CathepsinKCre mediated deletion of βcatenin results in dramatic loss of bone mass by targeting both osteoclasts and osteoblastic cells

It is well established that activation of Wnt/βcatenin signaling in the osteoblast lineage leads to an increase in bone mass through a dual mechanism: increased osteoblastogenesis and decreased osteoclastogenesis. However, the effect of this pathway on the osteoclast lineage has been less explored. Here, we aimed to examine the effects of Wnt/βcatenin signaling in mature osteoclasts by generating mice lacking βcatenin in CathepsinK-expressing cells (Ctnnb1f/f;CtsKCre mice). These mice developed a severe low-bone-mass phenotype with onset in the second month and in correlation with an excessive number of osteoclasts, detected by TRAP staining and histomorphometric quantification. We found that WNT3A, through the canonical pathway, promoted osteoclast apoptosis and therefore attenuated the number of M-CSF and RANKL-derived osteoclasts in vitro. This reveals a cell-autonomous effect of Wnt/βcatenin signaling in controlling the life span of mature osteoclasts. Furthermore, bone Opg expression in Ctnnb1f/f;CtsKCre mice was dramatically decreased pointing to an additional external activation of osteoclasts. Accordingly, expression of CathepsinK was detected in TRAP-negative cells of the inner periosteal layer also expressing Col1. Our results indicate that the bone phenotype of Ctnnb1f/f;CtsKCre animals combines a cell-autonomous effect in the mature osteoclast with indirect effects due to the additional targeting of osteoblastic cells.

Scientific RepoRts | 6:36201 | DOI: 10.1038/srep36201 catenin signaling in osteoblast and osteocytes upregulates the osteoclasts inhibitor factor osteoprotegerin (OPG) and thereby inhibits osteoclast formation and bone resorption 10,11 . Bone loss mediated by decreasing OPG expression is also observed in the conditional deletion of βcatenin in type II Collagen expressing cells 12 . Thus, a cellautonomous mechanism promoting proliferation and maturation of osteoblasts together with an indirect non cell-autonomous mechanism, mediated by an OPG-dependent reduction in osteoclast numbers, contribute to the high bone mass phenotype.
The involvement of osteoclasts as mediators of some of the effects of Wnt/β catenin signaling in the osteoblast lineage, the ubiquity of WNT proteins and the expression of components of the Wnt canonical pathway in the osteoclast lineage raised the question of whether this pathway could also play a cell-autonomous role in the osteoclast biology 13,14 . To explore this possibility, several attempts have been made to genetically manipulate βcatenin in the osteoclast lineage [15][16][17][18] . For example, the LysozimeMCre (LysMCre) deleter line, which drives recombination specifically in monocyte/macrophages and neutrophils 19 has been used to perform gain 17 and loss [15][16][17] of function of βcatenin in osteoclast precursors. βcatenin deletion in LysozimeM expressing cells leads to osteopenia by increasing osteoclastogenesis but without altering OPG levels. These results suggest that direct actions of Wnt/ β catenin signaling in the osteoclast lineage contribute to bone homeostasis [15][16][17] . However, in these models the genetic manipulation of βcatenin is not restricted to the osteoclasts but it also affects other cell types such as macrophages and neutrophils which could potentially contribute to the phenotype. In the present work, we aimed to study the function of Wnt/β catenin specifically in mature osteoclasts using the CtsKCre deleter line, which is commonly used to target osteoclasts at their late stage of differentiation [20][21][22] . CtskCre;βcatenin mice developed severe low-bone-mass caused by excessive osteoclast numbers. In vitro experiments showed that Wnt signaling decreased osteoclast numbers by reducing their life span in a cell-autonomous manner. We also identified a population of osteogenic cells located in the inner layer of the periosteum that express Ctsk postnataly. The CtskCre mediated removal of βcatenin in this population of osteoblasts, most probably, led to reduction in OPG production, thereby, contributing to the excessive osteoclast formation and dramatic loss of bone mass.

