Copy number variation analysis

We chose our panel of mammalian species based on their phylogeny, diet preference (broad vs. specialized), domestication status, and commensal relationship with humans. Overall, we compiled 204 DNA samples from 46 different species and 127 saliva samples from 22 different species. Detailed information about the samples used in this study and their sources can be found in Supplementary file 1. Briefly, DNA from various animals was collected using buccal swabs (PurFlock, Puritan Medical Products), saliva samples, or museum specimens (dried blood and tissues, Museum of Southwestern Biology, Division of Genomics Resources). Saliva samples were collected by suction using commercially available devices containing absorbent sponges in a syringe-like receptacle (Super-SAL and Micro-SAL, Oasis Diagnostics, Vancouver, WA) unless otherwise specified. DNA was extracted from swabs using a commercially available kit (ChargeSwitch gDNA Buccal Cell Kit, Invitrogen). For saliva samples, we used a commercially available extraction kit (BioWorld, Dublin, OH). Detailed information about sampling strategies employed for each species can be found in Collection of saliva samples section below.

Digital droplet PCR (ddPCR) was used to experimentally determine amylase gene copy numbers. If reference genomes were available for a given species, primers were designed specifically for use in these species. For species where reference genomes were unavailable, amylase coding sequences were chosen for primer design that were confirmed to be conserved in the two most closely related species. Further details about primer design and strategy are described in Primer design for digital PCR section below. The primer sets used for each species are listed in Supplementary file 2.

Translated amino acid sequences of the amylase gene copies were downloaded from NCBI reference genomes. Sequences were aligned and a phylogenetic output was generated using a custom Python code as described previously (https://github.com/duoduoo/VCFtoTree) (Xu et al., 2017; Pajic et al., 2016). We constructed a phylogenetic tree from the protein sequences by Randomized Axelerated Maximum Likelihood (RAxML) (Stamatakis, 2014) using the LG substitution model (Le and Gascuel, 2008), bootstrapping it with 1000 replicates for branch support. Visualization was performed using the FigTree software (Rambaut, 2012).

Previous work utilized lineage-specific retrotransposons to estimate the timing of amylase gene duplications and to distinguish between salivary and pancreatic amylase genes in humans (Samuelson et al., 1990). Using this approach, the salivary AMY1 gene could be traced back to a great ape ancestor (Samuelson et al., 1990). Later studies used the sequence of the great-ape specific retrotransposon to label AMY1 gene copies in humans by fiber-FISH (Perry et al., 2007). Building upon these findings, we searched 5 kb upstream and downstream of the amylase copies in mouse, rat, pig, and dog reference genomes for the existence of lineage-specific retrotransposition markers. Specifically, we searched for relatively recent L1 elements having sW scores of more than 1000 and being located at nearly identical distances to the 5’ or 3’ ends of amylase gene copies. Using this approach, we detected a relatively small number (less than 10) of distinct L1 retrotransposons in each species (Figure 2B). To ensure that these retrotransposons were duplicated specifically in the amylase locus, we conducted a BLAST analysis to search for the existence of these same retrotransposons outside of the amylase locus. We found that the retrotransposons within the amylase locus are highly similar to each other (>90%) and we found no similarly close matches for these retrotransposons elsewhere in the reference genomes of these species (Supplementary file 3). Next, we verified that these retrotransposons were indeed lineage-specific by showing that there were no close matches in reference genomes of other species. The most parsimonious explanation for our observations is that these L1 elements inserted into the proximity of an amylase gene copy and then duplicated along with additional copies of that gene, thereby suggesting lineage-specificity of duplication events.

We used two different methods to measure enzyme activity of amylase in saliva. First, we conducted a direct estimate of enzyme activity using a traditional starch lysis agar plate assay following a previously described protocol (Kilian and Nyvad, 1990). In brief, holes were punched in a starch-containing agar and filled with saliva. After 24 hr incubation at 37°C, the undigested starch remaining in the agar was stained with iodine and the diameters of the lysed clear rings were measured. Enzymatic activity was extrapolated from serial standard dilutions of purified α-amylase from human saliva (Sigma) measured in the same assay. In parallel, we measured the samples using a high-sensitivity (detection limit 2 mU/ml) colorimetric in-solution assay (EnzCheck Ultra Amylase Assay Kit, Invitrogen) following the manufacturer’s protocol with the same human α-amylase as the standard. Concentrations of total protein in saliva were determined by the bicinchoninic acid (BCA) assay (micro-BCA, BioRad) using bovine serum albumin as the standard. Optical density measurements were performed using a Nanodrop 2000 spectrophotometer (Thermo Fisher). Amylase activities were calculated as units of enzymatic activity normalized per mg of total salivary protein.

