31° South: Dietary niche of an arid-zone endemic passerine

Balancing energy budgets is thought to be challenging for birds living in arid ecosystems because food supplies are low and unpredictable, and climatic conditions extreme. Thus, to ensure they obtain sufficient energy to fuel daily energetic budgets, birds may need to adjust their diets and become less selective (generalist) as conditions become harsher. To test this hypothesis, we used DNA metabarcoding to characterize both the prey availability (from pitfall traps) and the dietary content (from fecal samples) of several conspecific populations of a semi- and arid-endemic insectivorous bird, the Karoo-scrub-robin (Cercotrichas coryphaeus) across a climatic gradient. Our results showed that Coleoptera, Hymenoptera, Orthoptera, and Lepidoptera were the main prey. When accounting for their presence as available prey, Coleoptera and Hymenoptera were preferred in all regions, whereas robins avoided Orthoptera and Lepidoptera in all but the most arid region. Although the different populations live in regions that vary with regards to productivity and thermoregulatory demands, we found that the dietary niche breadth (Bs) of the three populations was intermediate to low, and did not differ significantly. As a whole, our findings show that regardless of environmental harshness these insectivores have similar dietary niches, suggesting that large dietary plasticity is fundamental for their survival in energy-depauperated ecosystems.


INTRODUCTION 23
Food acquisition is a basic need of all animals: it provides the energy and nutrients required to 24 sustain life in any given environment. Because of the intricate relationship of animals with the 25 environment (where food is obtained from), there has long been an interest in describing patterns 26 of dietary variation in order to predict the underlying causes. The classical Optimal Foraging 27 Theory model (MacArthur and Pianka 1966) (Levins and MacArthur 1969) (Schoener 1971) 28 predicts that animals seek to maximize energy intake, so as to meet their daily energy costs. 29 Consequently, both prey availability in the environment and animal physiological status can 30 affect diet variation (e.g.: (Burgar et al. 2014). 31 Endotherms living in landscapes characterized by dramatic change in rainfall and temperature 32 face two main problems: firstly, frequent changes in food webs; and secondly, thermal challenges 33 that require them to adjust their physiology (energetically expensive tasks) to maintain body 34 temperatures within tolerable limits. As a consequence of both, some populations may undergo 35 dietary niche shifts in order to maximize fitness (Roches et al. 2016). In response to limited 36 resources, selection can either lead to generalist strategies such as dietary niche flexibility to use 37 the panoply of resources available (Grant et al. 2008), or drive the evolution of specialization so 38 as to exploit underused resources (Martin and Pfennig 2009). Moreover, if resources vary greatly 39 through time or space, being a generalist may allow fitness to be maintained (Stephens and Krebs 40 1986). 41 Arid-zone endemic birds are good models for testing the role of environmental temperature and 42 energetic physiology in dietary strategies, because they face continual challenges from their 43 energy-depauperated ecosystems, while seeking to fulfill the dietary requirements needed to fuel 44 the physiological processes associated with thermoregulation. In this study we molecularly 45 characterized both the diet and available arthropod prey community (Pompanon et Table S2), by first downloading all arthropod COI sequences from South 129 Africa archived on GenBank in January 2016, then designing them by hand in Geneious 7.0.6 130 (Biomatters, New Zealand) to meet the following criteria: a) 20-22 bp length; b) produce a 131 fragment of 150-210 bp (including primers); and c) C/G as the first two 3' end bases (alignment 132 provided as Appendix). 133 In addition, because robins have been seen eating berries (P. Lloyd, pers. comm.), we targeted the 134 chloroplast rbcL gene using primers rbcL-h1aF/rbcL-h2aR (Poinar et al. 1998). The ZBJ-135 ArtF1c/ZBJ-ArtR2c primers amplify a 211 bp region in the 5' section of COI gene, while the 136 FormiF/FormiR primers target a 168bp region in the 3' of same gene. All primers were modified 137 to include a unique 6 to 8 nucleotide sequence (nucleotide multiplex identifiers; MIDs) on the 5' 138 end to allow individual identification following (Binladen et al. 2007 included in all reactions. The optimal number of PCR cycles was established using a real-time 155 PCR assay in a subset of samples as the inflection point in the amplification curve and it 156 represents a compromise to obtain sufficient prey DNA while reducing clonality and consequent 157 bias in prey diversity. Real time PCR was performed in 20ul reaction containing 1-2 µl of 158 template, 1x buffer, 2.5 nM MgCl2, 2 nM each dNTP, 0.5U Amplitaq Gold, 1 nM of each 159 primer, and 1 µl SYBR Green/Rox mix (Invitrogen, USA). PCR amplifications were detected by 160 gel electrophoresis on 2% agarose gel stained with Gelred. All PCR preparations used aerosol 161 resistant filter tips, and were prepared in UV-sterilized laminar flow hoods. We note that 162 arthropods and fecal samples were processed independently, and the same applied to summer and 163 winter samples. This differential processing of samples was implemented as to avoid cross-164