Generation of mice with specific βcatenin deletion in Ctsk-expressing cells. Conditional inactiva-
tion of βcatenin (MGI: Ctnnb1) in mature osteoclasts was performed using the βcatenin floxed allele (Ctnnb1 f ; 23 ) and the CathepsinK;Cre line (CtsKCre; 21 ) that harbors the Cre recombinase at the CathepsinK locus. We generated two cohorts (males and females) of mice lacking βcatenin in CtsK-expressing cells, hereafter referred to as Ctnnb1 f/ f ;CtsKCre or experimental mice (n = 13 males and 17 females), and they were compared with the corresponding Ctnnb1 f/f littermate animals, considered as controls (n = 21 males and 20 females). Ctnnb1 f/f ;CtsKCre mice were born at the expected Mendelian ratio and their gross external aspect and size were indistinguishable from control mice. However, during the second month of postnatal life, the experimental Ctnnb1 f/f ;CtsKCre mice became clearly distinct from their control littermates due to an abnormal gait, malposition of the hindlimbs and sparse fur appearance as shown for 12-weeks old animals in Fig. 1b. Differences in body weight were also seen in males at this age (Fig. 1a). All experimental animals died between 14 and 15 weeks of age, a time when animals appeared sick.
Excision of the βcatenin gene was assessed by qRT-PCR in ex vivo cultures of bone marrow (BM)-derived macrophages and osteoclasts obtained from the femur of experimental and control 4-6 week-old mice. βcatenin mRNA levels in macrophages from Ctnnb1 f/f ;CtsKCre mice were indistinguishable from controls while in osteoclasts βcatenin levels were reduced by approximately 60% (Fig. 1c). The magnitude of the reduction in βcatenin was similar to the one seen with other Cre-deleter models 10,11 . The remaining βcatenin seen in cultures from experimental mice was, most probably, due to the presence of non-CtsK-expressing cells, of the hematopoietic or mesenchymal lineages. βcatenin deletion in CtsK-expressing cells causes a severe loss of bone mass. The hindlimb malposition and abnormal gait of Ctnnb1 f/f ;CtsKCre mice pointed to a skeletal phenotype. To start characterizing this phenotype we performed a sequential X-ray examination of the hindlimbs and spine of both experimental and control animals. As early as 6 weeks of age, experimental animals exhibited cortical thinning and alterations in shape including widening of the distal third of the femur and proximal third of the tibia (Fig. 2a, left panels). Furthermore, the secondary ossification center in the distal femur and proximal tibia were markedly misshapen in experimental animals. Poor mineralization of the vertebral bodies was also evident at 12 weeks of age (Fig. 2a, right panels). In line with the X-ray analysis, monthly DEXA bone mineral density BMD analysis showed a reduction of bone mass in the spine of Ctnnb1 f/f ;CtsKCre females, in the second analysis performed at 8 weeks and in both the spine and femur of females and males at 12 weeks (Fig. 2b).
To examine the bone microarchitecture in Ctnnb1 f/f ;CtsKCre mice, we performed micro-computed tomography (μ CT) of the femur and fifth lumbar vertebra (L5) of 6 and 12 week-old animals. The μ CT images showed a clear bone loss in the distal femur in 6 week-old animals (Fig. 3a). Compared with respective controls, experimental female and male mice had decreased trabecular number (TbN) and increased trabecular separation (TbSp) (Fig. 3b). However, the bone volume fraction (BV/TV) was not significantly different from controls ( Fig. 3b). At 12 weeks of age, μ CT scans of the distal femur revealed a dramatic loss of both cortical and trabecular bone (Fig. 3c) and both female and male experimental animals showed 50% decrease in BV/TV (Fig. 3d). However, no significant changes were detected in trabecular number and spacing but trabecular thickness was significantly increased in females (Fig. 3d). Moreover, despite the enormous damage seen in cortical bone at the distal end, the average cortical thickness (Th.) at the midshaft was unaffected in both male and female Ctnnb1 f/f ;CtsKCre mice ( Fig. 3d).
Like in the femur, cortical and trabecular bone were dramatically reduced in vertebra (L5) of 12-week-old experimental mice (Fig. 3e). Specifically, female and male mice exhibited a decrease in BV/TV, trabecular number, and an increase in trabecular spacing (Fig. 3f). Trabecular thickness was significantly higher in the experimental females (Fig. 3f).
Scientific RepoRts | 6:36201 | DOI: 10.1038/srep36201 Together, these results indicate that βcatenin deletion in CtsK-expressing cells causes a severe loss of both cortical and trabecular bone that started after the first month of postnatal life and worsen progressively.

