Unpredictability of nectar nicotine promotes outcrossing by hummingbirds in Nicotiana attenuata


(e-mail baldwin@ice.mpg.de).


Many plants use sophisticated strategies to maximize their reproductive success via outcrossing. Nicotiana attenuata flowers produce nectar with nicotine at concentrations that are repellent to hummingbirds, increasing the number of flowers visited per plant. In choice tests using native hummingbirds, we show that these important pollinators learn to tolerate high-nicotine nectar but prefer low-nicotine nectar, and show no signs of nicotine addiction. Nectar nicotine concentrations, unlike those of other vegetative tissues, are unpredictably variable among flowers, not only among populations, but also within populations, and even among flowers within an inflorescence. To evaluate whether variations in nectar nicotine concentrations increase outcrossing, polymorphic microsatellite markers, optimized to evaluate paternity in native N. attenuata populations, were used to compare outcrossing in plants silenced for expression of a biosynthetic gene for nicotine production (Napmt1/2) and in control empty vector plants, which were antherectomized and transplanted into native populations. When only exposed to hummingbird pollinators, seeds produced by flowers with nicotine in their nectar had a greater number of genetically different sires, compared to seeds from nicotine-free flowers. As the variation in nectar nicotine levels among flowers in an inflorescence decreased in N. attenuata plants silenced in various combinations of three Dicer-like (DCL) proteins, small RNAs are probably involved in the unpredictable variation in nectar nicotine levels within a plant.


Floral nectar frequently contains toxic compounds (Kerner von Marilaun, 1879; Adler, 2000), which have mainly deterrent functions, and either repel unwanted flower visitors such as nectar robbers and nectar thieves (Stephenson, 1981; Kessler et al., 2008) and inefficient pollinators (Johnson et al., 2006), or increase nectar shelf-life by antimicrobial properties (Carter and Thornburg, 2004). The existence of nectar toxins was long thought to be an unavoidable consequence of selection for resistance traits in response to folivorous herbivores and pathogens, and the accumulation of toxins in nectar and their repellent effects on pollinators may be an example of collateral damage resulting from deploying chemical defenses, commonly referred to as the ‘pleiotrophy hypothesis’ (Adler, 2000). However, recent studies suggest that nectar toxins may also increase pollinator fidelity. For example, high levels of nectar alkaloids were suggested to benefit plants via increased pollen export (Irwin and Adler, 2008).

Flowers of the native tobacco, Nicotiana attenuata, accumulate the alkaloid nicotine, which is synthesized in the roots and transported throughout the plant, where it provides resistance against herbivores (Steppuhn et al., 2004; Steppuhn and Baldwin, 2007). Nectar laced with nicotine protects flowers against nectar robbing and florivores, and changes the behavior of an important pollinator, hummingbirds (Archilochus alexandri) (Kessler et al., 2008). Although nicotine is a repellant that decreases nectar removal by floral visitors (Kessler and Baldwin, 2007), the presence of this toxin increases plant fitness through both maternal and paternal means (Kessler et al., 2008). The proposed mechanism for the increase in plant fitness is through a change in the nectaring behavior of hummingbirds, which visit more flowers per plant if a plant produces nicotine (Kessler et al., 2008). Why hummingbirds continue to visit flowers that contain repellent nectar and whether toxic nectar benefits the plant by increasing outcrossing rates remain important unanswered questions. Knowledge of the mechanisms responsible for the transport and accumulation of nicotine into nectar after its synthesis in roots (Hildreth et al., 2011) may also help to address the pleiotropy hypothesis (Adler, 2000).

Outcrossing is thought to benefit self-compatible plants such as N. attenuata by providing the genetic variability that is important for coping with variable and unpredictable environmental challenges. For example, plants germinated from outcrossed flowers have increased herbivore resistance compared to plants germinated from self-pollinated flowers (Mescher et al., 2009; Bello-Bedoy and Nunez-Farfan, 2011). Moreover, outcrossing increases seed production in some self-compatible plants (Yang et al., 2005). Outcrossing rates may be determined independently of floral traits (Kameyama and Kudo, 2009), or by groups of traits such as flower size or color. However, the influence of single floral chemical traits, such as nectar constituents, on outcrossing rates remains largely unstudied.

One of the most powerful means to quantify outcrossing within a plant species is using microsatellites (Ashman and Morgan, 2004). The genomic abundance (van Treuren et al., 1997) and polymorphism of microsatellites within and between species (Tautz and Renz, 1984; Queller et al., 1993) has facilitated their use as molecular markers in population genetics (Goldstein et al., 1999), systematics (Goldstein and Pollock, 1997) and genome mapping (Weissenbach et al., 1992). Moreover, microsatellite sequences possess high mutation rates, estimated at 10−2 to 10−3 per locus and gamete and generation (Tautz, 1989), which makes them a powerful tool for analysis of mating systems (Awadalla and Ritland, 1997; Korbecka et al., 2011) and kinship (Queller et al., 1993; Southworth, 2009).

