Introduction

Angiogenesis is required for the corpus luteum (CL) formation1,2. After the surge of luteinizing hormone (LH), endothelial cells in the theca cell layer proliferate and start to invade the granulosa cell layer during ovulation. This is the first process to form the CL3. Endothelial cells rapidly form blood vessels and vascular network in the CL, which makes CL highly vascularized organs. This event is necessary for CL continuing to produce progesterone for establishing and maintaining pregnancy. In previous studies, angiogenic factors that are responsible for luteal development have been identified4. Vascular endothelial growth factor (VEGF) is one of them and is highly expressed in rat and human CL5. VEGF in the ovary is involved in physiologic development, maintenance and maturation of blood vessels and contributes to rapid vascularization during CL formation. Therefore, VEGF is one of important genes, which leads to CL development6,7. The LH surge changes a number of gene expressions in granulosa cells (GCs) during ovulation8. Although VEGF gene expression is increased during ovulation, the detailed molecular mechanism of this regulation has been unclear. It is well known that hypoxia inducible factor 1-alpha (HIF1α) induced by hypoxic conditions increases VEGF expression in many types of cells9,10,11. However, it is unclear whether HIF1α also regulates VEGF expression in GCs undergoing luteinization.

CCAAT/enhancer-binding protein beta (C/EBPβ) is a transcription factor involved in follicle rupture and following CL formation12,13. In C/EBPα and β knockout mice, the expressions of angiogenesis-related genes are suppressed and vascular development in CL is defective12. These findings led us to hypothesize that Vegf expression is regulated by C/EBPβ in GCs during ovulation.

Not only transcription factors, but also epigenetic mechanisms regulate gene expression14,15,16. Histone modification is one of them and affects the chromatin structure, which regulates the accessibility of transcription factors into gene promoter or enhancer region17,18,19. In previous studies of GCs in rats undergoing luteinization13,20,21, we reported that changes of histone modifications and chromatin remodeling are involved in the regulation of three luteinization-related genes: steroidogenic acute regulatory protein (StAR), aromatase (Cyp19a1) and cytochrome P450 cholesterol side-chain cleavage enzyme (Cyp11a1). These findings raised the possibility that Vegf expression is regulated not only by transcription factors but also by an epigenetic mechanism.

In this study, we showed that C/EBPβ, but not HIF1α, regulates Vegf expression in rat GCs undergoing luteinization.

Results

Vegf mRNA expression

Human chorionic gonadotropin (hCG) was injected to rats to induce luteinization. GCs were isolated from ovarian follicles. Vegf mRNA levels in the GCs gradually increased and reached a peak at 12 h after hCG injection (Fig. 1).

Figure 1
figure 1

Vegf mRNA expression in rat GCs undergoing luteinization. The relative mRNA levels of Vegf in GCs before (0 h) and 4, 8, 12, 24 h after hCG injection were assessed by qPCR. Gapdh was used as an internal control. Data were shown as a ratio of those of 0 h. Each value represents the mean ± SEM of 6 animals. a, P < 0.05 vs. 0 h. mRNA expression was also assessed by RT-PCR. The images of representative gels are shown.

Full size image

C/EBPβ and HIF1α protein expressions

We examined by Western blot analyses whether C/EBPβ and HIF1α protein expression levels were increased in rat GCs after hCG injection. C/EBPβ protein (Fig. 2A) and HIF1α protein (Fig. 2B) were expressed in rat GCs before luteinization (0 h) and were up-regulated by hCG injection (12 h).

Figure 2
figure 2

C/EBPβ and HIF1α protein expressions in rat GCs undergoing luteinization. Whole cell lysate were isolated from the GCs (0 h and 12 h). Protein levels of C/EBPβ (A) and HIF1α (B) were examined by Western blotting. Protein levels of histone H3 were also assessed to ensure equal amount of proteins. Western blotting was repeated 3 times, and the representative immunoblots are shown. Quantification of bands were performed by using ImageJ and normalized with histone H3 levels. Data were shown as a ratio of those of 0 h. Each value represents the mean ± SEM of 3 independent experiments. a, P < 0.05 vs. 0 h.

