This study was carried out in the southern portion of the Chihuahuan desert known as Queretano-Hidalguense semi-desert, in the wilderness area protected by the Regional Botanical Garden of Cadereyta de Montes “Ing. Manuel González de Cosío”, Querétaro, Mexico. Its geographic coordinates are 20°41’15.8”N, 99°48’17.7”W, with an elevation of 2,046 m a.s.l., the vegetation type is xerophytic crassicaulous scrub. The climate is semi-dry, temperate with summer rains (Köppen climate group BS1 kw (w) modified by García 2004). The average annual temperature ranges between 12 and 19°C and the average annual precipitation is about 550 mm (Chávez-Martínez and Hernández-Magaña 2009).

Opuntia streptacantha is an arborescent or shrubby plant, up to 4 m tall, its stems (cladodes) are flattened and racket-shaped, and the flowers are yellow or orange, 5–7 cm long. The fruits are 5 cm long and 3 cm wide, globose to obovoid and usually wine-coloured when ripe. The glochids are short, the pulp is red, and the seeds are 3.8–4.5 mm long by 2.6 mm wide (Bravo-Hollis and Sánchez-Mejorada 1978; Arias et al. 2012; Galicia-Pérez et al. 2023).

This species is endemic to Mexico and is commonly known as “cardón”, “cenizo”, “chaveño”, or “nopal cardón” (Arias et al. 2012), and is distributed in the states of Aguascalientes, Mexico City, Durango, Mexico, Guanajuato, Hidalgo, Jalisco, Michoacán, Nuevo Leon, Oaxaca, Puebla, Queretaro, San Luis Potosi, Tamaulipas, Tlaxcala, and Zacatecas (Hunt et al. 2006). It grows at elevations of 1,600 to 2,400 m a.s.l. (Arias et al. 2017).

Opuntia streptacantha is generally found in xerophytic scrublands and with other co-dominant Opuntia species, they form a vegetation type called “nopaleras” (Arias et al. 2012). According to Bravo-Hollis and Sánchez-Mejorada (1978), it is a wild species of remarkable value because of its edible fruits and stems.

Opuntia streptacantha populations contain individuals with two different floral colour morphs: with yellow flowers (YFM) and with orange flowers (OFM) (Fig. 1). The presence of these colour morphs appears to be consistent throughout the distribution range of the species. Historically, the first description of the colour of the flowers of this species was made by Schumann (1899), who mentioned that the flowers were yellow. However, Bravo-Hollis and Sánchez-Mejorada (1978) observed that the flowers vary from yellow to orange within a population.

Twenty mature cladodes (including those with lateral cladodes or reproductive structures such as buds, flowers, or fruits) were measured, comprising two cladodes per individual, from ten reproductive individuals of O. streptacantha for each floral colour morph. Similarly, 20 young cladodes (considering only lateral cladodes as young) were measured, also two per individual. The sampled individuals were checked to ensure that they had fruits, buds, or flowers, or a combination of them, and were approximately 3 meters tall. The parameters used in the work of Muñoz-Urias et al. (2008) were measured: cladode length (cm), cladode width (cm), maximum distance from the apical to the widest part (cm), maximum distance from the basal to the widest part (cm), number of series of areoles, areole size (mm), distances between areoles (mm), and distance between lines (mm) (Fig. 2). To determine differences in the spines of floral morphs, the type, colour, and number of spines on the central areole of three cladodes per individual were determined for each floral morph (n = 10) (López-Borja et al. 2017; Galicia-Pérez et al. 2023).

Thirty-three undamaged flowers were collected at the time of maximum flower opening and during peak flowering from different individuals of each flower colour morph of O. streptacantha and fixed in FAA (formaldehyde, alcohol, acetic acid) (Kiernan 2002). Following Martínez-Ramos et al. (2017), 17 morphological characters were measured to the nearest mm: corolla aperture set in FAA, perianth segment length, total flower length, pericarp length, pericarp width, distance between anther and stigma, stigma width, stigma length, style length, longest stamen length, shortest stamen length, equatorial diameter of the ovarian chamber, polar diameter of the ovarian chamber, and we counted the number of stamens, number of ovules, number of lobes, and number of pollen grains in one anther (Fig. 3).

Morphometric data of cladodes and flowers were tested for differences between floral morphs using Generalized Linear Model (GLM) with Poisson distribution for discrete counts and t-tests for continuous variables. For the spines, a paired t-test was carried out.

We collect two fruits from ten different reproductive individuals per flower colour morph (n = 20). We assessed fruit diameter (mm), fruit length (mm), number of spiral series, and number of seeds (López-Borja et al. 2017). A paired t-test for continuous variables and a GLM with Poisson distribution for discrete variables were used to find differences between morphs. In addition, the external colour of the fruit, colour of the pulp, colour of the glochids using the HTML colour code, and the shape of the fruit were taken based on the classification given by Moreno (1984).

A sample of ten seeds was randomly selected from each fruit (n = 200 seeds per floral morph), photographed, and measured for size with length and width of each seed in mm using Adobe Photoshop CS6. A paired t-test was used to compare between morphs.

All statistical tests were performed in R v.4.2.2 (R Core Team 2022) with the packages stats v.4.2.2 (R Core Team 2022) and emmeans v.1.10.1 (Russell 2021).

