Variation in the production of plant tissues bearing extrafloral nectaries explains temporal patterns of ant attendance in Amazonian understorey plants
Item Type Article
Authors Nogueira, Anselmo; Baccaro, Fabricio B.; Leal, Laura C.; Rey, Pedro J.; Lohmann, Lúcia G.; Bronstein, Judith L.
Citation Nogueira, A., Baccaro, F. B., Leal, L. C., Rey, P. J., Lohmann, L. G., & Bronstein, J. L. (2020). Variation in the production of plant tissues bearing extrafloral nectaries explains temporal patterns of ant attendance in Amazonian understorey plants. Journal of Ecology. https://doi.org/10.1111/1365-2745.13340
DOI 10.1111/1365-2745.13340
Publisher WILEY
Journal JOURNAL OF ECOLOGY
Rights Copyright © 2019 British Ecological Society.
Download date 29/09/2021 00:25:45
Item License http://rightsstatements.org/vocab/InC/1.0/
Version Final accepted manuscript
Link to Item http://hdl.handle.net/10150/637522 1 RESEARCH ARTICLE
2 Variation in the production of plant tissues bearing extrafloral nectaries explains temporal
3 patterns of ant attendance in Amazonian understory plants
4 Anselmo Nogueira1*, Fabricio B. Baccaro2, Laura C. Leal3, Pedro J. Rey4, Lúcia G. Lohmann5 and
5 Judith L. Bronstein6.
6 1Centro de Ciências Naturais e Humanas, Universidade Federal do ABC, 09606-070 - São Bernardo
7 do Campo, SP, Brazil.
8 2Instituto de Biologia, Departamento de Biologia, Universidade Federal do Amazonas, Manaus,
9 69060-020, Amazonas, Brazil.
10 3Departamento de Ecologia e Biologia Evolutiva, Universidade Federal de São Paulo. Av. Conceição,
11 215, Diadema, 09972-270, São Paulo, Brazil.
12 4Departamento de Biología Animal, Biología Vegetal y Ecología, Universidad de Jaén, Paraje Las
13 Lagunillas s/n, E-23071, Jaén, Spain.
14 5Departamento de Botânica, Instituto de Biociências, Universidade de São Paulo, Rua do Matão, 277,
15 São Paulo-SP, 05508-090, Brazil.
16 6Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona 85721,
17 USA.
18
19 Running title: Temporal variation of ant-plant interactions
20
21 *Corresponding author: [email protected] ; [email protected]
22
23
24
25
26
27
1 28 ABSTRACT (350 words)
29 1. Information on direct and indirect drivers of the temporal variation of ant-plant interactions is still
30 scarce, compromising our ability to predict the functioning of these interactions at ecological and
31 evolutionary timescales.
32 2. Here we explicitly investigated the roles of precipitation, ant activity, abundance of young plant tissues
33 bearing extrafloral nectaries (EFNs), and EFN phenotypes in the establishment of ant-plant interactions
34 throughout the year at a site in Amazonia, Brazil. We hypothesized that the frequency of ant-plant
35 interactions follows a predictable seasonal pattern, being higher in wetter periods, during which plants
36 invest more in the production of new plant tissues bearing EFNs, ultimately promoting ant attendance.
37 We surveyed and tagged every understory Bignonieae plant rooted inside 28 500-m² plots, recording
38 ant-EFN interactions on each plant five times throughout the year. We also sampled ants with honey
39 baits to estimate temporal variation in ant activity.
40 3. The proportion of plants tended by ants in each plot was higher in drier months. Ant attendance was
41 indirectly and negatively related to precipitation, which was attributed to a decrease in the proportion
42 of plants producing new plant tissue bearing extrafloral nectaries during the wetter period.
43 Additionally, variation in ant activity did not explain temporal patterns of plant attendance. At the plant
44 level, ant attendance was positively and strongly related to the number of recently formed shoot nodes,
45 and ants almost never attended plants with few or none young tissues. Among the twelve most
46 abundant Bignonieae species, the amount of young tissues was the most important predictor of ant
47 attendance, secondarily explained by the EFN secretory area per plant node.
48 4. Synthesis. Our results highlight that seasonal variation in the production of new plant tissues bearing
49 EFNs is the primary driver of the temporal patterns of EFN-plant attendance by ants in this system.
50 Contrary to our expectations, production of new plant tissue is intensified in the drier months of the
51 year, boosting the frequency of interactions between ants and EFN-bearing plants. These results
52 highlight the role of plant phenology in the remarkable variation encountered in ant visitation to EFNs
53 in both space and time.
54
55 KEYWORDS: Ant-plant interactions, ecological interactions, herbivory, indirect defence, plant-herbivore
56 interactions, mutualism, tropical rainforest.
57 58
2 59 INTRODUCTION
60 In recent decades, it has become increasingly clear that mutualistic interactions are
61 highly variable across time and space (Bronstein, 2015). Several studies have highlighted
62 remarkable variation in the frequency of interactions between species pairs in different
63 habitats, both within and across years (e.g., Horvitz & Schemske, 1990; Di Giusto et al.,
64 2001; Burns, 2004; Olesen et al., 2008; Carnicer et al., 2009; Heath et al., 2010). The sign
65 and magnitude of the outcome of such interactions also vary widely over spatial and temporal
66 gradients (Chamberlain et al., 2014; Hoeksema & Bruna, 2015). Identifying factors driving
67 such variation is a fundamental step towards an improved understanding of the ecological and
68 evolutionary processes that shape mutualisms. However, most studies have focused on the
69 spatial drivers of such variability, relatively neglecting the role of the temporal drivers and
70 compromising the general understanding of how these interactions function at different time
71 scales.
72 Interactions between ants and plants bearing extrafloral nectaries (EFNs) are good
73 examples of mutualistic interactions that vary across both temporal and spatial scales
74 (Bronstein 1998, Rico-Gray & Oliveira 2007). In such interactions, plants secret nectar from
75 glands located outside of flowers that are highly attractive to ants. These ants forage on the
76 plants, repelling or consuming potential herbivores (Marazzi et al., 2013). In these systems,
77 plant investment in the production of extrafloral nectar generally decreases as EFNs age, with
78 more active nectaries being found on younger plant tissues (Bentley 1977, Rico-Gray &
79 Oliveira 2007), eventually ceasing secretion as they senesce (Dreisig, 2000). As the
80 production of new plant tissues occurs in brief intervals throughout the year (Fenner, 1998),
81 we should expect ant-plant interactions to vary over time in response to plant phenology.
82 Indeed, ecologists have long used the age of plant parts as a criterion for when to study ant-
83 EFN interactions (e.g. Di Giusto et al., 2001, Nogueira et al., 2015), suggesting that the
3 84 initiation of new plant tissues can be a major factor regulating ant-plant-herbivore
85 interactions through time.
86 In tropical forests around the world, temporal patterns of water uptake determine
87 biological cycles of several organisms, including plants (Chaves et al., 2002; Zeppel et al.,
88 2014). Temporal patterns of water uptake may also predict temporal patterns of interaction
89 between ants and plants bearing EFNs. In general, the frequency of ant-plant interactions
90 mediated by EFNs is higher in wet than dry seasons (e.g. Rico-Gray, 1993; Dias-Castelazo &
91 Rico Gray, 2004; Belchior et al., 2016). Although seasonal variation in water uptake can
92 partially explain such patterns, the proximate causes of the establishment of ant-plant
93 interactions seem related to the variation of plant resource availability and ant consumer
94 densities (Holland et al., 2005). Therefore, it is possible that the production of new EFN
95 bearing plant tissue, and, complementarily, local ant densities, can be the two major factors
96 regulating the temporal patterns of ant-plant-herbivore interactions throughout tropical
97 ecosystems.