contamination. 165
All PCR products were visualized on a gel to confirm amplification success, and then pooled at 166 an approximately equimolar ratio into 11 pools (5 for summer and 6 for winter). We included the 167 extraction blanks (from both fecal and arthropod pitfall extractions) in the pools, using 5 µl, 168 despite no actual product visualized in agarose. 169 We removed unspecific fragments and primer dimer using Ampure XP SPRI beads, following a 170 double-sided protocol. Shortly, by adjusting the ratio of beads to DNA (0.7x) we first captured 171 the fragments larger than 300bp, keep the supernatant, which was in turn used to sequester 8 fragments larger than 120bp (beads:DNA ratio = 1.8x). The concentration of the purified pools 173 was determined in Qubit and efficiency of purification evaluated in and 2% agarose gel stained 174 with Gelred. 175 The PCR pools were converted into sequencing libraries using the NEBNext 6070 blunt end 176 library build kit following the manufacture's protocol. Index PCRs were carried out in 25 µl 177 reaction volumes, and each library was amplified in three replicates to maximize amplicon fell below the thresholds or for which a match was not found in the NCBI database was 219 categorized as "unknown". Not all MOTUs could be assigned to species level due to the 220 incompleteness of the reference database for our study area. Therefore we present results 221 identified to family-level. We contend this still provides a good grasp of the dietary components 222 as well as the arthropod communities. 223 224

Statistical Analyses 225
All analyses were implemented in R v3.3.2 (R Development Core Team). We estimated 226 Shannon-Wiener's (H) and Simpon's (1-D) indices to quantify both consumed and available prey 227 diversity using vegan package (Oksanen 2017). Diversity indices were compared among sites 228 using the non-parametric Kruskal-Wallis test, as the data was not normally distributed (Shapiro-229 Wilk test, p < 0.05). Changes in composition of prey were inspected using a non-metric 230 multidimensional scaling (NMDS) with Bray-Curtis dissimilarity. To test whether region 231 (Coastal, Central, Inland; proxy for primary productivity and energetic demands) is shaping 232 diversity in available and consumed prey we used a Mantel test (Mantel 1967) with 999 random 233 permutations as implemented in vegan (Oksanen 2017). Briefly, we compared the observed 234 dissimilarity matrix (Bray-Curtis dissimilarity) against a conceptual matrix for regional effects, 235 coded as zero for same region comparisons and one for different region comparisons. We 236 highlight that we did not test for seasonal effects. Although we contend that seasonal variation in 237 weather and a bird´s life-stage (breeding vs non-breeding) may have implications relating to 238 available and consumed prey, respectively, and despite our effort to catch birds, the non-balanced 239 sample sizes hinder a formal analysis. 240 Trophic niche breadth was estimated using a standardized Levin's index (Bs; (Levins 1968). (specialist) and closer to one for a generalist diet. We created Bs 95% confidence intervals for 246 each region by bootstrapping with 10,000 replicates. In addition, to test whether the observed Bs 247 values were significantly different from a scenario of "no regional difference" we used a 248 permutational approach. All occurrences (pi) were pooled regardless of their regional origin, then 249 we randomly sampled n arthropod families (from the observed range: 37 -45) and finally 250 estimated Bs_random. This procedure was repeated 9999 times to generate the null distribution. 251 The estimated probability of obtaining a result that exceeded the observed value under the null 252 hypotheses of "no region difference" was estimated as p = [(number Bs_random > 253 Bs_observed)/total number randomizations]. 254 For the most common arthropod MOTUs in the diet of robins, we tested whether selectivity of 255 prey changed among regions, by calculating Ivlev's electivity index (Ivlev 1961) for each region 256 as implemented in selectapref package. This index varies from "-1" (avoidance of prey) to "+1" 257 (total reliance on prey), and "0" indicates prey selection is proportional to availability.

