Scent Chemistry Is Key in the Evolutionary Transition Between Insect and Mammal Pollination in African Pineapple Lilies

Research

Scent chemistry is key in the evolutionary transition between insect and mammal pollination in African pineapple lilies

Petra Wester1,2,3 , Steven D. Johnson1 and Anton Pauw2 1School of Life Sciences, University of KwaZulu-Natal Pietermaritzburg, Private Bag X01, Scottsville 3209, South Africa; 2Department of Botany and Zoology, Stellenbosch University, Private Bag X1, Stellenbosch 7602, South Africa; 3Institute of Sensory Ecology, Heinrich-Heine-University, Universit€atsstr. 1, 40225 Dusseldorf,€ Germany

Summary Author for correspondence: Volatile emissions may play a key role in structuring pollination systems of plants with mor- Petra Wester phologically unspecialised flowers. Here we test for pollination by small mammals in Eucomis Tel: +49 1729590360 regia and investigate whether its floral scent differs markedly from fly- and wasp-pollinated Email: [email protected] congeners and attracts mammals. Received: 22 August 2018 We measured floral traits of E. regia and made comparisons with insect-pollinated con- Accepted: 31 December 2018 geners. We observed floral visitors and examined fur and faeces of live-trapped mammals for pollen. We determined the contributions of different floral visitors to seed set with selective New Phytologist (2019) 222: 1624–1637 exclusion and established the breeding system with controlled pollination experiments. Using doi: 10.1111/nph.15671 bioassays, we examined whether mammals are attracted by the floral scent and are effective agents of pollen transfer. Key words: Asparagaceae, Eucomis regia, Eucomis regia differs from closely related insect-pollinated species mainly in floral scent, nectar, nonflying mammal pollination, polli- with morphology, colour and nectar properties being similar. We found that mice and ele- nator group, pollinator shift, scent, sulphur phant-shrews pollinate E. regia, which is self-incompatible and reliant on vertebrates for seed compounds. production. Mammals are strongly attracted to the overall floral scent, which contains unusual sulphur compounds, including methional (which imparts the distinctive potato-like scent and which was shown to be attractive to small mammals). The results highlight the important role of scent chemistry in shifts between insect and mammal pollination systems.

2015; Hobbhahn et al., 2017). It has been suggested that these Introduction plants rely mainly on deployment of scent cues to attract pollina- Specialised pollination systems are mostly characterised by flow- tors (Wester et al., 2009; Johnson et al., 2011). This is well sup- ers with complex architecture, allowing only a specific group of ported by studies of bat pollination systems in South America flower visitors access to rewards (Johnson & Steiner, 2000). where various aliphatic compounds and the sulphur compound However, there is increasing evidence that the chemistry of scent dimethyl disulphide (DMDS) emitted by flowers have been and nectar can mediate specialisation in pollination systems by shown to be attractive to flower-visiting bats (Knudsen & functioning as highly specific cues or filters (Johnson et al., 2006; Tollsten, 1995; von Helversen et al., 2000). There is still limited Nicolson et al., 2015; Sch€affler et al., 2015). These chemical traits evidence for a role of scent in pollination systems involving appear to be particularly important in determining the visitor ground-dwelling mammals. These pollination systems are partic- fauna in morphologically unspecialised flowers that potentially ularly well developed in Australia (involving mainly marsupials) allow a broad range of animal visitors to reach nectar. Evolution- and South Africa (involving mainly rodents and elephant-shrews) ary transitions in scent chemistry may, even in the absence of (Rourke & Wiens, 1977; Carthew & Goldingay, 1997; Letten & morphological changes, result in an evolutionary change in the Midgley, 2009; Wester et al., 2009; Wester, 2010, 2011). In a pollinator fauna with important implications for the evolution of study of the African parasitic plant Cytinus visseri (Cytinaceae) it reproductive isolation and for speciation (Peakall et al., 2010; was shown that a volatile ketone (3-hexanone) plays an important Waterman et al., 2011; Okamoto et al., 2015). role in attraction of mice to the flowers (Johnson et al., 2011), Flowers pollinated by mammals often have relatively open but no comparisons were made among related species to establish morphology owing to the lack of specialised mouthparts of these if this compound is uniquely associated with mammal animals. They are mostly pollinated at night and have drab col- pollination. oration (Rourke & Wiens, 1977; Johnson et al., 2001; Wester The African genus Eucomis L’Her. (Asparagaceae, previously et al., 2009; Wester, 2010, 2011; Zoeller et al., 2017; but see Hyacinthaceae) consists of 12 species characterised by structurally Johnson & Pauw, 2014; Johnson et al., 2011; Melidonis & Peter, unspecialised greenish flowers with easily accessible nectar

1624 New Phytologist (2019) 222: 1624–1637 Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust New Phytologist Research 1625

(Zonneveld & Duncan, 2010). Previous studies have revealed the Study sites existence of scent-mediated specialisation for insect pollination in Eucomis. Species pollinated by spider-hunting wasps (Pompili- Observations and experiments took place between 2009 and dae) differ in scent from those pollinated by carrion flies, but 2014 at 16 sites that were selected to span the natural distribution wasp- and fly-pollinated species are otherwise very similar in of the two subspecies of E. regia (Table S2; Fig. 1). terms of floral morphology, colour and nectar properties (Shut- tleworth & Johnson, 2009, 2010). Addition of sulphur com- Pollinator observations pounds (DMDS and dimethyl trisulphide (DMTS)) to flowers of the wasp-pollinated species resulted in pollination by carrion We used both direct observations and videography to observe flies (Shuttleworth & Johnson, 2010). pollinator visits. Direct observations of small groups of flowering Eucomis regia (L.) L’Her. is the only Eucomis species in the E. regia plants were made among boulders on a dolerite ridge at winter-rainfall region of South Africa (Zonneveld & Duncan, the Nieuwoudtville site (7 h, about 20 plants) and in Renoster- 2010). Flowers of this species are very similar in appearance to veld vegetation at the Overberg site (14 h, about 50 plants; see those of other Eucomis species known to be pollinated by insects. Table 1 for details). To avoid disturbing mammals, plants were However, during preliminary observations we noticed that flow- illuminated with a flashlight covered with red plastic film. Video ers of E. regia have a highly distinctive potato-like scent and are observations were made at the Windhoek farm site. Five plants not visited by insects. The peculiar scent, along with the presence growing among boulders were monitored with three video cam- of fresh mammal faeces on leaves, led us to hypothesize that an corders (HDR-XR520, HDR-XR550; Sony, Tokyo, Japan) evolutionary transition in scent chemistry has occurred in con- using infrared lights (one to three 1-A SMD LEDs emitting junction with an ecological shift to pollination by small ground- 940 nm light), using 12 V/18 A.h lead-acid batteries as a power dwelling mammals. source. Camcorders and light sources operated continuously for The aims of this study were: to identify the pollinators of 6–13 h (86 h in total) and were placed 70–100 cm from the near- E. regia; to determine the breeding system in order to establish if est plants (see Table S2 for details). the plants depend on pollinators for reproduction; to evaluate Foraging behaviour of captured mammals was observed in ter- whether selective exclusion of vertebrates results in reduced seed raria (13 animals for c. 38 h in total, see Table S2 for details). production; to determine floral morphology, colour, nectar and Each terrarium was equipped with sand, stones as shelter, food, scent properties and compare these to those of insect-pollinated water, one animal and one to two flowering E. regia plants. congeners; and to experimentally test whether the overall scent and individual compounds are attractants for small mammals. Pollen transfer simulation Experiments were conducted using living mice to test whether Materials and Methods they could effect pollen transfer between E. regia inflorescences. Two E. regia inflorescences, one of which had anthers dusted Study taxa with dye powder, were placed in terraria with one live mammal E. regia is native to the winter rainfall zone of the Cape Floral at a time (three Namaqua rock mice Micaelamys namaquensis, Region of South Africa (Northern and Western Cape; Zonneveld four Four-striped field mice Rhabdomys pumilio, one Cape rock & Duncan, 2010). It occurs as two subspecies with E. regia (L.) elephant-shrew Elephantulus edwardii; Table S2). The stigmas of Gawl. ex Reyneke ssp. regia in the north, flowering from August flowers of undyed inflorescences were subsequently inspected for until October, and E. regia ssp. pillansii (L. Guthrie) Reyneke in the presence of dye powder. the south of its distribution area, flowering from July until September (Fig. 1; Reyneke, 1972; Crouch, 2010). The inflores- Trapping of mammals and pollen loads cences of E. regia ssp. pillansii are taller than those of ssp. regia (20 vs. 10 cm), but the subspecies have very similar flower dimen- To investigate which small mammal species occur near the sions (Supporting Information Table S1). plants and whether they carry pollen of E. regia, rodent traps, E. regia plants are robust with several large leaves lying flat on baited with a mixture of peanut butter and rolled oats, were laid the ground and an inflorescence with 11–76 flowers being topped out at the study sites (22–128 traps over 17 days/nights; see by large upper bracts hiding at least some of the flowers from Table S2 for details). We waited until snouts of the captured view (Fig. 2a,b,e). The flowers are arranged in the axils of small mammals protruded through a small hole in the corner of a bracts along the thick and stable inflorescence axis (Table S1; plastic bag and then swabbed the snouts with a small block of Fig. 2a–c,e,f). The robust flowers are almost sessile with six fuchsin glycerine gelatine (Beattie, 1971) which was later melted curved and sturdy but flexible stamens that are slightly asymmet- onto a microscope slide. Animal faeces retrieved from the live rically arranged around the ovary (Fig. 2c,f). The carpels and sta- traps were crushed in liquid fuchsin glycerine gelatine and mens are about 10 mm in height (Table S1). The broadened mounted on a slide. Under the microscope, pollen grains were filaments are fused at their base, forming a bowl which contains counted for the snout and faeces (the latter over three scans of nonviscous nectar in a gap between the filaments and the ovary the length of a coverslip, representing c. 20% of the sample) at (Table S1; Fig. 2c,f). The fruits are capsules. 1009 magnification.