Although the occurrence of toxic nectar constituents is widespread in angiosperms (Adler, 2000), their variability and the mechanisms responsible for this variability within species has not been examined. Such information may be very helpful in understanding how nectar alkaloids change the behavior of pollinators (Kessler et al., 2008). We measured nectar nicotine concentrations of individual N. attenuata flowers from plants from many different populations grown under controlled conditions. We found that this floral trait was highly variable and that no other measured floral or developmental trait predicted nectar nicotine concentrations. We conducted choice tests using populations of native hummingbirds, and found that hummingbirds avoided the high end of the range of nectar nicotine concentrations (> 50 μm) when given a choice, but tolerated these concentrations if alternatives were not available. These choice tests revealed that the variance in nectar nicotine strongly influences foraging behavior, and suggested that the presence of this alkaloid in nectar may influence outcrossing rates. In field experiments with N. attenuata plants silenced for expression of a biosynthetic gene for nicotine production (Napmt1/2), we quantified the influence of nectar nicotine on outcrossing rates using a set of 16 polymorphic microsatellite markers optimized to resolve the paternity of seeds in the native populations. As this alkaloid, through its effects on hummingbird foraging behavior, increased outcrossing rates, we next asked how this unpredictable variance is generated, and to this end, examined the variance in nectar nicotine levels in plants silenced for expression of various combinations of three Dicer-like (DCL) proteins known to process small RNAs. Without a heritable mechanism, the unpredictability of this important floral trait could not evolve.


Nectar nicotine concentrations are highly variable

When nectar nicotine levels were measured in 322 flowers from 49 plants from ten native populations under controlled glasshouse conditions using a highly sensitive and accurate LC-MS3 analytical procedure, nicotine levels were found to be highly variable (Figure 1). No predictable spatial patterns with regard to ontogeny or floral meristem location were found. Also no differences in nicotine concentrations between night-opening flowers and morning-opening flowers were found (paired Student’s t test, t37 = 1.20, = 0.24). Similarly, no significant correlation between nectar nicotine concentration and nectar volume (= 1.0401x + 17.16, R2 = 0.004), nectar sugar concentration (= −0.0557x + 18.40, R2 = 0.012) or emission of benzylacetone (= −0.0371x + 21.89, R2 = 0.00 001), the flower’s most abundant scent, were found. Nicotine concentrations for flowers were highly variable for flowers within a single inflorescence (estimate of variance, s2 = 456.5), for plants within a population (s2 = 20.9) and among populations (s2 = 404.0). While one plant population (population 7; Figure 1) had high nicotine concentrations, the between-flower variance tended to equal the between-population variance (nested anova, F9,268 = 1.13, = 0.46). However, the variance among plants within a population was significantly lower than the variance within an inflorescence (F39,268 = 21.86, < 0.001) or among populations (F9,39 = 19.34, = < 0.001). The minimum difference among flowers within an inflorescence was 0.7 μm, and the maximum difference was 259.5 μm, while the minimum difference among plants within populations was 5.2 μm and the maximum difference was 74.2 μm. Most flowers (236) contained nectar nicotine below the 50 μm rejection threshold observed for native hummingbirds (see below), 37 flowers (11%) had higher concentrations, and 49 flowers (15%) had values between 25–50 μm.

Figure 1.

 Nectar nicotine levels are highly variable. Values are nectar nicotine concentrations per flower (mean ± SD) from four or five randomly chosen Nicotiana attenuata plants (A–E), germinated from individual-plant collections from ten native populations (1–10). As an example of within-plant variance, nicotine concentrations for single flowers of plant 7D are shown in the inset according their positions within the inflorescence (S0–S4). The horizontal lines indicate the 25 and 50 μM levels that represents the critical tolerance thresholds for hummingbirds (see Figure 2). For the plants whose results are shown in the inset, four of the six flowers had nectar nicotine levels above the 50 μM rejection threshold for the hummingbirds.

Hummingbirds avoid nicotine but tolerate it

When a native breeding population of A. alexandri was provided with nectar of various nicotine concentrations in artificial flowers, they did not distinguish between nicotine-free flowers or naturally occurring concentrations of nicotine of either 10 or 25 μM (Figure 2a). However, at 50 μm, the hummingbirds were clearly repelled, frequently ‘spitting’ and vocalizing after nectaring (Figure 2a). This rejection threshold may be altered by gradually increasing the concentrations in the feeders over longer periods of time. When provided for 1 week with 10 μm nicotine, followed by a 2nd week of 25 μm nicotine, hummingbirds were no longer repelled by 50 μm nicotine (Figure 2b). However, this effect was short-lived, lasting only as long as a few individuals had not discovered the nicotine-free nectar. Once discovered, the nicotine-free solution was strongly preferred over the 50 μm nicotine solution, usually within 30 min of the start of a trial and after much vocalization (Figure 2b). These results suggest that hummingbirds do not become addicted to nicotine, but are clearly able to tolerate it after long-term exposures in the absence of alternative food sources.

Figure 2.

