Phenotypic Differences in Alpha 1 Antitrypsin-Deficient Sibling Pairs May Relate to Genetic Variation

Abstract

Alpha-1-antitrypsin deficiency is associated with variable development of airflow obstruction and emphysema. Index patients have greater airflow obstruction than subjects detected by screening, but it is unclear if this reflects smoking differences and/or ascertainment bias, or is due to additional genetic factors. In this study 72 sibling pairs with alpha-1-antitrypsin deficiency were compared using lung function measurements and HRCT chest. Tag single nucleotide polymorphisms to cover all common variation in four genes involved in relevant inflammatory pathways (Tumour necrosis factor alpha, Transforming growth Factor beta, Surfactant protein B and Vitamin D binding protein) were genotyped using TaqMan® technology and compared between pairs for their frequency and relationship to lung function. 63.5% of non-index siblings had airflow obstruction and 59.5% an FEV₁ < 80% predicted. Index siblings had lower FEV₁ and FEV₁/FVC ratio, a higher incidence of emphysema (all P ≤ 0.001) and lower gas transfer (P = 0.02). There was no correlation of FEV₁ between siblings but KCO was significantly correlated (r = 0.42, P = 0.002). Quantitative analyses against lung function showed that a polymorphism in Surfactant protein B was associated with FEV₁ (P = 0.002). This result was replicated in a non-sibling group (P = 0.01). Our results show that clinical differences in families with alpha-1-antitrypsin deficiency are not solely explained by smoking or ascertainment bias and may be due to variation within genes involved in inflammatory pathways.

Key words: :

• Chronic Obstructive Pulmonary Disease
• Alpha 1-Antitrypsin Deficiency
• Genetics
• Population

View correction statement:

Erratum

INTRODUCTION

Patients with alpha-1-antitrypsin deficiency (AATD) are susceptible to the development of airflow obstruction and early-onset emphysema in adulthood ([1]). The most important risk factor for the development of lung disease in AATD is active smoking ([2]) and differences in exposure to smoke may account for some of the variation in lung function in patients with this condition ([3]). Patients are primarily diagnosed with AATD after presentation to health services with respiratory symptoms. However alleles of the alpha-1-antitrypsin (AAT) gene are inherited in a co-dominant manner thus siblings of an index PiZZ case (the most common genotype giving severe deficiency) who share heterozygote parents have a 1 in 4 chance of having severe deficiency. Hence, family screening of a known index case often identifies other affected siblings. Index patients have been shown to have more severe airflow obstruction than non-index patients ([4]) although differences in these populations may be influenced by ascertainment bias and by differences in smoking. Genetic modifiers may be contributing to phenotypic differences and a recent family based study has identified some single nucleotide polymorphisms (SNPs) that may influence FEV₁([5]).

Individual sibships of PiZZ patients have been described in case reports ([6], [7]). The UK national registry, established as part of ADAPT (Antitrypsin Deficiency Assessment and Programme for Treatment), includes over 700 well-characterised PiZZ patients, including 72 sibling pairs. We have taken this opportunity to compare clinical features of index and non-index siblings and to develop a correction factor for FEV₁ based on an individual's smoking history. We have applied this correction factor to siblings to determine whether differences between them can be explained by smoking alone. Finally, we have compared the lung function of a subset of seven pairs of non-index siblings, attempting to overcome any ascertainment bias. Residual differences after correction for smoking suggest that there may be other genetic factors influencing lung function. Inflammation is more marked in AATD than usual chronic obstructive pulmonary disease (COPD) ([8]), so variation in inflammatory genes may be of greater importance. In addition, interactions between inflammatory cytokines and AAT ([9]) could amplify the genetic effect seen. We therefore chose to investigate inflammatory genes known to contain SNPs associated with usual COPD and replicated by independent groups ([10],[11],[12],[13],[14],[15],[16],[17],[18]).

MATERIAL AND METHODS

Study subjects

Seventy-two Caucasian sibling pairs with AATD of the PiZZ genotype were studied. Samples suitable for DNA extraction were available on 52 of these pairs. All patients had a serum alpha-1-antitrypsin (AAT) level of < 11 μ M, and PiZZ genotype confirmed by specific PCR (Heredilab, Salt Lake City, UT, USA). None of the subjects had ever received AAT replacement therapy. A non-sibling group of PiZZ patients from the UK national registry were selected for use as a potential replication data set for the genetic study. The family data set was chosen for the primary analysis as there would be no risk of population stratification affecting results in this group. ADAPT has ethical approval to carry out clinical and genetic studies relating to registered subjects (South Birmingham LREC, 3359 and 3359a).