Ó 2019 The Authors New Phytologist (2019) 222: 1624–1637 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com New 1626 Research Phytologist

newly opened flowers of each plant (at least four flowers per treatment per plant): cross-pollination of emasculated flowers with pollen from a plant that was at least 35 m away as a positive control, self-pollination to test for self-compatibility, and no manip- ulation to test for autogamy. Stigmas were saturated with pollen applied directly from excised anthers held with forceps. After pollination, plants were carefully rebagged and later the number of fruits and seeds per flower in each treatment was counted.

Floral traits Morphology To compare the morphology of E. regia with its insect-pollinated congeners, we measured tepal dimensions of E. regia using plants at the ﬁeld sites (Table S2) and herbarium specimens (Table S3). Tepal dimensions were calculated from measurements of three representative ﬂowers per individual.

Colour Spectral reflectance (300–700 nm range) was determined for tepals, lower and upper bracts, and leaves of six E. regia plants from each of the Nieuwoudtville and Fairfield Fig. 1 Distribution of the two Eucomis regia subspecies (grey) and study sites (dots) in the Northern and Western Cape of South Africa (distribution populations. Measurements were made using an Ocean Optics area after Reyneke, 1972; Crouch, 2010). Study sites from NW to SE: (Dunedin, FL, USA) USB4000 spectrometer with PX-2 pulsed E. regia ssp. pillansii: Kamiesberg (Windhoek, Kamiesbergpass, Bovlei), xenon light source and fibre-optic reflection probe (UV/VIS Bokkeveld (Nieuwoudtville: above, Papkuilsfontein: below), Roggeveld 400 lm), the latter held at 45° c. 5 mm from the surface of the (Soekop, Quaggasfontein, Blesfontein); E. regia ssp. regia: Westcoast plant organ. A mean spectrum was calculated for each plant (Waterklip), Overberg (Goedvertrouw, Swartrivier, Drayton, Fairfield, opposite Skurwekop, Skurwekop, Vogelgezang). organ. The similarity of the flowers as perceived by insects was determined by plotting the reflectance spectra in the bee colour hexagon (Chittka, 1992) and the blowfly colour vision model Seed set after selective exclusion (Troje, 1993). The bee colour hexagon is thought to be representative of colour vision across most higher Hymenoptera (includ- To investigate whether small mammals are important for seed ing pompilid wasps that pollinate related Eucomis species; production, plants were covered at the bud stage with hexagonal Shuttleworth & Johnson, 2010). 9 wire mesh (17 21 mm) to exclude mammals but not insects (in To detect possible UV patterns that might influence bee total 6929 flowers of 188 plants, Table S2). The Pygmy mouse behaviours, photos were taken with a Lumix GH-1 camera (Mus minutoides), one of the smallest mammal species world- (Panasonic, Osaka, Japan) without low-pass filter in front of the wide, is able to get through the holes in this mesh, but is a minor sensor in combination with a UV-transmissible Ultra-Achr- component of local mammal assemblages. Fruit set in caged omatic-Takumar 1 : 4.5/85 quartz glass lens (Pentax, Tokyo, plants was compared to that of uncovered control plants and Japan) and a UV filter (transmitting at 320–380 nm) or, for stan- those that were completely covered with fine mesh (to exclude all dard photos, a UV/IR cut filter (transmitting at 400–700 nm, visitors). In addition, we recorded natural fruit and seed set for both Baader Planetarium, Mammendorf, Germany). Illumina- plants at the different study sites (Table S2). tion was achieved by means of a UV-torch (UV 365 nm; U301, MTE, Shenzhen, China). Brightness was adjusted using a grey Breeding system standard.

Controlled hand-pollination experiments were conducted to Nectar The standing crop of floral nectar was measured in the determine the breeding system and, in particular, the dependence Nieuwoudtville and Swartrivier populations (Table S2). Volume of plants on pollinator visits for seed production. Ten plants of was measured with microcapillary tubes (Brand, Wertheim, Ger- E. regia at the Fairfield field site were covered with fine mesh at many) and sugar concentration with hand-held refractometers the bud stage to exclude flower visitors. When flowers had (Eclipse 45-81 and 45-82: 0–50% and 45–80% sucrose (w/w), opened, three treatments were applied to randomly selected Bellingham and Stanley, Tunbridge Wells, UK). Nectar was

Fig. 2 Small mammals visiting flowers of Eucomis regia (a–c, ssp. pillansii; e, f, ssp. regia). (a) Micaelamys namaquensis visiting flowers of E. regia growing between rocks near Nieuwoudtville. (b) Elephantulus edwardii licking nectar with its long tongue. (c) Micaelamys namaquensis lapping nectar and becoming dusted with pollen. (d) Rhabdomys pumilio after E. regia flower visits with a thick layer of pollen around its nose and snout. (e) Flower-visiting Mus minutoides touching the anthers. (f) Rhabdomys pumilio with snout inserted in the flower, touching anthers and becoming dusted with pollen. Bars, 1 cm. Pictures taken in terraria except (a) (photographs: P. Wester).

New Phytologist (2019) 222: 1624–1637 Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust New Phytologist Research 1627

(a) (b)

(e) (f)

Ó 2019 The Authors New Phytologist (2019) 222: 1624–1637 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com New 1628 Research Phytologist

Table 1 Percentage contribution of specific scent compounds to the all substances that were not present in the background control average similarity (based on Bray–Curtis coefficient; SIMPER, similarity samples. percentages) between Eucomis inflorescence samples grouped after pollinators (compounds shown represent the first 50% of overall similarity). Choice experiments Mammal- Fly-pollinated Wasp-pollinated pollinated (E. bicolor and (E. autumnalis To determine whether the mammals would respond to the scent a a Compound (E. regia) E. humilis) and E. comosa) of E. regia in the absence of visual cues, we placed pairs of exo-Brevicomin 27.7 inverted plastic flower pots with small holes near their top and 2-Heptanone 11.0 base, through which the scent could escape, into terraria. One Methional 7.7 pot contained an E. regia inflorescence with flowers and the other 3-Methyl-3- 6.4 contained an E. regia inflorescence with flowers removed (con- buten-2-one trol). The behaviour in the terraria of eight M. namaquensis indi- Linalool 22.0 24.4 Dimethyl disulphide 8.4 viduals and one E. edwardii individual caught near E. regia plants 3,5-Dimethoxytoluene 4.7 9.0 in Nieuwoudtville was recorded. Sniffing at the holes of a pot or Dimethyl trisulphide 4.1 attempting to get in or under a port was considered a positive (E)-Linalooloxide 4.0 4.9 response. At least five sessions with each animal were carried out. Hotrienol 4.0 11.0 In each session, an animal was given one opportunity to choose Benzaldehyde 4.0 (E)-Ocimene 5.4 between the two pots. In total, we obtained 448 choices with Total 70.9 51.2 54.7 M. namaquensis and 99 choices with E. edwardii. Average similarity 41.1 57.1 60.3 To determine whether mammals would respond to methional,

