1 Obligate herbivory in an ancestrally carnivorous lineage: the 2 giant panda and bamboo from the perspective of nutritional 3 geometry 4 5 Yonggang Nie a, Zejun Zhang a, David Raubenheimer b, James J. Elser c, Wei Wei a, Fuwen 6 Wei* a 7 a 8 Academy of Sciences, 1-5 Beichenxi Road, Chaoyang, Beijing 100101, China 9 10 11 Key Lab of Animal Ecology and Conservation Biology, Institute of Zoology, Chinese b The Charles Perkins Centre and Faculty of Veterinary Science and School of Biological Science, The University of Sydney, Sydney, Australia c School of Life Sciences, Arizona State University, Tempe, AZ 85287-4501, USA 12 13 14 Running Headline: Nutritional geometry of the giant panda 15 16 17 18 19 20 21 22 23 24 25 *Corresponding Author: Fuwen Wei; E-mail: weifw@ioz.ac.cn 26 Address: Key Lab of Animal Ecology and Conservation Biology, Institute of Zoology, 27 Chinese Academy of Sciences, 1-5 Beichenxi Road, Chaoyang, Beijing 100101, China 1 28 29 30 31 Summary 1. Herbivores face various nutritional challenges in their life cycles, challenges that may become increasingly acute under ongoing environmental changes. 2. Here, focusing on calcium, phosphorus, and nitrogen, we used nutritional geometry 32 to analyze individual-based data on foraging and extraction efficiencies, and 33 combined these with data on reproduction and migratory behavior to understand how 34 a large herbivorous carnivore can complete its life cycle on a narrow and seemingly 35 low quality bamboo diet. 36 3. Behavioral results showed that pandas during the year switched between four main 37 food categories involving the leaves and shoots of two bamboo species available. 38 Nutritional analysis suggests that these diet shifts are related to the concentrations 39 and balances of calcium, phosphorus and nitrogen. Notably, successive shifts in 40 range use and food type corresponded with a transition to higher concentrations 41 and/or a more balanced intake of these multiple key constituents. 42 4. Our study suggests that pandas obligatorily synchronize their seasonal migration and 43 reproduction with the disjunct nutritional phenologies of two bamboo species. This 44 finding has potentially important implications for habitat conservation for this 45 species and, more generally, draws attention to the need for understanding the 46 nutritional basis of food selection in devising management plans for endangered 47 species. 48 Key words: feeding strategy, giant panda, nutritional geometry, life cycle, reproductive 49 timing, Right-angled mixture triangles, seasonal migration 2 50 51 Introduction It is widely accepted that herbivores face nutritional challenges, including low digestive 52 efficiency of food due to high fiber (Milton 1979), plant-produced toxins (Rosenthal & 53 Berenbaum 1991), and nutritionally imbalanced foods (Ritchie 2000; Elser et al. 2000, 2007). 54 Such nutritional challenges may grow worse under ongoing climate change that shifts the 55 range, timing, and physiological conditions of forage plants (Tuanmu et al. 2012). In response 56 to such challenges, animals have evolved behavioral, developmental and physiological 57 adaptations that interact across timescales to facilitate homeostasis and maintain performance 58 (Mayntz et al. 2005; Rothman et al. 2011). The science of nutritional ecology aims to 59 understand the ways that these interactions mediate the relationships between nutrient needs 60 and ecological constraints (Raubenheimer et al. 2009). This helps to inform our understanding 61 of the ecological and evolutionary processes that have shaped the diversity of animal foraging 62 modes and to devise management strategies for endangered species and their habitats (Moore 63 & Foley 2005; Raubenheimer et al. 2012). 64 An important requirement for maintaining fitness in the face of changes in nutrient supply 65 or demand (e.g. at different stages in the life cycle) is compensatory homeostatic adjustment 66 whereby foraging behavior and physiological processing of nutrients counter-act the changes 67 (Raubenheimer et al. 2012). Such changes typically involve balancing the gain of several 68 nutrients (e.g. usable energy, protein and amino acids, minerals), and are highly significant for 69 individual fitness and also population dynamics (Pyke 1984; Lewis & Kappeler 2005; Taillon 70 et al. 2006). In particular, females have specific nutritional requirements to support the 71 demands of reproduction, and animals that live in seasonal environments must often adjust the 3 72 timing of reproduction and the pattern of foraging behavior to meet these nutritional demands 73 (Goldizen et al.1988; Rubenstein & Wikelski 2003). 74 The giant panda (Ailuropoda melanoleuca) is an endangered, obligate herbivore that 75 diverged early within an otherwise carnivorous clade (Qiu & Qi 1989; Wei et al. 2012; Zhao 76 et al. 2013). Uniquely within the order Carnivora, pandas specialize (~99%) on various 77 species of bamboo, resulting in a diet that is generally believed to be of poor quality due to 78 low protein and high fiber and lignin contents, contributing to its low dry matter digestibility 79 (Schaller et al. 1985; Hu et al. 1990; Wei et al. 1999; Zhu et al. 2011). Despite being 80 exclusively herbivorous, the giant panda retains the simple stomach and short gastrointestinal 81 tract typical of carnivores (Dierenfeld et al. 1982), and consequently needs to eat large 82 amounts of poorly digestible foods (Hu et al. 1990). This high degree of specialization on 83 large quantities of low quality food taken from a small number of plant species renders the 84 giant panda vulnerable to extinction in the face of environmental change (Colles et al. 2009). 85 Further, its recently acquired herbivorous lifestyle provides an especially interesting 86 opportunity for evaluating how shifts in herbivore behavior allow a species to cope with food 87 quality challenges involving the levels and balance of essential nutrients. 