|Home | About | Journals | Submit | Contact Us | Français|
A South American cactus species, Echinopsis ancistrophora (Cactaceae), with dramatic among-population variation in floral traits is presented.
Eleven populations of E. ancistrophora were studied in their habitats in northern Argentina, and comparisons were made of relevant floral traits such as depth, stigma position, nectar volume and sugar concentration, and anthesis time. Diurnal and nocturnal pollinator assemblages were evaluated for populations with different floral trait combinations.
Remarkable geographical variations in floral traits were recorded among the 11 populations throughout the distribution range of E. ancistrophora, with flower lengths ranging from 4·5 to 24·1 cm. Other floral traits associated with pollinator attraction also varied in a population-specific manner, in concert with floral depth. Populations with the shortest flowers showed morning anthesis and those with the longest flowers opened at dusk, whereas those with flowers of intermediate length opened at unusual times (2300–0600 h). Nectar production varied non-linearly with floral length; it was absent to low (population means up to 15 µL) in short- to intermediate-length flowers, but was high (population means up to 170 µL) in the longest tubed flowers. Evidence from light-trapping of moths, pollen carriage on their bodies and moth scale deposition on stigmas suggests that sphingid pollination is prevalent only in the four populations with the longest flowers, in which floral morphological traits and nectar volumes match the classic expectations for the hawkmoth pollination syndrome. All other populations, with flowers 4·5–15 cm long, were pollinated exclusively by solitary bees.
The results suggest incipient differentiation at the population level and local adaptation to either bee or hawkmoth (potentially plus bee) pollination.
The degree to which flower–pollinator relationships are specialized (or not) has stimulated much research and discussion during the past decade (Waser et al., 1996; Johnson and Steiner, 2000). One outcome of this debate has been a reconsideration of what it means to be specialized. Obligate mutualism (e.g. between yuccas and yucca moths or between senita cactus and senita moth) is a relatively rare form of specialization, largely limited to nursery pollination systems in which a subset of fertilized ovules are consumed by the progeny of the pollinator (Holland and Fleming, 1999). More commonly, plants are specialized to pollination by guilds or functional groups of pollinators, e.g. sphingid moths with certain proboscis dimensions (Moré et al., 2006), or oil-collecting bees (Cosacov et al., 2008). One problem is that floral specialization on certain visitor guilds or species is often visible only at second glance, as different types of filters (e.g. chemical filters as in Johnson et al., 2006; morphological filters as in Moré et al., 2007) can act on the (potential) visitor spectrum. Furthermore, only a subset of floral visitors may act as pollinators, and only a subset of pollinators may account for most pollen transfer (Bloch et al., 2006). For pollination systems based on animals that forage for nectar or pollen rewards, specialization has traditionally been thought to evolve as an adaptation to the most effective pollinator (Stebbins, 1970). Alternatively, Aigner (2001) has suggested that specialization even to a relatively inefficient pollinator can occur if it results in a net fitness increase. For example, Bloch et al. (2006) have shown that high visitation frequency can compensate for low pollination efficiency (i.e. number of conspecific pollen grains deposited on a stigma per visit) by butterflies at Dianthus flowers, an example of what Herrera (1989) has termed the ‘quantity component of pollination’.
To test hypotheses about floral adaptation, Aigner (2001) has advocated the study of model systems with a range of floral phenotypes. However, the study of adaptive potential in floral evolution has been hampered by the fact that traits under directional or stabilizing selection tend to lose their variance in natural populations (Fenster, 1991; Herrera, 1996). Ironically, natural, heritable variation in floral colour, scent or shape, considered the sine qua non for adaptive evolution via natural selection (Endler, 1986), often is found to have neutral or non-significant effects on plant reproductive fitness (e.g. Ackerman et al., 1997; Jersakova et al., 2006). Thus, it can be difficult to explore the potential for pollinator-mediated selection if there is not sufficient floral variation left upon which selection can act. There are several opportunities to observe incipient phenotypic sorting:
These approaches share the common goal of measuring the extent to which variation in floral traits could be moulded by pollinator-mediated natural selection.