Full size image

Binding activity of C/EBPβ and HIF1α to the Vegf promoter

The JASPAR database lists consensus binding sequences for several transcription factors in the Vegf promoter, including C/EBPβ (−1115 bp to −1106 bp) and HIF1α (−913 bp to −906, −714 bp to −707 bp, −434 bp to −423 bp) (Fig. 3A). Because C/EBPβ and HIF1α protein levels were increased after hCG injection in rat GCs (Fig. 2), we hypothesized that C/EBPβ and HIF1α bind to the Vegf promoter region and regulate Vegf expression. Therefore, we designed chromatin immunoprecipitation (ChIP) primers surrounding these sites, −1148 bp to −1024 bp (C/EBPβ binding region), −976 bp to −857 bp (HIF1α binding region-1), −724 bp to −645 bp (HIF1α binding region-2) and −470 bp to −369 bp (HIF1α binding region-3), respectively (Fig. 3A). The binding activity of C/EBPβ to the Vegf promoter region was significantly increased in GCs 12h after hCG injection (Fig. 3B). The binding activities of HIF1α were observed in all HIF1α binding regions before luteinization (Fig. 3C, 0 h). However, they were not increased 12 h after hCG injection (Fig. 3C, 12 h). These results indicate that C/EBPβ, but not HIF1α, regulates Vegf gene expression by binding to the promoter region in rat GCs undergoing luteinization.

Figure 3
figure 3

Binding activities of C/EBPβ and HIF1α to the Vegf promoter region in rat GCs undergoing luteinization. (A) Location of the binding sites for transcription factors. Primers for ChIP assay were designed to surround the C/EBPβ binding site (−1148 bp to −1024 bp) and the HIF1α binding sites (−976 bp to −857 bp; region-1, −724 bp to −645 bp; region-2, −470 bp to −369 bp; region-3). Binding activities of C/EBPβ (B) and HIF1α (C) before (0 h) and 12 h after hCG injection were analyzed by ChIP assay. GCs from three rats were pooled at each time point to use for ChIP assay. ChIP assay was performed with antibodies to C/EBPβ, HIF1α and control IgG. The relative binding activities of C/EBPβ and HIF1α were assessed by qPCR. The ratio of IP DNA to INPUT DNA sample was calculated. Data were shown as a ratio of those of 0 h. Each value represents the mean ± SEM of 3 independent experiments. a, P < 0.05 vs. 0 h. PCR products from IP DNA and INPUT DNA were also electrophoresed. The images of representative gels are shown.

Full size image

Effect of C/EBPβ knockdown on VEGF mRNA expression

To elucidate the involvement of C/EBPβ in Vegf mRNA expression, we examined the effect of C/EBPβ knockdown on Vegf mRNA expression. Because the efficiency of knockdown in primary GCs is quite low, we used KGN cells (human granulosa tumor cells) that are widely used for experiments on luteinization22. Because KGN cells do not respond to hCG stimulation, cAMP, a second messenger of hCG, was used to induce luteinization in KGN cells as reported previously22,23. First, we confirmed that VEGF mRNA was increased by cAMP stimulation in KGN cells (Fig. 4A). Knockdown of C/EBPβ by siRNA decreased the protein expression levels of basal (control) and cAMP-increased C/EBPβ (Fig. 4B). The expression of cAMP-increased VEGF mRNA was suppressed by the knockdown of C/EBPβ (Fig. 4C). The basal VEGF mRNA expression in KGN cells without cAMP stimulation was also significantly decreased by the knockdown of C/EBPβ (Fig. 4C). These results suggest that C/EBPβ is involved in not only cAMP-induced Vegf expression but also basal expression of Vegf.

