The use of reference gene selection programs to study the silvering transformation in a freshwater eel Anguilla australis: a cautionary tale
© Setiawan and Lokman; licensee BioMed Central Ltd. 2010
Received: 24 February 2010
Accepted: 22 September 2010
Published: 22 September 2010
Quantitative real-time PCR (qPCR) has been the method of choice for the quantification of mRNA. Due to the various artifactual factors that may affect the accuracy of qPCR, internal reference genes are most often used to normalize qPCR data. Recently, many studies have employed computer programs such as GeNorm, BestKeeper and NormFinder in selecting reference genes, but very few statistically validate the outcomes of these programs. Thus, in this study, we selected reference genes for qPCR of liver and ovary samples of yellow (juvenile), migratory (silver) and 11-KT treated juveniles of New Zealand shortfinned eels (Anguilla australis) using the three computer programs and validate the selected genes statistically using REST 2009 software and the Mann-Whitney test. We also tested for the repeatability of use for the best reference genes by applying them to a data set obtained in a similar experiment conducted the previous year.
Out of six candidate genes, the combination of 18 s and eef1 was found to be the best statistically validated reference for liver, while in ovary it was l36. However, discrepancies in gene rankings were found between the different programs. Also, statistical validation procedures showed that several genes put forward as being the best by the programs were in fact, regulated, making them unsuitable as reference genes. Additionally, eef1 which was found to be a suitable - though not the top ranked - reference gene for liver tissues in one year, was regulated in another.
Our study highlights the need for external validations of reference gene selections made by computer programs. Researchers need to be vigilant in validating and reporting the rationale for the use of reference gene in published studies.
Quantitative real-time PCR (qPCR) has by far been the most widely used method for the measurement of transcript abundance. The method is applicable across most tissues and targets of interest, sensitive, relatively inexpensive and quick compared to Northern blotting. Most importantly, when employed appropriately, it allows for accurate quantification over a wide dynamic range of template amounts.
However, the level of accuracy depends upon various factors. The integrity of RNA contained in the original sample, storage conditions, the efficacy of various reagents and enzymes used in RNA extraction, purification, reverse transcription and the actual qPCR, and the type of thermal cycler used are examples of factors that affect the accuracy of qPCR data. In order to take these sources of variation into account, an internal control (reference gene) is commonly used .
Reference genes need to have uniform transcript abundance across the different groups being compared (e.g, across treatments, developmental stage, conditions, etc). As such, normalizing transcripts abundance of target genes over that of a reference gene should, in theory, eliminate artifactual variation. It is clear however, that to date, not a single universal reference gene has been found, if it even exists at all . Very often, genes that are stably expressed between treatments in one study system are found to be regulated in another. As such, careful consideration must be given to selecting and validating candidate reference genes .
More recently, many studies have employed a combination of multiple reference genes as normalizers [3–5]. This method commonly utilizes software packages, e.g., GeNorm , BestKeeper  and NormFinder , that are specifically designed to determine the most stable normalizer (i.e., with the least transcript abundance variability) amongst a set of candidates, and/or the most stable combination of genes based on their geometric averages. It is argued that the variation in the geometric average of several acceptably stable genes is usually smaller than that of any one of those genes alone, thus increasing normalizer stability . Unfortunately, most of these studies only mention which genes are considered best by the software without validating if the genes are 'good enough'.
At the heart of the issue for most researchers, is whether or not reference genes are differentially regulated between compared groups (e.g., treatment, developmental stage, etc) . In this study, we will illustrate the importance of validating the outcomes of these three reference gene selection programs. We do this by testing the transcript levels or normalization factor values (NF calculated in GeNorm) of software-selected reference genes for differences between sample groups using traditional non-parametric statistics (Mann-Whitney) and REST 2009 (Corbett Research Pty. Ltd. and Pfaffl, 2009). REST 2009 is a software application specifically created to statistically test qPCR data for pairwise differences between groups. For our purpose, we use the New Zealand shortfinned eel (Anguilla australis) as our study system.
