ALTERNATE FOOD-CHAIN TRANSFER OF THE TOXIN LINKED TO AVIAN VACUOLAR MYELINOPATHY AND IMPLICATIONS FOR THE ENDANGERED FLORIDA SNAIL KITE (ROSTRHAMUS SOCIABILIS)

A novel neurologic disease, avian vacuolar myelinopathy (AVM), was described after an epizootic in Bald Eagles (Haliaeetus leucocephalus) and American Coots (Fulica americana) at DeGray Lake, Arkansas, US, in the winters of 1994–96 (Thomas et al. 1998). It has now been implicated in the deaths of over 170 Bald Eagles, thousands of American Coots, and at least six other species of waterbirds in five southeastern states (Rocke et al. 2002; Augspurger et al. 2003; Fischer et al. 2003). Birds affected by AVM exhibit a characteristic loss in motor coordination that can cause death in affected birds that become unable to swim, fly, or obtain food. These behaviors are consistent with damage to the central nervous system, and this is the only consistent microscopic abnormality apparent in diseased birds. Neurospinal damage manifests as vacuolation of white matter in the brain and spinal cord, known as intramyelinic edema, and is usually most severe in the optic lobe (Thomas et al. 1998). Field monitoring and sentinel studies demonstrated that the disease was site-specific and seasonal, because diseased birds are observed and collected only during the winter (Rocke et al. 2002; Augspurger et al. 2003; Fischer et al. 2006). Initial field and laboratory studies failed to link AVM to any known disease agents or anthropogenic toxicants associated with water, sediments, or vegetation in these locations (Thomas et al. 1998; Dodder et al. 2003; Rocke et al. 2005).

Comprehensive surveys revealed that all AVM sites are dominated by dense infestations of one of three types of invasive exotic submerged aquatic vegetation (SAV): hydrilla (Hydrilla verticillata), Brazilian elodea (Egeria densa), and Eurasian watermilfoil (Myriophyllum spicatum) (Wilde et al. 2005). Additionally, SAV in each site serves as a substrate for the newly described genus and species Aetokthonos hydrillicola (Wilde et al. 2014). Aetokthonos hydrillicola dominates the epiphyte community on the leaves of SAV in AVM sites but is rare or not present in sites where AVM deaths have not occurred (Wilde et al. 2005, 2014; Williams et al. 2007). Feeding studies have implicated food chain transfer (via ingestion of the gastrointestinal tract of diseased birds) of a toxin produced by a cyanobacterium associated with submerged aquatic vegetation growing in waterbodies where AVM-positive birds were previously diagnosed (Fischer et al. 2003; Birrenkott et al. 2004; Lewis-Weis et al. 2004; Rocke et al. 2005; Wilde et al. 2005, 2014). Thus, the working hypothesis of all ongoing AVM investigations is that waterbirds develop AVM by ingesting SAV and A. hydrillicola either directly or via predation on other birds that have recently fed on this material.

While feeding trials indicate that fish and reptiles fed hydrilla collected from AVM sites develop vacuolar lesions, it has not been demonstrated that they can transmit the AVM toxin to their predators (Haynie et al. 2013; Mercurio et al. 2014). In a field and laboratory study, triploid grass carp (Ctenopharyngodon idella), which are frequently used as a biological control agent for dense SAV, were fed hydrilla from an AVM site. The grass carp developed brain lesions, but chickens that consumed those grass carp tissues in the lab did not develop AVM (Haynie et al. 2013). Possible transfer of the AVM toxin to birds through herbivorous invertebrates has not been investigated.

In 2009, we identified colonies of A. hydrillicola growing on hydrilla in Lake Tohopekaliga (Toho), in Kissimmee, Florida, US, and confirmed its identity with modifications to a previously validated PCR protocol (Williams et al. 2007; Wilde et al. 2014). In a 2007 survey of cyanobacterial epiphytes in 47 Florida lakes that included Toho, A. hydrillicola was confirmed by PCR only in Lake Huntley, a small lake approximately 160 km south of Toho (Williams et al. 2009). Toho supports dense infestations of hydrilla and an exotic herbivorous snail, the island applesnail (Pomacea maculata), previously Pomacea insularum (Hayes et al. 2012,), which is known to consume large amounts of hydrilla (Baker et al. 2010). Toho also supports a nesting population of the federally endangered Florida Snail Kite (Rostrhamus sociabilis). Fewer than 500 breeding pairs of Snail Kites remain in Florida, and a large percentage of the remaining birds nest in the Kissimmee Chain of Lakes, especially Toho (Bowling et al. 2012; Reichert et al. 2012). The Florida Snail Kite historically fed almost exclusively on the native Florida applesnail (Pomacea paludosa) but now readily consumes the larger exotic P. maculata (Sykes 1987; Darby et al. 2007; Cattau et al. 2010). Although AVM has never been documented at Toho, the presence of A. hydrillicola may pose another threat to Snail Kites if their primary food source, the exotic apple snail, accumulates AVM toxin.