 Hummingbirds tolerate high-nicotine nectar, but prefer low-nicotine nectar. (a) Hummingbird (Achilochus alexandri) feeding time per visit (mean ± SE) in a choice assay between a pure sugar solution (con) or the same sugar solution containing either 10, 25 or 50 μM nicotine. (b) Feeding durations (mean ± SE) for A. alexandri on a pure sugar solution (con) compared with the same sugar solution containing 50 μM nicotine after a period of experiencing no nicotine in the feeders (pre-exposure) or after experiencing one week of 10 μM nicotine/sugar solution followed by a week of 25 μM nicotine/sugar solution in the feeders (post-exposure). The results were separated into two time intervals, 0-30 min and 30-60 min, during which the hummingbirds collectively discovered the control nectar feeder. Asterisks indicate significant differences as determined by Student’s t test (< 0.05).

Nectar nicotine promotes outcrossing

To test whether the larger number of hummingbird visits per flower results in more diverse pollen loads and higher outcrossing rates for plants, we transplanted plants in which expression of a biosynthetic gene for nicotine production (Napmt1/2) was silenced (PMT plants) and control empty vector (EV) plants with normal levels of nicotine (27.4 μM ± 17.3, mean ± SD) into two native populations in the Great Basin Desert, Utah. The plants with emasculated (antherectomized) flowers were covered at night to exclude nocturnal pollinators, and hummingbirds were the only daytime pollinators observed for these populations. In total, 69 flowers of each genotype were treated on four different days, from which 16 capsules on EV plants and only seven capsules on PMT plants matured, a difference that reflects the beneficial effect of nicotine on maternal fitness. All viable seeds were germinated and genotyped using 16 microsatellite markers that allowed seedling paternity to be assigned to pollen donors in the surrounding native population. Seeds from capsules from EV plants were sired by a mean of 5.1 paternal genotypes compared to 3.6 paternal genotypes for the seeds of capsules from PMT plants (Figure 3a,b: repeated-measures anova, F1,5 = 8.56, = 0.033). Due to the plant size and the pollinator exclusion methods used, we were restricted to a maximum of three flowers per inflorescence in this experiment. In larger plants with a greater number of flowers per inflorescence, we would expect a larger beneficial effect of nectar nicotine on capsule set and outcrossing.

Figure 3.

 Silencing nicotine accumulation in the nectar reduces outcrossing rates compared to isogenic nicotine-producing plants grown in experimental native N. attenuata populations, antherectomized and exposed only to hummingbird pollinators. (a) Number of paternal genotypes (mean ± SE) per capsule for nicotine-containing empty vector (EV) plants or nicotine-free PMT plants pollinated by hummingbirds in two native populations in the Great Basin Desert. The asterisk indicates a significant difference as determined by repeated-measures anova (< 0.05). (b) Distribution of paternal genotypes found in seedlings germinated from seeds of individual capsules produced from antherectomized flowers of EV or PMT plants using microsatellite markers. The same colors represent identical paternal genotypes, but different individuals, that were resolved using one to eight microsatellite markers.

Are small RNAs involved in creating the variability in nectar nicotine within an inflorescence?

Using N. attenuata plants silenced for three Dicer-like (DCL) proteins (NaDCL2, NaDCL3 and NaDCL4), and various combinations, we examined whether small RNAs are involved in the variation in nectar nicotine among flowers. These DCL endoribonucleases process small RNAs before incorporation into effector complexes for use in various biological processes (Meister and Tuschl, 2004). Silencing the expression of NaDCL3 and NaDCL4 was found to regulate nicotine accumulation in leaves during herbivore attack in N. attenuata (Bozorov et al., 2012). Analysis of nectar nicotine levels revealed that EV transformants have similar median nectar nicotine concentrations to the DCL lines, except for DCL4 and DCL3 × 4, which have reduced levels of nectar nicotine (Figure 4a, one-way anova, F1,19 = 45,46, < 0.001). However, the within-inflorescence variability in nectar nicotine levels of EV plants was significantly greater in comparison with that of the three DCL-silenced lines and their crosses (Figure 4b, F1,19 = 11.60, < 0.003), suggesting a role for small RNAs in generating the variance in nectar nicotine levels.

Figure 4.

 Dicer-like (DCL) protein-silenced lines of N. attenuata have less variable nectar nicotine concentrations than isogenic EV-transformed plants. (a) Median nectar nicotine concentration (mean ± SD) for 10 flowers per plant (= 3) for EV, DCL2, DCL3 and DCL4 plants or crosses of the various DCL lines.
(b) Quotient of the interdecile range (difference between the 10% and 90% percentiles) and the median (mean ± SD), indicating the variability in nicotine content between flowers from one plant.Different letters indicate significantly different means as determined by one-way anova: < 0.001 in (a) and < 0.003 in (b).


This paper presents a rigorous survey of the variation in a single nectar chemical constituent and provides evidence that a nectar toxin increases outcrossing rates by altering the behavior of floral visitors, results that are consistent with the pollinator fidelity hypothesis (Adler, 2000). Furthermore, the analysis suggests that small RNAs are involved in generating the unpredictable variation in nectar nicotine concentrations within an inflorescence.

