Human leucocyte antigen class II association in idiopathic bronchiectasis, a disease of chronic lung infection, implicates a role for adaptive immunity
R. Boyton, MRC Clinician Scientist, Lung Immunology Group, National Heart and Lung Institute, Sir Alexander Fleming Building, South Kensington Campus, Faculty of Medicine, Imperial College, London SW7 2AZ, UK.
The aetiology of idiopathic bronchiectasis, a lung disease where chronic inflammation and bacterial infection leads to progressive lung damage, is unknown. A possible role for natural killer cells has been highlighted previously. However, a role for adaptive immunity is suggested by the presence of CD4 and CD8 T cells in diseased lung tissue. Evidence of a human leucocyte antigen (HLA) class II disease association would further implicate a role for adaptive immunity. To establish if there is any HLA association, we analysed HLA-A, HLA-B, HLA-DQA1, HLA-DQB1 and HLA-DRB1 alleles in patients with idiopathic bronchiectasis and controls. Genomic DNA from 92 adults with idiopathic bronchiectasis and 101 healthy controls was analysed by polymerase chain reaction with sequence-specific primers. We found an increase in the prevalence of HLA-DRB1*01 DQA1*01/DQB1*05 genes in idiopathic bronchiectasis; that is, the HLA-DR1, DQ5 haplotype (odds ratio 2·19, 95% confidence interval 1·15–4·16, P = 0·0152) compared with control subjects. The association with HLA-DR1, DQ5 implicates a role for CD4 T cells restricted by these molecules in susceptibility to the progressive lung damage seen in this disease. This may operate either through influencing susceptibility to specific pathogens or to self-reactivity and requires further investigation.
Bronchiectasis is defined clinically as a chronic, progressive lung disease in which long-standing airway inflammation leads to irreversible bronchial dilatation and destruction with sputum overproduction . It is a common structural end-point with several known underlying causes, including abnormal host defence (ciliary dyskinesia and common variable immunodeficiency), foreign body obstruction and genetic abnormalities such as in cystic fibrosis (CF). There is a link with autoimmunity, as bronchiectasis features in a subset of patients with rheumatoid arthritis and inflammatory bowel disease .
A diagnosis of idiopathic bronchiectasis is made when there is bilateral, predominantly lower-lobe disease and chronic rhinosinusitis, but no known underlying cause . The aetiology of idiopathic bronchiectasis is unknown. Progressive lung damage is thought to result from a cycle of recurrent bacterial infection and a poorly regulated inflammatory response. Recent immunogenetic evidence supports a hypothesis that there may be a link between the level of natural killer (NK) cell activation and disease susceptibility; that is, a predisposing role for innate mechanisms in susceptibility [4,5]. This hypothesis is based partly on the involvement of bronchiectasis in the lungs of patients with transporter for antigen presentation deficiency syndrome, a rare disease in which impaired human leucocyte antigen (HLA) class I expression encompasses dysregulated NK and γδ cells , and inferred following combinational analysis of HLA-C/killer immunoglobulin-like receptors (KIR) genotypes in patients with idiopathic bronchiectasis. High-resolution HLA-C and KIR analysis shows that HLA-Cw*03 alleles and HLA-C group 1 homozygosity is associated with an increased risk of idiopathic bronchiectasis. Further analysis of the relationship between HLA-C and KIR genes suggests a shift of balance to activatory NK cell function, as has been proposed in other diseases including autoimmune psoriasis [5,7–9].
There is evidence that adaptive immunity has a role in idiopathic bronchiectasis, as CD4 and CD8 T cells are found in diseased lung tissue [10,11]. Furthermore, patients are infected commonly with particular bacterial pathogens such as Haemophilus influenzae, H. parainfluenzae, Pseudomonas aeruginosa and Streptococcus pneumoniae. As with susceptibility to any other microbial pathogen, it can be assumed that there is likely to be immune response gene variability in terms of the ability of particular HLA class I and class II alleles to present immunodominant epitopes to T cells.
Historically, many HLA disease associations were first reported in the field of autoimmunity. However, there have been major efforts in recent years to identify HLA alleles predisposing to particular infectious diseases . A number of HLA associations are now documented, including those in malaria, hepatitis C infection, human immunodeficiency virus (HIV) progression, leprosy and tuberculosis. Susceptibility genes for bacterial infections have been rather less studied. In the case of HIV progression, the strong contributory effects both of HLA-B polymorphisms and of KIR subtypes are regarded to indicate the contributions of both innate and adaptive mechanisms of resistance [14,15].
