Recent studies have suggested a direct association between low serum 25-hydroxy-vitamin D3 [25(OH)D3] levels and the presence, severity, and prognosis of several liver diseases.1-7 Targher et al.1 and Manco et al.2 reported low serum 25(OH)D3 concentrations among adults and children affected by nonalcoholic steatohepatitis (NASH). Furthermore, we demonstrated the existence of a strong association between nonalcoholic fatty liver disease (NAFLD) and low 25(OH)D3 levels in a large adult population with normal serum liver enzymes.3 Recently, other studies have shown the presence of low serum 25(OH)D3 levels in patients affected by chronic hepatitis C (CHC) along with a failure to achieve sustained antiviral responses after standard therapy in CHC subjects with hypovitaminosis D.4-6 Vitamin D is present in the diet and dietary supplements, but its primary source is the photo-mediated conversion of 7-dehydrocholesterol in the skin.8 To become biologically active, vitamin D requires 25-hydroxylation in the liver and subsequent 1-hydroxylation in the kidney.9, 10 The 25-hydroxylation occurs exclusively in hepatocytes and is mediated by CYP27A1 and CYP2R1, two liver-expressed cytochromes characterized by different intracellular location, specificity, and affinity for vitamin D3.11, 12 Low serum 25(OH)D3 levels and hypovitaminosis D–related diseases have been shown to be associated with CYP2R1 polymorphisms by some authors,13, 14 but only one study to date has investigated liver 25-hydroxylase expression.4 The biological effects of 1,25(OH)2 vitamin D3 are mediated by the vitamin D receptor (VDR), an NR1I family receptor with ligand-activated transcription factor activities.15 VDR polymorphisms have been associated with liver diseases such as primary biliary cirrhosis,16, 17 autoimmune hepatitis,17 alcohol-related hepatocarcinoma,18 and HBV.19 VDR expression in normal liver cells has been demonstrated in several studies.20-22 Gascon-Barre et al.20 described VDR reactivity mainly in nonparenchymal and biliary epithelial liver cells, whereas other studies have reported VDR expression in primary human hepatocytes.21-23 Because the majority of the evidence suggests a tight association between low 25(OH)D3 levels and both NAFLD/NASH and CHC, the goal of this study was to evaluate the hepatic expression of VDR, CYP2R1, and CYP27A1 in patients affected by NASH or CHC and their relationship with histological features of the hepatopathy and serum 25(OH)D3 levels.
Evidence suggests an association between low serum 25-hydroxy-vitamin D3 [25(OH)D3] levels and the presence and prognosis of liver disease. Vitamin D receptor (VDR) has been widely detected in the liver, but its expression in the course of liver disease has never been investigated. We evaluated the hepatic expression of VDR along with that of vitamin D 25-hydroxylases in patients with nonalcoholic steatohepatitis (NASH) or chronic hepatitis C (CHC) and its relationship with hepatic histological features and serum 25(OH)D3 levels. We evaluated 61 patients (25 NASH and 36 CHC) who had undergone liver biopsy for clinical purposes and 20 subjects without liver disease. Serum 25(OH)D3 was measured via colorimetric assay. Expression of VDR, CYP2R1, and CYP27A1 was evaluated via immunohistochemistry in hepatocytes, cholangiocytes, and liver inflammatory cells. Parenchymal and inflammatory cells from liver biopsies of patients with NASH and CHC expressed VDR, CYP2R1, and CYP27A1. In NASH patients, VDR expression on cholangiocytes was inversely correlated with steatosis severity (P < 0.02), lobular inflammation (P < 0.01), and nonalcoholic fatty liver disease score (P < 0.03). Moreover, expression of CYP2R1 in hepatocytes correlated strongly with VDR positivity on liver inflammatory cells. In CHC subjects, fibrosis stage was associated with low hepatic CYP27A1 expression, whereas portal inflammation was significantly higher in patients with VDR-negative inflammatory cells (P < 0.009) and low VDR expression in hepatocytes (P < 0.03). Conclusion: VDR is widely expressed in the liver and inflammatory cells of chronic liver disease patients and its expression is negatively associated with the severity of liver histology in both NASH and CHC patients. These data suggest that vitamin D/VDR system may play a role in the progression of metabolic and viral chronic liver damage. (HEPATOLOGY 2012;56:2180–2187)
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Patients and Methods
Sixty-one patients who had undergone liver biopsy for clinical purposes at the Campus Bio-Medico Hospital of Rome were included in the study. There were 36 patients with CHC (20 men, 16 women; age, 55.4 ± 12.4 years; body mass index [BMI], 21.61 ± 3.4 kg/m2) who had received biopsy for grading and staging of liver disease and 25 subjects with suspected NAFLD based on clinical and instrumental data (13 men, 12 women; age, 48.7 ± 13.3 years; BMI, 30.5 ± 5.5 kg/m2) who had undergone liver biopsy for diagnostic confirmation and for staging of the disease. For purposes of comparison, we included 20 subjects (9 men, 11 women; age, 40.8 ± 12.9 years; BMI, 35.8 ± 8.4 kg/m2) with no history of liver disease who were selected from patients who underwent surgery for non–liver-related reasons between January 2012 and March 2012. All of these subjects underwent an intraoperative liver biopsy that revealed no evidence of hepatic injury; their work-up included physical examination and blood sampling to assess biochemistry and serum 25(OH)D3 levels.
