Clinical Endocrinology

Corticosteroid-binding globulin gene polymorphisms: clinical implications and links to idiopathic chronic fatigue disorders

Authors


David Torpy, Endocrine and Metabolic Unit, and the Hanson Institute, Royal Adelaide Hospital, Adelaide, South Australia, Australia. E-mail: dtorpy@mail.rah.sa.gov.au

Summary

Corticosteroid-binding globulin (CBG) binds cortisol with high affinity, facilitating transport of cortisol in blood, although tissue-specific CBG–cortisol interactions have long been postulated. There are three heritable, human CBG gene mutations that can reduce CBG–cortisol binding affinity and/or reduce circulating CBG levels. In some families, fatigue and low blood pressure have been associated with affinity altering or CBG level reducing mutations. The limited numbers of reports raise the possibility of ascertainment bias as many cases presented with features suggesting cortisol deficiency. The recent description of a genetically CBG-deficient mouse listed fatigue, manifest as reduced activity levels, as part of the phenotype, which also included immune aberrations. Severe CBG mutations may produce fatigue, but one study suggests that these are a rare cause of idiopathic fatigue. A mechanism for the effect of CBG mutations on fatigue is not readily apparent because free cortisol levels are normal, although we speculate that CBG may have an effect on cortisol–brain transport.

Introduction

CBG is the 50–60-kDa high-affinity plasma transport glycoprotein for cortisol.1 It is secreted from hepatocytes as a 383-amino-acid peptide after cleavage of a 22-amino-acid signal peptide and circulates at concentrations ranging from 30 to 52 pg/ml. Each CBG molecule contains five active glycosylation sites. Variable bi- and triantennary glycosylation leads to CBG size (neutral sugars) and charge (sialic acid) heterogeneity. Glycosylation at Asp238 appears to be necessary for the adoption of a tertiary structure and creation of an active steroid binding site.2 However, near-complete deglycosylation does not alter steroid binding.1 Each CBG molecule has a single steroid binding site involving the Trp371 residue as demonstrated by photoaffinity-labelling and site-directed mutagenesis (Fig. 1).3 CBG glycosylation shifts to a predominantly triantennary pattern in pregnancy with increased sialic acid groups, slowing clearance through the liver sialo-glycoprotein receptor, contributing to the two- to threefold rise in circulating CBG of pregnancy.4 It has been proposed that nonlectin binding pregnancy-specific CBG glycoforms may facilitate transplacental cortisol transport based on glycoform-specific enhanced syncytiotrophoblast membrane binding.5 It is not known whether the glycosylation changes in pregnancy are equivalent to those resulting from exogenous oestrogen.

Figure 1.

A protein model for CBG was developed by Dr Christopher J. Bagley (Hanson Institute) using the Homology module of the InsightII software suite (Accelrys, San Diego, CA). The structures of antichymotrypsin (PDB code) and antitrypsin (PDB code) were used as templates for the model, which was refined by energy minimization using the Discover module. The cartoon was prepared using MOLSCRIPT.54 The backbone trace of the structure is shown with the ribbon form used to denote helices and beta strands. The site of cleavage of the activation loop is shown as an arrowhead. Residues that are mutated in natural human variants (Ser224, Asp376 and Leu93) or thought to participate in binding steroids (Trp371 and Cys228) are shown as CPK spheres. A ring denotes the approximate region of cortisol binding.