Candidate reference genes and their functions
Catalysation of GTP-dependent binding of amynoacyl-total RNA to the ribosome; translational factor
Component of the cytoskeleton and mediates cell motility
Ornithine decarboxylase 1
Involved in the urea cycle as rate limiting enzyme in polyamine synthesis, carboxylating L-onithine into diamine putrescine
18 s ribosomal RNA
Part of the small ribosomal subunit
60 S ribosomal protein L36
Smallest protein in the large subunit of the ribosome
Nucleolar Protein 14
Processing pre 18 s rRNA and export of 40 s pre-ribosomal unit to cytoplasm
qPCR assay validations
All qPCR assays except that for eef1 of 2008 ovary samples (efficiency of 94%) had amplification efficiencies between 95% and 105%, and R2 values of ≥ 0.985. All qPCR assays produced a single amplicon as shown by single peaks during melting curve analyses. The identities of qPCR products were further confirmed through sequencing.
Analyses of transcription stability
Ranking of reference gene candidates based on overall transcription stability.
GeNorm*[M value] (without 18s)
eef1-18 s [0.316] < l36 < nop14 < actb < odc1
(eef1-l36 [0.330] < nop14 < actb < odc1)
eef1-actb [0.188] <l36 < nop14 < odc1 < 18s
BestKeeper [SD x-fold]
18 s [1.13] <eef1 [1.19] < l36< nop14< actb < odc1
l36 [1.28] <nop14 < actb < eef1 < 18s < odc1
Normfinder [Stability value] (best pair)
18 s [0.183] < nop14 [0.189] actb < l36 < eef1 < odc1
nop14 [0.125] < actb < l36< eef1 < 18s < odc1
For ovary samples, there is only a moderate agreement between the three programs (Table 2). actb was ranked in the top half by all programs, and nop14 in the top two selected by BestKeeper and NormFinder. In the ovary samples, 18 s is consistent in being the worst or second-worst performing gene
Validation of reference genes
Pair-wise statistical analyses of the inter-group stability of selected reference genes.
Yellow vs 11-KT
Yellow vs Silver
11-KT vs silver
p based on REST 2009
p based on Mann-Whitney
p based on REST 2009
p based on Mann-Whitney
p based on REST 2009
p based on Mann-Whitney
eef1 - 18 s ( GeNorm )
18 s (BestKeeper, Normfinder)
eef1 ( BestKeeper )
l36 ( BestKeeper )
eef1 - 18 s (2009 best)
eef1 (2009 2nd best)
18 s ( BestKeeper )*
l36 (2009 best)
Most of the genes or pairs of genes selected by the three programs were not sufficiently stable among liver samples, showing strong regulation with treatment (yellow vs 11-KT) or developmental stage (yellow vs silvers) (Table 3). Only the eef1-18 s combination and eef1 by itself were sufficiently stable across the three groups, and no significant pairwise differences were detected by REST 2009 or Mann-Whitney (Table 3). As indicated by the higher p values in two pairwise comparisons, the eef1-18 s combination is more stable than eef1 by itself. However, the high transcript abundance of 18 s may render it unsuitable as reference for lowly expressed genes. In these cases, eef1 would be a suitable reference.
For ovary, all of the genes or pairs of genes selected by the programs were sufficiently stable to act as references, except for nop14 which seems to be down-regulated in silver eels (Table 3; Figure 2). Among the selected reference genes, a single gene - l36 - was shown to perform best in terms of stability by yielding the lowest probability of a statistical difference between two pairwise comparisons.
The best reference genes identified from 2009 data were applicable for 2008 data for both liver and ovary (Table 3). The 2008 eef1-18 s combination for liver and l36 for ovary were suitable for use as reference genes. However, we found significant differences in two sets of pairwise comparisons for liver eef1, making it unsuitable as a reference gene on this data set. To find out if any other reference genes are suitable, we ran the reference gene selection programs on the 2008 liver data and validated the selection statistically. The results indicated that 18 s used as a single reference gene was the only other viable option for 2008 liver samples, as validated by REST 2009. However, it should be noted that Mann-Whitney analyses showed a significant difference in 18 s mRNA levels between yellow and silver groups.
In this study, we have for the first time selected reference genes that are suitable for use in qPCR of ovarian and liver samples from an anguillid eel. Our results show that for studies on the silvering of freshwater eels and the involvement of 11-KT in this event, the transcript levels for l36 and the combination of eef1 and 18 s are suitable for use as references in ovary and liver tissues, respectively. More importantly, we have for the first time demonstrated a method to validate the conclusions made by reference gene selection programs. Indeed, our validation method has surprisingly shown that a single gene (l36 in ovary) can perform better in terms of intergroup stability than a combination of genes.
