Sexual and Breeding Systems in a Xerophytic Shrubland

Reproductive systems are fundamental attributes for understanding life cycle and regeneration processes and provide information about seed production and genetic diversity. Analyses of reproductive strategies within communities and their associations with functional groups can indicate how physical and biological characteristics may influence the reproductive ecology of such communities. The main goal was to determine if the reproductive systems and their associated functional groups have particular characteristics related to extreme conditions and disturbance within xerophytic shrubland. Floral morphology analysis and four experimental tests were conducted to determine the reproductive systems of species and their associations with the life form, succulence, carbon metabolism, dispersal syndrome, pollination, and disturbance. Of the 144 plant species studied, 72.9% were hermaphrodite, 22.9% were monoecious, and 4.2% were dioecious. Dioecy was associated with woodiness, frugivory and undisturbed areas, while monoecy was more common in herbs. Adichogamy, protandry and herkogamy were more frequent than dichogamy, protogyny and no herkogamy, respectively. Xenogamous species tend to be woody and grow in undisturbed areas, while partially xenogamous species were mainly herbs occurring in disturbed areas. The majority of species were partially self-incompatible. High levels of outbreeding strategies tended to occur mainly in woody K-strategy species from undisturbed areas, mixed breeding strategies occurred in disturbed areas and overall community, and inbreeding strategies were associated with mostly herbaceous r-strategy primarily in disturbed areas.


Introduction
Drylands comprise large areas of terrestrial ecosystems [1]. In Venezuela, xerophytic areas are located mainly in the northern part of the country and are commonly associated with coastal zones. These stressful areas are characterized by high temperatures, low precipitation and low availability of soil nutrients.
Under this regime, plant species exhibit many xeromorphic modifications and adaptations related to their life cycle, such as slow growth and regeneration [2].
In addition, many xerophytic areas are frequently exposed to episodic disturbance driven by torrential rainfall. This together with steep topography produces soil erosion and discontinuous vegetation cover [3], where colonizing pioneer species is very common. However, specific xeromorphic adaptations and reproductive trait associations have not been investigated in detail.
Plant reproductive systems are fundamental attributes for understanding life cycle and regeneration processes and provide information about seed production and genetic diversity. Analyses of reproductive systems within communities and associations with functional groups can indicate how ecological properties may influence the reproductive ecology and evolution of such communities.
Functional groups have an implicit relationship with reproductive and demographic processes and in this context are defined as any trait at the individual level that is directly related to reproductive performance or fitness measured by fertility and survival, among other fitness parameters [4]. In addition, the relationship between reproductive systems and regeneration processes allows us to understand how communities persist over time as a whole. The diversity of reproductive strategies associated with different functional groups shows multiple combinations in disturbed and undisturbed environments of the communities [5] [6] [7], which is an approximation to explain the characteristics of the plants in the communities, functional diversity and biodiversity.
Flowering plants exhibit remarkable diversity in their reproductive system, which reflects their adaptation to biotic and abiotic environments. Studying reproductive systems and their correlates at a community scale is very important to an understanding of how environmental factors drive the evolution of the sexual organization and breeding systems. Previous studies have found that abiotic factors contribute to the evolution of dioecy [8] [9] [10]. However, other studies suggest hermaphrodites are likely to occur in stressful environments where selfing can provide reproductive assurance [11] [12] [13] [14] [15]. How reproductive diversity varies with climate, especially with water availability, remains controversial. Two contrasting and extreme plant reproductive strategies have been described: outbreeding and inbreeding. In nature, however, there is a continuum from outbreeding to inbreeding strategies, where various possible combinations of sexual systems, dichogamy, herkogamy and breeding systems, exist [5] [6] [9] [16]. Despite the great diversity of mechanisms promoting outbreeding (see Table 1 for the definition of reproductive terms), including unisexuality, dichogamy, herkogamy, and self-incompatibility, a substantial number Table 1. Glossary of reproductive terms used. The definitions agree with those proposed by Cardoso et al. (2018), but differ in the organization of the groups.

