Anatomical and morphological spine variation in Gymnocalycium kieslingii subsp. castaneum (Cactaceae)

Abstract Although spine variation within cacti species or populations is assumed to be large, the minimum sample size of different spine anatomical and morphological traits required for species description is less studied. There are studies where only 2 spines were used for taxonomical comparison amnog species. Therefore, the spine structure variation within areoles and individuals of one population of Gymnocalycium kieslingii subsp. castaneum (Ferrari) Slaba was analyzed. Fifteen plants were selected and from each plant one areole from the basal, middle and upper part of the plant body was sampled. A scanning electron microscopy was used for spine surface description and a light microscopy for measurements of spine width, thickness, cross-section area, fiber diameter and fiber cell wall thickness. The spine surface was more visible and damaged less in the upper part of the plant body than in the basal part. Large spine and fiber differences were found between upper and lower parts of the plant body, but also within single areoles. In general, the examined traits in the upper part had by 8–17% higher values than in the lower parts. The variation of spine and fiber traits within areoles was lower than the differences between individuals. The minimum sample size was largely influenced by the studied spine and fiber traits, ranging from 1 to 70 spines. The results provide pioneer information useful in spine sample collection in the field for taxonomical, biomechanical and structural studies. Nevertheless, similar studies should be carried out for other cacti species to make generalizations. The large spine and fiber variation within areoles observed in our study indicates a very complex spine morphogenesis.


Introduction
Spines may be considered one of the most characteristic morphological structures of the Cactaceae family. Cactus spines are the modifi ed bud scales of an axillary bud, originating from primordia which are morphologically indistinguishable from the leaf primordia (Mauseth 2006). Spines contain just two cell types, which never occur in long-shoot leaves of cacti: libriform fi bers and sclerifi ed epidermis cells (Mauseth 2006). Th e cactus spine epidermis lacks stomata. In a few species, some spine epidermis cells elongate outward as trichomes (Sotomayor and Arredondo 2004, Mauseth 2006, Řepka and Gebauer 2012. Th e epidermis can be continuous, divided into single cell elements or transversely fi ssured, and such fi ssures extend deeply into the underlying sclerenchyma (Barthlott 1979).
Spines are not only a lifeless part of the plant body but have several important functions. Th ey provide defense against herbivores (Norman andMartin 1986, Gibson andNobel 1986), protect the sensitive meristems from freezing temperatures (Loik andNobel 1993, Mauseth 2006) and shade the plants to avoid temperature stress (Gibson andNobel 1986, Mauseth 2006). Consequently, each cactus must evolve its spine coverage pattern in order to maximize its photosynthetic effi ciency within its own habitat (Menezes et al. 2015). Th e spine surface of Opuntia is also constructed as an effi cient system to collect fog and to drive water droplets towards the spine base, where they are absorbed (Ju et al. 2012). Moreover, the areole position on the plant body may reveal past physiological and climatic variation since new spines develop on the top of the cactus body whereas the oldest spines are situated in the basal part (English et al. 2007). Th e exception are tree-like species of Opuntia and Quiabentia, which continue producing spines from their lower areoles even in old age (Rowley 2003).
Th e spine diversity within the family is truly spectacular and spine anatomical and morphological traits are useful tools for taxonomists (e.g. Hunt et al. 2006, Mosco 2009, Řepka and Gebauer 2012. Although, there is evidence that spine development is infl uenced by genetic (Mihalte and Sestras 2012) and environmental conditions as water availability or solar radiation can modifi ed spine growth (e.g. Peharec et al. 2010, Menezes et al. 2015. Nevertheless, to the best of our knowledge, there is no prior study analyzing spine and fi ber structure variation within areoles, individuals or populations, with the exception of variation in spine number and length (e.g. Schmalzel et al. 2004, Hunt et al. 2006, Baker and Butterworth 2013, Menezes et al. 2015. Although the variation in spine anatomical and morphological traits within a population or species is assumed to be large (Mauseth 2006), there are studies where only 2 spines were used for Turbinicarpus species comparison (Mosco 2009) or for the assessment of spine stiff ness (Schlegel 2009).
Spine and fi ber variation within areoles and individuals was studied in one Gymnocalycium kieslingii subsp. castaneum (Ferrari) Slaba population with the intent to solve three questions: (1) does areole position on the plant body play an important role in spine and fi ber variation?; (2) are spine and fi ber traits less variable within an areole than between areoles and individuals?; (3) how many spines need to be collected for an analysis of specifi c traits? Our results will provide useful information for spine sample collection in the fi eld for taxonomical, biomechanical, physiological or structural studies.