The South American cactus genus Echinopsis contains several highly variable taxa, including the E. ancistrophora species complex, in which different natural populations vary in floral shape, colour, scent, nectar and time of anthesis. Owing to its dramatic phenotypic variability, numerous names have been assigned to different populations of E. ancistrophora and its relatives, until Hunt (1999) circumscribed E. ancistrophora to encompass four subspecies, including taxa with white or colourful flowers and short (approx. 5 cm) to very long (>20 cm) flowers. Remarkably, almost all flower colour × length combinations can be found within this group, often departing dramatically from the conventional trait combinations outlined by the floral syndrome concept for cacti and Angiosperms in general (Porsch, 1938, 1939; Faegri and van der Pijl, 1971). Here we focus on the white-flowered populations, i.e. E. ancistrophora sensu Lowry (2002) and Hunt (2006), that include the full extent of floral morphological variation. Preliminary molecular phylogenetic analyses confirm the monophyly of E. ancistrophora (sensu Lowry, 2002 and Hunt, 2006), and indicate that further subdivision of these entities into separate taxa (e.g. subspecies) is not supportable (B. O. Schlumpberger, unpubl. res.). Thus, the plants studied here represent a cluster of closely related entities undergoing unusual phenotypic sorting, and present a rare opportunity to study floral trait combinations not often observed in nature.
Flowers of the populations included in this study are white (to slightly pink in some cases) and funnel-shaped with highly variable tube length, and thus have been assumed to bloom nocturnally and to be moth-pollinated (Lowry, 2002). However, as the shortest flowers do not resemble typical hawkmoth-pollinated cactus flowers (e.g. Raguso et al., 2003), and diurnal anthesis has been suggested for some of these populations (Backeberg, 1959), we predicted that pollinator affinities would vary among populations, offering the chance to study intraspecific divergence and potential adaptation to different functional groups of pollinators.
We tested specific predictions concerning floral trait combinations and associated pollinator affinities among 11 populations of E. ancistrophora in northern Argentina, as follows.
If floral variation is not adaptive, we should fail to reject the null hypothesis of random association between floral traits and visitor assemblages.
Alternatively, if flowers of different populations are adapted to different pollinator groups, anthesis time and nectar production should vary in concert with floral length as follows:
In contrast, we have no clear predictions based on pollination syndromes for two populations of E. ancistrophora with intermediate floral tube lengths and anthesis times. One possibility is for bimodal pollination by sphingid moths and bees seeking nectar and pollen, respectively (see Barthell and Knops, 1997; Miyake and Yahara, 1998). A second possibility is for pollination by a novel class of pollinator or functional group (see Fenster et al., 2004), whose presence would not have been predicted by classical syndromes (e.g. hummingbirds on white cactus flowers of Echinopsis atacamensis ssp. pasacana; Schlumpberger and Badano, 2005). A third possibility is that unusual combinations of floral traits may be discriminated against by potential pollinators due to incompatibility of floral rewards or mechanical fit with visitors (e.g. Wilson et al., 2004), which would perhaps explain why such combinations are not observed more commonly in nature. These populations were of specific interest, as they allowed us to evaluate the potential constraints associated with transitional stages between discrete syndromes (e.g. Baker, 1963) versus the possibility of adding an additional pollinator with no reduction in fitness (e.g. Aigner, 2001).
Echinopsis ancistrophora grows in the northern Argentine provinces of Tucuman, Salta and Jujuy along the eastern Andean slopes or on pre-Andean hills above the Chaco forest zone. Field studies took place during the flowering period of E. ancistrophora from mid October to mid December in 2002 and 2003. Sixteen populations of E. ancistrophora were studied, covering the entire range of its distribution in the Argentine provinces Tucuman, Salta and Jujuy. However, only 11 populations for which sufficient sample sizes of floral trait data and flower visitors could be gathered are included in the present analysis, representing a mostly latitudinal transect, which also encompasses the small longitudinal extension of the species (Table 1). Plant identifications were determined by the first author. Voucher specimens are deposited in the herbarium of the Museo Botánico de Córdoba (CORD), Argentina. Bees were determined by Arturo Roig-Alsina, Museo Argentino de Ciencias Naturales ‘Bernardino Rivadavia’ (MACN), Buenos Aires, Argentina, where voucher specimens were deposited as well. Sphingids were determined by the second and third authors, and all vouchers were deposited at the Laboratorio de Biología Floral, Instituto Multidisciplinario de Biología Vegetal (CONICET-UNCba), Córdoba, Argentina.
Flower characters were evaluated in the field. Additional observations on cultivated plants confirmed that floral traits are stable under differing environmental conditions (data not shown). Flower depth was measured with a ruler (to the nearest millimetre) during full anthesis. When flower depth was measured, we also recorded stigma position in these flowers, in order to control for potential correlation of flower length and stigma position, and four different positions were distinguished: (1) exserted, above upper anthers; (2) nested among or below upper anthers; (3) central in flower; and (4) nested among or below lowest anthers. We chose to distinguish four functional categories rather than measuring style length, because precise measuring of the styles would have required cutting the flowers open, and functional (relative) stigma position is more informative than absolute style length, especially in flowers with such grossly different lengths. We calculated the Spearman's rank correlation coefficient to test for correlation between flower length and the four distinguished stigma positions.
