The Ant-Pollination System of Cytinus Hypocistis (Cytinaceae), A Mediterranean Root Holoparasite
The Ant-Pollination System of Cytinus Hypocistis (Cytinaceae), A Mediterranean Root Holoparasite
The Ant-Pollination System of Cytinus Hypocistis (Cytinaceae), A Mediterranean Root Holoparasite
Received: 22 November 2008 Returned for revision: 6 January 2009 Accepted: 20 January 2009 Published electronically: 3 March 2009
† Background and Aims The genus Cytinus is composed of rootless, stemless and leafless parasites whose flowers
are only visible during the reproductive period when they arise from the host tissues. Most of the taxa occur in
Madagascar and South Africa, where mammal pollination has been suggested for one species. There is only one
species in the Mediterranean region, and its pollination system has been unknown. Here, a long-term field obser-
vation study is combined with experimental pollination treatments in order to assess the pollination biology and
reproductive system in the Mediterranean species Cytinus hypocistis.
† Methods Field studies were carried out in six populations in southern Spain over 4 years. Temporal and spatial
patterns of variation in the composition and behaviour of floral visitors were characterized. Pollen loads and
pollen viability were observed, and exclusion and controlled-pollination treatments were also conducted.
† Key Results Cytinus hypocistis is a self-compatible monoecious species that relies on insects for seed pro-
duction. Ants were the main visitors, accounting for 97.4 % of total floral visits, and exclusion experiments
showed that they act as true pollinators. They consistently touched reproductive organs, carried large pollen
loads and transported viable pollen, although the different ant species observed in the flowers differed in their
pollination effectiveness. The abundance of flying visitors was surprisingly low, and only the fly Oplisa aterrima
contributed to fruit production and cross-pollination.
† Conclusions Mutualistic services by ant are essential for the pollination of Cytinus hypocistis. Although this
parasite does not exhibit typical features of the ‘ant-pollination syndrome’, many other characteristics indicate
that it is evolving to a more specialized ant-pollination system. The striking interspecific differences in the polli-
nation systems of Mediterranean Cytinus (ant-pollinated) and some South African Cytinus (mammal-pollinated)
make this genus an excellent model to investigate the divergent evolution of pollination systems in broadly
disjunct areas.
Key words: Ant, breeding system, Cytinus hypocistis, Cytinaceae, insects, flies, Mediterranean Basin, parasitic
plant, pollination, Rafflesiaceae.
of insects foraging in flowers during 15-min-long watching Breeding system and reproductive success
periods (‘censuses’ hereafter) separated by 15-min intervals
The breeding system of C. hypocistis was studied by hand-
throughout daytime (0800– 2200). Given that many floral visi-
pollinations in plants from five populations (Cl1, Cs1, Cs2,
tors were small ants, we made four additional censuses of
Hh1 and Hh2) using within-plant self-pollination (n ¼ 212
5 min each hour at 0.5 m from the plants in order to register
TA B L E 1. Characteristics and behaviour of all Cytinus hypocistis pollinators that accounted for at least 0.2 % of total floral visits
(pooled across all populations and years).
1
Presence or absence of metapleural glands.
2
Pollen loads are given for the three most abundant species. Values are means + s.e. (range in parentheses).
3
Reward collected: P, pollen; N, nectar; TS, tepal secretions.
Insects that accounted for less than 0.2 % of total floral visits were Apis mellifera (Apidae, 0.02 % of total floral visits), Bombus terrestris (Apidae, 0.04 %),
an unidentified Braconidae (0.04 %), an unidentified Camilidae (0.02 %), and an unidentified Rinophoridae (0.06 %).
visited was compared among insect species and between type % of the nocturnal censuses, but only accounting for 0.3 % of
of visitors (flying and ants) by means of a GLZ with a Poisson total visits. These four ant species and Plagiolepis schmitzii
distribution and a log link-function. foraged on C. hypocistis flowers in all the years of the study,
Pollen loads on insects were compared among species by their abundance being consistent across years, but none
means of a one-way ANOVA, and differences in pollen ger- of them appeared in all the populations studied (Table 1).
