A model of digestive tooth corrosion in lizards: experimental tests and taphonomic implications

Corrosion patterns induced by gastric fluids on the skeleton of prey animals may depend on the nature of the corrosive agents (acid, enzymes) as well as on the composition of the hard parts and the soft tissues that surround them. We propose a framework for predicting and interpreting corrosion patterns on lizard teeth, our model system, drawing on the different digestive pathways of avian and non-avian vertebrate predators. We propose that high-acid, low-enzyme systems (embodied by mammalian carnivores) will lead to corrosion of the tooth crowns, whereas low-acid, high-enzyme systems (embodied by owls) will lead to corrosion of the tooth shafts. We test our model experimentally using artificial gastric fluids (with HCl and pepsin) and feeding experiments, and phenomenologically using wild-collected owl pellets with lizard remains. Finding an association between the predictions and the experimental results, we then examine corrosion patterns on nearly 900 fossil lizard jaws. Given an appropriate phylogenetic background, our focus on physiological rather than taxonomic classes of predators allows the extension of the approach into Deep Time.


Supplementary Note
Bones or other mineralized parts of prey are found in the feces or intestine of sharks 1,2 , actinopterygian fish 3 , African Coelacanth 4 , African 5 and Australian 6 lungfish, stem amphibians 7 , mammals 8,9 including carnivorous bats 10,11 , carnivorous squamates [12][13][14][15] , carnivorous turtles [16][17][18] , and crocodylians 19 . In these taxa, prey carcasses normally pass unidirectionally through the digestive tract. Similarly, coprolites indicate that bones were normally passed into the intestine in dinosaurs as far crownward as Tyrannosaurus rex 20 , and none of the thousands of specimens of the dromaeosaur Microraptor preserves a pellet 21 . In contrast, pellets have been documented in the troodontid Anchiornis 21 and in enantiornithine birds [21][22][23] , suggesting that this propensity arose on the branch separating Dromaeosauridae and Troodontidae + birds. Pellet production is of course well documented in numerous clades of extant birds.
In vertebrates, gastric acidity does not correlate well with proteolytic activity. For birds, Duke et al. 24 showed that diurnal raptors had relatively low gastric pH (1.3-1.8), whereas the two owls they monitored had pH of 2.2 and 2.5, closer to the pH in the gizzard of many other birds 25 .
The Barn Owl appears to have the highest pH (lowest acidity), generally greater than 3.0 26 .
Proteolytic activity in the stomach, however, is high in both diurnal raptors and owls, as noted above. Carnivorous birds usually show higher gastric proteolytic activity than herbivorous birds 24,[27][28][29] .
Among mammals, published data on even zoo animals are surprisingly meager; data on morbid or experimental animals are easier to come by. Normal beagle dogs had an average baseline pH of 1.8 30 or less 31 . Christiansen et al. 32 , while noting large differences among animals taken at random (with regard to feeding state), suggested that a stomach pH of about 2.0 was necessary for adequate digestion in their two seal species. Data in Fox et al. 33 34 ]. Proteolytic activity in the stomach, as noted above, is low in examined carnivoran mammals, including those that bracket Ferrets phylogenetically.
Intestinal reflux has been documented in a wide variety of extant birds, including members of both Palaeognathae and Neognathae 35 .

Supplementary Figure S1
Micro-computed tomography reconstruction of a cheek tooth of the pleurodont lizard Iguana iguana in transverse section. The enamel (white) is confined to the tooth tip, whereas the rest of the tooth comprises a 'dentine cone'. Courtesy: M. Wirkner (Senckenberg Research Institute) and T.

Supplementary Figure S2
Corrosion induced on teeth of Green Anoles (Anolis carolinensis) by artificial gastric solutions (Supplementary Table S1). (a-d) whole lower jaw, whole teeth, and close-ups of crown and midshaft on specimen #2 (control); (e-f) whole lower jaw and whole teeth of specimen #8; (g-j) whole lower jaw, whole teeth, and close-ups of crown and mid-shaft on specimen #25; (k-l) whole lower jaw and whole teeth of specimen #26; (m) whole lower jaw of specimen #24; (n-q) whole lower jaw, whole teeth and close-ups of crown and mid-shaft of specimen #12; (r-v) whole lower jaw, whole teeth and close-ups of crown, upper shaft and lower shaft on specimen #32; (w) whole lower jaw of specimen #29. Scale bars: left panels 1 mm, middle panels 300 µm, right panels 10 µm.

