Leaf and flower consumption modulate the drinking behavior in a folivorous-frugivorous arboreal mammal

Water is vital for the survival of any species because of its key role in most physiological processes. However, little is known about the non-food-related water sources exploited by arboreal mammals, the seasonality of their drinking behavior and its potential drivers (including diet composition, temperature, and rainfall). We investigated this subject in 14 wild groups of brown howler monkeys (Alouatta guariba clamitans) inhabiting small, medium, and large Atlantic Forest fragments in southern Brazil. We found a wide variation in the mean rate of drinking among groups (range=0-16 records/day). Streams (44% of 1,258 records) and treeholes (26%) were the major types of water sources, followed by bromeliads in the canopy (16%), pools (11%), and rivers (3%). The type of source influenced whether howlers used a hand to access the water or not. Drinking tended to be evenly distributed throughout the year, except for a slightly lower number of records in the spring than in the other seasons, but it was unevenly distributed during the day. It increased in the afternoon in all groups, particularly during temperature peaks around 15:00 and 17:00. We found via generalized linear mixed modelling that the daily frequency of drinking was mainly influenced by flower (negatively) and leaf (positively) consumption, whereas fruit consumption, fragment size, rainfall, and mean ambient temperature played negligible roles. The influence of leaf consumption is compatible with the ‘metabolite detoxification hypothesis,’ which states that the processing of this fibrous food requires the ingestion of larger volumes of water to help in the detoxification/excretion of its metabolites. In sum, we found that irrespective of habitat size and climatic conditions, brown howlers seem to seek a positive water balance by complementing preformed and metabolic water with drinking water, even when it is associated with a high predation risk in terrestrial sources.

3 57 Introduction 58 59 Water is an essential chemical substance for all animals, not only because it represents a large 60 percentage of whole-body mass, but because it is the medium within which the chemical 61 reactions and physiological processes of the body take place [1][2][3]. This substance is involved in a 62 myriad of vital processes, such as secretion, absorption, and transport of macromolecules (e.g. 63 nutrients, hormones, metabolites, antibodies, and neurotransmitters), electrolyte homeostasis, 64 transmission of light and sound, and thermoregulation [2][3][4]. Therefore, water intake is essential 65 for animal health and survival, particularly in the case of terrestrial vertebrates [3,[5][6][7]. 66 In all terrestrial mammals, water inputs come from three major sources -water ingested 67 within consumed foods, metabolic water resulting from macronutrient oxidation, and water composition that influence their importance and reliability as water sources, thereby influencing 83 the animals' need to drink [9]. Herbivorous mammals inhabiting dry environments, such as desert 84 rodents and camelids, can reach water balance by relying on preformed (i.e. water in plant items) 85 and oxidation (i.e. metabolic water resulting from macronutrient oxidation) water during dry 86 periods [2,10]. In addition to these water sources, animals inhabiting wetter environments also 87 rely on another major source, drinking water [2,7,11]. Drinking is rare (e.g. giraffe, Giraffa 88 camelopardalis) or presumably nonexistent in mammals that rely on succulent diets [2]. Arboreal 89 folivores once believed to obtain all their water demands from food have been reported to drink  While ground-living species drink water from rarely-depletable sources (e.g. rivers, 93 streams, and lagoons), highly arboreal mammals depend on depletable arboreal reservoirs, such 94 as bromeliads and treeholes (primates [15][16][17][18]), or on short lasting rain water on tree branches 95 and leaves (koalas [14], sloths [19]). However, the exploitation of terrestrial water reservoirs by 96 these mammals tends to be rare because their vulnerability to predators likely increases when 97 they descend to the forest floor, as has been observed for other tropical primates [15,[20][21][22][23] [11]). These monkeys meet their 102 water needs primarily via preformed water [15,30], although they also drink from arboreal 103 reservoirs or, to a lesser extent, terrestrial sources [15][16][17]20]. the presence and reliability of water sources. Therefore, it is important to identify the factors that 123 modulate their drinking behavior to better understand how habitat patch size and spatial 124 restriction resulting from land use changes can affect their health and survival.

125
Here we investigate the drinking behavior in wild groups of brown howler monkeys (A. 126 guariba clamitans) inhabiting Atlantic Forest fragments in southern Brazil as models of 127 folivorous-frugivorous arboreal mammals. Specifically, we assess (i) the arboreal and terrestrial 6 128 water sources that these monkeys exploit and how they drink, (ii) the daily frequency and 129 seasonal distribution of drinking records, and (iii) the influence of fragment size, season, ambient 130 temperature, rainfall, and the contribution of fruits, leaves, and flowers to the diet on drinking. 131 We predicted that brown howlers would complement the preformed water obtained from their 132 diet with water from arboreal and terrestrial reservoirs, if available, because the availability of 133 fleshy fruits, flowers and young leaves vary seasonally in the study region [37]. We also 134 predicted a within-day increase in drinking in the afternoon in response to an increase in water 135 demands resulting from higher ambient temperatures and the daily water loss via digestion, 136 excretion, breathing, and sweating [1,8]. Finally, we predicted that diet composition, climatic 137 variables, and fragment size influence the frequency of drinking. While the TDH will receive 138 support if ambient temperature and rainfall are good predictors of the frequency of drinking, a 139 positive influence of leaf ingestion on water consumption will support the MDH.   (Table 1, Fig 1). We classified the fragments in three 155 size categories: small (<1 to 10 ha), medium (>10 to 100 ha) and large (>100 to 1,000 ha; sensu 156 [38]). Small and medium fragments in Porto Alegre (S1-S3 and M1) and Viamão (S4-S6; Figure   157 1) were surrounded by anthropogenic matrices comprised of small human settlements, pastures, Brazilian laws, CISM is impacted by a lower human pressure than the unprotected study sites.