Due to the widespread use of qPCR, it is unsurprising that reference gene selection programs are commonly used and form a critical element of publication of qPCR data. However, this study has highlighted several reasons for why external validations of the conclusions drawn from such programs are needed.
First, just as there is no single universal reference gene, no single program could provide the optimal reference genes for all situations. As can be expected from the different algorithms and assumptions employed, the three programs often do not agree on the best gene or rankings of genes . The optimal gene(s) for liver and ovary, and even liver itself between seasons, were selected by different programs. We highly recommend that researchers interested in computer-assisted reference gene selection use multiple programs. It is encouraging that this approach has been adopted by many similar studies [16–20].
Second, the optimal reference genes selected by a program can only be as good as their candidates. In our case, ovary had greater overall transcript stabilities than liver for all but one gene (18s). This is consistent with our findings that most program-selected reference genes performed satisfactorily for ovary, while the opposite was found for liver where only two out of six selected genes passed statistical validation.
Third, differences in study design or context may affect the performance of reference genes, even within the same species. In this study, we show that year and season affected the performance of eef1 as a reference gene. It performed satisfactorily for samples taken in summer 2009 but showed a strong effect of treatment in the autumn 2008. Changes in reference gene suitability due to differences in study design (e.g., in vivo vs in vitro) or developmental stage have been reported in many studies (e.g., locusts , cerebral ischaemia in rats , channel catfish ). Thus researchers need to be vigilant in validating the reference genes used in every study.
All three potential problems described above could be mitigated by using an independent statistical validation method. REST 2009 was written by the same group that produced BestKeeper, and its conclusions very closely matched the independent Mann-Whitney tests which strongly suggest that it is a suitable validating program. We therefore, recommend the use of REST 2009 because its specific design for qPCR data analysis circumvents the difficulties of parametric statistical analyses on data based on proportions, such as those generated by qPCR . However, other methods of qPCR data analyses are available (reviewed by Pfaffl et al 2009) and may be equally or more suitable, such as when analyses other than pair-wise comparisons are required.
There has been a realization that a great number of publications of qPCR data do not include sufficient information to allow for evaluation of reliability of the results [22, 24]. One of the key description pertains to the selection of reference genes (selection procedure, validity, etc), yet, this information is often not provided. Fewer still indicate whether or not reference genes are stably transcribed across the compared groups, with many simply citing previous research that used the same genes in a similar study system. Our findings convincingly show that such approaches are inadequate as a matter of due diligence.
We suggest that all reporting of normalized qPCR data include the following information: 1) Justification/rationale for the selection of the reference genes used (e.g., selected using computer programs, previous use in a similar study system, etc), and 2) whether or not the mRNA levels of reference genes are different between the experimental groups.
The main sampling was conducted in the austral summer of 2009 (February - March 2009). Female yellow eels (n = 20) and silver eels (n = 10), weighing between 800-1200 g, were caught by fyke net from Lake Ellesmere, Canterbury, New Zealand. Within 48 h of capture, they were intraperitoneally implanted with passive-integrated-transponder tags. At the same time, the silver and half of the yellow eels were given placebo implants (30 mg, 95% cholesterol, 5% cellulose), while the remaining yellow eels received 11-KT implants (same size and composition as placebo, but with 1 mg of 11-KT). Eels were then transferred to a salmon raceway (35 m × 5 m; New Zealand King Salmon Hatchery, Tentburn, Canterbury, New Zealand) for 28 days to allow the effects of 11-KT to be manifested in treated eels. Water was kept at a depth of c.30 cm with a unidirectional gentle flow. At the end of the exerperimental period, all eels were sacrificed with an overdose of anaesthetic (benzocaine) and liver and ovary samples taken and flash frozen in liquid nitrogen until storage in a -70°C freezer.
Our 2009 study was a refinement of one conducted in 2008. This earlier study was carried out later in autumn (April - May 2008), with a shorter captivity period (21 days). Data from the 2008 study were used to evaluate if selected reference genes (on the basis of 2009 data) were suitable for use between seasons.