Laboratory studies have confirmed that AVM toxin can be transferred up the food chain from herbivorous waterbirds to predatory raptors (Fischer et al. 2003; Birrenkott et al. 2004; Lewis-Weis et al. 2004). We predicted that Snail Kites would be susceptible to the toxin that causes AVM similar to Bald Eagles, Red-tailed Hawks (Buteo jamaicensis), and Great Horned Owls (Bubo virginianus) that developed AVM lesions after consuming prey from an AVM site (Lewis-Weis et al. 2004). Because the AVM toxin is uncharacterized, an avian bioassay is the only available method to illustrate toxicity of SAV material containing A. hydrillicola. Through a series of laboratory chicken feeding trials, we tested the hypothesis that the AVM toxin can be acquired via an alternative food web: from herbivorous island applesnails to their avian predators. We conducted an initial feeding trial using only hydrilla with or without A. hydrillicola to ensure that the AVM toxin was present in the treatment hydrilla collection. We then investigated whether island applesnails feeding on the hydrilla/A. hydrillicola accumulated a sufficient amount of toxin to cause AVM in a predatory bird model. The final feeding trial tested whether hydrilla collected directly from Toho induced lesions in avian feeding trials. Our overall objective was to evaluate whether the already imperiled population of Florida Snail Kites is at risk of developing a fatal neurologic disease if this toxin transfer pathway is valid and the AVM toxin exists in the Snail Kite's increasingly limited range.

Although other species of native and invasive SAV can host A. hydrillicola, we used hydrilla because this is the dominant vegetation in all sites where AVM bird deaths have been documented since 2000 (Wilde et al. 2005). Because we propose that the toxin is only associated with the combination of SAV and A. hydrillicola, material containing both is considered “treatment hydrilla” and hydrilla lacking the A. hydrillicola, where disease has never been documented, is considered “control hydrilla.” Treatment hydrilla was collected on 8 December 2010 from J. Strom Thurmond Reservoir (JSTL), South Carolina (33°42′22″N, 82°20′35″W), a 28,700-ha reservoir on the Savannah River with a history of yearly AVM epornitics since 1998 (Fischer et al. 2006). The specific collection site on Lake Thurmond was selected on the basis of an initial survey of relative density of A. hydrillicola. Control hydrilla was collected from Lake Hatch (Lake Hatchineha, Florida; 28°1′18″N, 81°24′19″W), a 2,697-ha lake where no AVM deaths have ever been documented and A. hydrillicola has not been detected. Hydrilla was also collected from 30 sites on Lake Toho (28°13′7″N, 81°23′48″W) in February 2012. Larger collections were then made from the sites with the highest density of A. hydrillicola colonies for the avian bioassay. Hydrilla from all sites was collected at depths of 0–1 m using a throw rake, kept on ice while transporting to the laboratory, and frozen at −20 C for use in the snail and avian feeding trials. Random samples of hydrilla leaflets from each collection site were examined under light microscopy for A. hydrillicola colonies to confirm that treatment hydrilla contained A. hydrillicola and that control hydrilla did not. Subsamples of hydrilla to be used in the feeding studies from both sites were validated with a PCR probe recently modified from a previously published method for detecting A. hydrillicola (Williams et. al 2007; Wilde et. al 2014).

Approximately 150 adult island apple snails were collected from a small pond in St. Mary's, Georgia (28°13′7″N, 81°23′48″W) and transported to the University of Georgia Whitehall Fisheries Laboratory, Athens, Georgia, in plastic bins filled with continuously aerated water. Snails were divided into treatment and control groups of equal number (∼75) and held separately in large tanks filled with dechlorinated tap water with continuous aeration. Water changes (50% to 75% exchange) were performed daily to maximize the amount of vegetation the snails could consume without compromising water quality. During water changes, solid wastes were siphoned from the bottom of the tanks, collected over a screen, and autoclaved before disposal. The snails received a diet comprising solely treatment or control hydrilla over a 7-d feeding period to ensure that the snails retained only hydrilla in their gastrointestinal tracts. After 7 d of feeding, the snail tissues were removed from the shells, coarsely chopped, frozen, and lyophilized to facilitate delivery and increase palatability in the chicken feeding trial. The uncharacterized AVM neurotoxin has been previously shown to be stable and active after freezing and lyophilization (Wiley et al. 2009).