Plant material
A single representative of the nominate subgenus of Gymnocalycium, G. kieslingii subsp. castaneum was chosen for the study. Th is is an endemic taxon of the Argentinian province of La Rioja, which is taxonomically clear with relatively low morphological variability. It grows in 13 populations occupying fairly narrow ecological niche. It grows on the poor and highly permeable sandy soils without humus, blended with skeleton basement rock, generated by the disintegration of granitoids on the slopes of the Sierra de Velasco Mts. Th e climate is semi-arid and the mean annual precipitation is 360 mm. Th e mean annual temperature is 20 °C with 2,800 hours of sunshine (http://www.arquinstal.com.ar/atlas/datos/larioja.html). Plants are uniformly found under/or outsides of Larrea cuneifolia Cav. in the phytogeographical district Monte (Cabrera 1976) at altitudes of 1250-1550 m. Distribution area is relatively small (cca 760 km 2 ) compared to other species of the subgenus. It lies between city of Aimogasta in the north and city of Villa Sanagasta in the south.
Th e plant usually forms fl attened spherical bodies 60-90 mm in diameter. It rarely achieves a greater height than width (Till 1990). Th e plant material was collected from one population at the sandy depression of Bolsón de Huaco, SSW of the village of El Huaco in the northern part of the La Rioja Province, at altitude of 1260 m (67°07'S; 29°22'W). Fifteen plants were randomly selected on a ca. 700 m line along the road (north-south), at distances of 50 m from each other. From each plant, one areole from the basal, middle and upper part on the northern face of the plant body were sampled (i.e. 45 areoles in total). Additional areoles were sampled from 5 plants at each examined position for scanning electron microscopy (i.e. 15 areoles). Only fully developed areoles were collected.

Scanning electron microscopy (SEM)
Th e whole areole was mounted on specimen stubs, sputter-coated with gold, and observed with high-vacuum SEM using a VEGA TS 5130 instrument (Tescan, Czech Republic) operating at 15 kV. Images of the whole areole and detailed images of the spine surface in the middle and top spine parts were made. From these images, epidermis characteristics, shape of epidermis cells, presence and type of trichomes and presence of fi ssures were determined.

Light microscopy
A 5% solution of hydrochloric acid was used to soften spines before sectioning. In this solution, two days of soaking was suffi cient to soften the spines for anatomical analysis. Th e spine length (S l ) was measured before making cross-sections. Handmade crosssections were taken from the spine base and were examined under a bright fi eld microscope (Olympus BX51, Olympus Czech Group, Czech Republic) at magnifi cations up to 400× and were photographed using a digital camera (Olympus E-330, Olympus Czech Group, Czech Republic) connected to a computer with the QuickPhotomicro 2.3 software (Promicra, Czech Republic). Spine width (S w ), spine thickness (S th ), spine cross-section area (S a ), spine circumference (S c ), fi ber maximum and minimum diameter (F max and F min ), and cell wall thickness (CW th ) were measured using the ImageTool 3.00 software (UTHSCSA, USA) ( Fig. 1). Ten fi bers were measured in each spine. Only the largest fi bers were measured as fi bers at the end taper. Spine and fi ber roundness (S r and F r , respectively) were defi ned as the S w /S th and F max /F min ratio, respectively, with a ratio of 1 denoting a perfectly round cross-section and larger ratios indicating a more ellipsoidal shape. In total, 245 spines were examined.