All input data used for creating the main figures are provided in Supplementary file 1. Information about the dietary preferences of individual species was acquired from extensive literature research presented in Categorization of starch consumption section below. All figures were produced using the R statistical package (https://www.r-project.org/). For calculating the independent phylogenetic contrasts shown in Figure 4C, we used the approach outlined by Felsenstein (1985). For this analysis, we used the subset of species available through the Hg19 100way conservation alignment (http://genomewiki.ucsc.edu/index.php/Hg19_100way_conservation_alignment). Using the phylogenetic distance provided in this dataset, we first normalized the differences in amylase gene copy number and salivary enzyme activity between any two species by the square roots of the phylogenetic distance between them. Using these normalized values and applying the non-parametric Kolmogorov–Smirnov test, we tested the null hypothesis that the phylogenetic contrasts between species consuming a specialized diet is not different from the phylogenetic contrasts between species consuming the different types of diet. Among the species consuming a broad-range diet, we further tested that the phylogenetic contrasts between high and moderate starch consuming species are not different from those between species consuming moderate levels of starch.

Saliva samples and buccal swabs from deer mice (Peromyscus spp.) were provided by Danielle Garneau (SUNY Plattsburgh). Mice were trapped in the wild by Sherman live traps (Garneau et al., 2012). After restraint by scruffing behind the neck, a glass capillary tube was introduced to the animal's mouth and was moved about the lower lip and cheeks to collect saliva. The tube was introduced at an angle such that gravity would help draw down the sample into the tube. The capillary tube was placed in an Eppendorf tube and a pipet pump was used to force air to drive the rest of the sample from the capillary tube into the Eppendorf tube for storage at −20°C and shipment on dry ice.

Saliva from house mice (laboratory strain C57BL10/SNJ) was kindly provided by Jill Kramer (University at Buffalo) using a collection procedure as previously described (Kiripolsky et al., 2017).

Saliva from woodrats was kindly provided by Michelle Skopec (Weber State University). To collect saliva, woodrats were scruffed and Micro-Sal collection (Oasis Diagnostics, Vancouver, WA) devices were placed in their mouths. The woodrats were allowed to chew on the absorbent sponge part of the device and, then, their tongues and cheeks were swabbed to retrieve residual saliva. Collection devices were centrifuged and saliva samples were stored at −20°C before shipment on dry ice.

Saliva from Long Evans hooded rats was kindly provided by Ann-Marie Torregrossa (University at Buffalo). As described previously (Martin et al., 2018; Torregrossa et al., 2014), rats were conditioned to salivate when a pipette was inserted into the mouth and saliva was collected. Approximately 50 µl saliva was retrieved by suction from below and around the tongue where it pools naturally.

Saliva from dogs, cows, sheep, goats, horses, pigs, and giant African pouched rats was provided by Erin Daugherity and Luce E. Guanzini (Cornell University). Animals were not allowed to eat or drink prior to the collection to ensure the oral cavity was free of food and other debris. Saliva from giant African pouched rats was collected opportunistically while animals were anesthetized for an unrelated clinical procedure. The collection was performed using a commercially available device (Micro-Sal, Oasis Diagnostics). Large animals were gently restrained and a larger collection device (Super-Sal, Oasis Diagnostics) was placed under the tongue for up to three minutes, or until fully soaked. Devices were stored at −20°C before shipping on dry ice.

Saliva from female wild boars and castrated domestic pigs was provided by Anja Globig (Friedrich-Loeffler-Institut, Insel Riems - Greifswald, Germany). For the collection of saliva a commercial collection device, consisting of an absorbent cotton swab in a tube, was used (Salivette, Sarstedt, Nümbrecht, Germany). The swab was inserted in the animal’s mouth and fixated with a forceps until it was drenched with saliva. After placing the swab back in the tube, saliva was extracted by centrifugation. Samples were lyophilized before international shipping.

Saliva from wolves was kindly provided by Karen Davis (Wolf Park, Battle Town, IN). The wolves housed in this facility are well socialized, which allowed saliva collection by inserting Super-Sal (Oasis Diagnostics) devices into the mouths of adult wolves willing to participate. Swabs were kept in the animals’ mouths as long as they would tolerate it or until fully soaked. Samples from juvenile wolves could be collected while they were resting by inserting the swabs into their mouths. Samples were stored at −20°C before shipment on dry ice.