Ctnnb1 f/f ;CtsKCre mice exhibit normal bone development up to 4 weeks of age. A sequential
Hematoxylin-eosin examination of femur and vertebra sections was performed to further investigate the onset of the phenotype and the bone structure of experimental animals. Up to 4 weeks of postnatal development, both femur and vertebra appeared normal in experimental animals (Fig. 4a,c). However, a dramatic reduction in both the trabecular and cortical bone compartments was evident by 12 weeks old Ctnnb1 f/f ;CtsKCre mice ( Fig. 4a-c) with cortical bone absent from wider areas of the distal femur (arrows in Fig. 4a) and vertebra (arrowheads in Fig. 4c) in 12 weeks old experimental animals. The loss of bone in distal femur was obvious in the fresh bone (Fig. 4b) and accompanied by extensive bone marrow adiposity (Fig. 4d). It should be noted, that the long bone phenotype was not restricted to the distal femur, as similar findings were observed in the proximal tibia (not shown) and humerus (Fig. 4e).
The results of the histological analysis confirm that Wnt/β catenin signaling in CtsK-expressing cells is not necessary for normal skeletogenesis up to the end of the first month of postnatal life and suggest that excess bone destruction, starting between 4 and 6 weeks, might be the primary mechanism involved in the bone loss phenotype. Ctnnb1 f/f ;CtsKCre mice exhibit increased osteoclast numbers. To determine whether altered bone resorption was the cause of the bone phenotype in Ctnnb1 f/f ;CtsKCre mice, osteoclast numbers were evaluated in histological sections of femur and vertebra. Tartrate-resistant acid phosphatase (TRAP) specific staining revealed an increase in osteoclasts in the distal femur and lumbar vertebra of 4-week-old experimental mice as compared to littermate controls (Fig. 5a,b). Indeed, quantitation of osteoclasts in representative vertebral sections confirmed that osteoclast numbers were significantly increased in experimental mice ( Fig. 5c) at 4 weeks of age just before the phenotype was apparent. However, and in line with the μ CT results, the bone area per total area did not show differences between control and experimental mice (Fig. 5d), confirming the late onset of the osteopenic phenotype. Osteoclasts were barely detectable in 12-week-old experimental animals, likely because a great portion of the cancellous bone was already lost (Fig. 5a,b). Overall, our results strongly suggest that an increase in osteoclast numbers is the primary cause of the bone defects in Ctnnb1 f/f ;CtsKCre mice.

Wnt/βcatenin signaling promotes osteoclast apoptosis and attenuates osteoclast numbers in vitro.
We also assessed the number of osteoclast progenitors in the BM by culturing BM-cells obtained from the femur of control and experimental mice in the presence of RANKL and M-CSF for 5 days. The number of osteoclasts developed in cultures from Ctnnb1 f/f ;CtsKCre mice was significantly higher than in cultures from littermate controls (Fig. 5e). It should be noted that βcatenin deficient osteoclasts exhibited normal morphology indicating that βcatenin is dispensable for the intracellular cytoskeleton architecture and cell-cell adhesion of osteoclasts (Fig. 5e). As suggested in previous studies, plakoglobin may substitute for β catenin in these functions 6,24,25 .
In accordance with this observation, addition of recombinant WNT3A dose-dependently decreased the number of osteoclasts derived from BM of wild type mice in the presence of M-CSF and RANKL (Fig. 6a). Thus, stimulation of the canonical Wnt signaling attenuates the number of osteoclasts derived from BM cells in vitro. Because the genetic manipulation in our experimental mice occurs in a late phase of osteoclast differentiation, most likely when osteoclast are already postmitotic, we considered the alternative possibility of Wnt signaling modifying their life span.  Osteoclasts have a short life span and die by apoptosis, therefore, changes in their life span may alter osteoclast numbers. We next examined the role of Wnt/β catenin signaling in osteoclasts apoptosis. Addition of WNT3A to osteoclast cultures from BM of control littermates mice stimulated osteoclast apoptosis, as determined by the TUNEL assay (Fig. 6b). The pro-apoptotic effect of WNT3A was abrogated in osteoclast cultures obtained from Ctnnb1 f/f ;CtsKCre mice ( Fig. 6b), confirming β catenin requirement for this effect. Addition of DKK1 to the cultures also blunted the pro-apoptotic actions of WNT3A in osteoclasts as determined by caspase 3 activity (Fig. 6c). These results show that canonical Wnt/β catenin signaling promotes osteoclast apoptosis in vitro.  models, the discrepancy in the phenotype suggests the involvement of additional cell targets in one or both models. Indeed, several observations pointed to additional contribution of non-osteoclast cells to the phenotype of Ctnnb1 f/f ;CtsKCre mice. First, the similarity of the bone phenotype with that of mice in which βcatenin was removed from osteoblasts and osteocytes, or chondrocytes suggested a common causal mechanisms 11,12,24,26 . Second, the evaluation of βcatenin mRNA expression level in the experimental bone (assessed in tibia from 4-week-old animals) was reduced by 70%, much more than expected if only osteoclasts, which represent only a scanty 3-5%, proportion of bone cells, were responsible for this decrease (Fig. 7a).
