Macondo crude oil from the Deepwater Horizon oil spill disrupts specific developmental processes during zebrafish embryogenesis

Gross morphological deformations

Water accumulated fractions (WAF) of the Macondo crude oil were vortex mixed in embryo medium at a 1:10 dilution in accordance with conventional methods [40]. Analysis of different mixing procedures and serial dilution experiments were performed and are described in Additional files 1 and 2. Based on our dose response analysis we conducted all embryo treatments at the 100% concentration of the vortex-mixed, 1:10 WAF stock solution, which will from here on be referred to as WAF. Only during our analysis of locomotor behavior was 50% WAF solution used. These WAF preparations serve to model the portions of the Macondo crude oil sampled from the riser insertion tube during the Deepwater Horizon oil spill that are capable of dissolving in embryo medium. It is important to acknowledge that according to communications with BP during the period of 2 and 3 May 2010 when this sample was being collected Nalco EC9323A defoamer was being applied topside and methanol with VX9831 oxygen scavenger/catalysts solution injected subsea. While we cannot rule out the possibility that these additional agents in the vicinity of the spill site could be present in this sample, it is highly unlikely due to the method of direct sampling through the riser tube.

Embryos exposed to WAF starting at 3.5 hours post-fertilization (hpf) until a maximum of 5 days post-fertilization (dpf) caused an array of morphological phenotypes that included dorsal tail curvature, cardiac edema, cyst formation, reduced head structures and brain hemorrhages (Figure 1). More specifically, WAF-treated embryos showed moderate (Figure 1F) to severe cardiac edema (Figure 1C-E, arrows) that, in some cases, also included yolk sac edema (Figure 1C, right most arrow). In addition, a dorsal curling of the tail and caudal cyst formation were variably observed within a treated clutch of embryos but were consistently present across treatments (Figure 1E, F, double arrows; Figure 1C, D, F, arrowhead). Qualitatively, the overall size of the brain and eyes were often reduced, which was most apparent in the ventral jaw structures (Figure 1G, H, brackets). Equally obvious was the average 28% of embryos with brain hemorrhages present in the forebrain, midbrain, or hindbrain regions (Control 0.0%, n = 452; WAF, n = 441; Figure 1H-J, arrowheads). While the eyes of some WAF-treated embryos were noticeably smaller, the mean measurements of the area and perimeter of the retina and lens showed only subtle, but statistically significant, reductions in size (Figure 1K-M). These varied, but relatively specific, morphological deformations suggest that certain embryonic processes may be affected by exposure to Macondo crude oil.

Chemical analysis of water accumulated fraction

To determine what components of the crude oil were released into the WAF mixture and thus exposed to the embryos during experimental treatments, WAF samples were analyzed within one hour of being made using solid phase microextraction (SPME, 100 mm polydimethylsiloxane) and gas chromatography mass spectrometry (GCMS, Agilent 7890A GC/5975C MSD). Chemical analysis confirmed the presence of n-hexane, toluene, xylene, benzene, naphthalene and ethylbenzene that were reported by BP to be present in this source oil B. This analysis also revealed a variety of additional components, of which the most prominent were aromatics and alkanes (Figure 2A), as observed in chemical analyses of Gulf of Mexico waters after the Deepwater Horizon blowout [41–43]. Aromatic concentrations were similar over the four WAF solutions sampled (Figure 2A). We interpret the large variation in heavier alkane concentrations (Figure 2A) to the inclusion of a non-aqueous phase, into which these poorly soluble components fractionate. Such partitioning of these less-soluble components into deep-plume oil droplets [42] and into the surface slick [43] was observed in the Gulf of Mexico after the BP Oil Spill. For a full detailed report of the chemical analysis obtained see Additional file 2.

To determine how component concentrations changed with time, samples from a WAF solution were taken every five hours from petri dishes maintained under the same conditions as the treated zebrafish embryos. As demonstrated by a few selected representative components, there is a dramatic drop-off in concentrations with time (Figure 2B). For aromatics (blue), the drop-off is less extreme for the heavier napthalenes (75 to 80% drop in five hours), compared to the lighter alkylated benzenes (93 to 95% drop in five hours). The initial concentrations for the heavy alkanes (red) are beyond reported hydrocarbon solubilities, such that the WAF solution is likely to have been initially fully saturated with these poorly soluble components (as discussed further in Additional files 1 and 2). Assuming initial saturation, the drop-off for tetradecane is similar to the heavier napthalenes, at 75% within five hours. These results suggest that while embryos are initially exposed to high levels of hydrocarbon compounds, within the 5- to 10-hour window, they are experiencing hydrocarbon concentrations that approach levels observed in underwater plumes in the Gulf of Mexico [41, 42] and those along the Louisiana marshes capable of altering gene expression in adult killifish [44]. For example, within 10 hours, our concentrations of toluene, ethylbenzene and total xylenes fall below ranges observed in underwater plumes [42].

Cell proliferation and cell death

We next wanted to characterize the gross morphological phenotypes we observed at higher resolution to glean better insight into whether exposure to Macondo oil affected specific developmental processes during zebrafish embryogenesis. WAF-treated embryos exhibit subtle reductions in brain and eye size and changes to the symmetric elongation of the tail (Figure 1), all of which could indicate reductions in cell numbers. No obvious greying of tissue, indicative of cell necrosis, was ever observed in WAF treated embryos; reduction in tissue size could be attributed to either reduced cell proliferation or increased programmed cell death.

Embryos treated with WAF from 3.5 hpf to 30 hpf were immunolabeled for Phosphorylated Histone H3 (PH3) to visualize all cells undergoing mitosis [45]. Pooled averaging of three replicate WAF treatments did not show significant changes in the number of PH3-positive cells as compared to untreated controls (Figure 3A, B; control, 79.9; WAF, 76.2; t-test, P = 0.18). To control for the presence of outlying values, the lack of statistical significance was confirmed with a non-parametric test (Mann-Whitney, P = 0.0511). If mitotic rates are not altered in WAF-treated embryos, then reductions in tissue size may be due to increased apoptosis or programmed cell death.

