Search tips
Search criteria 


Logo of hmgLink to Publisher's site
Hum Mol Genet. Nov 15, 2010; 19(22): 4409–4420.
Published online Sep 6, 2010. doi:  10.1093/hmg/ddq364
PMCID: PMC2957326
Knockdown of zebrafish Lgi1a results in abnormal development, brain defects and a seizure-like behavioral phenotype
Yong Teng,1 Xiayang Xie,2,3 Steven Walker,2,3 Grzegorz Rempala,1,4 David J. Kozlowski,2,5 Jeff S. Mumm,2,3 and John K. Cowell1*
1MCG Cancer Center,
2Department of Cellular Biology and Anatomy,
3Vision Discovery Institute,
4Department of Biostatistics, and
5Program in Developmental Neurobiology, Institute of Molecular Medicine and Genetics, Medical College of Georgia, School of Medicine, 1120 15th Street, Augusta, GA 30912,, USA
*To whom correspondence should be addressed. Tel: +Phone: 1 7067214381; Fax: +1 7067211670; Email: jcowell/at/
Received June 7, 2010; Revised August 4, 2010; Accepted August 24, 2010.
Epilepsy is a common disorder, typified by recurrent seizures with underlying neurological disorders or disease. Approximately one-third of patients are unresponsive to currently available therapies. Thus, a deeper understanding of the genetics and etiology of epilepsy is needed to advance the development of new therapies. Previously, treatment of zebrafish with epilepsy-inducing pharmacological agents was shown to result in a seizure-like phenotype, suggesting that fish provide a tractable model to understand the function of epilepsy-predisposing genes. Here, we report the first model of genetically linked epilepsy in zebrafish and provide an initial characterization of the behavioral and neurological phenotypes associated with morpholino (MO) knockdown of leucine-rich, glioma-inactivated 1a (lgi1a) expression. Mutations in the LGI1 gene in humans have been shown to predispose to a subtype of autosomal dominant epilepsy. Low-dose Lgi1a MO knockdown fish (morphants) appear morphologically normal but are sensitized to epilepsy-inducing drugs. High-dose Lgi1a morphants have morphological defects which persist into adult stages that are typified by smaller brains and eyes and abnormalities in tail shape, and display hyperactive swimming behaviors. Increased apoptosis was observed throughout the central nervous system of high-dose morphant fish, accounting for the size reduction of neural tissues. These observations demonstrate that zebrafish can be exploited to dissect the embryonic function(s) of genes known to predispose to seizure-like behavior in humans, and offer potential insight into the relationship between developmental neurobiological abnormalities and seizure.
Epilepsy is a clinically heterogeneous disorder estimated to affect 1–2% of individuals in the USA (1,2). Over the past decade, genetic predisposition to various types of epilepsy has been linked to mutations in a variety of different genes and, not surprisingly, the majority of these genes encode structural components of ion channels (3,4). In 1995, Ottman et al. (5) described a rare form of epilepsy with mild seizures associated with acoustic auras. This genetic disorder was described as autosomal dominant partial epilepsy with auditory features (ADPEAF), also referred to as autosomal dominant lateral temporal lobe epilepsy (6). Affected members of these families were shown to have mutations in the LGI1 gene (7). LGI1 (8), which is a secreted protein (911), contains a leucine-rich repeat (LRR) domain flanked by cysteine clusters at the N-terminal end and a beta-propeller repeat in the C-terminal region of the protein. Both of these motifs indicate protein–protein interaction functions. LGI1, therefore, was the first epilepsy susceptibility gene identified that did not directly encode an ion channel protein.
The discovery of a non-ion channel gene associated with epilepsy provides the opportunity to define new mechanisms that lead to seizures. Indeed, although several reports now implicate LGI1 in synapse transmission (1215), there are suggestions that it may also influence brain development. In humans, malformations in the lateral temporal lobe were described in ADPEAF patients (16), and a reduction in hippocampal size with an altered shape of the hippocampal gyrus was seen in 10% of cases. MRI analysis of similar patients described mild atrophy of the right hemisphere and mild temporal horn asymmetry (17), whereas in others, abnormalities were recognized in the hippocampus in ADPEAF (17,18). Recently, voxel-based diffusion tenor imaging has revealed abnormalities in the left lateral temporal cortex, which were not detectable using MRI (19). In mice, overexpression of a truncated LGI1 protein resulted in abnormal structural changes in granule cell dendrites typified by incomplete pruning of apical dendrites, suggesting a role for Lgi1 in structural remodeling of the dendrite branches and spines (20). Our in vitro data suggest that LGI1 may also be involved in axon guidance pathways (21). These results support a role for LGI1 in the normal development of the brain, which may contribute to, or even underlie, the seizure phenotype in ADPEAF patients. To investigate this possibility further, we sought a model system which facilitates studies of both embryology and behavior.
Zebrafish provide an experimentally tractable animal model for vertebrate developmental neurobiology and genetically linked behaviors (2223). They are also an established vertebrate animal model for epilepsy, exhibiting seizure-like phenotypes following treatment with epilepsy-inducing drugs. As reported by several groups, zebrafish (2426) and Xenopus tadpoles (27) treated with pentylenetetrazole (PTZ) showed distinct, dose-dependent behavioral changes characteristic of seizure-like activity, which progressed through three stages: (i) rapid swimming activity, (ii) a circular ‘whirlpool’ swimming activity and (iii) progressive cyclonic seizures leading to loss of postural control. Electrophysiological analysis demonstrated ictal and inter-ictal electrophysiological discharges in neurons in the forebrain and optic tectum (24,28). Concomitant treatment of fish with PTZ and more than 15 anti-epileptic drugs abrogated the seizure phenotype in a dose-dependent manner (24,25). Given these observations, zebrafish is an attractive model for defining the role of genes that underlie hereditary, seizure-related disorders.
To investigate the function of lgi1a during early embryonic development in zebrafish, and determine whether loss of lgi1a results in seizure-like behavior, we used antisense morpholino (MO) oligonucleotides to disrupt Lgi1a expression. High-dose MO-knockdown embryos showed distinct seizure-like behaviors that were similar to those of the early stages in PTZ-treated fish. Loss of Lgi1a was also associated with reduced brain and eye size correlated with increased apoptosis in the developing brain. Lowering the MO concentration that was injected produced embryos with no gross morphological phenotype but which showed increased sensitivity to PTZ-induced seizures. Apoptosis in the brains of these fish was seen throughout the brain and was MO dose-dependent. This highly correlated relationship between loss of gene expression and behavioral phenotypes in fish and humans suggests that lgi1a knockdown fish may provide novel insights into a better understanding of the fundamental roles of LGI1 in embryonic development and seizures in adult animals.