Since a decline in OPG is the main culprit for the increased resorption in mice in which βcatenin was deleted from osteoblast but not from osteoclast lineage cells, we measured Opg mRNA expression in the whole tibia in control and Ctnnb1 f/f ;CtsKCre mice. Opg mRNA levels were significantly decreased in experimental mice (Fig. 7b). In contrast, RANKL expression was indistinguishable from littermate controls (Fig. 7b). The decrease in OPG in Ctnnb1 f/f ;CtsKCre mutant mice provides an additional cause for the increase in osteoclast numbers and strongly suggests that CtsK-Cre targets cells other than osteoclasts.
To identify this putative cell type we analyzed the mRNA expression pattern of CtsK by in situ hybridization in tissue sections of femur and vertebra from embryonic day (E) 13.5 to 4 weeks of postnatal development (Supplemental Fig. 1 and not shown). CtsK expression was first detected at E16.5 and, during this and subsequent developmental stages, the pattern of CtsK-expressing cells coincided with the distribution of TRAP positive cells, as expected ( Fig. 7c and Supplemental Fig. 1).
Importantly, the expression of CtsK was also found in TRAP negative cells located in the inner layer of the periosteum (red arrows in Fig. 7c). The periosteum is a thick sheet that covers the outer surface of bones and is composed of two distinct layers: an outer fibrous layer that lends structural integrity and an inner layer that has osteogenic potential. The fibrous layer is primarily composed of fibroblasts, collagen, and elastin while the inner layer, also called the cambium layer, is highly cellular and includes differentiated osteogenic progenitors, adult mesenchymal progenitor cells and osteoblasts [27][28][29] .
The periosteal CtsK-expressing cells appeared scattered in the cambium layer, not in direct contact with the bone surface, but adjacent to the line of cuboid osteoblasts (Fig. 7c,d). At postnatal (P) day 7, these cells were negative for Collagen type1 (Col1), Bone sialoprotein (Bsp), and sclerostin (Sost) (Fig. 7c). Most interestingly, at P15, these cells also expressed Col1 but not Bsp or Sost, suggesting that they are a subpopulation of undifferentiated osteogenic cells (Fig. 7d). It should be noted that we could not detect CtsK expression in osteocytes or in the Ranvier groove at this stage (Fig. 7). Together, these results strongly indicate that the CtsKCre line targets cells other than osteoclasts in bone and that deletion of βcatenin in these cells contributes to the loss of bone mass in Ctnnb1 f/f ;CtsKCre mice.

Discussion
The aim of this work was to examine the function of Wnt/β catenin signaling specifically in mature osteoclasts. Surprisingly, the removal of βcatenin at a late stage of osteoclast differentiation using the CtsKCre line caused a much more severe low-bone-mass phenotype than when βcatenin is removed from early myeloid precursors  using the LysMCre line [15][16][17] . While Ctnnb1 f/f ;LysMCre mice show a mild reduction in the femur trabecular bone 15 , the Ctnnb1 f/f ;Ctsk-cre mutants show a dramatic loss of both trabecular and cortical bone in both long bones and vertebra. Because in both animal models the mature osteoclast is deficient in βcatenin, it is reasonable to assume that the discrepancies in the phenotype rely on additional targets, distinct from osteoclasts, present in one or both of these models. Also, it should be considered that Ctnnb1 f/f ;CtskCre mice lack a functional allele of Ctsk, one of the most potent collagenases 30 . This raises the question of whether the degradation of ECM proteins, such as collagen and elastin, could be impaired in our experimental mice and contribute to the phenotype. However, this possibility can be discarded because Ctsk heterozygous mutant have no bone phenotype 31,32 and because the Ctnnb1 f/f ;CtskCre phenotype is characterized by excessive rather than limited osteoclasts function.