Figure 4
figure 4

Effect of C/EBPβ knockdown on VEFG mRNA expression. (A) KGN cells were treated with or without cAMP (0.5 mM) for 24 h. The relative mRNA levels of VEGF were assessed by qPCR. GAPDH was used as an internal control. Data were shown as a ratio of the control treatment sample. Each value represents the mean ± SEM of 3 independent incubations. a, P < 0.05 vs. control sample. mRNA expression was also assessed by RT-PCR. The images of representative gels are shown. (B) KGN cells were transfected with a siRNA targeted against C/EBPβ or with a nontargeting siRNA as a control. After 24 h of transfection, cells were treated with or without cAMP for 24 h. Whole cell lysates were prepared and protein levels of C/EBPβ were examined by Western blotting to confirm the C/EBPβ knockdown. Protein levels of β−tubulin were also assessed to ensure equal amount of proteins. Western blotting was repeated 3 times, and the representative immunoblots are shown. Quantification of bands were performed by using ImageJ and normalized with β−tubulin levels. Data were shown as a ratio of the control treatment sample. Each value represents the mean ± SEM of 3 independent experiments. a, P < 0.05 vs. control treatment; b, P < 0.05 vs. cAMP treatment in the control siRNA. (C) The relative mRNA levels of VEGF were assessed by qPCR. GAPDH was used as an internal control. Data were shown as a ratio of the control treatment in the control siRNA. Each value represents the mean ± SEM of 3 independent incubations. a, P < 0.05 vs. control treatment in the control siRNA. b, P < 0.05 vs. cAMP treatment in the control siRNA. mRNA expression was also analyzed by RT-PCR. All PCR products were electrophoresed on the same gel and the ethidium bromide-stained gel is a representative of 3 independent incubations.

Full size image

Transcriptional activity

The transcriptional activity of the C/EBPβ binding site in the rat Vegf promoter region was measured by a luciferase assay using KGN cells transfected with different rat Vegf promoter constructs (Fig. 5A). When cells were transfected with the promoter construct (−1171 bp to +115 bp), cAMP significantly increased the luciferase activities. Deletion of the C/EBPβ binding site completely blocked it (the construct of −976 bp to +115 bp) and the cAMP-induced luciferase activities were also decreased to the basal level (without cAMP stimulation). Mutation of the HIF1α binding sites (the mutated construct of −1171 bp to +115 bp) did not affect either cAMP-increased or basal luciferase activities. We further examined the luciferase activities of the rat Vegf promoter region under the condition of C/EBPβ knockdown. Both the construct of −1171 bp to +115 bp and siRNA were transfected to KGN cells and then cells were treated with cAMP. As shown in Fig. 5B, the increase in luciferase activities by the stimulation of cAMP was completely blocked by the knockdown of C/EBPβ. We also examined the effect of CoCl2, which induces HIF1α protein expression24, on the luciferase activities. CoCl2 increased the luciferase activities of the Vegf promoter region, and mutation of HIF1α binding sites blocked it (Supplementary Figure 1). These findings show that HIF1α binding sites in the rat Vegf promoter region have the transcriptional activities under CoCl2 stimulation, but not under cAMP stimulation.

Figure 5
figure 5

Transcriptional activities of the C/EBPβ binding site in rat Vegf promoter region. (A) The reporter constructs, −1171 bp to +115 bp, −976 bp to +115 bp (ΔC/EBPβ) and −1171 bp to +115 bp (mutation of HREs) were transfected into KGN cells. After 24 h of transfection, cells were treated with and without cAMP for 24 h. The firefly luciferase activity was normalized according to Renilla luciferase activities. Values of the luciferase activities were expressed as a ratio of control treatment with −1171 bp to +115 bp. Each value represents the mean ± SEM of 3 independent incubations. a, P < 0.05 vs. control treatment of the construct of −1171 bp to +115 bp. b, P < 0.05 vs. control treatment of the construct of −1171 bp to +115 bp. c, P < 0.05 vs. cAMP treatment of the construct of −1171 bp to +115 bp. d, P < 0.05 vs. control treatment of the construct of −1171 bp to +115 bp (mutation of HREs). (B) The reporter construct of −1171 bp to +115 bp and siRNA (C/EBPβ siRNA or non-targeting siRNA) were transfected into KGN cells. After 24 h of transfection, cells were treated with and without cAMP for 24 h. The firefly luciferase activity was normalized according to Renilla luciferase activities. Values of the luciferase activities were expressed as a ratio of control treatment with −1171 bp to +115 bp. Each value represents the mean ± SEM of 3 independent incubations. a, P < 0.05 vs. control treatment. b, P < 0.05 vs. cAMP treatment in the control siRNA.

Full size image

Histone modifications, chromatin structure, and binding activities of enhancer of zeste homolog 2 (EZH2)

Because not only transcription factors but also epigenetic mechanisms regulate gene expression14,15,16, we examined in rat GCs whether hCG stimulation changes histone modification levels of the C/EBPβ binding region in the Vegf promoter. The levels of H3K9me3 and H3K27me3, which are correlated with the suppression of transcription, significantly decreased at 12 h after hCG injection. The level of H3K4me3, which is correlated with the activation of transcription, was not affected by hCG injection (Fig. 6A).

