Application of droplet digital PCR in the analysis of genome integration and organization of the transgene in BAC transgenic mice

Transgenic (Tg) mice containing bacterial artificial chromosome (BAC) DNA are widely used for gene expression analysis and gene therapy models because BAC transgenes provide gene expression at physiological levels with the same developmental timing as endogenous genes. To ensure correct interpretation of transgene functions, investigation of the genomic organisation and integration of the BAC transgene is required. Here, we describe a reliable method based on droplet digital PCR (ddPCR) and inverse PCR to estimate copy number, genomic organisation and insertion sites of BAC transgenes in the mouse genome. We generated BAC Tg mice containing fragments of BAC clone RP23-59P20. ddPCR and iPCR analysis showed that the transgene consisted of five fragments of the BAC clone containing the Mkrn3 gene region, and that the transgene was inserted into Bckdhb, homozygous deletion of which causes the maple syrup urine disease phenotype. The ddPCR method described here should prove useful for analysis of genomic organisation and integration of BAC transgenes.

these clinical features remain poorly understood. The PWS critical region on chromosome 15q11-13, containing MKRN3, MAGEL2, NDN, SNRPN and noncoding SnoRNA clusters (SNORD115 and SNORD116), is largely conserved on chromosome 7 C in mice. Many knockout mice containing a single-or multi-gene deletion in the PWS critical region have been generated, most of which show similar features to the human PWS phenotype, but display incomplete phenotypic spectrums 7,8 . Therefore, to obtain a mouse model with all aspects of the PWS phenotype, a mouse strain with a corresponding loss of all paternal gene expression must be constructed. However, in mice, the loss of all paternally-expressed genes in the critical region results in severe failure to thrive in early life, yielding neonatal lethality [8][9][10] . To rescue these mice, exogenous gene expression in the PWS critical region is needed, but has not yet been achieved. Therefore, to rescue the neonatal lethality seen in PWS model mice, we attempted to generate Tg mice using BAC clone RP23-59P20, which contains Ndn, Magel2 and Mkrn3 (Fig. 1). In the process of generating these Tg mice, one of the mouse lines showed neonatal lethality. The genomic organization of the transgene was analysed using the ddPCR and iPCR method, revealing a complicated genomic organisation whereby the transgene was inserted into the branched-chain keto acid dehydrogenase E1 beta polypeptide gene (Bckdhb), homozygous mutations in which result in maple syrup urine disease (MSUD).

Results
Generation of Tg mice. We generated BAC Tg mice by injecting BAC clone RP23-59P20 into the pronuclei of fertilised C57BL/6 N mouse eggs. Twelve founder mice were genotyped using PCR primers targeting the 5′ and 3′ BAC vector-genomic DNA boundaries (i.e. the SP6 and T7 boundaries). Of these, one line was positive for both the SP6 and T7 boundaries, while another line (10 MT) was only positive for the T7 boundary. These two lines were bred as hemizygous Tg for several generations to ensure that no further segregation of unlinked transgenes occurred. During breeding to generate homozygous Tg mice, line 10 MT demonstrated neonatal lethality in homozygous Tg mice within 48 h of delivery. This line was therefore investigated further.
Copy number validation of the transgene by ddPCR. Because mouse line 10MT was PCR-positive for the T7 boundary but negative for the SP6 boundary, we postulated that fragments of BAC molecules containing the T7 boundary had inserted into the mouse genome. To assess the genomic localisation of the transgene, we Instead of the 87.9-kb deletion, including exons 9 and 10 of Bckdhb, the 42.5-kb transgene, consisting of five fragments (yellow boxes) of the BAC clone, is inserted into Bckdhb intron 8.
Scientific REPORTS | (2018) 8:6638 | DOI:10.1038/s41598-018-25001-x employed a ddPCR strategy. We designed PCR primers targeting the Ndn, Magel2 and Mkrn3 genes as well as the T7 boundary across the BAC. The primers successfully produced a clear distinction between positive and negative partitions in the two-dimensional analysis (Fig. 2a), and the concentrations of each of the genes and the T7 boundary were calculated (Fig. 2b). Concentrations (copies/μl) of RNaseP, located on mouse chromosome 19, were used as a control, with a known concentration of two copies per diploid genome in both wild-type and Tg mice. The ddPCR results showed that hemizygous Tg mice had three copies of Mkrn3 per diploid genome, two copies of Ndn and Magel2, and one copy of the T7 boundary. In contrast, wild-type mice contained two copies per diploid genome of all of the genes examined and no copies of the T7 boundary containing T7 sequence. We estimated that the three copies of Mkrn3 per diploid genome detected in hemizygous Tg mice consisted of two endogenous copies and one Tg copy.