Here, we have focused on the causes of the severe low-bone-mass in Ctnnb1 f/f ;CtsKCre mice and show that it associates with excessive osteoclast numbers evident at the end of the first month of postnatal life and coincident with the onset of the phenotype. The increase in osteoclast numbers results from the combination of two mechanisms. First, the removal of βcatenin in mature osteoclasts extends their life span in a cell-autonomous manner and second, the additional activity of the CtsKCre line in a sub-population of osteoblasts leads to increased osteoclastogenesis through the reduction in OPG production.
An essential role for βcatenin in controlling osteoclastogenesis has been previously demonstrated. It has been shown that WNT3A reduces osteoclast numbers by sustaining M-CSF-induced proliferation of osteoclasts precursors and preventing their differentiation into mature osteoclasts 15,17 . Recently, it has been shown that WNT3A suppresses RANKL-mediated NFATc1 expression in osteoclast precursors, while activates the cAMP/PKA pathway, as mechanisms by which WNT3A blunts osteoclast differentiation 18 .
Our in vitro experiments provide compelling evidence for a role of cell-autonomous Wnt/βcatenin signaling in reducing the life span of osteoclasts. The pro-apoptotic effect of Wnt/β catenin in osteoclasts may occur all along the osteoclast differentiation life as an additional mechanism together with the attenuation of differentiation, and our experiments suggest that this effect is still present at the end of the osteoclastic differentiation process. Several studies have demonstrated that modulation of osteoclasts life span leads to changes in bone mass 33 . For example, estrogens protect the skeleton, in part, via pro-apoptotic effects in osteoclasts 21,34 . Indeed, deletion of ERα in the osteoclast lineage cells, similar to deletion of βcatenin, increases osteoclast numbers and decreases bone mass. In contrast to estrogens, glucocorticoids cause loss of bone mass, at least in part, via prolongation of osteoclast life span. Interestingly, and similar to estrogens and glucocorticoids, Wnt/β catenin signaling exerts an opposite effect on the life span of osteoblasts and osteoclasts i.e. it increases the life span of osteoblasts and decreases the life span of osteoclasts 35 . Similar to our findings in osteoclasts, WNT3A/β catenin mediate the effects of hypoxia reoxygenation on rat cardiomyoblasts cell death 36 . Moreover, Wnt/β catenin signaling decreases melanoma cell invasiveness by a mechanism that enhances the expression of pro-apoptotic proteins 37 .
Several findings in Ctnnb1 f/f ;CtsKCre mice pointed to the involvement of cells other than osteoclasts in the bone phenotype of these mice. First, the reduction in the expression of βcatenin in bone is much higher than expected from the sole removal in osteoclasts, a minority cell type in bone. Second, the expression of Opg, a well-known target of βcatenin in osteoblasts, is dramatically reduced in bone. Third, the bone phenotype, including the excessive bone marrow adipocytes, is reminiscent of the phenotype exhibited by mice in which βcatenin is deleted in osteoblasts, osteocytes or chondrocytes 11,12,24,26,38 . These observations, together with the identification in the inner periosteal layer of cells that are TRAP-negative but concomitantly express CtsK and Col1, support the conclusion that a sub-population of osteogenic cells contributes to the phenotype via indirect control of osteoclastogenesis. Because the CtsK-expressing cells identified here are located in the inner layer of the periosteum and express Col1 at P15, we propose that they are a sub-population of osteoblastic cells. Although it has been reported that osteocytes can express CtsK under certain circumstances 39 , our study failed to detect cells that concomitantly expressed Sost and CtsK during the period studied.
Besides Wnt/β catenin, Bmp signaling is another important pathway regulating both bone formation and bone resorption and it is known that both pathways interact in complex and varied ways 2,40 . Thus, Bmp signaling, through the Bmpr1a, positively regulates sclerostin expression as well as canonical Wnt signaling 41,42 . In addition, Wnt/β catenin signaling is required for the Bmp osteogenic effects 43 and Bmp2 directly enhances osteoclastic differentiation and survival by regulating expression of RANKL and CSF1 44,45 . While less is known on Wnt/β catenin modulation of Bmp signaling in bone, recent studies have shown upregulation of Bmp ligands by Wnt/β catenin signaling 46 . Thus, it is possible that Bmp signaling is altered in Ctnnb1 f/f ;CtsKCre mice, an issue that certainly deserves further analysis.