Nectar characters were evaluated in all populations for which sufficient numbers of flowers were obtained. Nectar quantity was measured towards the end of anthesis, i.e. in the afternoon for flowers with morning anthesis and in the morning for flowers with evening anthesis, with calibrated micropipettes (5, 10 or 20 µL). Subsequently, sugar concentration (% Brix = sucrose equivalents) was identified to the nearest 0·5 % with a temperature-controlled hand refractometer (Atago Co., Ltd, Tokyo, Japan), when nectar quantity was sufficient (>1 µL). In flowers with smaller volumes, the nectar of several flowers had to be pooled for refractometer measurement. Owing to the narrowness of the flower tubes the flowers were sampled destructively in order to extract nectar, but care was taken to collect only floral nectar and not fluids from wounded floral tissues. We tested for a correlation between nectar quantity and quality (i.e. sugar concentration) using Spearman's rank correlation coefficient.
Limited numbers of flowers were available for only a few days each year, and were needed for the other studies described in this paper and in Schlumpberger and Raguso (2008). Thus, in some cases nectar had to be taken from open, unmanipulated flowers exposed to possible visitors, at the end of their anthesis time. However, observing nectar production in bagged flowers of several populations (all flowers in Puente del Toro, one-third of the flowers in El Mollar, and some flowers in Escoipe and Volcán, respectively) suggested that using open flowers did not strongly bias the results, as 70–75 % of flowers in these populations were not visited (see below). Furthermore, the fact that the longest flowers were unbagged and still yielded by far the highest nectar volumes indicates that visitor frequency was low. In the worst case, the substantial nectar production in those flowers was underestimated. The low nectar volumes measured in the relatively long flowers from Volcán (see Fig. 1 below) were confirmed by both large sample size and the inclusion of bagged flowers. In order to test for reliability of the nectar data obtained in the field, we sampled nectar from some additional flowers of plants cultivated in the Munich Botanical Garden, Germany, descended from (a) the El Fuerte population, (b) the outskirts of our Campo Quijano population and (c) a population near Valle Grande, Jujuy, Argentina, with long flowers with evening anthesis, which was not part of our field study.
Anthesis times were observed in the field; additional data were obtained from cultivated plants by direct observation or by means of a surveillance camera. In the latter case, plants were illuminated with infrared diodes, and photographs were taken automatically by ‘EyeCU2’ surveillance software (JPC Software, Northport, AL, USA) and directly stored on a computer.
In order to evaluate the potential for hawkmoth pollination in E. ancistrophora, we combined two indirect methods to study hawkmoth visitation.
In the seven populations in which no indication of hawkmoth pollination was found, we studied the spectrum of diurnal flower visitors by two methods.
Due to the rarity and brevity of flowering events, the long-tubed populations were not watched extensively for bee visitation on the mornings following anthesis.
In order to understand the role of moth-adapted E. ancistrophora flowers in their communities, we evaluated flower lengths of all sympatric night-blooming flowers likely to be pollinated by moths in both the El Fuerte and the Termas de Reyes populations.
In E. ancistrophora, flower length varied little within populations, but dramatically among populations, with the shortest flowers measuring 4·5 cm in the Puente del Toro population (mean 6·40 cm, s.d. = 0·89, n = 14), and the longest flower tubes in El Fuerte measuring up to 24·1 cm (mean 20·80 cm, s.d. = 2·09, n = 10; Fig. 1C). Observation of cultivated plants revealed that open flowers continue to grow after anthesis; haphazardly chosen individuals increased their total floral length between opening and wilting by 3–9 %. Stigma position was highly variable within all populations, with no correlation to flower length (Spearman's rank correlation coefficient r = −0·08, P = 0·14, flowers from all populations pooled). The continued elongation of flowers after anthesis also changed the relative position of the stigma over time, giving the appearance that the stigma retreated further into the floral tube. Flower measures obtained from cultivated plants originating from five study populations were found to be within the distribution of field-collected data (not shown).