Because the ddPCR assay clearly and simply showed the copy number variation within the diploid genomes, primers located between Magel2 and Mkrn3 were designed for further ddPCR analysis to identify the candidate region of the boundary, downstream of Mkrn3 between three copies and two copies per diploid genome. After several rounds of ddPCR analysis, we narrowed the candidate region to a 500-bp segment between the Mk + 1.3 and Mk + 3 loci (Fig. 2b,c). Because the results were reproducible using half volumes (10 μl) of the reaction mixtures (Fig. 2b, Supplementary Fig. S1), half volumes were routinely used for copy number validation in the genome.
Identification of unknown sequences of the transgene using inverse PCR (iPCR). After narrowing the candidate regions of the boundaries to within several hundred base pairs, iPCR was performed to identify the unknown sequences connected to the known transgene fragments. We individually used restriction endonucleases NlaIII, MseI and Sau3AI, each of which recognizes a 4-bp sequence (CATG, TTAA and GATC, respectively), and digests, on average, every several hundred bases. Digested DNA fragments from wild-type and hemizygous Tg mice were self-ligated and then amplified by nested PCR. To identify the unknown sequence connected to known fragments between the Mk + 3 and Mk + 1.3 loci, the Sau3A1-digested fragments were amplified using an inner primer, and then an aliquot of the product from the first PCR assay was amplified using outer primers ( Supplementary Fig. S2). Although agarose gel electrophoresis showed that similarly sized products were amplified from wild-type and Tg DNA (Fig. 3a), direct sequencing of the PCR products revealed overlapped sequences of the unknown connected fragment, starting at chr7: 62,268,615 in UCSC Genome Browser on Mouse Dec. 2011 (GRCm38/mm10) (Fig. 3b, Supplementary Fig. S2b), 148 kb centromeric to Mkrn3 (Fig. 1a).
After confirming the sequence of the joined fragment by PCR using primers Mk + 1.3 R and Probe80KR (Fig. 1a, Supplementary Fig. S3), the neighbouring fragments were further investigated by ddPCR, followed by iPCR. Results showed that the transgene consisted of five fragments (total length 42.5 kb) of the BAC clone, and was inserted in the branched-chain keto acid dehydrogenase E1 beta polypeptide gene (Bckdhb) on chromosome 9 ( Fig. 1, Supplementary Fig. S3). Instead of insertion of the desired transgene, an 87.9-kb region, including the last two exons of Bckdhb, was deleted (Fig. 1b).
Gene expression in the Tg mice. Northern blot analysis has previously shown that Mkrn3 is ubiquitously expressed in almost all tissues in adult mice, with the highest levels found in the testes 11 . Recently, hypothalamic Mkrn3 mRNA levels were shown to increase in the arcuate nucleus of both male and female mice during the infantile and early juvenile periods, but decrease between postnatal days 12-15 (P12-15), prior to the onset of puberty 12 . We first analysed Mkrn3 expression in adult tissues using RT-PCR. Expression was detected in the cerebrum, spinal cord, ovaries and testes, but was not observed in the liver, kidney, spleen or heart of wild-type mice (Fig. 4a). To see if transgenic Mkrn3 expression is regulated by imprinting mechanism, because endogenous Mkrn3 is imprinted in various tissues, where it is expressed exclusively from the paternal chromosome, Mkrn3 expression analysis was performed by qPCR, on neonatal brains of Tg mice inheriting the transgene either maternally or/and paternally. Regardless of maternally or paternally inheritance of the transgene, total expression level of Mkrn3 in hemizygous Tg mice was 1.8~2.3-fold higher in the neonatal brain, while that in homozygous Tg mice was 2.7-fold higher than in the wild-type brain ( Supplementary Fig. S5). Mkrn3 expression in various tissues in hemizygous Tg mice was analysed to see if the transgenic Mkrn3 gene was expressed with the same tissue specificity as the endogenous Mkrn3 gene. qPCR analysis revealed that total expression levels were approximately 2-fold higher in all tissues examined than levels observed in wild-type mice (Fig. 4b).