88 Previous studies have highlighted a number of interesting ecological and life history 89 characteristics that might be associated with the ability of panda populations to survive as 90 bamboo specialists (Schaller et al. 1985; Pan et al. 2001). Nevertheless, these relationships 91 remain poorly understood, largely because of the difficulty of obtaining intensive behavioral 92 data for these secretive animals. Other bears give birth to unusually small, altricial young, a 93 trait that has been associated with hibernation (Garshelis 2004). However, pandas do not 4 94 hibernate and yet they have the shortest gestation period (3 – 5.5 months) and give birth to 95 offspring that are the smallest of any bear species (Garshelis 2004; merely 0.1% of the 96 mother’s weight). As with other bears, pandas have an embryonic diapause, known as 97 seasonal delayed implantation, in which the embryo remains suspended in the uterus in a state 98 of arrested development until it attaches and resumes growth, sometimes months later 99 (Schaller et al. 1985). While the adaptive significance of this remains uncertain (Thom et al. 100 2004), delayed implantation is believed to be an ecological adaptation to adjust the timing of 101 mating and the rearing of offspring to different seasonal environments (Sandell 1990). 102 In this study we used a combination of direct behavioral observations and 103 individual-based characterization of food intake and egestion to assess the nutritional 104 consequences of the seasonal food choices of pandas. We relate these choices to the timing of 105 altitudinal migration and other major life history events including seasonal mortality, mating, 106 gestation, parturition and lactation. We focused our analysis on the mineral nutrients calcium 107 and phosphorus, as well as nitrogen as a proxy for protein, because of the critical roles that 108 these nutrients play in growth and reproduction of animals, including mammals (White 1993; 109 McDowell 1996; Moen et al. 1999; Sterner & Elser 2002). Importantly, mammalian 110 requirements for calcium and phosphorus, principally for bone growth, are critically 111 inter-dependent (Van Soest 1994; Underwood & Suttie 1999) and thus of special interest in 112 the study of nutritional ecology of vertebrate herbivores. To examine the roles of these 113 multiple dimensions of nutritional quality and their inter-dependencies, we organize our 114 analysis using nutritional geometry, an approach for modeling the inter-active effects of 115 nutrients on animals (Raubenheimer 2011). Our objectives were: i. to elucidate the 5 116 relationships between the distinctive adaptations of these newly obligate herbivores, seasonal 117 habitat choice, food selection and nutrient gain in the extreme nutritional environment to 118 which they have become specialized, and ii. to learn whether the respective foods and habitats 119 of the giant pandas are inter-changeable (alternative sources of the same resources), or 120 complementary (provide different combinations of essential nutrients). The study thus 121 provides fundamental insight into the nutritional ecology of a highly unusual and ecologically 122 threatened herbivore, as well as critical information for the management and conservation of 123 panda habitat. 124 Methods 125 Study site and animals 126 This study was conducted in Foping Reserve, a key panda reserve, in the Qinling 127 Mountains, China. The Qinling Mountains contain a high density of wild giant pandas with a 128 population of 273 individuals (State Forestry Administration- China 2006). Two bamboo 129 species are the main diet resource of the pandas there, wood bamboo (Bashania fargesii) and 130 arrow bamboo (Fargesia qinlingensis), which grow at mean elevations of 1600 m and 2400 m, 131 respectively. These two bamboo species have different life histories. Wood bamboo (WB) 132 produces shoots in May and the shoots begin to sprout abundant new leaves in August. In 133 contrast, arrow bamboo (AB) produces shoots in early June and its shoots sprout a limited 134 number of new leaves in the following spring and considerably more new leaves in summer. 135 The leaves of WB persist year-round while the AB leaves drop off in winter. 136 137 With approval from the State Forestry Administration in China (2009-261), a total of six pandas, three adult females and males, respectively, were fitted with GPS/VHF collars (Lotek 6 138 Wireless Inc., Ontario, Canada; Nie et al. 2012a.b, Zhang et al. 2014). This made it possible 139 to conduct intensive behavioral observations and collect food samples, enabling us to 140 determine individual-level seasonal food intake and obtain paired food- fecal samples for 141 chemical analysis and assessment of relative digestive extraction efficiencies. 142 Observations of foraging behaviour and sample collection 143 Using the GPS collars, over six years we tracked pandas from short distances (usually 144 10-20 m) to examine their seasonal pattern of food selection. Food and dung samples were 145 collected in the four foraging seasons of 2009 and 2010. During this period we also tracked 146 pandas to conduct behavioral observations at intervals of 3-5 days for each individual, except 147 when inclement weather prevented this. These observations enabled us to determine which 148 bamboo species, tissues, and ages were chosen by pandas. Paired food and fresh feces 149 samples were collected at each observed feeding patch during different foraging periods 150 year-round. We defined a feeding patch as an area with a size of ~300 x 300 m within which a 151 panda was observed feeding for at least 24 hours (because the gut passage time is usually 152 around 10-12 hours; Schaller et al. 1985). Bamboo leaf and shoot samples were collected 153 according to the age of plants; that is, one- and multi-year old leaves, and new and old shoots, 154 respectively. All food and fecal samples were coded by the feeding patch, dried in the field 155 station, and the plant samples were sorted by different bamboo species and tissues. The dried 156 plant and fecal samples were stored in zip-lock bags in the field for transport to laboratory. 157 Life cycle and mortality data collection 158 To examine the possible relationship between nutrition and reproduction strategy, we 7 159 conducted a study of the reproductive ecology of giant pandas by tracking collared animals 160 over six years from 2007 to 2012 (Nie et al. 2012a.b). We also analyzed long-term (37 years) 161 historical data of panda death and illness events in the wild from Foping Reserve records. 162 These data were used to explore the potential effect of food resource quality on the individual 163 lifespan and population dynamics of this endangered species. 164 Laboratory analyses 165 A total of 263 plant and fecal samples were collected in the field, including 66 shoots 166 and 47 fecal samples during shoot foraging season, and 100 leaf and 50 fecal samples in leaf 167 foraging season. All samples were ground to powder with a common multi-functional 168 laboratory mill and oven-dried at 70℃ and then weighed before laboratory analyses. We used 169 the micro-Kjeldahl method (Bremner 1996) to analyze N concentrations (% of dry mass). P 170 contents (% of dry mass) were measured by the ammonium molybdate method after 171 persulfate oxidation (Kuo 1996), standardized against known reference materials. Ca contents 172 (% of dry mass) were determined using an atomic absorption spectrometer after hydrofluoric 173 acid oxidation (Langmyhr & Thomassen 1973). 