Factors Affecting Leaf Selection by Foregut-fermenting Proboscis Monkeys: New Insight from in vitro Digestibility and Toughness of Leaves

Free-living animals must make dietary choices in terms of chemical and physical properties, depending on their digestive physiology and availability of food resources. Here we comprehensively evaluated the dietary choices of proboscis monkeys (Nasalis larvatus) consuming young leaves. We analysed the data for leaf toughness and digestibility measured by an in vitro gas production method, in addition to previously reported data on nutrient composition. Leaf toughness, in general, negatively correlated with the crude protein content, one of the most important nutritional factors affecting food selection by leaf-eating primates. This result suggests that leaf toughness assessed by oral sensation might be a proximate cue for its protein content. We confirmed the importance of the leaf chemical properties in terms of preference shown by N. larvatus; leaves with high protein content and low neutral detergent fibre levels were preferred to those of the common plant species. We also found that these preferred leaves were less tough and more digestible than the alternatives. Our in vitro results also suggested that N. larvatus were little affected by secondary plant compounds. However, the spatial distribution pattern of plant species was the strongest factor explaining the selection of the preferred leaf species.

Choice between preferred and common species. Properties of common and preferred leaves are shown in Table 3. The preferred leaves generally produced more gas than the common leaves, with more  Table 1. Summary of model selection using linear models to investigate whether the leaf toughness was affected by chemical properties (only the models with ΔAICc score ≤ 2 are shown).
consistent differences between the in vitro assays with and without PEG (Fig. 2). The preferred leaves were also less tough (Fig. 3 or see more detailed results in Appendix I).
As shown in Table 4, the chemical properties, and especially CP (positive effect), were more useful than other factors in the predictions of preference for young leaves. In other words, N. larvatus selectively fed on young leaves with high CP content, as has been shown previously 6 . However, the ΔAIC values of some other models were also quite small (i.e. < 2.0), suggesting that other factors like the digestibility (positive effect) and leaf toughness (negative effect) can still be considered useful factors in predictions of leaf preference. For such predictions, the digestibility tested with PEG was more useful than digestibility without PEG.
The fact that the plant species abundance was always correlated negatively with the preference supports the observation that the N. larvatus does not feed opportunistically but preferentially select certain foods.
Selection within the preferred plant species. The GLMs predicting the percentages of the time spent by N. larvatus eating young leaves of the preferred species indicated that the leaf toughness was the factor most clearly related to this measure (Table 5). However, the trend was unexpected: the animals spent more time eating tougher leaves. Similarly, more time was spent feeding on leaves with more NDF and low digestibility than eating the other types of leaves. In contrast to the choice between common and preferred leaves, the positive relationship of the percentages of the time spent by N. larvatus eating each plant species of young leaves with abundance showed that, within their preferred plant species, N. larvatus fed opportunistically, not selectively. Nonetheless, these tendencies seemed to be driven by the two "super-abundant" plant species (Fig. 4), i.e. Mallotus muticus (Euphorbiaceae) and Lophopyxis maingayi (Lophopyxidaceae), as described in Matsuda, et al. 6 . Other factors were also selected in several models, indicating that the monkeys spent more time eating young leaves containing low levels of CP and CL compounds but with a high content of TA compounds.

Discussion
We analysed the mechanisms of food selection by a leaf-eating primate, N. larvatus. We considered not only the chemical properties and abundance of leaves but also the toughness and digestibility which are rarely taken into account.
We found that the toughness of young leaves was negatively correlated with the CP content, one of the most important nutritional factors affecting leaf selection by colobine monkeys e.g. 2,3,5,6 . This indicates that the leaf toughness, assessed by oral sensation, arising from mechanical properties 17,36 , might be a proximate cue for the protein content for leaf-eating primates. We did not examine the effect of tannins on the leaf toughness. Toughness and chemical toxicity increase as leaves age 37 , suggesting that toughness is a cue to reduce tannin intake, which negatively affects digestion. The toughness of plant tissues depends on relative amounts of cell wall components, i.e. NDF 17,18 . However, we detected a negative correlation between the fibre content (NDF) and toughness. The relationship between toughness and NDF content varies for different tissues, such as the midribs, petioles and laminae 13,18 . Kitajima, et al. 38 have shown that leaf toughness increases with increasing total bulk density and cellulose fraction, but decreases with the increasing hemicellulose and lignin content. These results suggest that we should not simply use the NDF as an indicator of toughness. The positive association between toughness and crude lipids levels might be caused by particularly thick, wax-covered cuticles in tough leaves 39 ; this relationship has rarely been investigated and deserves some future attention.