Strong differences in the timing of anthesis were found among populations. Matinal flower opening occurred in the five populations with the shortest flowers, and vespertine flower opening occurred in the four populations with the longest flowers (Fig. 1, top). Short flowers usually opened between 0900 and 1030 h, and long flowers opened between 1930 and 2100 h, but low temperatures, cloud cover or precipitation could delay opening. In the two populations with intermediate flower lengths, anthesis time was intermediate as well, with first flowers opening at 2300 h in Campo Alegre and at 0130 h in Volcán. In populations with short or long flowers, the process of flower opening was synchronous, i.e. all opening flowers were in similar opening stages at a given time. In contrast, flowers in Volcán opened during an extended period from 0130 h to sunrise, with different opening stages found at all times. Anthesis times of cultivated plants did not differ from times observed in the field (data not shown).
Flowers usually remained open for less than 24 h in long-flowered populations, but longer blooming duration (up to 48 h) was observed in short-flowered populations, depending on environmental conditions: field observations suggest that hot and dry conditions cause shorter flowering, and delayed wilting was observed in bagged flowers. For the short-flowered plants from Puente del Toro, El Mollar, Cuesta del Cebilar, we observed that flowers close in the afternoon or evening and re-open the next morning. On other occasions, flowers in the same populations (e.g. El Mollar) bloomed for only 1 d, depending on the weather. The wilting of flowers (or closing before re-opening on the second day in shorter flowers) began before 1000 h in long flowers (e.g. Termas de Reyes) and no earlier than 1200 h in short to medium flowers (e.g. El Mollar, Volcán).
Nectar production was low in flowers from the populations with short and intermediate flower lengths (i.e. those with diurnal or intermediate anthesis times), with highest nectar production in the Volcán population (mean 14·9 µL). No nectar was found in the shortest flowers from the Puente del Toro population, despite the fact that all flowers were bagged (Fig. 1A). Mean nectar production was more than ten-fold higher in the four populations with very long-tubed flowers and nocturnal anthesis than the average of the other populations (Fig. 1A). In the three populations for which sufficient sampling was performed, large nectar crops were present even in unbagged open flowers, with a maximum of 262 µL (mean 170 µL per flower) in Termas de Reyes (Fig. 1A).
Nectar sugar content ranged from 13·4 to 20·6 %, with no obvious relationship between sugar concentration and nectar quantity or flower length (Fig. 1A). There was a marginally significant, negative correlation between sugar concentration and nectar volume (Spearman's r = −0·377, P = 0·056, flowers from all populations pooled).
Results from cultivated plants mirror the patterns observed in the field: the diurnal flowers from plants from the Campo Quijano region have relative low nectar production (mean 16·3 µL, n = 5); flowers of descendants from El Fuerte produced on average 92·6 µL (n = 5), and one nocturnal flower from Valle Grande produced 99·7 µL. In three of the five flowers of cultivated plants from the El Fuerte population, we measured a concentration gradient from the top to the base of the nectar column. Sugar concentrations ranged from 12 % in the uppermost samples to 27·5 % in the lowest samples.
In three of the four populations of E. ancistrophora with the longest flower tubes and evening anthesis (Calderilla, Termas de Reyes, El Fuerte), 25–30 % of all stigmas had lepidopteran scales adhering to their surface, indicating moth visitation (Fig. 1B). In the fourth of these populations (Abra Sta. Laura) not enough stigmas could be examined, due to low levels of flowering. Among the seven populations with short to intermediate flower tubes, moth scales were found on stigmas in only three populations. Within these populations, the frequency of individuals with moth scales was low (1·8, 4·3 and 6·7 % of observed flowers; Fig. 1B).
In the three populations of E. ancistrophora with the highest frequency of moth scales on floral stigmas (Calderilla, Termas de Reyes, El Fuerte), we captured 476 sphingid moths belonging to 20 species from eight genera (Table 2), of which 19 individuals (three species from two genera) carried pollen of E. ancistrophora (Fig. 2). All but one of these moths (94·7 %) belonged to the genus Manduca (M. stuarti n = 4, M. tucumana n = 14), and were caught in the Termas de Reyes population (Table 3). These two species had the greatest proboscis lengths of all captured sphingids, with means of 79·3 mm for M. tucumana and 78·6 mm for M. stuarti. The mean proboscis length for all captured sphingids was 57·9 mm, and the mean for the subset of all Manduca moths was 71·1 mm. Only one specimen of Sphinx phalerata with E. ancistrophora pollen was caught in the El Fuerte population, with a proboscis 53·3 mm long. In addition to pollen of E. ancistrophora, the sphingids carried pollen of various other putatively sphingophilous plants belonging to the families Apocynaceae, Asteraceae, Martyniaceae, Onagraceae, Rubiaceae, Solanaceae and Verbenaceae (Fig. 3).
