Determination of novel reference genes for improving gene expression data normalization in selected canine reproductive tissues - a multistudy analysis

Abstract

Background: Real time RT-PCR (qPCR) is a useful and powerful tool for quantitative measurement of gene expression. The proper choice of internal standards such as reference genes is crucial for correct data evaluation. In female dogs, as in other species, the reproductive tract is continuously undergoing hormonal and cycle stage-dependent morphological changes, which are associated with altered gene expression. However, there have been few attempts published so far targeted to the dog aimed at determining optimal reference genes for the reproductive organs. Most of these approaches relied on genes previously described in other species. Large-scale transcriptome-based experiments are promising tools for defining potential candidate reference genes, but were never considered in this context in canine research.

Results: Here, using available microarray and RNA-seq datasets derived from reproductive organs (corpus luteum, placenta, healthy and diseased uteri) of dogs, we have performed multistudy analysis to identify the most stably expressed genes for expression studies, in each tissue separately and collectively for different tissues. The stability of newly identified reference genes (EIF4H, KDELR2, KDM4A and PTK2) has been determined and ranked relative to previously used reference genes, i.e., GAPDH, β-actin and cyclophillin A/PPIA, using RefFinder and NormFinder algorithms. Finally, expression of selected target genes (luteal IL-1b and MHCII, placental COX2 and VEGFA, and uterine IGF2 and LHR) was re-evaluated and normalized. All proposed candidate reference genes were more stable, ranked higher and introduced less variation than previously used genes.

Conclusions: Based on our analyses, we recommend applying KDM4A and PTK2 for normalization of gene expression in the canine CL and placenta. The inclusion of a third reference gene, EIF4H, is suggested for healthy uteri. With this, the interpretation of qPCR data will be more reliable, allowing better understanding of canine reproductive physiology.

Keywords: Dog (Canis lupus familiaris); Reference genes; Reproductive tract.

Fig. 1

Stably-expressed genes in canine placenta (a), corpus luteum (b) and uterus (c) including different treatments, cycle/pregnancy stages and/or health and treatment status, filtered by the following criteria: the coefficient of variation (CV) < 0.2; the base mean value of number of transcripts > 500. c Analysis of uterine genes was performed with (i) all uterine samples and (ii) without diseased tissues (healthy uterine samples). d Common stable genes identified in data sets from all examined tissues. No gene was found when all uterine samples were included. However, when excluding pathological uterine samples, 36 potential candidates were identified common for all remaining samples. Asterisk (*) indicates genes chosen for further validation

Fig. 2

Ranking of proposed and commonly (previously) used reference genes in placental, luteal and healthy uterine samples, generated by RefFinder (a) and NormFinder (b) applications. The NormFinder tool was used to calculate intragroup (group refers to a particular tissue: 1-CL, 2-placenta, 3-uterus) stability values. Both softwares ranked KDM4A, EIF4H and PTK2 as the most stably expressed genes. The comparison was made including previously used reference genes (GAPDH, β-actin, cyclophilin A/PPIA) (c) Determination of the optimal number of control genes for normalization in canine CL, placenta and uterus. Pairwise variation (V_n/n + 1) analysis was done to determine the number of control genes required for accurate normalization. A cut-off value 0.15 was applied (Vandesompele et al. 2002). Pairwise variation analysis shows that V2/3 values in canine CL and placenta, and V3/4 in the uterus, were lower than 0.15, indicating that two reference genes are suitable for gene normalization in CL and placenta, but 3 genes should be included for uterus

Fig. 3

Validation of reference genes in the expression of exemplary target genes known to vary greatly in target tissues: (a) Placental expression of cyclooxygenase-2 (COX2/PTGS2) and vascular endothelial growth factor A (VEGFA) during mid-gestation, antigestagen-induced and normal luteolysis; (b) Luteal expression of interleukin 1b (IL-1b) and major histocompatibility complex class II molecules (MHCII) during mid-gestation and normal prepartum luteolysis; (c) uterine expression of insulin-like growth factor 2 (IGF2) and luteinizing hormone receptor (LHR) in non-pregnant uterus (Embryo-) and during early pregnancy (Embryo+). All experiments were normalized with commonly (previously) used reference genes (Previous ref. gen.: GAPDH, β-actin, cyclophilin A/PPIA) or proposed reference genes (Proposed ref. gen.: KDM4A, EIF4H and PTK2), ranked as the best normalizers by RefFinder and NormFinder applications, as determined by real-time (TaqMan) qPCR. Data are presented as Xg ± geometric standard deviation (SD). Normalization with routinely used genes resulted in higher intragroup variation when compared to results normalized with newly proposed genes