RNA extraction and cDNA synthesis
Total RNA was extracted from frozen tissue samples (< 100 mg) using TRIZOL (Invitrogen) according to the manufacturer's instructions, and quantified using a spectrophotometer (NanoDrop ND 1000, Thermo Fisher Scientific). The quality of RNA samples were ascertained using an Agilent Bioanalyzer 2100 (Agilent Techonologies), according to the manufacturer's instructions. All RNA samples were shown to be of high integrity, with clear and distinct peaks at the 18 S and 28 S areas. Liver RNA samples all had RNA integrity numbers (RIN) of ≥ 9.0. However, we detected very high amounts of small RNA in ovary samples, particularly for those of yellow eels (see Additional file 1: representative electropherograms of ovary and liver RNA). This observation was consistent and was also encountered by a colleague (Dr Yuichi Ozaki) in his ovary samples. As a result, the software could not calculate RIN for ovary samples of yellow eels, but could do so for those of silver eels (RIN ≥ 8.0). Five micrograms of total RNA were then treated with Turbo DNA-free (Ambion) according to manufacturer's instructions to minimize potential genomic DNA contamination, and quantified using spectrophotometry. Information from the manufacturer stated that DNAse treatment were able to reduce DNA contamination in RNA samples by 5.4 million fold (Ambion 2009). Our own analyses conducted by comparing the Ct of DNAse treated and untreated liver cDNA showed a 512 fold reduction (99.8% reduction). One microgram of DNAse treated total RNA was then reverse-transcribed using High Capacity cDNA Reverse Transcription Kit with RNAse Inhibitor (Invitrogen) using random hexamer primers. The resulting 20 μl cDNA volume was finally diluted with 60 μl deionized water to make a cDNA concentration of 12.5 ng original total RNA/μl.
Cloning of target genes for sequencing and making of standards
Cloning parameters for candidate reference genes*
Primers (5 μM)
Anneal. temp. (°C)
Frag. size (bp)
Seq. identity (ref. species, accession number)
99% (A. anguilla, EU407825.1)
92% (A. Anguilla, DQ493907.1)
84% (Salmo salar, BT044794.1)
100% (A. Australis, FM946133)
Partial target gene cDNAs were amplified from ovary or liver templates using PCR (Bioline: Biotaq Red DNA Polymerase, 10× NH4 Buffer, 50 mM MgCl2 Solution). PCR products were electrophoresed, and amplicons of expected sizes extracted from the gel using the MinElute Gel Extraction Kit (Qiagen). Complementary DNAs were then ligated into pGEM T-Easy Vector (Promega) according to manufacturer's instructions. Ligated plasmids were transfected into Escherichia coli XL-1 Blue, grown overnight and single colonies amplified in 2YT medium. Plasmid was isolated using the Qiaprep Spin Miniprep Kit (Qiagen) and sequenced (Allan Wilson Centre Genome Service, Palmerston North, New Zealand) using M13 forward and reverse primers. Sequence identity was confirmed using the Basic Local Alignment Search Tool in the NCBI database http://blast.ncbi.nlm.nih.gov/Blast.cgi. Sequences were aligned with those known for the species or those of other teleosts. All cloned fragments were shown to have high identity with known sequences (Table 4).
Purified plasmids were linearized using restriction enzyme (Spe I or Nco I, Roche Diagnostics), and subsequently mixed with 10 volumes TE buffer and 10 volumes phenol:chloroform:isoamyl-alcohol (PCI; 25:24:1) to inactivate enzymes and remove protein residues. Following centrifugation and precipitation by sodium acetate and ethanol , plasmid-insert constructs were dissolved in TE buffer, quantified by spectrophotometry, and serially diluted for use as standards in qPCR (100 - 10-6 ng linearised plasmid/μl).
qPCR parameters for candidate reference genes
Forward primer (conc.)
Reverse primer (conc.)
cgaaagttgatagggcagaca (200 nM)
gagagcgagagaggctgaag (250 nM)
We confirmed the identities of the qPCR products by the following method. We first collected representative ovary samples at the end of a qPCR assay for each reference gene candidate, which were then gel-electrophoresed. Consistent with the results of the melting curve analyses, only single bands at the expected sizes were visible for each reference gene candidate (data not shown). DNA fragments were then extracted from the gel using the MinElute Gel Extraction Kit (Qiagen) and sequenced at the University of Otago Department of Anatomy and Structural Biology.
Selection of best reference genes
We used three software applications specifically designed to measure transcript stability and to identify the best candidate gene or best gene combination from a selection of candidate genes, GeNorm, Bestkeeper and NormFinder. All three programs were written as Microsoft Excel Visual Basic macros and are freely available from the internet.