a the dominant sulphur compound in headspace samples, a 6-cm- Data for the insect-pollinated species are from Shuttleworth & Johnson diameter Plexiglas Y-maze (15-cm main arm split at a 50° angle (2010). into two 20-cm arms) was used (Fig. S1). The main arm was connected to a terrarium containing one animal. The other two arms terminated in small metal chambers, connected to fans blowing À collected at different times during the day and night to encom- air (at c. 1000 ml min 1) into the Y-maze. One randomly pass the entire period of activity of the mammals. For the deter- selected chamber was equipped with a 1 : 1000 mixture of mination of free sugars, nectar was spotted onto filter paper, air methional and white mineral oil (to delay evaporation), the other dried, stored in a freezer and later examined by high-performance with pure white mineral oil (control). The choices of five liquid chromatography (HPLC) (for procedure see Steenhuisen M. namaquensis individuals were recorded with at least five runs & Johnson, 2012). per animal (Table S2). A choice was considered to be each case in which an animal entered the arm of the Y-maze and moved to Floral scent We used headspace sampling and GC-MS analyses the end of the arm. Animals were not offered rewards in the to determine the chemical composition and rates of emission of choice experiments or any prior training. floral scent. Floral scent was collected from E. regia inflorescences, and also directly from nectar (collected in 1.5 ml screw Statistical analyses neck vials), at most study sites (Table S2) at different times during the day and night. Flowers or nectar samples were enclosed in a Unless stated otherwise, data were analysed using generalized lin- polyacetate bag (Nalophan, Kalle, Wiesbaden, Germany) and air ear models implemented in SPSS 22 (IBM Corp., Armonk, NY, from the bags pumped through small cartridges containing Tenax USA). Potentially correlated data obtained from repeated mea- and Carbotrap (Supelco, Bellefonte, PA, USA) activated charcoal at sures of the same plant or animal were analysed using generalized – a flow rate of 100 ml min 1 for 3 h (nectar samples 2 h). Control estimating equations (GEEs). samples of background air were taken at each locality under the The effects of selective exclusion and breeding system on seed same conditions. The samples were thermally desorbed and anal- set per flower and fruit were analysed with GEEs with a negative ysed using a Varian CP-3800 gas chromatograph coupled with a binomial error distribution and log link function, while those on Varian 1200 quadrupole mass spectrometer equipped with a car- the proportion of ovules that set seed were analysed using a bino- bowax column (procedures as described by Shuttleworth & John- mial error distribution and logit link function. Plant was treated son, 2009). Compounds were identified using a Varian MS as the subject and incorporated an exchangeable correlation Workstation (Palo Alto, CA, USA) with the NIST11 mass spectral matrix. Significance was assessed using score statistics. Post-hoc library (Gaithersburg, MD, USA) and comparisons with retention comparisons among means were conducted using the Dunn–S times of authentic standards, where available, as well as comparisons idak procedure. Means and standard errors were back- between calculated and published Kovats retention indices (using transformed from the log scale for analyses that incorporated a the NIST11 retention index database). All reference compounds negative binomial error structure and from the logit scale for were obtained from Sigma Aldrich (Munich, Germany) except exo- analyses that incorporated a binomial error structure. Brevicomin (Contech, Delta, BC, Canada). The relative amount of To assess the preference of animals in the choice experiments, each substance was given as a percentage of the total ion count for the proportion of responses in favour of the arm with the test

New Phytologist (2019) 222: 1624–1637 Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust New Phytologist Research 1629 compound was analysed using GEEs with a binomial error distri- dawn and dusk). Pollen was deposited mostly at the tip or distal bution and logit link function (Table S2). Animal individual was half of the nose. Often the nose was already covered with pollen treated as the subject. Means with confidence intervals that did before flower visits, but during almost all observations the not overlap a proportion of 0.5 (equal choice) were considered to amount of pollen on the nose visibly increased. represent a significant preference. When only one animal was In the terrarium, five M. namaquensis individuals from the Nieu- involved in the experiment, a binomial test was used. woudtville site went without hesitation to the flowers of several To compare nectar volume and concentration among the two inflorescences (> 350 flower visits in total), showing the same subspecies and among species with different pollinators (data behaviour observed in the field. This behaviour was also evident for from Shuttleworth & Johnson, 2010), we used Mann–Whitney an individual of E. edwardii that visited more than 150 flowers dur- U-tests and Kruskal–Wallis tests (implemented in SPSS 22), ing the day and night (Fig. 2b). Because of the long and slender respectively, as these data did not fit a normal distribution. snout of E. edwardii, pollen was deposited mostly at the tip of the To determine whether variation in morphometrics and scent is nose (Video S5). Sometimes the animal only sniffed at specific flow- related to pollinator group, taxon or region, phenotype data were ers (or inflorescences) without visiting (probably older ones with lit- square-root transformed and plotted in two dimensions using non- tle nectar), but always visited fresh inflorescences presented metric multi-dimensional scaling (NMDS) based on Bray–Curtis afterwards. In the terrarium, five R. pumilio (Fig. 2d,f; Videos S6, similarity. Data for species not included in this study were obtained S7) and one M. minutoides (Fig. 2e) visited E. regia ssp. regia flowers from Shuttleworth & Johnson (2010). The resulting similarity (Swartrivier and Fairfield sites) in the same way described above matrices were the basis for analysis of similarities (ANOSIM; with (> 380 flower visits for R. pumilio including juveniles born in captiv- up to 10 000 permutations) in which the test statistic R compares ity, > 100 for M. minutoides). The southern African vlei rat (Otomys average rank similarities within and among groups. R values close to irroratus cf.) never visited E. regia flowers in the terrarium. unity indicate complete separation of groups while R values close to zero indicate minimal separation among groups. Scent compounds Pollen transfer simulation characterising the fragrance of each species, pollinator group or Dye powder was transferred among inflorescences by all mam- region were explored using SIMPER (similarity percentages). mals tested. In total, individuals of M. namaquensis transferred NMDS, ANOSIM and SIMPER were conducted using PRIMER pollen to stigmas of 54 flowers of nine inflorescences, E. edwardii 6.1.6 (Primer-E Ltd, Plymouth, UK). to six flowers of five inflorescences and R. pumilio to 50 flowers of seven inflorescences. Results

Floral visitors Trapped mammals and pollen loads We recorded small mammals as flower visitors of E. regia at three At the study sites, 135 small mammals were caught (Table S4). of the study sites (Table S2; Fig. 2). Apart from a few ants and E. regia pollen was abundant around the snouts in all examined sam- mites, no other insects or birds were observed on the flowers, ples of R. pumilio and M. minutoides and in almost all of the faeces despite these being common visitors to flowers of other species at samples from these species (Table S5). E. regia pollen represented the study sites. almost 100% of the overall pollen found (except 80% in the Direct observations were made of M. namaquensis visiting M. minutoides snout samples). The snouts of three R. pumilio indi- E. regia ssp. pillansii flowers in the field at the Nieuwoudtville site viduals, examined immediately after trapping, had conspicuous yel- (Figs 2a, S2). More than 15 flower visits were observed from low masses of E. regia pollen. Faeces and snout samples of the 18:15 h and 18:45 h shortly after dusk over a period of 3 d. The herbivorous rodent O. irroratus cf. contained no E. regia pollen. mice visited several flowers (up to 20 flowers per inflorescence) and inflorescences one after another, sometimes returning to the Seed set after selective exclusion same inflorescence after a few minutes. The mice spent several sec- onds at each flower, pushing their snout between the stamens and Exclusion of most small mammals from E. regia inflorescences licking nectar with their tongue (Figs 2a,c, S2; Video S1), thereby resulted in a significant decline in seed set, fruit set and seeds per touching the anthers and stigma, clearly becoming dusted with fruit in comparison with control plants at all three localities pollen on the nose and around the snout (Fig. 2c; Video S1). The (Fig. 3). Almost no seed set occurred in flowers from which all mice only lapped nectar, and did not feed on pollen directly or eat visitors were excluded. Natural levels of fecundity were high in or damage flowers. The mice sometimes groomed their fur after most populations (Fig. S3; Table S6). flower visits. Video observations of small mammals visiting E. regia ssp. Breeding system pillansii flowers were made at the farm Windhoek. E. edwardii were observed to visit more than 100 flowers of four different Cross-pollinated flowers of E. regia showed significantly higher plants (1–18 flowers per plant: mean 6.5 Æ 4.5 SE; 1–3 plants fruit and seed set compared with self-pollinated or control flowers consecutively: mean 1.3 Æ 0.7 SE; spending up to 5 s per flower, with a mean of seven seeds per fruit in cross-pollinated flowers vs mostly 2–3 s, Videos S2–S4) over two days and nights (including one in self-pollinated or control flowers (Fig. 4).

