Relaxin deficiency results in increased expression of angiogenesis- and remodelling-related genes in the uterus of early pregnant mice but does not affect endometrial angiogenesis prior to implantation

All animal experiments were approved by The University of Melbourne Animal Experimental Ethics Committee (AEEC #0911478.1) and conducted in accordance with the Australian Code of Practice and the National Health and Medical Research Council. This study used wild type (Rln ^+/+ ) and Rln ^-/- mice. Homologous recombination in embryonic stem cells was used to disrupt the relaxin gene by deleting a region essential for biological activity and replacing it with the neomycin transferase gene [25]. Rln ^-/- mice were then backcrossed to a C57/BLK6J background and the F14 generation relocated to the School of BioSciences. Genotypes were confirmed by PCR analysis of genomic DNA from ear clips as previously described [25]. Mice were maintained on an automated time cycle of 12 h light/dark at 20 °C, with laboratory chow (Barastock, Pakenham,VIC, Australia) and water available ad libitum.

Uterine tissues from non-pregnant transgenic mice with a modified Rxfp1 gene (n = 3) were kindly provided by Prof AI Agoulnik (Florida International University, Miami, FL, USA). The modification is a 4 kb fragment containing parts of exons 15 and 17 that was ligated into a SalI site of the pNTR-LacZ/PGKneo/loxP vector as described by Kamat et al [26]. During homologous integration, the targeting vector was inserted into exon 17 and created a duplication of the 4 kb genomic fragment. Uterine tissues were dissected from Rxfp1 ^TG mice after anesthetic overdose and were fixed in 2 % paraformaldehyde in 1X phosphate buffered saline/MgCl₂ for 30 mins at RT. Tissues were then embedded in Tissue-Tek OCT compound (Sakura Finetechnical Co. Ltd), and cryosections were cut at 10 μm and mounted on Superfrost slides. For histological analysis of LacZ expression, tissues sections were processed for X-gal staining using the LacZ Detection Kit for Tissues (InvivoGen) according to the manufacturer’s instructions. Nuclei were then counterstained with 0.1 % nuclear fast red and mounted in aqueous mounting medium. Control tissues were obtained from wild type littermates.

HES cells were isolated from a collagenase-treated uterus of a pre-menopausal woman who had a hysterectomy for reasons other than uterine disease, with informed consent and approval from the institutional review board [15, 27]. Stromal cells obtained from a normal uterus were cultured in 75 mm² flasks, and fed every two days with DMEM:F12 medium plus 10 % newborn calf serum and 2 mM L-glutamate. These cells have previously been shown to possess high affinity relaxin binding sites [15] and to produce cAMP in response to relaxin [27]. Once confluent, the cells were washed with phosphate buffered saline (PBS) and treated in serum-free DMEM:F12 medium with either vehicle alone or with recombinant human relaxin (rhRLX) (Connetics Corporation, Palo Alto, CA, USA) (10 ng/ml) for 24 h. This concentration and duration was shown to be bioactive in a previous study [15]. Total RNA was isolated using Trizol (Invitrogen) according to manufacturer’s instructions. Total RNA from 8 HES cell cultures treated with vehicle or with rhRLX were pooled to yield 100 μg RNA each. This was sent to Incyte Genomics (Palo Alto) for extraction of PolyA-RNA and subsequent microarray analysis [28]. Briefly, 600 ng of polyA-RNA of each sample was used for reverse transcription and attachment of the appropriate cyanine (Cy) dye. The Cy3- and Cy5-labelled probes were then mixed, and simultaneously hybridized to a single cDNA microarray. Following incubation, the array was washed in 3 consecutive washes of decreasing ionic strength, and scanned in both Cy3 and Cy5 channels using GenePix scanners (Foster City). Incyte GEMtools software was used for image analysis. Background-subtracted element signals are used to calculate Cy3:Cy5 ratios. Default criteria in GEMtools analysis are that the signal-to- background ratio > 2.5 for each element, and that the area for which pixel values are counted is at least 40 % of the element for both signals. The ratio of fluorescent intensities from the two dyes were used to assess relative transcript abundance in the two RNA samples. The minimum detectable fold change for differential expression, according to evaluation of the performance of the Incyte array, is 1.4-fold [29], i.e. expression in mRNA from rhRLX-treated cells 1.4-fold higher than in vehicle-treated cells, or vice versa.