Our field experiments revealed that, when nectar lacks nicotine, outcrossing rates decrease when hummingbirds are the main pollinators. Our experiments compared outcrossing rates of plants lacking nectar nicotine (0 μm; PMT) with plants whose flowers contained a mean nicotine concentration of 27.6 μm (EV), and hence may be interpreted as not directly testing the influence of the variance in nicotine concentrations. On average, in the inflorescence of an EV plant, only two of ten flowers will have nicotine concentrations above 25 μm, and hence be repellent to hummingbirds (although the plant shown in the inset to Figure 1 has four flowers with levels greater than 50 μm). As hummingbirds do not distinguish between 0 and 25 μm nicotine sugar solutions (Figure 2a), the alterations in hummingbird behavior that decreased outcrossing rates in PMT plants are probably a reflection of the unpredictable nicotine concentration differences among individual flowers in inflorescences of EV plants, something that is completely lacking in PMT plants. This unpredictable variability in nectar nicotine concentrations, particularly within an inflorescence, probably keeps hummingbirds searching for low-nicotine flowers within a plant, and hence is responsible for the greater movement among flowers (Kessler et al., 2008). Moreover, if pollinators are able to sense the presence of nectar and do not visit flowers without nectar, the residual nectar remaining in high-nicotine flowers would increase flower visits and outcrossing rates in flowers with high-nicotine nectar. The residual nectar may also attract nocturnal pollinators, which appear after the activity period of hummingbirds in the dusk, or vice versa, may attract hummingbirds in the morning after the visits of nocturnal pollinators. Additional detailed behavioral analyses are required to understand how variable nectar nicotine levels influence the behavior and pollen transfer by nocturnal pollinators. However, previous observations of the white-lined sphinx moth (Hyles lineata) suggested comparable positive effects on outcrossing in a nocturnal pollinator; these moths, as with the hummingbirds, showed a decreased number of visits to flowers on plants unable to produce nicotine (Kessler et al., 2008). Why hummingbirds visit fewer flowers on PMT plants than on EV plants (Kessler et al., 2008), particularly when they are known to prefer nicotine-free nectar, and why they nectar longer at PMT flowers and remove more nectar compared to visits to EV flowers (Kessler and Baldwin, 2007) remains an unanswered question that deserves additional research. Silencing PMT expression may have changed the abundance of other minor unmeasured nectar constitutents, or somehow changed the nectar microbial community, such that hummingbird behavior was altered. However, such explanations do not explain why hummingbirds remove more nectar from PMT flowers. Further investigations that examine hummingbird nectaring behavior within an inflorescence are required.

For methodological and regulatory reasons, we had to experimentally exclude geitonogamy when comparing outcrossing rates in EV and PMT plants in our field experiments. Flowers of the experimental plants were emasculated to exclude the possibility of geitonogamy, i.e. pollination of a flower with pollen from another flower on the same plant. However, this does not weaken the conclusions that we can draw regarding the effect of nicotine on outcrossing, even if outcrossing only accounts for a proportion of the seeds sired in a capsule produced from unemasculated flowers. Although visits within an inflorescence may increase the likelihood of geitonogamy, the increasing chance of a flower being visited multiple times also increases the outcrossing probability. Furthermore, the first flower probed in a floral display typically receives the majority of outcross pollen, and successive flowers receive increasing quantities of geitonogamous self pollen (Karron et al., 2009), increasing the number of outcrossed pollen per flower if flowers receive multiple visits. It is also likely that flowers are able to discriminate between self and outcrossed pollen in such a way that increases outcrossing in self-compatible plants, as most self-compatible plant species also outcross (Goodwillie et al., 2005). In unmanipulated N. attenuata plants growing in native populations, as many as 50% of seeds are sired by outcrossed pollen (Figure S5), highlighting the importance of outcrossing for this self-compatible species.

Nicotine is synthesized in the roots (Hibi et al., 1994; Shoji et al., 2009; Shoji and Hashimoto, 2011a) and transported to above-ground tissues, where it accumulates in leaves and stems in largely allometrically predictable patterns consistent with the ‘optimal defense’ theory (McKey, 1974): young leaves show higher concentrations than older leaves, unless damaged by herbivores, which increases the concentrations in attacked leaves and the allometric pattern of all above-ground parts, including entire flowers (Ohnmeiss and Baldwin, 1994, 2000; Baldwin and Karb, 1995; Diezel et al., 2011). Furthermore, alkaloid levels of floral tissues were found to be positively correlated with those of leaf tissues (Irwin and Adler, 2006), suggesting that alkaloids also function in the direct defense of flowers. The high variability in nectar nicotine levels among flowers within an inflorescence as found in the present study is in marked contrast to the more predictable pattern observed in leaf tissues, and even in the sepals subtending flowers. Nectar alkaloids were thought to be an unavoidable consequence of the defense of vegetative tissues (pleiotropy hypothesis), but also clearly serve as defenses against nectar robbers or florivores (Stephenson, 1981; Kessler et al., 2008). The unpredictable variability of nectar nicotine levels clearly suggests a function in addition to defense. Nectar nicotine levels were not correlated with any other measured floral trait such as nectar volume, nectar sugar concentration or emission of benzylacetone, the main floral attractant. These results are consistent with other studies that have found that floral traits evolve independently of each other in response to different floral visitors (Mazer and Hultgard, 1993; Frey et al., 2011). The high variance in nectar nicotine levels may be adaptive for the plant by increasing outcrossing rates, and, as such, falsifies the null hypothesis that nectar nicotine is a maladaptive consequence of defending vegetative tissues using this toxin (Adler, 2000).