To test whether cells were dying by apoptosis we immunolabeled embryos for Activated Caspase 3, which is a marker for cells undergoing programmed cell death [46]. Pooled averaging of four replicate experiments did show a statistically significant increase in the number of dying cells positive for Activated Caspase 3 along the trunk in WAF-treated embryos as compared to controls (Figure 3C; control, 1.94; WAF, 12.9; t-test, P = 0.0001). Apoptotic cells were present both inside and outside the spinal cord (Figure 3D-F). Interestingly, despite using the same crude oil source for each WAF preparation, the number of apoptotic cells in WAF-treated embryos decreased with each successive replicate experiment (Figure 3C). In the first treatment an average of 38.25 Activated Caspase 3-positive cells were seen in WAF-treated embryos (Control, 3.9; t-test P = 0.0001), and in the second treatment this average dropped to eight cells (Control, 2.05; t-test P = 0.0062). Furthermore, in the third and fourth replicate experiments the number of apoptotic cells in WAF-treated embryos was no longer statistically different from their respective controls (Rep3: control, 1.1; WAF, 3.05 t-test P = 0.0966; Rep4: control, 0.7; WAF, 2.15; t-test P = 0.1151). A total of 20 embryos were examined for each condition in each separate replicate experiment.

The prevalence of WAF-induced cell death during the initial experiments suggests our crude oil sample was changing and becoming less potent over time. Chemical analysis did show much of the aromatics and alkanes in the WAF drop off dramatically within five hours after exposure to the embryos (Figure 2B), thus it is possible that our crude oil source sample experienced an incremental loss of some compounds over the course of storage. To test this hypothesis directly, we treated embryos with WAF from 3.5 hpf to 18.5 hpf and then replaced this solution with a freshly mixed WAF solution and continued the experiment until 30 hpf. This WAF refreshing protocol did show a statistically significant increase in the number of Activated Caspase 3 cells (36.74 cells; n = 42) as compared to untreated (1.75 cells; n = 40) or non-refreshed WAF treated controls (11.97 cells; n = 38) (Kruskal-Wallis Test P < 0.0005) (Figure 3G-I). In fact, the number of apoptotic cells in the refreshed WAF experiment was similar to the initial WAF experiment (Rep1, 38.25 cells; P = 0.2276) (Figure 3C). This result suggests refreshing the WAF solution returned the cell death-inducing properties of the crude oil.

Circulation and vasculogenesis

We have demonstrated that embryos treated with WAF derived from the Macondo Oil exhibit cardiac edema and hemorrhaging in the brain (Figure 1). We next wanted to determine whether the Macondo Oil WAF likewise interferes with the proper development and physiology of the circulatory system that has been demonstrated previously for other crude oil types and components [9, 10, 36–39, 47, 48]. Closer examination for the presence of blood through whole-mount hemoglobin stained control and WAF-treated embryos showed a qualitative reduction in the amount of blood cells in treated embryos (Figure 4A-D). This reduction was particularly obvious in the vasculature of the pharyngeal arches (Figure 4A, B, brackets; C, arrowheads, D). Importantly, significant hemoglobin staining was still present in the common cardinal vein distal to and including the heart, but abruptly ended in the heart or bulbus artery prior to filling the aortic arches (Figure 4A-D, arrow).

The reduced blood volume specifically beyond the heart could be due to diminished hematopoiesis, improper vessel formation or altered heart function. To better discern between these possibilities, we conducted angiography with fluorescent quantum dots in 72 hpf live embryos treated with WAF starting at 3.5 hpf. This procedure was done in tg[fli:eGfp] transgenic larvae, which allowed for visualization of green endothelial cells as the red quantum dots flowed throughout the circulatory system, illuminating both the extent of circulatory function and blood vessel anatomy [49]. Imaging of the aortic arches revealed that in some cases the endothelial vasculature was present and capable of supporting circulation (Figure 4E, G, arrows and arrowheads), while in more severe cases the more posterior arch vasculature was lost and circulation was significantly reduced in the remaining vasculature (Figure 4F, H). In addition, there was often a build-up of the quantum dots in the heart atrium paired with a reduced flow of these quantum dots into the ventricle (H', dashed line). These results suggest that the endothelial vasculature and proper heart development and function are all variably compromised in WAF-treated embryos. Further supporting these results, real time imaging of blood flow through the dorsal aorta showed that individual blood cells from 60 different WAF-treated embryos (3 replicates) took, on average, 2.68 times longer to travel the distance of 7 somites (Figure 4I-K; Control, 1.44s; WAF, 3.85s; t-test, P = 0.0001).

Continued analysis of quantum dot circulation and endothelial vasculature in the trunk revealed additional phenotypes involving intersegmental blood vessel development. The most prevalent phenotype was varied frequency of reduced circulation of fluorescent quantum dots through the intersegmental blood vessels, such that vessels showed either a complete absence of flow (Figure 4L-O, arrowheads), truncated flow (Figure 4L-O, arrows), or normal circulation. Interestingly some vessels devoid of any quantum dots often showed ectopic branching and an overall improper vessel pattern (Figure 4N', O'). These phenotypes were consistent over three separate experimental replicates with at least 10 embryos imaged per replicate. As a whole, these results suggest the Macondo crude oil is capable of causing specific deformations in vasculogenesis in both the trunk and head of zebrafish that leads to reduced circulatory function.

Craniofacial development

Due to the vascular defects associated with pharyngeal arches, we next examined whether exposure to Macondo crude oil similarly causes defects in craniofacial development, a phenotype known to occur with specific PAH exposure [11, 36]. We treated embryos from 3.5 hpf to 4 dpf and then performed Alcian Blue staining to visualize cartilage [50]. Consistent with head vasculature phenotypes, WAF-treated larvae showed a range of head and jaw cartilage phenotypes (Figure 5A-F). WAF-treated embryos had a general size reduction in all cartilage elements with a significant lack of anterior extension of the most prominent jaw elements (Meckel's and ceratohyal cartilage) (Figure 5B, C, E, F). The most severely affected embryos showed dramatic reduction in all of the pharyngeal arch cartilage elements (ceratobranchial) with a complete loss of the most posterior three arches as compared to untreated control larvae (Figure 5A, B, E). In addition, we report here for the first time, in response to any crude oil treatment, a lack of anterior midline fusion of the basihyal cartilage (Figure 5E, arrowhead). Less severely affected embryos, however, did not display this basihyal phenotype (Figure 5F).