In Study 1, Rln ^+/+ mice were studied on days 1, 2, 3 and 4 of pregnancy to assess changes in gene expression in the wild type mice (n = 8/genotype/day). Based on data from Study 1, we then chose to examine pre-implantation phenotypes in Rln ^+/+ and Rln ^-/- mice only on days 1 and 4 of pregnancy (Study 2, n = 8/genotype/day). In Study 3, gene expression alone was examined in an additional group of mice on day 6 of pregnancy (n = 7/genotype) to examine if phenotypes persisted after implantation (Additional file 1: Figure S1). Female mice aged 2–4 months were housed with a stud male of matched genotype overnight and then checked for mating plugs the following morning. The presence of a mating plug was an indication of successful mating and was considered day 1 of pregnancy. Four hours prior to tissue collection, mice received a 500 μl i.p. injection of bromodeoxyuridine (BrdU; 2 mg/ml, Sigma-Aldrich Pty Ltd), enabling later visualization of proliferating cells by immunohistochemistry. After isofluorane anesthesia, blood was collected by cardiac puncture and the plasma stored at -80 °C. After cervical dislocation, the uterine tissues were removed with half fixed in 10 % neutral buffered formalin overnight before processing for paraffin sections, and the other half stored at -80 °C for gene expression analysis.

Total RNA was isolated from whole uteri using Trizol reagent (Invitrogen) according to the manufacturer’s instructions. The concentration of RNA was determined using a NanoDrop ND100 Spectrophotometer (NanoDrop Technologies Inc) with A260:A280 absorbance ratios of > 2.0 indicating sufficiently pure RNA for PCR analysis. First strand cDNA synthesis used 1 μg total RNA in a 20 μL reaction, using random hexamers (50 ng/μL), 10 mM dNTP and 200 U SuperScript III (Invitrogen). First-strand cDNA synthesis for all samples were performed simultaneously at 25 °C for 10 min, 50 °C for 50 min and 85 °C for 5 min. Quantitative reverse-transcription polymerase chain reaction (quantitative PCR) was performed to assess the relative expression of relaxin (Rln) and its receptor Rxfp1 and key genes chosen to represent angiogenesis (total VEGFA (VegfA), VEGF receptor 2 (Vegfr2)), steroid receptors (estrogen receptor (ER)-α (Esr1), ER-β (Esr2), progesterone receptor (Pgr)), tissue remodeling (matrix metalloproteinase 14 (Mmp14), Timp3)), and other key genes identified in our microarray analysis (ankryn repeat domain 37 (Ankrd37), egl-9 family hypoxia-inducible factor 1 (Egln1), hepatocyte growth factor (Hgf), hypoxia inducible factor 1α (Hif1α)). In addition to these genes investigated pre-implantation, interleukin 1β (Il1β) was also investigated post-implantation (Additional file 2: Figure S2). Forward and reverse primers and 6-carboxy fluorescein (FAM)-labeled TaqMan probes specific for mouse genes were designed to span an intron/exon junction (Biosearch Technologies, Inc). Primer sequences are shown in Additional file 3: Table S1. All quantitative PCR reactions for each gene were performed on one plate, in 10 μl or 20 μl 96-well reactions containing SensiFAST Probe Lo-Rox (BioRad) and performed in the ViiA™7 Real-Time PCR System (Life Technologies). Recent data suggest that glyceraldehyde 3-phosphate dehydrogenase (Gapdh) and 18 s ribosomal RNA (18 s) expression are upregulated in the rodent uterus in response to progesterone and estrogen [30, 31]. The same pattern of 18 s mRNA expression was observed in this study so we investigated peptidylprolyl isomerase A (Ppia), succinate dehydrogenase complex, subunit A (Sdha) and TATA box binding protein (Tbp) as reference genes; however, these were also significantly increased from days 1 to 4 of pregnancy (Additional file 2: Figure S2). Therefore we used the mean C_T value of the Rln ^+/+ mice on day 1 of pregnancy as the internal calibrator and subtracted this from the mean gene of interest triplicate C_T value (Ratio_{(test/calibrator)} = 2^ΔCT, where ΔC_T = C_{T(calibrator)} − C_T(test)) as has been described previously [30, 32]. These “normalized” data (ΔC_T) were presented as fold induction relative to day 1 of pregnancy in the Rln ^+/+ mice. The mean C_T values and range C_T values for each gene are presented in Additional file 3: Table S2.