Inclusion criteria for the patients and the comparison group were signed informed consent, liver biopsy performed during the winter, and availability of complete clinical data, aliquots of serum, and paraffin-embedded liver tissue from biopsy specimens. Exclusion criteria were history of current or past excessive alcohol intake (defined as an average intake of >30 g/day in men and >20 g/day in women), advanced liver cirrhosis (Child-Pugh class B and C), other causes of liver disease, presence or history of cancer, inflammatory bowel disease, treatment with drugs affecting vitamin D3 metabolism (including multivitamin supplements), and use of vitamin D– and/or calcium-fortified foods and drugs known to cause liver steatosis (e.g., corticosteroids, estrogens, methotrexate, tetracycline, calcium channel blockers, or amiodarone). Metabolic syndrome was defined according to modified National Cholesterol Education Program Adult Treatment Panel III criteria24 and diabetes mellitus according to American Diabetes Association 2009 criteria.25
The study protocol was reviewed and approved by the local Ethics Committees and conducted in conformance with the Helsinki Declaration.
Clinical and Laboratory Assessment.
All subjects underwent a complete work-up, including medical history, clinical examination, anthropometric measurements, laboratory tests, and liver biopsy. A 12-hour overnight fasting blood sample was obtained on the morning of the liver biopsy in all subjects to assess fasting blood glucose (mg/dL), total cholesterol (mg/dL), high-density lipoprotein cholesterol (mg/dL), triglycerides (mg/dL), aspartate aminotransferase (IU/L), alanine aminotransferase (IU/L), gamma-glutamyl transpeptidase (IU/L), alkaline phosphatase (IU/L), blood urea nitrogen (mg/dL), creatinine (mg/dL), serum calcium (mg/dL), and phosphorus (mg/dL).
The following laboratory tests were performed to rule out other causes of liver disease: hepatitis B surface antigen (HBsAg) positivity in patients with CHC; HBsAg and anti–hepatitis C virus (HCV) positivity in patients with NAFLD; and anti-HIV positivity, anti-nuclear antibody titer ≥1:80, anti–smooth muscle antibody titer ≥1:40, anti-mitochondrial antibody at any titer, reduced ceruloplasmin or α1-antitrypsin, and transferrin saturation ≥45%, in both groups.
Biochemical assessments were performed using standard laboratory methods. Insulin (μU/mL) was measured via radioimmunoassay (ADVIA Insulin Ready Pack 100; Bayer Diagnostics, Milan, Italy), with intra- and interassay coefficients of variation <5%.
Plasma adiponectin concentrations were measured using an RIA kit (reference range, 1.5-100 ng/mL; Linco Research, St. Louis, MO) with intra- and interassay coefficients of variation of 4.5% and 3%, respectively. The degree of insulin resistance was estimated by means of homeostasis model assessment of insulin resistance (HOMA-IR).
Vitamin D status in our population was evaluated measuring serum 25(OH)D3, the most stable circulating form of this molecule.26 25(OH)D3 (nmol/L) was measured by a validated colorimetric method (LAISON, DiaSorin) on sera frozen immediately after separation and stored at −25°C for less than two months.