Two tests were performed to compare flower, cladode, fruit, and seed characteristics between morphs (Sokal and Sneath 1963; Cuadras 1981). The first test was a principal component analysis (PCA), to reduce the variables to those that would give us the most taxonomic information. The morphological characters of all the previously mentioned measured structures were considered, the analysis was performed in R, with the packages FactoMinerR v.2.10 (Le et al. 2008), factorextra v.1.0.7 (Kassambara and Mundt 2020), psych v.2.4.3 (Revelle 2020), and Factoshiny v.2.5 (Vaissie et al. 2020). In addition, we performed a linear discriminant analysis, including only the relevant flower and cladode characters detected by PCA. In this analysis, data from two more species were added: Opuntia tomentosa Salm-Dyck and Opuntia cantabrigiensis Lynch (Galicia-Pérez et al. 2023), as both species are found in the same study site and present a very high flowering synchrony with the floral morphs of O. streptacantha (Martínez-Ramos 2019; Martínez-Ramos et al. 2024), the analysis was performed in R with the lda function of the stats v.4.2.2 package (R Core Team 2022).

Reproductive phenology (flowering and fruiting) was registered, taking monthly observations (April 2018 to March 2019) for ten individuals of each floral morph. The data were analysed with circular statistics to determine the flowering and fruiting peaks (Morellato et al. 2010), and the Rayleigh uniformity test (Zar 1999; Mendoza 2020) was calculated to identify if the distribution of the phenophases was uniform, and the non-parametric Mardia-Watson-Wheeler test (Batschelet 1981) was performed to determine differences between flowering and fruiting. Analyses were carried out in R with the circular v.0.5-0 package (Agostinelli and Lund 2022).

Two indices were evaluated to determine the flowering synchrony of the colour morphs. The Marquis (1988) index was evaluated, which considers the number of open flowers per census and the proportion that these flowers represent with respect to the total number of flowers, following the formula below:

S = ∑ t = 0 n x t ∑ t = 0 n x t × P t

where, S is the degree of synchrony, x_t the number of open flowers per census,

x t ∑ t = 0 n x t

is the proportion of open flowers to the total number of flowers, and Pt represents the proportion of the censused individuals at flowering during time t.

Flowering synchrony between YFM and OFM was calculated with the index of Mahoro (2002), modified by Osada et al. (2003). For the modified version, the relative number of open flowers in each individual at an interspecific level (in this case, between YFM and OFM) is considered, following the formula below:

s i = 1 2 2 - ∑ i = 1 n y A , j - y B , j

where s_i is the degree of synchrony of species A with species B, Y_A,j is the ratio of flowers in morph A, and Y_B,j is the ratio of flowers in morph B.

Both indexes take values from 0 to 1, where a value close to one represents perfect synchrony and a value close to zero represents asynchrony.

The morphometry of old cladodes differed significantly between YFM and OFM in two variables: cladode length (t = -2.62, p = 0.01) and cladode width (t = -2.23, p = 0.03), with OFM being the largest (Table 1). No significant differences were found for young cladodes. Both flower colour morphs have ovate cladodes and two types of spines, straight and subulate. The spines have a yellow or white colouration and the number of spines per areole in both morphs was similar (mean ± standard error; YFM = 3.63 ± 0.49, OFM = 3.93 ± 0.52, χ² = 0.3569, p = 0.55).

Flowers were actinomorphic in both morphs of O. streptacantha (Fig. 1), but we found significant differences in most of the characters (Table 2), with higher values for YFM. In contrast, the gynoecium was very similar for both morphs.

Both floral morphs had ovate fruits, a magenta pericarp with wine-coloured pulp, and opaque-golden glochids. Fruit length was the only difference, with YFM having longer fruits (Table 2). Both floral morphs have funiculate seeds, with an oval to amorphous shape and light brown colour. Significant differences in seed size (Table 2) were found between the morphs.

Of the morphological characters, those showing significant differences between floral morphs for PCA and discriminant analysis were considered. PCA shown seven principal components that explain 80% of the total variation. The first component explains 28.37% (total flower length), the second component 17.50% (number of stamens), the third component 11.01% (distance between anther and stigma), the fourth component 7.20% (number of pollen grains and style length), the fifth component 6.58% (equatorial diameter of the ovarian chamber), the sixth component 5.71% (width of the pericarpel), and the seventh component 4.05% (number of areole lines), considering only the first two components explain 45.87% of the total variation (Fig. 4; Supplementary material 1). It is important to emphasize that six of the seven components are flower morphometric variables.

Linear discriminant analysis explained 89.28% of the variation in the first two linear discriminant functions. Opuntia cantabrigiensis was completely separated from the other species, O. tomentosa was grouped with YFM, and OFM was almost separated from O. tomentosa but showed a small overlap with YFM (Fig. 5). The analysis was able to correctly assign 92% of the individuals within species. YFM had the fewest correctly classified individuals (Table 3).

Reproductive phenology differed between the floral morphs. Flowering for YFM was significantly (r = 0.9566, p = < 0.001) concentrated in four months, from March to June, with peak flowering in April (Fig. 6A). On the other hand, the OFM showed a significant (r = 0.9443, p = < 0.001) flowering season of five months, from February to June, with a single peak in flowering in May (Fig. 6B). The flowering patterns of the floral morphs were significantly different (W = 43.686, p < 0.0001).

Fruiting in YFM was significant (r = 0.9023, p = <0.001) concentrated in a period of five months (June to October), with peak fruiting in July (Fig. 6C). On the other hand, in the OFM, showed a significant fruiting season (r = 0.8749, p = <0.001), spanning seven months, from May to November, with peak fruiting in June (Fig. 6D). The fruiting patterns of the floral morphs were significantly different (W = 206.5, p < 0.0001).

According to the Marquis (1988) index, in the O. streptacantha population we studied, flowering synchrony is very high for both YFM (S = 0.94, EE = 0.25) and OFM (S = 0.91, EE = 0.21). For the Mahoro (2002) index modified by Osada et al. (2003), floral synchrony between YFM and OFM is also high (S = 0.86).