Ó 2019 The Authors New Phytologist (2019) 222: 1624–1637 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com New 1630 Research Phytologist

Fig. 3 Effects of selective pollinator exclusion of Eucomis regia at Fairfield, Swartrivier and Drayton: (a) seed set, (b) fruit set (percentage of flowers that set seed) and (c) seeds per fruit (back-transformed means Æ SE) after open pollination, cage exclosure and fine mesh exclosure (nopen pollination = Fairfield 42 (inflorescences)/1666 (flowers), Swartrivier 36/1154, Drayton 25/979; ncage exclosure = Fairfield 11/399, Swartrivier 10/336, Drayton 10/376; nfine mesh 2 exclosure = Fairfield 35/1364, Swartrivier 19/656 plants/flowers). Means that share letters are not significantly different. Seed set: Fairfield: v = 39.2, P = 0.000, Swartrivier: v2 = 42.7, P = 0.000, Drayton: v2 = 9.1, P = 0.003; fruit set: Fairfield: v2 = 54.2, P = 0.000, Swartrivier: v2 = 44.0, P = 0.000, Drayton: v2 = 8.1, P = 0.005; seeds per fruit: Fairfield: v2 = 23.8, P = 0.000, Swartrivier: v2 = 17.9, P = 0.000, Drayton: v2 = 5.7, P = 0.017.

especially according to pollinator groups (global R = 0.844, Floral characters P = 0.0001) with the greatest separation being between E. regia Morphology The floral architecture of E. regia is almost identi- and insect-pollinated species (Table S7; Fig. 5b). Scent composi- cal to that of its insect-pollinated congeners. Based on tepal tion was similar for the fly- and wasp-pollinated Eucomis species, dimensions, the different Eucomis species show only weak separa- with the main difference being the emission of the sulphur- tion in the NMDS (ANOSIM: global R = 0.245, P = 0.0001) containing compounds DMDS and DMTS in the fly-pollinated and even less according to pollinator groups (global R = 0.135, species. These compounds are also present at trace amounts in P = 0.0001; Fig. 5a; Table S7). The two subspecies of E. regia the scent of E. regia. The main compounds that characterise the show no significant difference concerning tepal dimensions floral scent of E. regia are exo-brevicomin, 2-heptanone, (ANOSIM: R = 0.032, P = 0.196) and there was also no separa- methional (found as a trace compound in the fly-pollinated tion according to geographical region (R = 0.059, P = 0.243; species) and 3-methyl-3-buten-2-one (Tables 1, S10). Fig. 5a; Table S8). The sulphur compounds methional and methionol are responsible for the boiled-potato-like smell of the flowers and of sepa- Floral scent When analysed according to chemical composition rated nectar (Table S8). Separated nectar also had a boiled- of scent (49 identified compounds and 64 unidentified com- potato-like taste. In plants with no or little nectar, no or only pounds; Table S9), Eucomis species separate into distinct clusters weak scent was detectable. About 50% of the scent compounds in the NMDS (ANOSIM: global R = 0.737, P = 0.0001, Fig. 5b) were also detected in nectar and additionally 18 compounds that

New Phytologist (2019) 222: 1624–1637 Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust New Phytologist Research 1631

(a)

(b)

Fig. 5 (a) Tepal dimensions (stress: 0.05) and (b) scent profiles (stress: 0.07) of the inflorescences of mammal-, fly- and wasp-pollinated Eucomis species plotted in two dimensions based on nonmetric multi-dimensional scaling (NMDS). E. autumnalis: closed triangle, E. comosa: open triangles, E. bicolor: closed circles, E. humilis: open circles, E. regia ssp. regia: closed squares (without symbol, Overberg; with cross, Westcoast; with dot, Heidelberg; with x, Robertson region), E. regia ssp. pillansii: open squares (without symbol, Roggeveld; with cross, Kamiesberg; with dot, Bokkeveld); photos are in the same sequence clockwise. Data from the insect-pollinated species are from Shuttleworth & Johnson (2010).

Fig. 4 (a) Seed set, (b) fruit set (percentage of flowers that set seed) and Geographical structure was evident in the multivariate analysis (c) seeds per fruit (back-transformed means Æ SE) of Eucomis regia plants after controlled pollination experiments (cross-pollination, self-pollination of the scent of E. regia. There was weak separation according to = = and unmanipulated controls) at Fairfield (ncross = 40, nself = 45, ncontrol = 50 subspecies in the NMDS (ANOSIM: R 0.273, P 0.004), and flowers, n = 10 plants). Means that share letters are not significantly strong separation by region (global R = 0.904, P = 0.0001) due to different. The values were significantly different between all three plants from the Bokkeveld region lacking mainly exo-brevicomin v2 = = v2 = = treatments. Seed set: 8.5, P 0.014; fruit set: 9.3, P 0.010; (and 2-heptanone) (Fig. 5b; Tables 2, S7, S11). seeds per fruit: v2 = 6.4, P = 0.042. occur in nectar only. Some compounds were also found in vege- Colour Spectral reflectance of the inconspicuous E. regia flowers was tative inflorescence parts or leaves (e.g. exo-brevicomin and endo- very similar to those of the bracts and leaves (Fig. S4) as well as the brevicomin, 2-heptanone, 2-heptanol, unidentified compound insect-pollinated congeners (Fig. 6). Colour loci showed overlap in (m/z: 70, 43, 86, 42, 41, 73, 54, 101, 39)) or were not found in both the colour hexagon and the blowfly colour vision models the scent of nectar (e.g. 2-heptanol, 2-heptanone, 2-nonanone, (Fig.S5).TheflowerslackedanyconspicuousUVpattern(Fig.S6). (E)-oct-3-en-2-one, unidentified compound (m/z: 58, 43, 41, 82, 71, 55); Table S8), indicating that they might derive from vegeta- Nectar The volume and concentration of nectar standing crop tive inflorescence parts. in E. regia flowers was similar across populations (concentration:

Ó 2019 The Authors New Phytologist (2019) 222: 1624–1637 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com New 1632 Research Phytologist

Table 2 Mean percentage of volatile compounds (selection) emitted by inflorescences of Eucomis regia of a specific region (identified by GC-MS from headspace samples).

Overberg Roggeveld Kamiesberg Bokkeveld Compound (ssp. regia) (ssp. pillansii) (ssp. pillansii) (ssp. pillansii)

exo-Brevicomin 65.5 47.5 22.5 0 3-Methyl-3-buten-2-one 5.3 0 0 13.0 Unidentiﬁed sulphur compound 4.4 0 0 15.4 (m/z: 43, 118, 71, 75, 61, 47) 2-Heptanone 3.8 26.9 42.9 0 Methional 3.2 2.0 0.3 26.5 2-Methylthio-2,3-dimethylbutane 2.8 0.1 0 17 endo-Brevicomin 2 0.9 0.4 0 Methionol 1.8 0.8 0.3 2.8 2-Heptanol 0 2.9 17.9 0