Study design

This is a family-based genetic association study in PiZZ sibling pairs with AATD. Detailed description of the family cohort is given prior to the genetic analyses. Quantitative analyses of lung function parameters have been conducted in the cohort, controlling for familial effects and all covariates by use of a family-based genetic association test, in order to show if phenotypic variation is due to underlying genetic differences. Families have been genotyped for four candidate genes, using tag single nucleotide polymorphisms (SNPs). The replication data set consists of unrelated, ethnically matched PiZZ AATD subjects. The replication dataset were genotyped for SNPs associated in the sibling pairs.

Methods

Clinical Phenotyping

A full clinical history was taken, including smoking history. Patients were classified as index siblings if they were the first family member to be diagnosed with AATD and non-index if they were diagnosed through family screening. Seven families revealed more than one PiZZ sibling identified through family screening, who were analysed separately attempting to overcome any ascertainment bias.

Patients performed lung function tests and underwent CT scanning as described previously ([19]). The presence of emphysema was determined using both the visual appearance of the scan and density mask analysis of slices at the level of aortic arch, representing the upper zone, and at the inferior pulmonary vein, representing the lower zone, using a threshold of -910HU. This is the most commonly used threshold level for emphysema detection, deemed optimum by Muller et al. ([20]) and validated against physiological measures in AATD ([21]). Patients whose voxel index exceeded values seen in normal subjects in either zone ([22]) were deemed to have emphysema. Phenotypic characteristics of the siblings and replication data set are shown in Table 1.

Table 1 Clinical characteristics of the sibling pairs and replication data set

		Sibling pair data set			Replication
		Index	Non-index	P value	data set
Gender	Male (%)	59.68	53.18	ns	53.18
	Female (%)	40.32	46.82		40.09
Age (years)		47.91 (1.2)	46.00(1.2)	ns	50.01 (0.52)
Smoking (pack years)		12.50 (1–23)	8.50 (0–18.5)	ns	14.00 (2.35–25.65)
FEV1(%pred)		31.00 (21–55)	68.00 (46–107)	< 0.001	37.15 (18.30–56.04)
FEV1/FVC (%)		37.00 (30–47)	59.00 (43–75)	< 0.001	38.40 (25.95–50.85)
KCO (%pred)		69.10 (3.2)	82.70 (2.5)	0.023	69.82 (1.18)
Emphysema (%)		82.50	55.70	0.001	79.94
UZVI (% voxels < -910HU)		35.35 (23.32–47.39)	16.02 (5.32–26.72)	< 0.001	31.09 (1.05)
LZVI (% voxels < -910HU)		53.90 (40.45–67.35)	27.92 (12.80–43.05)	< 0.001	44.59 (1.21)

Data are shown as median (IQR) for non-normally distributed parameters, and mean (standard error of the mean) for those normally distributed. For emphysema the percentage affected is shown. N = 72 pairs except N = 52 pairs for CT measures. Significance differences between the index and non-index cases are shown. The replication data set did not show any significant differences with the index data set.

Genotyping

Single nucleotide polymorphisms (SNPs) to tag the genes for Transforming Growth Factor Beta (TGFβ), Surfactant Protein B (SFTPB) and Vitamin D Binding Protein (GC) were chosen using data from HapMap phase II (Jan 2007)([23]) to capture all SNPs with a minor allele frequency > 0.05 and r² at least 0.8. Tags for the Tumour Necrosis Factor Alpha (TNFα) gene were chosen using linkage disequilibrium (LD) data for UK Caucasians ([24]).

DNA extraction was performed using a modified Nucleon Bacc II kit (Tepnel Life Sciences, UK) and the concentration checked using Picogreen® (Molecular Probes Inc, UK). Genotyping was carried out using TaqMan® genotyping technologies (Applied Biosystems, UK) on an ABI7900 HT. All TaqMan® assays are pre-validated by the company, this information being available from them on request.

