Microarray analysis of a salamander hopeful monster reveals transcriptional signatures of paedomorphic brain development
- Robert B Page†1,
- Meredith A Boley†1,
- Jeramiah J Smith1, 2,
- Srikrishna Putta1 and
- Stephen R Voss1Email author
© Page et al; licensee BioMed Central Ltd. 2010
Received: 16 March 2010
Accepted: 28 June 2010
Published: 28 June 2010
The Mexican axolotl (Ambystoma mexicanum) is considered a hopeful monster because it exhibits an adaptive and derived mode of development - paedomorphosis - that has evolved rapidly and independently among tiger salamanders. Unlike related tiger salamanders that undergo metamorphosis, axolotls retain larval morphological traits into adulthood and thus present an adult body plan that differs dramatically from the ancestral (metamorphic) form. The basis of paedomorphic development was investigated by comparing temporal patterns of gene transcription between axolotl and tiger salamander larvae (Ambystoma tigrinum tigrinum) that typically undergo a metamorphosis.
Transcript abundances from whole brain and pituitary were estimated via microarray analysis on four different days post hatching (42, 56, 70, 84 dph) and regression modeling was used to independently identify genes that were differentially expressed as a function of time in both species. Collectively, more differentially expressed genes (DEGs) were identified as unique to the axolotl (n = 76) and tiger salamander (n = 292) than were identified as shared (n = 108). All but two of the shared DEGs exhibited the same temporal pattern of expression and the unique genes tended to show greater changes later in the larval period when tiger salamander larvae were undergoing anatomical metamorphosis. A second, complementary analysis that directly compared the expression of 1320 genes between the species identified 409 genes that differed as a function of species or the interaction between time and species. Of these 409 DEGs, 84% exhibited higher abundances in tiger salamander larvae at all sampling times.
Many of the unique tiger salamander transcriptional responses are probably associated with metamorphic biological processes. However, the axolotl also showed unique patterns of transcription early in development. In particular, the axolotl showed a genome-wide reduction in mRNA abundance across loci, including genes that regulate hypothalamic-pituitary activities. This suggests that an axolotls failure to undergo anatomical metamorphosis late in the larval period is indirectly associated with a mechanism(s) that acts earlier in development to broadly program transcription. The axolotl hopeful monster provides a model to identify mechanisms of early brain development that proximally and ultimately affect the expression of adult phenotypes.
Darwin  proposed that evolution by natural selection is a gradual process that results in continuous phenotypic variation among species. However, there are many examples where discontinuous phenotypes are observed among related species and thus appear to evolve rapidly. That evolution could suddenly "leap forward" led to extensions of Darwin's theory to account for the rapid origin of novel phenotypes. One very old idea is that novel and dramatically different phenotypes originate via saltational evolution from mutations of genes that regulate key developmental or physiological processes during ontogeny. In particular, Goldschmidt  proposed that mutations occasionally yield individuals within populations that deviate radically from the norm and referred to such individuals as "hopeful monsters". If the novel phenotypes of hopeful monsters arise under the right environmental circumstances, they may become fixed, and the population will found a new species. While this idea was discounted during the Modern Synthesis , aspects of the hopeful monster hypothesis have been substantiated in recent years. For example, it is clear that dramatic changes in phenotype can occur from few mutations of key developmental genes and phenotypic differences among species often map to relatively few genetic factors [4–8]. These findings are motivating renewed interest in the study of hopeful monsters and the perspectives they can provide about the evolution of development [9, 10]. In contrast to mutants that are created in the lab, hopeful monsters have been shaped by natural selection and are therefore more likely to reveal mechanisms of adaptive evolution.
At least three lines of evidence led Goldschmidt  to cite the Mexican axolotl (Ambystoma mexicanum) as one of the original hopeful monsters. First, the axolotl follows a different ontogeny from other closely related tiger salamanders. Whereas some tiger salamanders undergo an obligatory metamorphosis during ontogeny that allows for a transition from an aquatic habitat to a more terrestrial habitat, the axolotl has a non-metamorphic life cycle that is often referred to as paedomorphic . This extreme example of discontinuous phenotypic variation supports a model of evolution by heterochrony: larval morphological traits of ancestral metamorphic forms are observed in the adult stages of derived paedomorphic forms. In the minds of early evolutionary biologists, these patterns were so clearly supportive of heterochrony that the Mexican axolotl became the exemplar of evolution by neoteny [12, 13]. The second reason Goldschmidt cited the axolotl was physiological - Huxley  had shown that a single molecule - thyroid hormone (TH) - was capable of rescuing metamorphosis in the axolotl. Thus, the axolotl seemed to be an example of evolution waiting around for the right macromutation to happen - simply block a single physiological step in TH regulation and a novel form is originated. The third reason Goldschmidt cited the axolotl was ecological. Previous researchers had noted that the axolotl was endemic to the high quality, permanent aquatic habitats of Xochimilco, which is near present day Mexico City . The evolution of paedomorphosis seemingly allowed the axolotl to exploit an empty niche in an environment that was devoid of predators.