GeNorm is by far the most widely used program among the three with 1128 citations, compared to 126 for BestKeeper and 92 for NormFinder as of July 2008 (Vandesompele 2009). GeNorm considers the best reference genes as those which show the most stable transcript levels (relative quantities based on standards or delta-Ct) across all samples, disregarding a priori comparative groupings. Stability is measured as the average pairwise variation (SD of log-transformed ratios) of a candidate gene with each of the other candidate genes (M value). Lower M value denotes higher stability. The program calculates the candidate gene with the highest transcript stability by a stepwise elimination of the least stable gene (highest M value) until only two candidate genes are left. GeNorm then calculates the geometric means of this pair of genes to become the normalization factors (NF) that can be used to normalize the transcript levels of target genes. GeNorm also calculates the optimal number of reference genes, which is useful when geometric means of the two best reference genes are not sufficiently stable, or when the addition of another gene increases stability.
Unlike GeNorm, BestKeeper uses raw Ct as inputs, rather than relative quantities. The program has as its assumption that genes that are stably transcribed, and thus suitable to act as references, should be highly correlated with each other. Therefore, in addition to ranking suitable reference gene candidate based on SD of Ct values, it also performs repeated pairwise correlation analyses of all candidate genes. The geometric means of Cts of genes that are highly and significantly correlated with each other are then calculated as an index (the equivalent of NF in GeNorm).
NormFinder uses a different approach to selecting the best reference genes. Instead of measuring overall transcript stability, it takes into account possible variation across the different sample groups of interest. It uses a "model based approach to estimation of expression variation" (Andersen et al 2004) to select suitable reference genes. In this algorithm, intra- and inter-group variations are estimated, combined into a stability value and the candidate genes ranked accordingly. NormFinder also calculates the optimal combination of two candidate reference genes for use as normalizers particularly with regard to intergroup stability. In this case, we used GeNorm to calculate NF for the gene combination.
All data were analyzed according to the instructions for each program. The best reference gene or combinations of genes that were selected by the programs were then statistically analyzed for differences between treatments or stages. The three groups being compared (yellow, silver and 11-KT) do not represent levels within a factor. The relationship between the three groups are that of stage (yellow vs silver), treatment (yellow vs 11-KT) and natural or artificial silvering (silver vs 11-KT). Therefore, it is not appropriate to use ANOVA or its equivalents.
Differences between groups were detected using non-parametric pairwise comparisons. We used REST 2009 (Corbett Research Pty. Ltd. and Pfaffl, 2009), a program specifically designed to conduct pairwise comparisons by using randomization and bootstrapping techniques. It allows the user to input multiple target genes, as well as reference genes. Based on its calculated normalized values of the target gene(s), the program then produces an output that tells the user the direction of difference between the groups, as well as the p value. The program is able to normalize the data over combinations of reference genes (based on the geometric averages). Therefore, REST 2009 is not a program originally designed to measure reference gene stability. However, it can provide a statistical output for stability between treatment groups for single genes input as 'reference', but not for combinations of reference genes.
Therefore, we had to 'trick' REST to measure the stability of reference gene combinations by including a dummy target gene variable which was given a uniform Ct value of 15. The candidate reference genes (e.g., actb, eef1, etc) were input as 'reference', the dummy gene was input as 'target'. Thus, the dummy gene was not meant to represent another normalizer, but merely a mathematical device to allow the stability of the combination of reference genes to be assessed. This method assumes that a stable combination of reference genes would not show a stably (uniformly) transcribed target gene - in this case the dummy gene - as being regulated. By testing for differences between groups for the dummy variable, we thus calculated the stability of the reference genes. In order to treat our analyses consistently, we utilized the dummy gene even when only one reference gene was being assessed (e.g., l36 for ovary). The results were virtually identical to when we did not use the dummy gene device (data not shown).
Due to this unorthodox method, we also utilized Mann-Whitney test in SPSS 17 to directly test for differences in transcript abundance or NF. Due to the multiple pair-wise comparisons, the Bonferroni correction dictates that statistical significance should be set at p < 0.016. However, since the onus is on minimizing Type II error, we have set our significance at a conservative level of p < 0.055. When 18 s was selected as the best gene by BestKeeper or NormFinder, the second best gene was also statistically validated, because 18 s may not be a suitable reference for many target genes.
Validation of results using 2008 data
Genes, or combination of genes that have been selected by the software applications as best were used as references in the 2008 data. We also tested their stabilities by statistically examining the differences in mRNA levels or NF between groups.