Notably, long flowers with nocturnal anthesis attracted dynastine scarab beetles of the genus Cyclocephala that fed on and destroyed flowers. These beetles were also attracted by our light-traps.
Populations of E. ancistrophora with short- or intermediate-length flowers and early morning anthesis were visited primarily by solitary bees. In total, 21 bee species from 14 genera and five families were observed (Table 4). Most abundant were bees of the family Halictidae, with 48 % of all collected bees (e.g. Dialictus sp. and Pseudagapostemon jensenii), followed by Apidae (28 %, e.g. Ceratina catamarcensis, C. morrensis and Diadasia sp.), Andrenidae (11 %, e.g. Anthrenoides spp. and Callonychium sp.), Colletidae (8 %, i.e. Brachyglossula communis) and Megachilidae (5 %, i.e. Megachile sp.). Sixteen per cent of all collected bees were males. Direct observations of visitor behaviour revealed that female bees visited flowers exclusively for pollen collection, whereas male bees visited flowers in search of mates (Table 4). Bees were not observed to probe for floral nectar. Only two of the observed bee taxa, namely Brachyglossula communis and Lithurgus sp., are known as cactus specialists. The overall frequency of bee visits was 8·1 bees per flower per hour for females only, and 11·5 bees per flower per hour if both genders were considered. Roughly two-thirds (64 %) of all bee visits resulted in stigma contact. If only female bees were considered, the frequency of stigma contacts increased to three in four visits (76 %). Visits by ants, beetles and flies were infrequent, and largely destructive and ineffective, as the observed visits rarely resulted in stigma contact.
Among northern Argentine populations of white-flowered Echinopsis ancistrophora, also referred to as E. ancistrophora ssp. ancistrophora (e.g. Hunt, 1999; Schlumpberger and Raguso, 2008), we document extreme variation of floral characters between populations, with only moderate variation within populations. Our first two predictions, driven by pollination syndrome expectations based on floral morphology, were largely supported by our results. First, the four populations with the longest (≥15 cm) flowers showed evening anthesis, presented copious amounts of nectar (>100 µL) and were pollinated by sphingid moths. Second, the five populations with short-tubed flowers (≤12 cm) showed morning anthesis, presented meagre (<10 µL) nectar volumes and were pollinated by diurnal bees. For populations with intermediate floral morphology, we predicted either a generalized pollination strategy (i.e. moths and bees) or a novel pollinator, but observed neither. The two populations whose flowers ranged from 12 to 15 cm opened between midnight and dawn, presented only slightly more substantial (10–20 µL) nectar volumes, and, despite their length, and also were pollinated by diurnal bees. In some populations, Cyclocephala scarab beetles were abundant and destructive nocturnal floral visitors whose presence was not predicted by floral syndromes. Below we discuss these patterns and their implications for floral evolution in Echinopsis cacti.
In natural populations of E. ancistrophora, flower length was associated with floral traits of visual display, reward presentation and pollinator fit. These floral traits also were associated with the observed flower visitor assemblage in patterns consistent with the expectations of pollination syndromes. Thus, we reject the null hypothesis of random association between floral traits and visitor assemblages, as hawkmoths did not heavily visit or pollinate short-tubed E. ancistrophora populations. Previously, naïve Manduca sexta were shown to be attracted to E. ancistrophora flowers of all lengths in wind tunnel assays (Schlumpberger and Raguso, 2008); here, we found significant amounts of moth scales only in long flowers. In addition, we found low amounts of moth scales on a few short-tubed individuals from El Mollar and Campo Quijano (n = 1 flowers each, i.e. 1·7 % of all short-tubed flowers). The combination of evening flower closure (see above), low scale frequency on stigmas (Fig. 1B) and poor mechanical fit for moths with long proboscides (Fig. 2, Table 2) suggests that hawkmoths are unsuitable or at best uncommon and inefficient pollinators of short-tubed E. ancistrophora. Instead, the presence of scales may indicate that diurnal butterflies are occasional visitors to these flowers, which are closed during the active periods for most hawkmoths in this region (Moré, 2008).