Supplementary Figure S3
Relationship between type 1 and type 2 corrosion patterns (symbols correspond to those in Fig. 1), where they occur (N = 20), and acid-enzyme conditions in the experiments with simulated gastric fluids. Where two specimens would plot at the same position, the symbols are separated and the position indicated with arrows. Type 2 corrosion is concentrated in the low-acid (pH ≥ 2.5), highenzyme (pepsin concentration ≥ 3.0 mg/ml) area of parameter space. Type 1 corrosion occurs more broadly but is the only pattern found at high-acid (pH ≤ 2.5), low-enzyme (pepsin concentration ≤ 3.0 mg/ml) area of parameter space. Where type 1 corrosion occurs in the low-acid, high-enzyme area, it is after very long exposure times (20 h).

Supplementary Figure S4
Corrosion damage to teeth of lizards ingested by a Barn Owl (Tyto alba). (a) whole lower jaw and close-up of Bridled Skink (Trachylepis vittata) from pellet #1239; (b) whole lower jaw and close-up of undetermined gecko from pellet #652. This "type 3" pattern of damage, in which the neck between the crown and shaft is preferentially corroded was not predicted by the model. Scale bars: whole jaws, 1 mm; close-ups, 0.5 mm.   Summary of corrosion damage to teeth of Anolis carolinensis exposed to artificial gastric fluids (without special treatments like shaking). Among them, five corresponded to one physiological end-member (high-acid, low-enzyme), and three corresponded to the other physiological endmember (low-acid, high-enzyme). The missing specimen in the right two columns for the first endmember showed approximately equal damage to crown and shaft.

Supplementary Table S5
Summary of corrosion damage to teeth of pleurodont lizards extracted from owl pellets. The parentheses indicate that while present, the corrosion was restricted to a very small number of the teeth in one of the jaws of the individual.

Supplementary Methods
As a control for the experiments, one anole carcass was macerated in 50 ml of de-ionized water (initially, pH = 7.0) for 7 days. Because even in this pure-water control treatment the fluid chemistry will change as a result of decomposition, a second control was prepared using dermestid beetles without exposure to any fluid. Controls were selected at random.
Our conceptual framework has two active components, HCl and enzymes.  Table S1). Experimental treatments of short duration (<4 hr) under high-acid (pH ≤ 2.0), low-pepsin (≤ 3 mg/ml fluid) conditions were meant to mimic mammalian carnivores, which represent the high-acid, low-enzyme physiologic end-member.
Gastric emptying times are the relevant expression of the duration of predigestion, since pH rises toward neutrality and pepsin activity ceases in the intestine. Half-emptying times vary greatly in dogs 34 , partly in dependence on diet, although it averages ~3 hr, with total emptying times of 4-6 hr 55 . For cats, a wider range has been documented for half-emptying times 34 , but durations of 4-6 hr also seem typical 56 . Although little data is available on wild or other carnivores, and both dogs and cats are domesticated species, they phylogenetically bracket the clade Carnivora 57,58 .
Experimental treatments of long duration (10-20 hr) under low-acid (pH ≈ 2.5), high-pepsin (≥ 5 mg/ml) conditions were meant to mimic owls, which represent the low-acid, high-enzyme physiologic end-member. Meal-to-pellet intervals (MPI), equivalent to gastric emptying times for mammalian predators, are much longer. Under a variety of conditions, MPI in examined owls species was usually >10 hr 59 and frequently up to 20 hr 60,61 . It may also be noted that in diurnal raptors, MPI of around 20 hr is the norm 24,60,62 . Thus, although MPI varies up to twofold among birds of prey, gastric residence time is always long compared to mammalian carnivores. This characteristic appears to be apomorphic of raptors, given the very short passage times for food in the digestive tract seen in many frugivorous and insectivorous members of Neoaves and Eight specimens were subjected to more extreme conditions to study other parameters.
Seven of these were placed in solutions with a pepsin concentration of 20 mg/ml (Supplementary   Table S1), considerably higher than has been documented in living predators. Four were cut into three pieces (at the neck and anterior to the pelvis) to simulate prey dismemberment. The flasks of six specimens were placed in a shaking water bath at an intentionally extreme rate of 120 revolutions per minute (rpm), where the temperature of the water bath was set to 40ºC, to simulate strong gastric motility. Notably, however, gastric motility is considered minor in owls, which is consistent with their weak gizzard walls 63,64 .
One dentary from each experimental and control specimen was extracted and studied using a binocular microscope. They were then affixed to stubs, coated in a gold-paladium alloy, and examined using a scanning electron microscope (SEM) at the Senckenberg Research Institute in Frankfurt. Qualitatively, we assessed whether the corrosion patterns corresponded to the predicted patterns, type 1 or type 2, and whether any other, unpredicted patterns emerged. Quantitatively, we counted in how many specimens corrosion was greater to the crown of the teeth than the shaft, and vice versa, as this anatomical distinction informed the conceptual model.