minability between control and ant-treated pollen were com- The other ant species recorded only appeared in some
pared using a GLZ with a Poisson distribution and a log years (Aphaenogaster senilis and Crematogaster scutellaris:
link-function. 2002, 2003 and 2005; Tapinoma nigerrimum: 2002, 2004;
Differences in fruit set after hand-pollination treatments were Tetramorium ruginode: 2002; and Tetramorium semilaeve:
compared using a GLZ with a binomial distribution and a logit 2002 – 2004).
link-function, and the same method was used to compare fruit The number of individuals foraging on female and male
set after exclusion experiments vs. control. Mean numbers of flowers of C. hypocistis varied greatly among ant species,
seeds per fruit between pollination treatments and populations ranging from 0.18 (C. pilicornis) to 12.05 individuals per
were compared by means of a factorial ANOVA, and differences 5 min (T. ruginode, Table 2). It is noteworthy that, although
among seeds produced after exclusion and control experiments the two Tetramorium species appeared in a low number of
were compared by a one-way ANOVA. All statistical analyses censuses, when they did appeared they were very abundant
were performed in Statistica 6.0 (StatSoft, 2001). Throughout (Fig. 2, Table 2).
the text, all means are shown + s.e. Flying visitors mainly foraged in C. hypocistis inflorescences
without ants; however, aggressive interactions between ants and
flying insects were not seen. The fly Oplisa aterrima (1.3 % of
RES ULT S overall floral visits, and 51.2 % of visits of flying insects;
Floral visitor assemblage Fig. 1F) and solitary female bees of Lasioglossum sp. (1.1 %
and 41.6 %, respectively) were the most abundant flying
Pollen traps showed a low presence of C. hypocistis pollen species (Table 1). Oplisa aterrima was observed in all the popu-
grains (0.9 + 0.23 grains per trap; n ¼ 90) indicating that lations and years studied, with its abundance being consistent
wind-pollination is unlikely in this species. across years, whereas Lasioglossum appeared in all populations
The flowers of C. hypocistis were visited by 17 insect species but only during 2005 (Table 1).
belonging to six families, with ants being much more abundant
than flying visitors and accounting for 97.4 % of total floral
visits (n ¼ 4638; Table 1, Fig. 1). The most abundant daytime
Floral visitor foraging behaviour
ant species were Pheidole pallidula (35.3 % of total floral
visits), Plagiolepis pygmaea (19.4 %) and Crematogaster Ants and flying insects visited both male and female flowers
auberti (13.4 %). During the night only the ant Camponotus and made contact with the plant reproductive organs when
pilicornis visited the flowers of C. hypocistis, appearing in 50 foraging for floral resources, acting as potential pollinators
de Vega et al. — Pollination in Mediterranean Cytinus 1069
A B C
F I G . 1. Flowers and pollinators of Cytinus hypocistis: (A) Cytinus hypocistis, (B) Aphaenogaster senilis, (C) Crematogaster auberti, (D) Pheidole pallidula,
(E) Tetramorium semilaeve and (F) Oplisa aterrima. Scale bars ¼ 5 mm.
(Fig. 1). Ants foraged for nectar, Lasioglossum foraged both however, such movements were much less for the nocturnal
for pollen and nectar, and Oplisa aterrima for nectar and C. pilicornis (40 %; Table 2). Oplisa aterrima and
secretions from tepal glandular trichomes; none of them Lasioglossum made approx. 60 % of flights within plants
destroyed floral organs. Some of the visitors foraged more fre- (Table 2).