Term Definition
Reproductive systems General term related to the processes of sexual reproduction, form of sexual organization, relationships between gametes, self-incompatibility (breeding systems or genetic reproductive systems) and various forms of asexual reproduction including agamospermy.

Herkogamy
Spatial separation anther-stigma whithin in the same hermaphrodite flower or unisexual flowers of monoecious species.

Dichogamy
Temporal separation of sexual functions by the sequential ripening of the androecium or gynoecium in hermaphrodite flower, or by different times of anthesis of staminate and pistilate flowers of monoecious species. Adichogamy: The adsence of dichogamy. [30]. Consistent with reproductive system characteristics, plant reproductive strategies in xerophytic areas represent diverse alternatives for plant regeneration. Xeromorphic adaptations of plant species growing in drylands could be related to specific reproductive traits.
The primary goal of sexual traits analyses has been to assess the relative importance of various selective pressures and understand how they interact in different situations [31]. The present study evaluates the community spectrum and diversity of reproductive systems (sexual organization and breeding systems) in xerophytic shrubland, including disturbed areas. It evaluates if stressful conditions inherent to xerophytic lands are associated with specific reproductive strategies. Additionally, an evaluation is made of whether sexual systems, dichogamy, herkogamy, and breeding systems are associated with functional groups (life form, succulence, carbon metabolism, dispersal syndrome, and pollination system specificity) and how such associations might influence the incidence of reproductive mechanisms promoting outcrossing or inbreeding in undisturbed and disturbed areas of the plant community.

Study Area
Fieldwork was conducted in the Venezuelan Central coastal zone on the Mamo plateau, including hill slopes (5 -20 m a.s.l.), situated in the Navy Base of Mamo district, Vargas State, in north Venezuela (10˚36'N and 67˚2'W). The expected vegetation type is a very dry tropical forest according to the climate regime of the Holdridge model [32]; however, some plant species from the tropical thorny shrubland also occur in the area (Figure 1), and for this reason Huber and Alarcon [33] classified vegetation as littoral xerophytic shrubland. The climate is characterized by two short precipitation peaks, the first between July and August, and the second between December and January. The total annual precipitation is 558 mm and the mean monthly temperature is 26.8˚C [34]  slopes produce soil erosion and, consequently, a discontinuous vegetation cover.
Two successional types were evident according to the degree of disturbance.
Disturbed areas were characterized mainly by perturbed soils and the development of vegetation comprising pioneer herbaceous species. In contrast, undisturbed areas, free of erosion or otherwise damaged by human activities, were dominated by long-lived woody species. Plant species were assigned to habitats during a census of the area.

Plant Species Selection and Phylogenetic Effect
The species investigated correspond to the area's flora recorded over three years by Castillo et al. [34]