As transgenic insertion in the 10MT mouse line resulted in deletion of an 87.9-kb region containing the last two exons of Bckdhb, we examined Bckdhb expression in the Tg mice. RT-PCR analysis showed that Bckdhb expression in the brain was decreased by almost 50% in hemizygous Tg mice, and no expression was observed in homozygous Tg mice (Fig. 4c). Because Bckdhb is involved in MSUD, a metabolic disorder affecting branched-chain amino acids, and homozygous Tg mice died within 48 h of delivery, we analysed the levels of branched-chain amino acids in neonatal blood samples (Fig. 4d). At P1, valine and leucine + isoleucine levels were nearly 5-and 9-fold higher, respectively, in homozygous Tg pups than in heterozygous and wild-type pups.

Discussion
In Tg mouse studies, analysis of copy number and integration sites of the transgenes are essential for downstream analysis of gene expression. In most cases, Southern blot analysis and quantitative PCR (qPCR) assays are used for copy number validation. Southern blotting is used for large-scale analysis of transgene structure, while qPCR assays, which can measure 1.25-to 1.5-fold differences in gene expression under ideal conditions, can be useful for evaluating copy number variations associated with gene/chromosomal dosage 13 . However, it is often difficult and laborious to establish ideal conditions with appropriate calibration curves for qPCR assays. Therefore, in this study, we used ddPCR in place of a qPCR approach. The ddPCR system partitions nucleic acid samples into thousands of nanolitre-sized droplets of water-oil emulsion. PCR amplification is carried out within each droplet, and the fraction of PCR-positive droplets is analysed using Poisson statistics to determine the target DNA template concentration in the sample. In ddPCR, the counting of the PCR-positive droplets is definitive and does not require the ideal conditions that are essential for qPCR, meaning that ddPCR results are more reproducible than those obtained by qPCR analysis. In addition, when compared with qPCR, ddPCR can accurately measure smaller fold differences in gene expression (<1.2-fold) 14 .
In the analysis of genomes modified with transgenes made up of fragmented BAC molecules, copy number evaluation at multiple loci across the BAC provides information about the transgene structure on a larger scale. Consistent copy numbers of more than three copies per diploid genome of all loci across the BAC indicates the genome contains one or more intact BAC molecules, while variable copy numbers between several loci across the BAC suggests that the genome contains fragmented BAC molecules. To compare copy numbers of multiple loci per diploid genome, the ddPCR method is a convenient and economical option because the QX200 ™ Droplet Generator can treat eight samples at a time using DG8 ™ cartridges, and half volumes of the reaction mixtures is enough for simple comparison of copy numbers between each sample (Fig. 2b). Tandem insertion of BAC fragments sometimes occurs in the mouse genome. To separate tandem gene copies and accurately determine copy numbers, DNA fragmentation by restriction digestion prior to droplet generation is recommended by the manufacturer. Four-base cutters that do not cut either the target or reference amplicon are preferred. In the present study, we did not perform DNA fragmentation by restriction digestion prior to droplet generation. However, when ddPCR results show ambiguous copy numbers per diploid genome, or copy numbers that are not whole numbers (e.g. copy numbers from 2.4 to 2.6 per diploid genome), DNA fragmentation should be carried out to confirm the tandem insertion of the BAC fragments, or to rule out the possibility of mosaicism of the transgenes.
In this study, we first focused on the Mkrn3 gene region, and could identify the candidate region of the border of the fragment containing Mkrn3 to within several hundred base pairs. To identify the sequence of the unknown neighbour fragment connected to the known fragment containing Mkrn3, we used an iPCR method, which involves a series of restriction digests and self-ligation steps, resulting in a looped fragment that can be primed for nested PCR from the known sequence. PCR-based chromosome walking techniques, including iPCR, are the main methods used to identify transgene-flanking sequences. The success of the PCR amplification in these techniques depends on the distance between the restriction digestion sites and the known region, meaning that the digested DNA fragments are sometimes too large for standard PCR amplification. Narrowing the candidate  Supplementary Figures (Fig. S4). (b) Quantitative RT-PCR analysis of Mkrn3 expression in tissues from wild-type and hemizygous Tg mice at P9. Error bars indicate the standard error of the means. Asterisks indicate significant differences (p < 0.05). Cer: cerebrum, HT: hypothalamus, Ov: ovary, Tes: testis. (c) Quantitative analysis of Bckdhb expression in brain tissue. Asterisks indicate significant differences (p < 0.05). (d) Amino acid analysis of neonatal blood at P1. Pup #7 was dead prior to blood sampling at P1. Wild: wild type mice, hemi: hemizygous Tg mice, homo: homozygous Tg mice.
