Diet Selection by a Lizard Ant-Specialist in an Urban System Bereft of Preferred Prey Open Access

Dietary specialization is presumed to increase foraging efficiency, resulting in high net nutritional gain (Beissinger, 1990) although such specialization is a biological trade-off. Morphological, physiological, and behavioral adaptations that are advantageous for hunting, capturing, handling, and assimilating specialized prey items (Lahti and Beck, 2008) may reduce hunting efficiency and extraction rates when specialists are forced to switch prey (Beissinger, 1990; Cooper and Secor, 2007). Rapid urbanization could be detrimental to the persistence of populations of native dietary specialists because their limited dietary plasticity may preclude the use of alternative prey.

In lizards, dietary specialization may change with life stage, fluctuations in prey abundance, or patchiness in prey distribution. Specialization may occur in localized populations (Fox and Morrow, 1981; Britt et al., 2006), or be a temporary phenomenon based on gape-size limitations of juveniles that enhance foraging efficiency through the selection of small, but numerous prey (Juanes and Conover, 1994; Lahti and Beck, 2008). Diets of juveniles may become less specialized as they grow and become less gape-limited (Drummond and Garcia, 1989); however, opportunistic specialization among dietary specialists that consume small insects, in response to seasonal and temporal changes in prey availability may occur (Roper, 1994; Elmhagen et al., 2000; Sales and Freire, 2015). When stressed, plasticity can lead to diet diversification in specialists (Beissinger, 1990) or diet contraction in generalists (Roper, 1994; Elmhagen et al., 2000).

Phrynosoma spp., or horned lizards, are typically cryptic, myrmecophage specialists that are distributed throughout the western United States and Canada through Mexico and Guatemala (Sherbrooke, 2003; Sherbrooke and Schwenk, 2008). Ant specialization has resulted in adaptations in skull morphology, tongue prehension, dentition, and pharyngeal processing (Pianka and Parker, 1975; Meyers et al., 2006; Vitt and Pianka, 2007; Sherbrooke and Schwenk, 2008). The relatively large size and expansive stomachs of Phrynosoma facilitate the consumption of large numbers of small, chitinous insects that are needed to fulfill energy requirements (Pianka and Parker, 1975).

Texas Horned Lizards (Phrynosoma cornutum) exist in relatively low-density populations in xeric habitats in western and central North America (Pianka and Parker, 1975). Harvester ants (Pogonomyrmex spp.) are the chief, even exclusive (Blackshear and Richerson, 1999; Eifler et al., 2012), prey of P. cornutum; and the presence of P. cornutum and harvester ants are strongly linked (Donaldson et al., 1994; Blackshear and Richerson, 1999; McIntyre, 2003). McIntyre (2003) suggested the use of the presence or absence of harvester ants for the determination of P. cornutum habitat suitability. Similarly, P. solare will persist in habitat patches across a range of sizes and disturbance levels, provided harvester ants are present (Sullivan et al., 2014).

Phrynosoma cornutum populations have experienced declines over much of their range because of habitat loss and destruction, as well as the aggressive use of pesticides to control ant populations (Donaldson et al., 1994). Furthermore, Sow et al. (2014) indicated that reptiles in arid environments may be especially vulnerable to climate change. Dietary specialization makes P. cornutum particularly sensitive to the removal of key prey species (Newbold and MacMahon, 2009), and the fate of P. cornutum populations is intimately linked to the prevalence and abundance of their food source (i.e., specific ant populations). In California, Phrynosoma blainvillii have steadily declined with the introduction and expansion of invasive Argentine ants (Lenepithema humile; Suarez et al., 2000). Another nonnative and non-prey ant species, red fire ants (Solenopsis invicta), have steadily expanded their range in the United States, sometimes out-competing native ant species on which P. cornutum forage (Blackshear and Richerson, 1999), and are known to depredate directly on reptiles (Reagan et al., 2000; Aresco, 2004). The red fire ant invasion has led to the widespread and indiscriminate use of pesticides to control all ant populations (Donaldson et al., 1994), which has resulted in the overall decline of ant prey abundance, having a deleterious effect on P. cornutum populations.