quently and spent more time in flowers of a given sex (Figs 2
and 3; Table 2). For each species, the number of visits and
Pollen loads and germination ability
mean time spent in a flower was highly variable throughout
the day (Fig. 3). All ants consistently touched the anthers with their bodies
The mean number of flowers foraged per plant by each indi- and carried pollen loads that were significantly different
vidual insect ranged from 2.62 to 3.65, the differences being among species (x2 ¼ 20.79, d.f. ¼ 2, P , 0.0001). Pollen
significant among species (x2 ¼ 20.574, d.f. ¼ 9, P ¼ 0.015). loads were not related to body size, since the smallest ant,
The mean number of flowers per plant foraged by ants P. pygmaea, carried the highest pollen loads (Table 1). In all
(2.83 + 0.05; all species combined) was similar to that for ant species pollen grains were attached to the head, thorax
winged pollinating insects (2.68 + 0.20, x2 ¼ 1.075, d.f. ¼ 1, and gaster. There was no detrimental effect to pollen germina-
P ¼ 0.299). Between 76.19– 96.97 % of movements made by tion after contact with the integument of C. auberti and
ants in the daytime were between flowers of the same plant; P. pallidula, either in the treatment of 10 min or 30 min
TA B L E 2. Behaviour of visitors while foraging on Cytinus hypocistis flowers. All species are ants except those marked with an
asterisk
Species Female flowers Male flowers Female flowers Male flowers No flowers visited Within-plant movements (%)
Aphaenogaster senilis 0.48 + 0.12 0.55 + 0.14 20.90 + 3.91 74.85 + 24.84 3.65 + 0.50 76.19
Camponotus pilicornis 0.18 + 0.07 0.24 + 0.08 9.00 + 3.76 8.67 + 5.17 3.00 + 0.52 40.00
Crematogaster auberti 2.88 + 0.34 1.51 + 0.166 70.05 + 11.70 60.71 + 8.59 2.97 + 0.11 95.87
Crematogaster scutellaris 1.03 + 0.18 1.03 + 0.17 69.00 + 35.13 24.33 + 17.84 2.64 + 0.17 92.13
Lasioglossum sp.* 0.15 + 0.11 1.18 + 0.18 15.48 + 4.86 13.71 + 2.11 2.62 + 0.29 58.62
Oplisa aterrima* 0.69 + 0.09 0.49 + 0.09 27.09 + 6.15 19.41 + 8.43 2.77 + 0.22 61.29
Pheidole pallidula 4.94 + 0.65 4.94 + 0.53 69.81 + 6.24 80.80 + 8.39 2.62 + 0.07 87.57
Plagiolepis pygmaea 0.60 + 0.06 1.00 + 0.07 41.01 + 4.38 43.43 + 3.99 2.83 + 0.09 98.61
Plagiolepis schmitzii 0.93 + 0.12 1.17 + 0.12 145.96 + 23.13 61.18 + 15.73 2.65 + 0.14 92.50
Tapinoma nigerrimum 0.76 + 0.18 0.78 + 0.17 37.05 + 17.52 16.34 + 3.86 2.82 + 0.18 96.97
Tetramorium ruginode 10.74 + 1.19 12.05 + 1.73 75.14 + 26.74 86.11 + 26.71 3.01 + 0.52 91.94
Tetramorium semilaeve 5.26 + 0.74 5.78 + 0.88 27.40 + 8.13 38.47 + 7.38 3.32 + 0.31 89.47
1·0 4 0·8
0·8 3 0·6
00
00
00
00
00
00
00
00
20
21
22
23
00
01
02
03
2·5 Crematogaster scutellaris 2·0 Lasioglossum sp.* 1·2 Oplisa aterrima*
No. individuals
2·0 1·0
per 5 min
1·5
1·5 0·8
1·0 0·6
1·0
0·4
0·5 0·5
0·2
0·0 0·0 0·0
8 1·6 1·6
per 5 min
6 1·2 1·2
4 0·8 0·8
2 0·4 0·4
0 0·0 0·0
30 Tetramorium ruginodis 20 Tetramorium semilaeve
2·5 Tapinoma nigerrimum
No. individuals
2·0 25 16
per 5 min
20 12
1·5
15
1·0 10 8
0·5 5 4
0·0 0 0
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
07
09
11
13
15
17
19
21
07
09
11
13
15
17
19
21
07
09
11
13
15
17
19
21
Time of day Time of day Time of day
F I G . 2. Insects visiting the flowers of Cytinus hypocistis during the day. Note that for each species the y-axis differs, and the x-axis is also different for the
nocturnal Camponotus pilicornis. Differences in the number of male and female flowers visited were observed for C. auberti (Z ¼ 3.96; P , 0.0001),
O. aterrima (Z ¼ 1.91; P ¼ 0.055) Lasioglossum (Z ¼ 3.69; P ¼ 0.0002), P. pygmaea (Z ¼ 7.71; P ¼ ,0.0001) and P. schmitzii (Z ¼ 1.97; P ¼ 0.049). All
species are ants except those marked with an asterisk.