Discussion

Small mammal pollination Our hypothesis that E. regia is pollinated by small mammals is con- firmed by observations and experimental evidence. The animals transfer pollen between flowers and had large E. regia pollen loads on their snouts and in their faeces (Table S5), the latter indicating ingestion after grooming as the animals do not appear to feed directly on pollen (see also Goldingay et al., 1987; Johnson & Pauw, 2014). Exclusion of small mammals resulted in a very strong decline in seed production (Fig. 3), indicating a dependence on these animals for pollination. It is remarkable that even isolated plantsshowedsomeseedset,indicating that small mammals might be reliable pollinators based on the fact that they are ubiquitous local residents in these populations (see also Kleizen et al., 2009). Fig. 6 Reflectance spectra of Eucomis regia (populations of Nieuwoudtville: ssp. pillansii and Fairfield: ssp. regia) tepals in comparison E. regia is one of several geoflorous small-mammal-pollinated with the wasp-pollinated E. autumnalis and E. comosa and the fly- monocots of the western part of South Africa. This group pollinated E. bicolor and E. humilis (data of the insect-pollinated species includes species of the tribe Hyacintheae (c. 500 species; of the are from Shuttleworth & Johnson, 2009, 2010). Asparagaceae subfamily Scilloideae; alternatively regarded as Hyacinthaceae subfam. Hyacinthoideae), several species of which have been shown to be visited for nectar by mice and elephant- U = 973, P > 0.1; volume: U = 1016, P > 0.2; Table S12). shrews (Massonia depressa Houtt., Massonia echinata L.f., Compared to the insect-pollinated congeners, there was no clear Whiteheadia bifolia (Jacq.) Baker; Johnson et al., 2001; Wester association of these nectar properties regarding pollinator groups et al., 2009; Wester, 2010, 2015; Flasch et al., 2016). The flower (concentration: F = 0.697, P = 0.07; volume: F = 0.218, P = 0.07; and inflorescence structure of E. regia is most similar to that of Table S12). The nectar of all species was dominated by hexose W. bifolia, but also similar to all other small-mammal-pollinated sugars (Table S12). species of the Hyacintheae and of the genus Colchicum (formerly Androcymbium; Colchicaceae). They all have flowers with easily accessible nectar and that are visually inconspicuous, lacking col- Choice experiments oration that plants usually use to attract insects or birds. The Mice and elephant-shrews strongly preferred the pots with species also differ from their insect- or bird-pollinated congeners hidden flowers over the empty pots (M. namaquensis: 82% of in their scent (Kleizen et al., 2008; Wester et al., 2009. choices in favour of pots with hidden flowers; v2 = 33.4, The mice species observed here are well known to feed on nec- df = 1, P < 0.001, n = 8 individuals that made 448 choices; tar of other geoflorous plant species (Wiens & Rourke, 1978; E. edwardii: 91%; P < 0.001; n = 1 individual that made 99 Kleizen et al., 2008; Biccard & Midgley, 2009; Wester et al., choices). In the Y-maze experiment, mice strongly preferred 2009; Melidonis & Peter, 2015). By contrast, the nectarivory of the arm with methional (87% of choices in favour of elephant-shrews has only been shown recently (Wester, 2010, methional; v2 = 11.1, df = 1, P < 0.001, n = 5 M. namaquensis 2011, 2015; Johnson et al., 2011; Flasch et al., 2016; Kuhn€ et al., that made 52 choices). 2017). Elephant-shrews (Macroscelidea) are not related to mice

New Phytologist (2019) 222: 1624–1637 Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust New Phytologist Research 1633

(Rodentia) and have long been interpreted as purely insectivorous unpublished). We do not know whether the absence of insects as (Perrin, 1997). floral visitors to E. regia is due to missing attractants or to repel- In addition to nocturnal visits by rodents, the flowers received lence by methional or other compounds. some visits during daylight hours by the rodent R. pumilio and by Methional is released during the cooking of potatoes (Peter- elephant-shrews. This phenomenon is also found in some other son et al., 1998) and occurs in many other materials such as mammal-pollinated plants (Johnson & Pauw, 2014; Johnson processed beef (Kerscher & Grosch, 1997). By contrast, et al., 2011; Melidonis & Peter, 2015; Hobbhahn et al., 2017). DMDS and DMTS are reminiscent of rotten meat, from However, the animals that visit during daylight are predomi- which DMDS is emitted at an early stage of the decay process nantly nocturnal or crepuscular with activity concentrated (Jurgens€ et al., 2013). Methional, methionol, DMDS and around dusk and dawn (Perrin, 1981; Schumann et al., 2005; DMTS are derivatives of the amino acid methionine, although Skinner & Chimimba, 2005) when scent seems to be more their biogenesis is not well understood (Landaud et al., 2008; important than visual attraction. Most rodents and probably ele- Liu et al., 2013). It is highly likely that the Eucomis species phant-shrews have a well-developed sense of smell (Stoddart, differ in the derivatisation pathway(s) of methionine, thus 1980; Skinner & Chimimba, 2005) and are thus expected to rely enabling pollinator shifts. An analogous scenario has been doc- mainly on olfactory cues. umented in Mitella (Saxifragaceae) in which the production of lilac aldehydes as metabolites of linalool functions to attract long-tongued fungus-gnats, but not short-tongued fungus-gnats Importance of floral scent (Okamoto et al., 2015). E. regia differs strongly from its insect-pollinated congeners in Scent composition varies geographically in E. regia (Fig. 5b, floral scent, but not in floral morphology or colour (Figs 5, 6, S6; Table S11), but this variation is not strongly partitioned between Tables 1, 2, S6, S9). In addition, nectar properties show no clear the two E. regia subspecies. Whereas methional and methionol association with any pollinator group and vary strongly among were found in all examined samples, other main compounds (e.g. insect-pollinated Eucomis species (Table S12). Nectar properties exo- and endo-brevicomin, 3-methyl-3-buten-2-one, 2-heptan- also vary greatly among different plant species adapted for polli- one, 2-heptanol, 2-methylthio-2,3-dimethylbutane, 2-methyl-1- nation by nonflying mammals (Wester et al., 2009; Johnson & hepten-6-one and an unidentified compound (m/z: 70, 43, 86, Pauw, 2014). Analysis of spectra using insect vision models 42, 41, 73, 54, 101, 39)) appeared to be absent in some popula- (Fig. S5) suggest that flowers of Eucomis are visually inconspicu- tions, including in some where small mammals were attracted by ous to many insects and thus probably rely mainly on scent to the flowers in the field and in the lab. Thus, it is unlikely that attract pollinators. The main distinguishing features of the scent these compounds are key attractants to small mammals. At least of E. regia are the emission of exo-brevicomin, 2-heptanone, some of these compounds (exo-brevicomin, 2-heptanone, 2- methional and 3-methyl-3-buten-2-one (Tables 1, S10). Our heptanol) were also found in vegetative parts of the plant that are experiments show that small mammals are able to locate E. regia also present in pre- and post-floral stages. Exo-brevicomin is inflorescences using scent cues alone and that methional, a sul- quite rare in flowers and is most widely known as a male attrac- phur compound that imparts the characteristic boiled-potato-like tant/aggregation pheromone component produced by female scent to the human nose (Belitz et al., 2009), is attractive to small bark beetles (Francke et al., 1995). 2-Heptanone is a common mammals. In addition to M. namaquensis (this study), our component of mammalian urine, occurring as an alarm unpublished data show that methional is attractive to other pheromone in mice urine and as a component of human urine mammals that pollinate E. regia (E. edwardii, R. pumilio), the (Novotny, 2003; Burger, 2005; Fortes-Marco et al., 2015). It also Round-eared elephant-shrews (Macroscelides proboscideus) that occurs in the floral scent of the related species, E. humilis, and visit flowers of other plants (Arena, 2018), as well as some non- plants pollinated by carrion flies as well as in carrion (Shuttle- pollinator mammals such as domesticated mice (Mus musculus worth & Johnson, 2010; van der Niet et al., 2011). domesticus) and Banded mongoose (Mungos mungo, Herpestidae, The volatile methional was present in all populations and Carnivora). Sulphur compounds have been implicated as attrac- seems to play a key role in attracting mammals. Although the tants in bat-pollination systems, but the evolutionary reason for amount of methional in E. regia plants differed among regions, this is not known (Bestmann et al., 1997; von Helversen et al., to the human nose, the typical boiled-potato smell (probably 2000; Pettersson et al., 2004). The sulphur compounds DMDS mainly due to methional) was very dominant in all plants. and DMTS are known to attract some bats in South America as Like humans and other mammals such as bats (von Helversen well as to attract carrion flies (von Helversen et al., 2000; Freder- et al., 2000), rodents are ultrasensitive to volatile sulphur com- ickx et al., 2011). These compounds also play a role in attraction pounds as they can indicate food sources (and whether food is of carrion flies to flowers of some Eucomis species (Shuttleworth rancid), conspecifics and the presence of predators (carnivores & Johnson, 2010), but are present only in trace amounts in the that have left scent marks) (Duan et al., 2012; Sarrafchi et al., floral scent of E. regia. Interestingly, DMDS was repellent to 2013). Humans can perceive methional at concentrations of as À small mammals (R. pumilio, M. namaquensis and M. proboscideus) little as 0.12 ng l 1 air (Belitz et al., 2009). Even if present in preliminary tests (P. Wester & S. Johnson, unpublished). only in trace amounts, sulphur compounds change the overall Linalool, the main compound of all insect-pollinated Eucomis olfactory impressions of fragrant mixtures (Goeke, 2007). congeners examined, is repellent to M. proboscideus (P. Wester, Thus, other, nonsulphur, main compounds such as exo-