CBG is a member of the serine protease inhibitor (SERPIN) structural family although it lacks intrinsic serine protease inhibitor activity. The 19-kb five-exon (four-coding) gene is located at 14q32·1, among several contiguous highly homologous genes that are thought to be derived from a common ancestor gene.6,7 The gene is also expressed in the kidney, placenta and pancreas.1,8 Levels of CBG are increased by oestrogen and pregnancy and decreased by insulin, or by glucocorticoids, such as prednisolone, or in Cushing's syndrome.1,10 The effects of glucocorticoids on CBG synthesis are glucocorticoid receptor dependent and those of oestradiol (and mitotane) are oestrogen receptor α dependent.11,12 High-affinity CBG–cortisol binding is saturated beyond cortisol levels of 400–500 nmol/l, hence the free cortisol levels rise exponentially at high cortisol concentrations.13 Under normal circadian conditions approximately 80% of cortisol is bound to CBG, 10–15% is bound to low-affinity albumin and 5–8% of circulating cortisol is unbound or free.14 Currently, only the free or unbound cortisol fraction is thought to be biologically active. Under conditions of stress, elevated cortisol levels saturate available CBG and increase the free cortisol fraction to above 20%.15

CBG is traditionally considered to have a transport role, distributing water-insoluble cortisol throughout the circulation, a buffer role perhaps blunting elevations of free cortisol during a secretory surge, or a reservoir function acting as a pool of cortisol during times of reduced cortisol secretion. Two lines of evidence have suggested a specific-tissue cortisol delivery role for CBG. These include the finding of a specific CBG interaction with human leucocyte elastase (HLE) and evidence for the presence of CBG receptors. HLE specifically cleaves CBG at the 344–345 residue leading to loss of a 39-amino-acid C-terminal fragment and almost complete loss of CBG–cortisol binding affinity.16 Hence, CBG may play a role in delivering cortisol specifically to inflammatory sites.17 One report of specific, saturable, high-affinity CBG binding sites in prostate tissue suggested the presence of a cell-surface CBG receptor, but no CBG receptor has been cloned.18 Recently, endocytic uptake of SHBG–sex steroid complexes by megalin, a low density lipoprotein (LDL) receptor analogue, has been described.19 It is not known whether a similar mechanism applies to CBG–cortisol. If present, such a receptor may play a role in tissue-specific CBG–cortisol delivery. Immunocytochemical detection of CBG has been reported in pituicytes and costaining with ACTH suggests that CBG is present selectively in corticotrophs.20 Reports of intracellular localization of CBG and differential expression of CBG mRNA according to gestational stage in various tissues have led to the proposition that CBG may play a role in modulating access of cortisol to the glucocorticoid receptor, thereby altering tissue glucocorticoid sensitivity.20,21

CBG participates actively in the stress response. Immune activation liberates interleukin-6 (IL-6), which stimulates cortisol secretion through activation of the hypothalamic CRH neurone.22 Concomitant direct inhibition of CBG gene transcription by IL-6 augments the stress response by increasing the free cortisol fraction, thereby increasing circulating glucocorticoid activity.23,24 In vivo, exogenous IL-6 reduces CBG concentrations by 50% in humans, and specific illnesses such as burns, sepsis and cardiac surgery are associated with similar drops in CBG levels, which correlate with the extent of IL-6 elevation.15,24–26 In post-traumatic stress disorder, a condition associated with low cortisol levels and elevated catecholamines, CBG levels have been shown to be elevated.27

Genetic variants of CBG

Three genetic variants of CBG with definite effects on CBG function have been described: CBG Leuven (1982), CBG Lyon (2000), and CBG Null (2001) (Fig. 2). There is a common CBG polymorphism, CBG Ser/Ala, that may affect CBG levels slightly,28 and a general heritability of CBG levels.29 A CBG intron 2 variant (2005) with as yet uncertain molecular effects may influence the risk of obesity.

Figure 2.

Schematic diagram of the CBG gene with the location of mutations indicated. Darker shaded regions of the exons are untranslated. The mutations include CBG null (G121A; Trp-12X), transcortin Leuven (T433A; Leu93His), CBG Lyon (G1254A; Asp367Asn) and Ser/Ala224 polymorphism (G825T).