Characterization of the zebrafish lgi1a gene
The main objectives of this study were to determine whether knockdown of lgi1 expression in zebrafish would result in seizure-like behavior and whether embryonic development, particularly of the brain, is affected. As a result of partial duplication within the zebrafish genome during evolution, there are now two copies of many genes, including lgi1. We therefore recovered the consensus sequence for the zebrafish lgi1a and lgi1b genes from the ZFIN database, and alignments of these proteins with the human and mouse sequences (Supplementary Material, Fig. S1) showed ~63 and 67% identity for lgi1a and lgi1b, respectively, with the human protein (LGI1). Significantly, at the genome level, the zebrafish lgi1 genes show exactly the same exon structure that we previously determined for human LGI1 (8), with each LRR domain occupying an individual exon. This high homology and conserved domain structure suggests a conserved function for the lgi1 genes in zebrafish. For this study, we have focused on the behavioral and developmental consequences of disrupting lgi1a expression, which was the closest to the human gene in the phylogenetic tree analysis (Supplementary Material, Fig. S1).
To knockdown lgi1a expression, we used modified antisense oligonucleotides termed ‘morpholinos’. Since MO targeting is more efficient under conditions of perfect homology with the target sequence, we first defined the sequence and structure of lgi1a in the wild-type strain used (see Materials and Methods). Separate MO-targeting strategies were designed to (i) create aberrant mRNA processing by interfering with splicing within the gene and thus generate a non-functional protein by interfering with splicing (MO-E3) and to (ii) interfere with translation by targeting the lgi1a initiation codon (MO-ATG). Perfect match MOs were designed (Fig. 1A) which would target either the initiation codon (MO-ATG) or the sequence at the 3′ end of exon 3 and the beginning of intron 3 (MO-E3). In addition, control MOs were designed, which carried five mismatch bases, MO-ATGmis and MO-E3mis (see Materials and Methods).
Figure 1.
Figure 1.
Antisense MOs affect normal splicing of lgi1a. (A) Schematic representation of a partial pre-mRNA map showing the predicted outcomes in Lgi1a morphants. The exons are shown in boxes, labeled with the corresponding exon number E1–E4, and the introns (more ...)
In situ hybridization analysis has shown previously that lgi1a is expressed in the developing eyes, ventral midbrain and hindbrain, and in the peripheral spinal cord of 24 h post-fertilization (hpf) embryos. By 48 hpf, mRNA expression was seen in the retinal ganglion cell layer of the eye, the diencephalon and ventral aspects of the hindbrain (29). We performed semi-quantitative PCR to determine relative expression levels of lgi1a during early development; lgi1a mRNA was expressed at high levels by 6hpf and was sustained at consistent levels over 7 days (Supplementary Material, Fig. S2).
lgi1a MOs were injected into one-cell stage embryos. To determine whether splicing was affected in the MO-E3 morphants, we analyzed mRNA size across exons 2–4 in the lgi1a gene using reverse transcription (RT)-PCR. Targeting the E3/I3 splice junction should result in abnormal exon splicing leading to a predictably smaller (by 72 bp) mRNA variant (Fig. 1A). As shown in Figure 1B, over a 96h period, the mRNA profile for lgi1a morphants was consistent with the loss of exon 3. This variant appeared after only 18 h and became the predominant species of mRNA after 24 h. The knockdown could be sustained for up to 96 h before the wild-type mRNA reappeared at low levels. These observations were confirmed using quantitative RT-PCR (qRT–PCR) (Fig. 1C) and demonstrate the disruption of proper lgi1a pre-mRNA splicing. In parallel experiments, the mismatch MO had no effect on splicing (Fig. 1B). To determine the effect of varying the concentration of the MO-E3 on the efficiency of splicing, and knockdown of gene expression, we injected embryos with 2, 3 and 4 ng of MOs and used semi-qRT-PCR to determine mRNA levels after 72 h. These studies demonstrated an approximately proportional relationship between MO concentration with ~50% knockdown achieved using 2 ng and >90% knockdown using 3 ng (Fig. 1D). Almost complete knockdown was seen using 4 ng, but this dose also resulted in significant premature death of the embryos.
Lgi1 morphants exhibit seizure-like behavior
Having established that the splice-blocking MO (MO-E3) generates an alternatively spliced mRNA, we next monitored the behavior of the MO-E3-injected fish. At 3 days post-fertilization (dpf), high-dose (3 ng) MO-E3 morphants showed a variety of hyperactive phenotypes, which consisted predominantly of an erratic swimming behavior typified by a tight circling motion and/or jerky directional swimming (see Supplementary Material, Movies M1–M4). In many respects, this phenotype was similar to that seen in the early stages of PTZ-induced seizures (24). This has been a consistent qualitative observation in over 10 independent experiments, demonstrating the reproducibility of the phenotype both within and between clutches. Mismatch morphants and untreated controls from the same clutch showed only normal swimming phenotypes. In parallel experiments, injection of 3 ng of the MO-ATG MO resulted in morphants displaying an even more intense hyper activity, although ~40% of these fish died by 2–3 dpf (Fig. 1E). Even when lower concentrations of the MO-ATG MO were used (2 ng), mortality was still high and was only prevented using 1 ng of the MO, where none of the behavioral/developmental abnormalities was obvious. The MO-ATGmis morphants did not show behavioral changes, and their development was normal even at high-dose (2–3 ng) treatments (data not shown). Because of the high mortality rates associated with the MO-ATG morphants, we focused largely on characterizing the seizure-like phenotype of the MO-E3 morphants.
Phenotypic characterization of morphant fish
To characterize the morphological abnormalities in the lgi1a morphants, we used a combination of confocal and conventional microscopy to image live embryos. Several phenotypes were consistently observed in the morphants (Fig. 2). First, using 3–4 ng MO (high doses), there was a decrease in the overall size and length of the MO-E3 morphants at fixed time points compared with MO-E3mis-injected siblings (Fig. 2A–C). Measurements of overall body size showed a highly significant difference between 3 ng MO-E3 morphants and their mismatch controls (Fig. 2F), which was not seen in the 2 ng morphants. The eyes were also considerably smaller in these morphants (Fig. 2B), and the brain tissue between the eyes was also reduced (Fig. 2B). Heart edema was also seen in these morphant embryos (Fig. 2C). To determine whether the difference in head and eye size was due to a general developmental delay, we stage-matched morphant and control embryos according to somite number at ~14hpf and compared morphology between 18 and 72hpf (Fig. 2D and E; see Materials and Methods). Under these circumstances, the same aberrant morphologies of the eyes, brains and overall size were noted in the morphants. These observations were consistent within and between different clutches, ruling out the possibility of a founder effect. The three phenotypes, involving eyes, brain size and overall size were all more severe in the MO-ATG morphants compared with the MO-E3 morphants (data not shown) and were virtually absent in the mismatch morphants.