Functional Groups
All 144 plant species were characterized according to life form, succulence, dispersal syndrome and type of habitat occupied. Furthermore, pollination system specificity was established for 113 previously studied species. Plant life forms were categorized according to habit, longevity, and stem lignification, height and ramification type. In the first instance, species were classified as perennial or short-lived. The life-span of herbaceous species was determined by observing a minimum of ten individuals per species over two years in both disturbed and undisturbed areas. Species in which more than 80% of individuals died during this period were considered short-lived or annual species. Species were also classified as succulents, having specialized fleshy tissue in a plant organ for the conservation of water, or non-succulents, and were further categorized according to the three main carbon assimilation pathways, C 3 , C 4 and CAM, following previously published data [30]. Additional information about carbon metabolism was obtained from the literature (see Appendix A). Species were also assigned to a successional status, based on where species grew in the community: 1) late seral or climax species, and 2) pioneer species. Late seral species grew in natural or undisturbed areas, while pioneer species occurred in disturbed areas, such as eroded sites, road edges, and water ponds constructed for domestic animals.
Observations on pollinators were made during three days of floral anthesis, and completed over three years of flowering periods. The activity of all types of floral visitors was described before visitors were captured. Pollinators were distinguished from other floral visitors using five criteria [40]: 1) presence of pollen, 2) if the body site where pollen is carried is available for pollination, 3) if pollen on the body of a vector could be transferred to a stigma (the pollen load made contact with the stigma during a visit), 4) relative abundance of each visiting species (if the relative abundance of each visiting species is significantly higher than zero), and 5) relationship between flower and visitor size. After that, plant species were categorized according to their pollination system specificity in relation to their pollen vectors (slightly modified from [41]. In this study, the following categories were used: 1) polyphyly-pollinated by different taxonomic orders of visitors, 2) oligophily-pollinated by more than one family of the same taxonomic order and 3) monophily-pollinated by only one species, one genus or different genera of the same taxonomic family. Occurrence of wind pollination was determined according to floral morphology [41] and in some cases, tested by enclosing flowers or inflorescences in 1 mm nylon mesh bags, which excluded most insects but allowed passage of airborne pollen [42].
Information on the morphological adaptation of dispersal units, fruits or seeds was obtained by field observations. Plants were classified according to four dispersal syndromes following Ramírez [9]: 1) abiotic dispersal, represented by

Sexual Organization
The distribution of sexual organs in flowers, individual plants, populations, and species as well as their spatial separation and relative timing in the maturation of sexual organs in flowers, inflorescences or individual plants (Table 1) was determined for the total plant species recorded in the study area. Plant species were initially categorized according to sexual systems as hermaphroditic, andromonoecious, gynomonoecious, monoecious, subdioecious, or dioecious (see Table  1 for the definition of reproductive terms), based mainly on floral morphology, including number ovule per ovary, information on literature specialized and functional criteria: experimental tests and fruit set. All hermaphroditic-dimorphic species were tested for cross-and self-pollination effectivity. On the basis of controlled crosses, fruit and seed sets, some morphologically hermaphrodite species were considered dioecious. In addition, morphologically hermaphroditic species were considered andromonoecious due to the absence of ovules in at least 20% of the flowers [9]. For comparative analyses, only three categories, hermaphrodite, monoecy (including andromonoecious and gynomonoecious species) and dioecy (including androdioecious, gynodioecious, and distylous-functional dioecious species) were considered. Plant species were classified as herkogamous and non-herkogamous (Table  1). Spatial separation between pollen presentation and pollen receipt within flowers of hermaphrodite species and hermaphrodite functional-dioecious species or between flowers of monoecious taxa was measured. In this study, ordered herkogamy was determined when the stigma was positioned at a statistically significant separation from anthers [43]. The null hypothesis tested was if the mean separation between stigma-anther is equal to zero (no herkogamy).
Temporal variation in sexual expression was determined following Ramírez [9]. All hermaphroditic, submonoecious, monoecious and hermaphrodite functional-dioecious species were examined to establish if individual flowers or inflorescences (when treated as pollination units) had synchronous or asynchronous male and female phases [44]. In most species, synchrony of sexual expressions was evaluated by observations at 2-h intervals from the start of anthesis until flower or inflorescence senescence, in a minimum of ten flowers or inflorescences per species. Maturation of stamens was determined by anther dehiscence or, in the case of poricidal anthers, by the time when pollen could be dislodged from anthers. Female maturity was determined by a shiny or moist stigmatic surface in taxa with wet stigmas, or by the elongation of the style and full  [41], protandrous (anther dehiscence occurring before stigmatic receptivity), or protogynous (stigmatic receptivity prior to anther dehiscence). The latter two categories may include species with posterior overlapping of the sexual phases (incomplete dichogamy, sensu [44]).

