Supplementary Information
TABLE S1: Number of crosses and controls realized per receiver population in the controlled crossing
experiment with Ranunculus bulbosus
Elevation of
receiver
population
Same
Different
Different
SameValley /
Valley /
Valley /
Valley /
Same
Different
Different
Same
Replicate
(short name)
Elevation
Elevation
Elevation
Elevation
Self
Bern
1200 (Fae)
18
20
19
19
19
19
114
1800 (Moo)
27
23
22
24
31
23
150
1200 (Ers)
5
6
5
5
10
4
35
1800 (Fes)
31
15
29
31
29
24
159
1200 (Hou)
15
13
11
12
15
13
79
1800 (Waa)
23
25
22
21
20
21
132
119
102
108
112
124
104
669
Wallis
West
TOTAL
NoPollen TOTAL
APPENDIX S1: Experimental validation of the crosses
Two control treatments were conducted in each of the six pollen receiver populations. First,
a self-pollination in absence of pollinating vector (“Self”; the closed bud was simply bagged), and
second, a no-pollination treatment (“NoPollen”; absence of any pollen, even its own: the
emasculated flower was bagged).
After harvesting and counting the parental seed-sets, we analysed the data in the R
environment v. 2.15.0 (R Development Core Team 2012). Since it showed an excess of zeroes, we
used hurdle models, which consist of a binary part applied to all data modelling the occurrence of
successes (counts > 0) and failures (counts = 0) and of a count part applied on non-zero data; these
two parts were modelled with the glmmPQL() function from the R-package “MASS” (Venables &
Ripley 2002). The data was modelled with the two main terms: relative elevation of receiver
population compared to elevation of pollen-donor population (factor “ElevationGroup”, with levels
“Same” and “Different”) and relative geographical origin of pollen (factor “ValleyGroup”, with levels
“Same” and “Different”), along with the two-way interaction as fixed effects; “Replicate” was used as
a random term. Each model was separately run on the 1200m a.s.l. and 1800m a.s.l. pollen-receiver
populations.
The ‘selfing’ control-treatments “Self” and “NoPollen” (sample size N = 124 and 104, respectively),
where mostly infertile (71%), but set seed in 66 cases, the number of seeds ranging from 1 to 15
(median=2) per plant. The probability of these two controls setting at least one seed was slightly
influenced by the elevation of the receiver population (ElevRecPop; p-value = 0.0395, df=224):
populations of 1200m a.s.l. set seed with a probability of 0.375, while the 1800m a.s.l. populations
had a probability of 0.243 of setting seed. But the number of seeds produced was independent of
the control type (“Self” or “NoPollen”) and elevation of the receiver population.
Comparing the seed-sets produced by the actual crosses to the controls showed that the
treatment type (control vs. cross) had a highly significant effect on the probability of setting seeds
(p=0.0000, df=664; prob(seeds>0) for crosses = 0.85 vs. 0.36 for controls). Likewise, the crosses
produced on average a 5.4 times higher seed number per flower than the controls (15.7 seeds vs. 2.9;
p=0.0000, df=413). The elevation of receiver population had about the same significant effect on the
probability of setting seeds in the crossing treatments as within the controls “Self” and “NoPollen”
(p=0.0092, df=664; for 1800m a.s.l. populations prob(seeds>0) = 0.25; for 1200, 0.36), but did not
have an effect on the seed number.
Given the results of the control treatments, we acknowledge that the offspring resulting from our
crosses may contain some selfed individuals. Yet, we are confident that this risk is negligible owing to
the very different probabilities of setting seed, and to the (potential) seed-set size. Also, the pollenload applied by the crossing treatment combined with tearing the anthers off may further reduce the
probability of selfing, as in many plants, selfing can act as a response to the lack of available pollen
(Nettancourt, 2001; Garcia et al., 2005). Eventually, the odds of randomly picking a selfed seedling
during the re-potting step of the hand-pollinated individuals (see Methods section in the main text)
are very small.
FIG.S1. Comparison of within-population crosses (dark grey) vs. between-populations crosses (light
grey) in Ranunculus bulbosus, for each elevation of receiver populations (1200m a.s.l. and 1800m
a.s.l.). Between-populations
crosses consist of the pooled combinations
SameValley /
DifferentElevation, DifferentValley / DifferentElevation, and DifferentValley / SameElevation.
Transformation of the data is indicated on the y-axis wherever applicable. Error bars are standard
errors. At the bottom of each bar the number of observations per group is indicated. Plots are
arranged in the same order as Fig.2.