Because of the corresponding deformations in both endothelial and cartilage cell development in the head, we next hypothesized that these phenotypes may be derived from an earlier defect in the differentiation or migration of neural crest cells, which are a common progenitor cell contributing to pharyngeal arch cartilage, smooth muscle of the pharyngeal arch arteries as well as portions of the heart such as the arterial pole and endocardial cushions [51–56]. To test this we conducted whole-mount in situ hybridizations for crestin transcripts in WAF-treated and control embryos at 27 hpf. crestin is a known marker of neural crest cells during their early specification, delamination from the dorsal neural tube, and subsequent migration into the characterized pathways of the trunk and head [57]. Interestingly, embryos treated with WAF from 3.5 hpf to 27 hpf of development showed qualitative reductions in crestin-labeled cells, specifically in the anterior migratory streams known to populate the pharyngeal arches (Figure 5G-L; H, J, L, arrowheads and bracket). In contrast, the amount and position of crestin-positive cells in the trunk qualitatively appeared normal (Figure 5G, I, K, insets).

To confirm that deformations in pharyngeal arch development were associated with defects in early cranial neural crest populations, we examined the presence of rostral migratory streams of neural crest cells in tg(fli:GFP) transgenic embryos and whether they appropriately expressed the cell specification marker dlx2 [58, 59]. A total of 76% of WAF treated embryos (n = 83) showed a specific loss of one of the more posterior pharyngeal arches (Figure 5M, N, arrows) (control = 8.6%; n = 83; P < 0.0005). Importantly, despite normal dlx2 expression in the forebrain, 89.4% of WAF-treated embryos (n = 94) showed a significant reduction of dlx2 expression in all of the rostral migratory streams with a near complete absence in the presumptive posterior-most arch locations (Figure 5O, P, bracket) (control = 15.6%; n = 122; P < 0.0005). These findings suggest that Macondo crude oil is compromising the early development of cranial neural crest cell differentiation, which results in reduced posterior pharyngeal arches and the more visible defects in the heart, arch vasculature and craniofacial development.

Locomotor behavior

During our many WAF treatments it was clearly evident embryos displayed irregular swimming behaviors. The locomotor escape response to stimuli is an important survival behavior that develops later in embryogenesis. Interestingly, previous studies examining the effect of PAHs on zebrafish swimming behavior did not reveal any significant phenotypes [36]. Therefore, we systematically tested whether exposure to Macondo crude oil WAF impacted swimming patterns and escape responses. To do this, we recorded the swimming behavior of individual 48 hpf larvae with a high-speed video camera (1,000 frames/second) following the administration of a specific touch stimulus [60]. WAF-treated embryos demonstrated abnormal swimming behavior and a failure to escape based on multiple criteria. WAF-treated embryos showed reduced sensitivity to touch stimuli, as demonstrated by 70% response rate for WAF-treated embryos as compared to a 99% response rate for untreated control embryos (n = 100 trials from 10 embryos each). When a response was produced in WAF-treated embryos they showed a significantly reduced frequency of body bends (Control, 39.10Hz; WAF, 18.82Hz; n = 10 each; t-test, P < 0.01) and swam for less time than untreated control embryos (Control, 875.8mS; WAF, 282mS; n = 10 each; t-test, P = 0.01) (Figure 6). The presence of locomotor behavior phenotypes suggests that there could either be a problem with neural transmission or a developmental problem resulting from an anatomical deformation in the nervous or musculature systems.

Neuronal development in the central and peripheral nervous system

Development of locomotor movement in response to touch stimuli requires a complex neural network that begins with an elaborate meshwork of Rohon Beard sensory axons at the periphery that originate from the dorsal neural tube. These bipolar sensory neurons make connections with a variety of interneurons that function to relay signals to Mauthner neurons, which serve as the major motor control center in the hindbrain. Mauthner neurons then send signals back down the spinal cord to stimulate coordinated motor neuron activation to achieve the alternating contraction of skeletal muscle required for swimming behaviors [61–63]. In order to determine what underlying developmental processes may be contributing to the visible defects in locomotor behavior, we used specific antibody markers to visualize each cellular step in the locomotor neural circuit.

Anti-Acetylated Tubulin (AT) labels all axonal pathways in the zebrafish as well as specifically marks the somas of Rohon Beard sensory neurons [64]. Anti-Islet-1 labels both the nuclei of Rohon Beard Sensory neurons and motorneurons, anti-Gaba labels a series of interneurons, and the 3A10 antibody marks Mauthner neurons in the hindbrain [65–67]. Surprisingly, using this spectrum of neuronal markers we found only one consistent and specific anatomical deformation that was restricted to the peripheral projections of sensory axons. The Mauthner neurons that serve as the central mediators for startle-reflex behavior in the hindbrain, showed absolutely no difference in number or axonal projections between WAF-treated and untreated control embryos (Figure 7A, B; n = 60) [67–69]. Likewise, no statistical difference was seen in the average number or position of Gaba-positive DoLa, CoSA, VeLD or KA interneurons (Figure 7C, D; Control, 41.1 cells, n = 60; WAF, 41.3 cells, n = 59; t-test, P = 0.855).

Investigation of motorneuron development did not reveal major differences in their cell differentiation. Specifically, anti-AT labeling of primary motor axons did not show consistent defects over the length of the trunk (Figure 7E, F, arrows). It should be noted that while occasional errors in motor axon pathfinding were detected, they were always associated with an affiliated muscle patterning defect that we describe in detail later. Quantification of three separate replicate experiments of Islet1 labeling for primary and secondary motorneurons at 30 hpf showed only minor reductions in the number of motorneurons, with only one of the three replicates actually exhibiting statistical significance (Figure 7G, H, lower bracket; Rep 1: 16% reduction (control n = 11, WAF n = 20), t-test, P = 0.0001; Rep 2: 4.6% reduction (control n = 18, WAF n = 18), Mann-Whitney test, P = 0.456; Rep3: 6.6% reduction (control n = 20, WAF n = 19), Mann-Whitney test, P = 0.043). These results suggest that in some limited way motorneuron number may be impacted by exposure to Macondo oil WAF. However, based on the relatively normal pattern of primary motor axons, very modest overall reductions in number, variable significance and time period of analysis correlating with the middle of secondary motorneuron birth [65], we attribute these minor decreases in motorneuron populations more likely to a mild developmental delay in treated embryos rather than any biologically relevant effect from the oil.