Since Goldschmidt, the axolotl has remained a quintessential hopeful monster . Speculation that the paedomorphic condition of the axolotl could have a simple mechanistic basis was supported when a quantitative trait locus (QTL) was identified for the segregation of metamorphic and paedomorphic phenotypes in interspecific crosses [8, 17–20]. Previous physiological studies had also established that axolotls do not produce a sufficient titer of thyroid hormone during larval development to initiate anatomical metamorphosis [[21, 22] reviewed in [23–25]]. The evolution of axolotl hypothyroidism is thought to be associated with a mechanism that affects the development and/or function of neuroendocrine axes that regulate the release of thyroid hormone from the thyroid glands [11, 26, 27]. Conceivably, this mechanism could function during early stages of development or it could function later in the larval period when metamorphosis is initiated. Regardless, whether paedomorphic and metamorphic larvae show similar or different patterns of neurological development and function has not been previously investigated.
In this study, microarray analysis was used to investigate transcription within whole brains (including the pituitary) of the paedomorphic axolotl and a closely related metamorphic species (A. tigrinum tigrinum). The primary objective was to identify patterns of gene expression during early ontogeny that could provide new mechanistic insights about paedomorphic and metamorphic modes of development. Transcripts were sampled from both species at four chronologically matched times post hatching to obtain temporal profiles of gene expression during the early larval period and during early stages of morphological metamorphosis in A. t. tigrinum. Hundreds of genes showed different or unique patterns of expression between the species, many of which were initiated very early in the larval period and prior to the onset of morphological metamorphosis. The results suggest considerable potential for transcriptional divergence between closely related vertebrate species and highlight the tiger salamander/axolotl model system for examining mechanisms in the developing brain that determine adult phenotypic outcomes.
Larval growth and metamorphosis
Differentially expressed genes identified independently from axolotls and tiger salamanders
Relatively few uniquely expressed axolotl genes were identified overall (N = 76) and thus only 3 broad GO terms were identified as statistically enriched (Figure 4): regulation of cellular process (n = 21, p = 0.004), regulation of biological process (n = 21, p = 0.011), and biological regulation (n = 22, p = 0.015). The unique axolotl genes are predicted to function in some but not all of the biological processes observed for the shared DEG list (Additional File 2). For example, six genes that function in apoptosis (srpbp1, anax1, anax5, mtch1, gstp1, pim1) were uniquely identified for axolotls. DEGs known to be associated with vertebrate brain development were identified, including genes that code for extracellular matrix constituents and cell adhesion (e.g., mmp1, dcn, col1a1, dpt, lgals4). Also, several biomarkers of mammalian brain pathologies were uniquely up regulated in axolotls (ctss, ogn, cd69). These and other uniquely expressed genes may be associated with the axolotl's paedomorphic mode of development.
The larger list of unique DEGs from tiger larvae yielded more biological process annotations and were statistically associated with 23 GO terms (Figure 4)(Additional File 3). As was observed for the shared gene list, the cell cycle GO term was significantly enriched and these genes showed decreasing mRNA abundances during larval development. However, several GO terms were identified that were not represented in the shared or unique axolotl list, including biological processes associated with chromatin organization and biogenesis. For example, lmx1b showed a pattern of decreasing transcript abundance, as did several other genes that function in chromatin organization, modification, and gene silencing (e.g. dnmt1, baz1b, baz1a, smarca5, hist1h1b, hist1hbj, hist2h2ac). It is possible that some of these unique DEGs are associated with the maturation of brain regions that orchestrate metamorphic events. For example, several genes that function to regulate the secretion of hypothalamic, pituitary, and interrenal hormones were uniquely expressed in tiger salamander larvae, including nr3c2, prl, and sstr5. In addition to these genes, pomc and crhr1 exhibited higher expression levels in tiger salamander larvae (see real-time PCR results below). Thus, the microarray analysis identified expression differences between axolotl and tiger larvae that may correlate with HPI axis regulation and function.
Direct comparison of transcription between axolotl and tiger salamander
Further investigation of gene expression using qPCR
This study used a functional genomics approach to detail larval brain transcription between the paedomorphic Mexican axolotl and metamorphic tiger salamander. The results show that larvae of these species have different transcriptional programs that are distinguishable in two important respects. First, although shared expression patterns were observed between the species, most of the genes that were identified as differentially expressed during the larval period showed species-specific patterns of expression. Gene expression was more similar between the species at earlier time points, with pronounced differences observed at 70 and 84 dph, which coincided with the onset of anatomical metamorphosis in a subset of the tiger salamander larvae. Second, the abundance of mRNAs tended to be higher for genes that were up regulated during tiger salamander development, relative to those that were up regulated during axolotl development. Approximately 31% of genes that could be reliably and directly compared between the species were differentially expressed and 84% of these showed higher mRNA abundances in tiger salamander larvae. Below, we discuss these primary results and explore their relationships to transcriptional programming that may correlate with metamorphic and paedomorphic modes of development.