(P . 0.25 in all cases; Table 3). In contrast, P. pygmaea and The mean number of seeds per fruit was 19 305 + 2171 in
C. pilicornis significantly reduced pollen germination self-pollinated flowers, 19 200 + 2333 in cross-pollinated
(Table 3); however, it is noteworthy that, even in the latter flowers, and 24 163 + 3847 in open-pollinated flowers.
two species, a considerable fraction of pollen remained Mean number of seeds per fruit did not vary among treatments
viable after contact. (F ¼ 0.96, d.f. ¼ 2, P ¼ 0.394) or among populations
Pollen loads were not measured on winged insects. (F ¼ 2.68, d.f. ¼ 4, P ¼ 0.083).
However, during the censuses it was observed that the fly
Oplisa aterrima and the bee Lasioglossum carried pollen
loads on the legs and body. Exclusion experiments
Flying-visitor exclusion experiments showed that ants act as
true pollinators. In the three populations studied, plants
Breeding system
exposed only to ants showed a fruit set of 80 – 87 %, and in
None of the permanently covered female flowers produced all flowers analysed pollen tubes were observed on the styles
fruit, indicating that pollinators are required to set seeds. and ovules.
Microscopic observations revealed many pollen tubes on the The fruit set of plants of the ‘ant-only’ treatment was stat-
style and also penetrating the ovules of all flowers examined istically lower than that of control flowers (x2 ¼ 43.07,
in both geitonogamous and cross-pollination, giving evidence d.f. ¼ 1, P , 0.0001), suggesting that flying insects also con-
of self-compatibility in this species. Cytinus hypocistis showed tribute significantly to fruit production in C. hypocistis
high values of fruit set after self-pollination (89.85– 98.11 %), despite the low quantitative importance of their visitation
similar to those observed after cross-pollination and open- rates. Although differences among populations were statisti-
pollination (Table 4). Significant differences were not found cally significant (x2 ¼ 7.53, d.f. ¼ 2, P ¼ 0.023), fruit set
either among treatments (x2 ¼ 2.59, d.f. ¼ 2, P ¼ 0.274) or after the ant-only treatment was similarly reduced relative to
populations (x2 ¼ 9.35, d.f. ¼ 4, P ¼ 0.063). Differences control plants in all populations ( population-by-treatment
among years for each population in the percentage of fruit interaction not significant: x2 ¼ 2.33, d.f. ¼ 2, P ¼ 0.311).
set after open-pollination were not significant (in all cases However, the mean number of seeds per fruit in plants
P . 0.6). exposed only to ants (19 418 + 2776) was not significantly
de Vega et al. — Pollination in Mediterranean Cytinus 1071
250 75 90
200 60 75
150 45 60
45
100 30
30
50 15 15
0 0 0
Pheidole pallidula Plagiolepis pygmaea Plagiolepis schmitzii
Time in flowers (s)
F I G . 3. Time spent by different pollinators on female and male flowers of Cytinus hypocistis during the day. Note that for each species the y-axis differs, and the
x-axis is also different for the nocturnal Camponotus pilicornis. Aphaenogaster senilis spent more time on male flowers (x2 ¼ 11.83; P ¼ 0.0006), whereas
P. schmitzii (x2 ¼ 22.01; P , 0.0001) and T. nigerrima (x2 ¼ 22.01; P , 0.0001) spent more time on female flowers. In the other species no differences
were found. All species are ants except those marked with an asterisk.
TA B L E 3. Percentage germination of Cytinus hypocistis pollen different to those produced after open-pollination (F ¼ 1.02,
without contacting ants (control) or after contact with ants of d.f. ¼ 1, P ¼ 0.321).
different species (ant-treated pollen) for either 10 or 30 min. Ripe seeds from ant-, cross-, open- and self-pollination
Sample sizes (total number of pollen grains examined) are seemed to be viable. All of them presented a small embryo
shown in parentheses composed of about ten cells surrounded by an endosperm com-
posed of larger cells, as is typical of C. hypocistis seeds (see de
Germination (%)
Contact time and
Vega and de Oliveira, 2007).
