A lethal combination
Although many human activities have clear negative effects on the natural world, there are also unforeseen consequences. Bald eagle mass death events in the southeastern United States may be one such downstream effect of human activity. After considerable effort, Breinlinger et al. identified the cause of these events as an insidious combination of factors. Colonization of waterways by an invasive, introduced plant provided a substrate for the growth of a previously unidentified cyanobacterium. Exposure of this cyanobacterium to bromide, typically anthropogenic in origin, resulted in the production of a neurotoxin that both causes neuropathy in animals that prey on the plants and also bioaccumulates to kill predators such as bald eagles.
Science, this issue p. eaax9050
Vacuolar myelinopathy (VM) is a neurological disease characterized by widespread vacuolization in the white matter of the brain. First diagnosed in 1994 in bald eagles, it has since spread throughout the southeastern United States. In addition to avian species such as waterfowl and birds of prey, VM has also been found to affect amphibians, reptiles, and fish. Despite intense research efforts, the cause of this mysterious disease has been elusive. Neither contagious agents nor xenobiotics were detected in deceased animals, but field and laboratory studies demonstrated that VM can be transferred through the food chain from herbivorous fish and wildlife to birds of prey.
Occurrence of VM has been linked to a cyanobacterium (Aetokthonos hydrillicola) growing on an invasive plant (Hydrilla verticillata) in man-made water bodies. Cyanobacteria are known to produce potent toxins, so we hypothesized that a neurotoxin produced by the epiphytic cyanobacterium causes VM.
Field studies in the southeastern United States confirmed that H. verticillata was colonized with A. hydrillicola in more than half of the watersheds. Wildlife VM deaths occurred only in reservoirs with dense H. verticillata and A. hydrillicola populations. Laboratory bioassays confirmed the neurotoxicity of crude extracts of A. hydrillicola–H. verticillata biomass collected during VM outbreaks, but neurotoxicity was not detected in samples from VM-free sites. Laboratory cultures of the cyanobacterium, however, did not elicit VM. A. hydrillicola growing on H. verticillata collected at VM-positive reservoirs was then analyzed by mass spectrometry imaging, which revealed that cyanobacterial colonies were colocalized with a brominated metabolite. Supplementation of an A. hydrillicola laboratory culture with potassium bromide resulted in pronounced biosynthesis of this metabolite. H. verticillata hyperaccumulates bromide from the environment, potentially supplying the cyanobacterium with this biosynthesis precursor. Isolation and structure elucidation of the metabolite revealed a structurally unusual pentabrominated biindole alkaloid, which we called aetokthonotoxin (AETX). Genome sequencing of A. hydrillicola allowed the identification of the AETX biosynthetic gene cluster. Biochemical characterization of a halogenase detected in the cluster demonstrated that it brominates tryptophan with the expected substitution pattern. AETX is highly toxic to the nematode Caenorhabditis elegans [median lethal concentration (LC50) 40 nM] and zebrafish (Danio rerio; LC50 275 nM). Leghorn chickens (Gallus gallus) gavaged with AETX developed brain lesions characteristic of VM, whereas no lesions were observed in control chickens. VM diagnosis in treated chickens was verified using transmission electron microscopy of brain tissue.
We confirmed that AETX is the causative agent of VM. AETX biosynthesis relies on the availability of bromide. Seasonal environmental conditions promoting toxin production of A. hydrillicola are watershed specific. The consequences of elevated bromide from geologic and anthropogenic sources (e.g., water treatment and power plants) on VM should be further investigated. Notably, integrated chemical plant management plans to control H. verticillata should avoid the use of bromide-containing chemicals (e.g., diquat dibromide). AETX is lipophilic with the potential for bioaccumulation during transfer through food webs, so mammals may also be at risk. Increased monitoring and public awareness should be implemented for A. hydrillicola and AETX to protect both wildlife and human health.
A. hydrillicola, growing in colonies on aquatic vegetation, produces the neurotoxin AETX. Waterbirds, tadpoles, aquatic turtles, snails, and fish consume this contaminated vegetation and develop VM. Predators develop VM when they consume animals that have been grazing on A. hydrillicola–covered plants.
