Control of mammary myoepithelial cell contractile function by α3β1 integrin signalling

α3β1 integrin deletion from myoepithelial cells does not interfere with functional differentiation of the mammary epithelium

To delete the gene coding for α3 integrin chain, mice harbouring floxed itga3 alleles (itga3^F/F) were crossed with transgenic mice expressing Cre recombinase under control of the keratin 5-promoter (K5Cre) active in the basal epithelial cells (Ramirez et al, 2004). As expected, Cre-driven recombination was detected in the epidermis and mammary gland (Figure 1A). Ductal branching morphogenesis was not affected in mutant mice, although itga3 was found to be deleted from the mammary basal epithelial layer of 5-week-old animals (Supplementary Figure S1A–C). Immunoblotting and immunofluorescence analyses confirmed efficient deletion of the α3 integrin chain from the mammary epithelium of 1 day-lactating mutant mouse mammary glands (Figure 1B and C).

Overall, mutant mammary tissue organization was similar to that of control animals (Figure 1C and D). The expression and distribution of various integrin subunits (α6, β1 and β4) and ECM components (processed and unprocessed γ2 laminin chain) were similar in control and mutant glands (Supplementary Figure S2A; data not shown). Ultrastructural analysis of the mutant mammary epithelium revealed no defect in basal myoepithelial cell adhesion to the basement membrane (Supplementary Figure S2B). These data suggest that, in the absence of α3β1, other integrin dimers present at the surface of the basal myoepithelial cells ensure adhesion to the ECM.

The functional differentiation of the mammary epithelium appeared normal in mutant glands, as shown by evaluation of the expression levels of milk and SM-contractile protein genes characteristic of luminal and myoepithelial cells, respectively, by quantitative RT–PCR (Figure 1E). Mammary epithelial cell proliferation was not affected in K5Cre;itga3^F/F glands, and hardly any apoptotic cells were detected with anti-cleaved caspase 3 antibody at the onset of lactation in control and mutant mammary epithelium (Supplementary Figure S2C–E; data not shown). Thus, the deletion of α3β1 integrin from myoepithelial cells did not affect the integrity and functional differentiation of the mammary epithelium.

K5Cre;itga3^F/F mice display lactation failure due to low rates of milk ejection

K5Cre;itga3^F/F mutant mice were viable and fertile. However, the females were unable to feed their litters. About half the pups nursed by K5Cre;itga3^F/F females died within 4 days, and the surviving pups displayed severe growth retardation (Figure 1F and G). Milk was visible in the stomachs of the pups fed by control females, whereas the pups nursed by K5Cre;itga3^F/F females appeared to have empty stomachs (Figure 1F). This defect was not intrinsic to the neonates, which developed normally after weaning or when nursed by control females. Control neonates fostered by lactating K5Cre;itga3^F/F females also failed to grow normally (data not shown). The mutant females presented no obvious abnormal maternal nurturing behaviour.

As milk was produced but not available to neonates, we hypothesized that the contractile activity of mutant myoepithelial cells was impaired, leading to a low rate of milk ejection. This hypothesis was supported by the results of morphological analyses of the lactating mouse glands. Most of the mutant glands analysed on days 7 and 18 of lactation contained large dilated ducts engorged with milk. In addition, collapsed alveoli and gland fragments displaying complete secretory tissue regression were found in mutant epithelium examined on day 18 of lactation (Figure 2A; data not shown). These morphological changes in the mutant glands were similar to those resulting from milk stasis, which is known to induce premature involution (Quarrie et al, 1996).

High myosin light-chain phosphorylation level is maintained in lactating K5Cre;itga3^F/F mouse mammary glands

The contraction of myoepithelial cells, like that of SM cells, is regulated by myosin light-chain (MLC) phosphorylation, which induces myosin ATPase activity, the binding of myosin to actin and contraction (Hartshorne et al, 1989). The subsequent dephosphorylation of the MLC by its specific phosphatase (MLCP) leads to relaxation. Immunofluorescence analysis of MLC distribution showed this protein to be present in large amounts in the myoepithelial cells of control and mutant mammary epithelium (Figure 2B, upper panels), whereas immunolabelling with antibodies against phosphorylated MLC was slightly stronger in mutant myoepithelial cells than in control cells (Figure 2B, middle and lower panels). Immunoblotting analysis performed with protein extracts prepared from control and mutant mammary epithelium on the first day of lactation clearly demonstrated that MLC phosphorylation levels were abnormally high in mutant myoepithelial cells (Figure 2C).