174 Data Analysis 175 We used Right-angled Mixture Triangles (RMTs, Raubenheimer 2011) to explore the 176 relationships among the proportional contents of nutrients in the foods and published 177 estimates of nutrient requirements. To estimate digestive extraction efficiencies of Ca, P and 178 N, we compared the proportional compositions of food samples with the associated feces 179 using RMTs. This method does not yield the absolute digestive efficiencies of separate 8 180 nutrients, for which measures are needed of the absolute intake and excretion of each nutrient. 181 Rather, by comparing the concentrations of nutrients in the food and matched feces we are 182 able to establish the relative extraction efficiencies of the focal nutrients (Raubenheimer 2011). 183 For example, if the concentration of P in the feces was half that in the matched foods, we 184 could not conclude that P was extracted with 50% efficiency, because we would not know the 185 extent to which the change in P concentration from food to feces was due to the extraction of 186 other nutrients (i.e., changes in the denominator rather than numerator in the concentration 187 ratio). However, if the Ca:P ratio in the food was twice the Ca:P ratio in the feces, then we 188 could conclude that Ca was extracted with higher efficiency (by a factor of 2) than P. Since in 189 this analysis we were interested in the relative extraction efficiencies of Ca, P and N, we 190 chose to use an RMT model in which each nutrient was expressed as a percentage of the sum 191 of the three nutrients [e.g., %Ca = Ca/(Ca+P+N) x100] rather than as a percentage of the total 192 sample mass (i.e., grams Ca/100g sample). This enables the relative digestive efficiencies of 193 all three nutrients to be compared in a single model, and also excludes from the denominator 194 unaccounted components that might otherwise confound the comparison of relative extraction 195 efficiencies (Raubenheimer 2011). We could then use as a baseline the null model in which all 196 three nutrients are extracted from the food with equal efficiency, indicated in RMTs as the 197 situation where the Ca-P-N ratio of feces and the associated food is the same (i.e., the 198 composition points for food and feces are superimposed). Alternative outcomes would be 199 indicated by the vectors of displacement of feces composition relative to food; for example, if 200 the Ca:P ratio in feces is lower than in food, this indicates that Ca was extracted with higher 201 efficiency than P. 9 202 We used t-tests to compare the concentrations and ratios of nutrients in the foods of 203 pandas, and to compare the ratios of nutrients in the foods with matched fecal samples to 204 establish relative digestive efficiencies (as explained above). Levene’s Test was used to test 205 for equality of variances, and where the null hypothesis of equal variances was rejected we 206 applied a modified t-test that does not assume equal variances. A Kolmogorov-Smirnov Test 207 was used to compare the observed monthly frequencies of mortality with the random null 208 model. One-sample t-tests were used to compare dietary calcium: phosphorus ratios with the 209 required ratios from the literature. All tests were performed using IBM SPSS v. 20. 210 Results 211 Seasonal migration, feeding, reproduction and mortality 212 Seasonal movement and foraging patterns 213 A mean of 52.5±6.5 observation days were collected year-round for each animal with a 214 range of 43 to 61 days. Over an annual cycle, pandas fed on two bamboo species located at 215 different elevations, which we refer to as winter habitat and summer habitat (Fig. 1, 2). All six 216 collared pandas in this study showed a similar pattern of foraging transition, switching to WB 217 shoots in early May (range: 30 April to 4 May) when the shoots reached a height (31.0±5.75 218 cm) sufficient that pandas could eat them. In early June (3 June to 11 June), pandas moved to 219 higher elevation and switched to AB shoots, within a short period from 3 June to 11 June at 220 the time that low-elevation WB shoots had grown tall (247±23.6 cm) and become lignified. 221 Similarly, pandas switched to AB leaves when the shoots of AB had grown tall (233±12.1 222 cm). The mean time of the migrations to low elevation was at the end of August, within the 223 period of 12 August - 10 September. Finally, pandas preferred the younger leaves during the 10 224 two leaf periods. Thus, based on the seasonal migration and the specific tissues eaten from 225 these two bamboo species, there were four distinct foraging periods, two in which leaves and 226 two in which shoots were eaten: wood bamboo leaf period (WBl), wood bamboo shoot period 227 (WBs), arrow bamboo shoot period (ABs) and arrow bamboo leaf period (ABl, Fig. 2). During 228 the ABl period, pandas sometimes ingested a small portion of the stems when eating leaves 229 but we did not include these infrequent occurrences in the data analysis. 230 Reproductive timing in giant pandas 231 Over several consecutive years, March (15 mating events; 68%) and April (7 mating 232 events; 32%) were the main months for mating for the Qinling Mountain pandas. Pregnancy 233 lasted 4-5 months (collared females: 137-143d; 146-151d; 132-138d), of which 1.5-2 months 234 comprised post-implantation gestation (Schaller et al. 1985). The females gave birth in a short 235 period between mid-August and early September: there were 4 births in mid-August (57%); 2 236 births in late August (29%) and one in early September (14%) during our study (Fig. 2). 237 Seasonal pattern of mortality of giant pandas 238 A total of 25 dead or ill pandas were observed in the wild over the past 37 years in 239 Foping Reserve. More than half (52%) of these occurrences were in March and April, a 240 frequency that was statistically greater than expected by chance (Z = 1.61; P = 0.01). This 241 period corresponds to the end of the longest time on any of the four diets—the WB leaf period 242 (Fig. 2). 