Scientific REPORTS | (2018) 8:6638 | DOI:10.1038/s41598-018-25001-x regions of the borders by ddPCR made it possible to select suitable four-base-cutter restriction endonucleases that digest, on average, every several hundred base pairs, leading to successful amplification by iPCR.
Finally, we confirmed that the transgene consists of five fragments of the BAC clone, with a total length of 42.5 kb, and was inserted into the Bckdhb gene located on chromosome 9. It is not clear whether the observed complicated genomic organization of the transgene is specific to the RP-23 59P20 BAC clone. We could not find any previous descriptions of complicated genomic organization of BAC transgenes made up of fragmented BAC molecules. A previous report described the generation of BAC Tg mice containing a transgene consisting of the Mkrn3, Magel2 and Ndn genes to rescue PWS model mice; however, the genomic organization and integration of the BAC transgene was not described 15 . The BAC clone was mapped to the large imprinted domain, where imprinted expression of the genes is regulated by a bipartite cis-acting imprinting centre (IC) located ~2.4 Mb downstream of Mkrn3. The long-range interactions between the IC and the imprinted genes in the domain are thought to form a chromatin holocomplex structure 16 similar to an active chromatin hub 17,18 . Such complexity in the chromatin conformation may affect the fragmentation and integration of the BAC molecule in the mouse genome.
Here we described a transgene with a complicated structure, but did not rule out the possibility of additional transgenes with small fragments sizes, where copy numbers were not analysed at loci across the BAC by ddPCR. Recently, successful use of a high-throughput next-generation sequencing (NGS) platform to characterise transgene integration has been reported 19 . The application of NGS to the characterisation of Tg animals will be significant but laborious, and will be a very expensive option. However, it may prove useful in cases where traditional PCR-based methods fail to provide enough information about the transgenes.
MKRN3/ZNF127 is one of the imprinted human genes identified locationally in the imprinted domain of the PWS critical region 20 . The mouse orthologue, Mkrn3/Znf127, is also imprinted in various mouse tissues, where it is expressed exclusively from the paternal chromosome 11 . Mkrn3 encodes a RING zinc-finger protein of the Makorin family, which presumptively possesses ubiquitin-protein ligase (E3) activity 21 , and may play a role in the imprinted phenotype of mouse models of PWS. However, since 2013, MKRN3 mutations have also been reported in families with central precocious puberty 12,21,22 . This suggests that the function of MKRN3 is relevant to pubertal initiation, although PWS patients usually have delayed or incomplete puberty despite the deletion of MKRN3.
In the process of generating BAC RP23-59P20 transgenic mice, the Mkrn3 Tg mouse line was produced. The transgene includes 9 kb of the upstream region and 6 kb of the downstream region of Mkrn3, as well as the major Mkrn3 transcript (~2.9 kb) 11 . Regardless of maternally or paternally inheritance of the transgene, the level of Mkrn3 expression from the transgene was almost equal to that from the endogenous gene. Such physiologically relevant gene expression from the transgene contrasts with results observed in modified BAC Tg mice containing a Ndn-eGFP BAC transgene, in which the Ndn coding sequence was replaced with an eGFP sequence under the control of the Ndn promoter 23 . The Ndn-eGFP BAC transgene showed transcriptional regulation equivalent to monoallelic expression of Ndn, suggesting the possibility of promoter competition for transcriptional activators, or a mechanism involving physical contact in trans between promoters. The original BAC clone used in the Ndn Tg mouse study, BAC603M20 (referred as BAC109) 24 , contains a 103-kb NotI insert comprising the Ndn and Magel2 genes. In comparison, BAC clone RP23-59P20 contains a 248-kb EcoRI insert, including the 103-kb NotI insert, which is completely deleted in our fragmented transgene. It remains unclear as to whether, in the presence of the Ndn promoter, Mkrn3 is independently expressed from the transgene, or whether it is transcriptionally affected by the Ndn promoter in the transgene. A chromatin holocomplex structure between the IC and the promoters of imprinted genes, especially Mkrn3 and Ndn, has been suggested 16 .