Liver Biopsy and Histology.
Liver biopsies undertaken for clinical purposes were obtained via percutaneous echo-assisted method by the same expert hepatologist. Subjects in the comparison group underwent intraoperative liver biopsy during surgery. Liver fragments were fixed in buffered formalin for 2-4 hours and embedded in paraffin with a melting point of 55°C-57°C. Three- to 4-μm sections were cut and stained with hematoxylin and eosin and Masson's trichrome stains. A single pathologist blinded to each patient's identity, history, and biochemistry read all of the slides. A minimum biopsy specimen length of 15 mm or at least the presence of 10 complete portal tracts was required.27 Liver biopsy samples were classified according to the presence of NASH by Brunt definition.28 NAFLD activity score (NAS)29 and a staging score based on the location and extent of fibrosis quantified according to Brunt diagnostic criteria were applied to provide numerical pathological scores for correlation purpose.
Necroinflammatory grading and fibrosis staging in patients with CHC were quantified using the Ishak scoring system.30
For immunohistochemical studies, the sections were mounted on glass slides coated with 0.1% poly(L-lysine). Each case was analyzed via immunohistochemistry for VDR, CYP2R1, and CYP27A1. After deparaffination and subsequent blockage of the endogenous peroxidase activity via incubation in 2.5% methanolic hydrogen peroxide (30 minutes), the endogenous biotin was blocked by the Biotin Blocking System (Dako, Milan, Italy) according to the manufacturer's instructions. The sections were then washed three times in phosphate-buffered saline. VDR (vitamin D receptor antibody, clone D-6, Santa Cruz Biotechnology, Inc.), CYP2R1 (CYP2R1 antibody, AbCam), and CYP27A1 (CYP27A1 antibody, AbCam) cellular expression was evaluated via immunohistochemistry. Anti-VDR antibody (diluted 1:100), anti-CYP2R1 antibody (diluted 1:50), and anti-CYP27A1 antibody (diluted 1:400) were used as primary antibodies. The sections were incubated for 1 hour at room temperature with anti-VDR and anti-CYP27A1, and overnight with anti-CYP2R1, after thermo-induced unmasking (98°C) with citrate buffer at pH 6.0 for 30 minutes. After three washes in phosphate-buffered saline, sections were incubated for 30 minutes with the appropriate biotinylated secondary antibody (labeled streptavidin-biotin, Dako). Negative controls were incubated with normal mouse antiserum instead of the primary antibody, which uniformly demonstrated no reaction. The sections were developed with 3,3-diaminobenzidine and counterstained with hematoxylin.
VDR expression has been tested and quantified on several hepatic cell lines in which the positivity was detected in the nucleus and the cytosol. For cholangiocytes evaluation, the percentage of VDR-positive (VDR+) cells was calculated among the total cholangiocytes present in the portal tracts. For hepatocytes and inflammatory cells, the VDR positivity was quantified by means of a semiquantitative scoring system according to the intensity of the staining (0: absence; 1: mild; 2: moderate; 3: intense positivity for VDR).
CYP2R1 and CYP27A1 immunohistochemical expression was evaluated in hepatocyte cytosol and quantified by means of a semiquantitative scoring system according to the intensity of the staining (0, absence of reactivity; 1, mild; 2, moderate; 3, intense reactivity).
SPSS version 17 was used to perform statistical analyses. Continuous variables are reported as the mean ± SD, and categorical variables are reported as percentages. Histological parameters are expressed by ordinal scales (NASH, NAS 0-8; CHC, staging 0-6, grading 0-3) according to the pertinent histological scoring systems.28-30 Comparisons between two different groups were performed using chi-square and Mann-Whitney tests for dichotomic and continuous variables, respectively. The association between clinical-biochemical parameters and liver histology was assessed by Spearman's rho correlation coefficient. Multivariate linear regression analysis was used to examine the relationships after adjusting for covariates and identify the best independent correlates of VDR expression in NASH patients and comparison group. Ordinal regression was used to detect the association between clinical and biochemical variables and the presence and degree of VDR, CYP2R1, and CYP27A1 cell expressions (0, absence; 1, mild; 2, moderate; 3, intense). P < 0.05 was considered statistically significant.