243 Nutritional composition of foods 244 Nitrogen-phosphorus relationships 11 245 Figure 3a shows the composition of the seasonal diets in terms of N and P and a range of 246 N to P ratios that are likely to encompass biomass requirements of mammals based on 247 proximate chemical composition and investment in muscle and bone for an animal of this 248 body size (Elser et al. 1996). The nutritional implications in relation to N and P for pandas for 249 the spring switch from old WB leaves to WB shoots are denoted by the solid arrow labeled 1 250 in Fig. 3a. Young shoots (early May) had substantially higher concentrations of N (P < 0.001) 251 and P (P < 0.001) than the leaves, with disproportionately more P and hence a lower N:P ratio 252 (P < 0.001). As the shoots matured through May and early June the concentrations of both N 253 (P < 0.0001) and P (P < 0.0001) dropped, but the N:P ratio remained unchanged (P = 0.097). 254 The arrow marked 2 in Fig. 3a shows nutritional changes associated with the switch from 255 old WB shoots to young AB shoots in June (see Fig. 2). The N:P ratio did not differ between 256 the foods (P = 0.98), but the concentrations of both N (P < 0.0001) and P (P < 0.0001) were 257 higher in AB shoots, indicating a shift to foods that allow greater intake of these nutrients. 258 By mid-July the concentrations of both N and P in the AB shoots had significantly 259 declined (P < 0.001), although the N:P ratio barely changed (P = 0.057). This corresponded 260 with a switch in July from the older shoots to young leaves of AB (solid arrow marked 3 in 261 Fig. 3a), which were higher both in N (P < 0.0001) and P (P < 0.0001), with an increased N:P 262 ratio (P < 0.0001). 263 Between mid-July and August, the concentrations of N (P < 0.0001) and P (P < 0.0001) 264 had decreased in AB leaves, and the N:P ratio had increased from 11.9±0.35 to 13.9±0.57 (P 265 = 0.005), whereupon female pandas migrated to the lower altitude foraging site and fed on 266 young leaves of WB. These leaves had a significantly higher P content (P = 0.016) with 12 267 similar N content (P = 0.997), and consequently a lower N:P ratio (P = 0.012) than the older 268 AB leaves. 269 From August to April, when the pandas once again switched to young shoots of WB 270 (arrow 1 in Fig. 3a), the concentrations of both N (P < 0.0001) and P (P < 0.0001) in WB 271 leaves decreased, and the N:P ratio increased (P = 0.002). 272 Calcium-phosphorus relationships 273 The relationships between Ca and P in the dietary transitions by pandas through the 274 annual cycle are shown in Fig. 3b. The significant increase in dietary P concentration (see 275 above) corresponding with the spring switch from leaves to shoots of WB (Fig. 2) was 276 accompanied by a reduction in Ca (P < 0.0001), and consequently a strong decrease in the 277 Ca:P ratio (P < 0.0001). Thereafter, as the shoots aged, the concentration of Ca dropped (P < 278 0.0001) together with the concentration of P (see above). However, Ca dropped more steeply 279 than P, resulting in a significant reduction in the Ca:P ratio (P < 0.0001). The Ca:P ratios of 280 shoots in the early, mid and late season were 0.26±0.02, 0.19±0.02 and 0.09±0.03, 281 respectively; all of these are substantially lower than the Ca:P ratio of 1-2 recommended in 282 the diets of mammals (van Soest 1994, Underwood & Suttie 1999; Buchman & Moukarzel 283 2000) (P < 0.0001). In contrast, the mean Ca:P ratio of mature WB leaves was 3.4±0.24, 284 which is significantly higher than the maximum of 2 recommended for mammals (P < 285 0.0001). 286 The switch from the old shoots of WB to young shoots of AB (arrow 2 in Fig. 3b) was 287 associated not only with a significant increase in P (see above), but also with a proportionately 288 larger increase in Ca (P < 0.0001). Consequently, the Ca:P ratio was higher (closer to the 13 289 recommended range) in young AB (0.25±0.02) than the older WB shoots (0.09±0.03, P < 290 0.0001). As AB shoots aged, there was a reduction in P (see above) and Ca (P < 0.0001), and 291 a decrease in the Ca:P ratio from 0.25 to 0.10 (P < 0.0001). 292 In mid-July, when the pandas switched from shoots to young leaves of AB (arrow 3 in Fig. 293 3b), the leaves were significantly higher in both P (above) and Ca (P < 0.0001), with a 294 substantially higher Ca:P ratio (P < 0.0001). At a value of 0.10±0.02, the Ca:P ratio of shoots 295 was an order of magnitude lower than the recommended minimum for mammals of 1 (P < 296 0.0001), while the value for leaves (2.49±0.08) was marginally but significantly greater than 297 the recommended maximum ratio of 2 (P < 0.0001). 298 By mid-August the leaves of AB had reduced in both P (see above) and Ca (P < 0.0001) 299 concentrations, but the Ca:P ratio remained unchanged (P = 0.84). At this point the female 300 pandas moved to a lower elevation and switched to the younger leaves of WB, which had a 301 significantly higher P content (above) and marginally higher Ca (0.51±0.04% vs.0.43± 302 0.02%; P = 0.08). The Ca:P ratio did not differ between the leaves of the two species (P = 303 0.97). 304 From mid-August, when female pandas started eating young WB leaves, to April when 305 they switched from the now older leaves to young WB shoots (arrow 1 in Fig. 3b), P 306 concentration in the leaves dropped (above) but there was no change in Ca (P = 0.63). 307 Consequently, the dietary Ca:P ratio of the WB leaves increased from 2.5±0.13 in 308 mid-August to 3.4±0.24 in April (P = 0.002). 309 Relative digestive extraction efficiencies 310 Our analysis (Table 1) showed that, relative to N, both Ca and P were enriched in feces 14 311 compared with shoots, whether the food species was WB (Fig. 4a) or AB (Fig. 4b). This 312 demonstrates that N was extracted from shoots with higher relative efficiency than either P or 313 Ca. Additionally, the Ca:P ratio in feces was higher than in shoots, indicating that Ca was 314 extracted with lower relative efficiency than P, thus exacerbating the effective deficit of Ca 315 relative to P in shoots. In both species of bamboo, the feces associated with leaves were 316 relatively enriched in Ca, whereas the proportional concentration of P was statistically 317 unchanged in leaves and feces (Table 1, Figs 4a and 4b). 