CBG Leuven

The first described CBG genetic variant, Leuven, was detected in seven members of a Polish kindred (two homozygous, five heterozygous) after screening 500 random blood donors using isoelectric focusing.30 CBG Leuven results from an exon 2 point mutation (T433A, Leu93-His) that reduces CBG–cortisol binding threefold. There are no apparent clinical implications of CBG Leuven.31

CBG Lyon

Now described in three kindreds, CBG Lyon results from a single base exon 5 point substitution and a single amino acid change (G1254A, Asp367–Asn). This mutation is close to the Trp371 cortisol binding site. Binding affinity for [3H]cortisol in serum free medium from Chinese hamster ovary (CHO) cells stably transfected with wild-type CBG Lyon DNA revealed that the Asn367 substitution reduces CBG–cortisol affinity fourfold (Ka = 0·15 vs. normal Ka = 0·66 l/nmol).32 In Lyon heterozygotes, there is an unexplained 20% reduction in immunoreactive (IR)-CBG levels, but there was a greater reduction in total cortisol levels (40%) due to low cortisol binding affinity.32 Low IR-CBG Lyon levels may be due to reduced CBG synthesis, or increased CBG degradation. The three families described include a Moroccan-French kindred (one homozygous, four heterozygous), a Brazilian kindred with ancestors from the Middle East (two homozygous, eight heterozygous) and an Australian kindred with ancestors from the Southern Italian region of Calabria (five heterozygous, of which two were also heterozygous for the null mutation). The Moroccan-French kindred were discovered after investigation for possible adrenal insufficiency in an index homozygous subject presenting with fatigue, depressed mood and low blood pressure (BP).32 The four heterozygous offspring were said to be healthy apart from obesity in the eldest (aged 21). The Brazilian kindred was discovered after detection of low cortisol levels in a proband presenting with fatigue, but fatigue was not present in any other affected family members and BP was normal in all individuals.33

Early reports of absent CBG

A large kindred where many members had a 50% reduction in cortisol binding capacity was reported in 1967.34 No phenotype or genetic basis was described. However, cortisol binding capacity assays cannot distinguish low CBG concentrations from reduced cortisol–CBG binding. A report of apparent complete lack of CBG in a 15-year-old boy with obesity and a sister with a normal appearance, and low levels in two parents, probably represented a genetic disorder of CBG synthesis. Again, this report was based on a cortisol binding capacity assay.35

Hence, until recently, no proven state of complete CBG deficiency had been described. A search for absent CBG in 10 124 random sera using a precipitin test found no subject without CBG.36 As complete CBG deficiency was never described, it had been thought that complete CBG deficiency may be incompatible with life for developmental reasons, as CBG is expressed selectively at precise epochs of embryonic development, perhaps modulating tissue glucocorticoid responses.1,21

CBG Null

In 2001, we reported a 39-member Italian-Australian family, including 21 heterozygotes and three homozygotes for the first-described complete loss-of-function (null) CBG gene mutation (Fig. 3).37 The exon 2 null mutation leads to a premature termination codon corresponding to residue –12 of the pro-CBG molecule (G121A). Surprisingly, the family also had five members with the previously reported CBG Lyon mutation. The null mutation leads to 50% reduced or undetectable CBG levels, in heterozygotes or homozygotes, respectively, whereas Lyon mutation CBG levels were within the reference range. The index case was investigated because of unexplained fatigue and low BP, suggesting glucocorticoid deficiency, and the finding of low plasma but normal urine cortisol values, suggesting CBG deficiency. To allow for age and gender effects on BP, measurements were related to normative data from a reference population and expressed as a z-score.38 Among 19 adults with the null mutation, the systolic BP z-score was 12·1 ± 3·5; 11 of 19 subjects (54%) had a systolic BP below the third percentile (e.g. systolic BP < 95 mmHg in females younger than 24 years or < 104 mmHg in males over 65 years). The mean diastolic BP z-score was 18·1 ± 3·4; eight of 19 (42%) had a diastolic BP z-score below 10. Idiopathic chronic fatigue was present in 12/14 adult null heterozygote subjects (86%) and in 2/3 null homozygotes. Fatigue questionnaires revealed scores of 25·1 ± 2·5 in 18 adults with the mutation vs. 4·2 ± 1·5 in 23 healthy controls (P < 0·0001, Fig. 4). In CBG-deficient individuals, free cortisol levels were normal and displayed normal circadian variation, confirming that it is free rather than total cortisol levels that are regulated by the hypothalamus–pituitary unit.14

Figure 3.