We studied an insular population of P. cornutum that exists in isolated patches of habitat within an urban environment (Wolf et al., 2013). Body size of prey and lizard are positively correlated in Phrynosoma spp. (Rissing, 1981; Suarez et al., 2000; Suarez and Case, 2002; Lahti and Beck, 2008), and P. cornutum may avoid small ants regardless of availability (Blackshear and Richerson, 1999). Therefore, because our study site lacked harvester ants, we predicted that the diet of P. cornutum would consist of a large proportion of relatively large-bodied, native ants. Furthermore, we predicted that the relatively large female lizards would consume larger proportions of large-bodied ants when compared to males, and we expected that juvenile lizards would consume a high proportion of smaller native ants attributable to gape-size limitations relative to adults.

Tinker Air Force Base (TAFB) is a largely urban base on the outskirts of Oklahoma City (35°24′58″N, 97°24′41″W; datum NAD83; Fig. 1). Approximately 500 ha of the 2,000-ha base are natural habitat that is dominated by mixed oak-hardwood forests and a mixture of native and nonnative grasslands. Core Reserve Area 3 (CRA3; known as Wildlife Reserve 3 in Endriss et al., 2007; Wolf et al., 2014) is a natural area (approximately 15 ha; Fig. 2) on the southwestern side of TAFB, dominated by grassland with patches of woody vegetation and gravel trails and includes residential housing and industrial facilities. Dominant vegetation on CRA3 include big bluestem (Andropogon gerardii), little bluestem (Schizachyrium scoparium), plains bluestem (Bothriochloa ischaemum), indiangrass (Sorghastrum nutans), sideoats grama (Bouteloua curtipendula), Maximilian sunflower (Helianthus maximiliani), tall fescue (Lolium pratense), and eastern redcedar (Juniperus virginiana) (Endriss et al., 2007).

We collected ants at 400 stations in 20-mL scintillation vials baited with millet and peanut butter in summers 2010 and 2011. We deployed vials at 200 bait stations (Bait Station Set 1) in May and July and at the other 200 bait stations (Bait Station Set 2) in June and August. We used the Hawth's Tools extension for ArcGIS 9.3 (ESRI, Redlands, CA) to randomly select locations throughout the study site for bait stations (Fig. 2). Vials were deployed between 0600 and 1200 h for 3–4 h on a single day per month on days with no precipitation. We preserved ants in 95% ethanol and later sorted and identified them to genus under a dissecting microscope. Ants were counted, cleaned, dried, and weighed using an Accu-124 analytical balance (Thermo-Fisher Scientific, Inc., Waltham, CA) accurate to 0.0001 mg. We estimated the mean mass of a single ant in each genus by regressing the mass of each sample on the number of ants in the sample, using the resulting regression line to extrapolate the mass of 100 ants, then dividing this number by 100.

We collected P. cornutum fecal samples (N = 75) opportunistically when found on the study site. Additionally, fecal samples (N = 49) were obtained from individuals held overnight for PIT tagging (12.5 mm tags; Biomark Inc., Boise, ID); these known scat samples were used as a reference in field collections. All samples (N = 124) were hydrated with 95% ethanol and contents examined for identification. Ant heads were counted and identified to genus. Other invertebrates found in bait stations and fecal samples were identified to order.

Phrynosoma cornutum hatchlings emerge during late summer and in the fall, develop into juveniles in their second year, and reach adulthood and sexual maturity in their third year (Montgomery and Mackessy, 2003). Phrynosoma cornutum at TAFB were categorized into age classes based on snout–vent length (SVL), with adult females measuring > 63 mm, and adult males > 49 mm (Endriss et al., 2007). Yet, because juveniles collected late in the year often were the same size as adults (and presumably had similar gapes and stomach capacity), we defined size classes as follows: large lizards with SVL > 50 mm (N = 26; adult males, adult females, and large juveniles); medium lizards with SVL between 30 and 50 mm (N = 15; juveniles); and small lizards with SVL < 30 mm (N = 8; hatchlings).

We tested for differences between use and availability of prey using log-likelihood ratio tests, also known as G-tests (Zar, 1999). We conducted an overall G-test with all scat samples pooled across lizard sizes and sexes to test for a population-wide difference between use and availability. Although bait stations were placed site-wide encompassing all of CRA3, we defined availability based only on bait stations within the area occupied by P. cornutum (Fig. 2). The lizard-occupied area was calculated with a 95% kernel density estimate (Kernohan et al., 2001) for all lizard locations, using the Home Range Tools Version 1.1 (Ontario Ministry of Natural Resources, Centre for Northern Forest Ecosystem Research, Thunder Bay, ON, Canada) extension for ArcGIS 9.3 with the reference smoothing parameter. We also used a G-test to test for differences in prey use between biologically relevant pairs of subgroups using G-tests of independence, such as differences between adult males and females, or between different size classes.