Plant Breeding Systems
A total of 73 species were characterized in their breeding systems: 62 species were experimentally evaluated in this study, six species come from previous studies in the same study area and five additional species presented morphological and functional characteristics that correspond to species without spontaneous self-pollination and xenogamy. The occurrence of agamospermy was only tested for 60 species in the present study; six additional reports come from previous studies.
Reproductive efficiency under experimental conditions was determined at two levels: 1) fruits developed per total number of flowers, and 2) a total number of non-abortive seeds produced by all fruits per total number of ovules (flower number multiplied by the average number of ovules per flower). Experimental pollination tests considered in this study were: 1) agamospermy test, as fruits and/or seeds produced from emasculated and isolated flowers; 2) spontaneous self-pollination test, as fruits and/or seeds produced from isolated and nonmanipulated flowers; 3) self-pollination test, as fruits and/or seeds produced from hand or assisted self-pollinated flowers; and 4) cross-pollination test, as fruits and/or seeds produced from hand outcrossed flowers. Nylon mesh bags were used to isolate flowers when this treatment was required. Breeding system data from previous studies in the same area for three Cactaceous species [25] [26], Melochia pyramidata var. pyramidata and Melochia tomentosa [27] and Coccoloba uvifera [45] were included in the general figure of plant community.
Four breeding system indexes (BSI) were determined at the fruit and/or seed level following [16]. Each BSI results from the quotient of two contrasting experimental tests, where the denominator is expected to be the largest referential value. In the case when the conclusion derived from both fruit and seed levels differed, it opted for the conclusion obtained at the seed level. Five categories for each breeding system index (Breeding Index Categories, BIC) were used [16] for all species: 1) BSI = 0, 2) 0 < BSI < 1.0, 3) BSI = 1.0, 4) 0 < (1/BSI) < 1.0 (when BSI > 1.0), and 5) 1/BSI ~ 0 (when BSI ~ ∞). This system of categories is a symmetrical model at both sides of value 1.0, positioning contrasting categories at the extremes: 0 (BSI = 0) and ∞ (1/BSI ~ 0) values, which represent opposite biological conditions. Intermediate values, below (0 < BSI < 1.0) and above (0 < (1/BSI) < 1.0) 1.0, but lower than the extreme conditions, correspond to intermediate or transitional biological categories. BSI = 1.0 denotes the referential value indicating that the experimental tests conforming to the index render approximately equal results. More details about the categorization of the BSI, as- In addition, some zoophilous pollination species in which spontaneous self-pollination is avoided as a result of morphological traits, sexuality, and dichogamy, were considered as non-spontaneous self-pollination (BSI = 0). These

Statistical Analysis
The t-test was employed to determine corresponding breeding system categories for the four indexes of each plant species. T-test, with degrees of freedom equal to n − 1 [50] was used to discriminate between Breeding System Index (BSI) values from 0 and 1.0 (see [16] for details). When BSI values were higher than ≥1.0 (up to infinite), the inverse value (1/BSI) was used instead of BSI, to make the statistical method symmetrical at both sides of BSI = 1. In order to calculate the four BSI that make the composite breeding system of a given species using fruit set or seed set data, the user can have access to an Excel spreadsheet that automatically calculates Log-linear analyses of frequency using two-way tables to determine dependence between reproductive (sexual system, dichogamy, herkogamy, and breeding system) and functional (life form, carbon metabolism, successional stages, pollination system, and dispersal syndrome) attributes were used. For example, comparing sexual systems and habitats, the frequencies of hermaphrodite, monoecious, and dioecious species that occurred in undisturbed and disturbed habitats were contrasted. In order to establish the level of dependence between reproductive variables and functional groups, log-linear analyses of frequency were performed using two-way tables [51]. The concept of interaction in log-linear analyses is analogous to that used in the analysis of variance. When the log-linear analysis of frequency was significant, residual frequencies (i.e., observed minus expected frequencies) were estimated for each cell of the two-factor comparison, and then standardized and tested for significance. This analysis established which pairs of variables deviated significantly from expected values [52], and therefore, made a larger contribution to the association. Significant and positive residuals indicated a strong association between both categories, and significant and negative residuals indicated an unusual occurrence.