Figure 2.
Figure 2.
lgi1a is essential for normal zebrafish development. (A) Lateral views of morphants at 18 and 24hpf. lgi1a MO-E3-injected embryos (3 ng) are smaller than the MO-E3mis-injected embryos. (B) Dorsal views of morphants at 60hpf. MO-E3-injected embryos show (more ...)
In addition to overall size reduction, another prominent phenotype in the high-dose MO-E3 morphants was a range of abnormal tail shapes which ranged from kinked to C-shaped (Supplementary Material, Fig. S3). Data presented in Figure 1 demonstrate that titrations of MO-E3 led to a proportional knockdown of lgi1a. Analysis of the abnormal tail phenotype at these different concentrations showed that the severity of the tail abnormality was also proportional to MO concentration, with virtually no tail shape abnormalities observed using 2 ng. These fish did not show overt hyperactivity either, even though there was an ~50% knockdown of full-length lgi1a mRNA levels.
Sensitization of Lgi1a morphants to PTZ-induced seizure-like behavior
Since the tail abnormalities described above would likely affect swimming behavior, it is possible that they contributed to the observed seizure-like phenotype, in particular circular swimming patterns. We therefore asked whether low-dose morphants were sensitized to PTZ-induced seizures. We first titrated PTZ treatments to define conditions which did not induce significant changes in swimming behaviors in 3dpf uninjected controls. PTZ concentrations between 1 and 2 mm did not produce significant hyperactivity in control fish. At 2.5 mm, a mild increase in activity was observed. This concentration, therefore, was used to test whether low-dose MO-E3 morphants displayed stronger reactivity to PTZ treatment. Any increased hyperactivity evident in 2.5 mm PTZ-treated lgi1a low-dose morphants would demonstrate a synergistic effect between Lgi1a knockdown and PTZ exposure. To test this hypothesis, we injected eggs with 2 ng of MO-E3 and selected larvae with normal tail morphology for behavioral studies at 3dpf. Selecting morphologically normal fish eliminated the possibility that aberrant tail shape was responsible for abnormal swimming behaviors. For the behavioral analysis, a ZebraBox system (ViewPoint) was used to simultaneously monitor the locomotor activities of large numbers of fish individually arrayed in multiwell plates. For each experiment, six different groups were analyzed; uninjected wild-type controls, MO-E3mis morphant controls and MO-E3 morphants, each being treated with either PTZ or control media. After a 30min ‘baseline’ observational period, either 2.5 mm PTZ or control media was added to all wells and locomotor activity was monitored over the next 2 h (see Materials and Methods). These data were visualized using MATLAB modeling software and analyzed using a custom biostatistical approach that allowed pooling data from independent experiments (see Materials and Methods).
Activity did not differ significantly between any of the three groups treated with control media (Fig. 3; uninjected versus MO-E3misinjected P= 0.15; uninjected versus MO-E3injected P= 0.57; MO-E3mis versus MO-E3 morphants P= 0.32). Interestingly, when fish were treated with 2.5 mm PTZ, marked differences in activity between control groups and MO-E3 morphants became evident (Fig. 3). Although PTZ induced a slight increase in activity in both control groups, MO-E3 morphants displayed a highly significant increase in hyperactivity compared with uninjected (P≤ 0.00001) and MO-E3mis-injected populations (P≤ 0.00001). These data demonstrate a clear synergistic effect and support the role of lgi1a in the seizure-like behaviors that is independent of gross morphological abnormalities observed with higher doses of MO.
Figure 3.
Figure 3.
Synergistic PTZ induction of seizure-like behavior in lgi1a morphants. (A) Summary of average activity over a 2h period using MATLAB modeling software time series plots. In the absence of PTZ (above) fish activity (n = 96 for each group) is relatively (more ...)
To further implicate lgi1a specifically in the seizure-like behavior, we performed mRNA rescue experiments. For this assay, a fourth group of embryos were co-injected with MO-E3 and full-length lgi1a mRNA, then introduced into the 2.5 mm PTZ-sensitivity regimen (see above). Because the injected mRNA does not require splicing, the MO-E3 should have no effect on this mRNA. This method allowed us to maintain Lgi1a expression levels, though ubiquitously. Fish from each group were selected, arrayed, treated with PTZ or control media and monitored as described above. In these experiments, rescued fish showed increased activity compared with the control or mismatch morphant fish (P< 0.0001), although their activity was significantly lower than that seen in the MO-E3 morphants after PTZ treatment (P< 0.0001). These data further support the specificity of the effect of lgi1a knockdown on behavior by showing that co-injection of processed mRNA with a splice blocking MO reduces PTZ-induced activity levels to a large extent (Fig. 3). Overall, these data provide supporting evidence that the loss of lgi1a expression is responsible for the seizure-like phenotype.
Increased apoptosis in the brain during development in lgi1a morphants
The smaller head size seen in the stage-matched comparison between high-dose MO-E3 and mismatch morphants was potentially an indication of a smaller brain size, which could result from the overall developmental delay or increased apoptosis in the developing brain. To investigate this issue further, we used acridine orange staining of stage-matched whole-mount fish to define the location and extent of apoptosis. For consistency, individual fish were compared over a range of developmental stages (Fig. 4A and B). In this analysis, it is clear that there is an increased incidence of apoptosis in the developing brain of the (3 ng) MO-E3 morphants compared with controls at three developmental stages between 10 and 35hpf. More detailed analysis shows that this apoptosis is found in the fore-, mid- and hindbrain regions. Morphological and histological analysis (Fig. 4C) shows that there are fewer cells and larger ventricles in the MO-E3 morphants compared with the MO-E3mis morphants. In addition, we performed cell death analysis using FLOW cytometry (Fig. 4D), which demonstrated an ~3-fold increase in the number of dead cells from the heads of the MO-E3 morphants. Thus, it appears that the reduced head size is related to smaller brain size, which results, at least in part, from increased apoptosis during development. When whole embryos were analyzed using TUNEL staining, an increased level of apoptosis was seen in the MO-E3 morphants compared with the MO-E3mis morphants (Fig. 4E). As shown above, addition of a rescue mRNA reduced hyperactivity and, as shown in Figure 4E, also resulted in a reduction in TUNEL staining in MO-E3 morphant embryos. Parenthetically, these rescue RNA experiments of high-dose morphants (n = 219) also virtually eliminated the developmental delay phenotypes in >60% of fish and overall development was now comparable with controls (Fig. 4E). These observations demonstrate that, even though an overall developmental delay is seen in the morphants, there is a specific increase in apoptosis in the brain, which may contribute to the seizure-like phenotype.