Arthropod availability 267
Our survey of potential Arthropoda prey availability across the gradient revealed 519 MOTUs, 268 with the majority (83%) identified to class Insecta (Appendix Table 3S). Coleoptera, Diptera, 269 Hymenoptera and Lepidoptera were the most abundant orders (Fig. 2). The most diverse 270 community was found in the Central region, the area with largest annual primary productivity 271 Simpson's 1-D = 0.94; Table 1). However, when these observations were tested statistically, we 275 were unable to reject the null hypothesis that there is no regional effect on arthropod community 276 (Mantel region , p = 0.430). 277 278

Arthropods and plants consumed 279
After data filtering criteria, 82% of fecal samples provided arthropod MOTUs (170). Insecta was 280 the richest Arthropoda class (127 MOTUs) with the orders Coleoptera, Hymenoptera, 281 Lepidoptera and Orthoptera being the most abundant (Fig. 2, Appendix Table 3S). Although the 282 Karoo scrub-robin is viewed as insectivorous, we detected plant DNA in 94% of the fecal 283

samples. Overall, 86 MOTUs were assigned to Spermatophyta (seed Plants) with Solanales and 284
Aspargales being the most abundant (Appendix Table 4S).

Trophic niche and prey selection 293
When accounting for available arthropod prey, we found that birds living in the Inland region -294 which was least productive and exerted high energetic demands on the robins -appeared to have 295 the narrowest dietary niche breadth (Bs = 0.372, 95% CI = 0.241 -0.677, Fig. 3A). This contrasts 296 with the Coastal population, whose individuals presented the widest niche (Bs = 0.479, 95% CI = 297 0.328 -0.667; Fig. 3A). However, again when statistically tested, we found that the observed Bs 298 values for each of the three study regions was not significantly different from the values expected 299 under the null hypothesis of no regional difference (mean Bs_random = 0.445, 95% CI = 0.247 -300 0.592, p Coastal = 0.658, p Central = 0.568, p Inland = 0.738; Appendix Fig. 2S). 301 An electivity analysis (that measures utilization of food items) restricted to the five most 302 abundant potential prey items (Coleoptera, Diptera, Hymenoptera, Lepidoptera and Orthoptera), 303 showed that robins from all regions positively selected for Hymenoptera:Formicidae (Ivlev´s 304 electivity > 0; Fig. 4). Electivity was, however, largest in the Inland population, with 55% of 305 sampled birds consuming ants. Although in Central region relatively more Coleoptera families 306 were potentially available as prey (Appendix Fig. 3S), robins did not positively selected on this 307 prey type (Fig. 4). We also found that Dipterans, another widely available potential prey source, 308 were negatively selected on in all regions (Ivlev´s electivity < 0; Fig. 4) 309 Karoo scrub-robins are generalist insectivores, some plant items seem to be included in its diet; 318 ii) dietary composition was not associated with region; iii) when accounting for available 319 arthropod prey the dietary niche breadth of each population was not different from a scenario of 320 no regional effect (i.e., similar productivity and thermoregulatory demands); iv) Formicidae 321 (ants) were positively selected in the three populations, yet Inland robins showed the largest 322 electivity. 323 324