We conducted three laboratory feeding trials with hydrilla or snail tissues as dietary treatments using 4–6-wk-old, specific pathogen-free Leghorn chicks at University of Georgia's Poultry Diagnostic and Research Center. Previous AVM feeding studies utilizing two to five chickens per treatment validated them as an appropriate laboratory surrogate for wild birds (Lewis-Weis et al. 2004). Chicks in both trials were randomly divided into control (n=5) and treatment (n=5) groups and housed in individual isolation units (Horsfal units) to prevent exposure to exogenous toxicants or stressors and to monitor diet consumption. At the conclusion of each trial, all remaining birds were euthanatized and necropsied and the tissues were prepared for AVM diagnosis as described in the upcoming text.

In the initial hydrilla/A. hydrillicola dietary treatment trial, five chicks received a diet of control hydrilla (from Lake Hatch) and five chicks were fed treatment hydrilla/A. hydrillicola (from JSTL) for 24 d. The birds received 35–40 mg hydrilla per gram of body weight (bw) per day in addition to 30 mg/g bw commercial poultry diet per day (Table 1). Treatment diet measurements were based on previous feeding studies (Lewis-Weis et al. 2004; Wiley et al. 2009). This initial trial was necessary to ensure the AVM toxin was present in the vegetation collection before proceeding with the snail feeding trial (Birrenkott et al. 2004; Wiley et al. 2007).

The snail tissue dietary treatment trial utilized the snails that had consumed either treatment or control hydrilla for 7 d. Five chicks received a diet of control snail tissue and five chicks were fed treatment snail tissues for 14 d. The birds received approximately 20 mg/g bw of either control or treatment snail tissue per day in addition to 30 mg/g bw commercial poultry feed per day (Table 1).

The final avian trial was to test for AVM toxin in Toho and was conducted using hydrilla collected from the two sites on Toho with the highest density of the A. hydrillicola colonies. The birds were randomly assigned to two treatment groups (n=5) and provided a nutritionally complete feed adjusted to contain one of two hydrilla collections: Goblet's Cove (28°13′7″N, 81°23′48″W, group A) or Big Grassy Island (28°13′7″N, 81°23′48″W, group B; Fig. 1).

Birds were monitored twice daily for clinical signs of AVM manifesting as ataxia (difficulty walking, standing, loss of balance, limb paresis), wing or head droop, and any other signs of neurologic impairment. On day 22 of the Thurmond hydrilla trial, one chicken in a group receiving treatment material displayed clinical signs of AVM and was euthanatized and necropsied. All other chickens were euthanatized and necropsied on day 24 of the Thurmond hydrilla trial, on day 14 of the snail trial, and on day 21 of the Toho hydrilla trial. At the conclusion of each chicken feeding trial, birds were euthanatized by CO₂ asphyxiation followed by cervical dislocation. Upon postmortem examination, brain tissues were removed and preserved in 10% neutral buffered formalin, processed, embedded in paraffin, sectioned at 5 μm, and stained with hematoxylin and eosin. Slides of brain sections were examined under light microscopy for AVM lesions.

The target cyanobacterial species, A. hydrillicola, was detected at 15 of 15 sites sampled at JSTL using morphologic identification and genetic analysis. The highest percent leaf coverage (>90%) by A. hydrillicola was found in Bussey Point Wildlife Management Area. This location was selected to conduct the large hydrilla collection required for avian and snail feeding trials.

Two of five treatment chickens in the Thurmond hydrilla study displayed clinical signs of AVM before the conclusion of the study. One chicken appeared lethargic on day 21 and, on day 22, displayed unsteadiness in gait and difficulty standing. A second chicken displayed mild ataxia beginning on day 22 and, by day 24, was very unsteady upon standing and unable to hold its tail erect. The remaining treatment chickens were euthanatized and necropsied on day 24. Vacuolar lesions unique to AVM were present in the optic tectum, brain stem, and cerebellum in all five chickens in the treatment group. Three of the five treatment chickens also had mild vacuolar lesions in the cerebrum. No vacuolar lesions were present in the brain tissue of any of the five chickens in the control group (Table 2).