318 Discussion 319 This is the first field study to use nutritional geometry to explore the relationship 320 between the balance of essential nutrients, selection of foraging habitat, and the life cycle of a 321 highly endangered herbivore species. Our data showed pandas experience marked seasonal 322 foraging changes with four primary foraging periods corresponding to the annual phenology 323 of the two bamboo species. The nutritional quality of the diet was heterogeneous through the 324 year, both in terms of the absolute concentrations of N, P, and Ca and the proportional 325 balance of these nutrients. Seasonal diet switches corresponded with shifts in quantities of 326 these key nutrients, as did the life cycle, reproduction and pattern of altitudinal migration. As 327 we will discuss, the close correspondence between animal life history events, shifting range, 328 and forage quality suggest that nutritional balancing is a contributing component that 329 maintains the population of this endangered species, an insight that may be crucial in its 330 conservation as well as that of other endangered species that have narrow dietary ranges. 331 332 Our analysis showed that the N-P-Ca composition of diets varied with bamboo species, plant part, and the age of the plant part. Age influenced primarily the concentration of 15 333 nutrients, which declined as both leaves and shoots matured (Figs 3a and 3b), most likely due 334 to an increase in plant structural components such as cellulose, hemicellulose, and lignin (Hu 335 et al. 1990). In the absence of reliable measures of daily food intake (Rothman et al. 2011), 336 which cannot readily be obtained for wild giant pandas in the field, it is difficult to interpret 337 the significance of the concentration of a nutrient in foods in relation to the animal’s 338 requirement for that nutrient. Invariably, however, diet switches by pandas corresponded with 339 a change to younger, more nutrient-rich alternatives, whether this involved a different plant 340 part (e.g., arrow 1 in Figures 3a and 3b) or species (e.g. arrow 2 in Figures 3a and 3b). This 341 preference for younger tissues could relate to their higher nutrient concentrations compared 342 with older tissues in which nutrients are diluted by greater concentrations of structural 343 components. Additionally, the biomechanical properties associated with plant structural 344 components can also reduce the nutritional quality of foods (Clissold et al. 2009). Although 345 we did not measure plant-produced allelochemicals, these might play a similar role 346 (Launchbaugh et al. 2001). 347 Nutrient concentrations also differed between species of bamboo, but these differences 348 were contingent on the plant part. Specifically, the young shoots of wood bamboo had 349 considerably higher N, P, and Ca concentrations than the young arrow bamboo shoots, 350 whereas the species difference was less marked and reversed for leaves (Figs 3a and 3b). The 351 greatest difference between plant parts, however, was in the balance of nutrients. Leaves had a 352 higher N:P ratio (Fig. 3a) and a substantially higher Ca:P ratio (Fig. 3b) than did shoots, and 353 this contrast applied for both bamboo species. Such differences in nutrient balance can be a 354 significant parameter of food quality, because nutrient balance determines the ways that 16 355 nutrients interact in their effects on consumers (Sterner & Elser 2002; Simpson & 356 Raubenheimer 2012). 357 An important consideration in inferring the functional significance of seasonal diet 358 switches in giant panda is therefore their implications for nutrient balance. Thus, the spring 359 switch by the giant pandas in our study from leaves to shoots of wood bamboo corresponded 360 with increased dietary N content, which could well be an important functional driver of the 361 switch (White 1993). Significantly, dietary P content increased to a proportionally even 362 greater extent than N, and consequently the dietary N:P ratio decreased. Since the N:P ratio of 363 wood bamboo leaves was higher than the optimal range, the decrease corresponding to the 364 switch to shoots brings the N:P ratio in the diet more in line with estimated requirements (the 365 shaded area in Fig. 3a). The higher P concentrations in the shoots of wood bamboo do, 366 however, have important consequences for giant pandas in relation to the dietary Ca:P ratio. 367 Dietary Ca:P ratios of 1:1 to 2:1 are recommended for mammals (Fig. 3b), with excesses 368 of either nutrient interfering with the absorption and metabolism of the other (Robbins 2001). 369 When the Ca:P ratio drops much below 1, P impedes absorption of the already limiting Ca, 370 resulting in Ca resorption from bones and ultimately osteomalacia (softening of the bones) 371 and associated diseases. In both human and animal studies, Ca:P ratios less than 0.5 have been 372 associated with reduced bone mass density and compromised bone strength (Calvo & Tucker 373 2013). These effects can be particularly acute in relation to reproduction because of its 374 increased calcium requirements for lactation and bone growth (Schulkin 2001). It is therefore 375 noteworthy that the high levels of P in young wood bamboo shoots in our study resulted in 376 Ca:P ratios of considerably less than 1 (0.2) . In contrast, Ca-P ratios in leaves were closer to 17 377 the recommended range for mammals, being marginally above 2 (Fig. 3b). In general, Ca:P 378 ratios higher than 2 (surplus Ca) are tolerated by herbivores to a greater extent than ratios less 379 than 1 (surplus P) (Robbins 2001). 380 It is important to note that measures of Ca:P ratios in plant foods might not accurately 381 represent the biologically effective Ca:P ratios, because a proportion of these elements might 382 be bound in molecular complexes that render them unavailable (Suttle 2010). For example 383 phytic acid is an important storage form of P in many plant tissues (especially seeds), and is 384 poorly digested by non-ruminant herbivores. However, if the relative excess of P in the panda 385 diet was significantly influenced by unavailable P in this way, then we would expect the feces 386 would be more highly P-enriched relative to Ca, but this was not the case. Rather, the feces 387 produced from bamboo shoots were enriched in Ca relative to P (i.e. the Ca:P ratio of feces 388 was greater than the Ca:P ratio of the shoots), and therefore the post-absorptive Ca:P ratio 389 associated with shoots was even lower than 0.2 (Fig. 4). The relatively high Ca levels in the 390 feces are consistent with the interfering effect of surplus P on Ca absorption discussed above. 391 By contrast with shoots, the feces associated with leaves were enriched in Ca but had 392 statistically similar P contents to the leaves (Fig. 4). This selective egestion of Ca would bring 393 the Ca:P ratio of leaves, which was marginally higher than 2, more closely in line with the 394 recommended range. 