Statistical analysis
Th e fi rst step was the calculation of fundamental descriptive characteristics using linear mixed eff ect models (LME). In these models, all of the traits come from a nested design; therefore, we used LME to avoid the problem of pseudoreplication (Hurlbert 1984;Pekár 2012). In the LME analyses, traits were treated as factors with fi xed eff ects, and individuals and areoles were treated as factors with random eff ects. LMEs were fi tted using the LME function in the NLME library of the R statistical program (R Development Core Team 2015). Results from the fi rst step were used for other calculations. Th e infl uence of spine position on the plant body was analyzed using one-way nested ANOVA. Plant position was treated as a factor with fi xed eff ects, and individuals and areoles were treated as factors with random eff ects. In the second step, minimal sample size (N) was calculated according to the following equation: where σ is the assumed standard deviation (SD) for the group, the (t 1-α/2 (n-1)) value is the quantile of the Student's t-distribution for n-1 degrees of freedom and D is the desired margin of error. Th e interval limits for minimum sample size were taken as 10, 15 and 20% diff erences of the mean. Only spines sampled from the middle part of plant body were used for minimum sample size calculation. Calculations were performed in the R software environment (R Development Core Team 2015) and STATISTICA v. 10 (StatSoft, USA).
All acronyms, abbreviations and symbols are defi ned in Table 1.

Spine surface structure in SEM
Spines in the basal part of the plant body were either completely covered with mineral deposition so that the surface structure was not recognizable or the surface was only partially visible (Fig. 2a, b). In rare cases, epidermal cells were still obvious (Fig. 2c). Spine tips were usually damaged, bent and deformed (Fig. 2d).
In the middle part of the plant body, the spine surface was only slightly damaged. Th e epidermal cells were clearly visible (Fig. 2e) and were only missing in a few cases. Th ey were generally rectangular in shape (major to minor axis diameter ratio 2:1). Epidermal cells were usually sharply bent upward in the upper part (Fig. 2e) and arranged in regular transverse rows with diff erent directions. In rare cases, deep fi ssures were observed.
Epidermal cells of spines in the upper part of plant body were very clearly visible and undamaged (Fig. 2f, g). Epidermal cells were irregular in both shape and size. Th ey were fl at on the spine tips and started to bend slightly to sharply at some distance from the spine tips (Fig. 2g). In some cases, they were also bent lengthwise in their central part (Fig. 2h). Th ey were usually arranged in regular rows. Th e rows were wavy in shape and ran slightly upwards. No fi ssures were observed.

Spine variation between individuals
Spine length (S l ) ranged from 3 to 16 mm and spine cross-section area (S a ) from 0.21 to 2.81 mm 2 (Table 2). S l and S a had the largest coeffi cient of variation (CV) between individuals (Table 3). In contrast, spine roundness (S r ) had the lowest CV (Table 3). In most cases, the shape was near-circular, as the mean S r was 1.25 (Table 2). Th e mean fi ber maximum diameter (F max ) and fi ber minimum diameter (F min ) were 13.7 and 10.03 μm, respectively (Table 2). Th e studied fi ber traits were less variable than spine traits between individuals (Table 3). Fiber roundness (F r ) was the least variable   (Table 3). Th e mean F r value was 1.38 (Table 2), indicating fi bers with a slightly ellipsoid shape. On the other hand, cell wall thickness (CW th ) had the highest CV of the fi ber traits (Table 3), ranging from 1.5 to 6.6 μm ( Table 2).

Spine variation between areoles and within the plant body
Th e fi ber traits were less variable than spine traits between areoles (Table 3). Although the CV of spine and fi ber traits between areoles was lower than that between individuals, it was higher than 30% in most cases (Table 3). Th e most and the least variable spine and fi ber traits between areoles were the same as between individuals. Th e CV of S a and S l was the highest, whereas the CV for S r and F r was the lowest (Table 3). Spine and fi ber traits were similar in the base and middle part of plant body. However, the upper part was signifi cantly diff erent from the other parts for almost all studied spine and fi ber traits (except S l , S r an F r ) ( Table 4). In general, studied traits were by 8-17% higher in the upper part than in the lower parts (Table 4).