How this unusual variability is generated is an interesting question, not only from a physiological perspective but also from evolutionary one. The nicotine uptake permease NUP1 (Hildreth et al., 2011), and other transporters involved in nicotine distribution throughout the plant (Kazufumi, 2006; Morita et al., 2009; Shitan et al., 2009), may provide regulatory targets that may be responsible for generating the unpredictably high variance in nicotine concentrations in the nectar. Our data on nectar nicotine levels in EV and DCL lines suggest that small RNAs are involved creating the variability within an inflorescence. Determining exactly how DCL-mediated small RNAs generate the variance in nicotine nectar levels requires a substantial amount of additional work. Small RNAs are known to process transposons (Girard and Hannon, 2008; Kuang et al., 2009), and ‘domesticated’ transposons are unusually common in floral tissues (Whitford et al., 2007; Benjak et al., 2008). Transposons that target proteins involved in nicotine transport in nectary cells may be able to alter the transport of nicotine to the nectar to produce the high variance in nicotine concentrations. However, many other scenarios are possible.

In this study, we found a direct correlation between the influence of nectar nicotine contents on hummingbird behavior and experimentally measured outcrossing rates. We propose that the high variation in nectar nicotine levels influences hummingbird behavior so that more flowers are visited, thereby increasing outcrossing rates. Many environmental factors are known to influence hummingbird behavior, which may influence the scenario we propose above. For example, the availability of alternative food sources may alter the response of hummingbirds to nectar nicotine (Hurly and Oseen, 1999, Bacon et al., 2010). Their nectar-foraging behavior may be cognition-mediated (Shafir et al., 2003; Drezner-Levy and Shafir, 2007) and influenced by prior feeding experiences. The response to one nectar constituent may influence the response to another. For example, variability in nectar sugar composition may alter the response to other nectar constituents (Gegear et al., 2007), and variability in nectar sugar alone is known to have dramatic consequences on pollinator behavior in relation to reward quantity (Bateson et al., 2003). In N. attenuata, the variance in nectar sugar concentrations (between 15.4 and 22.8% among plants of the ten populations surveyed) was dwarfed by the variation in nectar nicotine. As in humans, hummingbirds may vary in their ability to detoxify nicotine and hence may vary in their tolerance of this alkaloid. Assessing all these potentially complicating factors as to how nectar nicotine levels may influence hummingbird behavior and thereby outcrossing rates would require a major research effort. Directly disproving the hypothesis presented here would involve engineering a plant to produce uniform levels of nectar nicotine and comparing outcrossing rates with plants with normally variable nectar nicotine. However, this would require a better understanding of how this variance is created.

In summary, we propose that small RNAs regulate the nectar nicotine concentrations of flowers, leading to unpredictable nicotine concentrations and causing hummingbirds to visit more flowers in search of low-nicotine flowers, which in turn increases outcrossing rates. Hence, through a clever use of chemistry, plants manipulate the behavior of their floral visitors to maximize their reproductive success.

Experimental Procedures

Plant material

Wild-type (WT) N. attenuata plants obtained from seeds collected from a native population in 1989 at the DI Ranch (Santa Clara, UT) and subsequently inbred for 14 generations were transformed with Agrobacterium tumefaciens (strain LBA 4404) containing construct pRESC5PMT (Krügel et al., 2002; Steppuhn et al., 2004) to silence N. attenuata putrescine N-methyl transferase1. The vector construction and transformation procedures have been described previously. The same genotype, but of the 23rd inbred generation, was transformed to individually silence three Dicer-like (DCL) single-copy genes (NaDCL2–4). Partial cDNA fragments of NaDCL2, NaDCL3 and NaDCL4 were inserted into pSOL8 (Gase et al., 2011) and pRESC5 binary transformation vectors as inverted repeats (Bubner et al., 2006). Three lines, all fully characterized (Bozorov et al., 2012), with the strongest phenotypes were chosen: pSOL8DCL2 (A-09-584-10), pSOL8DCL3 (A-08-463-4) and pRESC5DCL4 (A-07-441-5). These were crossed with each other, i.e. DCL2 × DCL3, DCL4 × DCL2 and DCL4 × DCL3 (Bozorov et al., 2012).

All plants for nectar nicotine measurements were grown individually in 1 liter pots in the glasshouse at 26–28°C under 16 h of light supplied by Philips Sun-T Agro 400 (Philips, http://www.lighting.philips.co.uk/) sodium lights.