Young leaves preferred by N. larvatus contained more protein and less NDF than those of the common plant species, as previously reported by Matsuda, et al. 6 . Here, we found that those preferred leaves were less tough and more digestible than those of the common plants. Leaf toughness critically affects the feeding behaviour of colobines: ingestion rates (g/min) are negatively correlated, while masticatory investment (chews/g) is positively correlated with this trait 16 . Therefore, choosing young, tender leaves would be a reasonable strategy; it would increase the consumption of superior quality food with high protein content and, at the same time, ensure efficient ingestion. Nonetheless, if the toughness were the only proximate cue for the selection of young leaves, it would be difficult to select leaves with reduced NDF content. We found a negative correlation between fibre (NDF) and toughness, and a positive correlation between digestibility and toughness. These results suggest that N. larvatus may use another sense, such as vision 17 , to select good quality young leaves with high protein and low fibre content. As there is a variation in the colour of young leaves (e.g. light green and red), primates may account not only for leaf toughness but also leaf colour 40,41 . The leaf-colour factor should be included in the future investigations to give us a more comprehensive understanding of the dietary selection of leaf-eating primates.
The results of the digestibility tests with PEG were more useful for explaining the preferences of N. larvatus than the tests without PEG. This might indicate that these animals are capable of neutralising the tannins. Tannins in colobine forage significantly affect in vitro digestibility measured using rumen inoculum 29 . One typical strategy to improve digestibility is the production of tannin-binding salivary proteins; in ruminants, it is apparently linked to particularly large salivary glands 42 . Given that colobines, in general, have large salivary glands-an often-repeated statement based on the work of Kay, et al. 43 -they might use a similar strategy; however, this awaits further corroboration. Indeed, N. larvatus does not avoid young leaves of tannin-binding species 6 , which supports this assumption. There are many reports that tannin compounds do not affect leaf selection. This trend was found in colobine monkeys including N. larvatus as well as other non-human primates 5,30,[44][45][46][47][48] . Tannin-binding salivary proteins have been demonstrated in some primate species 23,49 .
We found that the most important factor affecting leaf selection among preferred young leaves was abundance, as shown by Matsuda, et al. 6 . In other words, N. larvatus was not highly selective in terms of leaf nutrition, toughness and digestibility; probably because of the relatively high quality of their preferred leaves in the forest areas adjacent to rivers. Forest areas near rivers might produce better leaves, with higher protein and lower fibre content than the primary inland forests due to the different soils and plant life histories 6,50-52 . Apart from its nutritional, mechanical and digestive properties, abundance of food is particularly important and this may be because the foraging cost saved by selecting common plants exceeds the benefit gained by selecting better quality leaves.   Our results show that the inclusion of more comprehensive measurements, beyond the quantification of crude nutrients, such as the leaf toughness or the in vitro digestibility, do not necessarily lead to a better understanding of dietary choices. This is an important conclusion for studies of primate diet selection, especially for comparisons between preferred and common leaf selections. CP remains probably the most easy-to-measure chemical proxy for the preferred diet items 7 . It should be noted, however, that available protein, i.e. CP minus the fibre-bound protein, may be the more relevant measure and could, in theory, yield better results. Because available protein and CP are generally positively correlated 53 , the use of CP appears nevertheless justified. However, when using the CP quantification, it is important to remember that it is not just a nitrogen-delivering nutrient. We should consider its relationships with other plant properties, including those related to the leaf age. The CP assays help to explain dietary choices as well as a more comprehensive set of predictors shown in Table 3. However, these results should not be interpreted as an indication that the other measures are not important, either conceptually  or physiologically. As the crude protein content is directly related to these factors, it remains a simple proxy for a complex phenomenon. Nonetheless, selection among preferred leaf species was rather difficult to explain based on such factors. Instead, the spatial distribution pattern of certain plant species, measured as abundance, was the strongest explanatory factor for diet choices of the N. larvatus at this location 5,6 .  6,31 . Continuous observations allowed the calculation of time budgets of adult monkeys, such as the proportion of the day spent feeding, and time spent feeding on individual food items. Overall, N. larvatus of the study group devoted 66% of feeding time to young leaves, 26% to fruits and 8% to flowers. The numbers of plant species providing young leaves, fruits and flowers were 182, 49 and 28, respectively 31 .

Sample collection and trait measurements.