Ó 2019 The Authors New Phytologist (2019) 222: 1624–1637 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com New 1634 Research Phytologist

brevicomin and 2-heptanone are probably overlaid by the sul- performing a pollination service. By contrast, scent can specifi- phur compounds and are of less importance for pollinator cally attract effective pollinators (see also Raguso, 2008; Shut- attraction. For example, compared to methional, exo- tleworth & Johnson, 2010). In this respect, the production of brevicomin, the scent compound with the highest percentage different sulphur compounds (e.g. DMDS and methional) in contribution responsible for differences to insect-pollinated Eucomis seems to have played a pivotal role in shifts between Eucomis congeners (Table 1), is hardly detected by the human wasp, carrion fly and mammal pollination (Shuttleworth & nose. Overall, the evidence points to the gain (or loss) of Johnson, 2010; this study). methional synthesis as a key step in the shift between insect and mammal pollination in the genus Eucomis. Acknowledgements In general, the large number of compounds we found in E. regia floral scent and nectar is consistent with findings for one We thank Cameron and Rhoda McMaster, Stuart Hall, Lita of the other species in the genus, E. humilis, for which more than Cole, Lizande Kellerman, Helga and Jac van der Merwe, Nicol 80 compounds were reported by Shuttleworth & Johnson (2010) and Marina van der Merwe and Sjirk Geerts for locality informa- and also for other nonflying mammal-pollinated plant species (P. tion, Marina, Michael, Liza and Peter Swart for generous hospi- Wester, unpublished). It is possible that some compounds are of tality, Ted Oliver for help with E. regia cultivation in his garden, microbial origin. For example, b-cyclocitral is a cyanobacterial Stella and Ernest Schulze for logistical support, the Swart family, metabolite (Zhang et al., 2013) and 2,3-dihydro-3,5-dihydroxy- Valerian van der Byl, Corneels Genis and the municipality of 6-methyl-4H-pyran-4-one is an antioxidant produced by some Nieuwoudtville for permission to work on their property, Adam Lactobacillus bacteria (Beppu et al., 2012; Hwang et al., 2013). Shuttleworth for providing raw data, Andreas Jurgens€ for scent However, both these compounds are also present in insect- analysis advice, Roman Kaiser for providing mass spectra, Sandy- pollinated Eucomis species (Shuttleworth & Johnson, 2009, Lynn Steenhuisen for help with HPLC, Hanneline Smit- 2010), which suggests that they have a floral origin. Robinson for elephant-shrew identification advice, Christian Ver- Volatiles present in the nectar may act as preservatives or have hoeven (Dusseldorf)€ for taking the UV pictures, Rob Raguso for other functions such as deterring other flower visitors (see also valuable comments on an earlier version of the paper, NBG and ^ Omura et al., 2000; Raguso, 2004; Johnson et al., 2006; Nicolson PRE for specimen loans, the NRF for support and the Claude & Thornburg, 2007). In general, nectar odour seems to be a com- Leon Foundation and UKZN, Pietermaritzburg for research fel- mon phenomenon (Raguso, 2004) and may serve as an honest sig- lowships of the first author, the Northern Cape Department of nal for food. It is unknown whether the floral scent elicits innate Tourism, Environment and Conservation (FAUNA 485/2009, attraction in the mammals (sensory exploitation/pre-existing bias) 1210/2014, FLORA 069&070/2009, 127&128/2011, 084&85/ and/or whether the preference is based on associative learning 2013) and Cape Nature (AAA005-00004-0035, 0028-AAA005- (Schiestl & D€otterl, 2012; Schiestl & Johnson, 2013). Some com- 00145, 0056-AAA007-00014, 0028-AAA007-00010) for the ponents of the floral scent of E. regia, such as the ketones, are simi- necessary permits, and the animal ethics committees of the lar to those emitted by seeds (Paulsen et al., 2013), suggesting that University of Stellenbosch (2009B01004) and UKZN (040/11/ granivorous rodents may associate volatiles in the floral scent with Animal, 020/12/Animal, 080/14/Animal) for ethical approval. sources of food. The positive responses of animals that are na€ıve This paper is dedicated to the memory of Cameron McMaster to the floral scent of E. regia (M. namaquensis, including the non- (1937–2018), devoted to South African bulbs and always helpful flower-visiting species occurring outside the range of E. regia, i.e. with sharing knowledge. R. dilectus and Mastomys natalensis; P. Wester & S. Johnson, unpublished) or to methional (R. pumilio, M. proboscideus, and the nonflower visiting species M. musculus domesticus and Author contributions M. mungo; P. Wester, unpublished) support the idea that the scent PW, SDJ and AP conceived the idea; PW designed the project elicits innate behaviour. and performed the study, analysed the data and wrote the manuscript; SDJ enabled scent analyses; and SDJ and AP con- Conclusion tributed to further manuscript revision. Evolutionary shifts in plants from one pollinator group to another are normally accompanied by coordinated changes of ORCID multiple traits such as flower shape, colour and scent, at least Steven D. Johnson https://orcid.org/0000-0002-5114-5862 in food-rewarding plants (Dell’Olivo & Kuhlemeier, 2013). Petra Wester https://orcid.org/0000-0002-6514-8159 By contrast, shifts mediated solely by floral scent seem to be rare (Shuttleworth & Johnson, 2010; Okamoto et al., 2015; this study) and are associated with flowers that are visually inconspicuous and morphologically generalised. Visual incon- References spicuousness (crypsis) in Eucomis may function to limit attrac- Arena G. 2018. Kraalaalwyn: an oasis in the drought. Veld & Flora 104:26–29. tion of nonpollinating animals, such as certain insects or birds Beattie AJ. 1971. A technique for the study of insect-borne pollen. Pan-Pacific that remove pollen, or the freely accessible nectar, without Entomologist 47: 82.

New Phytologist (2019) 222: 1624–1637 Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust New Phytologist Research 1635