Sexual Organization and Functional Groups Associated
Sexual systems and attributes associated. Plant sexual system was only significantly associated with seed dispersal syndromes (Table 2). Granivorechory and abiotic dispersal was the most frequent seed dispersal syndrome in monoecious species. Seed dispersal mediated by frugivory was the main syndrome found in dioecious plants. In spite of a non-significant association, the proportion of dioecious species was higher in plants with polyphilous pollination and late seral stage. Monoecy tended to be higher for herbaceous species, non-succulent plants, anemophilous and polyphilous pollination systems and pioneer seral stage.
Temporal variation in sexual expression did not exhibit significant relation with functional traits (Table 2); however, the proportion of protogyny was higher for herbaceous species, non-succulent, C 4 species and dispersed by granivores animals. In contrast, protandrous taxa were abiotically dispersed, and polyphilous and anemophilous pollination.
Herkogamy was not significantly associated with functional traits (Table 2); however, non-herkogamous species tend to be mostly herbaceous species, dispersed by granivorechory and epizoochory, polyphilous and anemophilous pollination, and frequently found in disturbed areas.

Plant Breeding Systems
Most plant species studied were non-agamospermous (N = 61; 92.4%) and 7.6% (N = 5) were partially agamospermous (see Table 1 for the definition of reproductive terms). These proportions were comparable for undisturbed and disturbed habitats. Partially agamospermous species were most numerous for herbaceous species from disturbed areas (Supplementary Material 3; Figure 2).
The five possible categories of the Index of Spontaneous Self-Pollination (ISSP) were recorded in the sample studied (see Table 1 for the definition of reproductive terms). Most species presented partially spontaneous self-pollination (46.6%), followed by non-spontaneous self-pollination (37.0%), partially constrained assisted self-pollination (13.6%), spontaneous self-pollination (Rhynchosia minima) and obligated spontaneous self-pollinated (Jacquemontia cumanensis) (Supplementary Materials 2, 3). However, obligated spontaneous self-pollination in Jacquemontia cumanensis was represented by a small fraction of flowers producing fruits and seeds by self-pollination (Supplementary Material 2).
Some trends for spontaneous self-pollination categories and functional groups were observed: non-spontaneous self-pollinated species corresponded to trees and lianas, followed by shrubs and perennial herbs, and only one species was an annual herb (Phyllanthus niruri). A substantial fraction of non-spontaneous self-pollinated species grows in undisturbed areas of the shrubland. Partially spontaneous self-pollinated species were annual herbs, polyphilous, growing in disturbed areas. Species with partially constrained assisted self-pollination were more frequent among herbs growing in disturbed areas.   Table 1 for the definition of reproductive terms). The highest frequency of xenogamy was found in trees and lianas; a large proportion of xenogamous species are dispersed by frugivorous animals. More than 50% of the xenogamous species grow in undisturbed areas. Partially xenogamous taxa were mostly shrubs and herbs found in disturbed areas and undergo polyphilous pollination. Partially endogamous species were herbs, mostly dispersed by granivorechory and epizoochory and grow in disturbed areas (Supplementary Information 3, Figure 2).
Four categories of the Index of Self-incompatibility (ISI) were found in the sample examined: partially self-incompatible (N = 46), self-incompatible (N = 9), partially cross-incompatible (N = 6), and four plant species were completely self-compatible (see Table 1 for the definition of reproductive terms