To determine whether Tg Mkrn3 expression is physiologically relevant to the endogenous expression, Mkrn3 expression patterns in adult tissues were analysed by RT-PCR. Although previous northern blot analysis showed ubiquitous expression of Mkrn3 11 , in the current study, Mkrn3 expression was detected only in neural and gonadal tissues by RT-PCR. These discrepancies between northern blot and RT-PCR analysis results may be caused by the large size of the RT-PCR product (764 bp), which would be difficult to amplify from Mkrn3 cDNA because the 3′ UTR of Mkrn3 is AU-rich, characterized by AUUUA elements, which are suggested to be unstable 10 . Despite this possible RNA instability, high levels of Mkrn3 expression in hypothalamus and gonadal tissues provide evidence that Mkrn3 is involved in the regulation of gonadal function at both the hypothalamic-pituitary level and the gonadal level in the hypothalamic-pituitary-gonadal axis. Mice lacking an active paternal allele of Mkrn3 have previously been described as normal in appearance 11 , and the Tg mice generated in the current study, with demonstrated overexpression of Mkrn3, are fertile without any pubertal abnormalities. However, careful investigation of phenotype, especially pubertal timing, will be carried out.
Insertion of the transgene in the current model resulted in the destruction of Bckdhb, which encodes the beta subunit of the BCKDH E1 component. Homozygous mutations of the human orthologue, BCKDHB, cause MSUD [MIM 248600], an inborn metabolic disease with an autosomal recessive pattern of inheritance. BCKDHB is one of four genes that control the formation of the BCKDH enzyme complex. Based on residual BCKDH complex activity, clinical presentation and onset age, MSUD can be divided into five forms 25 . The classic form, with less than 2% residual activity, is the most severe type, and accounts for almost 75% of cases. The classic form of MSUD (cMSUD) is mostly caused by mutations in BCKDHA, BCKDHB and DBT. To date, at least 93 mutations in BCKDHB have been described in MSUD patients, according to the Human Gene Mutation Database (www.hgmd.cf.ac.uk), most of which are missense mutations. Mice with targeted deletions of Dbt, which encodes the E2 subunit of BCKDH, have been used as a cMSUD model, with a phenotype mimicking that of patients with cMSUD. The present BAC Tg mice also show neonatal lethality and have markedly elevated levels of branched-chain amino acids, indicating that they may be used as another cMSUD model.
In summary, we developed a quick and reliable method for estimating copy number and integrity of BAC transgenes in a genome using a combination of ddPCR and iPCR. In BAC Tg mice containing fragments of BAC Scientific REPORTS | (2018) 8:6638 | DOI:10.1038/s41598-018-25001-x clone RP23-59P20, the transgene consisted of five fragments containing the Mkrn3 region, and was inserted into Bckdhb. Expression analysis revealed that Tg Mkrn3 expression levels are physiologically relevant, and that Bckdhb is not expressed in homozygous Tg mice, resulting in another cMSUD mouse model.

Materials and Methods
Generation of BAC Tg mice. All experiments, designed to minimise animal pain, were carried out with the approval of the Nagasaki University Institutional Animal Care and Use Committee. BAC Tg mice were generated by pronuclear injection of fertilised C57BL6/N eggs containing BAC DNA. BAC DNA (clone RP23-59P20) was diluted to 5 µg/ml in phosphate buffered saline, and then picolitre volumes of BAC DNA (5 µg/ml) in circular form were microinjected into the male pronucleus of fertilised oocytes 26 . Litters were screened for BAC integration into somatic cells by genotypic selection as described below, using DNA isolated from tail clips taken at the time of weaning 27 . Germ line transmission was established by mating tail-positive progeny with wild-type congenic mice. Detailed information on the insert in BAC clone RP23-59P20 is available from the UCSC genome browser (http://genome.ucsc.edu/index.html). Tg mice were maintained in their original genetic background (C57BL6/N), and the transgene was transmitted to the offspring from a hemizygous transgenic parent. Wild-type and hemizygous Tg littermates across multiple litters were used for the experiments. Homozygous Tg mice were generated by crossing of hemizygous littermates, but not maintained because of neonatal lethality.
Genotypic selection and PCR assays. Genotyping of the BAC Tg mice was performed by PCR analysis.