Similar expression patterns were observed for 99% of the genes that were commonly differentially expressed in axolotl and tiger salamander larvae as a function of time. This suggests that some biological processes are regulated similarly between axolotl and tiger salamander larvae during development. For example, it is possible that some of the genes that function in the specification and proliferation of neuronal cell types are similarly expressed during development in both metamorphic and paedomorphic salamanders. Genes associated with vertebrate brain development such as sox3 , msx1 , and npy  significantly increased in abundance in both species. Also clu, a gene expressed at low levels in the central nervous systems of embryonic mice before increasing during postnatal life , was similarly up regulated during the axolotl and tiger salamander larval periods. Many aspects of brain development and function are highly conserved among vertebrates. If these functions depend upon conserved patterns of gene transcription, then similarities are expected whether a salamander follows a metamorphic or paedomorphic mode of development.
Although many DEGs were expressed similarly between the species, approximately four times as many were uniquely differentially expressed in tiger salamander larvae. It is possible that many of these gene expression differences represent transcriptional responses in tiger larvae that are necessary for metamorphosis. Three lines of evidence support this idea. (1) More genes were uniquely expressed in tiger larvae and the majority of these showed larger fold changes later in the larval period, when larvae were undergoing anatomical metamorphosis. (2) Hundreds of genes were differentially expressed between axolotls and tiger salamanders throughout larval development, including the earliest time point sampled (42 dph). (3) Genes and gene functions that are likely associated with later metamorphic regulation were identified as DEGs. For example, several genes that encode chromatin structure and modifying proteins were uniquely identified from tiger larvae or were differentially expressed between the species (e.g. dnmt1, baz1a, smarca5). Also, several genes that may function in post-transcriptional modification of chromatin proteins were identified as differentially expressed between the species (e.g. sumo1, uba2, ube2e3, ube2I, ube2l3, ube2r2). It is well established that metamorphosis in amphibians and insects requires programming events that activate new transcriptional programs . Indeed, knocking out smt3 (homolog of sumo1) in Drosophila, a gene associated with chromatin remodeling by sumoylation, is known to extend the pupal stage and inhibit metamorphosis . In addition to chromatin-associated genes, the expression of several genes involved in cellular metabolic processes (e.g. cat, got2, adss, acp1, idh3g, ctps, eno3) and hormone pathways (e.g. sstr5, nr3c2, prl) increased or were expressed at higher levels (tef) during tiger salamander development. While it is not unexpected to discover gene expression differences between closely related species, the results are intriguing because transcriptional output was generally higher in tiger salamanders, for the majority of loci that showed differential expression. Moreover, some of the genes that were differentially expressed were identified early in the larval period, well before the onset of morphological metamorphosis. This suggests that metamorphic and paedomorphic modes of development are distinct in a transcriptional sense at very early stages of ontogeny, perhaps tracing back to embryogenesis .
Paedomorphosis is a heterochonic term that is classically defined as a change in the timing of development that leads to the retention of ancestral, juvenile characteristics in adults of evolutionarily derived lineages . The simplest model to explain such a pattern is a change that delays the overall rate of development. Patterns of gene expression were discovered that support the idea of developmental delay: axolotls maintained relatively constant hbe and hba transcript abundances, suggesting the maintenance of an embryonic hemoglobin expression profile throughout larval development. In comparison, transcripts for these genes declined precipitously midway during the larval period in tiger salamanders, perhaps coinciding with the initiation of early, metamorphic changes. However, developmental delay cannot explain all of the axolotl-tiger salamander expression differences. As was noted above, axolotls shared some expression patterns with tiger salamander larvae, which presumably present aspects of the ancestral metamorphic pattern. In addition, axolotls showed unique expression patterns not observed in tiger salamander larvae. Axolotls uniquely up regulated genes that are associated with vertebrate brain development and mammalian brain pathologies, including ctss and aging , ogn and pituitary cancer , and cd69 and Alzheimers . Also, an expression difference was identified that supports the idea of a depressed HPT axis in axolotls [21, 25]. Tef, a bZIP transcription factor implicated in the activation of mammalian TSHb , showed significantly lower abundances across all time points in axolotl larvae. Finally, consider the transcription of nr3c1 and nr3c2 during development. The glucocorticoid receptor (nr3c1) showed a dynamic temporal expression profile that was statistically indistinguishable in both species. Conversely, the mineralocorticoid receptor (nr3c2) was expressed at significantly lower levels in axolotl larvae. Thus, gene expression patterns in larval axolotls appear to be mosaic: some patterns are shared with tiger salamanders and some patterns are novel. Some of the novel patterns may be neutral in effect and fixed by genetic drift. However, in the case of transcription factors that may influence hypothalamic-pituitary development and activity (tef, nr3c2) or hemoglobins that allow for physiological adaptation to changing oxygen needs, the axolotl provides a model to study expression changes that have likely been selected to better suit large and reproductively competent "larval forms" for a totally aquatic life history.