Consistently, on the first day of lactation, the mutant myoepithelial cells presented extensions slightly shorter than those of control cells (data not shown) and, by day 5 of lactation, in most regions of the mutant glands, myoepithelial cells displayed a characteristic hypercontractile phenotype, with a compact cell body and short thick extensions (Figure 2D and E). In control tissue, most myoepithelial cells appeared more relaxed, with a significantly larger surface area and longer extensions (Figure 2D and E). Taken together, these observations suggest that the MLC kinase/phosphatase activity balance is altered in mutant glands.

Mammary myoepithelial cell contraction requires activation of RhoA/ROCK pathway

In non-muscle and SM cells, MLC phosphorylation is controlled essentially by MLCK (Ikebe and Hartshorne, 1985; Somlyo and Somlyo, 2003; Webb, 2003). The RhoA effector Rho kinase (ROCK) has also been reported to phosphorylate MLC in vitro, but the main effect of Rho/ROCK activation is the inhibitory phosphorylation of MLCP (Somlyo and Somlyo, 2003).

During lactation, mammary myoepithelial cell contraction is induced by pulsatile release of the cyclic peptide oxytocin (OT), which binds to a promiscuous G protein-coupled receptor (OTR). The downstream signalling pathways leading to myoepithelial cell contraction remain largely unclear (reviewed in Reversi et al (2006)).

To define whether ROCK has a role in myoepithelial cell contraction, fragments of mammary tissue dissected from 1 day-lactating mammary glands were treated with ROCK inhibitor and then incubated with OT. The changes in myoepithelial cell shape following OT treatment were analysed after labelling with anti-K5 antibody and phalloidin, by immunofluorescence microscopy. As previously reported, OT treatment induced major cell shape changes (Moore et al, 1987). Relaxed myoepithelial cells had thin elongated processes, whereas OT-treated contracted cells had a compact spoke-shaped body with short, thick extensions (Figure 3A and B). Pre-treatment with ROCK inhibitor (Y27632) abolished the myoepithelial cell response to OT (Figure 3C). Treatment with the inhibitor only did not lead to any visible cell shape changes (Figure 3D). Similar results were obtained with fragments of mammary tissue dissected from mutant glands (data not shown). These data suggest that OT-stimulated mammary myoepithelial cell contraction requires RhoA pathway activation.

Mammary myoepithelial cells depleted of α3β1 integrin are unable to complete the contraction/relaxation cycle

To define the molecular mechanisms involved in the regulation of mammary myoepithelial cell contraction by α3β1 integrin, we used a mouse myoepithelial cell line carrying conditional alleles of itga3 (MEC-α3+) established in our laboratory. These cells express myoepithelial markers, including K5, α-SMA and P-cadherin, and are negative for luminal markers, such as K8 (Supplementary Figure S3A–C). The adenoviral delivery of Cre recombinase, followed by cell sorting with anti-α3 antibody, yielded myoepithelial cells lacking α3β1 integrin (MEC-α3−) (Figure 4A and B; Supplementary Figure S3D). Flow cytometry analysis and adhesion assays revealed no major change in the surface expression of other integrin subunits or in adhesion properties of α3β1-deficient cells (Supplementary Figure S3D and E).

In agreement with the hypercontractile phenotype of mutant mammary myoepithelial cells observed in vivo, MEC-α3− showed an increased capacity to contract 3D collagen gels as compared with MEC-α3+ cells (Supplementary Figure S4A). However, both, MEC-α3+ and MEC-α3− contracted collagen gels regardless of the presence of OT (not shown) precluding us from further analysis of the molecular mechanisms underlying the abnormal contractile phenotype of α3β1-deficient cells, in this model.

The contractile response of MEC-α3+ and MEC-α3− to OT was analysed using time-lapse video microscopy. We have found that both, MEC-α3+ and MEC-α3− cells responded to OT by a decrease in surface area in a dose-dependent manner (Figure 4C and E, upper panels), and that most MEC-α3−, but not MEC-α3+ cells responded to lower doses of OT (Figure 4C). Analysis of OT receptor transcript levels by quantitative RT–PCR revealed no statistically significant difference between MEC-α3+ and MEC-α3− (Figure 4D), suggesting that the capacity of α3− cells to respond to lower doses of OT was not due to increased expression of OT receptor.