IMAGE CREDITS: GREENFROG TADPOLE, B. GRATWICKE; AMERICAN COOT, G. S. SEGLER; GRASS CARP, R. HAGERTY; SNAIL KITE, SIRKFISH; PAINTED TURTLE, U.S. FISH AND WILDLIFE SERVICE; BALD EAGLE, W. H. MAJOROS. IMAGES ARE ALL UNDER THE CREATIVE COMMONS ATTRIBUTION GENERIC LICENSE
Vacuolar myelinopathy is a fatal neurological disease that was initially discovered during a mysterious mass mortality of bald eagles in Arkansas in the United States. The cause of this wildlife disease has eluded scientists for decades while its occurrence has continued to spread throughout freshwater reservoirs in the southeastern United States. Recent studies have demonstrated that vacuolar myelinopathy is induced by consumption of the epiphytic cyanobacterial species Aetokthonos hydrillicola growing on aquatic vegetation, primarily the invasive Hydrilla verticillata. Here, we describe the identification, biosynthetic gene cluster, and biological activity of aetokthonotoxin, a pentabrominated biindole alkaloid that is produced by the cyanobacterium A. hydrillicola. We identify this cyanobacterial neurotoxin as the causal agent of vacuolar myelinopathy and discuss environmental factors—especially bromide availability—that promote toxin production.
Over the winter of 1994 to 1995, the largest undiagnosed mass mortality of bald eagles (Haliaeetus leucocephalus) in the United States occurred at DeGray Lake in Arkansas (1). More than 70 dead eagles were found in the next 2 years. The mysterious mortalities were characterized by a spongiform myelinopathy that had never been documented in wild avian populations (2, 3). Investigators of the Arkansas die-off began to notice eagles and waterbirds with similar neurological impairment throughout the southeastern states. By 1998, the emerging disease was termed avian vacuolar myelinopathy (AVM) and had been confirmed at 10 sites in six states (1). AVM has since been documented in numerous avian species across the southeastern United States during the fall and winter, most notably in waterbirds such as American coots (Fulica americana), ringnecked ducks (Aythya collaris), mallards (Anas platyrhynchos), and Canada geese (Branta canadensis) and in various birds of prey (3–7). All documented AVM cases were recovered on or near man-made water bodies with abundant aquatic vegetation that senesces during the late fall and winter months (3–7). The abundance of fish and avian prey associated with these aquatic plants attracts overwintering and nesting bald eagles and other birds of prey (8–10). AVM-afflicted wildlife are prone to injury and become easy prey for predators, as clinical signs of the disease include the severe loss of motor functions (movie S1 shows affected American coots) (2, 10). Although neurological impairment is a visual indication of disease, AVM diagnosis relies on histological confirmation of widespread vacuolization of the myelinated axons (intramyelenic edema) in the white matter of the brain and spinal cord (1).
Early sentinel field trials documented neuropathy and vacuolar lesions in wild coots and mallards within 5 days of release into a lake with an ongoing AVM epizootic (4–6). Initial chemical analysis of sediment and dead birds recovered from reservoirs where AVM cases were documented revealed no xenobiotic compounds known to induce intramyelenic edema in mammals and birds—e.g., hexachlorophene, triethyltin, or bromethalin (1, 11, 12). Early laboratory feeding trials using plants, water, and sediment collected from disease sites failed to induce the neuropathy and vacuolar lesions seen in wild birds (13, 14). Additionally, no contagious transfer, pathogens inducing myelinopathy, or neuroinflammation were documented in wild or experimental AVM-affected animals (12, 13). These early studies suggested that an unknown, seasonal, and environmental neurotoxin could be responsible (14, 15).
The search for the elusive source of this disease then focused on environmental conditions in AVM-positive water bodies, which revealed that all of them supported invasive submerged aquatic vegetation, primarily Hydrilla verticillata, with a previously unidentified epiphytic cyanobacterium—Aetokthonos hydrillicola—colonizing up to 95% of the plant leaves (16–18). Field and laboratory studies demonstrated that AVM could be transferred up the food chain. It is induced in herbivorous waterbirds after ingestion of H. verticillata colonized by A. hydrillicola and in birds of prey that consume the affected waterfowl (9, 16, 19, 20). Not only does AVM present an emerging threat to the Southeast’s avian species (5, 6), but subsequent field and laboratory H. verticillata–A. hydrillicola feeding trials have confirmed neuropathy and mortality in a wide variety of taxa, including amphibians, reptiles, and fish, as well as secondary disease transfer through the food chain. Thus, the disease is now referred to as vacuolar myelinopathy (VM) (21–24).
Cyanobacteria have long been associated with the production of toxins and other specialized metabolites (25–29). We hypothesized that a neurotoxin produced by the epiphytic cyanobacterium A. hydrillicola is the causative agent of VM. Here, we present our evidence that VM is caused by a cyanobacterial neurotoxin with notable structural features. In addition to discovering the neurotoxin, we have identified its biosynthetic gene cluster and present toxicity data on model birds, fish, nematodes, and crustaceans. Finally, we discuss environmental factors that promote toxin production.