Breeding Systems and Sexual Organization
The relationships between sexual organization and breeding systems were not significant, except for the self-fertility index categories and dichogamy ( Table 4). Regardless of non-statistical relationships, the sexual organization showed that most plant species examined were similarly distributed across the breeding system indexes for hermaphrodite, herkogamous and adichogamous species; however, frequencies of non-spontaneous self-pollinated-protogynous and partially spontaneous self-pollinated-protandrous species were relatively higher than their respective counterparts. In addition, non-spontaneous self-pollination was more frequent than partially spontaneous self-pollination for monoecious taxa. Xenogamous and partially xenogamous species were mostly adichogamous; however, frequencies of xenogamous-protogynous and partially xenogamous-protandrous species were relatively higher than their respective counterparts. Self-incompatibility was recorded in slightly higher frequency than partial self-incompatibility for adichogamous and herkogamous species. Protandry Table 4. Relationship between the most common breeding system index categories and morphological and temporal organization sexual traits.
Breeding system Sexuality Spatial separation between pollen-stigma

Discussion
The with less extreme environmental conditions. It is likely that convergences in reproductive attributes may respond, among many other variables, to regional or latitudinal patterns.

Sexual Systems
The frequency distribution of sexual systems in the xerophytic shrubland is concordant with results shown for many tropical communities [ [9], and under stressing conditions [15]. Interestingly, in the xerophytic shrubland, dioecy was more than four times higher in the late seral state than in the pioneer state. The low number of dioecious species in the latter may be related to predominantly herbaceous species occurring in disturbed areas generated by anthropogenic activity and by the effect of rainfall driven soil erosion and runoff, a common phenomenon in arid environments [3]. Dioecy is found in very low frequency in disturbed areas in some tropical communities [5] [9], because of the high number of colonizing herbaceous species and the well-recognized association between the woody condition and dioecy. Monoecy promotes cross-pollination by preventing within-flower selfing [58] [64] [65]. The proportion of monoecious species in undisturbed habitats (15.5%) found here is close to that in the mesothermic shrublands of the Gran Sabana Plateau (14%; [5], secondary deciduous forest remnant (18.7%; [7], and psamophylous (17.2%) and littoral meadows (13.9%) in the coastal plains of the Paraguaná Península [12]. This highlights two attributes influencing monoecy in undisturbed xerophytic shrublands: vegetation structure and dry coastal climate. In contrast, disturbed habitats exhibited a comparatively higher frequency of monoecious species (26.3%) related to the high number of herbaceous species. This figure suggests that permanent disturbance caused by humans and the natural erosion process may select monoecy as the main figure for cross-pollination throughout increment of herbaceous colonizing species. The high proportion of monoecy observed in disturbed habitats may be associated with the xerophytic environment, where the stressful condition is caused by water deficit. Separate sexes are favored in stressful environments [ [67]. Division of function in unisexual plants may increase male and female fitness due to a compensation effect [28], unless physiological constraints are so severe as to generate low plant density or lack of pollinators or reduced fertility. The association between monoecy colonizing species and C 4 carbon metabolism may enhance the capacity of herbaceous-C 4 species to reduce water loss in water limiting environments [68].
The high proportion of submonoecy found among monoecious species examined agrees with the results found in the Gran Sabana Plateau [5], and in the Venezuelan Central Llanos [9], and suggests that, in many cases, monoecy might have evolved from hermaphroditism. Male flowers in andromonoecous species may enhance male fitness by increasing pollen amount and pollen dispersal in Open Journal of Ecology the population and subsequently pollination efficiency, needed for many andromonoecious-polyphilous and -anemophilous species in the shrubland, where pollen required may be fulfilled by pollen produced by male flowers. In addition, seed dispersal by granivores and wind in monoecious-herbaceous species represents associations frequently found in herbs growing in disturbed areas [9] [24]. These dispersal syndromes may be considered opportunistic dispersal strategies for plant species colonizing disturbed habitats in the xerophytic shrubland where perturbations are continuous.