Clinical and biochemical characteristics of the study population according to the etiology of liver disease are shown in Table 1, along with the description of the comparison group.
|Number of subjects||25||36||20|
|Age, years||48.7 ± 13.3||55.4 ± 12.4||40.8 ± 12.9|
|BMI, kg/m2||30.5 ± 5.5||21.6 ± 3.4||35.8 ± 8.4|
|AST, IU/L||51.1 ± 23.5||54.2 ± 43.9||26.8 ± 12|
|ALT, UI/L||87.5 ± 46.6||88.7 ± 80.4||39.7 ± 24|
|GGT, UI/L||125.8 ± 96.9||52.3 ± 41.5||59.1 ± 40.2|
|FBG, mg/dL||106.3 ± 26.2||91 ± 14.14||93.6 ± 9|
|Total cholesterol, mg/dL||202.9 ± 40.9||168.8 ± 45.9||217 ± 45|
|HDL, mg/dL||48.4 ± 12.4||50.6 ± 15.2||46.2 ± 12.9|
|LDL, mg/dL||120.6 ± 34||100.3 ± 39.7||126 ± 32.7|
|Triglycerides, mg/dL||169.5 ± 80.9||93.4 ± 29.6||161.4 ± 76.2|
|25(OH)D3, nmol/L||54.7 ± 30.7||50.75 ± 26.7||52.9 ± 11.02|
More than 90% of cholangiocytes from subjects without liver disease expressed VDR. 85% of subjects from this comparison group had a degree of VDR expression ≥2 on hepatocytes, and more than 60% of this population had a degree of CYP2R1 and CYP27A1 ≥2. Table 2 shows the results of VDR, CYP2R1, and CYP27A1 expression in the liver from both patients and comparison subjects. The percentage of VDR+ cholangiocytes was significantly lower than that observed in subjects without liver disease (63.6 ± 26.7 % versus 92.2 ± 7.4%; P < 0.001).
|NASH (n = 25)||CHC (n = 36)||Controls (n = 20)||P*||P†|
|VDR+ cholangiocytes, %, mean ± SD||63.6 ± 26.7||56.6 ± 29.9||92.2 ± 7.4||<0.0001||<0.001|
|Degree of VDR+ ≥2 in hepatocytes||47%||43%||85%||0.01||NS|
|Degree of CYP2R1+ ≥2 in hepatocytes||40%||41%||61%||NS||NS|
|Degree of CYP27A1+ ≥2 on hepatocytes||48%||27%||72%||NS||<0.001|
NASH subjects had a mean serum 25(OH)D3 concentration of 54.7 ± 30.7 nmol/L, consistent with those reported in other studies.1, 2 Serum 25(OH)D3 levels were inversely correlated with intrahepatocyte ballooning (Spearman's coefficient, −0.84; P = 0.005). In this population, VDR was expressed on cholangiocytes, hepatocytes, and inflammatory cells, both in the cytosol and in nuclei, and its expression did not correlate with serum 25(OH)D3. The percentage of VDR+ cholangiocytes was significantly lower than that observed in the comparison group (63.6 ± 26.7% versus 92.2 ± 7.4%; P < 0.001) and was inversely associated with severity of steatosis (Spearman's coefficient, −0.6; P = 0.02), lobular inflammation (Spearman's coefficient, −0.6; P = 0.01) and, subsequently, NAS (Spearman's coefficient, −0.54; P = 0.03) (Fig. 1A,B). Similarly, the degree of VDR positivity on hepatocytes was lower than that seen in comparison subjects (Spearman's coefficient, −0.6; P = 0.01) and inversely associated with NAS (Spearman's coefficient, −0.56; P = 0.03) but did not correlate with any clinical and biochemical parameters.
Moderate to intense expression of CYP2R1 and CYP27A1 on hepatocytes was found in 40% and 48% of NASH patients (Fig. 2), respectively, which did not correlate with serum 25(OH)D3 levels or liver histology. CYP2R1 and CYP27A1 expression in NASH patients was lower than in the comparison group (61% and 72%, respectively) but these differences did not reach statistical significance.
To confirm the presence of an independent association between impaired liver VDR expression and the diagnosis of NASH, we compared clinical and metabolic parameters of patients affected by NASH with those obtained in the comparison subjects. The two groups were comparable for sex, age, and BMI. In all these subjects, we measured fasting blood glucose (mg/dL), basal insulin (μU/mL), and adiponectin, and calculated the HOMA-IR.