Figure 6
figure 6

Histone modifications, chromatin structure and binding activity of EZH2 to the C/EBPβ binding region in the Vegf promoter in rat GCs undergoing luteinization. GCs were obtained from rats treated with eCG before (0 h) and 12 h after hCG injection. GCs from three rats were pooled at each time point to use for ChIP assay and FAIRE-qPCR. (A) The levels of H3K4me3, H3K9me3 and H3K27me3 of the C/EBPβ binding region were analyzed by ChIP assay. ChIP assay was performed with antibodies to H3K4me3, H3K9me3, H3K27me and control IgG. The relative levels of H3K4me3, H3K9me3 and H3K27me3 were analyzed by real-time PCR. The ratio of IP DNA to INPUT DNA sample was calculated. Data were shown as a ratio of those of 0 h. Each value represents the mean ± SEM of 3 independent experiments. a, P < 0.05 vs. 0 h. All PCR products from IP DNA or INPUT DNA were electrophoresed on the same gel and the representative ethidium bromide-stained gels are shown at the right. (B) To determine the changes of the chromatin structure in the C/EBPβ binding region, a FAIRE-qPCR was performed with same primers used in ChIP assay. The ratio of FAIRE enrichment in the Vegf promoter region was calculated. Data were shown as a ratio of those of 0 h. Each value represents the mean ± SEM of 4 independent experiments. a, P < 0.05 vs. 0 h. (C) Binding activity of EZH2 to the C/EBPβ binding region was analyzed by ChIP assay. ChIP assay was performed with antibodies to EZH2 and control IgG. The relative levels of EZH2 binding activities were analyzed by real-time PCR. The ratio of IP DNA to INPUT DNA sample was calculated. Data were shown as a ratio of those of 0 h. Each value represents the mean ± SEM of 3 independent experiments. a, P < 0.05 vs. 0 h. PCR products from IP DNA and INPUT DNA were also electrophoresed. The images of representative gels are shown.

Full size image

The chromatin remodeling of the C/EBPβ binding region after hCG injection were examined by FAIRE (formaldehyde-assisted isolation of regulatory elements)-qPCR. The relative FAIRE enrichment ratio in this region increased by hCG injection (Fig. 6B), indicating that hCG stimulation loosens the chromatin structure of the C/EBPβ binding region.

We examined whether hCG stimulation changes the binding of EZH2, which induces H3K27me325,26, to the C/EBPβ binding region. The EZH2 binding activity to the C/EBPβ binding region, as measured by ChIP assay, significantly decreased after hCG injection (Fig. 6C).

Discussion

The present study demonstrated that C/EBPβ is an important transcription factor regulating Vegf expression in rat GCs undergoing luteinization during ovulation. C/EBPβ is well known to be transcription factor necessary for ovulation12, and belongs to the ERK-1/2 signaling pathway, which activates in GCs undergoing luteinization. We reported that C/EBPβ up-regulates the expression of StAR and Cyp11a1 and promotes the production of progesterone, which play central roles in follicle rupture and subsequent CL formation13,20. Our results show a new finding that Vegf is one of downstream target genes of C/EBPβ in rat GCs after the LH surge, suggesting that C/EBPβ contributes to angiogenesis for CL formation by up-regulating Vegf expression.

This is the first report to show the involvement of C/EBPβ in the regulation of Vegf gene expression in rat GCs. The promoter region of the rat Vegf gene has the consensus binding sequence of C/EBPβ, which is highly conserved among species27. However, this site has not been examined so far. C/EBPβ clearly binds to this site in rat GCs undergoing luteinization (Fig. 3B). In addition, by knockingdown C/EBPβ protein in KGN cells, we demonstrated that C/EBPβ regulates the basal and cAMP-increased Vegf mRNA expressions (Fig. 4C). Furthermore, the C/EBPβ binding site has transcriptional activity (Fig. 5). Taken together, these findings show that C/EBPβ up-regulates Vegf gene expression through the binding to the novel regulatory region in rat Vegf promoter region in GCs undergoing luteinization.