Statistical Analysis

Clinical data for the siblings, and all data from the replication data set, was analysed using SPSS version 12.0. Measures taken from each sibling were compared using the χ² test, Mann-Whitney U or t-test as appropriate. Spearman's coefficients were used for correlations of continuous variables between the sibling pairs. Linear regression was performed on all subjects to calculate the average effect of smoking on FEV₁. From the resultant equation, a predicted value for FEV₁ was calculated for each sibling based on their individual smoking history. The actual FEV₁ was then compared to the predicted value to calculate a residual for each patient (a low value indicating a worse FEV₁ than predicted).

Genetic effects were assessed using a family-based association test suitable for quantitative sibling pair data (ASSOC) implemented in SAGE ([25]) (see acknowledgements) accounting for gender, age and smoke exposure. The test is robust to population stratification effects, which are unlikely in this type of study design. It is a form of linear regression performed in the sibling cohort as a single group, using lung function as the outcome measure, which controls for familial relationships between siblings by use of estimated familial variance components, calculated according to the Elston-Stewart algorithm ([26]). The directionality of association is indicated by the Z statistic. For SNPs where a difference between siblings was seen the replication data set were assessed for similar phenotypic associations using regressions accounting for covariates as previously. All models assumed additive effects for the SNPs tested.

RESULTS

Patient characteristics and phenotypes

There was no significant difference in age, gender or smoking history between the index and non-index siblings. The index patients had smoked more than non-index cases, but this was not statistically significant. Comparisons of lung function between the two groups showed that index siblings had a significantly lower FEV₁, FEV₁/FVC ratio (P < 0.001) and gas transfer per unit volume (KCO) (P = 0.023). Index siblings were also more likely to have evidence of emphysema on CT scan compared to non-index siblings (P = 0.001) as shown by higher upper and lower zone voxel indices (UZVI and LZVI) in index cases (both P < 0.001). The replication data set showed no significant differences from the index siblings. These results are shown in Table 1.

Familial phenotypic correlations

There was no correlation between the FEV₁(% predicted) of the index and non-index siblings (Figure 1a), though KCO was significantly correlated (P = 0.002, Figure 1b). One patient had a markedly low gas transfer due to a combination of emphysema and thromboembolic disease but removal of this outlying subject did not alter the correlation's significance. Similarly for quantitative CT data, LZVI did not correlate between pairs (Figure 2a), but UZVI correlated well (P = 0.019, Figure 2b). Seven pairs of non-index patients showed no significant correlation for FEV₁ or KCO. CT images were not available for this subset.

Figure 1 Correlations of lung function between sibling pairs. The scatterplot in Figure 1a shows no significant correlation of FEV₁ (%predicted) between the siblings, whereas the KCO correlates well between pairs, as shown in Figure 1b.

Figure 2 Correlations of CT densitometry between sibling pairs. The scatterplot in Figure 2a shows no significant correlation between the lower zone voxel indices (LZVI) of the sibling pairs, whereas the upper zone voxel indices (UZVI) correlate well between the pairs, as shown in Figure 2b.

Linear regression showed that pack years smoked was associated with FEV₁, estimating an average loss equivalent to 0.8% of the predicted value per pack year smoked. This figure was used to calculate a predicted FEV₁ for each sibling based on their personal smoking history, in order to ascertain if sibling pair differences could be accounted for by smoking. The calculated predicted value was subtracted from the measured FEV₁ (% predicted value) to obtain a residual. Comparison of the residual figures between siblings showed no correlation (Figure 3). The same statistical process for LZVI showed a deleterious effect of smoking as before, and no correlation of the residuals (P = 0.101). KCO and UZVI remained correlated after adjustment for smoking (P = 0.035 and 0.046, respectively). A trend towards a correlation for emphysema zone (as measured by LZVI-UZVI) after adjustment for smoking was noted (P = 0.071).

Figure 3 Residual FEV₁ between sibling pairs. The scatter plot shows the residual FEV₁ value of the sibling pairs. This was calculated by subtracting the predicted FEV₁ based on their individual smoking history from the measured FEV₁ (%predicted). Negative values show a measured FEV₁ that is worse than predicted. There is no correlation between the pairs.

