rigida; Q-T -D.selloa; U-X -D.tomentosa; Y-B´ -D.choristaminea; C´-F´ -Bromelia antiacantha.Evidencing in A, E, I, M, Q, U, Y and C´ -the general appearance of plants; in B, F, J, N, R, V, Z and D´ -a leaf shape in transverse section under stereomicroscopy without staining; in G, O, S, and E´ -detail of adaxial leaf side showing spines and trichomes; in C, K. W and A´detail of abaxial leaf side showing spines and trichomes; in H, P and F´-detail of the adaxial leaf side and trichome arrangement; and in D, L, T, X and B´ -detail of the abaxial leaf side and trichome arrangement.(cl = chlorenchyma; sp = spine; wp = water-storage parenchyma; ap = armed parenchyma; t = trichomes.Scale bar: A, E, I, M, Q, U, Y and C´ = 2 cm; B, C, D, F, G, J, K, N, O, P, S, V, W, A´, D´, E´ and F´ = 1 mm; H, L, R, T, Z, X and B´= 0.5 mm.

Figure 4 .
Figure 4. Two-dimensional scatter plots obtained from morphological character analyses: on the left, graph of principal coordinates analysis (PCoA); on the right, discriminant analyses (DAs).The proportions of morphological variation captured by each of the two main dimensions of PCoA (D1 and D2) and the two discriminant functions (DF1 and DF2) are shown in the two scatter plots s, left and right, respectively.Colored circles indicate the analyzed species described inTable 1.

Table 1 .
Morphological and anatomical species sampled, with voucher information, number of registration in the Living Collection

Table 2 .
Mean ± standard deviation values for each morphological and anatomical character used in Dyckia species analyses and states for each categorical character.