HIF1α has been reported to regulate Vegf gene expression by binding to the promoter region in many types of cells28,29. HIF1α protein is expressed in granulosa cells before luteinization30,31 and is up-regulated by hCG stimulation10,32,33,34, which is consistent with our findings (Fig. 2B). Because both VEGF and HIF1α expression show the same trend toward luteinization, HIF1α was thought to be involved in the increase of Vegf expression by luteinization. However, the direct evidences showing that the increase in Vegf expression by luteinization is mediated by HIF1α have not been reported so far. We proved by ChIP assay that the binding activities of HIF1α were observed in all HIF1α binding regions in rat Vegf promoter region before luteinization (Fig. 3C, 0 h), but these binding activities were not increased after hCG injection (Fig. 3C, 12 h). This result suggests that HIF1α does not contribute to the up-regulation of Vegf expression by hCG stimulation. In addition, we showed by luciferase assay that hypoxia responsible elements (HREs) in rat Vegf promoter region are not involved in the regulation of Vegf expression by cAMP. In this study, we did not show the involvement of HIF1α in the regulation of Vegf in our experimental condition including the use of KGN cells. However, we do not exclude the role of HIF1α in Vegf induction during luteinization, because HIF1α may regulate Vegf expression at different time points of luteinization process, or bind to an unknown promoter or enhancer region of Vegf gene undergoing luteinization in GCs. On the other hand, Rico et al.35 showed that HRE deficient mice responded to LH with an increase in Vegf expression in granulosa cells in vivo, suggesting that Vegf expression by LH is not regulated by HIF1α in mouse granulosa cells, which supports our findings. Kim et al.10 used Echinomycin, which inhibits the binding ability of HIF1α to HREs, and found that Echinomycin inhibits hCG-increased Vegf mRNA expression in mice granulosa cells. Although they concluded that hCG up-regulates Vegf expression through HIF1α, it should be noted that Echinomycin directly affects many gene expressions. Taken together, we concluded that C/EBPβ is an important factor to regulate Vegf gene expression in rat GCs undergoing luteinization after the LH surge.

Although there are several reports showing that HIF1α is associated with Vegf mRNA expression in GCs, most of them are based on the data from in vitro experiments done under the conditions of hypoxia or CoCl2 stimulation. Martinez-Chequer et al.36 reported that CoCl2 increased VEGF expression in monkey GCs. Alam et al.37 reported that CoCl2 induced HIF1α protein expression in rat GCs. Kim et al.10 showed that HIF1α is involved in Vegf expression induced by CoCl2 in mouse GCs. Yalu et al.24 showed the involvement of HIF1α in the CoCl2-induced VEGF expression by knockingdown HIF1α in human GCs. Our results also showed that HIF1α contributes to Vegf expression under CoCl2 stimulation, but not under cAMP stimulation (Supplementary Figure 1). From these results, it is likely that HIF1α is involved in the regulation of Vegf expression in GCs under CoCl2 stimulation in vitro.

Our data also showed that epigenetic changes such as histone modifications and chromatin remodeling occurred in the Vegf promoter region in luteinizing rat GCs. These changes included decreases of H3K9me3 and H3K27me3 levels along with a decrease of EZH2 binding to the C/EBPβ binding region in the Vegf promoter all of which are associated with the activation of transcription by loosening the chromatin38. This result is consistent with our previous findings that stimulating rats with hCG caused decreases in H3K9me3, H3K27me3 and EZH2 binding of the Cyp11a1 promoter region in GCs and increases in their mRNA expressions20. Our results showed that hCG loosened the chromatin structure of the Vegf promoter region, allowing C/EBPβ to access its response element of the Vegf promoter region. These findings suggest that these epigenetic changes are closely associated with the up-regulation of Vegf expression in GCs undergoing luteinization. H3K4me3 is a histone modification related to active transcription39,40. In our study, the levels of H3K4me3 were not changed by hCG stimulation. This is not surprising, because the changes of H3K4me3 levels are not always correlated with those of other histone modifications18, and because gene expression is regulated by not only one histone modification, but also the combination of various histone modifications13.

In summary, the present study shows a molecular mechanism by which Vegf is up-regulated in rat GCs undergoing luteinization after the LH surge. We found that C/EBPβ, but not HIF1α, regulates Vegf gene expression by binding to the novel binding site in the rat Vegf promoter region. In addition to transcription factors, histone modifications and chromatin structure of the Vegf promoter region are involved in the regulation of Vegf expression. Because Vegf plays a key role in angiogenesis, our results should help to better understand the regulation of angiogenesis in GCs undergoing luteinization during ovulation.