Belitz H-D, Grosch W, Schieberle P. 2009. Food chemistry. Berlin, Germany: Johnson SD, Pauw A, Midgley J. 2001. Rodent pollination in the African lily Springer. Massonia depressa (Hyacinthaceae). American Journal of Botany 88: 1768– Beppu Y, Komura H, Izumo T, Horii Y, Shen J, Tanida M, Nakashima 1773. T, Tsuruoka N, Nagai K. 2012. Identification of 2,3-dihydro-3,5- Johnson SD, Steiner KE. 2000. Generalization versus specialization in dihydroxy-6-methyl-4H-Pyran-4-one isolated from Lactobacillus pentosus plant pollination systems. Trends in Ecology and Evolution 15:140– strain S-PT84 culture supernatants as a compound that stimulates 143. autonomic nerve activities in rats. Journal of Agricultural and Food Jurgens€ A, Suk-Ling W, Shuttleworth A, Johnson SD. 2013. Chemical mimicry Chemistry 60: 11044–11049. of insect oviposition sites: a global analysis of convergence in angiosperms. Bestmann HJ, Winkler L, von Helversen O. 1997. Headspace analysis of volatile Ecology Letters 16: 1157–1167. flower scent constituents of bat-pollinated plants. Phytochemistry 46: 1169– Kerscher R, Grosch W. 1997. Comparative evaluation of potent odorants of 1172. boiled beef by aroma extract dilution and concentration analysis. Zeitschrift fur€ Biccard A, Midgley JJ. 2009. Rodent pollination in Protea nana. South African Lebensmittel-Untersuchung und -Forschung 204:3–6. Journal of Botany 75: 720–725. Kleizen C, Midgley JJ, Johnson SD. 2008. Pollination systems of Colchicum Burger BV. 2005. Mammalian semiochemicals. Topics in Current Chemistry 240: (Colchicaceae) in southern Africa: evidence for rodent pollination. Annals of 231–278. Botany 102: 747–755. Carthew SM, Goldingay RL. 1997. Non-flying mammals as pollinators. Trends Kleizen C, Midgley J, Johnson SD. 2009. Variation in seed set amongst in Ecology and Evolution 12: 104–108. populations of a rodent pollinated geophyte, Colchicum coloratum. South Chittka L. 1992. The colour hexagon: a chromaticity diagram based on African Journal of Botany 75: 739–743. photoreceptor excitations as a generalized representation of colour opponency. Knudsen JT, Tollsten L. 1995. Floral scent in bat-pollinated plants: a case of Journal of Comparative Physiology. A, Neuroethology, Sensory, Neural, and convergent evolution. Botanical Journal of the Linnean Society 119:45–57. Behavioral Physiology 170: 533–543. Kuhn€ N, Midgley J, Steenhuisen S-L. 2017. Reproductive biology of three co- Crouch N. 2010. An adaptation of Reyneke’s key to the genus Eucomis. PlantLife occurring, primarily small-mammal pollinated Protea species (Proteaceae). 39–40:45–52. South African Journal of Botany 113: 337–345. Dell’Olivo A, Kuhlemeier C. 2013. Asymmetric effects of loss and gain of a floral Landaud S, Helinck S, Bonnarme P. 2008. Formation of volatile sulfur compounds and trait on pollinator preference. Evolution 67: 2023–3031. metabolism of methionine and other sulfur compounds in fermented food. Applied Duan X, Block E, Li Z, Connelly T, Zhang J, Huang Z, Su X, Pan Y, Wu L, Microbiology and Biotechnology 77: 1191–1205. Chi Q et al. 2012. Crucial role of copper in detection of metal-coordinating Letten AD, Midgley JJ. 2009. Rodent pollination in the Cape legume Liparia odorants. Proceedings of the National Academy of Sciences, USA 109: 3492– parva. Austral Ecology 34: 233–236. 3497. Liu R-S, Zhou H, Li H-M, Yuan Z-P, Chen T, Tang Y-J. 2013. Metabolism of Flasch L, von Elm N, Wester P. 2016. Nectar-drinking Elephantulus edwardii as L-methionine linked to the biosynthesis of volatile organic sulfur-containing a potential pollinator of Massonia echinata, endemic to the Bokkeveld plateau compounds during the submerged fermentation of Tuber melanosporum. in South Africa. African Journal of Ecology 55: 376–379. Applied Microbiology and Biotechnology 97: 9981–9992. Fortes-Marco L, Lanuza E, Martınez-Garcıa F, Augustın-Pavon C. 2015. Melidonis CA, Peter CI. 2015. Diurnal pollination, primarily by a single species Avoidance and contextual learning induced by a kairomone, a pheromone of rodent, documented in Protea foliosa using modified camera traps. South and a common odorant in female CD1 mice. Frontiers in Neuroscience 9:1– African Journal of Botany 97:9–15. 13. Nicolson SW, Lerch-Henning S, Welsford M, Johnson SD. 2015. Nectar Francke W, Bartels J, Meyer H, Schr€oder F, Kohnle U, Baader E, Vite JP. 1995. palatability can selectively filter bird and insect visitors to coral tree flowers. Semiochemicals from bark beetles: new results, remarks, and reflections. Evolutionary Ecology 29: 405–417. Journal of Chemical Ecology 21: 1043–1063. Nicolson SW, Thornburg RW. 2007. Nectar chemistry. In: Nicolson SW, Nepi M, Frederickx C, Dekeirsschieter J, Verheggen FJ, Haubruge E. 2011. Responses of Pacini E, eds. Nectaries and nectar. Dordrecht, the Netherlands: Springer, 215–263. Lucilia seicata Meigen (Diptera: Calliphoridae) to cadaveric volatile organic van der Niet T, Hansen DM, Johnson SD. 2011. Carrion mimicry in a South compounds. Journal of Forensic Sciences 57: 386–390. African orchid: flowers attract a narrow subset of the fly assemblage on animal Goeke A. 2007. Sulfur-containing odorants in fragrance chemistry. Sulfur Reports carcasses. Annals of Botany 107: 981–992. 23: 243–278. Novotny MV. 2003. Pheromones, binding proteins and receptor responses in Goldingay RL, Carthew SM, Whelan RJ. 1987. Transfer of Banksia spinulosa rodents. Biochemical Society Transactions 31: 117–122. pollen by mammals – Implications for pollination. Australian Journal of Zoology Okamoto T, Okuyama Y, Goto R, Tokoro M, Kato M. 2015. Parallel chemical 35: 319–325. switches underlying pollinator isolation in Asian Mitella. Journal of von Helversen O, Winkler L, Bestmann HJ. 2000. Sulphur-containing Evolutionary Biology 28: 590–600. ^ ‘perfumes’ attract flower-visiting bats. Journal of Comparative Physiology. Omura H, Honda K, Hayashi N. 2000. Floral scent of Osmanthus fragrans A, Neuroethology, Sensory, Neural, and Behavioral Physiology 186:143– discourages foraging behavior of cabbage butterfly, Pieris rapae. Journal of 153. Chemical Ecology 26: 655–666. Hobbhahn N, Steenhuisen S-L, Olsen T, Midgley JJ, Johnson SD. 2017. Paulsen TR, Colville L, Kranner I, Daws MI, H€ogstedt G, Vandvik V, Pollination and breeding system of the enigmatic South African parasitic plant Thompson K. 2013. Physical dormancy in seeds: a game of hide and seek? Mystropetalon thomii (Mystropetalaceae): rodents welcome, but not needed. New Phytologist 198: 496–503. Plant Biology 19: 775–786. Peakall R, Ebert D, Poldy J, Barrow RA, Francke W, Bower CC, Schiestl FP. Hwang G, Kim HY, Woo KS, Lee SH, Lee J, Jeong HS. 2013. Isolation and 2010. Pollinator specificity, floral odour chemistry and the phylogeny of identification of the antioxidant DDMP from heated pear (Pyrus pyrifolia Australian sexually deceptive Chiloglottis orchids: implifications for pollinator- Nakai). Preventive Nutrition and Food Science 18:76–79. driven speciation. New Phytologist 188: 437–450. Johnson CM, Pauw A. 2014. Adaptation for rodent pollination in Leucospermum Perrin MR. 1981. Notes on the activity patterns of 12 mice species of southern arenarium (Proteaceae) despite rapid pollen loss during grooming. Annals of African rodents and a new design of activity monitor. South African Journal of Botany 113: 931–938. Zoology 16: 247–258. Johnson SD, Burgoyne PM, Harder LD, Dotterl€ S. 2011. Mammal pollinators Perrin M. 1997. Cape rock elephant shrew, Elephantulus edwardii. In: Mills G, lured by scent of a parasitic plant. Proceedings of the Royal Society B: Biological Hes L, eds. The complete book of Southern African mammals. Cape Town, South Sciences 278: 2303–2310. Africa: Struik Winchester, 66. Johnson SD, Hargreaves AL, Brown M. 2006. Dark bitter-tasting nectar Peterson MA, Poll L, Larsen LM. 1998. Comparison of volatiles in raw and functions as a filter of flower visitors in a bird-pollinated plant. Ecology 87: boiled potatoes using a mild extraction technique combined with GC odour 2709–2716. profiling and GC-MS. Food Chemistry 61: 461–466.