98 Here, we investigated temporal patterns of ant-plant interactions and their direct and
99 indirect drivers in a site within the Amazon rainforest. Since abiotic factors such as water
100 availability could lead to temporal variation in the production of new plant tissues, including
101 the initiation of EFNs on the youngest plant parts (Wäckers & Bonifay, 2004), we
102 hypothesised that ant attendance on EFNs would follow a predictable temporal pattern, being
103 higher during wetter periods when plants invest more resources in the production of new
104 tissues. At the community level, such an increase in ant visitation to EFNs might be driven
105 either by an increase in EFN availability or by an increase in the local ant activity over time.
106 Hence, we explicitly evaluated if variation in ant-plant interactions resulted from temporal
107 variation in patterns of ant activity, from the temporal availability of young EFN-bearing
108 tissues, or both. If the availability of young EFN-bearing tissues predicts patterns of ant-plant
4 109 interactions at the community level, we also hypothesised that at the plant level, the
110 probability of an individual plant being attended by ants could also be explained by the
111 abundance of young tissues bearing active EFNs. To evaluate our hypotheses, we used a
112 species-rich plant clade that is abundant throughout the Amazon rainforest, the tribe
113 Bignonieae (Bignoniaceae) (Lohmann, 2006). Bignonieae have nectar-secreting trichomes on
114 younger plant tissues that function as EFNs (Nogueira et al., 2012a,b). These EFNs are
115 highly variable in abundance across species (Nogueira et al., 2013), and this variation was
116 also explicitly considered in tests of our hypotheses.
117
118 MATERIALS AND METHODS
119 Study site
120 We carried out this study at the Reserva Florestal Adolpho Ducke (Ducke Reserve)
121 located 26 km northwest of Manaus, Amazonas State (Brazil). This site characterised by a
122 mean annual temperature of ~26 °C and an annual rainfall of 2400 mm (Ribeiro et al., 1999).
123 The peak of the wet season is January-March, when rainfall averages > 300 mm per month;
124 the peak of the dry season is July-September, when rainfall averages < 100 mm per month
125 (Marques-Filho et al., 1981). The Ducke Reserve is covered by a terra-firme tropical
126 rainforest with vegetation characterised by a closed canopy and diverse understory species
127 with abundant subterranean-stemmed palms. The soils represent a continuum from clayey
128 latosols on the ridges becoming sandier as the altitude decreases (Chauvel et al., 1987).
129
130 Plant system and young tissues bearing secretory extrafloral nectaries
131 We choose the Bignonieae plants as our model group, a species-rich lineage of
132 neotropical lianas that bear EFNs on vegetative plant parts (Lohmann, 2006). These EFNs
133 attract ants capable of excluding herbivores on plants (e.g. Nogueira et al., 2012a, 2015). In
5 134 Bignonieae, EFNs are nectar-secreting trichomes derived from the differentiation of
135 epidermal cells after primary growth. These secreting trichomes occur in clusters on the
136 interpetiolar region of stems, in the prophylls of axillary buds, and leaflet bases (Seibert,
137 1948, Nogueira et al., 2012b; Fig. 1). In the field, clusters of active EFNs can be recognised
138 as yellowish-green and turgid structures secreting crystalline and viscous nectar (Nogueira et
139 al., 2012b, 2013).
140 Although EFN structure varies among vegetative plant parts in Bignonieae species,
141 EFNs are generally organised in subunits or modules composed by a recently formed shoot
142 node with four tiny prophylls and two opposite leaves (Nogueira et al., 2012b). The number
143 of new modules can be easily recognised in the field, representing a good descriptor of the
144 amount of young plant tissue within each individual. The young tissues are bright green,
145 recently formed shoots, lacking development of secondary tissues, and with incompletely
146 expanded leaves (Nogueira et al., 2012a,b; 2015). Therefore, individual plants can be
147 characterised by zero to multiple new modules of young tissues on the aboveground plant
148 body, depending on the phenological stage of the plant.
149 Plants produce new tissues in pulses over time (see Introduction), and Bignonieae
150 plants, in particular, can produce zero to many new modules bearing secretory EFNs each
151 year. Therefore, variation in EFN availability among plants can be generated by variation in
152 the production of new modules per plant (the amount of young tissues) or by variation in the
153 number/area of EFNs per module (see Nogueira et al., 2012b; 2013 for detailed
154 morphological descriptions of EFNs of 108 Bignonieae species). As our general hypothesis
155 predicts that availability of EFNs is the main factor driving the seasonally of the ant-plant
156 interactions, we explicitly explored both aspects of plant phenotype, investigating the impact
157 of the amount of young tissues and the secretory area of EFNs on ant-plant interactions over
158 time. Specifically, we evaluated if the temporal variation in the production of new plant
6 159 tissues would drive temporal patterns of ant-plant interactions at the community level (per
160 plot) and individual level (i.e. probability of each plant being tended by ants). Additionally,
161 we evaluated if the probability of individual plants being tended by ants would be driven by
162 the secretory area of EFNs in each recently formed shoot node.
163
164 Sampling design and plant descriptors
165 We systematically established 28 plots over a 9 km2 grid, formed by long trails
166 crossing north-south and west-east within the Ducke Reserve. Plots were 500 m2 (50 x 10 m)
167 and located at least 500 m from each other. We surveyed and tagged individual plants of
168 Bignonieae rooted inside the plot whose diameter at 30 cm from the rooting point (D30cm) was
169 ≥ 0.5 cm and whose vertical height was ≤ 3 m. Our sampling only included juvenile non-
170 reproductive individuals occurring in the understory. We focused on this group of plants
171 because they tend to be more visited by herbivores (A. Nogueira, pers. obs.) and less tolerant
172 of herbivore damage (Boege et al., 2007; Ochoa-López et al., 2015), exhibiting stronger anti-
173 herbivore responses in comparison to older reproductive plants. Plants > 3 m in height not
174 included in our sampling represented < 5% of Bignonieae individuals per plot and were either
175 reproductive lianas that access the canopy or emergent trees above the canopy.
176 We carried out five surveys on each tagged plant in each plot: October 2012 (Survey
177 1), November-December 2012 (Surveys 2 and 3), February 2013 (Survey 4), and May 2013
178 (Survey 5). In Survey 1, we recorded D30cm, length of the main stem, number of leaves and
179 number of recently formed shoot nodes of each tagged plant. In subsequent Survey 2, 3 and
180 4, we also measured the number of recently formed shoot nodes in each tagged plant. In the
181 final survey (Survey 5), we measured again all plant descriptors measured in the survey 1.
182 We measured the D30cm of each plant and the length of the main plant stem with callipers and
183 a tape measure, respectively.
7 184 Ant sampling
185 Before characterizing the abundance of ants attending EFNs, we first checked
186 whether diurnal observations alone could be a good proxy of 24-hour ant visitation patterns.
187 On a subset of 20 randomly selected plants, we performed ant censuses between 08:00 AM to
188 10:00 PM. Approximately 80 % of the plants had the same ant species attending the EFNs
189 during the day and night (Table S1). It suggests that observations taken during the day can
190 represent the temporal patterns of ant-plant interactions over 24h. Therefore, we performed
191 our censuses in all tagged plants from each plot between 8:00 AM and 5:00 PM on each of
192 the five censuses.
193 In each of the five surveys, we inspected each tagged plant for 10 min to identify
194 whether ants were visiting the active EFNs. Some tagged plants were small, with only 10
195 leaves. In these cases, we reduced the period of observation by 5 min, as that was sufficient
196 to count all ants visiting the EFNs on the entire plant. We counted the total number of ants
197 patrolling the entire plant in four of the five surveys, but not in the first, in which only
198 presence-absence data were collected.