Supplementary Data
A gradual rise in the pH was observed over the course of all treatments in the laboratory experiments. Especially for high-pH, extended-time treatments, these results and published activity curves for swine pepsin suggest that it would have become inactive later in the treatments.
However, since pH units are logarithmic, the pepsin may have still been active for most of the course of each treatment, not a small fraction.
Generally speaking, the poor digestion of the carcasses by the pepsin even at extremely high (biologically unrealistic) concentrations is noteworthy. The experimental specimens remained intact after exposure to the simulated gastric fluids. In some cases, the tongue, oral mucosa, throat and body wall exhibited extensive degradation. In one dismembered specimen that was weighed also at the end of the treatment, carcass mass was considerably reduced (by 35%), indicating that direct access to the body cavities aided the efficacy of the pepsin. Still, it is clear that mere exposure of a lizard carcass, intact or not, to pepsin will be insufficient to de-flesh it completely before the bones would be ejected as a pellet. This result suggests the importance of post-gastric enzymes for digestions in bird of prey.
The enamel surface of both the crown and shaft is smooth in the control specimens ( Supplementary Fig. S2a-d). Deep on the shaft, the surface is smooth or has a fine-scale texture consisting of low, rounded mounds ( Supplementary Fig. S2d).

High-acid, low-enzyme solutions
Solutions representing the high-acid, low-enzyme end-member covered pH ≤ 1.5 and pepsin concentrations of 1 mg/ml (Supplementary Table S1). These solutions produced significant corrosion to the lizard teeth, especially to the crowns. In many cases where specimens were exposed to these conditions for 2 hr or more (#21, 22,24,26,27), the crowns of the teeth above the gumline were completely decalcified, leaving at most the brown pliable protein matrix behind ( Supplementary Fig. S2k-l). This supports the hypothesis that the gingiva plays a protective role against corrosion of the tooth shaft by acid. As the cusps are principally composed of enamel, little crown morphology remained. The shafts of the same teeth, while sometimes appearing slightly corroded, are always much better preserved than the crowns. One specimen (#25), exposed to a solution with pH = 1.2 and pepsin = 1 mg/ml for 2 hr, produced considerable corrosion to the tooth crowns, which had been stripped away in patches ( Supplementary Fig. S2g-j).
Carcasses spending more than a short time in a very low-pH (~1.0) solution were heavily corroded, as expected given the findings of Fernández-Jalvo and Andrews 65 . Even 3 hr at such low pH was sufficient to decalcify the bones and teeth (remove the hydroxyapatite), leaving a brown, pliable structure probably made of a collagen matrix that curled upon drying (#24, 27). In a specimen (#23) that spent 18.5 hr in a pH = 1.0 solution, the mineral substance of the entire skeleton was destroyed ( Supplementary Fig. S2m).
Corrosion to the teeth that closely corresponds to the prediction "type 1" was observed in three (#21, 24, 26) of the five specimens specifically considered to represent the physiological endmember ( Supplementary Fig. S2k-l; Supplementary Table S1). Such corrosion was also observed in 11 other specimens, especially in the high-acid, low-enzyme area of parameter space. Note that we consider specimens in which the crown has been completely decalcified as representing type 1 corrosion, even if the protein matrix remains, because the matrix would quickly be destroyed in the diagenetic environment 9,19 . More generally, corrosion damage to the crown was greater than that to the shaft in four specimens, whereas the reverse was true in no specimen (Supplementary Table S4).