395 Our nutritional analysis therefore implies that, when pandas switch from a diet of old 396 leaves to shoots of wood bamboo at around the time of mating, they shift from a diet that is 397 low in both N and P with a N:P ratio that exceeds the maximum recommended for mammals 398 to a diet that is higher in both nutrients and has an N:P ratio within the recommended range 18 399 (Fig. 3a). With the subsequent switch to arrow bamboo shoots, the dietary concentrations of 400 both N and P were reduced but the N:P ratio was very close to the center of the expected N:P 401 range (~6.75, Fig. 3a). These high shoot N and P contents likely help support construction of 402 the placenta and the growing embryo during fetal development. During this period, however, 403 the dietary Ca:P ratio was considerably lower than considered necessary to support 404 reproduction in mammals, and was only restored with the subsequent switch to arrow bamboo 405 leaves (Figs 2 and 3b). 406 These dynamics lead us to suggest that perhaps delayed implantation provides a means 407 for pandas to postpone the Ca investment in lactation and bone growth, synchronizing these 408 more closely with a leaf-based diet that can support them. On the other hand, the relatively 409 low levels of both N and P, and the high N:P ratios in the autumn and winter diet, present 410 additional challenges for panda reproduction. Specifically, both N and P are required for 411 tissue growth, and there would be obvious fitness penalties for pandas that could not acquire 412 these in sufficient quantities for reproduction. An interesting possibility is that this could be 413 related to the evolutionary maintenance and enhancement of the short gestation period of 414 pandas, and the extremely small size of the offspring at birth (Garshelis 2004). Giving birth to 415 altricial young would ease the burden on the mother for acquiring limiting nutrients, by 416 enabling the offspring to start independent feeding (i.e. weaning) earlier. Having both mother 417 and offspring eating to meet their own respective nutrient needs would allow the pair to 418 process bamboo and acquire limiting nutrients at a greater rate than if the burden fell on the 419 mother alone. Barclay (1994) used similar reasoning to argue that the long development time 420 for flight, which delays independent foraging in flying vertebrates (bats and birds), might 19 421 impose constraints on Ca acquisition for bone growth and explain why these animals 422 generally have small litters. The peak in panda mortality in March and April is also consistent 423 with an interpretation that the extended low quality of the winter diet of leaves is nutritionally 424 stressful, highlighting the need for both mother and offspring to forage for limiting nutrients. 425 In summary, our analysis has shown that young shoots of wood bamboo were high in P 426 and N but had a Ca:P ratio markedly lower than is considered necessary to support bone 427 growth in mammals. In June, the levels of P, Ca and N dropped in the maturing shoots of 428 wood bamboo, whereupon the pandas migrated to higher elevation. This allowed them to 429 switch to arrow bamboo shoots, which had higher levels of P, N and Ca but again a 430 sub-optimally low Ca:P ratio. By early August, during the late stages of gestation, nutrient 431 levels in arrow bamboo shoots had dropped and the pandas switched to feeding on the leaves 432 of the same species. These had higher nutrient levels, in particular Ca, a more favorable Ca:P 433 ratio, and a better Ca:P absorption profile than bamboo shoots. In August, females returned to 434 the lower elevation feeding sites where birthing coincided with the availability of young wood 435 bamboo leaves, with high nutrient content and a high Ca:P ratio. However, as wood bamboo 436 leaves aged through the winter, their P content decreased, reaching their lowest level in the 437 period that coincides with the highest historical mortality rate. At that point, the pandas again 438 switched to young wood bamboo shoots once they became available. 439 Overall, our results also suggest that the two bamboo species are nutritionally 440 interchangeable but that different plant parts (shoots and leaves) are not. Rather, the shoots 441 and leaves are nutritionally complementary resources, with shoots providing primarily N and 442 P, but deficient in Ca, which is provided by the leaves. Both species of bamboo are, 20 443 nonetheless, critical for the pandas, because their asynchronous phenology, coupled with 444 seasonal altitudinal migration, enables the pandas to complete their life cycle on this low 445 diversity and highly specialized diet. Our insights into the phenological dynamics of panda 446 nutrition have important implications for managing the conservation of this charismatic 447 species in the face of climate change. Tuanmu et al. (2012) recently modelled likely future 448 distributions of wood and arrow bamboo in light of several IPCC climate projections for the 449 Qinling Mountain region (our study area). They noted strong potential for range contraction 450 and elevation shifts in these species, changes that are likely to lead to phenological 451 mismatches between the timing of panda life cycle events and the nutritional suitability of 452 bamboo. Such possibilities highlight the need for a systems-approach to panda conservation, 453 in which pandas, both species of their food plants, and their respective habitats, both current 454 and projected, are all afforded protection. More broadly, this work gives us a new insight into 455 animal nutritional ecology of potential benefit to further research in the field of animal 456 ecology and conservation biology, especially for species that face serious nutritional 457 challenges due to accelerating environmental change. 458 459 Acknowledgements 460 This study was supported by the National Natural Science Foundation of China (31230011； 461 31370414), the Knowledge Innovation Program of the Chinese Academy of Sciences 462 (KSCX2-EW-Z-4). Memphis Zoo provided additional support. JJE acknowledges support 463 from US National Science Foundation (DMS-0920744). The authors gratefully acknowledge 464 the support of the Wild Animal Experimental Platform of Chinese Academy of Science. We 21 465 are grateful to Drs. Marcel Holyoak, Michael Bruford, Ronald Swaisgood and Gabriel 466 Machovsky for their useful comments on a previous version of this manuscript. We also thank 467 XL Wang, QG He, and AX Zhang for their assistance in the field. 468 22 469 References 470 Barclay, R.M.R. (1994) Constraints on reproduction by flying vertebrates: energy and 471 calcium. American Naturalist, 144, 1021–1031. 