Pedigree of kindred with both CBG null and Lyon mutations segregating independently. Of those individuals with the CBG null mutation, three individuals were homozygous and 21 were heterozygous. Five individuals were heterozygous for the CBG Lyon mutation, two of whom were also heterozygous for the null mutation.

Figure 4.

Fatigue scores of individual subjects obtained from a questionnaire to adults with CBG null and/or Lyon mutations (n = 18) and healthy age-matched controls (n = 23). Inset: the mean ± SE fatigue score was 25.1 ± 2.5 among mutation subjects vs. 4.2 ± 1.5 in controls (P < 0.0001).

CBG mutations and chronic fatigue syndrome (CFS)

There are a number of potential links between the hypothalamic-pituitary-adrenal (HPA) axis and the idiopathic chronic fatigue states, such as chronic fatigue syndrome (CFS), idiopathic chronic fatigue and fibromyalgia. These links include the reported association between the CBG Lyon and Null mutations and chronic fatigue. Basal and stimulated cortisol levels in blood and urine also tend to be slightly lower in chronic fatigue states; typically one-third to one-half of patients have low cortisol levels (below 2 SD).39,40 In addition, family and twin studies suggest that CFS may be heritable.41–43 Apart from genetic CBG variants, described later, the only specific genetic link to CFS has been through isolated rare glucocorticoid receptor abnormalities.44,45

Hence, we hypothesized that CBG genetic variants may underlie a proportion of chronic fatigue cases. We recruited 248 CFS patients, 44% with a family history of CFS, and 248 controls (blood bank donors) for restriction enzyme analyses of the known CBG mutations, full CBG gene sequencing and measurement of CBG levels.28 The three severe CBG gene mutations were not detected in CFS patients. However, CFS patients exhibited an over-representation of the exon 3 polymorphism (G825T, Fig. 2) leading to homozygosity for the serine allele at amino acid 224 of the CBG molecule (Fig. 1), relative to controls (Table 1, χ2 = 5·31, P = 0·07, 2 df). By contrast, analysis of the two conservative single nucleotide polymorphisms, C467T and C1100CT, revealed no significant difference between CFS patients and controls, suggesting that the CBG Ser224 allele may be a direct genetic risk factor for CFS, rather than a marker for a physically linked causative mutation.28 The CBG Ser/Ala polymorphism has the potential to alter CBG function, as the amino acid change is from a hydrophobic to a hydrophilic residue, which may alter CBG structure and function. However, an in vitro study of Ser vs. Ala224 CBG from transfected CHO cells did not reveal altered CBG–cortisol binding.31 CBG and single morning cortisol levels were not different between CFS patients and controls (Table 2). Plasma CBG levels were significantly higher in controls with one serine allele and higher again in homozygotes. Among the CFS patients, this effect did not reach statistical significance.

Table 1.  Number and percentage of CBG serine/serine homozygotes, serine/alanine heterozygotes and alanine/alanine homozygotes due to the G825T polymorphism. The chronic fatigue syndrome (CFS) patients showed a trend towards an increased frequency of homozygosity for the serine/serine polymorphism, compared to controls (χ2 = 5·31, 2 df, P = 0·07). This trend was also apparent in those CFS patients with a positive family history of the disorder (χ2 = 5·14, 2 df, P = 0·08)
 Ser/SerSer/AlaAla/Ala
CFS patients (n = 248)88 (35%)111 (45%)49 (20%)
CFS, positive family history (n = 109)42 (39%) 45 (41%)22 (20%)
Controls (n = 248)66 (27%)134 (54%)48 (19%)
Table 2.  Single morning (0800–1000 h) levels of plasma cortisol (nmol/l) in the chronic fatigue (CFS) patients, and CBG (nmol/l) in controls, according to CBG genotype. The Ser/Ala224 amino acid change results from a single exon 3 base substitution (G825T). Plasma CBG levels in controls were higher in those with the serine allele
 Ser/SerSer/AlaAla/Ala
  1. CBG (controls): *Ser/Ser vs. Ala/Ala, P = 0·03; **Ser/Ala vs. Ala/Ala, P = 0·05.