One interesting finding was the apparent misinterpretation of functional floral depth among some of the Echinopsis populations studied. We were surprised to observe diurnal bee pollination in all examined populations with mean flower lengths ≤15 cm, despite their white corollas and relatively long, narrow floral tubes. Most non-cactus hawkmoth-pollinated plants growing in sympatry with E. ancistrophora populations had significantly shorter floral tubes or spurs (Fig. 3), several of them even shorter than the shortest bee-adapted E. ancistrophora flowers known (Fig. 1C). However, additional E. ancistrophora plants studied in cultivation, belonging to several other populations, confirm that the taxon is split into two groups: flowers shorter than 15 cm had predominantly diurnal anthesis and low nectar production (bee-pollinated group) and flowers longer than 15 cm had predominantly nocturnal anthesis and high nectar production (moth-pollinated group; data not shown). Similar patterns have emerged in an ongoing study on floral evolution in three other lineages of the genus Echinopsis (B. O. Schlumpberger, unpubl. res.). The paucity of nectar in most E. ancistrophora populations with flowers shorter than 15 cm (Fig. 1A), combined with the observed foraging behaviour of bee visitors, indicates that these flowers primarily offer pollen as a reward. Their functional depth probably does not include the floral tube, which may simply serve as a pedestal raising the corolla above the ground (see Cocucci, 1991), or could indicate a phylogenetic constraint reflecting the tubular floral architecture of this lineage's ancestor.
Another surprising finding was the variation in stigma position independent of pollination strategy, which we observed even within populations. For hawkmoth-pollinated flowers, both approach herkogamy (i.e. exserted stigma) and reverse herkogamy have been described (e.g. Raguso et al., 2003). It is unclear whether approach herkogamy vs. shorter stigmas should be more effective at receiving outcrossed pollen in trumpet-shaped, hawkmoth-pollinated flowers (see Kulbaba and Worley, 2008). The flower morphology of E. ancistrophora and proboscis lengths of the visiting hawkmoths suggest that the latter need to land and crawl into the flowers in order to deplete the nectar. Within-population variation in stigma position probably results from disruptive selection from different pollinators (bees vs. hawkmoths), as was suggested by Baker (1964) and Grant and Grant (1983). In our study plants, flower age does play a role as well, but does not explain the extensive observed variation.
We had expected a positive, linear correlation between floral depth and nectar volume among populations, given that all E. ancistrophora flowers are trumpet shaped and could offer nectar to different sized pollinators. Instead, the distribution of nectar standing crop volume among populations was a step function, with low (approx. 15 µL) to non-existent standing crops in populations with flowers <15 cm long, and copious mean volumes (≤170 µL) in the longest tubed flowers (Fig. 1). The observed differences in nectar quantity are not the result of enlarged nectaries in longer flowers, as nectar volume increased abruptly, while flower length increased linearly and intermediate nectar volumes were not observed (see Fig. 1). Therefore, these differences appear to represent adaptations to pollen-collecting bees vs. nectar-drinking moths in the absence of bimodally pollinated intermediate-length flowers. In nectar of long E. ancistrophora flowers, we occasionally measured an increasing sugar concentration from the top to the bottom of the floral tube. Such nectar gradients were described recently by Martins and Johnson (2007) from the floral spurs of sphingophilous orchids in Kenya. These authors suggested that nectar gradients might compel hawkmoths to probe more deeply into the flowers, enhancing physical contact with reproductive organs. Additional studies are needed to determine whether nectar concentration gradients in Echinopsis flowers result from active or passive secretion patterns within the nectar tube.
Interestingly, previous studies revealed that floral scent in this species complex does not covary with morphological characters in patterns predicted by floral syndromes. For reasons discussed elsewhere (Schlumpberger and Raguso, 2008), populations with hawkmoth-pollinated flowers (e.g. Termas de Reyes) did not have the strongest or most attractive floral scent, although the odours of long- and short-tubed flowers alike were sufficient to attract naïve Manduca sexta hawkmoths in wind tunnel bioassays. Further field experiments with bees and detrimental floral visitors (see below) will be needed to explore alternative explanations for the negative correlation between odour emission rates and floral tube length in this species complex.
Flowering in natural populations of E. ancistrophora occurred in a few short, synchronized waves during the year, but our observations were limited by the need to monitor 11 remote populations during several field trips. Additional observations made on cultivated E. ancistrophora plants grown under controlled conditions confirm pulsed flowering in only one to three events per year, each lasting 1–4 d (data not shown). The flowers usually last less than 24 h, possibly due to their size (e.g. high water investment), combined with water loss through evapotranspiration. We have observed similar flowering phenology in other Echinopsis taxa, as well as in more distantly related species in the genus Peniocereus (Raguso et al., 2003). From the standpoint of plant–pollinator food webs, Echinopsis flowers provide brief meals for their bee and hawkmoth pollinators, whose foraging demands are satisfied by other, more consistently available floral resources during the days to weeks of their adult lives (see Alarcón et al., 2008).