Similarly, no qualitative differences were seen in the position or morphology of Rohon Beard sensory neurons as detected with anti-AT labeling (Figure 7E, F asterisks). Quantification of Rohon Beard sensory neuron numbers with anti-Islet1 showed little to no effect (Figure 7G, H, upper bracket). Similar to motorneurons, only one of the three replicates revealed a statistically significant difference (Rep 1: 2% increase (control n = 11, WAF n = 20), t-test, P = 0.72; Rep 2: 6.3% reduction (control n = 18, WAF n = 18), t-test, P = 0.184; Rep 3: 15.6% reduction (control n = 20, WAF n = 19), t-test, P = 0.0001). Again, based on the normal morphology and position of Rohon Beard cells, with no statistical difference present in the first two replicates, and only a small effect observed in the third replicate, we conclude Rohon Beard sensory neuronal number to be unaffected by Macondo oil. However, analysis of the meshwork of sensory axonal projections in the periphery was consistently both visually and quantitatively significantly reduced (Figure 7I, J, Control axonal labeling intensity = 1274.1 pixels, n = 55; WAF axonal labeling intensity = 883.9 pixels, n = 59; t-test, P = 0.0001). These results suggest most of the anatomical development of the locomotor circuitry is unaffected by Macondo crude oil WAF with the exception of subtle but consistent deformations in the amount and/or branching dynamics of sensory axonal projections.

While the overall number of many key neuronal cell types was normal in WAF-treated embryos, we wanted to confirm that radial glial cells, which serve as the neural stem cell population in the developing spinal cord, were also unaffected [70–72]. Using an antibody to Glial fibrillary acidic protein (Gfap) to mark radial glia we show that radial glial cell staining throughout the spinal cord appeared normal (data not shown). Most relevant was the normal number of Gfap-positive cell somas located at the ventricular surface of the neural tube (Figure 7K, L, arrowheads; Control, 5.99 cells, n = 40; WAF, 5.67 cells, n = 40; t-test, P = 0.582). Anti-Gfap labeling fills radial glial cell bodies when they are undergoing mitosis at the ventricular zone (Johnson and Barresi, in preparation; [73]). Normal numbers of such dividing radial glia indicates that exposure to Macondo crude oil WAF does not affect neural stem cell proliferation during embryogenesis.

Somitogenesis and muscle fiber type development

Because our analysis of the nervous system showed specific deformation restricted to the axonal projections of the peripheral nervous system, it was possible that these defects were secondary to deformations in the paraxial mesodermal environment through which these axonal projections need to navigate. During somite development, radially migrating slow muscle precursor cells provide key guidance cues to the simultaneously outgrowing motor axons [74]. Therefore, any defects in the proper specification, migration and/or position and integrity of slow muscle fibers could influence the development of the peripheral nervous system. We exposed embryos from 3.5 hpf to 48 hpf in Macondo crude oil WAF and assayed for F59 labeling, which preferentially recognizes slow muscle myosin in zebrafish [75, 76]. As with most assays, we discovered embryos displaying a range of slow muscle development phenotypes from no observable defects to the loss of muscle fibers and segment boundary errors. Interestingly, similar to what was observed for our cell death analysis over time, embryos displaying slow muscle phenotypes were extremely severe during initial treatments but these phenotypes decreased significantly over the course of several months of repeated experiments using the same crude oil source for WAF preparation (Figure 8).

Initial experiments in November 2010 showed clear defects that represented several distinguishable phenotypes in the treated embryos (Figure 8A-D). The first was the sporadic instance of defects in proper somite boundary formation, such that a segment boundary in a portion of one somite would be absent as represented by slow muscle fibers extending from the anterior boundary of one somite to the posterior boundary of the neighboring somite (Figure 8B, C, arrowhead). Most interestingly, double labeling for motor axons showed pathfinding errors in the affected somite that were directly correlated to the somite/slow muscle defect (Figure 8B, asterisk, arrow). These results suggest that Macondo crude oil may not only be affecting slow muscle fiber development but also somitogenesis, the process by which segment boundaries are established in the paraxial mesoderm [77–79]. The second distinguishable phenotype was the sporadic loss and disorganization of slow muscle fibers in the superficial monolayer (Figure 8D), which suggests impairments in slow muscle differentiation and subsequent migration in WAF-treated embryos.

Later experiments revealed somitogenesis defects with corresponding motor axon pathfinding errors, but these deformations were not as dramatic as compared to our first experiments (Figure 8E-G) and subsequently absent from the final replicates (Figure 8H-M). Later experimentation still showed extremely sporadic loss and disorganization of the slow muscle monolayer (Figure 8H-M). Interestingly, high resolution imaging of regions showing reductions in slow muscle fibers revealed muscle phenotypes resembling muscle degeneration [80–85]. Variable sized fragments of slow muscle myosin were found present in locations devoid of morphologically normal slow-twitch fibers (Figure 8I, J, red). In contrast to the somite boundary defects, motor axon pathfinding appeared normal in areas showing these missing or degenerating muscle fibers (Figure 8I, J, green). Lastly, these phenotypes continued to diminish in severity upon our last round of treatment (Figure 8K-M).

The sporadic nature of muscle phenotypes both in a clutch of treated embryos and within an affected embryo paired with the gradual loss of these phenotypes over time suggests that the causative agent in the crude oil is not particularly potent and somehow loses its activity over time. Therefore, in an attempt to reproduce the severity of the muscle and somitogenesis defects observed initially, we exposed embryos from 3.5 hpf to 48 hpf in WAF that was refreshed every 15 h. Similar to our results of increased apoptosis in refreshed WAF treatments, more severe muscle deformations were present in refreshed WAF-treated embryos as compared to untreated controls and non-refreshed WAF-treated embryos (Figure 9). The sporadic nature of these deformations was consistent, but the severity was now qualitatively similar to our initial experiments. Remarkably, somitogenesis defects were detected again, showing slow muscle fibers crossing in locations that would normally have a somitic boundary (Figure 9D, D', arrowheads) as well as the presence of irregularly shaped somites (Figure 9D', arrows). In addition, significant slow muscle loss and potential degeneration was also observed following this WAF refreshing protocol. Specifically, slow muscle myosin fibrils were seen in varying stages of degeneration (Figure 9E). This type of dose-response approach confirms that some component(s) within the Macondo crude oil WAF do directly disrupt somitogenesis and slow muscle development.