The presence or absence of relaxin (Rln) and Rxfp1 gene transcripts in the uterus and ovaries were assessed by reverse transcriptase polymerase chain reaction (RT-PCR) in the GeneAmp PCR system 2700 (Applied Biosystems). Reaction mixes consisted of GoTaq Green (Promega), mouse-specific oligonucleotide primers (100 ng/μl) designed to span an intron/exon junction (Additional file 3: Table S3) and 1 μl cDNA. Ovary RNA was extracted as described above.

As the study focused on pre-implantation vascular remodeling, we used immunohistochemistry to examine endometrial endothelial cell proliferation as a marker of endometrial angiogenesis. We used a dual immunohistochemistry protocol with antibodies against CD31 (pan-endothelial cell marker) and BrdU (labels proliferating cells). After dewaxing and rehydration, sections (5 μm) were incubated with 0.1 % (1 mg/ml) pepsin in 3 % acetic acid for 10 min at 37 °C, followed by serum free protein block (Dako) for 15 min to prevent non-specific binding. Sections were incubated with rat monoclonal anti-mouse CD31 (5 mg/ml in 1 % bovine serum albumin in PBS (BSA/PBS) (BD PharMingen)) overnight at 4 °C. Sections were then incubated for 1 h at room temperature in biotinylated goat anti-rat IgG (1:200 in 1 % BSA/PBS, Chemicon), followed by incubation in alkaline phosphatase conjugated streptavidin (Dako) for 15 min. CD31-positive immunostaining was visualized using the vector blue alkaline phosphatase chromagen (10 min, Vector Laboratories). Sections were then incubated in 0.1 M HCl for 45 min at room temperature, followed by a further incubation 3 % H₂O₂ in PBS for 10 min at room temperature to quench endogenous peroxidase. Serum free protein block was applied for a further 15 min. Sections were incubated for 1 h at room temperature in monoclonal sheep anti-BrdU (8 μg/ml in 1 % BSA/PBS, Abcam). This was followed by a further incubation for 1 h at room temperature in donkey anti-sheep IgG (4 μg/mL in 1 % BSA/PBS, Jackson Immuno Research Laboratories). Sections were then incubated in streptavidin horseradish peroxidase (Dako) for 15 min, followed by 5 min in DAB chromagen (3,30-diaminobenzidine, Sigma-Aldrich) to visualize BrdU immunostaining. Each slide contained a negative isotope matched control section that was prepared by replacing the CD31 primary antibody with rat IgG2a (BD Biosciences) and the BrdU primary antibody with sheep IgG (Sigma-Aldrich) at the same concentrations as that of the primary antibodies. For the main analysis, we analyzed one section per uterus (n = 7/group) taken from the mid-point between the cervix and ovaries. The number of proliferative endothelial cells and the number of blood vessel profiles present in the endometrial tissue on each section were counted. Data are presented as the proportion of vessel profiles containing one or more proliferating endothelial cells, and as the number of proliferating endothelial cells per mm² (identified using a × 40 objective lens).