In considering how discontinuous phenotypes evolve via Darwinian means, Gould  proposed an answer in the case of the axolotl:
"the problem of reconciling evident discontinuity in macroevolution with Darwinism is largely solved by the observation that small changes early in embryology accumulate through growth to yield profound differences among adults.... Delay the onset of metamorphosis and the axolotl of Lake Xochimilco reproduces as a tadpole with gills and never transforms into a salamander."
Although once a controversial idea, it is accepted now that evolution can act on early stages of development to yield novel phenotypes . There is a lengthy temporal disconnect between embryogenesis and the time a tiger salamander larva first shows morphological changes indicative of metamorphosis (after 56 DPH). During this time, there is ample time for genetic and environmental factors to affect brain development in ways that alter hypothalamic-pituitary activity late in the larval period. The results show that many genes in the brains of axolotl larvae are transcribed at lower levels than they are in tiger salamander larvae. These include genes that function in the regulation of hypothalamic-pituitary activities that orchestrate anatomical metamorphosis. This suggests the following hypothesis: an axolotl's failure to undergo metamorphosis late in the larval period traces to mechanisms that act early in development to broadly program transcription. This hypothesis can be tested by over-expressing tiger salamander genes in axolotl embryos that function to program gene expression in the brain during early development. Other hypotheses can be tested in axolotls to investigate mechanisms that direct brain development in a predictable manner, towards a hopeful monster outcome.
This study shows that axolotl and tiger salamander larvae present different brain transcriptional programs and these programs diverge early in development. These early transcriptional differences include genes whose functions associate with a number of biological processes, including cell cycle, apoptosis, chromatin structure and remodeling, cellular metabolism, transcription, post-translational modification, neural development, and regulation of the HPI and HPT axes. Studies of other metamorphic and paedomorphic species of salamander are needed to disentangle species-specific gene expression responses from those that distinguish metamorphic and paedomorphic modes of development.
A single fertilized clutch of A. t. tigrinum was obtained from Charles D. Sullivan Co. Inc and the A. mexicanum were sibs deriving from a Voss lab axolotl strain. Larvae from both species were reared individually following hatching in 40% Holtfretter's solution at 20-22°C and fed brine shrimp napulii (Artemia sp., Brine Shrimp Direct, Ogden, UT) twice daily for three weeks. After three weeks, larvae were fed California blackworms ad libitum (Lumbriculus sp., J.F. Enterprises, Oakdale, CA). At 28, 42, 56, 70, 84, and 98 dph, salamanders were anesthetized in 0.01% benzocaine and whole brains and attached pituitaries were flash-frozen in liquid nitrogen immediately following collection. Observations and measurements were collected on larval to monitor the progression of tiger larvae towards metamorphosis. Snout-vent-length (SVL) was recorded for the salamanders from which tissues were collected. A general linear model of the following form was fit to the SVL data: SVLtij = β0 + St + Ti + (ST)ti + T2i + (ST2)ti + εtij where β0 corresponds to the intercept term for axolotls, St corresponds to the intercept term for tiger salamanders, Ti corresponds to the linear regression coefficient for axolotls, (ST)ti corresponds to the linear regression coefficient for tiger salamanders, T2i corresponds to the quadratic regression coefficient for axolotls, (ST2)ti corresponds to the quadratic regression coefficient for tiger salamanders, and εtij corresponds to the error term of the jth individual from species t sampled at time i. Animal care and use was approved by the University of Kentucky Animal Care and Use Committee (IACUC protocols # 01087L2006 and #00907L2005).
Three tissue pools were developed for each time point. Whole brains and attached pituitaries were used because of the small brain size of early larvae, which yielded low amounts of RNA. Each tissue pool contained the brains of three different individuals. Total RNA was isolated using TRIzol (Invitrogen, Carlsbad, CA) and RNA samples were further purified using Qiagen RNeasy mini-columns. RNA samples were quantified via UV spectrophotometry (NanoDrop, ND-1000) and qualified via an Agilent BioAnalyzer (Agilent Technologies).
Gene expression profiling
Genome-level expression profiling was conducted using a custom Affymetrix GeneChip [28–30]. Three replicate RNA pools for each of four time points (42, 56, 70, 84 dph) were labeled, hybridized, and scanned by the University of Kentucky Microarray Core Facility according to standard Affymetrix protocols. Additional gene expression profiling for selected genes was conducted for a broader range of time points (28, 42, 56, 70, 84, and 98 dph) using qPCR. Primers (Additional File 5) were designed using Primer3 . When possible, axolotl and tiger salamander orthologs for each gene in Additional File 5 were aligned via BLAST to identify gene regions that corresponded to the same nucleotides covered by Affymetrix probe-sets. When the orthologs were not 100% identical in the target regions, separate primers were designed for each species (see Additional File 5). A BioRad iScript Select cDNA synthesis kit (Hercules, CA, USA) was used to synthesize cDNA from 1 μg of total RNA and primer efficiencies were estimated separately for axolotl and tiger salamander via linear regression on dilution series. A reference gene (tif1; probe-set L_s_at; Additional File 5) was demonstrated to be invariant across all species by time combinations and relative expression ratios were calculated according to Pfaffl . All expression ratios are relative to the mean expression of axolotl at 28 dph and normalized to tif1. All PCRs were 10 μl reactions consisting of 4 ng cDNA, 16.4 ng of forward and reverse primers, and Roche FastStart Universal SYBR Master (Rox) Mix (Roche Diagnostics, Indianapolis, IN). PCRs were conducted on an Applied Biosystems StepOnePlus real-time PCR system. Reaction conditions were as follows: 10 minutes at 95°C, 40 cycles of 15 seconds at 95°C followed by 1 minute at 55°C, 15 seconds at 95°C, and 1 minute at 55°C. Melting curves were generated to ensure amplification of a single product for each reaction. All reactions were run on 48 well plates and blocked by sampling time and species (i.e., for a given time point, both species were present on the plate). At least two template free controls were present on each plate .