Defects in size and shape through induction of apoptosis

WAF-treated zebrafish embryos had visible reductions and morphological changes in the size and shape of the head and trunk. These phenotypes could be explained by a reduction in cell proliferation or increase in programmed cell death. We found there was no reduction in the number of mitotic cells (Figure 3A, B), which was further confirmed by no change in the number of dividing neural stem cells (Figure 7K, L). These data suggest cell division rates were unaffected by Macondo crude oil WAF. However, we did detect a statistically significant increase in the amount of apoptosis (Figure 3C-I), supporting a cell death regulatory role for some component in the crude oil. This component could actively induce cell death or play a role in the repression of a survival factor. There are relevant data to support this hypothesis; crude oil, fuel oils, or specific PAHs have been documented to up-regulate known apoptotic proteins in juvenile cod, to increase programmed cell death in cultured dolphin renal cells, and to trigger apoptotic DNA fragmentation in ovarian and liver cells of the juvenile channel catfish [96–98]. We observed an increase in apoptotic cells present inside and outside of the nervous system (Figure 3), suggesting activation of programmed cell death was not necessarily tissue specific. Systematic chemical analysis in zebrafish could help discern which components of the Macondo oil are responsible for induction of programmed cell death and whether induction is cell type specific.

Selective impairment of neural crest cells is at the 'heart' of the problem

While increased cell death may play a role in the defects contributing to reductions in embryo size and shape, reductions in head size may be more directly associated with a lack of proper jaw formation. Defects in craniofacial development in response to crude oil or specific PAHs have been demonstrated previously [36]. The most prevalent and severe craniofacial defect we observed in response to Macondo crude oil WAF was the preferential loss of posterior pharyngeal cartilage elements (Figure 5A-F). As one might expect, the vasculature associated with these same pharyngeal arches was also reduced (Figure 4A-H). Most crude oils, or their components, cause cardiac edema, heart morphogenesis defects, and reduced circulatory function during embryogenesis of several fish species [9, 36, 38, 47]. We hypothesized that the defects we observed in pharyngeal cartilage, vasculature development and heart morphogenesis were linked by an earlier disruption in the proper development of cranial neural crest cells, which are contributing precursors for all of these tissues.

Neural crest cells function as multipotent stem cells that actively delaminate from the dorsal neural tube and migrate along separate pathways across the entire anterior-posterior axis of an organism. During migration to their final destinations neural crest cells differentiate into a variety of cell types, some of which contribute to the development of pigment cells, peripheral nervous system, head cartilage, endothelial and smooth muscle vasculature, and portions of the heart [52–56]. Interestingly, recent investigations in amniotes have suggested pharyngeal arches and the heart are derived from the same "vagal" domain of neural crest cells lying at the intersection between the head and trunk axial positions (reviewed in [52]). This is of particular relevance since the most significant phenotypes we observed were restricted to the pharyngeal arch cartilage, arch vasculature and heart (Figure 4). Importantly, cardiac neural crest cells specifically migrate through the posterior pharyngeal arch pathways on route to the developing heart fields, where they differentiate into smooth muscle and endothelial cell derivatives that contribute to the morphogenesis of the outflow tract, septum, and valves of the heart [51, 55].

We discovered that Macondo WAF-treated embryos show a specific reduction in one of the posterior arches, and a similar restricted reduction was seen in both crestin and dlx2 gene expression by cranial neural crest cells associated with the posterior pharyngeal arches (Figure 5G-P). Initially cranial neural crest cells migrate in three streams toward the presumptive arches, and the posterior-most stream undergoes a branching process to form three additional streams that establish the posterior most pharyngeal arches [99]. Therefore, we hypothesize that the causative toxins in the Macondo WAF may be affecting neural crest cell specification as it relates to their ability to carry out this branching step during posterior pharyngeal arch development. This spatially restricted defect in neural crest development is a remarkably targeted effect by crude oil, which suggests there are equally specific molecular pathways directly influenced by the components of the Macondo WAF; such as, the known neural crest regulators Wnt, Fibroblast growth factor, or the Bone Morphogenic Protein signaling pathway [51, 100, 101]. These pathways may be activated directly or independent of the aryl hydrocarbon receptor 1 and 2, which has been shown to be required for cardiac edema and heart morphogenesis defects in response to selective PAHs [37].

Despite the apparent lack of neural crest defects in the trunk of WAF-treated embryos, there was reduced circulation in the intersegmental blood vessels, and those vessels devoid of any circulation often showed excessive vascular branching (Figure 4L-O'). These elaborate and often forked branching patterns (Figure 4N', O', arrow) were reminiscent of an angiogenic process called intussusception, in which a vessel splits along its longitudinal axis and undergoes vascular remodeling [102, 103]. While intussusception has not yet been described in zebrafish, others have demonstrated mouse retinal vasculature responds to hypoxic conditions by forming characteristic "vascular loops" [104]. In WAF-treated embryos vascular remodeling is seen only in intersegmental blood vessels that are devoid of circulation, which could represent an extreme but focused hypoxic event during which vascular looping might be observed (Figure 4N', bracket). We, therefore, hypothesize that the vascular remodeling observed in WAF-treated embryos is an indirect effect caused by reduced circulation in the intersegmental blood vessels due to a primary disruption in early neural crest-mediated heart development and function.

Defects in patterning the peripheral nervous system impacts locomotor escape behaviors

Like most teleosts, zebrafish larvae evolved early stereotypical swimming patterns in response to touch stimuli that enable fast escape locomotor behaviors [105, 106]. The inability to properly react to touch would have adverse consequences to larva survival. We observed that larvae exposed to the Deepwater Horizon crude oil had significantly reduced sensitivity to touch and disorganized swimming patterns relative to untreated controls (Figure 6). A previous study that examined the effects of specific PAHs on zebrafish locomotor behaviors did not detect any irregularities [36], suggesting that either the Macondo crude oil WAF possesses a unique component or a particularly toxic combination of known components impaired proper locomotor function. This could have unfortunate implications for the species of the Gulf of Mexico and Atlantic Ocean where even subtle reductions in larval or adult escape responses can be deadly. There is some precedent for this, as delayed escape behaviors have been documented in fiddler crabs in response to chronic exposure to No. 2 fuel oil polluting the sediment of Wild Harbor in Buzzards Bay, MA [107]. We sought to determine the developmental origin of the locomotor phenotype in WAF-treated embryos by assessing the anatomical organization of the neural circuitry and skeletal muscle necessary to respond to touch and yield functional swimming behaviors. Considering the complexity of the locomotor system, we were surprised to only detect specific deformations in the peripheral nervous and muscular systems; the neurons and astroglia within the central nervous system were normal in quantity and position (Figures 7, 8).