Plasma progesterone and corticosterone concentrations were measured by triple quadrupole mass spectrometry (Eric Thorstensen, Liggins Institute, University of Auckland, Auckland, New Zealand). Estrogen could not be measured as there was insufficient plasma for accurate detection. An internal standard solution (60 ng mL^-1 corticosterone-d8 and 0.2 ng mL^-1 estradiol-d3 in water) was added to each plasma sample (100 μL standard to 200 μL plasma). Steroids were then extracted using 1 mL of ethyl acetate (Merck). After removal of the organic supernatant to a clean tube, samples were dried by vacuum concentration (Savant SC250EXP, Thermo Scientific), resuspended in 60 μL of mobile phase (45 % methanol and 55 % water) and transferred to HPLC injector vials. Samples (12 μL) were injected onto an HPLC mass spectrometer system consisting of an Accela MS pump and autosampler followed by an Ion Max APCI source on a Finnigan TSQ Quantum Ultra AM triple quadrupole mass spectrometer all controlled by Finnigan Xcalibur software (Thermo Electron Corporation). The mobile phase was a gradient of methanol and water, flowing at 400 μL.min^-1 through a Hypersil Gold 1.9 μm C18 50 ×2.0 mm column at 40 °C (Thermo Fisher Scientific). Retention time was 6.2 min for progesterone and 2.3 min for corticosterone. Ionization was in positive mode for progesterone, Q2 had 1.2 mTorr of argon. The mass transitions followed were: progesterone 315.5 → 109.2 at 22 V and corticosterone 347.2 → 121.0 at 27 V.

Proliferating endothelial cell number and circulating steroid hormone concentrations were assessed by two-way ANOVA (genotype v day of pregnancy) and Tukey post-hoc tests (SPSS Inc.). Changes in gene expression during early pregnancy in the Rln ^+/+ mice were assessed by one-way ANOVA (day of pregnancy) and Tukey post-hoc tests. Two-way ANOVA with genotype and day of pregnancy as main effects assessed any interaction between genotype and day of pregnancy in the Rln ^+/+ and Rln ^-/- mice followed by Tukey post-hoc tests to compare means of all groups. T-tests were used to examine gene expression differences between Rln ^+/+ and Rln ^-/- mice on day 6 of pregnancy. If not normally distributed, raw data were log transformed and non-parametric analysis was performed using Mann-Whitney with Bonferroni correction. All figures are graphed using the untransformed fold change data.

Reverse transcriptase PCR confirmed the presence of two gene transcripts for Rxfp1 representing both isoforms in the uterus of both Rln ^+/+ and Rln ^-/- mice in early pregnancy (Fig. 1a) (raw C_T values in Additional file 3: Table S2). There was a significant (p = 0.006) increase in Rxfp1 mRNA in Rln ^+/+ mice on days 3 and 4 of pregnancy compared with day 1 (Fig. 1b). A comparison between genotypes revealed a significant (p = 0.003) increase in Rxfp1 expression on day 1 in Rln ^-/- mice compared to day 1 in Rln ^+/+ mice but expression on day 4 of pregnancy was similar between Rln ^+/+ and Rln ^-/- mice (Fig. 1b). Rln mRNA was detected at all stages in the ovaries of the wild type mice (Fig. 1c). Rln mRNA was also detected in whole uterus, with significantly increased expression on days 3 and 4 of pregnancy relative to day 1 (p < 0.05) (Fig. 1c and d).

Strong positive staining for Rxfp1-specific β-galactosidase activity was identified in the myometrium of Rxfp1 ^TG mice, particularly in the inner circular muscle layer (Fig. 2). Upon closer examination of the endometrium, positive staining for Rxfp1-specific β-galactosidase activity was seen in the stroma towards the luminal epithelium. The negative control used in this study was the uterus of a wild type mouse, which showed no endogenous or background β-galactosidase activity (Fig. 2). This confirmed that the β-galactosidase activity observed in the Rxfp1 ^TG mouse uterus was specific and represents localization of Rxfp1.

A total of 321 genes were upregulated (by 1.4-fold or more) in HES cells incubated with rhRLX for 24 h, whereas 203 genes were down-regulated by at least 30 %. 64 well characterized genes increased with relaxin treatment (Table 1), of which glycerol kinase showed the largest increase (2-fold). A number of genes whose expression is associated with regulation of angiogenesis were among those modulated by rhRLX. Timp3 (1.9-fold), Hgf, (1.5-fold) Hif1α (1.5-fold) and angiogenin (1.4-fold) were increased. The expression of VEGF was analyzed as 1.2-fold higher in relaxin-treated cells, within the noise of the assay. In contrast, VEGF-D and CD36 were decreased by 30 %. CD36 thought to be an inhibitor of angiogenesis [33] was down-regulated by 30 % after rhRLX treatment.