Quality control and low-level analyses of the Ambystoma GeneChip
All arrays were subjected to quality control (QC) at the individual probe level by inspecting box-plots, histograms, pair-wise M versus A plots of replicate GeneChips, pseudo-images of probe level models, and an RNA degradation plot that allows for visualization of the 3' labeling bias across all GeneChips simultaneously [46, 47]. Background correction, normalization, and expression summaries were obtained via the robust multi-array average (RMA) algorithm . Two RMA expression matrices were generated separately for each species (see below) and a third RMA expression matrix was computed from all arrays from both species (see below). Upon implementing the RMA algorithm, the probe-set level data from each of these three matrices were subjected to further QC by inspecting pair-wise M vs. A plots of replicate GeneChips and examining correlation matrices among replicate GeneChips (minimum mean r for a given species by time point combination across all three of the RMA expression matrices = 0.989). Probe-sets from the two species-specific RMA matrices that were classified as "absent" on > 75% of the GeneChips were filtered .
Identification of identical probe-sets
A total of 1604 (~33%) of the 4844 probe-sets on the Ambystoma GeneChip were designed from contigs that have predicted orthologs in axolotl and tiger salamander. The sequences of the probe-sets were used as queries in BLAST searches of axolotl and tiger salamander EST contigs . BLAST alignments were used to extrapolate the number of mismatches (MM) between microarray probes designed to axolotl and orthologous EST contigs from tiger salamander, and vice versa. These data were then used to calculate the number of probes in each probe-set that had > 0 MM and the sum of MM across each probe-set. Probe-sets that had > 0 MM between species were filtered before conducting statistical analyses that directly compared expression values between axolotl and tiger salamander larvae (see below).
Identification of differentially expressed genes
Two statistical approaches were used to identify DEGs. First, RMA matrices were generated for each species and quadratic regression  was used to identify genes that changed as a function of time. This approach also classified genes into nine different temporal profiles based on the values of the estimated regression coefficients (Figure 3; see also ). The "flat" profile describes genes that do not show transcript abundance changes (null results). The LU, LD, QLVU, and QLCD profiles described genes that show linear (LU, LD) or nonlinear (QLVU, QLCU, QLCD, QLVD) changes in transcript abundance across sample times. The QV and QC expression profiles described genes that show transient changes. Statistical correction for multiple testing was done separately for each species by evaluating α0 at a false discovery rate (FDR- ) of 0.05. α1 was set to 0.05. In addition to statistical criteria, genes were only retained if they exhibited ≥ 1.5 fold changes relative to 42 dph (baseline) at one or more of the other time points (56, 72, or 84 dph).
The second statistical approach used the global RMA matrix to directly compare axolotl and tiger salamander expression levels/profiles for genes known to exhibit zero sequence divergence in the regions encompassed by Affymetrix probe-sets (see above). This analysis was conducted using the maSigPro software package  that is available from bioconductor http://www.bioconductor.org for the R statistical computing environment http://www.r-project.org. In short, maSigPro was used to fit second order (i.e., quadratic) regression models, in which the species term is identified by a dummy variable, in a gene-by-gene manner. Correction for multiple testing was achieved by evaluating the over-all model P-values according to the algorithm of Benjamini and Hochberg  at an FDR of 0.05. A backward selection procedure was then used to eliminate non-significant (α = 0.05) terms from significant models. In order for genes from this analysis to be considered "identified" they had to meet the following criteria: (1) over-all model P-values lower than the FDR adjusted threshold, (2) significant species, time × species, or time2 × species terms, (3) an R 2 ≥ 0.50, and (4) a ≥ 1.5 fold difference between axolotl and tiger salamander at one or more of the sampling times (42, 56, 70, or 84 dph).
Identification of statistically enriched biological processes
To identify biological processes that were statistically enriched in our lists of DEGs, we conducted EASE analyses using the database for annotation visualization and integrated discovery (DAVID). For all analyses, the 3728 genes on the Ambystoma GeneChip with established orthologies to humans were used to generate expected values (i.e., as the background). The count threshold was always set to two and the EASE threshold was always set to 0.05. The list of significant GO terms was manually inspected to remove redundant terms.