No significant differences were found between metabolic parameters of NASH patients and the comparison group [25(OH)D3, 54.7 ± 30.7 nmol/L versus 52.9 ± 11.02 nmol/L, P value not significant; basal insulin, 18.4 ± 11 μU/mL versus 19.8 ± 12.4 μU/mL, P value not significant; HOMA-IR, 5.5 ± 4.7 versus 3.87 ± 2.6, P value not significant; adiponectin (p = 0.25), 16.03 ± 8.7 ng/mL versus 17.8 ± 13.3 ng/mL], excluding fasting blood glucose (106.3 ± 26.2 mg/dL versus 93.6 ± 9 mg/dL, P = 0.038). The bivariate correlation analysis showed the presence of a negative association between VDR expression in cholangiocytes and in hepatocytes and the diagnosis of NASH (Spearman's coefficient, −0.5, P = 0.001, and Spearman's coefficient, −0.4, P = 0.01, respectively), whereas no correlation was found between liver VDR positivity and basal insulin, HOMA-IR, adiponectin, or BMI. Stepwise multiple regression analysis considering liver VDR expression as dependent variable and sex, age, BMI, HOMA-IR, adiponectin, and the diagnosis of NASH as independent variables confirmed the presence of a strong association between low liver VDR expression and the diagnosis of NASH independently from all other metabolic parameters (unstandardized β coefficient, −18.7; standardized β coefficient, 0.51; P = 0.027).
In the CHC population mean serum 25(OH)D3 levels were comparable with those observed in patients affected by NASH (50.75 ± 26.7 versus 54.7 ± 30.7 nmol/L; P value not significant) and correlated with CYP2R1 expression on hepatocytes (Spearman's coefficient, 0.50; P = 0.02). VDR expression was found in both the nuclei and cytosol of cholangiocytes and hepatocytes and was strongly associated with hepatic reactivity for CYP27A1 and CYP2R1 (Table 3), but did not correlate with either serum 25(OH)D3 levels or other clinical and biochemical parameters. The degree of VDR expression on cholangiocytes and the expression of CYP27A1 were significantly reduced compared with those observed in the comparison group (Spearman's coefficient, −0.69, P < 0.001, and Spearman's coefficient, −0.5, P < 0.001, respectively), as shown in Table 2.
|Parameter||VDR+ Hepatocytes||VDR+ Cholangiocytes|
|Spearman's rho||P||Spearman's rho||P|
|VDR+ inflammatory cells||0.43||0.01||0.23||NS|
Interestingly, the portal inflammation was inversely correlated with VDR positivity on inflammatory cells (Spearman's coefficient, −0.55; P < 0.009), and on hepatocytes (Spearman's coefficient, −0.43; P < 0.03), according to the relevant section in the Ishak score (Fig. 1C,D).
Moreover, the presence of higher VDR expression on inflammatory cells was significantly associated with increased CYP2R1 hepatic expression (Spearman's coefficient, 0.49; P < 0.005).
In the CHC population, fibrosis staging was associated with male sex, serum aminotransferases, and low hepatic CYP27A1 expression, as shown in Table 4.
|Scoring (Ishak System)|
Figure 3 summarizes the pathogenesis of liver injury over the course of different hepatopathies together with the hypothetic role played by VDR.
This study demonstrates for the first time the presence of VDR expression in liver biopsies from patients with NASH or CHC. Although VDR is known to be a nuclear receptor, VDR expression in our study was detected both in the nucleus and in the cytoplasm of liver cells. The explanation for this reactivity pattern has been provided by experimental data showing that VDR translocates from the cytoplasm into the nucleus after binding its specific ligand,31 where it constitutes a heterodimer with the retinoid X receptor α and binds specific DNA response elements before exerting its biological activity.23
In addition, serum 25(OH)D3 levels correlated inversely with hepatocyte damage, as expressed by cellular ballooning, in patients affected by biopsy-proven NASH. In these patients, VDR expression on cholangiocytes and hepatocytes was significantly lower than that observed in the comparison group without liver disease and was negatively associated with a more severe NAS. Furthermore, the multivariate regression analysis confirmed the presence of a strong association between reduced liver VDR expression and the diagnosis of NASH independently from other metabolic determinants, such as BMI, insulin resistance, and adiponectin.