Three of five treatment chickens that ate treatment snail material displayed clinical signs of AVM, including limb paresis before the conclusion of the 14-d study. From days 4–14, three chickens displayed varying signs of ataxia. Two chickens were frequently sedentary and displayed unsteadiness in gait. On day 14, both exhibited unsteadiness in gait and rested on the metatarsi when attempting to stand. A third chicken displayed unsteady gait and drooping wings from days 4–14 of the trial. All chickens were euthanatized and necropsied on day 14. Vacuolar lesions were present in the optic tectum, brainstem, and cerebrum of all five chickens in the treatment group (Fig. 2). Four of five chickens also had lesions in the cerebellar tissue (Fig. 3); the cerebellar tissue of the fifth chicken was not examined because of a processing error. Sparse medium-sized vacuoles were present in the white matter of the cerebellum of one of the five control chickens. No vacuolar lesions were present in the brain tissue of any of the other chickens from the control group (Table 2).

The target cyanobacterial species, A. hydrillicola, was detected at 14 of 30 sites sampled in Toho using morphologic identification and genetic analysis (Fig. 1). The highest percent leaf coverage (>50%) by A. hydrillicola was found in Goblet's Cove and Big Grassy Island, two additional sites had >40% average coverage. Two sites had moderate coverage (>20%), and A. hydrillicola coverage was low, but detected at eight additional sites (Fig. 1).

All birds in the Toho hydrilla feeding trial remained in good body condition, and average weights increased by 23% (Goblets Cove group) and 24% (Grassy Island group) per week during the trial. The birds readily consumed the treatment diet, and commercial feed was supplemented beginning day 8. No birds in either group consistently displayed clinical signs of neurologic impairment. Histopathology revealed diagnostic lesions in the brain tissues of all 10 birds that consumed hydrilla with the A. hydrillicola collected from Toho (Fig. 4). Vacuolar lesions were present in the optic tectum and brainstem, and cerebrum of all five chickens in the Goblets Cove group and all five in the Grassy Island group (Table 2). Four of five chickens in Goblets Cove and three of five in the Grassy Island group had lesions in the cerebellar tissue (Table 2).

All chickens in this study fed tissue from island applesnails that consumed hydrilla/A. hydrillicola hydrilla developed microscopic lesions in the white matter of the brain consistent with AVM. These results confirm that the putative toxin can be transferred to birds through the herbivorous island applesnail (Fig. 5). Laboratory feeding trials have shown that predatory birds ingest the toxin and develop AVM through direct consumption of the gastrointestinal tract of AVM-affected birds (Fischer et al. 2003). We have demonstrated that the toxin can transfer in an invertebrate-to-bird food chain.

The presence of A. hydrillicola in Toho may be a threat to the resident nesting Snail Kite population. Colonies of A. hydrillicola were initially documented only on hydrilla collected in a cove in the northeastern section of Toho, but an October 2011 survey identified colonies on hydrilla throughout the lake (S.B.W. unpubl. data). Furthermore, our results demonstrate that chickens fed hydrilla colonized by A. hydrillicola collected from Toho developed the characteristic AVM lesions, indicating that A. hydrillicola produces toxin in that environment. Epornitics of AVM have only been documented in winter at man-made reservoirs and ponds in the Piedmont region of the southern US. Since 1998, most new AVM sites have been discovered and confirmed through focused monitoring and sentinel research conducted from Georgia to North Carolina. Confirmation that the hydrilla/A. hydrillicola collected from Florida contains the putative toxin significantly extends the potential geographic range of AVM. It seems likely that additional regions of the country, even with environmental and hydrologic conditions vastly different from southern Piedmont reservoirs, may also harbor toxin producing A. hydrillicola.