199 In Surveys 2, 3, 4 and 5, we used honey baits to estimate the ant activity on each plot
200 over time. To accomplish this, we added ~1 ml of 50 % water-diluted honey in baits available
201 in 2 ml Eppendorf tubes positioned on plants at 0.6-1.5 m height. This procedure allowed us
202 to estimate local ant activity with a standardized resource independent of the natural
203 availability of extrafloral nectar. We systematically distributed 10 honey baits (one per plant)
204 within each plot. We disposed one bait on the surface of five randomly chosen Bignonieae
205 plants lacking young tissues with active EFNs, and on the surface of the closest five non-
206 Bignonieae species lacking EFNs. We did not observe differences in the patterns of ant
207 attendance of baits located on Bignonieae and non-Bignonieae plants (Figure S1). Hence, we
208 did not consider it separately in our analyses. We also inspected the baited plants to make
8 209 sure that no other sugar resources that could interfere with ant activity were present (e.g.,
210 honeydew-secreting insects). We collected all ants visiting the baits after 80-90 min of
211 observation. Subsequently, all ants from surveys and baits were counted and identified in the
212 laboratory. Ant species were classified as dominant or subordinate based on Baccaro et al.
213 (2010). We categorized our ant species in these groups to assess the potential quality of anti-
214 herbivore defence provided by the ants tending plants over time. Dominant ant species
215 rapidly recruit workers that can displace other ant species from resources, allowing them to
216 monopolize food sources (Parr 2008). These species tend to be more aggressive towards
217 possible herbivores, being considered better bodyguards to plants bearing EFNs than the
218 subordinate ants (Xu & Cheng 2010).
219
220 Statistical analyses
221 To investigate the period of the year in which ant-plant interactions in the understory
222 are most frequent, we performed a generalized linear mixed model (GLMM) with the
223 proportion of tagged plants attended by ants in each plot as our response variable, and the
224 survey month as the fixed categorical factor. We also explored the temporal pattern of young
225 tissue production and of ant activity using additional GLMMs. Response variables in those
226 cases were the proportion of plants with young tissues and the proportion of honey baits
227 occupied by ants in each survey, respectively. In all analyses, we used the binomial error
228 distribution in which the response variables were treated as the number of failures (i.e., plants
229 not attended by ants) and successes (i.e., plants attended by ants) as a two-vector response
230 variable, as suggested by Crawley (2007). Plot identity was included in the models as a
231 random variable. Temporal patterns would be consistent with our hypothesis if wetter
232 periods, in which ant attendance to EFNs is higher matched the periods in which plants invest
233 more in the production of new tissues bearing extrafloral nectaries.
9 234 Since survey date cannot be considered the causal factor explaining variation in ant-
235 plant interactions throughout the year, we also constructed a structural model expressed as a
236 direct acyclic graph to relate our three variables (i.e., proportion of plants with young tissues,
237 proportion of honey baits occupied by ants, and proportion of plants tended by ants per plot at
238 each survey date) to seasonal variation of precipitation. The multivariate causal model
239 according to our hypothesis incorporates precipitation, potentially determining local ant
240 availability, and/or the number of plants producing new tissues but does not directly
241 incorporate ant-plant interactions. Precipitation was expressed as the proportion of days with
242 rain in the month before each survey. We used this measure of water availability because it
243 describes more accurately how rainfall is distributed over the days before each sampling than
244 measures of total accumulation of rainfall. Independently of the precipitation pattern, our
245 main objective here was to evaluate if patterns of EFN attendance over the year would be
246 better explained by the variation in the proportion of plants with young tissues, variation in
247 the ant activity, or both.
248 Our structural model (Fig. 2, upper panel) was an alternative confirmatory path
249 analysis consisting of a generalization of Shipley’s d-sep test that could incorporate
250 hierarchical structures (e.g., resampling throughout time) and variables with different
251 sampling distributions (Shipley, 2009). Path analyses were performed in five steps, as
252 suggested by Shipley (2009): (i) construct hypothetical causal relationships between variables
253 in the form of a directed acyclic graph; (ii) list each pair of variables without a direct link on
254 the acyclic graph; (iii) define a full set of the k independence claims [i.e. (Xi, Xj)|{Z}] that
255 formed the basis set BU; (iv) for each element in this basic set, obtain the probability, pk, that
256 the pair (Xi, Xj) is statistically independent conditional on the variables {Z}; (v) combine the
257 k probabilities (pi) using the C equation. The C value was compared to a chi-squared
258 distribution with 2k degrees of freedom. We rejected the causal models whenever the C value
10 259 was unlikely to have occurred by chance. After confirming or rejecting the hypothetical
260 causal model, the path coefficients were obtained by regressing each variable within each of
261 its direct causes as detailed in Shipley (2009).
262 Considering each plant individually, we also asked if plants with a higher numbers of
263 recently formed shoot nodes (and thus more young tissues per plant) had a higher probability
264 of being attended by ants. We performed a generalized linear mixed model (GLMM) with a
265 binomial distribution (logit link function) including all individual plants together. We
266 modelled the presence/absence of ants on EFNs as the response variable, including the
267 number of recently formed shoot nodes per plant and surveys as fixed factors. We also
268 included the identity of the species and ‘individual plants within plots’ as random terms in
269 this model. The random variable ‘individual plants within plots’ explicitly described two
270 aspects of our sampling design: the spatial aggregation of plants per plot and the repeated
271 measurements of plants across surveys.
272 In the second set of analyses, to test whether the probability of ant attendance as a
273 function of the number of recently formed shoot nodes varies across Bignonieae species, we
274 applied a similar GLMM model, adding the categorical variable plant species as fixed factors
275 jointly with the number of recently formed shoot nodes. We included in the model the
276 individual plants within plots as a random variable. In these analyses, we only included the
277 12 species with more than 10 individuals across all 28 plots. Alternatively, we replaced the
278 identity of the plant species in the GLMM by a mean descriptor of the EFN phenotype of
279 each species as a fixed factor. In this case, to quantitatively describe differences of EFNs per
280 shoot node among species, we characterized the secretory area of EFNs (mm2), standardizing
281 the sampling effort to a single shoot node per plant in all cases. First, we estimated the
282 secretory area of EFNs quantifying the number and size of EFNs in each cluster and the
283 number of clusters per shoot node in at least five individual plants per species (see Nogueira
11 284 et al., 2012b for details about clustered EFNs in Bignonieae species). Then, we multiplied the
285 total area occupied by each EFN by the number of EFNs in one cluster and added up the total
286 secretory area by the number of clusters per shoot node (Table S2).
287 To explicitly describe differences in the relationship between young tissue availability
288 and ant attendance between species, we also performed individual analyses for each of the 12
289 most abundant Bignonieae species, using the same GLMM models described above
290 (binomial distribution and logit link function). Among them, four species produced only a
291 single recently formed shoot node at a time, showing low variation in the abundance of
292 young tissues among individuals, which could compromise the fitting of our GLM models.
293 Therefore, for each plant species, we also performed a binomial proportion test, evaluating if
294 proportions of plants attended by ants were the same between plants with and without young
295 tissues. In this analysis, we only considered whether plants did or did not have young tissues
296 in each period of the survey. We applied the Bonferroni correction in the binomial tests
297 owing to the false replication in our raw dataset. The binomial tests were considered
298 significant if p values were equal to or smaller than 0.01 (µ = 0.05/5 surveys).
299 All statistical analyses were performed in R, v. 3.4 (R Development Core Team,
300 2009) with standard and additional packages, as follows: glmmADMB (Bolker et al., 2012),
301 lme4 (Bates et al., 2016), nlme (Pinheiro et al., 2016), vegan (Oksanen et al., 2016), MASS
302 (Ripley et al., 2015), and glmmML (Broström 2018). If the frequency of ant-plant interaction
303 is higher in the wettest periods, we expected a higher proportion of Bignonieae plants
304 attended by ants in February, our wettest survey. Additionally, if temporal patterns of ant-
305 plant interaction are driven by plant production of new tissues or ant activity, we expected
306 that the proportion of plants attended by ants increases in periods in which the proportion of
307 plants producing new plant tissues and/or the proportion of honey baits occupied by ants also
308 increases. Our expectations included the indirect positive effect of precipitation, and the
12 309 direct positive effects of young plant tissues and ant density, on ant-plant interactions.