Statistical analysis of the qPCR data
General linear models were fit to qPCR estimates of mRNA abundance to determine if gene expression differed in magnitude and/or temporal profile between the species. These models took the form: Log2(R)tij = β0 + St + Ti + (ST)ti + T2i + (ST2)ti + T3i + (ST3)ti + εtij where β0 = the intercept term for axolotl, St = the intercept term for tiger salamander, Ti = the linear regression coefficient for axolotl, (ST)ti = an additional linear regression coefficient for tiger salamander, T2i = the quadratic regression coefficient for axolotl, (ST2)ti = an additional quadratic regression coefficient for tiger salamander, T3i = the trinomial regression coefficient for axolotl, (ST3)ti = an additional trinomial regression coefficient for tiger salamander, and εtij= the error term associated with jth RNA pool from species t and time i. When necessary, these models were simplified via a backward selection scheme that removed non-significant terms (P > 0.05).
We thank James Monaghan, John Walker, Phil Crowley, and Bruce O'Hara for helpful discussion and comments. Donna Walls and the University of Kentucky Microarray Core Facility processed samples for array analysis. The research was supported by grant R24-RR016344 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of NCRR or NIH. The project was also supported by funds from the Kentucky Spinal Cord and Brain Injury Research Trust. The Spinal Cord and Brain Injury Research Center and the NSF supported Ambystoma Genetic Stock Center (DBI-0443496) provided resources and facilities.
- Darwin C: On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life. 1859, John Murray, LondonGoogle Scholar
- Goldschmidt R: The Material Basis of Evolution. 1940, Yale University Press. New HavenGoogle Scholar
- Simpson GG: Tempo and Mode in Evolution. 1940, Columbia University Press, New YorkGoogle Scholar
- Hoekstra HE, Hirschmann RJ, Bundey RA, Insel PA, Crossland JP: A single amino acid mutation contributes to adaptive beach mouse color pattern. Science. 2006, 313: 101-104. 10.1126/science.1126121.View ArticlePubMedGoogle Scholar
- Cresko WA, Amores A, Wilson C, Murphy J, Currey M, Phillips P, Bell MA, Kimmel CB, Postlethwait JH: Parallel genetic basis for repeated evolution of armor loss in Alaskan threespine stickleback populations. Proc Natl Acad Sci. 2004, 101: 6050-6065. 10.1073/pnas.0308479101.PubMed CentralView ArticlePubMedGoogle Scholar
- Schemske DW, Bradshaw HD: Pollinator preference and the evolution of floral traits in monkeyflowers (Mimulus). Proc Natl Acad Sci. 1999, 96: 11910-11915. 10.1073/pnas.96.21.11910.PubMed CentralView ArticlePubMedGoogle Scholar
- Doebley JA, Stec A, Hubbard L: The evolution of apical dominance in maize. Nature. 1997, 386: 485-488. 10.1038/386485a0.View ArticlePubMedGoogle Scholar
- Voss SR, Shaffer HB: Adaptive evolution via a major gene effect: paedomorphosis in the Mexican axolotl. Proc Natl Acad Sci. 1997, 94: 14185-14189. 10.1073/pnas.94.25.14185.PubMed CentralView ArticlePubMedGoogle Scholar
- Theissen G: The proper place of hopeful monsters in evolutionary biology. Theory Biosci. 2006, 124: 349-369. 10.1016/j.thbio.2005.11.002.View ArticlePubMedGoogle Scholar
- Theissen G: Saltational evolution: hopeful monsters are here to stay. Theory Biosci. 2009, 128: 43-51. 10.1007/s12064-009-0058-z.View ArticlePubMedGoogle Scholar
- Shaffer HB, Voss SR: Phylogenetic and mechanistic analysis of a developmentally integrated character complex: Alternate life history modes in ambystomatid salamanders. Am Zool. 1996, 36: 24-35.View ArticleGoogle Scholar
- Kollman J: Das Ueberwintern von europäis-chen Frosch- und Triton-Iarven und die Umwand-lung des mexikanischen Axolotl. Verhandlungen der Naturforschenden Gesellschaft in Basel. 1885, 7: 387-398.Google Scholar
- Gould SJ: Ontogeny and Phylogeny. 1977, Belkap Press, Cambridge, MAGoogle Scholar
- Huxley JS: Metamorphosis of axolotl caused by thyroid feeding. Nature. 1920, 104: 2618-10.1038/104435b0.View ArticleGoogle Scholar
- Gadow H: The Mexican axolotl. Nature. 1903, 67: 330-332. 10.1038/067330b0.View ArticleGoogle Scholar
- Gould SJ: The Return of Hopeful Monsters. Nat Hist. 1977, 86: 22-30.Google Scholar
- Voss SR: Genetic basis of paedomorphosis in the axolotl, Ambystoma mexicanum: a test of the single gene hypothesis. J Hered. 1995, 86: 441-447.Google Scholar