PCR assays were designed using primers targeting the T7 and SP6 sequences in BAC vector pBACe3.6, and primers 59P20T7qR and 59P20SP6qR, which were specific to the 5′ and 3′ boundaries of mouse genomic DNA within the BAC clone. All primers used for PCR, qPCR and ddPCR analysis are listed in Supplementary Table 1.
Copy number analysis of the transgene using ddPCR. Genomic DNA extracted from the wild-type and BAC Tg mice was diluted with distilled water to 10 µg/ml and then used for all ddPCR experiments to normalise the DNA concentration between samples. As recommended for ddPCR, primers were designed to amplify products of 100-200 bp. The reaction mixtures contained ddPCR EvaGreen Supermix (Bio-Rad Laboratories, Hercules, CA, USA), primers (final concentration, 100 nM; Supplementary Table 1) and template DNA (2 μl) in a final volume of 20 μl. Each reaction was then loaded into a sample well of an eight-well disposable cartridge (DG8 ™ ; Bio-Rad Laboratories) along with 70 μl of droplet generation oil (Bio-Rad Laboratories). Droplets were formed using a QX200 ™ Droplet Generator (Bio-Rad Laboratories) as per the manufacturer's instructions.
Droplets were then transferred to a 24-well PCR plate, heat-sealed with foil and amplified to the end point using a conventional thermal cycler (95 °C for 5 min, followed by 40 cycles of 94 °C for 30 s and 60 °C for 1 min, and final extension at 98 °C for 10 min). The annealing temperature and cycle number were optimised for incremental separation between positive and negative partitions in ddPCR. The resulting products were scanned on a QX200 Droplet Reader (Bio-Rad Laboratories), and the data analysed using QuantaSoft ™ software (Bio-Rad Laboratories). When a half volume (10 μl) of the reaction mixture was applied, 40 μl of droplet generation oil was used to form droplets. iPCR and sequencing. Extracted genomic DNA was also digested with Sau3AI, MseI or NlaIII at 37 °C for 2 h. The digested DNA was then heat-inactivated at 65 °C for 20 min, and circularized by self-ligation with an equal volume of Instant Sticky-end Ligase Master Mix (New England Biolabs, Beverly, MA, USA). Nested amplification around the circular DNA using primers oriented in a direction opposite to that of conventional PCR was performed using reaction mixtures containing DNA template, 1 × Ex Taq buffer (containing 2.0 mM of MgCl 2 ), 250 μM of dNTP mix, 0.5 μM of each primer, and 0.25 U/μl of Ex Taq DNA Polymerase (Takara Bio, Otsu, Japan). Thermal cycler parameters used in the first round of amplification were as follows: 94 °C for 6 min, followed by 35 cycles of 94 °C for 30 s, 55 °C for 30 s and 72 °C for 2 min, and a final extension at 72 °C for 10 min. One microlitre of product from the first round of PCR was then used as template for a second round of amplification using the same cycling conditions. PCR amplifications were performed in an ABI9700 Thermal Cycler (Applied Biosystems, Foster City, CA, USA), and PCR products were sequenced using an ABI3130 Sequencer (Applied Biosystems). cDNA synthesis. Total RNA was isolated from tissues using an RNeasy Mini Kit (QIAGEN, Valencia, CA, USA) according to the manufacturer's protocol. The RNA was treated with amplification-grade DNaseI (Invitrogen, Grand Island, NY, USA) to degrade any genomic DNA present in the sample. cDNA was then generated from total RNA using Superscript III RNaseH Reverse Transcriptase (Invitrogen) primed with random hexamer primers. First-strand cDNA was synthesised at 42 °C for 50 min. mRNA-cDNA chains were denatured and the reverse transcriptase activity was arrested by incubating the mixtures at 70 °C for 15 min. Duplicate reactions minus the reverse transcriptase were included as negative controls.
Quantitative analysis of gene expression by RT-PCR. The cDNA was used as template for RT-PCR assays to quantify the expression of Mkrn3 and Bckdhb in the wild-type and Tg mice. Assays were carried out using SYBR Premix Ex Taq (Takara Bio) and an ABI Prism 7900 RT-PCR system (Applied Biosystems). Each sample was analysed in at least triplicate to control for PCR variation. To standardise each experiment, the results were represented as a percentage of expression, calculated by dividing the average value of the expression of the target gene by that of an internal control gene, Gapdh. Each experiment was repeated two or three times using independent RNA samples. All data were presented as the mean ± standard deviation of the means. The p-values were calculated using a two-tailed t-test.