310 Finally, if the probability of an individual plant being attended by ants can be explained by
311 the amount of young tissues bearing active EFNs, we expected that the probability of ant
312 attendance increases as the number of recently formed shoot nodes also increases. Similarly,
313 if patterns of ant-plant interaction are also mediated by EFN secretory activity in each
314 recently produced plant module, we expected that Bignonieae species with a higher secretory
315 area of EFNs per shoot node should have a higher probability of ant attendance.
316
317 RESULTS
318 We tagged 431 understory plants from 22 species of Bignonieae across 28 plots (3-31
319 individuals per plot) (Table S3). Averaging across surveys, 10 ± 5% (mean ± SD) of the
320 tagged plants were attended by ants in each plot, reaching a maximum of 14% in the October
321 survey. A total of 33 species of ants distributed among 19 genera were recorded. Overall,
322 46.5% (range 39-61 % across surveys) of Bignonieae plants and 33% (29-38% across
323 surveys) of the ant-occupied honey baits were occupied by six ant species: Crematogaster
324 brasiliensis, C. limata, C. tenuicula, Ochetomyrmex semipolitus, Pheidole biconstricta and
325 Wasmannia auropunctata.
326
327 Seasonal patterns of ant-plant interactions and the role of young plant tissues bearing
328 EFNs
329 The proportion of plants attended by ants changed over time, showing a sharp decline
330 during the peak of the wet season (Fig. 3). Similarly, the proportion of plants with recently
331 formed shoot nodes (young tissue) was lower during the peak of the wet season (February: 17
332 ± 8%) compared to all drier months (42 ± 20%) (Table 1). The proportion of baits occupied
333 by ants over the survey period exhibited a similar pattern, with the proportion of baits
13 334 occupied by ants being lower at the peak of the wet season (February: 30 ± 15%) compared
335 to drier months (52 ± 8%).
336 Precipitation had an indirect and negative relationship with ant-plant interactions at
337 the plot level, mediated by the decrease of the proportion of plants with young tissues in
338 wetter periods (Fig. 2; Table S5). Precipitation before each survey was negatively related to
339 the proportion of plants with young tissues (Z = -6.96; p < 0.001) and to the proportion of
340 baits occupied by ants (Z = -4.18; p < 0.001) (Fig. 2; Table S4 and Table S5). Additionally,
341 plots with a higher proportion of plants with young tissues had, on average, a higher
342 proportion of plants attended by ants (Z = 8.73; p < 0.001). In contrast, the proportion of baits
343 occupied by ants was unrelated to the proportion of plants attended by ants in each plot (Z =
344 1.89; p = 0.064) (Fig. 4).
345
346 Variation in the amount of young plant tissues bearing EFNs explaining ant attendance
347 At the individual plant scale, the probability of the plants being attended by ants was
348 positively related to the quantity of young tissues per plant (solid line in Fig. 5; Z = 4.9; p <
349 0.001). This relationship varied marginally across surveys (different dashed lines in Fig. 5;
350 Survey effect: z= -0.6; p = 0.318). In general, plants with two or more recently formed shoot
351 nodes were more likely to be attended by ants than were plants bearing only one recently
352 formed shoot node. Plants with shoots at the initial stage of elongation (i.e., possessing very
353 little young tissue) exhibited more variable pattern of ant attendance, while plants lacking
354 young tissues bearing EFNs were almost never occupied by ants (< 1%) (Fig. 5). However,
355 ant attendance was not always positively related to the number of recently formed shoot
356 nodes across Bignonieae species (Fig. 6; Table 2, first model).
357 Species with higher EFN secretory area per node had a higher probability of being
358 attended by ants than did species with smaller secretory areas (Fig. 1S; Table 2, second
14 359 model). However, more of the variation in the occurrence of ant-plant interactions across
360 Bignonieae species was explained by the quantity of young tissues than by variation in the
361 EFN secretory area per shoot node (Fig. S2; Table 2). In six species (four species of
362 Adenocalymma, one species of Fridericia, and Callichlamys latifolia), ant attendance was not
363 predicted by the number or by the presence of recently formed shoot nodes on the plants (Fig.
364 6). On average, these species had less young tissue and a smaller EFN secretory area per
365 node, characterizing a group of plants that were almost never attended by ants. The exception
366 was Adenocalymma tanaeciicarpum, which had a large secretory area, but few recently
367 formed shoot nodes. For this species, we only detected a relationship between ant attendance
368 and the amount of young tissue in the binomial proportion test (barplot in Fig. 6). In contrast,
369 in the other six species (three species of Adenocalymma, one Anemopaegma, one Fridericia,
370 and one Pachyptera), ant attendance was positively related to the number of recently formed
371 shoot nodes (Fig. 6). On average, these species had the highest number of recently formed
372 shoot nodes per plant and a higher EFN secretory area in each shoot node (Fig. 6). We
373 detected no direct relationship between the amount of young tissue per plant and the
374 secretory area of EFNs in each shoot node estimated per species (Z = 0.55; p = 0.58).
375
376 DISCUSSION
377 Our results suggest that the proportion of plants attended by ants in the understory at
378 one tropical rainforest site is indirectly and negatively related to the precipitation over time.
379 This pattern is the opposite of the pattern proposed by our hypothesis and of the temporal
380 pattern of ant-plant interaction observed in other ecosystems (e.g. Dias-Castelazo & Rico
381 Gray, 2004, Belchior et al., 2016). This unexpected result was primarily explained by
382 seasonal differences in the availability of young EFN-bearing tissues, rather than by variation
383 in the local activity of ant consumers. Likewise, differences in the number of recently formed
15 384 shoot nodes (young tissues) at the plant level were the primary driver of variation in the
385 probability of ant attendance among individual plants and species. The variation in the EFN
386 secretory area per plant node secondarily explained patterns of ant attendance across
387 Bignonieae species. Altogether, our results reinforce the greater importance of plant
388 phenology and the availability of new plant tissues than the EFN phenotype per se. Below,
389 we discuss the importance of young EFN-bearing tissues for the temporal dynamics of ant-
390 plant protective mutualisms in an understory tropical rainforest at both the community and
391 individual plant scales.
392
393 Temporal variation of ant attendance on EFNs throughout the year
394 Our results clearly show that ant attendance at EFN-bearing plants was higher in the
395 drier months of the year at the Ducke Reserve. This finding contrasts with the temporal
396 patterns of ant-plant interactions described in other studies, in which ant attendance at EFN-
397 bearing plants is more frequent in the wet season in tropical (e.g., Rico-Gray, 1993; Dias-
398 Castelazo & Rico Gray, 2004; Kersch & Fonseca, 2005; Dutra et al., 2006; Lange et al.,
399 2013; Belchior et al., 2016) and subtropical sites (e.g., Aranda-Rickert et al., 2014). It is
400 tempting to conclude that water availability is the main abiotic factor driving temporal
401 patterns at previously studied sites, whereas ant-plant interactions at our study site are less
402 dependent on this driver. However, our results also showed that drier months had the highest
403 proportion of plants producing EFN-bearing tissues, as well as the highest local ant activity.
404 Therefore, it is possible that the mismatch between the seasonal patterns of ant-plant
405 interactions described here from what has been reported elsewhere primarily reflect
406 ecosystem-specific variation in temporal patterns of bud activation and plant phenology.