In the context of chronic respiratory disease, there is the added complexity of trying to elucidate a genetic susceptibility not to infection with a given pathogen per se, but to an organ-specific disease phenotype that may, in some cases, encompass several pathogens. Nevertheless, HLA associations have been identified for risk of diseases such as diffuse panbronchiolitis, bacterial infection in the context of CF and pulmonary infection with Mycobacterium avium-intracellulare[16–18].
In this study, we examine HLA associations in idiopathic bronchiectasis in Caucasian British patients. We consider that the immunogenetics of adaptive immunity are likely to be informative because idiopathic bronchiectasis is ultimately a disease of the respiratory immune response, albeit a poorly regulated one, to common lung pathogens. The allele frequencies of HLA-A, HLA-B, HLA-DRB1, HLA-DQA1 and HLA-DQB1 in idiopathic bronchiectasis patients and controls are studied.
Materials and methods
Blood samples were collected from patients with idiopathic bronchiectasis attending the Royal Brompton Hospital, London, UK. Ninety-two unrelated and randomly selected white Caucasian patients (mean age 55 + 1·4 years; 32% male) with idiopathic bronchiectasis gave written, informed consent. The patients underwent a detailed protocol of investigations before a diagnosis of idiopathic bronchiectasis was made . Idiopathic bronchiectasis is diagnosed in individuals showing predominantly bilateral lower-lobe bronchiectasis on high-resolution computed tomography (CT) and chronic rhinosinusitis where no underlying cause is found. Controls consisted of 101 UK heart/lung transplant organ donors (mean age 24·8 + 1·4 years; 73% male). The majority of controls had died unexpectedly following a road traffic accident, head injury or cerebrovascular event. The study was approved by the Royal Brompton, Harefield and NHLI Ethics Committee, UK.
Diagnosis and evaluation of pulmonary disease
A detailed protocol of investigations was carried out before a diagnosis of idiopathic bronchiectasis was made. Idiopathic bronchiectasis was diagnosed in individuals showing predominantly bilateral lower-lobe bronchiectasis on high-resolution CT and chronic rhinosinusitis where no underlying cause is found. Diagnosis and pulmonary disease severity at presentation was evaluated by chest and sinus radiographs, high-resolution thin-section CT scan, respiratory function tests, blood investigations including levels of immunoglobulin G (IgG), IgM, IgA, IgE and testing for rare immunodeficiencies in selected patients where indicated clinically, aspergillus rodioallergosorbent test and precipitins, rheumatoid factor, anti-nuclear antibodies and α1-anti-proteinase, sputum microscopy, culture and sensitivities, smear and culture for acid-fast bacilli. Skin tests were carried out for aspergillus, and a sweat test proceeding to CF genotyping if abnormal. CFTR genotyoping was performed to detect the following mutations: delta F508, G551D, G542X, 621 + 1G>T, R553X, 1717-G>A, W1282X, N1303K, R117H, R1162X, R334W and 3849 + 10kbC>T. These account for about 83% of mutations in the UK Caucasian population. Nasal mucociliary clearance, exhaled nasal nitric oxide was measured, proceeding to full cilia studies where indicated. In selected patients, fibreoptic bronchoscopy, barium swallow, respiratory muscle function tests, semen analysis and tests for associated conditions were also carried out if indicated clinically. Consequently, any patients with known underlying causes of bronchiectasis such as CF, immunoglobulin deficiency and primary ciliary dyskinesia were excluded prior to a diagnosis of idiopathic bronchiectasis being made.
Analysis of HLA-A, HLA-B, HLA-DQA1, HLA-DQB1 and HLA-DRB1 alleles
Genomic DNA from all subjects and controls, extracted from peripheral blood, was HLA typed for HLA-A, HLA-B, HLA-DQA1, HLA-DQB1 and HLA-DRB1 alleles by polymerase chain reaction with sequence-specific primers using a cycler plate system (Protrans; Quest Biomedical, Solihull, UK).