Importantly, the planar cell surface area did not change in response to OT, if MECs were pre-treated with an inhibitor of myosin ATPase, blebbistatin (Supplementary Figure S4B), strongly indicating that the decrease of cell surface area can serve as the readout of cell contraction.

Consistent with the results obtained with lactating mouse mammary tissue fragments, the OT-induced contraction of MEC-α3+ and MEC-α3− depended largely on ROCK. Indeed, pre-treatment with the ROCK inhibitor abolished OT-induced contraction (Figure 4E and F; Supplementary Movie 3). Thus, regardless of the presence of α3β1 integrin, the myoepithelial cell contractile response to OT involved activation of the RhoA pathway.

Treatment with 10⁻⁹ M OT induced a contractile response in most MEC-α3+ and MEC-α3−. This concentration was therefore chosen for further experiments. No significant difference was observed between MEC-α3+ and MEC-α3− in terms of the time course and intensity of the primary contractile response, and a minimal surface area of 54.8±1.6 and 51.9±1.4% of the initial area, for MEC-α3+ and MEC-α3−, respectively, was reached about 6 min after stimulation (Figure 4E and F; Supplementary Table I). Following contraction, MEC-α3+ and MEC-α3− gradually recovered their initial surface area, with mean relaxation times of 20.2±1.1 and 20.4±0.9 min, respectively. After relaxation, the surface area of most MEC-α3+ remained constant, whereas about half the MEC-α3− failed to maintain the state of post-contraction relaxation and underwent additional contractions (Figure 4E and F; Supplementary Table I; Supplementary Movies 1 and 2). These data are in keeping with the observation that high levels of MLC phosphorylation were maintained in lactating K5Cre;itga3^F/F mouse mammary gland. Thus, α3β1 integrin-mediated signalling is required for the accomplishment of the contraction cycle, i.e., MLC dephosphorylation and relaxation.

To study whether the regulation of the cell contractile function by α3β1 integrin involves its engagement at the cell surface, we pre-incubated MEC-α3+ cells with a function-blocking anti-β1 integrin antibody before OT treatment. The anti-β1 integrin antibody induced additional contractions in MEC-α3+ cells, phenocopying α3-deficient cells, and confirming that integrin interaction with its ECM ligands is essential for the control of the contraction/relaxation cycle in the myoepithelial cells (Supplementary Figure S4C; Supplementary Table 1).

α3β1-mediated stimulation of the FAK/Rac/PAK pathway is required for the inhibition of MLCK and the accomplishment of the myoepithelial cell contraction/relaxation cycle

Focal adhesion kinase (FAK) controls Rho-family GTPase activation following integrin engagement (reviewed in Vincente-Manzanares et al (2009) and Huveneers and Danen (2009)). We therefore analysed FAK phosphorylation levels in the immunoprecipitates obtained from 1 day-lactating mouse mammary gland protein extracts, with anti-FAK antibodies. The levels of total FAK phosphorylation (as estimated with an anti-PY antibody) and FAK autophosphorylation at tyrosine 397 (Y397) were considerably lower in mutant glands than in control tissue (Figure 5A), suggesting an impaired Rho-family GTPase activation.

Previous studies in keratinocytes have shown that α3β1-mediated signalling may lead to Rac activation (Choma et al, 2004, 2007; Hamelers et al, 2005). In the extracts obtained from 1 day-lactating K5Cre;itga3^F/F mouse mammary glands, the level of active Rac was found to be significantly lower than in control tissue, whereas the level of Rho activation was higher in mutant gland extracts than in control (Figure 5B). In untreated and OT-treated MEC-α3−, Rac activation levels were found to be significantly decreased as compared with MEC-α3+ (Figure 5C). As expected, OT stimulation did not affect Rac activation but resulted in significant increase of active Rho levels in control MEC-α3+ and MEC-α3−. However, prior, as well as after OT treatment, active Rho levels in MEC-α3− were similar to those of MEC-α3+ (Figure 5C). Western blot analyses of extracts prepared from 1 day-lactating mouse mammary glands and cultured MECs did not reveal any α3β1-dependent significant differences in Rac and Rho protein levels (Supplementary Figure S5A and B; data not shown). Altogether, these data indicate that the Rho/Rac activity balance is altered in α3β1 integrin-deficient mammary myoepithelial cells.