407 Water supply is weakly seasonal in most wet tropical forests, including the Reserva
408 Ducke, and covaries temporally with other abiotic factors that exhibit more marked variation
16 409 over the year and that directly influence plant phenology. Light irradiance (Saleska et al.,
410 2016), for example, is a limiting factor for bud activation and shoot elongation in Amazonian
411 rainforest plants (Fenner, 1998), especially for the ones occurring in the light-limited
412 understory. In the transition between wet and dry season, tree species from forest canopies
413 replace many of their leaves (Wright and Van Schaik, 1994), creating a gradient of light
414 availability in the forest understory. Furthermore, drier seasons generally have a lower cloud
415 cover. Thus, in drier periods of the year, a higher light level reaches the forest understory,
416 favouring leaf production in recently formed shoot nodes (Aide 1988). This increase in light
417 availability during the driest period of the year can act as a signal for the activation of shoot
418 apical meristems (Fenner, 1998), increasing photosynthetic plant rates (Wu et al., 2016) and
419 starting biochemical pathways controlling nectar secretion in EFN-bearing plant species (e.g.,
420 Radhika et al., 2010). Cumulatively, these aspects might explain the higher proportion of
421 plants with young tissues bearing active EFNs during the driest months in our study site,
422 creating a favourable temporal window in which associations between ants and EFN-bearing
423 plants can develop. Therefore, the same factors that drive the production of new tissues
424 bearing EFNs might indirectly regulate the patterns of ant attendance to EFNs over time.
425
426 Abundance of young EFN-bearing tissues determine patterns of ant attendance
427 In our study, both the number of recently formed shoot nodes and local ant activity
428 increased in the dry season. Theoretically, any increasing in these two factors are expected to
429 equally increase the probability of EFN-bearing plants attendance by ants. However, our
430 confirmatory path analyses showed that the proportions of EFN-bearing plants attended by
431 ants over time were only explained by the proportion of plants with young tissues per plot.
432 The abundance of young tissues not only explained temporal patterns of ant-plant interaction
433 at the community level, but also the probability of ant attendance at the individual plant level.
17 434 Individual plants with a higher number of recently formed shoot nodes showed a higher
435 probability of ant attendance. Therefore, ant attendance to EFN-bearing plants was consistent
436 with our initial hypothesis, reinforcing the importance of the phenological stage of plants
437 tended by ants (Bentley, 1977, Rico-Gray & Oliveira, 2007).
438 Many studies have shown that secretion of extrafloral nectar is greatest during periods
439 of rapid vegetative growth, in which new leaves bearing EFNs are produced in flushes (Elias,
440 1983; Koptur, 1992; Rico-Gray & Oliveira, 2007). Bentley (1977) critically stressed that
441 active extrafloral nectaries secreting sugar-rich resources are found on younger portions of
442 the plant, usually above the first 5-10 recently formed shoot nodes below an actively growing
443 meristem. Therefore, during periods in which EFN-bearing plants are investing more in the
444 production of new tissues, ants will be supplied with more extrafloral nectar. Although this
445 pattern can be modified by selection decoupling EFN activity and plant tissue age (e.g.,
446 McKey 1984), much of the temporal pattern of interactions with plant bodyguards (mostly
447 ants) is explained by investment in new plant tissues bearing secretory EFNs (Dreisig, 2000;
448 Heil et al., 2000; Rico-Gray & Oliveira, 2007). This increase in the uptake of extrafloral
449 nectar is expected to fuel ant activities (Aranda-Rickert et al., 2014), boosting both ants
450 patrolling on the plant surface, as well as in the number of EFN-bearing plants locally
451 explored by ants (e.g., Ness et al., 2009). Therefore, it is likely that the increased activity of
452 ant species foraging in the vegetation during the driest months is a response of the ant
453 community to the increasing availability of extrafloral nectar produced on the young plant
454 tissues, also more abundant in this period of the year.
455 In extensive surveys of ant-plant interactions, plant species are commonly categorized
456 according to the presence or absence of EFNs (e.g. Dattilo et al., 2014). However, our results
457 highlight that most EFN-bearing plant species behave like plants without EFNs, depending
458 much more on the phenological status of plants than on the EFN phenotype per se. EFN-
18 459 bearing plants producing young tissues not only become predictable sources of food to ants
460 but sources of energy that increase foraging success and constancy of ant attendance
461 (Nogueira et al., 2015). Consequently, EFN-bearing plants supporting more recently formed
462 shoot nodes simultaneously are more attractive to ants, being more likely to be tended by
463 ants. In fact, the six Bignonieae species for which ant attendance probabilities were unrelated
464 to the amount of young tissues were the same species that had the fewest recently formed
465 shoot nodes over our five surveys. In these species, the amount of extrafloral nectar offered to
466 ants should be low, compromising the quality of the plant as partners to ants.
467 In our study, using a conservative estimate of dominant ant species based on Baccaro
468 et al. (2010), six dominant aggressive ant species occupied about half of the plants producing
469 young EFN-bearing tissues, independent of the sampling period. As dominant ants are more
470 aggressive towards potential herbivores (Buckley & Gullan, 1991; Xu & Chen, 2010), it
471 means that most Bignonieae plants occurring in our study site is interacting with high-quality
472 ant partners. This association has the potential to intensify biotic defence on young tissues
473 that are preferred by herbivores (Coley 1980), even though the herbivory pressure is likely to
474 be slightly lower in the dry season (Aide 1988; Wolda 1988; Coley & Barone 1996). In this
475 scenario, it is tempting to say that the temporal match between the increase in plant
476 attendance by ants and the production of new tissues benefit the plants, as such young and
477 vulnerable new tissues face a higher probability to be tended by more efficient ant defenders.
478 Such implication should be interpreted with caution, mainly because the direction and
479 magnitude of ant effect on plant fitness depend on the herbivore pressure. Indeed, EFN-
480 bearing plants interacting preferentially with dominant ant species in an arid ecosystem in
481 Brazil had lower reproductive success than plants preferentially tended by subordinate, less
482 aggressive ant species (Melati & Leal, 2018). Dominant ant species forage in groups,
483 recruiting more workers on plants (Cerdá et al., 2013), and impose high ecological and
19 484 energetic costs to the plant partner compensated when plants are under high herbivore
485 pressure (Melati & Leal, 2018). Therefore, any discussion about the net effect of ant-plant
486 protection mutualisms is only possible with more information about how the herbivore
487 pressure varies on Bignonieae plants over the year.
488
489 Concluding remarks
490 Temporal patterns of ant attendance at EFNs coincide with the seasonal appearance of
491 young EFN-bearing tissues, on which EFNs appear to be more active. Therefore, ant
492 attendance at both the community and individual levels over time can be driven by the
493 amount of young tissues bearing EFNs. The contrasting temporal patterns of ant-plant
494 interactions across seasonal and aseasonal tropical sites can be explained by the variation in
495 factors triggering the changes into the phenological stages of EFN-bearing plants in different
496 habitats. In tropical seasonal environments, most plant species produce new tissues at the
497 onset of rainy seasons (Levings, 1983), when water availability to plant growth is higher.
498 However, in the understory of tropical rainforest, plants have available a relatively high
499 supply of water all over the year. Then, plant growth season in aseasonal environments
500 should be triggered by other environmental factors, such the light that is more abundant in the
501 understory of tropical forests in the dry season. Differences in ant attendance among species
502 in the understory were only poorly explained by the secretory area of EFNs per shoot node,
503 highlighting the remarkable role of plant phenology and the availability of young plant tissue
504 in explaining the establishment of ant-plant interactions. Plant phenology could be especially
505 important and even critical in seasonally drier areas where growing seasons are shorter, and
506 EFN-bearing plants are more common than in wet environments (Leal & Peixoto, 2017). In
507 such environments, EFN-bearing plants should experience not only a more marked temporal
20 508 pattern of ant attendance but also a narrower window in which ants can more efficiently
509 protect plants.