Ó 2019 The Authors New Phytologist (2019) 222: 1624–1637 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com New 1636 Research Phytologist

Pettersson S, Ervik F, Knudsen JT. 2004. Floral scent of bat-pollinated species: West Africa vs. the New World. Biological Journal of the Linnean Society 82: Supporting Information 161–168. Raguso RA. 2004. Why are some floral nectars scented? Ecology 85: 1486–1494. Additional supporting information may be found online in the Raguso RA. 2008. Wake up and smell the roses: the ecology and evolution of Supporting Information section at the end of the article. floral scent. Annual Review of Ecology, Evolution and Systematics 39: 549–569. Reyneke WF. 1972. ‘n Monografiese studie van die genus Eucomis L’Herit. in Suid Fig. S1 Y-maze used for choice experiments to test for the small Afrika. Department of Botany, University of Pretoria. Unpublished MSc thesis. mammals’ responses to methional (diluted in white mineral oil) Rourke JP, Wiens D. 1977. Convergent floral evolution in South African and Australian Proteaceae and its possible bearing on pollination by nonflying and a control (white mineral oil). mammals. Annals of the Missouri Botanical Garden 64:1–17. Sarrafchi A, Odhammer AME, Hernandez Salazar LT, Laska M. 2013. Fig. S2 Micaelamys namaquensis (Namaqua rock mouse) visiting Olfactory sensitivity for six predator odorants in CD-1 mice, human subjects, several flowers of a Eucomis regia ssp. pillansii inflorescence. and spider monkeys. PLoS ONE 8: e80621. Sch€afflerI,SteinerKE,HaidM,vanBerkelSS,GerlachG,JohnsonSD,Wessjohann L, D€otterl S. 2015. Diacetin, a reliable cue and private communication channel in a Fig. S3 Natural levels of seed set, fruit set (percentage of flowers specialized pollination system. Scientific Reports 5: 12779. that set seed) and seeds per fruit (back-transformed means Æ SE) Schiestl FP, D€otterl S. 2012. The evolution of floral scent and olfactory of Eucomis regia at different localities. preferences in pollinators: coevolution or pre-existing bias? Evolution 66: 2042– 2055. Fig. S4 Reflectance spectra of tepals, lower bracts, upper bracts Schiestl FP, Johnson SD. 2013. Pollinator-mediated evolution of floral signals. Trends in Ecology and Evolution 28: 307–315. and leaves of two Eucomis regia populations: the Nieuwoudtville Schumann DM, Cooper HM, Hofmeyr MD, Bennett NC. 2005. Circadian population (ssp. pillansii), and the Fairfield population (ssp. rhythm of locomotor activity in the four-striped field mouse, Rhabdomys regia). pumilio: a diurnal African rodent. Physiology and Behavior 85: 231–239. Shuttleworth A, Johnson SD. 2009. A key role for floral scent in a wasp- Fig. S5 Floral colours of the Eucomis species in bee and fly per- pollination system in Eucomis (Hyacinthaceae). Annals of Botany 103: 715–725. Shuttleworth A, Johnson SD. 2010. The missing stink: sulphur compounds can ceptional colour space. mediate a shift between fly and wasp pollination systems. Proceedings of the Royal Society B-Biological Sciences 277: 2811–2819. Fig. S6 Inconspicuous Eucomis regia ssp. regia flower lacking any Skinner JD, Chimimba CT. 2005. The mammals of the Southern African conspicuous pattern, shown in daylight and under UV illumina- subregion, 3rd edn. Cambridge, UK: Cambridge University Press. tion. Steenhuisen S-L, Johnson SD. 2012. Evidence for beetle pollination in the African grassland sugarbushes (Protea: Proteaceae). Plant Systematics and Evolution 298: 857–869. Table S1 Selected floral characteristics for Eucomis regia of the Stoddart DM. 1980. The ecology of vertebrate olfaction. London, UK: Chapman & Nieuwoudtville and Overberg populations Hall. Troje N. 1993. Spectral categories in the learning behaviour of blowflies. Table S2 Eucomis regia localities in South Africa (E. regia ssp. Zeitschrift fur€ Naturforschung 48c:96–104. Waterman RJ, Bidartondo MI, Stofberg J, Combs JK, Gerbauer G, Savolainen V, pillansii: Kamiesberg, Bokkeveld and Roggeveld and E. regia ssp. BarracloughTG,PauwA.2011.The effects of above and belowground mutualisms regia: Westcoast and Overberg), type and year of observation on orchid speciation and co-existence. American Naturalist 177:E54–E68. Wester P. 2010. Sticky snack for Sengis: The Cape rock elephant-shrew, Elephantulus Table S3 Eucomis regia herbarium specimens used for data on edwardii (Macroscelidea), as a pollinator of the Pagoda lily, Whiteheadia bifolia tepal dimensions (Hyacinthaceae). Naturwissenschaften 97:1107–1112. Wester P. 2011. Nectar feeding by the Cape rock elephant-shrew Elephantulus edwardii (Macroscelidea) – a primarily insectivore pollinates the parasite Hyobanche Table S4 Small mammals caught at the field sites near flowering atropurpurea (Orobanchaceae). Flora 206: 997–1001. Eucomis regia plants Wester P. 2015. The forgotten pollinators – First field evidence for nectar- feeding by primarily insectivorous elephant-shrews. Journal of Pollination Table S5 Pollen loads of mice captured near Eucomis regia ssp. Ecology 16: 108–111. Wester P, Stanway R, Pauw A. 2009. Mice pollinate the Pagoda Lily, regia plants at Swartrivier and Fairfield (Overberg) Whiteheadia bifolia (Hyacinthaceae) – First field observations with photographic documentation of rodent pollination in South Africa. South Table S6 Eucomis regia plant density and examined plants at dif- African Journal of Botany 75: 713–719. ferent localities Wiens D, Rourke JP. 1978. Rodent pollination in southern African Protea spp. Nature 276:71–73. Zhang K, Lin TF, Zhang T, Li C, Gao N. 2013. Characterization of typical taste Table S7 R values and significances for pairwise comparisons and odor compounds formed by Microcystis aeruginosa. Journal of (ANOSIM) of different Eucomis species and species grouped after Environmental Sciences 25: 1539–1548. pollinators regarding floral morphometrics and fragrance Zoeller KC, Midgley JJ, Johnson SD, Steenhuisen S-L. 2017. Floral biology and breeding systems of geoflorous Protea species (Proteaceae). South African Table S8 R values and levels of significance for pairwise compar- Journal of Botany 112: 452–459. Zonneveld BJM, Duncan GD. 2010. Genome sizes of Eucomis L’Her. (Hyacinthaceae) isons (ANOSIM) of Eucomis regia ssp. regia and ssp. pillansii and a description of the new species Eucomis grimshawii G.D.Duncan & Zonneveld. from different regions as to floral morphometrics and fragrance Plant Systematics and Evolution 284:99–109.

New Phytologist (2019) 222: 1624–1637 Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust New Phytologist Research 1637

Table S9 Relative amounts (%) of volatile compounds identified Videos S3 Elephantulus edwardii visiting Eucomis regia ssp. by GC-MS from headspace samples of Eucomis regia inflores- pillansii in the field (near Kamieskroon, Namaqualand) lapping cences and separated nectar nectar at dusk

Table S10 Mean percentage of volatile compounds (selection) Videos S4 Elephantulus edwardii visiting Eucomis regia ssp. emitted by inflorescences of both mammal- and insect-pollinated pillansii in the field (near Kamieskroon, Namaqualand) lapping Eucomis species (identified by GC-MS from headspace samples) nectar at night (infrared footage)

Table S11 Percentage contribution of specific scent compounds Videos S5 Elephantulus edwardii lapping nectar at Eucomis regia to the average similarity (based on Bray–Curtis coefficient, ssp. pillansii flowers with its long tongue SIMPER: similarity percentages) between Eucomis regia inflorescence samples representing a specific region (compounds shown Videos S6 Rhabdomys pumilio (Four-striped field mouse) lapping represent the first 50% of overall similarity) nectar at Eucomis regia ssp. regia flowers

Table S12 Nectar properties for Eucomis regia and insect- Videos S7 Young Rhabdomys pumilio (c. 2 wk old, born in captiv- pollinated Eucomis species ity) also licking nectar from ﬂowers of Eucomis regia ssp. regia

Videos S1 Micaelamys namaquensis (Namaqua rock mouse) lick- Please note: Wiley Blackwell are not responsible for the content ing nectar at Eucomis regia ssp. pillansii flowers or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be Videos S2 Elephantulus edwardii (Cape rock elephant-shrew) directed to the New Phytologist Central Office. visiting Eucomis regia ssp. pillansii in the field (near Kamieskroon, Namaqualand) lapping nectar during the day

New Phytologist is an electronic (online-only) journal owned by the New Phytologist Trust, a not-for-proﬁt organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews and Tansley insights.

Regular papers, Letters, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged. We are committed to rapid processing, from online submission through to publication ‘as ready’ via Early View – our average time to decision is <26 days. There are no page or colour charges and a PDF version will be provided for each article.

The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for table of contents email alerts.

If you have any questions, do get in touch with Central Office (np-centraloffi[email protected]) or, if it is more convenient, our USA Office (np-usaoffi[email protected])

For submission instructions, subscription and all the latest information visit www.newphytologist.com

Ó 2019 The Authors New Phytologist (2019) 222: 1624–1637 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com