Spine variation within areoles
Th e CV of spine traits within an areole was lower than between areoles and individuals (Table 3 and Suppl. material 1). Nevertheless, a CV value higher than 15% was observed for S a (91% of sampled areoles) and for S l (60% of sampled areoles) (Suppl. material 1). Th e most and the least variable spine traits within an areole were almost the same as those between the areoles. On the other hand, the variation in fi ber traits within an areole diff ered from the variation of fi ber traits between the areoles. Contrary to the variation between areoles, F r was the most and CW th the least variable trait within an areole. A CV value higher than 15% was found in most cases for F r (62% of sampled areoles) and for F max (44% of sampled areoles). On the other hand, a CV of CW th higher than 15% was only found for 7% of the sampled areoles. Spines within an areole were distributed only marginally with radial arrangement (Fig. 3). All spines were straight or slightly curved to the body. Th e lower one pointing downward Th e number of spines per areole range from 3 to 7 spines and the most frequent number of spine per areole was 5 (51% from all areoles) (Suppl. material 1). Only 11% from all areoles had 3 and 4 spines.

Minimum sample size
In general, calculated minimal sample sizes corresponded with the CVs of the studied traits. To study spine traits with 10% diff erences in the mean value, we should measure at least 52 spines (Table 5). However, if we studied only S a or S c , 25 spines would be needed and even fewer spines would be needed for S r (Table 5). In the case of fi ber traits, the most variable trait was CW th , for which 70 spines should be used to obtain results within 10% diff erences in the mean value (Table 5). To study fi ber diameter, 43 spines should be used and no more than 6 spines are needed for F r . Th e minimal sample size for the studied parameters would be on average 3.7 times lower, if 20% diff erences in the mean value were used (Table 5).

Spine surface
In the present study, the spine epidermal cells were usually bent upward, but fl at shapes were also observed. Epidermal cells were usually arranged in regular transverse rows. Th e spine epidermis and mesophyll of several cacti have deep fi ssures, created during normal development (Mauseth 2006). Th ese fi ssures are believed to play an important role in water absorption, but it has not yet been investigated what eff ect this process may have on the overall water balance of the plants (Mosco 2009). In G. kieslingii subsp. castaneum such fi ssures had only 9% of observed spines. Trichomes observed in other cactus species (Mauseth 2006, Řepka andGebauer 2012) were also absent in G. kieslingii subsp. castaneum. It has been reported that diff erences in surface structure are related to age: young spines (situated in the upper part of the plant body) of Mammillaria scrippsiana var. armeria and Echinopsis sp. have a smooth surface, whereas older spines (situated in the lower parts of plant body) show a broken surface. We did not observe such diff erences, which may be specifi c characteristics of a particular genus of cacti. However, there were diff erences in spine surface damage and visibility. Th e spine surface was more visible and less damaged in spines from the upper part than in the lower parts. In the basal part, the spine surface was almost completely covered with mineral deposition (Fig. 2a). Th us, spines situated in the upper part of plant bodies are the most appropriate for surface analysis and taxonomic use.

Spine variation between individuals and areoles
Spines develop from lateral buds (areoles) and vary considerably across species in number, length, width and thickness (Mauseth 2006, Mosco 2009). Th is is an extreme form of leaf modifi cation (Mauseth 2006). Th e number of spines per areole of G. kieslingii subsp. castaneum is described as ranging from 5 to 7 (Ferrari 1980). In the present study, areoles with 3 and 4 spines were also found in 11% from all areoles. Th e spines shape was the least variable trait and it was mostly circular. Th is is in accordance with Mauseth (2006), who mentioned that cactus spines are frequently circular in crosssection. On the other hand, S l and S a were the most variable spine traits. In another study, spine length was also found to be the most variable trait in 30 Aylostera and Rebutia (Cactoideae) hybrids (Mihalte and Sestras 2012). It has already been found that spine variation occurs even within a single species, refl ecting environmental conditions during cactus growth (e.g. Nobel 1988, English et al. 2007, Menezes et al. 2015. For example, S l was found to be positively correlated with rainfall (Menezes et al. 2015). It is obvious that longer and thicker spines will require more photosynthates for its development. On the other hand, such a spine increment would reduce the interception of photosynthetic active radiation, which would then reduce photosynthetic productivity (Nobel 1983). However, abundant spines shade the photosynthetic cortex from intense insolation and UV radiation to avoid high-temperature extremes (Loik 2008). Th us, the large variation of S l and S a found in our study may be partly explained by a sensitive plant regulation system that balances between positive and negative spine functions. However, the factors controlling morphogenesis in the basal meristems of spines are still unknown (Mauseth 2006) and the identifi cation of genes and their expression will be an important step towards our understanding of the spine development.
Two main fi ber shapes in spine cross-sections (i.e. folded and pillar) were described by Schlegel (2009). Fibers of G. kieslingii subsp. castaneum were mostly oval and had a pillar structure. In our study, the mean fi ber maximum and minimum diameters were in the range reported for Echinocactus grusonii (Huang and Guo 2013), Opuntia fi cusindica (Vignon et al. 2004) and Turbinicarpus sp. div. (Mosco 2009), which had spines composed of fi bers with diameters of 5-15, 6-10 and 6-18 μm, respectively. Th ere are only two studies including measurements of cell wall thickness of fi bers (CW th ). Turbinicarpus species had a mean CW th of 0.7 to 5.9 μm (Mosco 2009), and Escobaria species had a mean CW th of 3.3-4.0 μm (Řepka and Gebauer 2012). Th ese values are slightly lower than for G. kieslingii subsp. castaneum.