Saliva from dogs was kindly provided by Barbara McCabe (Buffalo, NY). Samples were obtained from diverse breeds of dogs including Boxers, Pitbulls, Golden Retrievers, and Labradors, along with several mixed breeds (see Supplementary file 1 for details). Super-Sal devices (Oasis Diagnostics) were placed in the mouth of dogs for 1–5 min, or until swab was damp. The swabs were stored at −20°C until transfer to our laboratory.

Saliva from Ring-tailed Lemur samples was kindly provided by Erin Ehmke (Duke Lemur Center). Samples were collected using commercially available absorbent strips (SalivaBio Children's Swabs, Salimetrics, Carlsbad, CA). Saliva-soaked swabs were immediately centrifuged and the collected saliva was frozen at −80°C and shipped on dry ice.

Saliva from humans was collected by passive drooling following the protocol approved by the University at Buffalo Human Subjects IRB board (study # 030–505616). Informed consent was obtained from all human participants. Saliva from chimpanzees and gorillas was collected in a noninvasive manner following the protocol approved by the University at Buffalo IACUC committee (IACUC ID# AR201800024). Chimpanzees were trained by the caretaker to voluntarily expectorate into a plastic cup. Gorilla (Western lowland gorilla) saliva was collected by the animal caretakers with a soft disposable plastic Pasteur pipette (VWR, Radnor, PA) from individuals who were previously trained to open their mouth upon request. Saliva from Rhesus macaques was provided by the Southwest National Primate Research Center, San Antonio, TX, and by the Yerkes National Primate Research Center, Atlanta, GA. All samples were immediately transferred into a polypropylene tube chilled on ice. Aliquots were stored at −80°C and shipped on dry ice.

All input data used for creating the main figures are provided in Supplementary file 1. Information about the dietary preferences of individual species was acquired from the Michigan Animal Diversity Web (https://animaldiversity.org/), unless other studies were cited. With regard to starch consumption, the literature was limited. Thus, we undertook the following steps to construct a categorization of starch consumption among the species that we used in our analysis presented in Figure 4A and B.

Based on the information available on Michigan Animal Diversity Web, we first identified animals with specialized diets (carnivores: cat, polar bear, cougar, and wolf; herbivores: sheep, bison, snow sheep, cow, goat, horse, bighorn sheep, ibex, yak, wild goat, zebra, sheep, and donkey). We assumed that starch makes up a negligible percentage of these animals’ diet.

For the animals with broad-range diets, which presumably have considerable starch content in their diet, we conducted a wider literature research. Most information about starch consumption is available for present-day human populations (Bright-See and Jazmaji, 1991), and it has been suggested that humans consume a higher percentage of starch in their diet than great apes (Perry et al., 2007; Zohary et al., 2012). Indeed, chimpanzees and bonobos primarily consume ripe fruits (poor in starch), and in the scarcity of ripe fruits, they prefer piths (also poor in starch) (Hohmann, 2009; Wrangham et al., 1998). Orangutans and gorillas primarily consume ripe fruit and leaves (Hohmann, 2009). However, it was suggested that they also consume seeds and cambium, an observation that led Janiak (Janiak, 2016) to argue that gorillas and orangutans have a relatively higher starch content in their diets than chimpanzees and bonobos.

Old World monkeys show remarkable diversity in their diets. Specifically, cercopithecines (represented in our dataset by baboons, rhesus macaques, vervet monkeys, green monkeys, and Allen’s swamp monkeys) consume considerable amounts of unripened fruit (higher starch content [Lambert, 1998]), especially when no ripe fruit is available (Mau et al., 2010; Wrangham et al., 1998). In contrast, most colobinae (represented by the colobus monkeys in our dataset) are primarily leaf-eating and, thus, likely have little starch in their diet (Oates, 1994).

New World monkeys (represented by capuchin monkeys, owl monkeys, and marmosets in our dataset) also show diversity in their likely starch consumption. Capuchin and owl monkeys primarily consume fruits, even though they supplement their diets with flower foraging and insects (Lambert, 2017; Kinzey, 1997), which likely indicates starch consumption similar to chimpanzee and bonobos. Marmosets differ in their dietary habits as their primary food intake comprises gum and other exudates (which are not starch sources) from various trees and vines, and scarcely involve fruits (Soini, 1982). Lemurs (represented by ring-tailed lemur in our dataset) have been reported to primarily consume Tamarind fruit, which is rich in non-starch polysaccharides (Gould et al., 2003). Overall, primates depend on a wide variety of food sources, including starch-based foods. However, humans are the only primate species consuming a diet unusually high in starch content.