For the fieldwork with nicotine-silenced plants, we chose the plant lines that, according to studies of plants grown in the glasshouse and field, had completely normal growth and the strongest reductions in leaf nicotine accumulation, i.e. A-03-108-3. Flowers of PMT plants did not differ from flowers of EV plants in terms of nectar volume, nectar sugar concentration, nectar composition or floral volatile emission (Kessler and Baldwin, 2007). Three seasons of field work with this line revealed that its nectar is preferred and removed at a greater rate by the native floral visitor community than from wild-type and EV plants. As a control, we used EV line A-04-266-3 transformed with pSOL3NC (Bubner et al., 2006), which is known to be completely comparable to wild-type plants (Schwachtje et al., 2008). Seeds of the transformed N. attenuata lines were imported under US Department of Agriculture Animal and Plant Health Inspection Service (APHIS) notification number 07-341-101n, and the field experiments were performed under notification number 06-242-02r and 06-242 03r. Seed germination was performed as described by Krügel et al. (2002). At the seedling stage, plants were transferred into previously hydrated 50 mm peat pellets (Jiffy 703 http://www.jiffygroup.com/jiffy) 15 days after germination, and were gradually adapted to the environmental conditions of high sun and low relative humidity of the Great Basin Desert habitat over 14 days by keeping the seedlings in the shade. Adapted size-matched seedlings were transplanted into two native N. attenuata populations in the vicinity of the Lytle Ranch Preserve (Santa Clara, UT). Seedlings were watered every other day until roots were established. To obtain a collection of native genotypes reflecting the genetic diversity in the local area for experimental outcrossings, nectar measurements and development of microsatellite markers, seeds from 17 native populations were collected between 1993 and 2009 within a 200 km radius of the native populations in which the releases were performed. Ten of these populations were selected to assess the variance in nicotine nectar levels (described below). From each population (1–10; Figure 1), seeds of five single-plant collections (A–E) were germinated on separate GB5 (Duchefa, http://www.duchefa.com/) plates, and five plants of each genotype were transferred into the glasshouse in 1 liter pots. One collection of population 9 did not germinate and was excluded from the analysis. To develop the microsatellite markers for paternity analysis of seeds and perform experimental outcrossings, bulk collections from seven populations growing within a 40km radius of the experimental populations were bulk-germinated, and 20 plants (G1–G20) were selected for microsatellite development (described below).

Nectar nicotine concentration

Nectar was collected in the morning (6:00–8:00 am) from newly opened flowers, using a 25 μl glass capillary and a refined technique (Kessler and Baldwin, 2007) that avoids damage to the ovary and contamination of the nectar. Nectar from eight flowers per plant was collected separately in 1.5 ml Eppendorf tubes. Then 2 or 1.5 μl nectar (for flowers with low nectar volume) were transferred into a new tube containing 400 μl water and 20 ng of the internal standard nicotine-D3. Particles were removed by centrifugation (10 min at 12 000 g) and the supernatant was transferred into GC vials (http://www.wicom.de). A 10 μl aliquot of the solution were analyzed using a Varian 1200 triple quad LC-MS system (http://www.varianinc.com) connected to an ESI source with a capillary voltage of 35 V. Solvent A comprised ultrapure (18.2 MΩ) Millipore H2O (Millipore Corporation, http://www.millipore.com/) + 0.1% ammonium hydroxide solution 25%, pH 10; solvent B was methanol. The gradient of 0 min/5% B, 0.5 min/5% B, 2 min/80% B, 6.5 min/80% B, 8.5 min/5% B, 10 min/5% B was used on a Phenomenex Gemini NX 5 × 2 mm column (http://www.phenomenex.com), particle size 3 μm. The transition of the precursor ion nicotine [M+H]+ = 163 and nicotine-D3 [M+H]+ = 166 to the fragment (m/z) = 130 at a collision energy of 14.5 V was recorded for quantification.

To test the reproducibility of the mean and within-plant and population variance, four plants from two populations (from Figure 1) were re-grown from seeds in the glasshouse 2 years after the first analysis, and nectar nicotine was again quantified. Both the means and the variance structure were found to be highly reproducible (Student’s t test, > 0.05; Figure S1).

Hummingbird assays

The bioassays were performed at the Lytle Ranch Preserve using resident hummingbirds (Archilochus alexandri) in June 2009. All experimental measures were performed between the two main activity periods in the morning and evening (11:00–17:30) to avoid the typical territorial behavior that occurs during peak activity periods. Birds were provided with a 28% sucrose solution (the mean sugar concentration of N. attenuata flowers in the field) from three-port feeders, and test birds were self-recruited to the experimental arena in groups of 6–12 individuals. The circular experimental arena had six equally spaced feeding stations, each equipped with a white star-shaped lid surrounding a 3 mm hole (Kessler and Baldwin, 2007). Nicotine at concentrations of 0 (control) and 10, 25 or 50 μm was dissolved in 10 ml of 28% sucrose solution, and treatment and control solutions were used in alternate feeding stations. Each test period lasted 1 h, and the duration (in sec) of each nectaring bout (time for which the birds’ beaks were immersed in the test solutions) at each feeding station was recorded. Nectar removal rates are strongly correlated with nectaring times of A. alexandri (= 0.0087x + 0.1701; R= 0.83; Kessler et al., 2008). When performing these choice experiments, we repeatedly observed that feeding time on control solution depended on the repellence of the treatment solution it was paired with (Figure 2a) (Kessler and Baldwin, 2007). The more repellent a treatment solution, the longer a hummingbird nectared on a control solution and vice versa.