Allele and genotype frequency was determined by direct counting. Statistical analyses were performed using the χ2 or Fisher's exact test. The odds ratio (OR) and 95% confidence interval (CI) were calculated. A P-value of less than or equal to 0·05 was considered significant. In addition, for each HLA class analysed one single χ2 analysis was performed encompassing all of the alleles within that class. Subsequently, post hoc cell contribution analyses were performed to determine which allele, if any, was significantly either over- or under-represented within the given class. Again, a P-value of less than or equal to 0·05 was considered significant.
Human leucocyte antigen class I-A and -B alleles in individuals with idiopathic bronchiectasis
Ninety-two patients with idiopathic bronchiectasis and 101 controls were typed for alleles of HLA-A. Only minor differences in gene frequencies were observed, none of which was statistically significant (Table 1). A single χ2 analysis was performed encompassing all the HLA-A alleles. This confirmed that there was no significant difference between any of the alleles (overall analysis P = 0·581).
Table 1. Human leucocyte antigen (HLA)-A allele frequency in subjects with idiopathic bronchiectasis and controls.
|01||A*01011-09||32 (17·4)||41 (20·5)||0·82||0·49-1·36||n.s.|
|02||A*02011-51||63 (34·2)||54 (27·0)||1·41||0·91-2·18||n.s.|
|03||A*03011-09||22 (12·0)||28 (14·0)||0·83||0·46-1·52||n.s.|
|11||A*11011-07/09/10||17 (9·2)||16 (8·0)||1·17||0·57-2·39||n.s.|
|23||A*2301-08N||3 (1·6)||4 (2·0)||0·81||0·18-3·68||n.s.|
|24||A*2402-11/13-33||14 (7·6)||19 (9·5)||0·78||0·38-1·61||n.s.|
|25||A*2501-03||2 (1·1)||3 (1·5)||0·72||0·12-4·37||n.s.|
|26||A*2601-17||1 (0·5)||5 (2·5)||0·21||0·02-1·84||n.s.|
|29||A*2901-04||5 (2·7)||8 (4·0)||0·67||0·23-2·09||n.s.|
|30||A*3001-04/06-11||1 (0·5)||5 (2·5)||0·21||0·02-1·84||n.s.|
|31||A*31012-05||5 (2·7)||2 (1·0)||2·77||0·53-14·43||n.s.|
|32||A*3201-06||7 (3·8)||8 (4·0)||0·95||0·34-2·67||n.s.|
|33||A*3301/03-06||2 (1·1)||1 (0·5)||2·19||0·20-24·32||n.s.|
|66||A*6601-04||1 (0·5)||0 (0·0)||–||–||–|
|68||A*6801-20||9 (4·9)||6 (3·0)||1·66||0·58-4·77||n.s.|
Typing for HLA-B alleles showed three different alleles with significantly altered frequencies in the disease group compared with controls (Table 2). However, because of the very large number of HLA-B alleles each present in the population at low frequency, the impact of changes to the frequency of any given allele will be small, except in rare circumstances. The idiopathic bronchiectasis group showed a raised frequency of the HLA-B60 allele (P = 0·015) (data not shown). We have reported previously that patients with idiopathic bronchiectasis show an increased frequency of HLA-Cw*03, the allele frequency in patients being 19·8% compared with 9·9% in control subjects . The increased HLA-B60 allele frequency in the disease group appears to be attributable to an extended haplotype with HLA-Cw*03. The effect of HLA-B60 appears to be secondary to the HLA-Cw*03 effect, as the allele frequency of HLA-Cw*03 in patients with idiopathic bronchiectasis is 19·8%, and HLA-B60-positive individuals account for approximately two-fifths of these. We also noted a reduced frequency of HLA-B51 and increased frequency of HLA-B52 in the patient group (P = 0·018 and P = 0·012 respectively). A single χ2 analysis was performed encompassing all the HLA-B alleles. This confirmed that there was a significant difference between the alleles (overall analysis P = 0·006). Subsequent post hoc cell contribution analyses confirmed that HLA-B51 was under-represented and HLA-B52 was over-represented in disease (P = 0·017 and P = 0·012 respectively). We found no evidence that the associations with HLA-B51 and -B52 were the consequence of an extended haplotype encompassing the HLA-Cw*03 alleles described previously in this population .