Nonetheless, 25-hydroxylase expression, although low compared with subjects without liver disease, was relatively well preserved and did not affect serum 25(OH)D3 concentrations.
On the other hand, a significant association between hepatic VDR, CYP27A1, and CYP2R1 expression was found in our population affected by CHC. Expression of CYP2R1 but not CYP27A1 in the liver correlated with 25(OH)D3 levels and VDR expression on the inflammatory infiltrate. The finding of an association between serum 25(OH)D3 levels and CYP2R1 expression in CHC patients is in line with several genetic studies that have identified CYP2R1 as having a key role in vitamin D 25-hydroxylation.12, 32
In addition, liver CYP27A1 expression was significantly reduced in CHC patients compared with subjects not affected by liver disease, and was inversely associated with the fibrosis stage in CHC patients as reported,4 but did not affect serum 25(OH)D3 levels.
Our study also demonstrated for the first time that low VDR reactivity in liver infiltrating inflammatory cells is closely associated with worst inflammatory grading in patients affected by CHC, independent of serum 25(OH)D3 concentrations.
In the present study, we observed strong VDR expression in cholangiocytes of subjects without liver disease, which was partially reduced in both NASH and CHC patients. Several authors have demonstrated that lithocholic acid is a physiologic ligand of VDR33 and modulates bile acid detoxification. Han et al.22 identified VDR protein and messenger RNA in primary cultures of human hepatocytes and demonstrated that this receptor plays a critical role in the inhibition of the synthesis of bile acids, protecting the hepatocytes from cholestatic injury. VDR can be activated by either lithocholic acid acetate or 1α,25(OH)2D3 and exerts its activity through the transcriptional inhibition of CYP7A1, the initial and rate-limiting enzyme of bile acid synthesis, reducing the synthesis of bile acids in human hepatocytes.21 Interestingly, in NASH patients, we found that VDR expression on cholangiocytes was inversely associated with NAS, suggesting a possible role of VDR, expressed on biliary cells, in modulating the inflammatory process in course of liver disease. Studies in animal models and in patients with biliary disorders and CHC have shown that the ductal epithelium can express several profibrogenic and chemotactic proteins, the latter capable of attracting and activating inflammatory and fibrogenic cells.34-36
In this study, we demonstrated that liver expression of both CYP2R1 and CYP27A1 is preserved in NASH patients. This observation may question the hypothesis of a loss of hydroxylation capacity of hepatocytes in the course of NASH. Conversely, low 25(OH)D3 levels could favor, along with known risk factors, the intrahepatic accumulation of lipids, insulin resistance, progressive hepatic steatosis, and the development of steatohepatitis.
Overall, the present study suggests that vitamin D may influence the inflammatory response to chronic liver injury both in NASH and in CHC patients by means of its specific VDR, widely expressed on hepatic cell lines. In addition to the immunomodulator and antiproliferative activities on inflammatory cells, it is plausible to hypothesize that vitamin D exerts its action on cholangiocytes, in which the expression of VDR is particularly pronounced. Low hepatic VDR expression, closely associated with more severe liver histology in this study, could represent the primary event leading to progression of hepatitis. VDR polymorphisms have been investigated in the context of chronic liver diseases such as primary biliary cirrhosis and autoimmune hepatitis, where they seem to contribute to the risk of liver disease development.16, 17 Indeed, because serum 25(OH)D3 levels in our population of NASH patients are comparable to those observed in obese subjects without liver disease, it is plausible that VDR polymorphisms affecting liver VDR expression may play a role in the development and progression of NASH independently from serum vitamin D status. Nonetheless, reduced VDR expression in patients with a worse degree of NASH may also represent a consequence of the underlying liver disease, which may impair cellular metabolism and protein synthesis/expression.
In conclusion, although the cross-sectional design of this study does not allow us to establish a causal relationship, our data suggest that the vitamin D–VDR system may play an important role in the response of the liver to chronic damage induced by different pathogenic stimuli. Longitudinal studies are warranted to investigate the role of hypovitaminosis D and/or its supplementation in the pathogenesis, progression, and prognosis of chronic metabolic and viral liver disease.
We thank Professor Edwin Gale for critically revising the manuscript.