- Voss SR, Shaffer HB: Evolutionary genetics of metamorphic failure using wild-caught vs. laboratory axolotls (Ambystoma mexicanum). Mol Ecol. 2000, 9: 1401-1407. 10.1046/j.1365-294x.2000.01025.x.View ArticlePubMedGoogle Scholar
- Voss SR, Prudic K, Oliver J, Shaffer HB: Candidate gene analysis of metamorphic timing in ambystomatid salamanders. Mol Ecol. 2003, 12: 1217-1223. 10.1046/j.1365-294X.2003.01806.x.View ArticlePubMedGoogle Scholar
- Voss SR, Smith JJ: Evolution of salamander life cycles: A major-effect quantitative trait locus contributes to discrete and continuous variation for metamorphic timing. Genetics. 2005, 170: 275-281. 10.1534/genetics.104.038273.PubMed CentralView ArticlePubMedGoogle Scholar
- Ducibella T: The occurrence of biochemical metamorphic events without anatomical metamorphosis in the axolotl. Dev Biol. 1974, 38: 175-86. 10.1016/0012-1606(74)90268-1.View ArticlePubMedGoogle Scholar
- Galton VA: Thyroid hormone receptors and iodothyronine deiodinases in the developing Mexican axolotl, Ambystoma mexicanum. Gen Comp Endo. 1991, 85: 62-70. 10.1016/0016-6480(92)90172-G.View ArticleGoogle Scholar
- Rosenkilde P: The role of hormones in the regulation of amphibian metamorphosis. Metamorphosis. Edited by: Balls M, Bownes M. 1985, Oxford: Clarendon Press, 221-259.Google Scholar
- Kuhn ER, Jacobs GFM: Metamorphosis. Developmental Biology of the Axolotl. Edited by: Armstrong J, Malacinski G. 1989, Oxford, Oxford University Press, 187-197.Google Scholar
- Rosenkilde P, Ussing AP: What mechanisms control neoteny and regulate induced metamorphosis in urodeles?. Int J Dev Biol. 1996, 40: 665-673.PubMedGoogle Scholar
- Tompkins R: Genic control of axolotl metamorphosis. Amer Zool. 1978, 18: 313-319.View ArticleGoogle Scholar
- Gould SJ: Change in developmental timing as a mechanism of macroevolution. Evolution and Development. Edited by: Bonner JT. 1981, Berlin, Springer-Verlag, 333-346.Google Scholar
- Monaghan JR, Walker JA, Page RB, Putta S, Beachy CK, Voss SR: Early gene expression during natural spinal cord regeneration in the salamanders Ambystoma mexicanum. J Neurochem. 2007, 101: 27-40. 10.1111/j.1471-4159.2006.04344.x.View ArticlePubMedGoogle Scholar
- Page RB, Monaghan JR, Samuels AK, Smith JJ, Beachy CK, Voss SR: Microarray analysis identifies keratin loci as sensitive biomarkers for thyroid hormone disruption in salamanders (Ambystoma). Comp Bioch Physiol Part C. 2007, 145: 15-27.Google Scholar
- Page RB, Monaghan JR, Walker JA, Voss SR: A model of transcriptional and morphological changes during thyroid hormone-induced metamorphosis of the axolotl. Gen Comp Endo. 2009, 162: 219-32. 10.1016/j.ygcen.2009.03.001.View ArticleGoogle Scholar
- Rifkin SA, Kim J, White KP: Evolution of gene expression in the Drosophila melanogaster subgroup. Nat Genet. 2003, 33: 138-44. 10.1038/ng1086.View ArticlePubMedGoogle Scholar
- Nikcević G, Savić T, Kovacević-Grujicić N, Stevanović M: Up-regulation of the SOX3 gene expression by retinoic acid: characterization of the novel promoter-response element and the retinoid receptors involved. J Neurochem. 2008, 107: 1206-15. 10.1111/j.1471-4159.2008.05670.x.View ArticlePubMedGoogle Scholar
- Bach A, Lallemand Y, Nicola MA, Ramos C, Mathis L, Maufras M, Robert B: Msx1 is required for dorsal diencephalons patterning. Development. 2003, 130: 4025-4036. 10.1242/dev.00609.View ArticlePubMedGoogle Scholar
- Agasse F, Bernardino L, Kristiansen H, Christiansen SH, Ferreira R, Silva B, Grade S, Woldbye DP, Malva JO: Neuropeptide Y promotes neurogenesis in murine subventricular zone. Stem Cells. 2008, 26: 16636-1645. 10.1634/stemcells.2008-0056.View ArticleGoogle Scholar
- Charnay Y, Imhof A, Vallet PG, Hakkoum D, Lathuiliere A, Poku N, Aronow B, Kovari E, Bouras C, Giannakopoulos P: Clusterin expression during fetal and postnatal CNS development in mouse. Neuroscience. 2008, 155: 714-724. 10.1016/j.neuroscience.2008.06.022.View ArticlePubMedGoogle Scholar
- Tata JR: Gene expression during metamorphosis: an ideal model for post-embryonic development. Bioessays. 1993, 15: 239-48. 10.1002/bies.950150404.View ArticlePubMedGoogle Scholar
- Talamillo A, Sánchez J, Cantera R, Pérez C, Martín D, Caminero E, Barrio R: Smt3 is required for Drosophila melanogaster metamorphosis. Development. 2008, 135: 1659-1668. 10.1242/dev.020685.View ArticlePubMedGoogle Scholar