Relative susceptibility of different avian species to AVM is proving more dependent on their diet than taxonomic classification, including herbivorous waterfowl or birds of prey that consume herbivores. Species that specialize on dietary items that contain high levels of AVM toxin are most at risk (e.g., coots [hydrilla], and eagles [coots]; Fischer et al. 2003; Birrenkott et al. 2004; Lewis-Weis et al. 2004). The density of A. hydrillicola colonies observed on Toho hydrilla is comparable to those in AVM reservoirs in South Carolina and Georgia (Wilde et al. 2005). Exotic applesnails are distributed lake-wide in Toho, including areas where the highest densities of A. hydrillicola have been observed. Exotic applesnails are also more abundant than the native Florida applesnail in all areas of Toho most frequently used by Snail Kites for nesting and foraging (Cattau 2008). In our feeding study, each chicken was fed snail tissue to equal approximately one adult snail per day. However, adult Snail Kites at Toho can consume an average of 1.1 exotic applesnails per hour while foraging (Cattau et al. 2010). Thus, the kites' average rate of applesnail consumption should be more than sufficient to induce AVM because A. hydrillicola is producing toxin at Toho, and the snails are expected to consume this material.

Although no Snail Kites have yet been documented exhibiting neurologic signs consistent with AVM, it is possible that diseased birds have not been detected. In AVM studies in Arkansas, Georgia, and South Carolina, mortality of Bald Eagles and American Coots occurred within 1–5 d of the animals being observed with clinical signs or as quickly as 5 d after being introduced to a disease site (Thomas et al. 1998; Rocke et al. 2002; Haynie et al. 2013). There are also many cases of bald eagle mortalities where AVM was the suspected cause, but never confirmed, because the carcasses were not recovered in time for postmortem examination and diagnosis (Fischer et al. 2006). Wild animals are stoic when diseased and will often seek heavy cover, making carcass recovery difficult or impossible even when an active monitoring and recovery protocol is in place (Haynie et al. 2013).

Everglades National Park and surrounding wetlands potentially provide the largest area of continuous Snail Kite habitat and historically supported the majority of the Snail Kite population and breeding. Since 2007, the Kissimmee Chain of Lakes, primarily Toho and East Lake Toho, have had the highest number of nests and fledglings (Reichert et al. 2012). Kites nesting on Toho are dependent on the exotic applesnail, and these snails are relying on exotic vegetation, primarily hydrilla. Researchers continue to investigate whether hydrilla provides kites easier access to exotic applesnails near the water's surface in the hydrilla canopy, leading to higher nest success of kites at Toho (Reichert et al. 2012). Managing exotic hydrilla and snails may provide short-term enhancement of Snail Kite nesting; it could also increase the risk of kites' exposure to A. hydrillicola and the associated AVM toxin. Because neurotoxicity has been confirmed in Toho hydrilla/A. hydrillicola material, lake managers might need to incorporate targeted chemical treatments of hydrilla to mitigate risks to avifauna in this habitat.

The interaction of invasive species and native or introduced parasites plays a role in the emergence or spread of infectious disease in freshwater ecosystems (Okamura and Feist 2011; Poulin et al. 2011). The role of biological invasions in the occurrence of noninfectious diseases caused by biotoxins is complex and not well understood. Aquatic snails of the genus Pomacea have been shown to bioaccumulate cyanobacterial toxins, but the transfer of these toxins to other wildlife has not been studied (Berry and Lind 2010). Avian vacuolar myelinopathy may represent a unique case of multiple aquatic invaders increasing the prevalence of wildlife disease, although it does not involve an infectious disease agent that can be transmitted from host to host. The invasive macrophyte hydrilla has a known role in the transmission of AVM to wildlife, and exotic applesnails may provide a second mechanism by which wild birds can acquire the disease. Invasive species management will likely be central in the prevention of additional disease pathways that may ultimately affect native freshwater biodiversity.

Primary funding was provided by Florida Fish and Wildlife Conservation Commission (FFWCC) and from cooperative agreements coordinated through the Piedmont–South Atlantic Cooperative Ecosystem Studies Unit with the US Fish and Wildlife Service and the Army Corps of Engineers Research and Development Center. The chicken feeding trials were completed with the help of fellow graduate students of S.B.W., including Brigette Haram and Jamie Morgan, and funding from US Department of Agriculture McIntire-Stennis helped to support graduate students. Gwen Kerce (Poultry Diagnostic and Research Center) was a key assistant with postmortem examination and histologic slide preparation. We thank Don Schmitz and Jeff Schardt of FFWCC for advice and field support. Mike Netherland (Army Corps of Engineers and University of Florida) was a guiding force on the Toho research, and we could not have completed the fieldwork without the tireless boat driving and applesnail/hydrilla collecting by Dean Jones (University of Florida). Mac Stone Photography graciously contributed the Snail Kite photograph in the model of toxin transfer (Fig. 5). These studies were approved by the University of Georgia IACUC under the AUP A2010 11-579-Y3-A8.