Consistent with prior studies [34], there was a modest increase in VegfA on day 4 of pregnancy compared with day 1 in the wild type mice in study 2 (p = 0.012) (Fig. 3a). A comparison between genotypes demonstrated a significant (p = 0.026) increase in VegfA on day 1 of pregnancy in the Rln ^-/- mice compared with the same time point in Rln ^+/+ mice. Similarly, there was a significant (Day 3: p = 0.006; Day 4: p = 0.002) increase in Vegfr2 on days 3 and 4 of pregnancy in Rln ^+/+ mice compared with day 1 but there was no effect of genotype (Fig. 3b).

Because relaxin is known to modulate steroid receptor expression [9], Esr1 and Esr2 transcripts were examined (Pgr was also examined, see below). Esr1 but not Esr2 was significantly (Day 2: p = 0.01; Day 3: p = 0.002; Day 4: p = 0.004) increased in Rln ^+/+ mice on days 2, 3 and 4 of pregnancy. Only Esr1 expression was significantly (pregnancy: p ≤ 0.001; genotype: p = 0.002) affected by both pregnancy and genotype (Fig. 3c and d).

Mmp14 and Timp3 were also significantly (Mmp14: Day 3: p = 0.012; Day 4: p = 0.003, Timp3: Day 3: p = 0.019; Day 4: p = 0.004) upregulated on days 3 and 4 compared with day 1 in pregnant Rln ^+/+ mice (Fig. 4). Interestingly, Mmp14 was significantly (p = 0.001) higher in Rln ^-/- mice on day 1 of pregnancy compared with the Rln ^+/+ counterpart, whereas there was no significant effect of genotype on Timp3 expression. However Timp3 was significantly (pregnancy: p ≤ 0.001) higher on day 4 of pregnancy compared with day 1 in both genotypes in study 2.

Based on the data from the microarray study of HES cells (Table 1), we analyzed four other genes that were altered by relaxin treatment. In wild type mice, there was no significant increase in Ankrd37 from days 1 to day 4 of pregnancy (Fig. 5a). A comparison of genotypes demonstrated a significant (p = 0.011) increase in Ankrd37 on day 1 in the Rln ^-/- mice compared with Rln ^+/+ . Egln1 was significantly (p = 0.015) upregulated on day 4 of pregnancy in Rln ^+/+ mice compared to days 1 and 2 (Fig. 5b). There was also a significant (p ≤ 0.001) difference between the genotypes on day 1. Both Hgf and Hif1α were significantly (p < 0.05) increased on days 3 and 4 in the wild type mice compared to day 1 of pregnancy (Fig. 5c and d). Hif1α expression was significantly (p = 0.008) upregulated in the Rln ^-/- mice on day 1 compared with Rln ^+/+ but there was no difference in Hgf between the genotypes.

We also examined expression of the genes noted above in uteri from day 6 pregnant mice, but no significant differences were observed between Rln ^+/+ and Rln ^-/- mice (Additional file 1: Figure S1).

Endothelial cell proliferation occurred in the endometrium on day 4 but not day 1 of pregnancy in wild type mice, consistent with a previous report [6]. Proliferation in the Rln ^-/- mice also occurred on day 4 (Fig. 6). However, there was no significant difference in the percentage of vessel profiles containing proliferating endothelial cells or the number of proliferating endothelial cells in the endometrium per mm² between the genotypes. Additionally, there was no difference in the number of vessels per mm² (data not shown).

As seen in previous studies [6], there was a significant (p ≤ 0.001) effect of day of pregnancy on circulating progesterone levels in the wild type and relaxin deficient mice from day 1 to day 4 of pregnancy (Fig. 7a). Plasma progesterone concentrations were significantly (p = 0.013) different between the two genotypes on day 4, with an increase in progesterone in the Rln ^-/- mice relative to the Rln ^+/+ mice. There was a significant (Day 1: p = 0.017; Day 4: p = 0.005) increase in Pgr in Rln ^-/- mice compared to Rln ^+/+ mice on days 1 and 4 of pregnancy (Fig. 7b). Circulating corticosterone levels were significantly (p = 0.003) higher on day 1 compared with day 4 of pregnancy in both Rln ^+/+ and Rln ^-/- mice (Fig. 7c) but there was no difference between the genotypes.