We described prey use with an electivity index (Lechowitz, 1982) that was defined as u_i /a_i, where u_i is the proportion of prey genus i used out of the total proportion of prey items in all scat samples, and a_i is the proportion of genus i available out of the total sample of available prey items. We calculated electivity indices for each prey genus using the availability sample based on the area occupied by P. cornutum. Although scaling each electivity index to the overall sample often is recommended (Lechowitz, 1982), the extreme skew in our availability and use data caused a few prey genera to dominate analyses. Hence, we used unscaled electivity indices for each genus. Therefore, our indices reflect whether a given prey genus was used or not used in relation to the availability of that genus but not in relation to other genera.

Of the five most common ant genera found in lizard-occupied areas, four were found within the scats of lizards and with differing degrees of importance. We examined a total of 34,231 native ants (30–1,429 ants per sample) from 168 bait stations. In descending order of importance, the most common genera found included Monomorium (69%), Forelius (11%), Pheidole (10%), Crematogaster (7%), and Tapinoma (2%; Table 1), whereas harvester ants (Pogonomyrmex spp.) were infrequently sampled (0.0003%). During the same time frame, we collected 31,070 native ants from 124 P. cornutum scat samples. Again, in decreasing order of importance, the five most frequently encountered genera were Crematogaster (81%), Pheidole (12%), Formica (6%), and Monomorium (1%; Table 1).

Lizards consumed prey species in different proportions than were available in their environments. Log-linear tests of independence of data (G² = 60695.75, P < 0.0001) indicated that Crematogaster, Pheidole, and Formica were consumed in greater proportions than their availability and that Monomorium, Forelius, and Tapinoma were used proportionately less than their availability (Table 1). Prey use by adult P. cornutum differed between the sexes (G² = 295.58, P < 0.0001). Females consumed Pheidole ants much more frequently than did males (Fig. 3) although there was high dietary overlap of Formica and Crematogaster in males and females (Fig. 3). The three size classes of lizards differed in their use of ant genera (G² = 3.88, P < 0.0001). Hatchlings used Pheidole to a greater extent than did medium and large lizards, and Formica was used more by large lizards when compared to lizards in the smaller size classes (Fig. 4).

Phrynosoma cornutum's disproportionate use of large native ants also was reflected in an electivity index. Values for Ivlev's electivity index (E_i = [u_i – a_i] / [ u_i + a_i]) ranged from 1 (strong usage) to −1 (not used), whereas 0 indicated use in proportion to availability (Lechowitz, 1982). The most readily consumed ant prey represented the largest ant prey available at our study site, that is, Formica (E_i = 1.00), followed by Crematogaster (E_i = 0.84), and Pheidole (E_i = 0.09), whereas smaller ant species (e.g., Forelius, Monomorium, and Tapinoma) were not used (Fig. 5). Lizard size and ant mass were positively related, with a minimum ant threshold of 0.70 mg for consumption (Fig. 6).

Phrynosoma cornutum persisted on our urban study area (Wolf et al., 2014) without harvester ants, a primary and sometimes presumed exclusive prey species in other P. cornutum populations (Blackshear and Richerson, 1999; Eifler et al., 2012). Consistent with our predictions, P. cornutum disproportionately consumed large-bodied native ants. Prey consumption differed by sex and size class and was influenced by ant size, availability, and abundance.

Using biomass as a surrogate for nutritional value (Golley, 1961; Griffiths, 1975), P. cornutum focused their foraging activities on prey items that yielded the highest nutritional return, when balanced against foraging and digestive costs. Our results mirrored work on P. platyrhinos where ant consumption was regulated by encounter rate (i.e., prey abundance) and caloric intake (prey size [Newbold and McMahon, 2009]). In our study, Formica was the largest native ant and was consumed when available, even though it was seldom encountered at our bait stations, possibly attributable to low colony sizes (Fisher and Cover, 2007), inappropriate sampling (Newbold and MacMahon, 2009), or solitary foraging strategy (Trager and Johnson, 1985; King and Trager, 2007; Trager et al., 2007). Solitary foraging ant species may be more vulnerable to Phyrnosoma predation, because they are less prone to mobbing behavior that could attract predators to the normally cryptic lizard (Rissing, 1981). Whitford and Bryant (1979), however, found that P. cornutum in southern New Mexico preyed on both solitary foragers (Pogonomyrmex desertorum) and column foragers such as Pogonomyrmex rugosus. Unlike Formica, Crematogaster form populous colonies and proved to be an abundant food resource among our study lizards (Fig. 5), similar to Phrynosoma populations in the Great Basin of the western United States (Newbold and McMahon, 2009).