Next, we investigated the effect of Rac inhibition on the contractile response of MEC to OT. The treatment of control MEC-α3+ with Rac inhibitor NSC23766 before OT stimulation induced additional contractions after the primary contractile response (Figure 5D and E; Supplementary Table I; Supplementary Movie 4). Similarly, the expression of dominant-negative Rac (Rac^N17) in MEC-α3+ led to additional contractions (Figure 5D and E; Supplementary Figure S5C; Supplementary Table I; Supplementary Movie 5). The p21-activated kinase (PAK) is a downstream effector of Rac signalling (Bagrodia and Cerione, 1999). Expression of dominant-negative PAK (PAK^DN) or its silencing with shRNA (PAK-sh) leading to approximately two-fold decrease of initial PAK protein levels, induced additional contractions in control MEC-α3+ cells (Figure 5D and E; Supplementary Figure S5D and E; Supplementary Table I; Supplementary Movie 6). Thus, inhibition of the Rac/PAK pathway phenocopied the lack of α3β1 integrin in MEC. The pre-treatment of MEC-α3− with Rac inhibitor had only a minor effect on the response to OT, suggesting that Rac activity might already be inhibited in the absence of α3β1 (Supplementary Figure S5F). We then obtained MEC-α3− expressing a constitutively active form of Rac, Rac^V12, or a catalytically active form of PAK, PAK^CA (Supplementary Figure S5G and H). Expression of Rac^V12 or PAK^CA in MEC-α3− prevented additional contractions, with about 80% of the cells remaining stable after the primary contractile response (Figure 5F and G; Supplementary Table I; Supplementary Movies 7 and 8). Thus, activation of the Rac/PAK signalling pathway rescued the abnormal phenotype caused by a lack of α3β1 integrin.

Several studies have demonstrated that PAK can affect myosin contractile function by phosphorylating and inhibiting MLCK (Sanders et al, 1999; Goeckeler et al, 2000; Murthy et al, 2003; Wirth et al, 2003). We, therefore, investigated whether reducing MLCK activity restored the relaxation capacity of MEC-α3− cells. As expected, pre-treatment of MECs with 10 μM of the MLCK inhibitor ML7, prevented the OT-induced contraction (data not shown). However, treatment with a low dose of ML7 (1 μM) did not interfere with primary contractile response of MEC-α3+ and MEC-α3− and had only minor effect on MEC-α3+ post-contraction relaxation (data not shown; Figure 5H; Supplementary Table I). Strikingly, treatment of MEC-α3− with 1 μM ML7 led to the restoration of their relaxation capacity, such that ∼70% of the cells remained stable after the primary contraction (Figure 5H; Supplementary Table I; Supplementary Movie 9). These results suggest that α3β1 integrin-deficient cells display an increase in MLCK activity, leading to abnormally high MLC phosphorylation levels and impaired relaxation.

In concert, these data strongly indicate that, in the absence of α3β1 integrin, Rac activation is impaired, leading to the maintenance of MLC phosphorylation and impeding myoepithelial cell relaxation. This incomplete contraction/relaxation cycle should prevent subsequent full-strength contractions, accounting for the lower milk ejection rates observed in K5Cre;itga3^F/F mice.

Mouse strains

All animal experiments were conducted in accordance with the French veterinary guidelines and Council of Europe guidelines for experimental animal use (L358-86/609EEC). The K5Cre mice were kindly provided by Dr J Jorcano (Ramirez et al, 2004). The generation of mice carrying conditional α3 integrin alleles (itga3^F/F) has been described elsewhere (Margadant et al, 2009). Wild-type, itga3^F/+, itga3^F/F mice were used as control. Only females with litters of six to eight pups were selected for the analysis of mammary phenotype at lactation.