As for rodents, we first considered species which are primarily human-commensal (represented by house mice as well as by brown and black rats in our dataset). These species have considerable variation in their diets depending on the ecological context they are living in. For example in the wild, house mice have been reported to eat primarily insects and seeds, the latter containing significant amounts of starch (Badan, 1986; Roux et al., 2002). However, house mice normally live in human-influenced habitats and consume agricultural grains and other starch-rich human-produced food and human food leftovers (Clark, 1982; Gardner-Santana et al., 2009; Hulme-Beaman et al., 2016; Pocock et al., 2004; Schein and Orgain, 1953; Singleton et al., 2003). Other rodents with less human-commensal interactions (represented by birch mouse, deer mouse, pouched rat, and woodrat in our study) also consume diverse diets, including insects, seeds, grains, and flowers (Ajayi, 1977; Baker, 1991; Everett et al., 1978; Juskaitis, 2000). However, their access to starch-rich grains and seeds is seasonal, and these foods are not necessarily their primary caloric source.

The other mammals with broad-range diet in our dataset were dogs, pigs, boars, and bears. Brown and black bears eat a wide range of foods including leaves, fruits, grains, as well as meat from hunting or scavenging (Bojarska and Selva, 2012; Graber and White, 1983; Torgersen et al., 2001). However, the same studies documented that starch-containing foods, such as grains, make up only a small portion of the bears’ diet. Boars also have a diverse preference in their diet, including mushrooms, roots, fruits, and insects. However, unlike bears, boar diets include substantial amounts of roots and tubers (Baubet et al., 2004; Massei and Genov, 2004) and, if available, human agricultural crops (Herrero et al., 2006). A close relative of the boar, the domesticated pig thrives primarily on human-produced starch-rich food sources, such as corn or potatoes. Overall, pigs and boars have higher starch content in their diets than bears. In fact, their diet can be comparable to early human diets (Hatley and Kappelman, 1980; Miller and Ullrey, 1987). Another mammal consuming a broad-range diet that was included in our study was the dog, which has been discussed within the context of recent adaptation to human-derived starch-rich diets (Arendt et al., 2016; Axelsson et al., 2013). Based on this literature, we presume that dogs, pigs, and boars have higher starch content in their diet, while starch makes up a smaller portion of the diet of the bears.

Overall, all the mammals consuming a broad-range diet mostly have considerable levels of starch in their diet. However, our literature search indicates that humans, pigs, boars, dogs, mice, and rats (both brown and black) stand out in that their diet predominantly depends on starch-rich foods (grains, roots, and tubers). Thus, we grouped them under the ‘higher starch’ consuming category whereas we grouped the other species under the ‘lower starch’ consuming category (Figure 4A and B).

For digital droplet PCR experiments, we used two primer/probe sets. One targeted the amylase copies and the other targeted a conserved ‘reference’ sequence, which was found to be a single haploid copy in mammals with known reference genomes (SRSF7 gene). For the reference sequence, we used a primer/probe set that targets one of the exons of the SRSF7 gene. The sequence is highly conserved across species and unique (i.e. a single haploid copy) in all mammalian reference genomes we investigated. We have checked that the sequence is 100% conserved in species that we considered and for which reference genomes were available from the UCSC genome portal. To capture as many amylase gene copies as possible, we carefully designed primers and probes for each species where a reference genome was available to match (100% as assessed by BLAST alignment) all of the reference amylase copies. Primer and probe sequences are listed in Supplementary file 2. It is possible that we underestimated or missed some of the amylase copies that are not represented in the reference genomes. However, digital PCR is robust to 1–2 mismatches and in most species, ddPCR results were highly concordant with copy number estimations based on BLASTx and BLASTp analysis (Figure 1—figure supplement 1, Supplementary file 2). Therefore, we surmise that the main trends we observed in this study are reliable.

To decide which primer/probe sets to use for species where no reference genome was available, we designed primer/probe sets that work in the phylogenetically most closely related species for which reference genomes were available. For example, for zebra, we used a primer/probe set designed for the horse reference genome. To ensure that this primer/probe set was appropriate, we first made sure that it also worked in the donkey reference genome. As such, we surmised that our approach should work unless there is rapid, species-specific sequence divergence in zebra as compared to horse and donkey. An analogous approach was used for all the other species for which reference genomes were not available (Figure 1—figure supplement 1, Supplementary file 2, for substitute species genomes chosen). Of course, this approach might be prone to undercalling the number of amylase gene copies in species where reference genomes are not available. Although we are confident about these estimates, none of the major conclusions of this study depends on data from such species.