CFS patients   (n = 34)   (n = 66)   (n = 23)
 Total cortisol 367 ± 38·6 386 ± 19·3 425 ± 28
 Free cortisol13·8 ± 1·917·1 ± 1·914·9 ± 1·9
 IR-CBG 849 ± 36 825 ± 17 810 ± 28
Controls   (n = 31)   (n = 56)   (n = 21)
 IR-CBG 862 ± 34* 792 ± 19** 763 ± 32

There was a trend towards excess numbers of patients homozygous for the Ser224 polymorphism among the CFS group, and this allele is also associated with increased CBG levels, perhaps due to slightly lower degradability in vivo or increased synthesis and secretion. The data raise the possibility that the serine CBG allele may act as a weak genetic risk factor for CFS, perhaps manifest only when mild central hypocortisolism develops. Data such as these, if confirmed, increase the evidence that the endocrine stress system may play a role in the pathogenesis of CFS.46,47

CBG, obesity and insulin resistance

Serum CBG levels are reduced in subjects with insulin resistance, possibly due to a direct effect of insulin on CBG gene transcription.48 Severe acute caloric restriction reduces CBG levels, although CBG levels in anorexia nervosa are normal.49,50 The acute studies raise the possibility that CBG levels may fall with weight loss, thereby increasing circulatory free cortisol and predisposing to weight regain.

Altered lean/fat tissue ratios have been described in pigs with CBG gene polymorphisms.51 In humans, it has been proposed that an intron 1 CBG gene polymorphism consisting of a (GTTT)n microsatellite repeat may increase glucocorticoid sensitivity when present in the heterozygous form.52 This finding relied on an enhanced correlation between dexamethasone (0·25 mg) suppressed salivary cortisol and waist–hip ratio in those with the polymorphism. If confirmed, these findings may link CBG–cortisol associations to obesity and the metabolic syndrome.

The CBG-deficient mouse model

Recently, a mouse with a knockout CBG allele in homozygous form was produced.53 The phenotype is subtle and includes reduced activity levels, consistent with the fatigue noted in some humans with null and Lyon mutations. Several other findings in this model suggest reduced cortisol–CBG activity in selected tissues, particularly reduced activity of hepatic glucocorticoid target genes and increased inflammatory mediators such as IL-1β, mononuclear cell count and granulocyte-colony stimulating factor (G-CSF). These immune effects may underlie the reduced survival of CBG null mice after bacterial lipopolysaccharide injection. In contrast to humans, however, CBG null mice have elevated ACTH and free corticosterone levels. We have not observed an increased susceptibility to infection in our human studies. The results suggest an effect of CBG or cortisol–CBG at specific tissue interfaces. These effects may be different between humans and mice, with the most readily demonstrable difference relating to ACTH and cortisol/corticosterone levels. The immune parameters shown to be elevated in the CBG null mice have not been studied in humans.

Conclusions

Recent phenotypic observations of naturally occurring CBG mutations in humans and experimentally induced mutations in mice suggest a broader role for CBG, beyond cortisol transport. Severe CBG mutations may produce fatigue, but these are probably rare in idiopathic fatigue states. The effects of severe CBG mutations and the Ser/Ala polymorphism require validation in larger groups of individuals, in the former case through founder group population studies. An understanding of the mechanisms underlying these observations will require new data on CBG–cortisol–tissue interactions.

Ancillary