BrdU-incorporation test, tissue preparation, whole-mount staining, histology and immunostaining

Mice were injected intraperitoneally with bromodeoxyuridine (BrdU), 0.25 mg/g of body weight, 2.5 h before tissue collection. Dissected mammary glands were either embedded into cryoprotectant (Tissue-Tek^® O.C.T) and processed for cryosections (7 μm) and immunostained as previously described (Raymond et al, 2005) or spread onto a glass slide, fixed in MethaCarn and, either stained with Carmine alum in whole-mount, or embedded in paraffin. Sections (5–7 μm thick) were used for haematoxylin/eosin or immunofluorescence staining (Teulière et al, 2005). Cells grown on glass coverslips were processed for immunostaining as previously described (Raymond et al, 2005). Cell nuclei were stained with DAPI. The list of primary antibodies used for immunostaining is provided in the Supplementary data. All secondary antibodies and phalloidin were Alexafluor conjugated (Molecular Probes). The microphotographs shown in Figure 2D were acquired with Leica SP5 confocal microscope. Leica DM6000B microscope was used for other immunofluorescence images.

Evaluation of the OT effects on alveolar myoepithelial cell shape

Mammary glands were processed as described elsewhere (Moore et al, 1987). Briefly, glands were chopped finely with a scalpel in a buffer containing 136 mM NaCl, 1.2 mM NaH₂PO₄, 1.2 mM MgSO₄, 5 mM KCl, 1.7 mM CaCl₂, 1.1 mM glucose, 0.03 mM Na₂ EDTA, 10 mM Hepes, pH 7.2 and, when indicated 10 μM Y27632 (Calbiochem) for 10 min and then treated with 50 nM oxytocin (Tebu-bio) for 2 min at 37°C. Tissue fragments of ∼1 mm³ in size were subsequently fixed for 30 min in 4% paraformaldehyde, then for 5 min in 100% acetone at −20°C, re-hydrated in PBS containing 1% BSA and incubated with the anti-keratin 5 antibody, phalloidin and DAPI. After incubation with the secondary antibody and washing, tissue fragments were mounted in Thermo Scientific Shandon Immu-Mount medium for observation under a Leica DM6000B microscope.

Quantitative RT–PCR

RNA was isolated from the third mammary glands of 1 day-lactating females with the RNeasy kit (Quiagen Gmbh). Reverse transcription and quantitative PCR were performed as previously described (Taddei et al, 2008). The sequences of the primers used are provided in the Supplementary data and in Taddei et al (2008). The values were normalized to glyceraldehyde phosphate dehydrogenase expression levels using the comparative CT method.

Protein extracts, immunoprecipitation and immunoblotting

Proteins extracts were prepared from the fifth mammary glands of 1 day-lactating females. Tissue (125 mg/ml) was homogenized in lysis buffer (50 mM, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 0.1% SDS, 4 mM EDTA, 20 mM NaF, 2 mM sodium orthovanadate) containing a cocktail of protease inhibitors (Sigma-Aldrich). The extracts were cleared by centrifugation at 10 000 g at 4°C for 15 min. For immunoprecipitation assays, 1 mg of protein extracts was pre-absorbed with gammaBind Plus Sepharose beads (GE Healthcare) for 1 h at 4°C on a vertical rotator, incubated for 1 h with 4 μg of primary antibody and for 2 h with gammaBind Plus Sepharose beads. The beads were then washed four times with lysis buffer, resuspended in reducing sample buffer for electrophoresis and boiled for 5 min. The supernatants were run on a NuPAGE^® Novex^® 4–12% Bis Tris gel (Invitrogen). After electrophoresis, proteins were transferred onto nitrocellulose. Membranes were successively incubated with 5% BSA in TBST (10 mM Tris, pH 8, 150 mM NaCl, 0.1% Tween 20) for 1 h at room temperature to block non-specific binding sites and with primary antibodies in TBST, overnight at 4°C. Secondary antibodies coupled to horseradish peroxidase were purchased from Amersham. Detection was performed by chemiluminescence (Super Signal^® west Pico chemiluminescent substrate, Pierce) using a Fuji Imager LAS3000. The list of the antibodies used for immunoprecipitation and immunoblotting is provided in the Supplementary data. Protein extracts from four control and mutant glands were analysed. Protein extracts from MEC were prepared as previously described (Raymond et al, 2005) and immunoblotting was performed as described above.