472 Bremner, J.M. (1996) Nitrogen: total. In: Methods of Soil Analysis. Part 3: Chemical Methods 473 (eds Sparks, D. L., A. L. Page, P. A. Loeppert, P. N. Soltanpour, M. A. Tabatabai, and C. 474 T. Johnston, et al.). Soil Science Society of America and American Society of 475 Agronomy, Madison, pp. 1085–1123. 476 477 478 Buchman, A.L. & Moukarzel, A. (2000) Metabolic bone disease associated with total parenteral nutrition. Clinical Nutrition, 19, 217–231. Calvo, M.S. and Tucker, K.L. (2013) Is phosphorus intake that exceeds dietary requirements a 479 risk factor in bone health? In Dietary Phosphorus Excess and Health (Uribarri, J., ed), pp. 480 29-35 481 Clissold, F.J., Sanson, G.D., Read, J. & Simpson, S.J. (2009) Gross vs. Net income: How 482 plant toughness affects performance of an insect herbivore. Ecology, 90, 3393–3405. 483 Colles, A., Liow, L.H. & Prinzing, A. (2009) Are specialists at risk under environmental 484 change? Neoecological, paleoecological and phylogenetic approaches. Ecology Letters, 485 12, 849–863. 486 487 Dierenfeld, E., Hintz, H., Robertson, J., Van Soest, P. & Offeda, O. (1982) Utilization of bamboo by the giant panda. Journal of Nutrition, 112, 636–641. 488 Elser, J.J., Dobberfuhl, D.R., MacKay, N.A. & Schampel. J.H. (1996) Organism size, life 489 history, and N:P stoichiometry: toward a unified view of cellular and ecosystem 490 processes. BioScience, 46, 674–684. 23 491 Elser, J.J., Fagan, W.F., Denno, R.F., Dobberfuhl, D.R., Folarin, A., Huberty, A., Interlandi, 492 S., Kilham, S.S., McCauley, E., Schulz, K.L., Siemann E.H. & Sterner, R.W. ( 2000) 493 Nutritional constraints in terrestrial and freshwater food webs. Nature, 408, 578–580. 494 Elser, J.J., Bracken, M.E.S., Cleland, E.E., Gruner, D.S., Harpole, W.S., Hillebrand, H., Ngai, 495 J.T., Seabloom, E.W., Shurin, J. B. & Smith, J.E. (2007) Global analysis of nitrogen and 496 phosphorus limitation of primary producers in freshwater, marine and terrestrial 497 ecosystems. Ecology Letters, 10, 1135–1142. 498 Garshelis, D.L. (2004) Variation in ursid life histories: Is there an outlier? Comparative 499 ecology of giant pandas in the five mountain ranges of their distribution in China. Chapter 500 4 in: Lindburg, D. G., K. Baragona, and G. B. Schaller, (eds), Giant Pandas: Biology and 501 Conservation. University of California Press, Berkeley, California. pp.53–73. 502 Goldizen, A.W., Terborgh, J., Cornejo, F., Porras, D.T., & Evans, R. (1988) Seasonal food 503 stortage, weight loss, and the timing of births in saddle-black tamarins (Saguinus 504 Fuscicollis). Journal of Animal Ecology, 57, 893–901. 505 506 507 Hu, J., Wei, F., Yuan, C. & Wu, Y. (1990) Research and Progress in Biology of the Giant Panda. Sichuan Publishing House of Science and Technology, Chengdu. Kuo, S. (1996) Phosphorus. In: Methods of Soil Analysis. Part 3: Chemical Methods (eds 508 Sparks, D.L., A. L. Page, P. A. Loeppert, P. N. Soltanpour, M. A. Tabatabai, and C. T. 509 Johnston, et al.). Soil Science Society of America and American Society of Agronomy, 510 Madison, pp. 869–920. 511 512 Langmyhr, F.J. & Thomassen, Y. (1973) Atomic absorption spectrophotometric analysis bydirect atomization from the solid phase. Zeitschrift für Analytische Chemie, 264, 24 513 514 515 516 122–127. Launchbaugh, K.L., Provenza, F.D. & Pfister. J.A. (2001) Herbivore response to anti-quality factors in forages. Journal of Range Management, 54, 431–440. Lewis, R.J. & Kappeler, P.M. (2005) Seasonality, body Condition, and timing of reproduction 517 in propithecus verreauxi verreauxi in the Kirindy Forest. American Journal of 518 Primatology, 67, 347–364. 519 Milton, K. (1979) Factors influencing leaf choice by howler monkeys: a test of some 520 hypotheses of food selection by generalist herbivores. American Naturalist, 114, 521 363–378. 522 523 524 525 526 527 528 529 530 Mayntz, D., Raubenheimer, D., Salomon, M., Toft, S. & Simpson, S.J. (2005) Nutrient-specific foraging in invertebrate predators. Science, 307, 111–114. McDowell, L.R. (1996) Feeding minerals to cattle on pasture. Animal feed science and technology, 60, 247–271. Moen, R.A., Pastor, J. & Cohen, Y. (1999) Antler growth and extinction of Irish Elk. Evolutionary Ecology Research, 1, 235–249. Moore, B.D. & Foley, W.J. ( 2005) Tree use by koalas in a chemically complex landscape. Nature, 435, 488–490. Nie, Y., Zhang, Z., Swaisgood, R.R. & Wei, F. (2012a) Effects of season and social 531 interaction on fecal testosterone metabolites in wild male giant pandas: implications for 532 energetics and mating strategies. European Journal of Wildlife Research, 58, 235–241. 533 Nie, Y., Swaisgood, R., Zhang, Z., Liu, X. & Wei, F. (2012b) Reproductive competition and 534 fecal testosterone in wild male giant pandas (Ailuropoda melanoleuca). Behavioral 25 535 536 537 538 539 540 541 542 543 544 545 546 Ecology and Sociobiology, 66, 721–730. Pan, W., Lu, Z., Zhu, X., Wang, D., Wang, H., Fu, D., et al. (2001) A chance for lasting survival. Peking University Press, Beijing. Pyke, G.H. (1984) Optimal foraging theory: a critical review. Annual Review of Ecology and Systematics, 15, 523–575. Qiu, Z. & Qi. G. (1989) Ailuropoda found from the late Miocene deposits in Lufeng, Yunnan. Vertebrata PalAsiatica, 27, 153–169. Raubenheimer, D., Simpson, S.J. & Mayntz, D. (2009) Nutrition, ecology and nutritional ecology: toward an integrated framework. Functional Ecology, 23, 4–16. Raubenheimer, D. (2011) Toward a quantitative nutritional ecology: the right-angled mixture triangle. Ecological Monographs, 81, 407–427. Raubenheimer, D., Simpson, S.J. & Tait, A.H. (2012) Match and mismatch: conservation 547 physiology, nutritional ecology and the timescales of biological adaptation. 548 Philosophical Transactions of the Royal Society B-Biological Sciences, 367, 1628–1646. 549 550 551 552 553 554 Ritchie, M.E. (2000) Nitrogen limitation and trophic vs. abiotic influences on insect herbivores in a temperate grassland. Ecology, 8, 1601–1612. Robbins, C.T. (2001) Wildlife Feeding and Nutrition and Edition. Academic Press, New York. Rosenthal, G.A. & Berenbaum, M.R. (1991) Herbivores: their interaction with secondary plant metabolites. The Chemical Participants. Vol.1. Academic Press, San Diego. 