Herkogamy and Dichogamy
Herkogamy was twice as often as dichogamy in the xerophytic shrubland. A similar result has been reported in three other Venezuelan plant communities with contrasting species compositions and structures [5] [6] [7] [9]. The parallelism in the frequency of dichogamy and herkogamy observed between different geographic areas and plant communities, suggests convergent evolution in mechanisms that help avoid pollen-stigma interference and promote cross-pollination, irrespective of the taxonomic composition and ecological characteristics of plant communities. In the xerophytic shrubland, the frequency of dichogamy and herkogamy was not significantly associated with functional groups; however, there were some important trends, including disturbance. Herkogamy is a critical strategy for outcrossing in undisturbed xerophytic shrubland, but has slightly less importance in disturbed areas. These associations are concordant with a number of mechanisms that promote cross-pollination in late seral stages, mainly woody species [9] [24]. The abundance of non-herkogamous species in herbaceous and disturbed areas suggests that selfing strategies may represent an important adaptation for autogamous colonizing species, mostly granivorechorous dispersal and polyphilous and anemophilous pollination. Several surveys indicate that protandry is more common than protogyny [5] [7] [9] [69]. The proportion of protandry was approximately two times the proportion of protogyny in an extensive survey of intra-floral dichogamy [70]. In the xerophytic shrubland, the frequency of protandry was 1.4 times the frequency of protogyny, which is less than the ratio found in the Venezuelan Central Plain [9] and herbaceous-shrubby communities in the Gran Sabana plateau [5]. The frequency of protandry and protogyny was dependent on the successional stage found in the xerophytic shrubland: the highest protandry/protogyny ratio was noteworthy in primary vegetation (3.0), compared to disturbed vegetation (1.6). Protandry may act as a non-rigid mechanism in the undisturbed xerophytic shrubland and suggests a more versatile way of allogamy or a mixed-breeding system under environments characterized by low precipitation and high temperatures. Additionally, the highest frequency of protandry in the undisturbed xerophytic shrubland was non-significantly associated with abiotic dispersal and polyphily and anemophily pollination system. In this context, Barrett [71] pointed out that dichogamy is an exceptional widespread floral strategy occurring in many outcrossing species, regardless of the pollination system, which

Sexual Organization Associations
The sexual system, herkogamy and dichogamy may be in such a combination that each other's partial effectiveness is reinforced, cross-pollination promoted and pollen-stigma interferences avoided. For instance, the presence of dichogamy associated with herkogamy in the xerophytic shrubland represents a significant fraction (74.64%), similar to that found in the herbaceous-shrubby communities in the Gran Sabana Planteau [5]. Herkogamous-dichogamous species avoid self-pollination and may be considered the first step in the evolution of delayed selfing to provide reproductive assurance [72]. Hermaphrodite-herkogamous species tend to promote cross-pollination through herkogamy (79.8%) and less frequently through dichogamy (20.2%) in the xerophytic shrubland, similar to that found in the herbaceous-shrubby communities in the Gran Sabana Plateau [5].
Such outcomes show that sexual organization promote cross-pollination and avoids pollen-stigma interference in the xerophytic shrubland, and there is only a small proportion of plant species without adaptation for cross-pollination, represented by adichogamous-non-herkogamous species.