In E. ancistrophora we found that pollination by hawkmoths occurred only in populations whose floral morphology fits the sphingophilous pollination syndrome, with abundant nectar, maximum depth and anthesis at dusk. In contrast, bees pollinate not only populations with short flowers and purely diurnal anthesis, but also those with flowers of intermediate length and intermediate opening time, which appeared (superficially) to be adapted to sphingids and may represent transitional states.
Our evaluation of the spectrum of hawkmoth-adapted flowers growing in sympatry with two long-tubed populations of E. ancistrophora showed that only Craniolaria argentina has flowers of similar functional length (Fig. 3). At first glance, long-tubed E. ancistrophora flowers might be considered the most specialized moth flowers in these communities, with nectar being entirely accessible only to moths with the longest proboscides. Even moths with proboscis lengths up to 100 mm (Manduca spp., Fig. 2) must enter the upper, funnel-shaped part of the flowers in order to access the nectar resources fully (see Haber and Frankie, 1989). However, our evidence for moth pollination in the four longest-tubed populations of E. ancistrophora stems from the presence of moth scales on the stigmas of wilted flowers (Fig. 1) and the presence of pollen on light-trapped moths in long-tubed populations (Tables 2 and and3).3). Limited observations in the Termas de Reyes population revealed halictid bees gathering pollen in the wilting flowers. It is likely that pollen-seeking bees provide additional pollination to long-flowered, moth-pollinated E. ancistrophora by gathering pollen in the morning, as was recently found in three other moth-adapted Echinopsis species (B. O. Schlumpberger, unpubl. res.). Thus, the most conservative conclusion at present is that the four longest-tubed populations of E. ancistrophora have bimodal pollination strategies, with nectar-seeking moths and pollen-collecting bees as nocturnal and diurnal pollinators, respectively. Such a reproductive assurance strategy would be especially beneficial in the genus Echinopsis, most species of which are obligate outcrossers (Berger, 1929). Per-visit effectiveness experiments combined with measurements of gene flow are needed to assess how plants serviced by hawkmoths as well as bees might differ in fruit set and outcrossing rates.
Bimodal pollination involving bees has been described in other apparently hawkmoth-adapted species (Barthell and Knops, 1997; Miyake and Yahara, 1998). Examples for mixed pollination in cacti with predominantly nocturnal anthesis by both diurnal and nocturnal visitors were found in Echinopsis atacamensis ssp. pasacana (Schlumpberger and Badano, 2005) and a few other columnar cacti (Sahley, 1996; Dar et al., 2006; Rivera-Marchant & Ackerman, 2006). Such adaptation to sets of different pollinators is thought to have especially evolved near the northern and southern distributional limits of columnar cacti, as a response to low reliability of specialized pollinators, i.e. nectarivorous bats (Fleming et al., 2001). Aigner (2001) has argued that the attraction of additional pollinator taxa should be favoured if the required phenotypic trade-offs do not reduce the plant's net reproductive fitness. If long-tubed flowers evolved from a short-tubed, bee-pollinated ancestry in E. ancistrophora, changes associated with effective hawkmoth pollination (e.g. extended nectar tube and supplemented nectar volumes) would be unlikely to affect pollen foraging bees. In this case, we would have increased apparent floral specialization in respect to nectar accessibility, coupled with a decrease in the specificity of the potential pollinator spectrum. Similarly, Armbruster (2006) distinguished between evolutionary and ecological specialization. Although the floral phenotype of plants may evolve towards specialization to a subset of available pollinators or pollinator groups, their net ecological specialization may actually decrease (Aigner, 2001). If, instead, the direction of evolution in this lineage is from a long-tubed, nectar-rich ancestry to shorter-tubed, nectarless flowers, then some populations will have specialized in bee pollination through the loss of traits that attract sphingids and translate their visits into effective pollen transfer. Phylogenetic analysis (in progress) will be used to determine which of these scenarios most accurately depicts the order of evolution in E. ancistrophora.