510
511 ACKNOWLEDGEMENTS
512 We are grateful to Carlos André Nogueira and Caian Gerolamo for assistance during
513 fieldwork, and Emerson C. Merkel helping us to sort ants in the laboratory. Otávio G. Silva
514 provided useful comments on earlier versions of this manuscript. We are also grateful to
515 INMET (Instituto Nacional de Metereologia) for providing meteorological data from Manaus
516 station (http://www.inmet.gov.br/). This work was partially funded by two fellowships from
517 the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP 2012/02110-5 and
518 2013/04591-3) to A.N. and a Postdoctoral Fellowship from the Conselho Nacional de Apoio
519 à Pesquisa (CNPq 234000/2014-7) to A.N. Additional funds were provided to L.G.L. by the
520 Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) through a Pq-1B
521 grant (310871/2017-4).
522
523 AUTHORS’ CONTRIBUTIONS
524 A.N. and F.B.B. conceived the ideas, experimental design and collected the data; A.N.
525 analysed the data and led the writing of the manuscript. All authors contributed to the writing
526 and gave final approval for publication.
527
528 DATA ACCESSIBILITY
529 Data will be available from the Dryad Digital Repository:
530
531 ORCID
532 Anselmo Nogueira https://orcid.org/0000-0002-8232-4636
21 533 Fabricio B. Baccaro https://orcid.org/0000-0003-4747-1857
534 Laura C. Leal https://orcid.org/0000-0003-1570-8901
535 Pedro J. Rey https://orcid.org/0000-0001-5550-0393
536 Lúcia G. Lohmann https://orcid.org/0000-0003-4960-0587
537 Judith L. Bronstein https://orcid.org/0000-0001-9214-1406
538
539 LITERATURE CITED
540 Aide, T. M. (1988). Herbivory as a selective agent on the timing of leaf production in a
541 tropical understory community. Nature, 336, 574-575.
542 Aranda-Rickert, A., Diez, P., & Marazzi, B. (2014). Extrafloral nectar fuels ant life in
543 deserts. AOB Plants, 6, 1–14.
544 Baccaro, F. B., Ketelhut, S. M., & Morais, J. W. (2010). Resource distribution and soil
545 moisture content can regulate bait control in an ant assemblage in Central Amazonian
546 forest. Austral Ecology, 35, 274–281.
547 Bates, D., Maechler, M., Bolker, B., Walker, S., Christensen, R. H. B., Singmann, H., Dai,
548 B., Grothendieck, G., & Green, P. (2016). lme4: Linear mixed-effects models using
549 ‘Eigen’ and S4. R package version, 1, 1–11.
550 Belchior, C., Sendoya, S. F., & Del-Claro, K. (2016). Temporal variation in the abundance
551 and richness of foliage-dwelling ants mediated by extrafloral nectar. PLOS ONE, 11,
552 e0158283.
553 Bentley, B. L. (1977). Extrafloral nectaries and protection by pugnacious bodyguards. Annual
554 Review of Ecology and Systematics, 8, 407–427.
555 Boege, K., Dirzo, R., Siemens, D. & Brown, P. (2007). Ontogenetic switches from plant
556 resistance to tolerance: minimizing costs with age? Ecology Letters, 10, 177-187.
22 557 Bolker, B., Skaug, H., Magnusson, A., & Nielsen, A. (2012). Getting started with the
558 glmmADMB package. R package, 1–12.
559 Bronstein, J. L. (1998). The contribution of ant-plant protection studies to our understanding
560 of mutualism. Biotropica, 30, 150-161.
561 Bronstein, J. L. (2015). The study of mutualism. In J. L. Bronstein (Ed.), Mutualism (pp. 3–
562 19). New York, NY: Oxford University Press.
563 Broström, G. (2018). glmmML: Generalized linear models with clustering. R package version
564 1.0.3, 1–14.
565 Buckley, R., & Gullan, P. (1991). More aggressive ant species Hymenoptera: Formicidae
566 provide better protection for soft scales and mealybugs Hommoptera: Coccidae,
567 Pseudococcidae. Biotropica, 23, 282–286.
568 Burns, K. C. (2004). Scale and macroecological patterns in seed dispersal mutualisms. Global
569 Ecology and Biogeography, 13, 289-293.
570 Carnicer, J., Jordano, P., & Melián, C. J. (2009). The temporal dynamics of resource use by
571 frugivorous birds: a network approach. Ecology, 90, 1958-1970.
572 Cerdá, X., Arnan, X., Retana, J. (2013). Is competition a significant hallmark of ant
573 (Hymenoptera: Formicidae) ecology? Myrmecological News, 18, 131–47.
574 Chamberlain, S. A., Bronstein, J. L., & Rudgers, J. A. (2014). How context dependent are
575 species interactions? Ecology Letters, 17, 881–890.
576 Chaves, M. M., Pereira, J. S., Maroco, J., Rodrigues, M. L., Ricardo, C. P. P., Osório, M. L.,
577 Carvalho, I., Faria, T., & Pinheiro, C. (2002). How plants cope with water stress in the
578 field. Photosynthesis and growth. Annals of Botany, 89, 907–916.
579 Chauvel, A., Lucas, Y., & Boulet, R. (1987). On the genesis of the soil mantle of the region
580 of Manaus, Central Amazonia, Brazil. In: multi-author review. The dynamics of the
581 amazonian terra firme forest. Experientia, 43, 234–241.
23 582 Coley, P. D. (1980). Effects of leaf age and plant life history patterns on herbivory. Nature,
583 284, 545–546.
584 Coley, P. D., & Barone, J. A. (1996). Herbivory and plant defenses in tropical forests. Annual
585 Review of Ecology and Systematics, 27, 305–335.
586 Crawley, M. J. (2007). The R Book. Wiley, John Wiley & Sons, UK, 951 p.
587 Dattilo, W., Marquitti, F. M. D., Guimarães, & P. R., Izzo, T. J. (2014). The structure of ant-
588 plant ecological networks: Is abundance enough? Ecology, 95, 475-485.
589 Díaz-Castelazo, C., & Rico-Gray, V. (2004). Extrafloral nectary-mediated ant-plant
590 interactions in the coastal vegetation of Veracruz, Mexico: Richness, occurrence,
591 seasonality, and ant foraging patterns. Ecoscience, 11, 472–481.
592 Di Giusto, B., Marie-Charlotte, A., Dounias, E., & McKey, D. B. (2001). Variation in the
593 effectiveness of biotic defence: the case of an opportunistic ant-plant protection
594 mutualism. Oecologia, 129, 367–375.
595 Dreisig, H. (2000). Defense by exploitation in the Florida carpenter ant, Camponotus
596 floridanus, at an extrafloral nectar resource. Behavioral Ecology and Sociobiology, 47,
597 274–279.
598 Dutra, H. P., Freitas, A. V. L., & Oliveira, P. S. (2006). Dual ant attraction in the Neotropical
599 shrub Urera baccifera (Urticaceae): the role of ant visitation to pearl bodies and fruits in
600 herbivore deterrence and leaf longevity. Functional Ecology, 20, 252–260.
601 Elias, T. S. (1983). Extrafloral nectaries: their structure and distribution. In: B. L. Bentley &
602 T. S. Elias (Eds.), The biology of nectaries (pp. 174-203). New York, NY: Columbia
603 University Press.
604 Fenner, M. (1998). The phenology of growth and reproduction in plants. Perspectives in
605 Plant Ecology, Evolution and Systematics, 1, 78–91.
24 606 Heath, K. D., Stock, A. J., & Stinchcombe, J. R. (2010). Mutualism variation in the
607 nodulation response to nitrate. Journal of Evolutionary Biology, 23, 2494-2500.
608 Heil, M., Fiala, B., Baumann, B., & Linsenmair, K. E. (2000). Temporal, spatial and biotic
609 variations in extrafloral nectar secretion by Macaranga tanarius. Functional Ecology,
610 14, 749–757.