All experiments were undertaken in compliance with the University of Otago Animal Ethics Committee and New Zealand national standards for animal research.
We are grateful to the New Zealand Foundation for Research Science and Technology for funding a postdoctoral fellowship to ANS (Ref: UOOX0715), and Graeme Davidson and other staff at New Zealand King Salmon Co. Ltd., for the use of their raceway. We are grateful for the assistance provided by Matthew Wylie, Steven Impey, Alexandra Bowles, and the technical staff at the Department of Zoology, University of Otago.
- Vandesompele J, Kubista M, Pfaffl MW: Reference gene validation software for improved normalization. Real-time PCR: Current Technology and Applications. Edited by: Logan J, Edwards K, Saunders N. 2009, 47-64. Norfold: Caister Academic PressGoogle Scholar
- Andersen CL, Jensen JL, Ørntoft TF: Normalization of real-time quantitative reverse transcription-pcr data: a model-based variance estimation approach to identify genes suited for normalization, applied to bladder and colon cancer data sets. Cancer Research. 2004, 64: 5245-5250. 10.1158/0008-5472.CAN-04-0496View ArticlePubMedGoogle Scholar
- Scharlacken B, de Graaf DC, Goosens K, Brunain M, Peelman LJ, Jacobs FJ: Reference gene selection for insect expression studies using quantitative real-time PCR: The head of the honeybee Apis melifera, after bacterial challenge. Journal of Insect Science. 2008, 8: 33-Google Scholar
- Olsvik PA, Søfteland L, Lie KK: Selection of reference genes for qRT-PCR examination of wild populations of Atlantic cod Gadus morhua. BMC Research Notes. 2008, 2008: I:47-View ArticleGoogle Scholar
- Ouakad M, Bahi-Jaber N, Chenik M, Dellagi K, Louzir H: Selection of endogenous reference genes for gene expression analysis in Leishmania major developmental stages. Parasitology Research. 2007, 101: 473-477. 10.1007/s00436-007-0491-1View ArticlePubMedGoogle Scholar
- Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, Speleman F: Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biology. 2002, 3: 0034.0031-10.1186/gb-2002-3-7-research0034. 10.1186/gb-2002-3-7-research0034View ArticleGoogle Scholar
- Pfaffl MW, Tichopad A, Prgomet C, Neuvians TP: Determination of stable housekeeping genes, differentially regulated target genes and sample integrity: BestKeeper - Excel-based tool using pair-wise correlations. Biotechnology Letters. 2004, 26: 509-515. 10.1023/B:BILE.0000019559.84305.47View ArticlePubMedGoogle Scholar
- Inoue JG, Miya M, Miller MJ, Sado T, Hanel R, Hatooka K, Aoyama J, Minegishi Y, Nishida M, Tsukamoto K: Deep-ocean origin of the freshwater eels. Biology Letters. 2010Google Scholar
- Aarestrup K, Okland F, Hansen MM, Righton D, Gargan P, Castonguay M, Bernatchez L, Howey P, Sparholt H, Pedersen MI, McKinley RS: Oceanic spawning migration of the European eel (Anguilla anguilla). Science. 2009, 325: 1660- 10.1126/science.1178120View ArticlePubMedGoogle Scholar
- Chow S, Kurogi H, Mochioka N, Kaji S, Okazaki M, Tsukamoto K: Discovery of mature freshwater eels in the open ocean. Fisheries Science. 2009, 75: 257-259. 10.1007/s12562-008-0017-5. 10.1007/s12562-008-0017-5View ArticleGoogle Scholar
- Lokman PM, Vermeulen GJ, Lambert JGD, Young G: Gonad histology and plasma steroid profiles in wild New Zealand freshwater eels (Anguilla dieffenbachii and A. australis) before and at the onset of the natural spawning migration. I. Females. Fish Physiology and Biochemistry. 1998, 19: 325-338. 10.1023/A:1007719414295. 10.1023/A:1007719414295View ArticleGoogle Scholar
- Aoyama J, Miller MJ: The silver eel. Eel Biology. Edited by: Aida K, Tsukamoto K, Yamauchi K. 2003, 107-117. Tokyo: Springer-VerlagView ArticleGoogle Scholar