Methods

This study was reviewed and approved by The Committee for the Ethics on Animal Experiment in Yamaguchi University Graduate School of Medicine. All experiments were performed in accordance with relevant guidelines and regulations.

Animal models

Female Sprague Dawley rats (21- to 24-day old) were purchased from Japan SLC (Hamamatsu, Japan). They were injected subcutaneously with 15 IU of equine chorionic gonadotropin (eCG) (Sigma, St. Louis, MO, USA) to promote follicular growth followed by 15 IU of hCG (Sigma) injection to induce ovulation and luteinization. The ovaries were obtained before hCG (0), and 4, 8, 12, and 24 hours (h) after hCG injection. The follicles were punctured to isolate GCs. Cells were centrifuged and washed in PBS and provided for the experiments.

Cell culture

KGN cells (Nikon, Tokyo, Japan), which were derived from a human granulosa cell tumor, were cultured in DMEM/Ham’s F-12 medium.

Quantitative real-time PCR

RNA was extracted with RNeasy Mini (QIAGEN, Chatsworth, CA, USA) and was converted to cDNA with reverse transcriptase (Invitrogen, Waltham, MA, USA) as reported previously41. Quantitative real time PCR reaction was carried out as described previously42,43. Primer sequences are listed in Table 1.

Table 1 Primers Used in This Study.
Full size table

Western blotting

Total protein were extracted and subjected to SDS-PAGE and then transferred to membrane as described previously18,44. The membranes were incubated for overnight with the following antibodies: C/EBPβ (Santa Cruz Biotechnology), HIF1α (Novus Biologicals, USA) and Histone H3 (Cell Signaling Technology, Tokyo, Japan). Amersham ECL Western Blotting Detection Reagent and hyperfilm-ECL (GE health care UK Ltd., Buckinghamshire, UK) were used to detect immunoblot bands.

ChIP assay

The histone modification levels and recruitments of transcription factors were examined by ChIP assays as described previously45 with some modifications. The following antibodies were used in this assay: H3K4me3 (Upstate Biotechnology, Lake placid, NY, USA), H3K9me3 (Abcam, Cambridge, UK, USA), H3K27me3 (generous gift from Dr. Kimura), EZH2 (Cell Signaling Technology), C/EBPβ, HIF1α and non-specific normal IgG (Cell Signaling Technology). Real-time qPCR was performed with primers listed in Table 1 to calculate the relative enrichment.

Transfection of siRNA

Transfection of siRNAs was performed as reported previously46. After 24 h of transfection, KGN cells were incubated with or without dibutyryl-cAMP (cAMP, 0.5 mM) (Sigma) for 24 h.

Luciferase assay

Three 5′-flanking regions of the rat Vegf gene; 1) the construct including both C/EBPβ binding site and HIF1α binding sites (−1171 bp to +115 bp), 2) the construct lacking the C/EBPβ binding site (−976 bp to +115 bp) and 3) the construct that four HIF1α binding sites are mutated (−1171 bp to +115 bp), were amplified using the primer sets in Table 1. These constructs were inserted to pGL3 basic vector (Promega, Tokyo, Japan). Transfection of the constructs were performed as described20 and then cells were incubated with or without cAMP (0.5 mM) or CoCl2 (0.5 mM) for 24 h. Luciferase activities were examined as described20. Luciferase assays combined with C/EBPβ knockdown were performed as reported previously23. The reporter construct (the construct of −1171 bp to 115 bp) and siRNA (C/EBPβ siRNA or non-targeting siRNA) were transfected simultaneously. Then, KGN cells were incubated with or without cAMP and luciferase activities were examined. Luciferase assays were done in triplicate and repeated three times.

FAIRE-qPCR

To examine the chromatin accessibility of C/EBPβ binding region in the Vegf promoter, FAIRE-qPCR was performed as described47,48 with some modifications. Real-time qPCR was performed to determine and calculate the FAIRE enrichment with same primers used in ChIP assay (Table 1).

Statistical analyses

Unpaired student t-test was used to compare the mean values in two groups. To analyze differences between groups, one-way ANOVA followed by Tukey-Kramer test was used. P < 0.05 was considered significant.