611 Hoeksema, J. D., & Bruna, E. M. (2015). Context-dependent outcomes of mutualistic
612 interactions. In J. L. Bronstein (Ed.), Mutualism (pp. 181–202). New York, NY: Oxford
613 University Press.
614 Holland, J. N., Ness, J. H., Boyle, A., & Bronstein, J. L. (2005). Mutualisms as consumer-
615 resource interactions. In: P. Barbosa, Ed. Ecology of Predator-Prey Interactions. Oxford,
616 Oxford University Press, 17-33.
617 Horvitz, C. C., & Schemske, D. W. (1990). Spatiotemporal variation in insect mutualisms of
618 a neotropical herb. Ecology, 71, 1085-1097.
619 Kersch, M. F., & Fonseca, C. R. (2005). Abiotic factors and the conditional outcome of an
620 ant-plant mutualism. Ecology, 86, 2117–2126.
621 Koptur, S. (1992). Plants with extrafloral nectaries and ants in everglades habitats. The
622 Florida Entomologist, 75, 38-50.
623 Lange, D., Dáttilo, W., & Del-Claro, K. (2013). Influence of extrafloral nectary phenology on
624 ant-plant mutualistic networks in a neotropical savanna. Ecological Entomology, 38,
625 463–469.
626 Leal, L. C., & Peixoto, P. E. C. (2016). Decreasing water availability across the globe
627 improves the effectiveness of protective ant-plant mutualisms: a meta-analysis.
628 Biological Reviews, 92, 1785–1794.
25 629 Levings, S. C. (1983). Seasonal, annual, and among-site variation in the ground ant
630 community of a deciduous tropical forest: some causes of patchy species distributions.
631 Ecological Monographs, 53, 435-455.
632 Lohmann, L. G. (2006). Untangling the phylogeny of neotropical lianas Bignonieae,
633 Bignoniaceae. American Journal of Botany, 93, 304–318.
634 Marazzi, B., Bronstein, J. L. & Koptur, S. (2013). The diversity, ecology and evolution of
635 extrafloral nectaries: current perspectives and future challenges. Annals of Botany, 111,
636 1243-1250.
637 McKey, D. (1984). Interaction of the ant-plant Leonardoxa africana (Caesalpiniaceae) with
638 its obligate inhabitants in a rainforest in Cameroon. Biotropica, 16, 81-99.
639 Marques-Filho, A. D. O., dos Santos, H. M., & dos Santos, J. M. (1981). Estudos
640 climatológicos da Reserva Florestal Ducke-Manaus, AM, IV- Precipitação. Acta
641 Amazonica, 4, 759–768.
642 Melati B. G., Leal, L. C. (2018). Aggressive bodyguards are not always the best: preferential
643 interaction with more aggressive ant species reduces reproductive success of plant
644 bearing extrafloral nectaries. Plos One, 13, e0199764.
645 Ness, J. H., Morris, W. F., & Bronstein, J. L. (2009). For ant-protected plants, the best
646 defense is a hungry offense. Ecology, 10, 2823–2831.
647 Nogueira, A., Guimarães, E., Machado, S. R., & Lohmann, L. G. (2012a). Do extrafloral
648 nectaries present a defensive role against herbivores in two species of the family
649 Bignoniaceae in a Neotropical savanna? Plant Ecology, 213, 289–301.
650 Nogueira, A., Rey, P. J., & Lohmann, L. G. (2012b). Evolution of extrafloral nectaries:
651 adaptive process and selective regime changes from forest to savanna. Journal of
652 Evolutionary Biology, 25, 2325–2340.
26 653 Nogueira, A., El Ottra, J. H. L., Guimarães, E., Machado, S. R., & Lohmann, L. G. (2013).
654 Trichome structure and evolution in Neotropical lianas. Annals of Botany, 7, 1331–1350.
655 Nogueira, A., Rey, P. J., Alcántara, J. M., Feitosa, R. M., & Lohmann, L. G. (2015).
656 Geographic mosaic of plant evolution: extrafloral nectary variation mediated by ant and
657 herbivore assemblages. PLoS ONE, 10, e0123806.
658 Ochoa-López, S., Villamil, N., Zedillo-Avelleyra, P., Boege, K. (2015). Plant defence as a
659 complex and changing phenotype throughout ontogeny. Annals of Botany, 116, 797-806.
660 Olesen, J. M., Bascompte, J., Elberling, H., & Jordano, P. (2008). Temporal dynamics in a
661 pollination network. Ecology, 89, 1573-1582.
662 Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. R., O’Hara, R. B.,
663 Simpson, G. L., Solymos, P., Stevens, M. H. H., & Wagner, H. (2016). Vegan:
664 Community Ecology Package. R package version 2, 3–4.
665 Parr, C. L. (2008). Dominant ants can control assemblage species richness in a South African
666 savanna. Journal of Animal Ecology, 77, 1191-1198.
667 Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., & R Core Team. (2016). nlme: Linear and
668 nonlinear mixed effect models. R package version 3.1–126.
669 R Development Core Team. (2009). R: A language and environment for statistical
670 computing. R Foundation for Statistical Computing. Available at URL: http://www.R-
671 project.org.
672 Radhika, V., Kost, C., Mithöfer, A., & Boland, W. (2010). Regulation of extrafloral nectar
673 secretion by jasmonates in lima beans is light dependent. PNAS, 107, 17228–17233.
674 Ribeiro, J. E. L. S., Hopkins, M. J. G., Vicentini, A., Sothers, C. A., Costa, M. A. S., Brito, J.
675 M., Souza, M. A. D., Martins, L. H. P., Lohmann, L. G., Assunção, P. A. C. L., Pereira,
676 E. C., Silva, C. F., Mesquita, M. R. & Procópio, L. C. (1999). Flora da Reserva Ducke:
27 677 guia de identificação das plantas vasculares de uma floresta de terra-firme na Amazônia
678 Central. Manaus-AM: DFID.
679 Rico-Gray, V. (1993). Use of plant-derived food resources by ants in the dry tropical
680 lowlands of coastal Veracruz, Mexico. Biotropica, 25, 301–315.
681 Rico-Gray, V., & Oliveira, P. S. (2007). The ecology and evolution of ant-plant interactions.
682 Chicago: The University of Chicago Press.
683 Ripley, B., Venables, B., Bates, D. M., Hornik, K., Gebhardt, A., & Firth, D. (2015). MASS:
684 Support functions and datasets for Venables and Ripley's MASS. R package version 7. 3–
685 45.
686 Saleska, S. R., Wu, J., Guan, K., Araujo, A. C., Huete, A., Nobre, A. D., & Restrepo-Coupe,
687 N. (2016). Dry-season greening of Amazon forests. Nature, 531, E4–E5.
688 Seibert, R. J. (1948). The use of glands in a taxonomic consideration of the family
689 Bignoniaceae. Annals of the Missouri Botanical Garden, 35, 123–137.
690 Shipley, B. (2009). Confirmatory path analysis in a generalized multilevel context. Ecology,
691 90, 363-368.
692 Wäckers, F. L., & Bonifay, C. (2004). How to be sweet? Extrafloral nectar allocation by
693 gossypium hirsutum fits optimal defense theory predictions. Ecology, 85, 1512–1518.
694 Wolda, H, 1988. Insect seasonality: Why? Annual Review of Ecology and Systematics, 19, 1-
695 18.
696 Wright, S. J., & Van Schaik, C. P. (1994). Light and the phenology of tropical trees. The
697 American Naturalist, 143, 192–199.
698 Wu, J., Albert, L. P., Lopes, A. P., Restrepo-Coupe, N., Hayek, M., Wiedemann, K. T., Guan,
699 K., Stark, S. C., Christoffersen, B., Prohaska, N., Tavares, J. V., Marostica, S.,
700 Kobayashi, H., Ferreira, M. L., Campos, K. S., Silva, R., Brando, P. M., Dye, D. G.,
701 Huxman, T. E., Huete, A. R., Nelson, B. W., & Saleska, S. R. (2016). Leaf development
28 702 and demography explain photosynthetic seasonality in Amazon evergreen forests.