Figure 4.
Figure 4.
lgi1a knockdown results in apoptosis in early embryonic development. Z-stack confocal images (A: lateral view; B: dorsal view) showing cell death in living embryos following acridine orange (AO) staining at different developmental stages (scale bars: (more ...)
To define apoptosis levels in the absence of developmental abnormalities, we performed confocal studies in MO-E3 morphants treated with 2 ng MO compared with MO-E3mis morphants. We demonstrated above, in the locomotor studies, that this concentration of MO did not produce seizure-like behavior or developmental abnormalities but sensitized the fish to PTZ-induced hyperactivity. Apoptosis cells in the fore-, mid- and hindbrain regions were quantitated as described in the Materials and Methods in an analysis of approximately 50, serial, 3 µm optical sections through individual fish. As shown in Figure 5, there is a highly significant increase in apoptosis between the MO-E3 and MO-E3mis morphants.
Figure 5.
Figure 5.
Analysis of apoptosis in low-dose E3 morphants. (A) Z-stack confocal image from MO-E3 morphants compared with mismatch controls (MO-E3mis) shows a significant increase in the apoptotic cells in the MO-E3 morphants. (B) Comparison of apoptotic cells in (more ...)
In this report, we provide the first example where knockdown of a gene known to be responsible for epilepsy in humans results in seizure-like behaviors in zebrafish. Several independent observations suggest that specific loss of lgi1a is directly responsible for this phenotype: (i) targeting lgi1a function using two different approaches (translation and splicing) results in the same phenotype; (ii) a mismatch control MO does not alter mRNA splicing or mRNA levels and does not result in seizure-like behavior; (iii) co-injection of processed lgi1a mRNA with the MO-E3 MO significantly reduces seizure-like activity. The lgi1a morphants, therefore, provide a tractable model to investigate the involvement of this and other genes in the development of seizures.
LGI1 mutations in humans predispose to epilepsy, and there is some evidence that this is due in part to impaired synaptic transmission (1215). Other evidence suggests that LGI1 may have a role in brain development (1620, 30) but it is not known whether abnormal development underlies LGI1-linked seizure events. Imaging studies have suggested focal abnormalities in brains of some ADPEAF patients, although this phenotype appears heterogeneous. Previous in vitro molecular studies suggested that LGI1 might affect axon guidance pathways (21), which would have a consequence for normal brain development. Abnormal axonal targeting, if it resulted in failed synapse formation, for example, could lead to apoptosis in neurons. Increased apoptosis was observed in the brains of fish with loss of lgi1a function. In addition, hypocellularity in the brain of lgi1a morphants supports the idea that LGI1 may play a fundamental role in embryonic brain development in vertebrates.
Although hyperactivity was a common phenotype in Lgi1a morphants and PTZ-treated zebrafish, there were clearly behavioral differences between the two groups. These differences could depend on protocol differences such as the age of the fish at the time of treatment as well as the relative doses of PTZ used. In our study, we first detected increased hyperactivity using 2.5 mm PTZ, which was also reported by Baraban et al. (24), although no further details were given at this dose. Instead, this and other studies (2426) routinely used doses that were either 15 or 20 mm, in order to accentuate the seizure response in a short time frame. This high dose, therefore, could produce a phenotype severity beyond that induced by lgi1a knockdown. By comparison, the Lgi1a morphants only demonstrated a behavior that would have been scored between stages 1 and 2 in PTZ-treated fish, resembling the rapid swimming/whirlpool phenotype. Of particular relevance to Lgi1a morphants, however, was the phenotype reported in the Xenopus tadpoles (26) which, even though 15 mm PTZ was used, was described as ‘swimming in tight circles’ during the earliest stage-1 phenotype. This was exactly the phenotype seen in the Lgi1a morphants. Furthermore, a C-shaped tail movement was also described in early stages of tadpole treatment, which was also seen in the Lgi1 morphants. Thus, it appears that the seizure-like phenotype in the lgi1a morphants more closely resembles the earliest stages of PTZ-induced seizure activity, which may only be transient in the presence of high-dose PTZ. At higher MO doses, seizure-like phenotypes were more pronounced and their interpretation was complicated by the confounding observation that morphants also show pronounced tail deformities. The swimming behavior, therefore, was potentially a byproduct of improper tail development, potentially preventing conventional tail propulsion. The synergy studies, however, demonstrated that low-dose morphants with normal tail shape were more susceptible to PTZ-induced seizures, demonstrating a predisposing effect of the knockdown of lgi1a. The hyperactivity we saw following PTZ treatment of low-dose lgi1a morphants may reflect a kindling effect that sensitized the fish to drug-induced seizures. A similar effect was seen in the recently described Lgi1 mutant null mice (14,15), which showed an early onset of seizures followed by premature death, as in the zebrafish model described here. In one study (15), heterozygous Lgi1 null mice, which did not show overt seizures, showed increased susceptibility to PTZ-induced seizures compared with wild-type littermates. In this respect, the Lgi1a morphant fish mimic the phenotype seen in Lgi1 mutant mice.
Another possible explanation for the slightly different seizure-like behavior in Lgi1a morphants and PTZ-treated zebrafish is their age at the time of exposure to the seizure-inducing agent. In all reports of PTZ-only treatment, the drug was given to 6–7-day-old fish, when the majority of morphological development and primary neurogenesis was complete. In contrast, Lgi1a morphants were treated at 3dpf in order to ensure that sensitization tests were performed at a time point where knockdown of Lgi1a expression was likely to be in effect. Therefore, it may not be surprising that the specific seizure-like behavioral phenotype was different in the two paradigms. In one report (28), however, electrographic analysis of 3dpf fish treated with 15 mm PTZ demonstrated seizure activity, which was less complex than that seen in 7dpf fish. The age differences, however, may be particularly relevant if low-dose lgi1a knockdown results in abnormal brain development. Notwithstanding, the similarities in seizure-like behavior, together with the kindling of PTZ-induced seizures as a result of 50% reduction of wild-type lgi1a expression, provide convincing evidence for a causative role for Lgi1a in zebrafish seizure-like behavior.