Similar to our findings of touch response defects, the peripheral sensory and motor axon defects and slow muscle patterning phenotypes, to the best of our knowledge, have never previously been documented. Specifically, Macondo crude oil WAF caused reduced sensory axonal branching along the entire trunk, and sporadic motor axon pathfinding errors directly associated with corresponding deformations in slow muscle fiber development. Reductions in sensory neuronal branches likely cause reductions in touch responses [108, 109], which suggests that WAF-induced reductions in sensory axonal arbors contribute to the reduction in touch sensitivity. However, when a touch response was elicited in a WAF-treated larva, they exhibited disorganized swimming behaviors. While problems in sensory branching may contribute to these swimming errors, the cause is more likely a problem with stimulus transduction controlled by the downstream neural circuitry and muscular output [110].

WAF-treated embryos exhibit deformations in motor axon pathfinding and slow muscle development (Figure 8). These specific defects could definitely lead to impaired muscle contractions and swimming behaviors. Early muscle contractions have been shown to be required for the proper ventral trajectory of pathfinding sensory axons and their ability to exhibit appropriate self-avoidance behaviors to establish the mesh-like pattern of sensory branching [111]. However, this is not likely a significant influence as muscle and motor axon defects were not consistently seen throughout the trunk of WAF-treated embryos unlike the sensory axonal defects, nor were longitudinal pathfinding errors or significant axon to axon contact seen that are characteristically found following muscle contraction loss [111].

Motor axon pathfinding errors were only found in somites that exhibited corresponding slow muscle patterning defects, which strongly suggests motor axon pathfinding errors are not direct effects of the oil but rather indirect phenotypes in response to inappropriate guidance cues derived from earlier problems with somitogenesis and slow muscle patterning. This is in contrast to the sensory neuron branch reductions that are uniformly present throughout the trunk and, thus, likely a direct affect of exposure to the Macondo crude oil. Slow muscle cells have been shown to provide critical axon guidance cues that direct the proper pathfinding of motor axons [74, 112]. This primary defect in early skeletal muscle development can be interpreted as two separable processes, in which there are changes to the highly stereotypical pattern of segmentation and then improper slow muscle positioning.

Within the musculature of the trunk we observed losses of somitic boundaries as well as the occurrence of inappropriate boundaries within the same somitic region (Figure 8B, C, F, G as examples). At this point we can only speculate how Macondo crude oil might be affecting somitogenesis. Segmentation in vertebrates is controlled by the precise coordination of a Notch-Delta mediated molecular clock that determines when a boundary will form and this process is paired with opposing anterior and posterior morphogenic gradients of Retinoic acid and Fibroblast growth factor that define the location of a segment boundary [77, 79, 113–116]. It is possible that some component within the Macondo crude oil might be impacting this somite molecular clock mechanism; however, it is also plausible that our WAF treatments were affecting the terminal step in boundary formation that involves the process of epithelialization and formation of the myotendinous junction rather than the timing of somitogenesis [117, 118].

After a segment boundary has formed in the paraxial mesoderm of zebrafish, "adaxial" slow muscle precursor cells located adjacent to the notochord undergo substantial morphogenesis and movement to the outer-most edge of a somite to form a monolayer of elongated, slow-twitch muscle fibers [76, 119, 120]. While alterations in proper somite boundary formation would yield irregularly elongated slow muscle fibers, it has not been documented to cause alterations in the medial to lateral positioning of fibers, change the parallel positioning of the slow muscle array, nor cause early slow muscle loss [118]. Therefore, we hypothesize that in addition to somite formation defects, Macondo crude oil WAFs may be independently affecting some aspect of slow muscle specification and migration.

Hedgehog signaling is required for proper slow muscle cell specification and Cadherin cell adhesion molecules are required for proper slow muscle morphogenesis and movement, and their loss causes slow muscle positioning phenotypes similar to what we observed following WAF treatments [120–122] (Figure 8D, L, M as examples). While these conclusions related to muscle development are speculative, they provide a basis for a series of future experiments aimed at analyzing the effects of specific Macondo crude oil compounds on somitogenesis and muscle development, as well as examining the role of the Aryl hydrocarbon receptor signaling pathway in these processes [37].

Aside from these early somitogenesis and muscle fiber type patterning defects, a separate and later forming slow muscle degeneration phenotype was also variably present in some WAF-treated embryos. We found a number of embryos treated with the Macondo oil WAF had sporadic muscle fiber breaks and single sided detachment from the lamina, which subsequently exhibited cell death morphologies (Figure 8I, J). We interpret this phenotype as a specific, WAF-mediated muscle degeneration, as it is nearly identical to the muscle pathology seen in zebrafish genetic models of muscular dystrophy [80–85]. Importantly, this late muscle degeneration defect does not have corresponding motor axon pathfinding errors because it occurs after motor axons have already successfully reached their target cells. While these varied early and late muscle phenotypes have real consequences for embryonic health and locomotor function, the sporadic nature of these muscle phenotypes suggests they cannot fully account for the consistent errors in swimming behavior exhibited by WAF-treated embryos.

The abnormal locomotor behaviors exhibited by WAF-treated larvae could be due to defects in central nervous system function. Our level of analysis did not reveal any neuroanatomical defects; however, fine mapping of neural circuits or analysis of neuronal network activity could help elucidate specific cellular and molecular mechanisms disrupted by crude oil exposure. Moreover, it will be equally important to characterize the quality of myelin wrapping by oligodendrocytes and schwann cells, which contribute significantly to how well signals are conducted between neurons.

Phenotypic changes over time

An unexpected finding of our research was the phenotypic reduction in certain defects over successive and identical experimental replicates. Specifically, apoptotic cell death and skeletal muscle phenotypes decreased in severity over the course of multiple experiments, while the phenotypes associated with neural crest development and sensory neuronal branching remained consistent. This observation suggests that different components within the Macondo crude oil might cause these phenotypes, and that some specific component(s) changed over the course of the use and storage of our Macondo crude oil sample. Importantly, we were able to reproduce the severity of both cell death induction and the severity of all the skeletal muscle phenotypes by exposing embryos to freshly mixed WAF preparations several times over the course of a single experiment. This confirms that Macondo crude oil is responsible for these phenotypes and, while the unidentified compound has reduced potency over time, it is still present and capable of impairing embryonic development. A recent study demonstrated that dissolved PAHs, rather than oil particles, are toxic to zebrafish embryos and cause edema, hemorrhaging, developmental delays, and abnormalities in cardiac function [38]. We propose a model in which the more readily dissolved PAHs are responsible for the neural crest derived phenotypes that lead to cardiac and craniofacial deformations, whereas the cell death and skeletal deformations may be caused by novel Macondo oil components, such as smaller ringed hydrocarbons that are more easily released from the WAF as a gas, or much heavier hydrocarbons that readily fall out of solution. Our findings will enable logical candidate approaches to analyze the role of individual components of Macondo oil in regulation of specific developmental processes during fish embryogenesis.