ant species Control Ant-treated F P
10 min DISCUSSION
Crematogaster 64.05 + 7.94 (2305) 58.45 + 5.75 (2358) 0.327 0.574
auberti
This study has shown that Cytinus hypocistis is a self-
Pheidole 65.52 + 7.65 (2207) 53.56 + 7.19 (2226) 1.298 0.269 compatible species consistently exhibiting a high fruit set
pallidula and extensive seed production in natural conditions (i.e. open-
30 min pollination). The species relies on insects for seed production
Crematogaster 57.92 + 6.50 (2203) 47.85 + 3.98 (2128) 1.744 0.203 since neither wind pollination nor agamospermy were
auberti
Pheidole 60.75 + 7.19 (2204) 51.15 + 6.37 (2181) 0.999 0.331 recorded. The long-term field observations and exclusion
pallidula experiments strongly suggest that ants are the main pollinators
Plagiolepis 63.39 + 9.94 (1088) 26.71 + 5.29 (1319) 11.725 0.008 in C. hypocistis, accounting for most flower visits in all
pygmaea populations and years, and yielding a fruit set close to 80 %
Camponotus 63.28 + 4.48 (1263) 14.76 + 9.27 (1264) 22.192 0.001
pilicornis
when other potential visitors were excluded. Accordingly,
this study joins a growing body of evidence highlighting the
1072 de Vega et al. — Pollination in Mediterranean Cytinus
TA B L E 4. Percentage fruit set of Cytinus hypocistis after self-pollination, cross-pollination, open-pollination, ant-only pollination or
total exclusion of flower visitors (sample sizes are shown in parentheses)
prominent role that ants can play in some plant – pollinator excellent or a poor pollinator, the same as for a bee or a but-
systems (Wyatt, 1981; Peakall, 1989; Peakall and Beattie, terfly, and so generalizations about the ability and quality of
1989; Gómez and Zamora, 1992, 1999; Garcı́a et al., 1995; ants as pollinators should be carefully considered.
Ramsey, 1995; Gómez et al., 1996; Puterbaugh, 1998; Even though C. hypocistis flowers offer abundant nectar and
Gómez, 2000; Schürch et al., 2000; Ashman and King, pollen, the diversity and abundance of flying insects foraging
2005; Sugiura et al., 2006). However, contrary to most pre- in this parasite is surprisingly low, in spite of many bees,
vious studies in which single ant species were involved in pol- flies, butterflies, moths and beetles being observed foraging
lination (but see Gómez et al., 1996; Ashman and King, 2005), on the flowers of other plant species nearby. Similar obser-
in C. hypocistis as many as ten species can be considered as vations were done in more than 50 populations elsewhere in
true pollinators. Spain and Morocco (C. de Vega, CSIC, Sevilla, Spain,
The effectiveness of a given pollinator not only depends on unpubl. res.). Since winged-visitors were only observed in
their abundance and frequency of flower visitation but also on inflorescences with few or no ants, it is likely that the presence
the efficiency with which they remove and deposit pollen of ants deterred flying pollinators from visiting the flowers, as
(quantity and quality components of the plant – pollinator inter- has been suggested in other species (McDade and Kinsman,
action, respectively; Herrera, 1987, 1989). Ants were the most 1980; Schürch et al., 2000; Philpott et al., 2005). Among the
abundant pollinators during both the day and night, accounting flying insects, only the fly O. aterrima was a predictable
for 97.36 % of floral visits, and while foraging for nectar they visitor of C. hypocistis, appearing in all populations and
frequently touched the anthers (each male flower may produce years, whereas other winged insects only appeared during
more than one million pollen grains; de Vega, 2007) and the the drought year of 2005, when the dramatically low availability
wide stigmatic surface (4.2 2.5 mm in size). In addition, of flowers of other species in the populations probably forced
we have demonstrated that ants carried large pollen loads them to search for alternative feeding resources. Temporal
and that they spent a long time foraging at each flower, variation in the identity and importance of pollinators has
which increases the probability for them to contact the repro- been observed in other plant species (Herrera, 1988; Gómez
ductive organs of C. hypocistis, and thus promoting pollina- and Zamora, 1999; Thompson, 2001; Ivey et al., 2003);
tion. Furthermore, pollen germination experiments showed nonetheless, for C. hypocistis, despite the differences found in
that the ants’ metapleural secretions did not play a major pollinator composition between years, the pollinator assemblage
role in inhibiting pollen viability, in contrast to the hypothesis that accounted for most visits remained remarkably constant.