A. hydrillicola distribution
Sampling of submerged aquatic vegetation in lakes, reservoirs, and other water bodies throughout the southeastern United States has revealed a complex and widespread pattern of A. hydrillicola distribution (16–18). As of fall 2019, we documented H. verticillata colonized with A. hydrillicola in 31 of 69 sampled watersheds (Fig. 1 and table S1). Water bodies include large (>10,000 ha) hydropower or water-supply reservoirs, county water-source reservoirs, suburban recreational lakes, and farm ponds. Given the difficulty of documenting animals dying from VM and the extensive spread of invasive H. verticillata, our current map of A. hydrillicola distribution is certainly underestimating the prevalence of the cyanobacterium and its threat to endemic wildlife, fish, and freshwater resources.
Watersheds where VM has been diagnosed (indicated by black crosshatching). Watersheds where H. verticillata has been confirmed to be colonized with A. hydrillicola are shown in red, and watersheds where A. hydrillicola has not yet been observed on H. verticillata are shown in yellow. Watersheds not yet screened for A. hydrillicola, but where H. verticillata occurs, are shown in green. Base map: copyright 2013 from the National Geographic Society.
Discovery and production of the putative toxin
In 2011, we collected H. verticillata with epiphytic A. hydrillicola from the J. Strom Thurmond Reservoir (in Georgia and South Carolina) to isolate the cyanobacterial strain for mass cultivation and subsequent isolation of the putative cyanotoxin. As a result of challenges in establishing culture conditions for this epiphytic colonizer, it took 2 years to generate sufficient A. hydrillicola biomass for a first feeding trial. The identity of the strain was confirmed as A. hydrillicola by 16S ribosomal RNA sequencing. However, chickens gavaged with this A. hydrillicola biomass did not develop VM, which failed to support that A. hydrillicola was producing a VM-inducing toxin.
Hypothesizing that the cyanobacterium produces the hypothetical toxin only when growing on H. verticillata, but not in laboratory culture, we collected additional samples of A. hydrillicola growing on H. verticillata at confirmed VM sites. The cyanobacterial colonies on the H. verticillata leaves were analyzed by atmospheric-pressure matrix-assisted laser desorption/ionization mass spectrometry imaging (AP-MALDI-MSI) to detect cyanobacteria-specific metabolites in situ. Using AP-MALDI-MSI, we could colocalize the cyanobacterial colonies with a metabolite with the sum formula C17H6Br5N3 (Fig. 2, A to D), which was not detectable in laboratory cultures by high-performance liquid chromatography–mass spectrometry (HPLC-MS). Neither commercial (Dictionary of Natural Products 28.2, SciFinder) nor in-house natural product databases revealed an entry for this elemental composition, which suggests that it is a novel natural product. The fact that the metabolite contains five bromine atoms is notable, as polyhalogenated synthetic compounds, such as bromethalin or hexachlorophene, are known to induce VM-like brain lesions in birds and mammals (1, 5, 11, 30).
(A) Micrograph of A. hydrillicola colonies on H. verticillata leaf. Autofluorescence (excitation, 395 to 440 nm; emission, 470 nm) was used to acquire the image. Regions of interest for evaluation of the subsequent MSI experiments are shown bordered in red (leaf without cyanobacterium) and blue (cyanobacteria colony on the leaf). (B) Comparison of mean mass spectra of the two regions of interest. The enlarged region shows the characteristic isotope pattern of the pentabrominated metabolite at a mass/charge ratio (m/z) of 645 ([M − H]−). This molecule is exclusively found to be associated with the cyanobacterial colony. (C) AP-MALDI image showing the spatial distribution of the distinct feature m/z 649.6382 ± 2 parts per million (ppm) ([C17H679Br381Br2N3 − H]−). Intensity is scaled from 0 (violet) to 5 × 104 (yellow). (D) Overlay of micrograph and m/z feature 649.6382 ± 2 ppm.
The presence of bromine in the putative toxin presented a potential explanation for why our laboratory cultures of the cyanobacterium did not cause VM. Our standard cultivation medium, BG11, does not contain any bromide, which is critical for the biosynthesis of the toxin. Supplementation of the cultivation medium with potassium bromide resulted in the pronounced biosynthesis of this pentabrominated metabolite. We found a nonlinear relation between bromide concentration in the medium and the production of the pentabrominated metabolite by the cyanobacterium. We determined that the optimum bromide concentration for productivity is between 0.1 and 0.5 mM KBr (fig. S1).