To accustom the hummingbirds to nicotine, the three-port feeders were filled for 1 week with a 28% sucrose solution containing 10 μm nicotine, followed by a week when the feeders were filled with the same sucrose solution containing 25 μm nicotine. After these 2 weeks, a choice test with the 50 μm nicotine solution as the test solution was repeated as described above.

Field outcrossing experiment

Plants in which expression of a biosynthetic gene for nicotine production (Napmt1/2) was silenced (PMT plants) and control empty vector (EV) plants were planted into two experimental native populations at the Lytle Ranch Preserve during the 2009 field season. The Ut-WT accession that was transformed was originally collected from a native population growing 20 km north in the same watershed as the experimental populations. When the transplanted transformed plants started flowering, we emasculated three flowers by antherectomization, labeled these flowers with a small white string at the base of the pedicle, and removed all remaining flowers (Kessler et al., 2008) from pairs of size-matched EV and PMT plants at dawn. By covering plants at sundown, we ensured that only hummingbirds could visit the flowers and pollinate, and this was confirmed by visual inspection. Day-active bees were not observed at these populations. The experimental populations of N. attenuata were isolated, and hummingbirds had only limited access to other food sources than N. attenuata during the experiments. A. alexandrii is frequently found to nest in the vicinity of N. attenuata populations, even in highly isolated N. attenuata populations where other nectar sources are not available, which suggests that N. attenuata nectar is a primary energy source for these hummingbirds. Each of the two experimental populations contained approximately 50 reproductively mature plants that served as potential pollen donors. Approximately 2 weeks after emasculation, capsules were collected and stored in the laboratory.

Determination of seed paternity by microsatellite markers

A set of 16 microsatellite markers were developed to distinguish 11 accessions of N. attenuata collected from various native populations in Utah (G1–G10 are from within 200 km radius of the experimental natural populations, and G11–G20 are from more distant locations throughout the Great Basin Desert). We screened the microsatellite markers developed for linkage mapping in N. tabacum (Bindler et al., 2007) in anticipation of cross-species amplification in N. attenuata. Positive amplicons from 45 primer pairs were sequenced to confirm the repeat motifs. The Simple Sequence Repeat Identification Tool (SSRIT, http://www.gramene.org/db/markers/ssrtool) was used to identify perfect microsatellites according to the following criteria: dinucleotides with five or more repeats and trinucleotides with three or more repeats (Temnykh et al., 2001). Twenty-eight primer pairs were confirmed to amplify SSR motifs in N. attenuata, but only 16 polymorphic microsatellite markers were identified in the 11 native accessions. Twenty seeds of each EV or PMT capsule sired from antherectomized flowers and pollen transferred by hummingbirds from the surrounding natural genotypes were tested for paternity. Maternal genotypes (EV and PMT) were included to enhance the resolution of the paternity analysis. Offspring and maternal genotype data were analyzed in silico together with microsatellite allele sizes of potential paternal genotypes to deduce the paternity of the seeds with the maximum confidence.

To determine the paternity of seeds sired by pollen from the natural genotypes G1–G11, multiplex PCR reactions were performed after DNA extraction from the germinated seedlings using an Agencourt Chloropure DNA isolation kit (Beckman Coulter, https://www.beckmancoulter.com/) according to the manufacturer’s instructions. To check fidelity and allele dropouts (if any) in the multiplexed reactions. we simultaneously performed regular uniplex PCR amplifications with single primer pairs. Purified genomic DNA was amplified using a Type-It microsatellite PCR kit (Qiagen, http://www.qiagen.com/) with a set of eight primer pairs labeled with either FAM or HEX at the 5′ end of the forward primer (Table S1). The primers were grouped in three multiplex groups using MultiPLX 2.0 (http://bioinfo.ut.ee/multiplx) by analyzing PCR primer compatibility and optimal multiplexing (grouping) to avoid possible unwanted pairings between PCR samples (Table S2). The amplifications were performed using 50 ng DNA template and 0.2 μM of each primer, with an initial activation step of 5 min at 95°C (for HotStarTaq Plus DNA polymerase) followed by 30 three-step cycles of 30 sec denaturation at 95°C, 90 sec annealing at 60°C and 30 sec extension at 72°C, concluding with a final extension for 30 min at 60°C. The amplified products were purified using a QIAquick 96 PCR purification kit (Qiagen). Approximately 250 ng purified PCR product was mixed with 0.5 μl GeneScan™ 500 ROX™ size standard (Applied Biosystems, http://www.appliedbiosystems.com/) and diluted to 10 μl to enable resolution of sizes on an ABI 3100 genetic analyzer (Applied Biosystems). Allele size and genotypes were determined using GENEMAPPER® software version 3.7 (Applied Biosystems) for each multiplex group (Figure S2) according to the manufacturer’s instructions. Seed paternity was determined using COLONY software version 2.0 (Jones and Wang, 2010) by providing information on markers, potential paternal/maternal genotypes (when available) and offspring genotypes. The software has the power to deduce paternity from offspring genotypes alone; however, to provide additional confidence in the paternity determinations, microsatellite allele sizes, determined from uniplex and multiplex PCR with at least three replications, for the 11 natural accessions of N. attenuata were used to validate potential paternal genotypes.