The lgi1a morphants also demonstrate that lgi1a is important in normal embryological development in fish, since the high-dose MO-ATG morphants show severe defects in organs such as the eyes and brain and abnormal trunk development. Similar, but less severe, phenotypes were seen in the MO-E3 morphants. As expected, the penetrance of these effects is directly related to the levels of lgi1a knockdown. As suggested in other studies (31), we assume that the severe phenotype in the 3–4 ng MO-ATG fish is due to the fact that this MO targets both zygotic and maternal lgi1a mRNA, creating more of an expression null in these fish, compared with the MO-E3, which only targets zygotic mRNA. This suggestion is reinforced by rescue experiments, where a preprocessed mRNA can largely overcome the consequences of knocking down de novo Lgi1a. The spatial and temporal expression of LGI1 (29), together with the constitutional predisposition to seizures in humans, and suggestion of reduced brain size in fish and humans having reduced lgi1a/LGI1 expression levels, further points to the possibility that abnormal development of the brain may contribute to the seizure phenotype.
In summary, we describe a zebrafish model of seizure-like behavior as a result of the knockdown of the lgi1a gene, which is responsible for epilepsy in humans. In addition, we provide evidence that lgi1a may also play a role in normal development of zebrafish, particularly in the brain. This zebrafish model offers the opportunity to extend these observations to other genes known to interact with lgi1a/Lgi1 or involved in pathways downstream of the action of lgi1a/Lgi1.
Zebrafish strains and husbandry
Wild-type zebrafish of the Tü strain were maintained at 28.5°C (except as indicated below to regulate their rate of development) and bred according to a 14 h-on/10 h-off light cycle (32). Staging of zebrafish embryonic development was performed using standard morphological criteria (33). All animal protocols used in this study were reviewed and approved by the Medical College of Georgia's Institutional Animal Care and Use Committee.
Multiple-sequence alignment and phylogenetic analysis
Nucleotide and predicted protein sequences for the lgi1a (Accession CU570800) and lgi1b (Accession CU463858) genes of zebrafish and other species were obtained from GenBank. Multiple alignments of amino acid sequences for LGI1 protein were obtained using the Vector NTI 9.0 alignment program and subsequently analyzed with the neighbor-joining method to construct a phylogenetic tree.
MO oligonucleotides and rescue mRNA injection
The MO-targeting strategy can significantly affect outcome. In the experiments described here, we targeted the E3/I3 junction in lgi1a. It has been shown that targeting internal exons will usually result in the deletion of the exon (34), whereas targeting the first exon–intron boundary or last intron–exon boundary usually results in the retention of the intron, which is predicted to create a frame shift leading to a premature termination codon. Our analysis of the lgi1a exon–intron boundary sequences identified E3/I3 as the best candidate for successful MO design. Although it might have been preferable to create a premature stop codon through intron retention, the sequences surrounding the first exon–intron boundary were not so favorable for MO design. In addition, MOs targeting the donor splice site are more effective because this tract is constrained by only two bases, compared with the more extensive variability tolerated in the polypyrimidine tract of the splice acceptor (34). Deleting exon 3, however, was predicted to cause a loss-of-function mutation, since an entire repeat of the critical LRR motif, which is considered essential for lgi1a function, is lost.
Before designing specific MOs, we derived the genomic sequence for the exon–intron boundaries of the lgi1a gene from the Tü strain of fish maintained in the MCG core facility. Primers were designed using the existing database sequence to amplify the genomic regions containing the start codon and the splice junctions for the first six exons of lgi1a which contain the LRRs. We generated PCR products for each region and sequenced both DNA strands. This sequence was then used to design the antisense oligonucleotide MOs which were synthesized by Gene Tools, LLC (USA). The translation-blocking MO targeting the lgi1a 5′ UTR (MO-ATG) is 5′-CGCCGCAAACACATCATCCCGGACA-3′. The splice-inhibiting MO of lgi1a (MO-E3) was designed against the splicing donor region of the intron between exon 3 and exon 4: 5′-ATTTACTGCTGTTACTCACAGATAC-3′. As controls, MOs that included five mismatching bases were used: MO-ATGmis—5′-CGgCcCAAAgACATCATCgCcGACA-3′; MO-E3mis—5′- ATTTAgTcCTcTTACTCAgAcATAC-3′ (mismatches in lower case).
Unless otherwise stated, in all experiments, 2 ng of lgi1a MO-ATG, 2 ng of MO-E3 (low dose), 3 ng of MO-E3 (high dose) and 4 ng of MO-E3mis were injected into one-cell stage embryos. For lgi1a mRNA rescue experiments, the full-length lgi1a mRNA was subcloned into the pCS2 + vector and message synthesized in vitro using the mMessage mMachine SP6 Kit (Ambion, USA), and 30pg of mRNA per embryo was injected together with the MO-E3.
RNA isolation and semi-qRT-PCR
Ten each of uninjected, MO-E3mis- and MO-E3-injected embryos were collected at the indicated developmental stages. Total RNA was isolated from embryos using TRIzol (Invitrogen, USA) according to the manufacturer's protocol and treated with DNase (Promega, USA) to remove genomic DNA contamination. The first-strand cDNA synthesis was performed using the SuperScript II First-Strand System (Invitrogen). For the detection of the lgi1a knockdown effect, RT-PCR reactions were accomplished using two pairs of primers (Fig. 1A): p1/p2 (designed to amplify between exon 2 and exon 3) and p1/p3 (designed to amplify between exon 2 and exon 4). β-Actin was used as a loading control. The primers used in this study were as follows: p1—5′-ATCATTCGTCAAATCCGGCT- 3′; p2—5′-AGATACTCCAGATGAGGGAG-3′; p3—5′-AGGTGAATTAAAGACTTGAG-3′; β-actin forward—5′-CGAGCAGGAGATGGGAACC-3′ and β-actin reverse—5′-CAACGGAAACGCTCATTGC-3′.
Single-embryo real-time qRT-PCR analysis
Single embryos were collected at the indicated developmental stages. Total RNA was isolated from embryos and purified using TRIzol with the PureLink RNA mini kit (Invitrogen). RT was performed as described above. qRT-PCR reactions were performed in triplicate using the BioRad iQ SYBR Green Supermix, and cDNA amplification was performed using 40 cycles on a BioRad iCycler equipped with an iCycler iQ Detection System. The same p1, p2 and β-actin primers described above were used in amplification reactions and β-actin was used to normalize target gene RNA expression levels. The comparative threshold cycle (Ct) method was used to calculate amplification levels as specified by the manufacturer.
