Peer Review #2 of "Elucidating the diet of the island flying fox (Pteropus hypomelanus) in Peninsular Malaysia through Illumina Next-Generation Sequencing (v0.2)"

1 Rimba, Bandar Baru Bangi, Selangor, Malaysia 2 Département Adaptations du Vivant, UMR MECADEV CNRS-MNHN 7179, Muséum National d’Histoire Naturelle, Brunoy, France 3 School of Environmental and Geographical Sciences, The University of Nottingham Malaysia Campus, Semenyih, Selangor, Malaysia 4 Department of Biological Sciences, University of Southampton, Southampton, United Kingdom 5 School of Science, Monash University Malaysia, Petaling Jaya, Selangor, Malaysia 6 Kenyir Research Institute, Universiti Malaysia Terengganu, Kuala Terengganu, Terengganu, Malaysia 7 Département Adaptations du Vivant, Muséum National d’Histoire Naturelle, UMR MECADEV CNRS-MNHN 7179, Muséum National d’Histoire Naturelle, Brunoy, France 8 Genomics Facility, Tropical Medicine and Biology Platform, Monash University Malaysia, Petaling Jaya, Selangor, Malaysia 9 School of Natural Sciences and Engineering, National Institute of Advanced Studies, Bangalore, India

). Prior to primer synthesis, partial Illumina adapter sequences were 232 added to the 5' end of the designed primers, rbcL-357F and rbcL-556R, to allow barcoding and 233 sequencing on the Illumina platform. 234 Individual droppings were pooled according to roost (n=5, 2 in Tekek and 3 in Juara) and 235 month (n=8), creating 40 separate mixtures for analysis. The tubes containing the daily samples 236 were first vortexed for 2 min to homogenise the content and subsequently, 1000 µL of the 237 sample was pipetted into another tube to form the mixture. Next, 100 µL of the mixture was used 238 for DNA extraction similarly using DNAeasy Plant Mini Kit (Qiagen, Halden, Germany) instead 239 of a stool-specific DNA extraction kit to improve the recovery of plant-derived DNA from faecal 240 samples.

241
PCR reaction was performed using IlluM_rbcLF and IlluM_rbcLR. The 20 µL PCR 242 cocktail consists of 10 µL Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolab, 243 Ipswich, MA), 1 µL each of 10 µM forward and reverse primer, 1 µL gDNA and 7 µL nuclease-244 free water. All reactions were performed in a Veriti® 96-Well Fast Thermal Cycler with the 245 following protocol: initial denaturation for 30 sec at 98 ºC, 25 cycles of 10 sec at 98 ºC, 30 sec at 246 55 ºC and 10 sec at 65 ºC, with a final 1 min extension at 65 ºC. The PCR product was purified 247 using 0.8x vol. ratio Agencourt Ampure XP beads (Beckman Coulter, Inc). Then, 1 uL of Index 248 1 and Index 2 primers from Nextera XT kit were added to 3 uL of purified PCR product and 249 combined with 5 uL of Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolabs, 250 Ipswich, MA). The PCR protocol was as followed: initial denaturation for 30 sec at 98 ºC, 8 251 cycles of 10 sec at 98 ºC and 1 min at 65 ºC, with a final 1 min extension at 65 ºC.   281 For the 10 dropping samples collected in May, we sent one set for NGS analysis (following the 282 protocol above) and used another set for microscope analysis. For the latter, we first manually 283 broke up the dropping contents in the tube to produce a relatively more representative liquid 284 sample. We then droppped 1-3 drops of this liquid onto a microscope slide using a pipette. 285 Fuchsin jelly was added to this in order to stain pollen grains within the dropping, a slip cover 286 was placed on top, and the jelly was then melted over an open flame, sealing the slip cover to the 287 slide. The slide was then cooled down in a conventional fridge in order to allow the jelly to 288 solidify again before examination.

289
Once cooled, we placed the slide under a conventional light microscope (Leica DM E) 290 and first examined it using 10/0.25 magnification in order to detect pollen grains and other plant 291 parts. We used a self-made reference collection as well as photos from Start (1974), S. 292 Bumrungsri (http://www.seabcru.org/seabcru-resources) and Mohamed (2014) to identify pollen 293 and other plant parts. When necessary, we used higher magnification (40/0.65) to view the pollen 294 grains. We used just 'presence/absence' to assess pollen to avoid quantification biases towards 295 species that naturally produce greater amounts of pollen. Following Thomas (2009) we 296 considered a species present if we found three or more pollen grains on a single slide.   315 With our newly designed rbcL primer, we were able to successfully extract, amplify, and 316 subsequently identify plant DNA from all of the collected flying fox droppings. Initially, a total 317 of 160 Operationally Taxonomic Units (OTUs) were recovered from the sequencing reads, of 318 which 29 OTUs (Table 1) were retained after filtering based on cumulative relative abundance 319 (>0.5%) and presence of stop codon(s) in reading frame. Using a conservative LCA approach, 320 we identified at least three different plant genera and at least 14 plant families from the 321 droppings (Table 1). In addition, 8 OTUs matched with specimens from the site-specific plant 322 reference collection. Based on sampling completeness (calculated using EstimateS) for OTU relative 324 abundance data from five roosts (data pooled over three days) per month using Chao 1 species 325 richness estimator (good for datasets skewed towards low abundance classes; Chao 1984), 326 sampling completeness was relatively high for the months March (99%), April (100%), June 327 (100%), August (100%), September (88%), and October (96%). However, sampling 328 completeness could be improved for May (55%) and July (79%). The month of May, which had 329 the lowest overall sampling completeness, also had the highest number of droppings collected Table S4).

331
The results from our NGS analysis of island flying fox droppings over eight months 335 Spatio-temporal patterns in the relative abundance of these four taxa in the diet were observed 336 during the sampling period (Fig 2). For example, OTU 5 appeared to be consumed in similar 337 proportions at both Juara and Tekek across all months whereas OTU 4 was consistently 338 consumed in low proportions in Tekek yet consumed irregularly in Juara over the same period 339 (Fig. 3). Even between different roosts in the same site, spatio-temporal differences were 340 observed, such as for OTU 7 (Supplementary Fig. S5).   363 We have demonstrated that identification of plant taxa to family level is generally possible based 364 on the partial sequence of rbcL using the LCA approach. In addition, some OTUs in our study 365 were successfully assigned to the genus level. In order to be conservative, however, we avoided 366 assigning most OTUs to species level, unless there were matches with BOLD/NCBI database 367 sequences and site-specific reference plant sequences. As species-level plant identification based   In our study, attempts to use microscope analysis to identify plant parts in droppings 411 proved to be challenging, as no pre-existing reference collection was available. Building our own 412 microhistological reference collection for Tioman was time-consuming and labour-intensive -413 and the resulting collection often did not match up with the plant parts found in the flying fox Manuscript to be reviewed 414 droppings. However, obtaining DNA from plant specimens is still necessary to narrow down the 415 identity of OTUs to species level. Indeed, 8 out of 29 OTUs had 100% matches to the sequences 416 of plant specimens collected from the study site, highlighting the importance of building a 417 comprehensive local sequence library beforehand, preferably specific to one's particular study 418 site.

419
It is important to note that NGS did not detect Durio in the same individual droppings as 420 those identified via microscope. This is likely due to the low abundance of this plant taxon in the 421 droppings affecting detection probability, especially since the NGS analysis used a more general 422 primer that was not specific to Durio. This pollen detection probability is another caveat to be