The validity of microsatellite markers in paternity analysis was determined by genotyping of seeds from controlled mixed pollinations. Five antherectomized flowers of three replicate plants from two natural accessions of N. attenuata (Ut-WT and Az-WT) were hand-pollinated using an equal pollen mixture from the 11 natural genotypes (G1–G11) that prior experiments had determined could be readily distinguished by microsatellite genotyping. In previous experiments, all 11 genotypes were found to be as fecund as Ut-WT pollen in terms of seed siring in single pollination experiments using antherectomized Ut-WT flowers. After the pollinated flowers had produced seed capsules, 96 randomly selected seeds from one randomly chosen capsule from each maternal genotype were genotyped by microsatellite genotyping.

To test the robustness of the microsatellite marker system, we evaluated the paternity of seeds by both uniplex and multiplex amplifications in three biological replicates. The resolution of the microsatellite markers was determined to be relatively high, with the Shannon–Weaver information index (I) (Shannon and Weaver, 1949) varying from 0.997 to 1.879 (Table S2). Uniplex and multiplex analysis produced comparable results (Figure S3) for the Shannon–Weaver information index (I) for both the Ut-WT (R= 0.9433, Figure S4a) and Az-WT (R= 0.9275, Figure S4b) accessions using the same samples, validating the similar discriminatory power of the uniplex and multiplex genotyping procedures. Therefore, all further paternity analyses were performed using multiplex amplifications.

To evaluate the ability of the microsatellite markers to resolve the paternity of seeds sired in a plant growing in a native population, we germinated and extracted DNA from 96 randomly selected seeds from a single open-pollinated capsule collected from a plant growing in the middle of a large N. attenuata population (> 100 000 plants) located on a burnt area near Apex Mine, Highway 91, Utah, USA in 1992. We assumed that the maternal genotype was the same for all the samples as all seeds were derived from a single capsule. Probable paternity was deduced in silico by providing potential paternal genotype data. The seeds were found to be sired by eight natural genotypes in varying proportions (3–47%), with the most abundant (P1) likely to result from selfing (Figure S5).

The multiple paternity that we observed by microsatellite analysis of the seeds from the open-pollinated capsule from the native population is consistent with earlier inter-simple sequence repeat (ISSR) and amplified fragment length polymorphism (AFLP)-based and observational work on the species that suggested a substantial amount of outcrossing (Bahulikar et al., 2004) (Figure S5). It could not be firmly established that the 47% of seeds genotyped as P1 were sired by self pollen, but the predominant selfing nature of the species is consistent with this conclusion.

Statistical analysis

To partition the variation in nectar nicotine among single plants, among plants of a single population and among populations, nested anova was used. Flowers were therefore nested in plants and populations (flowers/plants/populations). Analysis was performed using Microsoft Excel version 2007 (Redmond, WA). In order to estimate the variance in nicotine concentrations within flowers, plants or populations, s² was estimated from a random sample that was used to estimate the variance of the population from which the sample is drawn. s² represents the sum of the squared deviations around the mean of a random sample divided by the sample size minus one.

When analyzing differences in paternity between EV and PMT plants, pseudo-replication occurred by taking multiple flowers from each plant. We accounted for this by using a split-plot anova. Here individual plants (the unit at which pseudo-replication occurred) were treated as a nesting factor in the error term of the model. Analysis was performed using R2.11.1 (R Development Core Team, 2007).

To evaluate differences in the nicotine concentration between EV and DCL lines, a one-way anova was used. To fulfil the assumption of an anova (normality and equal variances), data were log-transformed. Models were simplified by factor level reduction (Crawley, 2007), and the analysis was performed using R2.11.1 (R Development Core Team 2007).

Hummingbird choice assay data were analyzed using Student’s t test with StatView version 5.0 (SAS Institute Inc., Cary, NC).


This work was supported by the Max Planck Gesellschaft. We thank Chun-Mo Park for insightful discussions on heritable variability in nectar nicotine, M. Baldwin, J. Forbey, J. Gershenzon, D. Heckel and H. Vogel for critically reading an earlier draft, M. Kallenbach, T. A. Bozorov, and especially G. Kunert for help with the statistical analysis, Brigham Young University for use of the Lytle Ranch Preserve field station, and the US Department of Agriculture Animal and Plant Health Inspection Service (APHIS) for constructive regulatory supervision of the genetically modified organism releases.