Drug treatment and behavioral monitoring
Stock solutions of PTZ (Sigma, USA) were prepared at a concentration of 1 M in 0.3× Danieau's solution (pH 6.8). In an effort to account for developmental delays incurred with MO injection, uninjected fish were incubated at 25°C from 6 to 24hpf. Embryos from each group were then stage-matched by somite number and all fish subsequently incubated at 28.5°C. At 3dpf, representative individual larvae from experimental and control groups were separated into single wells of a 48-well plate in a semi-randomized pattern such that each group was evenly distributed. Plates were then placed inside a ZebraBox and monitored using the ZebraLab video activity system (ViewPoint Life Sciences, Montreal, Canada).
Activity software such as Ethovision tracks movement by calculating the distance a fish travels in a set time period, which is then extrapolated into overall activity. This approach, however, will not accurately record hyperactivity (twitching, rotating or shaking) in fish that remain stationary or only move short distances. Viewpoint software, on the other hand, offers a method which captures all forms of activity. Instead of using distance/time calculations, this software measures fluctuations in the shift in the intensity of the pixels surrounding the zebrafish over time and converts this value into an activity measure. Each individual pixel shade can have a value between 0 and 255 (black to white). As the fish moves, these values change across the pixels in the image. The moving object within the chamber is identified as larger than a preset size (number of adjacent pixels), which eliminates background objects below this threshold. Similarly, preset values are also used to set the number of adjacent pixel shifts that are required before activity is assigned. Within this window, Viewpoint then calculates the activity of the fish based on overall pixel shade changes over time. Preset values are also used to distinguish low activity from more intense activity. Using this approach, measurements of pixel shade changes capture activity associated with both movement over large distances and shaking, rotating and twitching, which constitute more localized activity (see Supplementary Material, Movies M1–M4). The activity plots then simply display the activity changes over time based on the above calculations.
Fish behavior was monitored before treatment for 30 min to obtain an activity baseline and after PTZ treatment for 2h. Experimental fish were treated with PTZ and controls were treated with 0.3× Danieau's solution (see Results). Viewpoint activity data were processed and analyzed in MATLAB 7.6. The activity data generated by the Viewpoint system was parsed into three levels of activity: rest, mid and burst. User-defined activity parameters for the 3-day-old fish were: freezing = 3; burst = 6; detection sensitivity =16; thus, 0–3 was scored as no activity (rest), 3–6 as mid, and >6 as burst. Analyses of mid activity alone, burst alone and mid + burst activity levels gave qualitatively similar results. For consistency, mid activity levels were used to quantify all behavioral activity studies.
Statistical analysis of activity data
Measurements of fish locomotion, as described above, included increases in activity punctuated by periods of relative inactivity of the fish, which generated non-symmetrical longitudinal data distributions that were not suitable for direct ANOVA analysis. The pooled data for each experimental condition, therefore, were tested to determine whether the statistical characteristics of this activity changed with time (statistical stationarity) using procedures described by Said and Dickey (35) and Perron (36), as well as non-parametric circular bootstrap testing, with varying window size developed by Ostaszewski and Rempala (37). These analyses demonstrated the stationarity of the activity data which allowed us to use smoothing methods (thinning) to obtain approximately uncorrelated longitudinal activity measurements at 10s intervals. Using an autocorrelation assessment (38), we were then able to demonstrate the absence of lag in the data which permitted the use of the Wilcoxon two-sample, non-parametric pair-wise testing (39) for equality of distributions based on the magnitude of the observed measurements (ranks), to assess the differences between pooled measurements across the various experimental conditions. The analysis was performed using components from the statistical software R (40).
Acridine orange staining and whole-mount TUNEL assay
Live embryos (<48hpf) were dechorionated and placed in 5 ml of a 1 µg/ml solution of acridine orange (Sigma) in 10% Hank's saline solution at room temperature. After 30 min in the dark, embryos were washed four times for 10 min with 10% Hank's saline and mounted. For TUNEL assays, embryos were staged and fixed overnight in 4% PFA in PBS at 4°C, then washed in PBS/0.1% Tween 20 (PBST), dechorionated, dehydrated stepwise in methanol and stored at −20°C. After rehydration, embryos were incubated in 10µg/ml proteinase K (Sigma) for 5 min at room temperature and re-fixed in buffered 4% PFA, followed by four rinses in PBST. Fragmented genomic DNA was identified using the TdT enzyme and fluorescein-dUTP (Roche, Canada) according to the manufacturer's instructions. Images were recorded using a zoom steromicroscope.
For the quantification of apoptotic cells, MATLAB was used to count cells staining with acridine orange in optical sections through the zebrafish brain obtained using confocal microscopy. Each optical section was separated from the Hyperstack and converted to a 16-bit format for automatic counting. Background autofluorescence was established across the image based on pixel intensity, and apoptotic cells were defined showing fluorescence levels above background. As a second criterion, apoptotic cells were also defined by object size, based on calculated pixel area of individual cells. In this way, small areas of intense fluorescence that were below the threshold size were excluded. Because some regions of the image showed lower inherent autofluorescence levels, the threshold levels and intensity levels were adjusted accordingly in these areas using MATLAB calculations. The apoptotic cells were counted in each region with criteria appropriate to the background levels. As verification, each optical section was also analyzed manually to review and confirm the MATLAB assignation of apoptotic cells to further reduce false-positive calls.
Flow cytometry
Upon removal of the chorion, 10 embryos per group were washed in PBS and decapitated. The heads were then transferred into a 35 mm culture dish with 2 ml PBS, containing 0.25% trypsin and 1 mm EDTA and incubated for 30–60 min at 28.5°C during which time they were triturated with a 200 µl pipette tip every 10 min. The trypsin digestion was stopped by adding CaCl2 to a final concentration of 1 mm and FBS to 10% and the cells were then passed through a 40 µm filter. Cells were centrifuged for 3 min at 1000g, rinsed once with PBS and stained in DNA buffer (100 mm Tris, pH 7.5, 154 mm NaCl, 1 mm CaCl2, 0.5 mm MgCl2, 0.2% BSA, 0.1% NP-40, 250 µg/ml RNase, 20 µg/ml propidium iodide) for at least 30 min at 4°C in the dark. DNA content was analyzed using a Cytomics FC 500 (Beckman Coulter, USA) flow cytometer.