Table 2. Human leucocyte antigen (HLA)-B allele frequency in subjects with idiopathic bronchiectasis and controls.
|07||B*07021-27||23 (12·5)||23 (11·4)||1·11||0·60-2·06||n.s.|
|08||B*0801-14||23 (12·5)||26 (12·9)||0·97||0·53-1·76||n.s.|
|13||B*1301-04/06/08/09||1 (0·5)||7 (3·5)||0·15||0·02-1·25||n.s.|
|14/64/65||B*1401-06||6 (3·3)||7 (3·5)||0·94||0·31-2·85||n.s.|
|15/62||B*1501-69||17 (9·2)||13 (6·4)||1·48||0·70-3·14||n.s.|
|63||B*1516/17/67||0 (0·0)||2 (1·0)||–||–||–|
|72||B*1503/29/46/47/49/52/53/54/61/64/68||0 (0·0)||1 (0·5)||–||–||–|
|18||B*1801-13||2 (1·1)||9 (4·5)||0·24||0·05-1·11||n.s.|
|27||B*2710-23||12 (6·5)||12 (5·9)||1·10||0·48-2·52||n.s.|
|35||B*3501-09/11/14/15/17-21/23-27/29/30/32-37||14 (7·6)||18 (8·9)||0·84||0·41-1·74||n.s.|
|37||B*3701-05||1 (0·5)||5 (2·5)||0·22||0·02-1·86||n.s.|
|38||B*3801-07||1 (0·5)||2 (1·0)||0·55||0·05-6·08||n.s.|
|39||B*3901-20/22-25N||2 (1·1)||0 (0·0)||–||–||–|
|40/60/61||B*4001-1-39||20 (10·9)||11 (5·4)||2·12||0·99-4·55||P = 0·05|
|44||B*4402-27||35 (19·0)||30 (14·9)||1·35||0·79-2·30||n.s.|
|45||B*4501-05||0 (0·0)||3 (1·5)||–||–||–|
|47||B*4701-04||1 (0·5)||1 (0·5)||1·10||0·07-17·69||n.s.|
|49||B*4901-03||5 (2·7)||1 (0·5)||5·61||0·65-48·51||n.s.|
|50||B*5001/02/04||4 (2·2)||2 (1·0)||2·22||0·40-12·28||n.s.|
|51||B*5101-09/11N-24/26/27N||2 (1·1)||11 (5·4)||0·19||0·04-0·87||P = 0·018|
|52||B*52011-03||8 (4·3)||1 (0·5)||9·14||1·13-73·77||P = 0·012|
|53||B*5301/02/04-06/08||2 (1·1)||2 (1·0)||1·10||0·15-7·88||n.s.|
|55||B*5501-03/05/07-10||1 (0·5)||3 (1·5)||0·36||0·04-3·50||n.s.|
|17/57/58||B*5701-5709 B*5801/02/04-06||4 (2·2)||12 (5·9)||0·35||0·11-1·11||n.s.|
Evidence for a class II association with HLA-DR1, DQ5
Disease association studies looking at HLA class II polymorphisms are confounded by the fact that linkagedisequilibrium makes it difficult to distinguish contributions of HLA-DR and -DQ polymorphisms . Indeed, it is often virtually impossible to separate the potential functional contribution to disease of gene products from the two loci . Looking initially at HLA-DRB1 alleles, we found an over-representation of HLA-DR1 alleles in the patient group (Table 3). The HLA-DR1 allele frequency was 17·8% for patients, but only 8·9% for controls (OR = 2·22, 95% CI = 1·19–4·12, P < 0·0105). A single χ2 analysis was performed encompassing all the HLA-DRB1 alleles (overall analysis P = 0·058). Subsequent post hoc cell contribution analyses confirmed that HLA-DRB1*01-08 was over-represented in disease (P = 0·0105). Similarly, as would be expected from the association with HLA-DR1, there is an association with HLA-DQ5. This is seen most strikingly in the increased frequency of HLA-DQ1*0101/04-06 alpha chains (Table 4). The allele frequency for HLA-DQ1*0101/04-06 is higher in idiopathic bronchiectasis patients compared with controls at 35·3% for patients compared with 21·5% for controls (OR = 1·99, 95% CI = 1·27–3·14, P = 0·0026). A single χ2 analysis was performed encompassing all the HLA-DQA alleles (overall analysis P = 0·0386). Subsequent post hoc cell contribution analyses confirmed that HLA-DQA1*0101/04-06 was over-represented in disease (P = 0·0015). This is, as expected, mirrored by an increase in the patient group of the associated beta chains with more patients carrying DQ5β chains, although in this case the result does not reach statistical significance (Table 5). A single χ2 analysis was performed encompassing all the HLA-DQB alleles. This confirmed that there was no significant difference between any of the alleles (overall analysis P = 0·481). Taken together, the evidence is of increased relative risk to HLA-DR1, DQ5 individuals, with an OR of 2·19 (95% CI 1·15–4·16, P = 0·0152). There is a suggestion that this association relates primarily to a functional role for HLA-DQ molecules falling within the old DQw1 serotype, as the presence of DR15, DQ6 individuals in the patient group is also raised, although not reaching significance.