Gape size may be an important factor in ant selection by Phrynosoma, affecting the consumption of Formica, which was used most frequently by large lizards and least by hatchlings. Size class of lizards is positively correlated with average prey size (Powell and Russell, 1984). Horned lizards seize prey with tongues and swallow ants whole (Sherbrooke and Schwenk, 2008), and gape-size limitations may prohibit P. cornutum hatchlings from consuming large Formica prey. A similar relationship between predator and prey size was observed in P. douglasii (Lahti and Beck, 2008), P. douglasii brevirostre (Powell and Russell, 1984), P. platyrhinos (Rissing, 1981), and P. blainvillii (Suarez et al., 2000; Suarez and Case, 2002). Furthermore, adult lizards may simply be superior competitors, dominating profitable static food sources (e.g., Formica ant mounds), while excluding smaller, inexperienced conspecifics. Although Pheidole have dimorphic castes (Wilson, 2003), all size classes of lizards exploited this food resource and gape size did not appear to be a limiting factor in Pheidole consumption.

Our results also suggested a minimum threshold in prey size for persistence of Phrynosoma populations. As a cryptic species, Phrynosoma may be more vulnerable to predation by selecting small ant species that would require a large foraging effort over extended periods (Suarez and Case, 2002). Despite their high relative abundance, Monomorium and Forelius were the smallest ants at our study site and were underused by P. cornutum. Suarez et al. (2000) found that P. blainvillii did not typically use small native ants (mean = 1.11 mm body length), and Whitford and Bryant (1979) found P. cornutum used only native ants measuring >5 mm in total length. Monomorium minimum (<2 mm) and Forelius pruinosus (∼2 mm) are diminutive (Fisher and Cover, 2007), and members of both genera are known to swarm food resources and exclude competitors with chemical defenses (Adams and Traneillo, 1981; Scheffrahn et al., 1984); hence, their high numbers at our bait stations may be exaggerated. The small body size of Monomorium make it an unappealing choice for P. cornutum, whereas Forelius are thermophilic, foraging at peak environmental temperatures (Fisher and Cover, 2007) that coincides with inactivity of P. cornutum. Temporal partitioning may exclude Forelius as suitable prey, as Whitford and Bryant (1979) observed in P. cornutum and large nocturnal native ants.

Although common throughout the entire study area, Tapinoma spp. were rare at our bait stations. Fellers (1987) found Tapinoma sessile to be a consistently subordinate competitor at bait stations; thus, Tapinoma may be more abundant at our study site than we detected. Approximately 20% smaller than Crematogaster, Tapinoma may fall under a minimum ant biomass threshold below which foraging becomes unprofitable for P. cornutum.

Phrynosoma cornutum was presumed to be a harvester ant specialist (Munger, 1984; Blackshear and Richerson, 1999; Eifler et al., 2012); however, our data suggested plasticity in foraging behavior when harvester ants are not available. Management efforts to conserve or re-establish P. cornutum populations should expand their habitat suitability assessments beyond simply the presence/absence of harvester ants and, instead, consider habitat fragments with abundant medium- and large-sized native ant communities.

This study was supported by the United States Department of Defense (U.S. Air Force, Tinker Air Force Base) via the Great Rivers Cooperative Ecosystem Unit, U.S. Army Construction Engineering Research Laboratory, and Southern Illinois University–Carbondale. All fieldwork was conducted under protocols approved by Southern Illinois University–Carbondale Institutional Animal Care and Use Committee (IACUC; Protocols 11-043, 08-039, and 05-063), and permits provided by the Oklahoma Department of Wildlife Conservation. We thank lab technicians K. Lursen, A. Misner, and F. Soveg, who assisted in this research. J. Lee (TAFB), A. Suarez (University of Illinois at Champaign–Urbana), and J. Trager (Missouri Botanical Gardens) assisted with ant identification. We acknowledge the support from the Natural Resources staff of Tinker Air Force Base and the staff of the Cooperative Wildlife Research Lab at Southern Illinois University.