Sampled Macondo crude oil

A one-liter sample of Macondo crude oil, MC 252 (Source Oil B; ID #: 20100617-121) from the Deepwater Horizon oil spill was provided by British Petroleum (BP) on September 8, 2010. This oil sample was collected from the riser insertion tube on the Enterprise by Entrix samplers during May 2-3, 2010. BP acknowledged that during sampling of this crude oil, Nalco EC9323A defoamer was being injected topside, while methanol with 10,000 ppm VX9831 oxygen scavenger/catalysts solution was being injected subsea. While the presence of these additional agents in the sample cannot be ruled out, it is unlikely they were incorporated into this crude oil sample due to the method of sampling directly through the riser tube. It was further acknowledged by BP that there was a variable amount of water in the oil samples collected in this manner. Material safety data sheets for the Source Oil B crude oil sample stated its major constituents were n-hexane, toluene, xylene, benzene, naphthalene, ethylbenzene, and hydrogen sulfide.

Water accommodated fraction (WAF)

Following previously accepted mixing procedures, we created water accumulated fractions (WAF) of Macondo oil with autoclaved E3 embryo medium [40, 123]. Specifically, the 1-liter container of Macondo crude oil was slowly inverted 20 times before collecting a sample to ensure consistent sampling over the course of the study. Oil dilutions of 1:10 in embryo medium were prepared in cap-sealed glass bottles that allowed for 20% air volume. Initial gas chromatography mass spectrometry (GCMS) analysis of WAF exposed to plastics detected n-butyl phthalate in solution, which is a plasticizer speculated to cause developmental effects on its own (data not shown, [124]). Therefore, only glass pipettes and glass petri dishes were used with WAF solutions. For initial characterization, WAF stock solutions were mixed with a magnetic stir bar in two different ways: (1) WAF stock was stirred at a slow, non-vortex inducing mixing speed [40]; and (2) WAF stock was stirred at a rate sufficient to produce a vortex equivalent to one-third the volume of the solution. Both solutions were mixed for 24 h, followed by a 2 h resting period before only the aqueous WAF phase was sampled. Unless otherwise noted, the embryos in the following experiments reported here were exposed to a 100% WAF solution derived from the 1:10 (crude oil to embryo media) vortex-mixed stock solution.

WAF chemical analysis

To determine what the chemical composition was of the resulting WAF solutions made in embryo medium, Solid phase microextraction (SPME,100 mm polydimethylsiloxane) and gas chromatography mass spectrometry (GCMS, Agilent 7890A GC/5975C MSD) were performed. WAF samples were analyzed within one hour after the mixing procedure was completed. The SPME fiber was immersed for six minutes in a 15 mL vial of diluted WAF (25-fold dilution), with constant stirring, before insertion into the GC (helium carrier gas; splitless mode for first six minutes; 220°C inlet; oven at 35°C for four minutes, ramped 8°C/minute to 275°C, held at 275°C for three minutes). For consistency among WAF samples, standards for selected components (cyclopentane, methylcyclohexane, n-hexane, n-decane, n-tetradecane, benzene, toluene, xylenes, naphthalene, anthracene, phenanthrene, all from Supelco) were also made in diluted embryo medium (25-fold dilution). Benzene-d6 (Supelco) was added as an internal standard. An expanded explanation of the chemical analysis is provided in Additional files 1 and 2.

Embryo collection and treatment

Fish lines were maintained in the ALAAC accredited Smith College Zebrafish Facility using standard husbandry techniques ([125]; Zebrafish International Resource Center). Fertilized eggs from wild type embryos (AB strain) and Tg[fli:eGfp] transgenic fish (TU/AB background, provided by the Lawson Lab, UMass-Medical School) were collected, washed in embryo medium, and incubated at 28.5°C. Unless otherwise stated, embryos were exposed to WAF solutions starting at 3.5 hours post-fertilization (hpf) and maintained in WAF solutions until a maximum of 5-days post fertilization (dpf). Control embryos were maintained in oil-free embryo medium and grown in a separate 28.5°C incubator to eliminate potential exposure to WAF gases. All treatments were carried out with either 20 mL of solution in 100 mm glass petri dishes or 10 mL of solution in 50 mm glass petri dishes. Prior to fixation or live embryo imaging, embryo chorions were removed by chemical digestion with Pronase (2 mg/ml, Sigma-Aldrich, St Louis, MO, USA) for three to five minutes.

Behavioral analysis

In an effort to perform a more conservative assessment of locomotor behavior, analysis was carried out using a 50% WAF solution. Embryos exposed to either 50% or 100% WAF had similar locomotor responses (data not shown). To characterize larval swimming behavior at 48 hpf, a light touch stimulus was applied to the head of larvae. The minimum stimulus required to elicit a response was determined using pressure-specific Von Frey filaments. The touch-response was recorded using a high-speed camera (Fastec Imaging, San Diego, CA, USA) at 1,000 frames per second (fps). A 35 mm lens (Nikon, Melville, NY, USA) was used for magnification. To illustrate the responses, single frames taken at 20 mS intervals were overlaid in Adobe Photoshop (San Jose, CA, USA).

For kinematic analysis, the head-to-tail angles were calculated for each frame using automated software developed by the Downes Lab (Biology Department, University of Massachusetts, Amherst, MA, USA) [126]. In brief, pixel density was used to identify three landmarks along the larval body: the tip of the nose, the border between the yolk extension, and the tip of the tail. These three points form an angle, which was plotted over time using Microsoft Excel (Microsoft Corporation, Redmond, WA, USA). To calculate body bend frequency, a full body bend was defined as two intervals of more than 50 degrees of opposite directions. To determine the duration of a swimming episode, we measured the beginning of a swimming episode until the final time the body was straightened to within 20 degrees of being straight (defined as 0 degrees).