In vivo confocal microscopy, image processing and histology
All live imaging was performed as described previously (41). Briefly, embryos were immersed in 40°C 0.5% low-melt agarose in 0.3× Danieau's solution containing phenythiourea (PTU, Sigma) and tricaine. Droplets containing embryos were then placed in Petri dishes and manually oriented. After the agarose gelled, Petri dishes were flooded with 0.3× Danieau's solution plus PTU and tricaine. For imaging, an upright confocal laser-scanning microscope (FV1000; Olympus, Tokyo, Japan) was used and the resulting multidimensional image stacks were analyzed using ImageJ 1.41 (NIH, USA) and Photoshop CS4 software. For histological sections, embryos were rinsed in PBS three times and embedded in 1.5% agarose/5% sucrose blocks, and submerged in 30% sucrose overnight at 4°C. Blocks were frozen on dry ice and mounted using OCT compound (Tissue Tek®, Sakura, the Netherlands), and 25 µm sections were cut and stained with hematoxylin and eosin.
This work was supported in part by grants from the National Institutes of Health NS046706 (J.K.C.), MH083614 (J.S.M.) and HD053767 (D.J.K.) and MCG start-up funds (J.S.M.). J.K.C. is a Georgia Cancer Coalition Distinguished Cancer Scholar.
The authors would like to thank the staff of the MCG Transgenic Zebrafish, in particular Dr Baozheng Yuan, for their support.
Conflict of Interest statement. None declared.
1. Noebels J.L. The biology of epilepsy genes. Ann. Rev. Neurosci. 2003;26:599–625. doi:10.1146/annurev.neuro.26.010302.081210. [PubMed]
2. Kobau R., Zahran H., Thurman D.J., Zack M.M., Henry T.R., Schachter S.C., Price P.H. Centers for Disease Control and Prevention (CDC). Epilepsy surveillance among adults—19 States, Behavioral Risk Factor Surveillance System, 2005. MMWR Surveill Summ. 2008;57:1–20. [PubMed]
3. Lerche H., Jurkat-Rott K., Lehmann-Horn F. Ion channels and epilepsy. Am. J. Med. Genet. 2001;106:146–159. doi:10.1002/ajmg.1582. [PubMed]
4. Gutierrez-Delicado E., Serratosa J.M. Genetics of the epilepsies. Curr. Opin. Neurol. 2004;17:147–153. doi:10.1097/00019052-200404000-00011. [PubMed]
5. Ottman R., Risch N., Hauser W.A., Pedley T.A., Lee J.H., Barker-Cummings C., Lustenberger A., Nagle K.J., Lee K.S., Scheuer M.L., et al. Localization of a gene for partial epilepsy to chromosome 10q. Nat. Genet. 1995;10:56–60. doi:10.1038/ng0595-56. [PMC free article] [PubMed]
6. Poza J.J., Sáenz A., Martínez-Gil A., Cheron N., Cobo A.M., Urtasun M., Martí-Massó J.F., Grid D., Beckmann J.S., Prud'homme J.F., et al. Autosomal dominant lateral temporal epilepsy: clinical and genetic study of a large Basque pedigree linked to chromosome 10q. Ann. Neurol. 1999;45:182–188. doi:10.1002/1531-8249(199902)45:2&lt;182::AID-ANA8&gt;3.0.CO;2-G. [PubMed]
7. Kalachikov S., Evgrafov O., Ross B., Winawer M., Barker-Cummings C., Martinelli Boneschi F., Choi C., Morozov P., Das K., Teplitskaya E., et al. Mutations in LGI1 cause autosomal-dominant partial epilepsy with auditory features. Nat. Genet. 2002;30:335–341. doi:10.1038/ng832. [PMC free article] [PubMed]
8. Chernova O., Somerville R.P.T., Cowell J.K. A novel gene, LGI1, from region 10q24, is rearranged and downregulated in malignant brain tumors. Oncogen. 1998;17:2873–2881. doi:10.1038/sj.onc.1202481. [PubMed]
9. Senechal K.R., Thaller C., Noebels J.L. ADPEAF mutations reduce levels of secreted LGI1, a putative tumor suppressor protein linked to epilepsy. Hum. Mol. Genet. 2005;14:1613–1620. doi:10.1093/hmg/ddi169. [PubMed]
10. Sirerol-Piquer M.S., Ayerdi-Izquierdo A., Morante-Redolat J.M., Herranz-Pérez V., Favell K., Barker P.A., Pérez-Tur J. The epilepsy gene LGI1 encodes a secreted glycoprotein that binds to the cell surface. Hum. Mol. Genet. 2006;15:3436–3445. doi:10.1093/hmg/ddl421. [PubMed]
11. Head K., Gong S., Joseph S., Wang C., Burkardt T., Rossi M.R., LaDuca J., Matsui S.-I., Vaughan M., Hicks D,G., Heinz N., Cowell J.K. The expression pattern of the LGI1 gene in tissues and organs from BAC transgenic mice demonstrate a neuronal and glial expression pattern as well as other distinct cell types in the adult animal. Mamm. Genome. 2007;18:328–337. doi:10.1007/s00335-007-9024-6. [PubMed]
12. Schulte U., Thumfart J.O., Klöcker N., Sailer C.A., Bildl W., Biniossek B., Dehn D., Deller D., Eble S., Abbass K., et al. The epilepsy-linked Lgi1 protein assembles into presynaptic Kv1 channels and inhibits inactivation by Kvbeta1. Neuron. 2006;49:697–706. doi:10.1016/j.neuron.2006.01.033. [PubMed]
13. Fukata Y., Adesnik H., Iwanaga T., Bredt D.S., Nicoll R.A., Fukata M. Epilepsy-related ligand/receptor complex LGI1 and ADAM22 regulate synaptic transmission. Science. 2006;313:1792–1795. doi:10.1126/science.1129947. [PubMed]
14. Yu Y.E., Wen L., Silva J., Li Z., Head K., Sossey-Alaoui K., Pao A., Mei L., Cowell J.K. Lgi1 null mutant mice exhibit myoclonic seizures and CA1 neuronal hyperexcitability. Hum. Mol. Genet. 2010;19:1702–1711. [PMC free article] [PubMed]
15. Fukata Y., Lovero K.L., Iwanaga T., Watanabe A., Yokoi N., Tabuchi K., Shigemoto R., Nicoll R.A., Fukata M. Disruption of LGI1-linked synaptic complex causes abnormal synaptic transmission and epilepsy. Proc. Natl Acad. Sci. USA. 2010;107:3799–3804. doi:10.1073/pnas.0914537107. [PubMed]