Dietary niche breadth 325
The concomitant assessment of available and consumed prey in conspecific populations revealed 326 that the foraging behavior of robins is not simply based on encounter rate: Coleopterans, 327 Hymenopterans and Orthopeterans were preferred prey. While coleopterans represent less than 328 16% of available prey (7%, 16% and 11% of MOUTUs in Coastal, Central and Inland, 329 respectively) they were clearly preferentially preyed upon (representing 33% of total dietary 330 MOTUs in the Coastal region, 23% in the Central region and 20% in the Inland region). A 331 similar pattern of preference was found for Hymenopterans in the Central and Inland regions: 332 Hymenopetans were consumed at a higher rate than their relative availability (comprising 11% 333 and 19% of the dietary MOTUs ,vs 7% and 11% of available MOTUs). The exception was found 334 in the area with intermediate primary productivity and temperature amplitude -the Coastal 335 region -where Hymenopetans were only 4% of available prey and 5% of consumed MOTUs. In 336 contrast, Diptera and Lepidoptera were consumed at a rate inversely proportional to their 337 presence, i.e. they were avoided by the birds. These results are in accordance to prey descriptions 338 based on Karoo scrub-robins stomach contents (invasive method) sampled throughout the year in 339 a region East of our Inland site (Free State; (Oatley 1970)). Specifically, Oatley (1970) revealed 340 that ants (Hymenoptera) and beetles (Coleoptera) were present in 55% and 14% of the stomachs, 341 respectively. We also noted that while terrestrial arthropods are the bulk of robin's prey, they 342 appear to supplement their diets with plant matter (94% of samples yielded plant DNA). 343 Although, this finding could in theory simply indicate that robins are plant secondary consumer 344 through herbivore arthropods, we contend the latter is unlikely given both the anticipated 345 degradation to the plant DNA that would be incurred through two rounds of ingestion and 346 degradation, and perhaps more importantly, two additional lines of evidence. Firstly, there are 347 unpublished observations (P. Lloyd, pers. comm.) of birds foraging on berries and plant shoots, 348 and secondly, seeds and berries were detected in the stomachs contents of robins (Oatley 1970). 349 Thus together, it suggests that that plant matter is not simply a serendipitous food item, but may 350 in fact be selected for. 351 Our study sites are embedded within the Southern African semi-and arid-regions, where seasonal 352 pulses of rainfall are known to determine arthropod densities (Dean and Milton 1999) and 353 subsequently food supplies for insectivores (Lloyd 1999). Surprisingly, despite the differences in 354 primary productivity among our study sites, the dietary niche breadth (Bs) among populations 355 was not different from a scenario of homogeneous environment (i.e., no difference in 356 productivity or thermal demands). This plasticity in diet may be favorable in environments where 357 prey is patchily distributed in space and time, such as the Southern African semi-and arid-358 regions (Dean and Milton 1999). Moreover, this finding corroborates the expectations of Optimal 359 Foraging Theory that animals can only afford to be specialist (small niche breadth) where 360 resources are readily available (Stephens and Krebs 1986). 361 362

Life in arid environments 363
Life in arid environments is harsh for endotherms in general, and in particular for ground-feeding 364 insectivore birds as they need to fuel their typically high avian metabolic rates in energy-365 depauperated ecosystems. We predicted the robins living in the most demanding regions (Inland) 366 would exhibit the most generalist feeding strategy, under the rationale that as this was the most 367 energy-depauperated ecosystem and energetically demanding environment, their diet should 368 become more generalist as ignoring potential sources of energy is not an option (Levins and 369 MacArthur 1969;Schoener 1971). Specifically, we expected robins living in the Inland region to 370 prey on the greatest diversity of ground arthropods, so as to obtain sufficient energy to fuel the 371 thermogenic demands of winter (burn energy to produce heat as a means to warm-up 372 (thermogenesis; (Hothola 2004) as well as actively cool down in summer through evaporative 373 cooling (Williams and Tieleman 2005). In contrast however, we found that the Inland robins prey 374 on an array of arthropods that is similar to robins living in the less demanding region (Coastal). 375 We also noted they exhibit a preference for ants (Formicidae), a finding that lends support to the 376 notion that dependable food sources such as ants are critical for arid endemic birds (Dean and 377 Milton 2017), yielding the necessary energy, nutrients and water (robins do not drink) to 378 overcome the challenges that the environment exerts. 379 We also hypothesize that supplementation of their diet with plant items such as berries (the most 380 common plant in diet, Solanaceae, produces berries) may provide the carbohydrates and hence 381 calories required to cover their general energetic budget, when sufficient animal protein and fat is 382 not available. Furthermore beside sugars, fruits would also provide extra water, a critical 383 component when dealing with hot temperatures: birds loose water to cool down and avoid 384 hyperthermia in the hot days. Although we did not test for seasonal differences, it is worth noting 385 that plant DNA was mostly found in samples collected in winter. We hypothesize this may be 386 related to high energy demands to deal with cold as well as short time-window to actively search 387 for arthropod prey on the ground (shorter periods of day light).