Table 3. Human leucocyte antigen (HLA)-DRB1 allele frequency in subjects with idiopathic bronchiectasis and control subjects.
|01||DRB1*0101-08||31 (17·8)||18 (8·9)||2·22||1·19-4·12||P = 0·0105|
|15||DRB1*15011-11||26 (14·9)||20 (9·9)||1·60||0·86-2·98||n.s.|
|16||DRB1*1601-05/07/08||0 (0)||4 (2·0)||–||–||–|
|17||DRB1*0301/04/06/15/16/18/19/20||21 (12·1)||31 (15·3)||0·76||0·42-1·37||n.s.|
|18||DRB1*0302/05/09/14||0 (0·0)||0 (0·0)||–||–||–|
|04||DRB1*04011-42||30 (17·2)||41 (20·3)||0·82||0·49-1·38||n.s.|
|07||DRB1*0701/03-05||19 (10·9)||27 (13·4)||0·79||0·43-1·49||n.s.|
|08||DRB1*0801-19/21-23||6 (3·4)||3 (1·5)||2·37||0·58-9·62||n.s.|
|09||DRB1*09012||2 (1·1)||3 (1·5)||0·77||0·13-4·67||n.s.|
|10||DRB1*10011/012||2 (1·1)||2 (1·0)||1·16||0·16-8·34||n.s.|
|11||DRB1*1101-42||13 (7·5)||25 (12·4)||0·57||0·28-1·16||n.s.|
|12||DRB1*12011-08||3 (1·7)||0 (0)||–||–||–|
|13||DRB1*1301-48||16 (9·2)||23 (11·4)||0·79||0·40-1·54||n.s.|
|14||DRB1*1401-39||5 (2·9)||5 (2·5)||1·17||0·33-4·10||n.s.|
Table 4. Human leucocyte antigen (HLA)-DQA1 allele frequency in subjects with idiopathic bronchiectasis and controls.
|01||DQA1*0101/04-06||65 (35·3)||43 (21·5)||1·99||1·27-3·14||P = 0·0026|
|0102||DQA1*0102||5 (2·7)||12 (6·0)||0·44||0·15-1·27||n.s.|
|0103||DQA1*0103||6 (2·3)||13 (6·5)||0·48||0·18-1·30||n.s.|
|0201||DQA1*0201||22 (12·0)||29 (14·5)||0·80||0·44-1·45||n.s.|
|03||DQA1*0301-03||34 (18·5)||43 (21·5)||0·83||0·50-1·37||n.s.|
|04||DQA1*0401||5 (2·7)||3 (1·5)||1·83||0·43-7·79||n.s.|
|05||DQA1*0501-05||40 (21·7)||55 (27·5)||0·73||0·46-1·17||n.s.|
|06||DQA1*0601||1 (0·5)||0 (0·0)||–||–||–|
Table 5. Human leucocyte antigen (HLA)-DQB1 allele frequency in subjects with idiopathic bronchiectasis and control subjects.
|05||DQB1*0501-0504||36 (19·6)||30 (15·0)||1·38||0·81-2·35||n.s.|
|06||DQB1*0601-17||46 (25·0)||41 (20·5)||1·29||0·80-2·09||n.s.|
|0301||DQB1*0301/04/09/10||38 (20·7)||43 (21·5)||0·95||0·58-1·55||n.s.|
|0302||DQB1*0302/0307/0308||14 (7·6)||23 (11·5)||0·63||0·32-1·27||n.s.|
|0303||DQB1*0303||5 (2·7)||9 (4·5)||0·59||0·20-1·80||n.s.|
|02||DQB1*0201-03||40 (21·7)||50 (25·0)||0·83||0·52-1·34||n.s.|
|04||DQB1*0401/02||5 (2·7)||3 (1·5)||1·83||0·43-7·79||n.s.|
Again, analysis of extended haplotypes yielded no evidence to suggest that the association with HLA-DR1, DQ5 was part of an extended haplotype with the HLA-Cw*03 association documented previously.