Plant Breeding Systems
The majority of species examined in the xerophytic shrubland were non-agamos permous. This pattern is consistent with the observed limited occurrence of agamospermy at the community level in many ecosystems [6] [7] [19] [36], with available records at the family level [73], and others [16]. Only 7.6% of plant species were partially agamospermous, which is equivalent to facultative agamospermy. This proportion is less than levels found in some isolated tropical communities [5] [18] and larger or similar to other tropical areas [12] [19] [74].
The occurrence of partially agamospermous species tends to be associated with herbaceous life forms growing mainly in disturbed areas [74]. The highest frequency of partially agamospermous species has been found in disturbed areas recorded in our study was partially self-incompatible. Partial self-incompatibility has been interpreted as evidence of high reproductive success, associated with mixed-breeding under the current scenario of pollination service in natural ecosystems [76], and 56 it is considered an optimal and evolutionary stable breeding strategy [77]. The frequency of partially self-incompatible species was similar in undisturbed and disturbed areas of the xerophytic shrubland; a larger than that recorded in the herbaceous-shrubby communities in the Gran Sabana Plateau [5]. Partial self-incompatibility and shrub and herb association in the xerophytic shrubland could bias the occurrence of partially self-incompatible species in undisturbed and disturbed areas.
The second most important frequency of non-spontaneous self-pollination and xenogamy in the xerophytic shrubland were correlated with specialized life history strategies: woody and perennial life forms from undisturbed areas, which are primarily related with the high incidence of xenogamy in woody species [5] [7] [78]. Additionally, a large proportion of xenogamous species are dispersed by frugivorous animals. Frugivory in xenogamous species is associated with late seral stages, where plant species have specialized reproductive strategies [22] [24]. Self-incompatibility was the second most significant category and similar to that found in the mesothermic shrublands [5]. Life form composition seems to be related with self-incompatibility frequency in the xerophytic shrubland. Trees and lianas tend to be predominantly self-incompatible in undisturbed areas, which agree with the recognized association between self-incompatibility and woodiness [5] [21] [60]. Woody life form may influence self-incompatibility occurrence, because perennial life history is generally associated with multiple reproductive episodes and consequently with the permanent contribution to reproductive success.
Among inbreeding strategies, a low frequency of partial constrained assisted self-pollination, partial endogamy and partial cross-incompatibility were recorded in disturbed areas and the overall community. Most of these species were herbs, dispersed by granivorechory and epizoochory, and growing in disturbed areas; the largest parts of these are pioneer species [13] [76]. Partial endogamy may occur under a variety of conditions, being more frequent for taxa growing in stressful environments, with reduced pollinator service [11] [12] [13], and with some specific traits, such as invasive-exotic or colonizing species [79] [80]. Herbaceous life forms and generalist dispersal syndromes suggest that partially endogamous species may be influenced by flexible reproductive attributes, mainly in herbaceous pioneer species in xeric environments.
Cross-incompatibility is a breeding system category poorly examined at the community level [16]. Partial cross-incompatibility was found in six predominantly herbaceous species, dispersed abiotically and by granivorous animals, and growing mostly in disturbed areas in the xerophytic shrubland, with frequency comparatively low to that found in the mesothermic shrublands (23% -31%,   5]). Ecological circumstances also play an important role in determining when selfing evolves [81]. Seed dispersal by granivorous animals or wind together with herbaceous life form could relate to inbreeding in some taxa from disturbed areas in the xerophytic shrubland, as has been recorded in a secondary deciduous forest remnant [7].

Sexual Organization and Breeding Systems
Most of the plant species were similarly distributed among breeding system categories for hermaphrodite, herkogamous and adichogamous species. Most of the herkogamous species were similarly distributed for spontaneous self-pollination index categories, self-fertility index categories, and self-incompatibility index categories (see Table 1 for the definition of reproductive terms) and suggest that herkogamy is the main floral attribute avoiding autogamy in this plant community such as reported previously [5] [6] [7] [82]. In contrast, dichogamy, protandry and protogyny, have a differential role in promoting cross-pollination and avoiding pollen-stigma interference. Protogynous species tend to be xenogamous while protandrous species are predominantly partially xenogamous, proposing that protogyny could be a more effective attribute than protandry to avoid pollen-stigma interference. Self-incompatibility and partial self-incom patibility were mostly associated with hermaphroditism, herkogamy, and adichogamy, which agrees with previous studies [5] [6] [7] [17], though dichogamy, has been found equally common among self-incompatible and self-compatible species [70]. However, protogyny was only found in partially self-incompatible species. This figure represents attributes that can promote cross-pollination in plant species where self-pollination is possible.
Supplementary Material 3. Frequency of breeding system categories according to some functional plant traits and seral states of the xerophytic community.