stating that these antimicrobial secretions mostly prevent the When pollinator assemblages are taxonomically hetero-
transfer of viable pollen (Beattie et al., 1984, 1985; Hull and geneous, determination of the behaviour of each species is
Beattie, 1988; but see Peakall and Beattie, 1989; Gómez and important for understanding their differential contribution to
Zamora, 1992; Garcı́a et al., 1995; Gómez et al., 1996). It plant reproduction (Herrera, 1987; Giménez-Benavides et al.,
may be possible that the high pollen production could ‘neutral- 2007; Li et al., 2008). Our findings suggest that despite their
ize’ the effects of ant metapleural secretions, as has been low proportional frequency, the fly O. aterrima might play a
suggested by Hull and Beattie (1988). Two of the most abun- key role in producing outcrossed progeny in C. hypocistis,
dant ant species, P. pallidula and C. auberti, have metapleural since it tends to visit few flowers per individual and move
glands, but they did not significantly reduce pollen viability. It more often among different plants. In contrast, ants exhibited
is noteworthy that although C. pilicornis reduced pollen ger- a restricted foraging area and repeatedly visited individual
minability (the same trend has been observed for other flowers and inflorescences, thus potentially enhancing the
Camponotus species; Beattie et al., 1985; Hull and Beattie, occurrence of geitonogamous selfing, as has been observed
1988), this ant species has no metapleural glands, a common in other ant-pollination systems (Svensson, 1985; Peakall and
feature in the genus (Hölldobler and Engel-Siegel, 1984). Beattie, 1991; Gómez and Zamora, 1992). Although selfing
Perhaps antibiotic substances are distributed throughout the often negatively affects plant reproductive success, self-
cuticle or may be secreted from a different gland (Beattie compatibility may be advantageous in species with small
et al., 1985; Hull and Beattie, 1988). Our findings have population sizes, subjected to strong stochastic demographic
shown that when different ant species are pollinating a given fluctuations (Stebbins, 1957; Schemske and Lande, 1985;
plant species they can differ in their pollination effectiveness, Pannell and Barrett, 1998) as it occurs in C. hypocistis
as is well known to occur in other pollinator guilds (Herrera, (de Vega, 2007; de Vega et al., 2008). In C. hypocistis pollina-
1987; Larsson, 2005; Fumero-Cabán and Meléndez- tion exclusively by ants resulted in lower fruit set than open-,
Ackerman, 2007). Consequently, an ant can be either an cross- and geitonogamous pollinations; however, it is unlike
de Vega et al. — Pollination in Mediterranean Cytinus 1073
that this decrease is due to the self-pollen carried by ants, the divergent evolution of pollination systems in broadly dis-
because geitonogamous hand-pollinations result in similar junct areas. Floral morphology, nectar characteristics and
levels of fruit set to cross-pollinations. Our field observations floral scents could be playing crucial roles in establishing
suggest that ant behaviour could account for this reduction in mutualistic pollination interactions in Cytinus, and may be
Garcı́a MB, Antor RJ, Espadaler X. 1995. Ant pollination of the paleoen- Molau U. 1995. Reproductive ecology and biology. In: Press MC, Graves JD
demic dioecious Borderea pyrenaica (Dioscoreaceae). Plant Systematics eds. Parasitic plants. London: Chapman and Hall, 141 –177.
and Evolution 198: 17–27. Nickrent DL. 2002a. Plantas parásitas en el mundo. In: López-Sáez JA,
Garcı́a-Franco JG, Rico-Gray V. 1997. Reproductive biology of the holopar- Catalán P, Sáez L eds. Plantas Parásitas de la Penı́nsula Ibérica e
asitic endophyte Bdallophyton bambusarum (Rafflesiaceae). Botanical Islas Baleares. Madrid: Mundi-Prensa Libros S.A., 7 –28.
de Vega C, de Oliveira RC. 2007. A new procedure for making observations Visser J. 1981. South African parasitic flowering plants. Capetown: Juta &
of embryo morphology in dust-like seeds with rigid coats. Seed Science Co.
Research 17: 63–67. Wallace LE. 2003. The cost of inbreeding in Platanthera leucophaea
de Vega C, Ortiz PL, Arista M, Talavera S. 2007. The endophytic system of (Orchidaceae). American Journal of Botany 90: 235–242.
Mediterranean Cytinus (Cytinaceae) developing on five host Cistaceae Webb DA. 1964. Cytinus. In: Tutin TG, Heywood VH, Burges NA, Valentine