16. Kobayashi E., Santos N.F., Torres F.R., Secolin R., Sardinha L.A., Lopez-Cendes I., Cendes F. Magnetic resonance imaging abnormalities in familial temporal lobe epilepsy with auditory auras. Arch. Neurol. 2003;60:1546–1551. doi:10.1001/archneur.60.11.1546. [PubMed]
17. Michelucci R., Poza J.J., Sofia V., de Feo M.R., Binelli S., Bisulli F., Scudellaro E., Simionati B., Zimbello R., D'Orsi G., et al. Autosomal dominant lateral temporal epilepsy: clinical spectrum, new epitempin mutations, and genetic heterogeneity in seven European families. Epilepsia. 2003;44:1289–1297. doi:10.1046/j.1528-1157.2003.20003.x. [PubMed]
18. Cendes F., Kobayashi E., Lopes-Cendes I. Familial temporal lobe epilepsy with auditory features. Epilepsia. 2005;46(Suppl. 10)):59–60. doi:10.1111/j.1528-1167.2005.00360.x. [PubMed]
19. Tessa C, Michelucci R, Nobile C, Giannelli M, Della Nave R, Testoni S, Bianucci D, Tinuper P, Bisulli F, Sofia V, et al. Structural anomaly of left lateral temporal lobe in epilepsy due to mutated LGI1. Neurology. 2007;69:1298–1300. doi:10.1212/01.wnl.0000277045.16688.b6. [PubMed]
20. Zhou Y.D., Lee S., Jin Z., Wright M., Smith S.E., Anderson M.P. Arrested maturation of excitatory synapses in autosomal dominant lateral temporal lobe epilepsy. Nat. Med. 2009;15:1208–1214. doi:10.1038/nm.2019. [PMC free article] [PubMed]
21. Kunapuli P., Lo K., Hawthorn L., Cowell J.K. Reexpression of LGI1 in glioma cells results in dysregulation of genes implicated in the canonical axon guidance pathway. Genomics. 2010;95:93–100. doi:10.1016/j.ygeno.2009.10.001. [PMC free article] [PubMed]
22. Lieschke G.J., Currie P.D. Animal models of human disease: zebrafish swim into view. Nat. Rev. Genet. 2007;8:353–356. doi:10.1038/nrg2091. [PubMed]
23. Ingham P.W. The power of the zebrafish for disease analysis. Hum. Mol. Genet. 2009;18:R107–R112. doi:10.1093/hmg/ddp091. [PubMed]
24. Baraban S.C., Taylor M.R., Castro P.A., Baier H. Pentylenetetrazole induced changes in zebrafish behavior, neural activity and c-fos expression. Neuroscience. 2005;13:759–768. doi:10.1016/j.neuroscience.2004.11.031. [PubMed]
25. Berghmans S., Hunt J., Roach A., Goldsmith P. Zebrafish offer the potential for a primary screen to identify a wide variety of potential anticonvulsants. Epilepsy Res. 2007;75:18–28. doi:10.1016/j.eplepsyres.2007.03.015. [PubMed]
26. Winter M.J., Redfern W.S., Hayfield A.J., Owen S.F., Valentin J.P., Hutchinson T.H. Validation of a larval zebrafish locomotor assay for assessing the seizure liability of early-stage development drugs. J. Pharmacol. Toxicol. Methods. 2008;57:176–188. doi:10.1016/j.vascn.2008.01.004. [PubMed]
27. Hewapathirane D.S., Dunfield D., Yen W., Chen S., Haas K. In vivo imaging of seizure activity in a novel developmental seizure model. Exp Neurol. 2008;211:480–488. doi:10.1016/j.expneurol.2008.02.012. [PubMed]
28. Baraban S.C., Dinday M.T., Castro P.A, Chege S., Guyenet S., Taylor M.R. A large-scale mutagenesis screen to identify seizure-resistant zebrafish. Epilepsia. 2007;48:1151–1157. doi:10.1111/j.1528-1167.2007.01075.x. [PMC free article] [PubMed]
29. Gu W., Gibert Y., Wirth T., Elischer A., Bloch W., Meyer A., Steinlein O.K., Begemann G. Using gene-history and expression analyses to assess the involvement of LGI genes in human disorders. Mol. Biol. Evol. 2005;22:2209–2216. doi:10.1093/molbev/msi214. [PubMed]
30. Nobile C., Michelucci R., Andreazza S., Pasini E., Tosatto S. C., Striano P. LGI1 mutations in autosomal dominant and sporadic lateral temporal epilepsy. Hum. Mut. 2009;30:530–536. doi:10.1002/humu.20925. [PubMed]
31. Eisen J. S., Smith J.C. Controlling morpholino experiments: don't stop making antisense. Development. 2008;135:1735–1743. doi:10.1242/dev.001115. [PubMed]
32. Westerfield M. The Zebrafish Book: A Guide for the Laboratory Use of Zebrafish (Brachydanio rerio) Eugene, OR: University of Oregon Press; 1993.
33. Kimmel C.B., Ballard W.W., Kimmel S.R., Ullmann B., Schilling T.F. Stages of embryonic development of the zebrafish. Dev. Dyn. 1995;203:253–310. [PubMed]
34. Morcos P.A. Achieving targeted and quantifiable alteration of mRNA splicing with morpholino oligos. Biochem. Biophys. Res. Com. 2007;358:521–527. doi:10.1016/j.bbrc.2007.04.172. [PubMed]
35. Said S.E., Dickey D.A. Testing for unit roots in time series. Biometrika. 1984;71:599–607. doi:10.1093/biomet/71.3.599.
36. Perron P. Trends and random walks in macroeconomic time series. J. Econ. Dyn. Control. 1988;12:297–332. doi:10.1016/0165-1889(88)90043-7.
37. Ostaszewski K., Rempala G. Emerging applications of the resampling methods in actuarial models. In: Shapiro A.F., Jain L.C., editors. Intelligent Techniques in the Insurance Industry: Theory and Applications. River Edge, NJ: World Scientific; 2003.
38. Venables W.N., Ripley B.D. Modern Applied Statistics with S. 4th edn. New York, NY: Springer-Verlag; 2002.
39. Hollander M., Wolfe D.A. Nonparametric Statistical Methods. New York: John Wiley & Sons; 1973. pp. 27–75.
40. R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2009.
41. Lohmann C., Mumm J., Morgan J., Godinho L., Schroeter E., Stacy R., Wong W.T., Oakley D.M., Wong R.O.L. Imaging the developing retina. In: Yuste R., Konnerth A., editors. Imaging in Neuroscience and Development. Cold Spring Harbor, NY: CSHL Press; 2005. pp. 171–184.
Articles from Human Molecular Genetics are provided here courtesy of
Oxford University Press