We compared the chemical properties of young leaves preferred by N. larvatus and those of the common species in the study site. We had planned to sample the young leaves of the top 25 preferred species (defined on the basis of the proportion of total feeding time) 6  Three tests in leaf toughness studies have commonly been used, i.e. shearing test: measuring the work to traverse a leaf; punch test: measuring the maximum force to punch out the leaf lamina; tearing test: measuring the maximum force to tear a leaf strip 55 . Those completely different methods can apparently be used for a broad comparative survey of the toughness of primate foods; however, for some specific questions like discerning the mechanisms by which plant tissues resist fracture under incisal or molar loading, the choice of a particular test is of relevance 56 . From September 2014 to June 2015, young leaves of the same plant species were collected to assess the toughness of the leaves via punch test (originally measured as kilogram-force; 1 kgf = 1 kg × 1 G = 1 kg × 9.8 0665 m/s 2 = 9.80665 N, to convert SI units to kPa). It was determined using the mass needed to penetrate a leaf, employing a penetrometer 57,58 with a column of 3 mm in diameter (digital force gauges: Imada Co., Ltd, Aichi Japan). The measurement was performed for each of the 30 leaves per plant species (collected from at least four individual plants in the vegetation survey area) and the results were averaged for each species.     degradability of the young leaf samples (previously submitted to chemical analysis). The inoculum was obtained from the rumen fluid of cattle fed a standardized diet. An inoculum typical for N. larvatus was not available. However, the use of a standardized inoculum source makes it possible to compare the results with in vitro results from other studies. The relative abundance of microbes and their taxonomic assignments in the forestomach of the snub-nosed monkey (Rhinopithecus roxellana)-one of the species phylogenetically closest to N. larvatus-are rather similar to those in the cattle, in contrast with the microbiome of the humans or the giant panda (Ailuropoda melanoleuca) 59 . Two hundred milligrams of milled plant tissue was weighed in airtight glass syringes together with the inoculum, as described previously 28,60 , and incubated at 39 °C for 72 h. Gas production (Gp) was recorded after 4,8,12,24,32,48,56 and 72 h. The gas produced during the fermentation reflects the extent of food degradation. It consists of nearly equal parts of the waste gases of fermentation and the CO 2 from the buffer (bicarbonate) reaction with the volatile fatty acids produced during the fermentation 61 . The leaves were analysed without and with 200 mg of polyethylene glycol (PEG) added to reduce the negative effects of tannins on the digestion (Makkar et al. 1995). Most of the samples were used in two tests on two different days (with two replicates each time). Four plant species were tested on two different days without replicates due to an insufficient size of the sample. The fermentation parameters, maximal degradation and relative degradation rate (the proportion of maximal gas production; % h −1 ), were obtained according to Blümmel and Becker 62 , via nonlinear regression using the equation where a and b were constants (a + b = maximal gas production) [ 64 . Therefore, collinearity of the explanatory variables did not affect the results. We examined a set of models with all possible combinations of the explanatory variables and ranked them using the corrected version of the AIC for small sample size, the AICc 65 . Following guidelines published for wildlife research, we selected as the best-supported models those with a ΔAIC(c) score ≤ 2, where ΔAIC(c) = AIC(c) -minimum AIC(c) within the candidate model set 65 . In other words, if AIC(c) in a model is less than 2 units larger than in the best model, it must also be discussed and reported. A linear model was also used to find out whether the leaf digestibility (Gp at 40 h) for all collected plants, i.e. 30 plant species, was affected by leaf traits such as their chemical properties and toughness. The digestibility (with and without PEG) of young leaves was log-transformed to achieve normally distributed response variables. The other factors were treated as explanatory variables. The values of the VIFs were smaller than the cut-off value. We examined a set of models with all possible combinations of the explanatory variables and ranked them using the corrected version of the AIC.
To see which factors might explain N. larvatus preferences, we employed generalised linear models (GLM) using the leaf chemical properties, toughness, abundance and digestibility (with and without PEG), applying the binomial regression family calculations to obtain the AIC. The leaf preference, i.e. the preferred or common types (excluding species listed as preferred young leaf species), was treated as the categorical response variable, and the other leaf traits were treated as explanatory variables. The VIFs were smaller than the cut-off value even including the explanatory variables for digestibility with and without PEG. We examined a set of models with all possible combinations of the explanatory variables and ranked them using the corrected version of the AIC.
We also investigated the effects of leaf traits of the preferred plant species on the percentage of feeding time [as determined by Matsuda, et al. 6 ], using GLM. The gamma family (link function: inverse, i.e. the calculated coefficient value reflects the inverse effect) was used to calculate the AIC. GLM including the explanatory variables for digestibility with and without PEG was run separately as the VIFs for these two variables were larger than the cut-off value. We examined a set of models with all possible combinations of the explanatory variables and ranked them using the corrected version of the AIC. All analyses were performed using R ver. 3.1.0 66 , employing the dredge function in the MuMIn package, ver. 1.9.13 67 .