- Van Ginneken V, Durif C, Balm SP, Boot R, Verstegen MWA, Antonissen E, Van den Thillart G: Silvering of European eel (Anguilla anguilla L.): seasonal changes of morphological and metabolic parameters. Animal Biology. 2007, 57: 63-77. 10.1163/157075607780002014. 10.1163/157075607780002014View ArticleGoogle Scholar
- Rohr DH, Lokman PM, Davie PS, Young G: 11-Ketotestosterone induces silvering-related changes in immature female short-finned eels, Anguilla australis. Comparative Biochemistry and Physiology A-Molecular and Integrative Physiology. 2001, 130: 701-714. 10.1016/S1095-6433(01)00402-0. 10.1016/S1095-6433(01)00402-0View ArticleGoogle Scholar
- Lokman PM, George KAN, Divers SL, Algie M, Young G: 11-Ketotestosterone and IGF-1 increase the size of previtellogenic oocyte from shortfinned eel, Anguilla australis, in vitro. Reproduction. 2007, 133: 955-967. 10.1530/REP-06-0229View ArticlePubMedGoogle Scholar
- Sirakov M, Zarrella I, Borra M, Rizzo F, Biffali E, Arnone MI, Fiorito G: Selection and validation of a set of reliable reference genes for quantitative RT-PCR studies in the brain of the cephalopod mollusc Octopus vulgaris. BMC Molecular Biology. 2009, 10:Google Scholar
- Van Hiel MB, Van Wielendaele P, Temmerman L, Van Soest S, Vuerinckx K, Huybrechts R, Broeck JV, Simonet G: Identification and validation of housekeeping genes in brains of the desert locust Schistocerca gregaria under different developmental conditions. BMC Molecular Biology. 2009, 10:Google Scholar
- Kessler Y, Helfer-Hungerbuehler AK, Cattori V, Meli ML, Zellweger B, Ossent P, Riond B, Reusch CE, Lutz H, Hofmann-Lehmann R: Quantitative TaqMan (R) real-time PCR assays for gene expression normalisation in feline tissues. BMC Molecular Biology. 2009, 10: 106- 10.1186/1471-2199-10-106View ArticlePubMedPubMed CentralGoogle Scholar
- Li YL, Ye F, Hu Y, Lu WG, Xie X: Identification of suitable reference genes for gene expression studies of human serous ovarian cancer by real-time polymerase chain reaction. Analytical Biochemistry. 2009, 394: 110-116. 10.1016/j.ab.2009.07.022View ArticlePubMedGoogle Scholar
- Hu RB, Fan CM, Li HY, Zhang QZ, Fu YF: Evaluation of putative reference genes for gene expression normalization in soybean by quantitative real-time RT-PCR. BMC Molecular Biology. 2009, 10: 93- 10.1186/1471-2199-10-93View ArticlePubMedPubMed CentralGoogle Scholar
- Gubern C, Hurtado O, Rodriguez R, Morales JR, Romera VG, Moro MA, Lizasoain I, Serena J, Mallolas J: Validation of housekeeping genes for quantitative real-time PCR in in-vivo and in-vitro models of cerebral ischaemia. BMC Molecular Biology. 2009, 10: 57- 10.1186/1471-2199-10-57View ArticlePubMedPubMed CentralGoogle Scholar
- Small BC, Murdock CA, Bilodeau-Bourgeois AL, Peterson BC, Waldbieser GC: Stability of reference genes for real-time PCR analyses in channel catfish (Ictalurus punctatus) tissues under varying physiological conditions. Comparative Biochemistry and Physiology B-Biochemistry & Molecular Biology. 2008, 151: 296-304.View ArticleGoogle Scholar
- Pfaffl MW, Vandesompele J, Kubista M: Data analysis software. Real-time PCR: Current Technology and Applications. Edited by: Logan J, Edwards K, Saunders N. 2009, Norfolk: Caister Academic PressGoogle Scholar
- Bustin SA, Benes V, Garson JA, Hellemans J, Huggett J, Kubista M, Mueller R, Nolan T, Pfaffl MW, Shipley GL: The MIQE guidelines: Minimum Information for publication of Q uantitative real-time PCR E xperiments. Clinical Chemistry. 2009, 55: 611-622. 10.1373/clinchem.2008.112797View ArticlePubMedGoogle Scholar
- Ausubel FM, Brent R, Kingston RE, Moore DH, Seidman JG, Smith JA, Still K, : Short protocols in molecular miology: a compendium of methods from Current protocols in molecular biology. 1995, New York: WileyGoogle 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.