- Wendt W, Lübbert H, Stichel CC: Upregulation of cathepsin S in the aging and pathological nervous system of mice. Brain Res. 2008, 1232: 7-20. 10.1016/j.brainres.2008.07.067.View ArticlePubMedGoogle Scholar
- Hu SM, Li F, Yu HM, Li RY, Ma QY, Ye TJ, Lu ZY, Chen JL, Song HD: The mimecan gene expressed in human pituitary and regulated by pituitary transcription factor-1 as a marker for diagnosing pituitary tumors. J Clin Endo Met. 2005, 90: 6657-6664. 10.1210/jc.2005-0322.View ArticleGoogle Scholar
- Kusdra L, Rempel H, Yaffe K, Pulliam L: Elevation of CD69+ monocyte/macrophages in patients with Alzheimer's disease. Immunnobiology. 2000, 202: 26-33.View ArticleGoogle Scholar
- Drolet DW, Scully KM, Simmons DM, Wegner M, Chu KT, Swanson LW, Rosenfeld MG: TEF, a transcription factor expressed specifically in the anterior pituitary during embryogenesis, defines a new class of leucine zipper proteins. Genes Dev. 1991, 5: 1739-1753. 10.1101/gad.5.10.1739.View ArticlePubMedGoogle Scholar
- Raff RA: The Shape of Life. Genes, Development, and the Evolution of Animal Form. 1996, University Chicago Press, ChicagoGoogle Scholar
- Rozen S, Skaletsky H: Primer3 on the www for general users and for biologist programmers. Methods Mol Biol. 2000, 132: 365-86.PubMedGoogle Scholar
- Pfaffl M: A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Research. 2001, 29: e45-10.1093/nar/29.9.e45.PubMed CentralView ArticlePubMedGoogle Scholar
- Bustin SA, Nolan T: Pitfalls of quantitative real-time reverse-transcription polymerase chain reaction. J Biomolec Techniques. 2004, 15: 155-66.Google Scholar
- Bolstad BM, Irizarry RA, Gautier L, Wu Z: Preprocessing high-density oligonucleotide arrays. Bioinformatics and Computational Biology Solutions Using R and Bioconductor. Edited by: Gentaleman R, Carey VJ, Huber W, Irizarry RA, Dudoit S. 2005, New York: Springer, 13-32. full_text.View ArticleGoogle Scholar
- Bolstad BM, Collin F, Brettschneider J, Simpson K, Cope L, Irizarry RA, Speed TP: Quality assessment of Affymetrix GeneChip data. Bioinformatics and Computational Biology Solutions Using R and Bioconductor. Edited by: Gentaleman R, Carey VJ, Huber W, Irizarry RA, Dudoit S. 2005, New York: Springer, 33-47. full_text.View ArticleGoogle Scholar
- Izarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, Speed TP: Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics. 2003, 4: 249-264. 10.1093/biostatistics/4.2.249.View ArticleGoogle Scholar
- McClintick JN, Edenberg HJ: Effects of filtering by present call on analysis of microarray experiments. BMC Bioinformatics. 2006, 7: 49-10.1186/1471-2105-7-49.PubMed CentralView ArticlePubMedGoogle Scholar
- Smith JJ, Putta S, Walker JA, Kump DC, Samuels AK, Monaghan JR, Weisrock DW, Staben C, Voss SR: Integrating new and existing ambystomatid research and information resources. BMC Genomics. 2005, 6: 181-10.1186/1471-2164-6-181.PubMed CentralView ArticlePubMedGoogle Scholar
- Liu H, Tarima S, Borders AS, Getchell TV, Gertchell ML, Stromberg AJ: Quadratic regression analysis for gene discovery and pattern recognition for non-cyclic short time-course microarray experiments. BMC Bioinformatics. 2005, 6: 106-123. 10.1186/1471-2105-6-106.PubMed CentralView ArticlePubMedGoogle Scholar
- Page RB, Voss SR, Samuels AK, Smith JJ, Putta S, Beachy CK: Effect of thyroid hormone concentration on the transcriptional response underlying induced metamorphosis in the Mexican axolotl (Ambystoma). BMC Genomics. 2008, 9: 78-10.1186/1471-2164-9-78.PubMed CentralView ArticlePubMedGoogle Scholar
- Benjamini Y, Hochberg Y: Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Royal Stat Soc, B. 1995, 57: 289-300.Google Scholar
- Conesa A, Nueda MJ, Ferrer A, Talon M: maSigPro: a method to indentify significantly differential expression profiles in time-course microarray experiments. Bioinformatics. 2006, 22: 1096-1102. 10.1093/bioinformatics/btl056.View ArticlePubMedGoogle Scholar
- Dennis G, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, Lempicki RA: DAVID: database for annotation, visualization, and integrated discovery. Genome Biology. 2003, 4: R60-10.1186/gb-2003-4-9-r60.PubMed CentralView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.