Although a minor production can readily be detected by HPLC-MS once the medium is supplemented with bromide, we observed a substantial increase in production under stress conditions. A drop in temperature (from cultivation at 28°C down to 21°C) or enhanced culture movement (shear stress) trigger metabolite production to an extent (>100-fold) that it becomes easily detectable by HPLC-UV (ultraviolet, detection at 286 nm) (figs. S2 and S3).
Hypothesizing that this pentabrominated metabolite was the putative toxin, we screened A. hydrillicola–H. verticillata assemblages collected from lakes during VM outbreaks using HPLC-MS and compared them with biomass from VM-free sites (where H. verticillata was not colonized with A. hydrillicola). We could detect this metabolite only in biomass collected from VM-affected lakes, which strengthened our hypothesis (fig. S4). Furthermore, the compound could be detected in the tissues of two deceased wild American coots collected during an AVM outbreak at the J. Strom Thurmond Reservoir (Georgia) in November 2014 (fig. S5), which confirmed that the compound is absorbed from the gut and accumulates in wild waterfowl. Additionally, we observed a seasonal variation of the toxin concentration in A. hydrillicola–H. verticillata biomass collected from J. Strom Thurmond Reservoir, with the peak concentration detected in November (fig. S6). This observation agrees with the finding that VM occurrences have been documented in late autumn in reservoirs, coinciding with seasonal water temperature declines and lake turnover.
As the biosynthesis of the pentabrominated metabolite requires bromide, we investigated bromide availability in A. hydrillicola habitats. Total bromine content in H. verticillata and sediments as well as the bromide concentration in water from VM-positive (containing H. verticillata colonized by A. hydrillicola) and VM-negative (containing only uncolonized H. verticillata) reservoirs was monitored seasonally. Colonized and uncolonized H. verticillata leaves contain significantly (P < 0.001) higher concentrations of bromine than sediments (~20-fold) and water (500- to 1000-fold) (fig. S7). In late summer, southeastern U.S. reservoirs are stratified, with warm, sunlit, and oxygenated water above and cool, dark, and anoxic water trapped below. During late fall, surface water temperatures cool, water layers mix, and H. verticillata senesces. We propose that this seasonal shift provides a bromide-enriched local environment, which ultimately triggers A. hydrillicola’s production of the elusive toxin.
Isolation and structure elucidation of aetokthonotoxin (AETX)
Cultivation of A. hydrillicola with bromide supplementation as well as field collections of A. hydrillicola–H. verticillata assemblages allowed us to isolate the compound in sufficient amounts for structure elucidation and bioactivity characterization. Because of the proton deficiency of the compound, extensive nuclear magnetic resonance (NMR) and infrared spectroscopic as well as high-resolution mass spectrometric analyses were required to elucidate its structure, which was confirmed by x-ray crystallography (Fig. 3, supplementary text, fig. S13, and tables S4 to S9). The structure has notable chemical features. Most prominent—also evident from the isotope pattern observed in mass spectrometry analyses of the compound—are the five bromo substituents. Brominated organic compounds are often found to be produced by marine organisms but are also found in plants, fungi, lichen, bacteria, and even humans (31–33). Several bromoindoles have been isolated from natural sources (34). Tyrian purple, one of the first brominated indole alkaloids to have been discovered, is not only the most famous example, it is also a 2,2′-biindole (35). Many brominated organic compounds exhibit strong bioactivity, ranging from antifungal to antimicrobial to antioxidant activity (36). Synthetic representatives have lately become infamous as environmental pollutants (37): Because of their lipophilicity, they tend to accumulate in sediment and biota, where they can pose a serious threat to ecosystems (38, 39). Another notable chemical feature of the compound is the connection of the two indole moieties via N1 and C2′; to date, no natural 1,2′-bi-1H-indole has been described. The two indole substructures present in AETX are rare in natural products. The 2,3,5-tribromoindole substructure has only been described from red algae of the genera Laurencia and Nitophyllum (40), the mollusk Aplysia dactylomela (likely because of its red-algal diet) (41), and the cyanobacterium Rivularia firma, which produces numerous structurally related brominated 1,3′-, 3,3′-, and 3,4′-bi-1H-indoles (42). 5,7-Dibromoindole-3-carbonitrile has not yet been found as a natural product substructure. Indole-3-carbonitrile without additional substituents has only been described once as a natural product, isolated from a halophilic bacterium probably belonging to the genus Bacillus (43). On the basis of the systematic name of the cyanobacterium, A. hydrillicola (which is Greek for “eagle killer residing on Hydrilla”), we called this compound aetokthonotoxin (AETX), or “poison that kills the eagle” [from the Greek άετός (áetós), eagle; κτείνω (kteínō), to kill; and τοξικόν (toxikón), toxin].
(A and B) Structure (A) and x-ray crystallography structure (B) of AETX.