Interestingly, CtsK expression has also been recently described surrounding the physis, in the anatomical structure called the groove of Ranvier, which contains osteochondro-progenitor cells that contribute to bone development 47 . However, despite careful analysis we could not detect CtsK expressing cells in such location, ruling out the participation of this structure in the phenotype of our experimental mice. Micro-computer tomography (μCT). μ CT analysis of the fifth lumbar vertebra and right femur of 10 mice per genotype was performed into 12.3 mm diameter scanning tubes and imaged (μ CT40, Scanco Medical, Basserdorf, Switzerland). Scans were integrated into 3-D voxel images (1024 × 1024 pixel matrices for each individual planar stack) and a Gaussian filter (sigma = 0.8, support = 1) was used to reduce signal noise. A threshold of 200 was applied to all analyzed scans. Scans were done at medium resolution (E = 55 kVp, I = 145 μ A, integration time = 200ms). The entire vertebral body was scanned with a transverse orientation excluding the pedicles and articular processes. Manual analysis excluded the cortical bone and the primary spongiosa from the analysis. All trabecular measurements were made by manually drawing contours every 10 to 20 slices and using voxel counting for bone volume per tissue volume and sphere filling distance transformation indices without assumptions about the bone shape as a rod or plate for trabecular microarchitecture. Cortical thickness was measured at the femur mid-diaphysis. Densitometry, and μ CT analyses were performed blind.

Methods
Bone histology. Three lumbar vertebrae (L2-4) and left femur of 6 animals for each genotype, were fixed in 4% paraformaldehyde overnight at 4 °C and decalcified in 9% EDTA (pH7.4) for a variable time period, depending on the stage, before embedding in paraffin. For the histomorphometric analysis, osteoclasts were stained for tartrate-resistant acid phosphatase (TRAP) and measurements performed with Fiji software. Data are reported using the nomenclature recommended by the American Society for Bone and Mineral Research.

Quantitative real-time PCR (qRT-PCR). Total RNA of 4 animals per genotype, was extracted with
TRIzol reagents (Life Tecnologies). 1μ g total RNA was used to produce first-strand cDNA using the m-MLV RT enzyme (Invitrogen). qRT-PCR was carried out in duplicates using the PreMix Ex taq (Takara) and the data were analyzed using the Biorad software. The primers and probes for βcatenin [Mm01350385_g1 (fam)], and gapdh [Mm99999915_g1 (vic)] were manufactured by the TaqMan Gene. Expression Assays service (Applied Biosystems).
Apoptosis assays. Osteoclasts derived from purified non-adherent bone marrow osteoclast precursors isolated from 3 animals for each genotype were treated with WNT3A once osteoclasts had developed in the culture plate in triplicates. After 24 hours of treatment, the cultures were fixed and subject to TUNEL and TRAP staining. The total number of osteoclasts and the number of apoptotic osteoclasts in the plate were quantified. Osteoclasts were considered apoptotic when at least one of their nuclei was TUNEL positive. The TUNEL method was performed using the FragEL DNA fragmentation detection kit (EMD Chemicals, San Diego, CA) before staining for TRAP. Multinuclear TRAP-positive and TUNEL-positive cells were enumerated. Caspase-3 activity was measured by determining the degradation of the fluorometric substrate DEVD (Biomol Research Laboratories, Plymouth Meeting, PA), and protein concentration was measured using a Bio-Rad detergent-compatible kit (Bio-Rad Laboratories, Hercules, CA).
In Situ Hybridization. In situ hybridization was performed in sections using digoxigenin-labeled antisense riboprobes. The probes used were Col1, Bsp and Sost 8 . A specific antisense RNA probe for CtsK was generated using the following primers: CtsK-F: gaccactgccttccaatacg and CtsK-R:ctgctggaggactccaatgt. A minimum of three mice of each stage were analyzed. Statistical analysis. All data are reported as the mean ± standard deviation. Group mean values were compared by Student's unpaired two-tailed t test.