Establishment of MEC line, viral transduction, stable transfection and flow cytometry

The established MEC of itga3^F/F genotype was routinely maintained in low serum medium consisting of DMEM:F12 (Gibco-BRL) buffered with 15 mM Hepes, pH 7.6 (Sigma-Aldrich) containing 10 μg/ml insulin (Sigma-Aldrich), 5 ng/ml epidermal growth factor (Sigma-Aldrich), 0.5 mg/ml bovine serum albumin (Sigma-Aldrich), 5 μg/ml linoleic acid complex (Sigma-Aldrich), 2% calf serum (HyClone), 50 μg/ml gentamycin (Sigma-Aldrich) and 20 U/ml nystatin (Sigma-Aldrich). The line was established as previously described (Medina and Kittrell, 2000). In brief, primary mammary epithelial cells obtained by collagenase/hyaluronidase digestion of inguinal glands from 16 day-pregnant itga3^F/F females were seeded into matrigel in low serum medium defined above. After 4 weeks of growth, the cells were recovered from matrigel using dispase, plated on collagen-coated plastic and maintained in culture. After ∼2 months, when most cells became senescent, and only few clones were capable of expanding, the proliferating cell clones were harvested and amplified. The cell clone stably expressing myoepithelial cell markers and negative for mammary luminal markers was chosen for further studies (Supplementary Figure S3). Cre-mediated deletion of itga3^F/F alleles was performed as described previously (Raymond et al, 2005). Retroviral transduction was performed as described elsewhere (Sterk et al, 2000), and infected cells were subsequently maintained under G418 selection (500 μg/ml; #ant-gn-5, Invivogen). The retroviral construct myc–Rac^V12 cloned in pLZRS-MS-IRES-Neo/pBR was kindly provided by Dr JG Collard. Stable transfections were performed with Lipofectamin™ 2000 (#11 668, Invitrogen) according to the manufacturer's instructions. The pCAG-Myc-PAK-CAAX-IRES-GFP, pCAG-Myc-PAK-R299-CAAX-IRES-GFP and pCDNA-GFP-Rac^N17 vectors provided by Dr M Nicolic (Jacobs et al, 2007) and Dr A Gautreau, respectively, were co-transfected with pGK puro plasmid. Transfected cells were selected using puromycin (10 mg/ml, #ant-pr-5, Invivogen), and the bulk population of cells was analysed in all cases. The αPAK shRNA and control shRNA plasmids were from Santa Cruz Biotechnology (036SC-270016-SH and sc-108060). Transfections were performed according to the manufacturer's instructions. Flow cytometry and cell sorting were performed, respectively, on BD FACSCalibur and FACSvantage (Becton Dickinson) flow cytometers, as previously described (Margadant et al, 2009).

Time-lapse video microscopy

For video microscopy, sparse cell cultures were grown on glass-bottomed microwell dishes (MatTek Corporation) for 16–20 h. Time-lapse video microscopy (one image captured per min) was performed on Time Lapse Imaging System Biostation IM (Nikon Corporation) with an integrated incubation system (37°C, 5% CO₂). MLCK inhibitor (ML7) was used at concentrations of 1 μM and 10 μM, ROCK inhibitor (Y27632) at 10 μM; Rac1 inhibitor (NCS23766) at 100 μM; Myosin ATPase inhibitor (blebbistatin) at 25 μM (all from Calbiochem). OT (1 nM, or as indicated; Tebu-bio) was added 45–120 min after inhibitor and the response was followed during 60 min. Planar surface area was measured as readout for contractile responsiveness with Image J software. Initial surface area was defined as a mean value over 15 min. Spontaneous cell surface area fluctuation did not exceed 9.6%. A decrease in surface area was considered as contraction when the cell surface area was ⩽72% of the initial area. Cells with a surface area of at least 86% of the initial surface area were considered as relaxed. Cells that were not spreading (showing refringent borders), abnormally large (usually polynucleated) or dividing, and cells that did not remain stable during the 15 min preceding the addition of OT were excluded from the analysis. The data obtained in four independent experiments are presented.

Rho-GTPases activation assay

Quantification of Rac and RhoA activation was performed using G-LISA Rac and RhoA Activation Assay Biochem Kits, following the manufacturer's instructions (Cytoskeleton, Inc.).

Statistical analysis

Statistical analysis was performed with Prism 4 software (GraphPad) using Mann–Whitney U test for the comparison of two groups except in Figure 2E where unpaired t-test was used. A value of ^*P<0.05, ^**P<0.01 or ^***P<0.001 denoted statistical significance.