Spine variation within areoles
Although a large variation of spine traits within single species has been described (Peharec et al. 2010), there is no extant report describing spine variation within an areole. Since the spines grow from the same lateral bud, low variation of anatomical and morphological parameters would be expected. Th is hypothesis was only partly supported in our study, as the spine and fi ber variation within areoles was lower than between areoles, but the CV for diff erent studied traits within an areole was still high. For example, the CV for S a was even higher than 40% in a few cases (Suppl. material 1). Th us, it seems that even within a single areole, the function of the basal meristem of the spine is very sensitive to environmental and internal stimuli.

Spine variation on the plant body
Th e areole position on the plant body could be related to age, since new spines develop on top of the cactus body, whereas the oldest spines are situated in the basal part. Th us, variation in spine traits from diff erent positions on the plant body can be expected due to diff erent environmental conditions during spine development. For example the stable isotope composition of spines produced serially from the apex of the long-lived columnar species Carnegiea gigantea revealed the past physiological and climatic variation (English et al. 2007). Th is corresponds with our results, as almost all spine and fi ber traits (except for S l , S r and F r ) had diff erent values in the lower parts than in the upper part of the plant body. Nevertheless, the growth of cacti is very slow, and determination of their age without direct observation is diffi cult (Martinez del Rio et al. 1995).

Minimum sample size
Although spine variation is known even in single species, spine sample sizes used by diff erent authors are very variable. For example, anatomical studies use 2 spines (Mosco 2009), morphological studies use 17 spines (Peharec et al. 2010), and mechanical studies usually use 2-22 spines (Schlegel 2009, Huang andGuo 2013). In our study, we found that the minimum sample size was largely infl uenced by the studied spine and fi ber traits, and ranged from 1 to 70 spines, if 10% diff erence in the mean value was taken in account. Th us, for spine sample collection, the spine/fi ber trait to be studied is crucial. In our case, fewer spines (taken from the middle part of the cactus body) would be needed to describe S r (1 spine), F r (6 spines), S a (23 spines) and S c (25 spines). In contrast, more spines should be collected to study CW th (70 spines), S l (52 spines), F max (43 spines) and S w (40 spines). Th ese sample sizes may, however, only apply to spines collected from diff erent plants within a single population. We should note that more similar studies on other cactus species or on the same species, but growing in diff erent environment conditions are needed to make generalizations for spine sample collection in the fi eld. Th is is important task for the correct delimitation and identifi cation of cactus species especially if cultivated specimens were used as sources of evidence in taxonomy (e.g. Schmalzel et al. 2004).

Conclusion
Our study of 15 cactus individuals, 45 areoles and 245 spines showed that spine and fi ber traits are highly variable. Th e areole position on the plant body was an important factor in most of the studied spine and fi ber traits (Question 1). Th e spine and fi ber variation within an areole was lower than between areoles, but the variation was still high (Question 2). Th e minimum sample size was largely infl uenced by the examined spine and fi ber trait, ranging from 1 to 70 spines (mean ± 10%) (Question 3). Th e large spine and fi ber variation between individuals and even within single areoles observed in this study indicates a very complex spine morphogenesis. We encourage a further research focus on the spine and fi ber variation in other cacti species, but also on the factors controlling the basal meristem function and gene expression in spines.