166
The predominant vegetation in all study fragments is subtropical semideciduous forest.   (Table 1). 179 We followed brown howler monkey groups ranging from 4 to 12 individuals in each 180 fragment (n=116 individuals, Table 1). All groups in small fragments were well-habituated to 181 humans before study, while we habituated the groups inhabiting medium and large fragments 182 during two to three months prior to their respective monitoring. Whereas most groups inhabited a 183 single forest fragment, S1, S2, and S7 (hereafter named by the acronym of their respective 184 fragments) used more fragments. S1 ranged outside of its most used fragment for about 35% of  Behavioral data collection 191 We studied the diet and drinking behavior of the groups during periods ranging from 12 to 21 192 months ( Table 1)    Statistical analyses 222 We performed Chi-square tests for proportions to compare the proportions of drinking records 223 per water source and season in each study group using the 'prop.test' function of R. We 224 calculated these proportions by dividing the number of records for each water source (or season) 10 225 by the total number of records for each group during the entire study period. We did not compare 226 fragments or groups because of their sampling effort differences (i.e. the number of sampling 227 months, days, or days per month varied between the five study periods, Table 1). We used the 228 same procedure above to calculate and compare the proportion of drinking records in each hour 229 of the day in those fragments with >10 drinking records. When we found significant differences, 230 we compared the proportion of records in each class using post-hoc proportion contrasts via the R 231 function 'pairwise.prop.test' with a Bonferroni correction because of multiple comparisons of the 232 same data sets. 233 We performed generalized linear mixed-effects models GLMM to assess the influence of 234 the contribution of fruits, leaves, and flowers to the diet, fragment size, ambient temperature, and 235 weekly rainfall on the daily number of drinking records (our response variable) using the function 236 'lmer' of the R package lme4. We set the Poisson error family for the response variable and we 237 specified group ID as a random factor to account for repeated-measures from the same groups. 238 We did not consider interactions between predictor variables to minimize overparameterization  Therefore, we included all variables in the global GLMM model. 244 We used the Akaike's Information Criterion for small samples (AICc) to select the 245 models that best explain the effects of the predictor variables on drinking behavior. According to Water sources 262 We obtained a total of 1,258 individual drinking records (range=4-322 records/group, Table 1) 263 distributed in 917 events of group drinking and 313 observation days (range=0-16 records/day, 264 Table 1). We did not record drinking in 66% of the study days (i.e. 596 out of 909 days).

265
The water sources were streams (44% of 1,258 records), followed by treeholes (26%), 266 Vriesea, Aechmea, and Tillandsia bromeliads (16%), pools (11%), and rivers (3%) (Fig 2a). The  Seasonal and daily patterns in drinking behavior 285 We found no clear pattern in the proportion of drinking records between or within seasons (Fig   286   2b, Supplementary Fig S3). We observed drinking in all seasons in seven fragments, in three 287 seasons in six fragments, and in two seasons in the remaining fragment (Fig 2b). For those 288 fragments where we found seasonal differences in the proportion of drinking records (n=11,  Factors driving the drinking behavior of brown howlers 302 We found six models that included all predictor variables, except weekly rainfall, with substantial 303 empirical support (i.e. ΔAICc<2; Table 2). Flower and leaf consumption were the only predictors 304 present in all models. The two best models for explaining the frequency of drinking contained 305 only these two variables (first), plus ambient temperature (second; Table 2). The averaged model 306 differed from the null model (likelihood ratio test: X 2 =22, df=5, P<0.001). Flower consumption 307 had an inverse relationship with drinking (β=-0.14, z-value=3.08, P<0.01), whereas leaf 308 consumption had a direct one (β= 0.14, z-value=2.35, P<0.05, Table 2). Fragment size, fruit 309 consumption, and ambient temperature had insignificant relationships with howler monkey 310 drinking (Table 2). We found that brown howlers drank water accumulated in bromeliads and treeholes in the 314 canopy, and that they also descended to the ground to drink from streams, rivers, and pools.

315
Drinking increased in the afternoon and was less frequent in the spring. Also, while howlers 316 drank more when their diet included more leaves and drank less when they ate more flowers, the 14 317 contribution of fruits to the diet, habitat size, mean ambient temperature, and rainfall did not 318 predict the frequency of drinking.

319
The exploitation of non-food arboreal and terrestrial water reservoirs supports our 320 expectation that oxidation and preformed water are insufficient for permanently satisfying 321 howlers' water needs, as reported for many terrestrial mammals [1][2][3]. The finding that streams 322 were the most used water sources by brown howlers differs from the greater importance of    In sum, we have found that both the TDH and the MDH can explain the drinking behavior 371 of brown howlers in response to short-term thermal environment and diet composition.

372
Extrapolating from brown howlers to arboreal folivorous-frugivorous mammals in general that 373 also lack adaptations to tolerate high levels of dehydration, we suggest that the higher the 374 proportion of leaves in their diet, the greater might be the challenges in fulfilling their water 375 requirements, particularly in habitats where terrestrial water reservoirs are scarce or absent, such 376 as some forest fragments. Therefore, highly folivorous species may be more sensitive to droughts 377 than more frugivorous ones. Despite the higher availability of leaves than flowers and fruits in 378 forests, highly folivorous mammals may also be more vulnerable to predators if they are forced to 379 descend to the ground to drink from terrestrial reservoirs, particularly in forest fragments   We thank Danielle Camaratta and João Claudio Godoy for logistical support and field assistance. 403 We thank the landowners of the study fragments in Porto Alegre and Viamão for giving us 404 permission to conduct this research on their properties. We thank Commandant Aluísio S.R.

405
Filho for giving us permission to work in the Campo de Instrução de Santa Maria (CISM).