Blood, circulation and vascular analysis

Visualization of blood in 3 dpf embryos was accomplished by staining hemoglobin with o-Dianisidine as previously described [127]. Blood circulation rates were determined by recording the amount of time a single blood cell took to travel through the dorsal aorta from somite 8 to somite 14. A total of three blood cells were recorded and their travel times averaged for each fish. Individual blood cells were easily visualized in live larvae at 3 dpf using an Olympus stereo microscope (Center Valley, PA, USA).

Microangiography was performed as previously published [49, 128] and described in detail online [129]. Briefly, 3 dpf live control and WAF treated tg[fli:eGfp] larvae were anesthetized with Tricaine (MS222, Argent Chemical Laboratories, Inc, Redmond, WA, USA) to prevent skeletal muscle contractions, mounted ventral side up in 5% methylcellulose, and microinjected with Qtracker 655 non-targeted quantum dots (Invitrogen, Life technologies, Carlsbad, CA, USA) into the sinus venosus. Larvae were then laterally embedded in 0.75% agarose on glass bottom petri dishes (MatTek, Ashland, MA, USA), and imaged with an inverted Leica SP5 Laser Scanning Confocal Microscope (Lieca, Buffalo Grove, IL, USA) with resonant scanning. Live cell imaging of the fluorescently labeled green endothelial cells and red quantum dot-filled vasculature was captured at 20X using bi-directional scanning at 8,000 hz, 16-times line averaging at a 1024 × 1024 resolution, and with a 1.7 × digital zoom. Z-stacks with an optical slice thickness of 1 μm were collected of the entire trunk or portions of the head. Z-stack analysis and maximum intensity projections were performed using Volocity software (PerkinElmer, Waltham, MA, USA). Adobe Photoshop was used to align projections and reconstruct the entire trunk vasculature of individual larva as illustrated in Figure 4L-O.

Immunohistochemistry and imaging

Embryos were fixed in 4% phosphate-buffered formaldehyde, except for anti-Islet1 labeling, which were fixed in 4% formaldehyde, 0.05% glutaraldehyde, 5 μM EGTA, 5 μM MgSO₄ and 0.1% Triton-X-100. Primary antibodies used were mouse IgG_2b anti-Acetylated Tubulin (AT, 1:800; Sigma); rabbit anti-Glial fibrillary acidic protein (Gfap, 1:500; Dako, Carpinteria, CA, USA), mouse IgG₁ anti-Phospho-Histone H3 (PH3, 1:1000, Cell Signaling Technology, Inc., Danvers, MA USA), mouse IgG₁ anti-slow muscle myosin (F59, 1:10; Developmental Studies Hybridoma Bank, University of Iowa, Department of Biology, Iowa City, Iowa, USA), rabbit anti-Gamma-aminobutyric acid (Gaba, 1:1000; Sigma), mouse IgG₁ anti-3A10 (1:50, Developmental Studies Hybridoma Bank), and rabbit anti-activated Caspase 3 (1:500; BD Pharmigen, San Jose, CA, USA). All primary antibodies were diluted in blocking solution (see below). Secondary antibodies used were Alexa-594 goat anti-rabbit IgG, AlexaFluor-594 goat anti-mouse IgG₁, AlexaFluor-647 goat anti-mouse IgG_2b, and AlexaFluor-488 goat anti-mouse IgG₁, all obtained from Invitrogen and diluted 1:200 in blocking solution. Immunocytochemistry was carried out as previously described [130] with minor modifications. Embryos were first dehydrated at -20°C for 20 minutes with 100% MeOH followed by 7 minutes with 100% Acetone. Further antibody penetration was achieved with protein digestion using Proteinase-K (10 μg/ml) for 2 minutes to 25 minutes depending on the age of the embryos. Embryos were incubated in blocking solution (0.5 g BSA; 25 ml Phosphate buffered saline, 0.1% Triton-X-100; 5% normal goat serum; 1% DMSO) prior to both primary and secondary antibody application.

Fluorescent, whole-mount images of fixed preparations were captured using the Zeiss AxioImager epifluorescent compound microscope with ApoTome for structural illumination. Embryos were dissected into separate trunk and head segments. Trunks were laterally mounted on a microscope slide, while heads for 3A10-labeled embryos were mounted dorsal side up. Imaging was performed with either 40× or 20× objectives and Z-stacks were collected with an averaging of 2 and 0.53 μm or 0.8 μm optical slice distance, respectively. An AxioCam Mrc (Carl Zeiss, Thornwood, NY, USA) was used to capture images on the AxioImager, and processing of Z-stacks and maximum intensity projections was done using Axiovision software (Carl Zeiss, Thornwood, NY, USA).

Whole-mount in situ hybridization at for crestin and dlx2 mRNA expression was completed as previously described [131]. Alcian Blue staining was performed to visualize cartilage development at 4 dpf and conducted as previously described [50]. crestin, dlx2, tg(fli:Gfp) labeled pharyngeal arches and Alcian Blue labeling were imaged on a Zeiss Lumar stereo microscope using an AxioCam HR and Axiovision software.

All figures were constructed using Adobe Photoshop. Any adjustments to image brightness, contrast or other properties were done to the entire image and identically performed between controls and WAF-treated data.

Quantitation and statistics

Data collection was obtained in at least 20 embryos per 3 separate replicate experiments for all analyses. One exception was made for the microangiography live embryo imaging where labor intensive microinjection and live embryo mounting restricted analysis to 10 embryos per control and WAF-treated for each replicate. However, at least eight separate locations were imaged along the anterior to posterior axis of each embryo. Cell counts for PH3, Caspase, Gaba, Islet1, and Gfap were done on maximum intensity projections and quantified using the "counting tool" in Adobe Photoshop CS Extended version. Sensory neuron processes were quantified in Image J [132]. Briefly, 8-bit images were converted into binary images to calculate pixel density, and the average number of pixels for each group was analyzed for statistical significance. The results for each replicate were pooled and means calculated for most statistical analyses except where noted in the results. The two-sample T-Test was used to compare treatments with a stringency of P < 0.01 for all analyses, except when the data did not fit a normal distribution, in which case a non-parametric analysis (Kruskal-Wallis or Mann-Whitney tests) was performed. Statistical significance for fli:Gfp labeled arch formation and dlx2 expression was determined using a Fisher's Exact Test.