703 Science, 351, 972–976.
704 Xu, F. F., & Chen, J. (2010). Competition hierarchy and plant defense in a guild of ants on
705 tropical Passiflora. Insectes Sociaux, 57, 343–349.
706 Zeppel, M. J. B., Wilks, J. V., & Lewis, J. D. (2014). Impacts of extreme precipitation and
707 seasonal changes in precipitation on plants. Biogeosciences, 11, 3083–3093.
708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
29 734 TABLES 735 736 Table 1 – Summary of the GLMM models evaluating the relationship between the proportion 737 of plants with young tissues, ant-occupied baits, and plants attended by ants over time. We 738 performed the GLMMs using binomial error distribution and plots identity as a random 739 variable. Response variables were treated as the number of failures (plants without) and 740 successes (plants with young tissues) in a two-vector response variable.
Response Fixed Estimated coefficient SE F Nplots p Graph variable (error factors in (December versus distribution) GLMM February)
Proportion of plants with young tissues Time -1.49 0.21 22.6 28 <0.001 -
(binomial)
Proportion of ant Time -0.93 0.19 13.1 28 <0.001 - baits occupied (binomial)
Proportion of plants Time -1.88 0.32 13.4 28 <0.001 Fig. 3B attended by ants (binomial)
741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
30 759 Table 2 – Summary of the generalized linear mixed models (GLMM) testing the relationship 760 between the number of recently formed shoot nodes (the amount of young tissues) in the plants 761 and the probability of plant attendance by ants in Bignonieae. The two first models considered 762 all individual plants and surveys, including the number of recently formed shoot nodes and 763 plant species as fixed factors, and the occurrence of ants on plants as the binary response 764 variable. Graphical representation of the relationship between number of recently formed shoot 765 nodes and the probability of plant attendance by ants per Bignonieae species can be viewed in 766 Figure 6.
Plant species Random Fixed factors Estimated z N p variables coefficient ± SE
All dataset (first model) Plot : individuals Number of recently formed 2.7 ± 0.3 9.7 1743 <0.001 shoot nodes Species - - 1743 <0.001
All dataset (second model) Plot : individuals Number of recently formed 4.6 ± 0.7 4.9 1743 <0.001 shoot nodes Secretory area of EFNs 0.9 ± 0.4 2.1 1743 0.036
Adenocalymma adenophorum Plot : individuals Number of recently formed - - - - shoot nodes
Adenocalymma bracteosum Plot : individuals Number of recently formed - - - - shoot nodes
Adenocalymma flaviflorum Plot : individuals Number of recently formed - - - - shoot nodes
Adenocalymma longilineum Plot : individuals Number of recently formed 5.9 ± 1.3 4.6 357 <0.001 shoot nodes
Adenocalymma moringifolium Plot : individuals Number of recently formed 4.1 ± 1.3 3.2 436 0.002 shoot nodes
Adenocalymma tanaeciicarpum Plot : individuals Number of recently formed 8.3 ± 6.9 1.2 55 0.231 shoot nodes
Adenocalymma validum Plot : individuals Number of recently formed 6.6 ± 2.0 3.2 436 0.001 shoot nodes
Anemopaegma robustum Plot : individuals Number of recently formed 17.9 ± 8.6 2.1 277 0.038 shoot nodes
Callichlamys latifolia Plot : individuals Number of recently formed - - - - shoot nodes
Fridericia nigrescens Plot : individuals Number of recently formed - - - - shoot nodes
Fridericia prancei Plot : individuals Number of recently formed 2.3 ± 1.2 1.9 68 0.049 shoot nodes
Pachyptera aromatica Plot : individuals Number of recently formed 14.9 ± 8.2 2.0 120 0.044 shoot nodes
767 *Species in which GLMM was prohibited because the excess of zeros in the response variable. 768
769
31 770 FIGURES
771
772 Figure 1 – Ant-plant interactions on young EFN-bearing tissues in different Bignonieae
773 species. Red and yellow arrows highlight ant attendance and nectar secretion on young
774 tissues, respectively. A: Sprouting branch with young tissues of a typical Bignonieae liana in
775 the Amazon rainforest with the first, second and third recently formed shoot node (plant
776 modules described in the Methods). B: Third recently formed shoot node with Camponotus
777 ants on active EFNs. C: First and second recently formed shoot node of Adenocalymma
32 778 tanaeciicarpum with active EFN cluster on the prophylls attended by Crematogaster ants. D:
779 Mature tissues in the shoot from the last growing season without active EFNs (inferior plant
780 part) and new tissues recently formed by the primary growth with active EFN cluster in the
781 prophylls (superior plant part) attended by Azteca ants. E: Typical EFN cluster on young
782 tissues accumulating nectar on the prophylls of axillary buds (two prophylls encompassing
783 each bud totalling four prophylls per shoot node). F: Typical EFN cluster on young tissues
784 accumulating nectar on the interpetiolar region of the stem (one EFN cluster in each plant
785 side totalling two clusters per shoot node). G-H: Pachyptera aromatica with active EFNs
786 clustered in the interpetiolar region on recently formed shoot nodes attended by
787 Crematogaster ants. I-J: Anemopaegma robustum with active EFNs clustered in prophylls
788 and young leaflets on recently formed shoot attended by Camponotus and Crematogaster
789 ants, respectively. K-M: Adenocalymma moringifolium with EFNs clustered in the prophylls
790 and young leaflets on recently formed shoot attended by Solenopsis ants.
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
33 806
807 Figure 2 – Confirmatory path analysis testing the direct and indirect effect of precipitation on
808 the proportion of plants with young tissues, ants on honey baits and plants attended by ants.
809 Precipitation directly decreases the number of plants with young tissues and the number of
810 honey baits occupied by ants (ant availability). Furthermore, precipitation indirectly
811 decreases ant attendance on EFNs mediated by the local change in the availability of plants
812 with young tissues. Solid lines indicate standardized coefficients with p £ 0.05. Dashed lines
813 indicate standardized coefficients with p £ 0.10. Statistics of each path coefficient are
814 detailed in Table S4.
815
816
817
818
34 819
820 821 Figure 3 – Temporal variation of precipitation (mm) and ant-plant interactions per plot from
822 August 2012 to July 2013. A: Monthly precipitation; dots in grey represent the monthly
823 precipitation between 2008 and 2011 (data from INMET); in black, the third order
824 polynomial describing the precipitation tendency of the current year under study. B:
825 Proportion of plants attended by ants showing higher average values in the dry season and
826 smaller values in the wet season. Statistics are detailed in Table 1.
827
828
829
830
831
832
833
35 834
835
836 Figure 4 – Partial regression describing the relationship between the proportion of plants
837 attended by ants and either (A) the proportion of plants with young tissues and (B) the
838 proportion of honey baits occupied by ants. Plots with a higher number of plants with young
839 tissues had more plants attended by ants, while ant availability, assessed here with the
840 experimental addition of honey baits in the field, was unrelated to plant attendance. Partial
841 regression plots show the expected effect of a variable when the other variable in the model
842 are statistically held constant.
843
844
36 845
846
847
848 Figure 5 – Predicted probabilities of ant attendance and the number of recently formed shoot
849 nodes (the amount of young tissues) per plant. Continuous black line describes the general
850 positive relationship between both variables without explicitly considering each sampling
851 period or plant species. Dashed lines describe marginal variations in this relationship between
852 surveys. Statistics are detailed in Table 2.
853
854
855
856
857
858
859
860
861
37 862
863
864
38 865 Figure 6 – Predicted probabilities of ant attendance and binomial proportion tests for 12
866 Bignonieae species with more than ten individuals sampled in the plots. Statistics are detailed
867 in Table 3. There is a wide variation of secretory area of EFNs per shoot node among
868 Bignonieae species following the homoplastic pattern of EFN evolution in this plant clade
869 (Nogueira et al., 2012b, 2013).
39