Low-acid, high-enzyme solutions
Solutions representing the low-acid, high-enzyme physiological end-member covered pH ≈ Table S1). The digestion was also allowed generally allowed to proceed for a longer time (10 or 20 hr). These solutions produced corrosion on the teeth of all of the specimens (Supplementary Table S4), but the distribution of the corrosion was different. In two of the specimens (#11, 12), the shaft was corroded but the crown was unaffected ( Supplementary Fig. S2n-q).

and pepsin concentrations = 5 mg/ml (Supplementary
Corrosion roughly corresponding with type 2 was induced on a total of 6 specimens, all of which are concentrated in the low-acid, high-enzyme area of parameter space ( Supplementary Fig.   S3). Three of these were subjected to special treatments like shaking, however (Supplementary  Fig. S2) yields positive (favoring type 2 over type 1) regression coefficients with significant P-values for pH (P = 0.00028) and pepsin concentration (P = 0.024), as predicted by the model.
In particular, specimens #5, 32 and 33 showed considerable corrosion of the shaft that extended up onto the crown ( Supplementary Fig. S2r-v). The enamel was partly removed from the crown, but apically (around the cusps) it is preserved and uncorroded. Thus, a strong step is present between uncorroded enamel and corroded dentine. Specimen #34 shows greater damage to the crown than the shaft. By comparison with #32 and 33, it is likely that the enamel had completely spalled off the crown, such that corrosive damage to the crown exceeded that to the shaft.

High-acid, high-enzyme solutions
It is expected that, since both acid and pepsin are known to corrode bony tissues, high-acid, high-enzyme conditions will produce greater corrosion than either of the end-members examined above. The teeth of three specimens (e.g., #19) exposed to such conditions were completely destroyed down to the bone, which does not itself appear corroded to the naked eye ( Supplementary   Fig. S2w). No other specimen did. It is likely that once the pulp cavity of the tooth has been breached, corrosive fluid may penetrate to the base of the tooth and destroy it from the inside without significantly affecting the tooth-bearing bone.

Low-acid, low-enzyme solutions
Only two treatments used low-acid, low-enzyme solutions: #7 and #8 (Supplementary Table   S1). The latter ran for 20 hr. Here, the enamel on the crown above the gumline is distinctly eroded in a shallow patch at the lingual base of the central cusp between the accessory cusps ( Supplementary Fig. S2e-f). The former, which ran for half the time (10 hr), shows a more advanced state of the same corrosion, where the corroded patches have expanded over much of the crown. In some places the enamel caps of the cusps have been entirely destroyed. In neither case do we see broad surfaces of enamel where corrosion is indicated by fine pitting over a large area; rather, corrosion is more discrete. Corrosion to the shaft is not seen in either.

Influence of dismemberment and shaking
It was expected that shaking would increase fluid access to the reaction areas on the tooth or gingiva and so would lead to greater corrosion. Three specimens (#33-35) were subjected to shaking and/or dismemberment but otherwise identical conditions: pH = 2.7, pepsin concentration 20 mg/ml, duration 10 hr (Supplementary Table S1). The specimen that was merely dismembered but not shaken (#33) indeed showed considerably less total corrosion than the specimens that were shaken, or dismembered and shaken. Due to low sample sizes, no definite conclusions can be drawn about the influence of shaking (and hence gastric motility).