Idiopathic bronchiectasis is a chronic lung disease resulting partly from the impact of a poorly regulated respiratory immune response to common lung pathogens. In this study, we attempt to determine if idiopathic bronchiectasis is associated with any HLA class I or II alleles that would implicate a role for innate and/or adaptive immunity in disease pathogenesis. Analysis of HLA class I allelic frequencies showed a significant association of this disease with the alleles HLA-B51 and -B52. While HLA class I associations may often be markers of susceptibility to viral infection, immune responses to bacteria and mycobacteria may, similarly, also involve HLA class I-restricted CD8 T cells . The rapid diversification of HLA-B alleles during recent human evolution marks them down as key players in the development of functional immunity to common pathogens . We noted an increased frequency of HLA-B52 in the patient group (P = 0·012). HLA-B52 is considered a risk factor for Behçet's disease [23,24]. The fact that in idiopathic bronchiectasis patients there is a raised frequency of this gene may point to some related aspect of pathogenesis, as both diseases involve dysregulated inflammation and can be associated with autoimmunity . In Behçet's disease, much attention has focused on a potentially pathogenic role of autoimmunity to the stress protein Hsp60 .
Within the HLA class II region, we found an association with HLA-DR1/DQ5. It can be extremely difficult in such cases to determine whether the primary, functional association is with HLA-DR or with HLA-DQ. Certainly, HLA-DR is expressed more strongly by most antigen-presenting cells and is therefore often considered to be the functionally dominant HLA class II molecule in antigen presentation to CD4 T cells . Dissection of the relative contributions of HLA-DR and HLA-DQ expression most commonly requires the ability to study the disease trans-racially, such that there is a greater chance of being able to consider haplotypes with recombinations between the given HLA-DR and DQ alleles . Alternatively, it can be illuminating to clone the functionally implicated CD4 T cells and determine their patterns of HLA class II restriction. For example, coeliac disease is an HLA-DQ2-associated disease, the association confirmed functionally by the fact that the gluten-specific CD4 T cells are HLA-DQ2 restricted . Clearly, such studies in bronchiectasis await the culture and characterization of antigen-specific cells from the lungs of patients. Occam's razor might suggest that the HLA class II associations observed here may reflect simple immune response gene effects in the ability to mount a protective CD4 T cell response to key epitopes of H. influenzae, H. parainfluenzae, P. aeruginosa or S. pneumoniae.
How can one reconcile the finding of an HLA-DR/DQ susceptibility factor as described here, implicating an effect of CD4 T cell responses, with the earlier observations, suggesting a possible role of NK cell dysregulation? There is an increasing appreciation of the fact that, for a particular immune response in a particular organ, there may be separate, functionally linked, innate and adaptive phases [14,15]. In the respiratory immune response to respiratory syncytial virus there is evidence for a very rapid NK cell response, this in turn creating the cytokine milieu for the ensuing CD8 T cell adaptive immune response [30,31].
We have identified allelic polymorphisms for HLA-B, HLA-C and HLA-DR/DQ, all associated with statistically increased risk of idiopathic bronchiectasis. While there are clearly strong linkage disequilibrium effects in the HLA region and an extent to which disease phenotypes can be associated with extended haplotypes covering several genes, we found no evidence for an increase in an extended haplotype covering HLA-Cw3, HLA-B51 or -B52 and HLA-DR1/DQ5. Rather, we consider it likely that idiopathic bronchiectasis is a complex phenotype relating to an inability to mount an appropriately tuned respiratory immune response to a range of pathogens. Further work is required to define the mechanistic underpinnings that these susceptibility genes confer on the individual in terms of CD4 T cell responses to self- and foreign antigens.
This study was supported by the Royal Brompton and Harefield NHS Trust Clinical Research Committee, the Welton Foundation and the Medical Research Council, UK. R. J. B. is supported by a Medical Research Council Clinician Scientist Fellowship. The authors would like to thank Professor Malcolm Green, Professor Anthony Newman-Taylor, Professor Tim Evans and Professor Maggie Dallman for their support and advice.