555 Rothman, J.M., Raubenheimer, D., Colin, A. & Chapman, C.A. (2011) Nutritional geometry: 556 gorillas prioritize non-protein energy while consuming surplus protein. Biology Letters, 26 557 558 559 560 561 562 563 564 565 7(6), 847–849. Rubenstein, D.R. & Wikelski, M. (2003) Seasonal change in food quality: a proximate cue for reproductive timing in marine iguanas. Ecology, 84, 3013–3023. Sandell, M. (1990) The evolution of delayed implantation. Quarterly Review of Biology, 65, 23–42. Schaller, G.B., Hu, J., Pan, W. & Zhu, J. (1985) The giant panda of Wolong. University of Chicago Press, Chicago. Schulkin, J. (2001) Calcium Hunger: Behavioral and Biological Regulation. Cambridge University Press, Cambridge. 566 Simpson, S.J. & Raubenheimer, D. (2012) The Nature of Nutrition: A Unifying Framework 567 from Animal Adaptation to Human Obesity. Princeton University Press, Princeton. 568 569 570 571 572 573 State Forestry Administration—China (2006) The third national survey report on giant panda in China. Science, Beijing (in Chinese) Sterner, R.W. & Elser, J.J. (2002) Ecological Stoichiometry: The Biology of Elements From Molecules to the Biosphere. Princeton University Press, Princeton, NJ. Suttle, N.F. (2010) Mineral Nutrition of Livestock, 4th Edition. CAB International, Wallingford. 574 Taillon, J., Sauve, D.G. & Cote, S.D. (2006) The effects of decreasing winter diet quality on 575 foraging behavior and life-history traits of white-tailed deer fawns. Journal of Wildlife 576 Management, 70, 1445–1454. 577 578 Thom, M.D., Johnson, D.D.P. & MacDonald, D.W. (2004) The evolution and maintenance of delayed implantation in the mustelidae (Mammalia:Carnivora). Evolution, 58, 175–183. 27 579 Tuanmu, M.N., Vina, A., Winkler, J.A., Li, Y., Xu, W., Ouyang, Z. & Liu, J. (2012) 580 Climate-change impacts on understorey bamboo species and giant pandas in China’s 581 Qinling Mountains. Nature Climate Change, doi:10.1038/nclimate1727. 582 583 584 585 586 587 588 Underwood, E. & Suttie, N. (1999) The mineral nutrition of livestock. 3rd ed. Wallingford, Oxon, United Kingdom: CAB International. pp. 614. Van Soest, P.J. (1994) The nutritional ecology of the ruminant. 2nd ed. Cornell University Press, Ithaca, NY, USA, pp. 476. Wei, F., Feng, Z., Wang, Z. & Li, M. (1999) Feeding strategy and resource partitioning between giant and red pandas. Mammalia, 63, 417–430. Wei, F., Hu, Y., Zhu, L., Bruford, M.W., Zhan, X. & Zhang, L. (2012) Black and white and 589 read all over: the past, present and future of giant panda genetics. Molecular Ecology, 21, 590 5660–5674. 591 592 593 White, T.C.R. (1993) The Inadequate Environment: Nitrogen and the Abundance of Animals, Springer, New York. Zhang, Z.J., Sheppard, J.K., Swaisgood, R. R., Wang, G., Nie, Y. G., Wei, W., Zhao, N. X., 594 Wei, F. W. (2014) Ecological scale and seasonal heterogeneity in the spatial behaviors of 595 giant pandas. Integrative Zoology, 9, 46–60. 596 Zhao, S.C., Zheng, P.P., Dong, S.S., Zhan, X.J., Wu, Q., et al. (2013) Whole genome 597 sequencing of giant pandas provides insights into demographic history and local 598 adaptation. Nature Genetics, 45, 67–71. 599 600 Zhu, L., Wu, Q., Dai, J., Zhang, S. & Wei, F. W. (2011) Evidence of cellulose metabolism by the giant panda gut microbiome. Proceedings of the National Academy of Sciences USA, 28 601 108, 17714–17719. 602 603 29 604 Table 1. Comparison of the N, P and Ca concentrations and Ca/P ratio in foods (shoots and 605 leaves) from two species of bamboo and the feces produced by giant pandas from eating the 606 respective foods. All concentrations are calculated as a proportion of the sum of the three 607 components [e.g. %Ca = (Ca/N+P+Ca) x 100]. Statistics are from independent samples t-tests 608 comparing the N, P, Ca composition of each food with the composition of the relevant feces. 609 Positive t-values indicate decreased concentration in feces relative to food, while negative 610 t-values indicate increased concentration in feces relative to food. 611 Bashania fargesii (wood bamboo) Component N P Ca Ca/P Leaves t(38.270) = 8.150, P <0.0001 t(38.810)=1.375, P=0.177 t(39.741)=-8.791, P<0.0001 t(36.798)=-6.936, P<0.0001 Shoots t(30.016) = 11.731, P <0.0001 t(37.258)=-9.095, P<0.0001 t(26.320)=-9.974, P<0.0001 t(33.803)=-6.873, P<0.0001 612 613 30 Fargesia qinlingensis (arrow bamboo) Leaves Shoots t(58)=4.197, t(23.083)=15.464, P<0.0001 P<0.0001 t(58)=0.640, t(23.782)=-10.914, P=0.524 P<0.0001 t(58)=-4.141, t(22.970)=-11.942, P<0.0001 P<0.0001 t(25.590)=-3.373, t(26.928)=-8.529, P=0.001 P<0.0001 614 Figure legends 615 Figure 1. Annual pattern of migration of giant pandas corresponding with a diet transition 616 between two bamboo species in our study area. We used the GPS location data of two of the 617 GPS collared pandas to exemplify the seasonal foraging migration pattern of giant pandas in 618 this area. All the six collared pandas in our study lived in the same winter (September-May) 619 habitat area at low elevation where wood bamboo (Bashania fargesii) is located. In summer, 620 two of them moved to summer habitat 1 while four moved to summer habitat 2, at high 621 elevation where arrow bamboo (Fargesia qinlingensis) is located. 622 Figure 2. Scheme relating the annual pattern of diet selection and habitat elevation to the 623 reproductive cycle of giant pandas. Foods: WBl = wood bamboo leaves; WBs = wood bamboo 624 shoots; ABs = arrow bamboo shoots; ABl = arrow bamboo leaves. Life history events: MS = 625 mating season, DII = delayed implantation interval, BS = birthing season. 626 Figure 3. Implications in terms of N:P (a), Ca:P (b) of seasonal diet shifts in giant pandas. 627 Green = wood bamboo (WB), orange = arrow bamboo (AB); circles = shoots, squares = leaves; 628 no border on symbols = young tissue; thin border = intermediate, thick border = old. Solid red 629 arrows show active switches by pandas between foods, dashed blue arrows = seasonal 630 changes in composition of foods being eaten. The shaded area denotes the recommended 631 range for N-P, Ca-P ratios in the diets of mammals. 632 Figure 4. Comparison of the Ca-N-P composition of foods (shoots and leaves) and the 633 associated feces in (a) wood bamboo (WB) and (b) arrow bamboo (AB). Circle-orange = 634 shoots; Circle-empty = shoot-associated feces; Oval-green = leaves; Oval-empty = 31 635 leaf-associated feces. Radials show mean Ca-P ratios, and negatively-sloped diagonals (%N 636 isolines) show the mean percentage of N relative to the sum of N, P and Ca in each sample. 637 Values for the % N isolines are shown in square brackets. Solid lines represent foods (leaves = 638 green, shoots = orange) and dashed lines represent feces associated with leaves (green) or 639 shoots (orange). The shaded area shows the range of Ca-P ratios recommended in the diets of 640 mammals. 32