The assemblages of flower-visiting bees differed among populations with short-tubed flowers, which may simply reflect limited sampling time spread across many ephemerally blooming populations. However, with few, short, synchronized flowering events these flowers are best considered as transients within the established flower–pollinator networks in their communities (e.g. Medan et al., 2006), and flower visitor assemblages may simply vary among flowering events within and among years. Interestingly, specialized cactus bees, which are abundant in these regions and visit Echinopsis and other cactus genera (Roig-Alsina and Schlumpberger, 2008) were only occasionally found in some of the studied populations (Table 4). We assume that the extremely short, pulsed flowering of E. ancistrophora together with relatively low cactus diversity at some of the sites cannot support viable populations of these specialists. Despite these observations, generalized bee pollination appears to be an effective strategy for most populations of E. ancistrophora, given the relatively high (mean 76 %) observed stigma contact frequency for female bees, and that specialized bees accounted for a low percentage of all observed visits. Pollination exclusively by generalized bees was observed in cacti from Córdoba, Argentina (Schlumpberger, 2002). In contrast, studies of cacti in southern Brazil revealed that specialized cactus bees (e.g. Arhysosage, Andrenidae) touched the stigmas in more than 90 % of their visits, whereas generalized bees (e.g. Dialictus, Halictidae) only contacted the stigmas in 6–29 % of their visits (Schlindwein and Wittmann, 1995). For North American bee-pollinated cacti, most studies documented specialized cactus bees as well, usually outperforming co-visiting polylectic bees (Simpson and Neff, 1987; Johnson, 1992; McIntosh, 2005).
Studies on the pollination of intermediate flowers are rare and offer unique opportunities to study floral syndrome evolution. One possible interpretation for the occurrence of intermediate populations in our study is that they are of hybrid origin, and offer too little nectar and open too late at night (see Gregory, 1964; Raguso et al., 2003) to enlist hawkmoths as pollinators. Another possibility is that they represent a transition from a long-tubed ancestry, in which nectar reduction mitigates physiological or ecological costs in the absence of selection for hawkmoth pollination. In Volcán, flower opening times were spread over more than 5 h, indicating a less precise timing than in populations with shorter or longer flowers respectively. Anthesis time varied more strongly than did flower length and nectar quantity at Volcán, which may indicate that it is a more plastic trait. In the case of E. ancistrophora, the two intermediate populations were not intermediate in nectar volume, which resembled the low values of short-tubed, bee-adapted populations. It is unclear whether poor nectar rewards or the shifted anthesis time later at night was most responsible for the lack of observed moth scales. These flowers were long enough (10–15 cm) to have ensured scale deposition on stigmas if they had been visited by the moths observed in our study (see Fig. 2, Table 3).
Nectar production in the quantities measured for the four hawkmoth-pollinated populations has obvious physiological costs in a desert environment (Carroll et al., 2001 and references herein). Our observations suggest that the ecological costs of making flowers apparent to florivores may be just as high, with nocturnal Cyclocephala beetles visiting long-tubed flowers of E. ancistrophora. Flowers adapted to these beetles as pollinators often are structurally reinforced so as to withstand their destructive behaviour (Young, 1988). However, the nocturnal flowers of E. ancistrophora have no such defences, and were damaged by the beetles (B. O. Schlumpberger, pers. observ.). Our previous study revealed an inverse relationship between floral length and scent emission rates among E. ancistrophora populations, against our expectations that hawkmoth pollination should require the strongest floral scents (Schlumpberger and Raguso, 2008). Perhaps Cyclocephala beetles have selected for reduced scent in populations with hawkmoth pollination. Manipulative field experiments will be needed to measure the magnitude and consequences of beetle responses to augmented floral scent.
Our results suggest incipient specialization in E. ancistrophora, with population-level differences in apparent flower–pollinator adaptation, either to bees or hawkmoths (potentially plus bees). However, the presence of destructive beetles in populations with nocturnal anthesis underlines the potential for enemies to influence the evolution of floral phenotype. The observed intraspecific variation of floral traits and flower visitors among isolated populations, including intermediate, potentially transitional stages, makes this plant group an ideal model for studying plant–pollinator interactions and their evolutionary transitions.
We are grateful to Roberto Kiesling, Martin Lowry, Roberto Neumann, Jörg Piltz and Mats Winberg for information on plant locations. We also cordially thank Mats Winberg, Jörg Piltz and Matthias Uhlig for providing cultivated plant material, Marina Arce-Miller for help with fieldwork, Roberto Kiesling for facilitating permits in Argentina, and Ian J. Kitching for help with the identification of the hawkmoths. The first author was supported by a Feodor Lynen fellowship of the Alexander von Humboldt-Foundation. Additional help was provided by the research fund of the Cactus and Succulent Society of America. Travel and field work by B.O.S., R.A.R. and M.M were supported in part by National Geographic Foundation grant no. 7534-03 and by US National Science Foundation grant DEB 0317217. M.M., A.N.S. and A.A.C. are fellowship holder or researcher of CONICET.