Genotypes

The tagging algorithms identified a total of 22 SNPs to tag SFTPB, TGFβ, GC and TNFα. In addition we chose 4 tag SNPs to cover areas adjacent to TGFβ where associations with COPD have been found previously ([15]). Other known functional variants within our candidate genes, for example the GC2 allele of GC, have not been specifically typed, as the tags chosen will pick these up within the relevant haplotype blocks. Genotyping failed in 2.25% of reactions performed. The reference sequence numbers, minor allele frequencies, Hardy-Weinberg equilibrium status and results of quantitative analyses against lung function are shown in Table 2.

Table 2 Sibling pair genotyping results

Gene	SNP	Hardy Weinberg equilibrium	MAF (%)	FEV1	P valueFEV1/FVC	KCO	Direction of association
SFTPB	rs1130866	No	48.3
	rs2118177	Yes	41.8	0.002	0.003	0.236	Increase
TGFB	rs8110090	Yes	7.4	0.117	0.186	0.329
	rs8179181	Yes	27.6	0.590	0.416	0.112
	rs4803455	Yes	44.0	0.978	0.927	0.468
	rs11466338	Yes	14.5	0.160	0.064	0.231
	rs11466345	Yes	8.0	0.613	0.595	0.828
	rs1800469	Yes	23.2	0.940	0.978	0.861
MGC4093	rs11083616	Yes	30.6	0.875	0.902	0.599
CCDC97	rs10417924	Yes	16.7	0.363	0.533	0.214
	rs8109167	Yes	16.6	0.068	0.119	0.280
	rs8109627	Yes	28.4	0.204	0.316	0.642
GC	rs222029	Yes	24.2	0.317	0.330	0.587
	rs222035	Yes	44.6	0.424	0.557	0.979
	rs1491718	Yes	17.2	0.182	0.116	0.604
	rs1352844	No	14.2
	rs2070741	Yes	11.7	0.530	0.405	0.523
	rs3755967	No	25.5
	rs3733359	Yes	13.0	0.183	0.321	0.054
	rs705117	Yes	25.0	0.265	0.284	0.934
	rs222023	Yes	7.6	0.158	0.257	0.491
	rs2298849	Yes	22.2	0.201	0.148	0.489
TNFA	rs1800629	Yes	25.0	0.980	0.991	0.735
	rs361525	Yes	8.7	0.919	0.960	0.847
	rs1799964	Yes	28.7	0.981	0.899	0.583
	rs3093662	Yes	9.2	0.746	0.840	0.709

SFTPB = surfactant protein B, TGFB = transforming growth factor beta, MGC4093 and CCDC97 = areas adjacent to TGFB coding for hypothetical proteins, GC = vitamin D binding protein, TNFA = tumour necrosis factor alpha, MAF = minor allele frequency. N = 52 pairs. Direction of association is only shown for significant results and pertains to the minor allele effect upon the associated quantitative phenotype.

*P values of < 0.1 are in bold type.

Three SNPs were not in Hardy-Weinberg equilibrium, thus were excluded from further analysis. One SNP (rs2118177) in SFTPB showed a significant difference between siblings in the analyses for FEV₁ (P = 0.002) and FEV₁/FVC (P = 0.003), such that both were higher in those with the minor allele (C). Quantitative CT data was available on 32 pairs; a SNP in TGFβ was linked to relative lower zone predominance of emphysema (rs8179181, P = 0.027), such that LZVI-UZVI was higher. No associations were seen with emphysema severity in either zone.

Having found some evidence of association in the sibling pair analysis we went on to genotype the replication data set and ascertain the odds ratio (OR) of disease conferred by associated SNPs. Since rs2118177 was associated with both FEV₁ and FEV₁/FVC in the siblings we split the replication data set into a group in whom both of these were impaired, and a group in whom they were normal—i.e., subjects with COPD (defined as FEV₁/FVC < 0.7 and FEV₁ < 80%predicted) and without, functioning as cases and controls. Since rs8179181 was associated with lower zone predominant emphysema, we compared subjects with and without this phenotype. This method was chosen as qualitative phenotypes may be more informative in AATD ([27]), provided the data set is large enough to have adequate power. In the replication data set we had greater than 80% power to detect an OR of 1.50 between the groups for rs2118177 and 1.83 for rs8179181, where α = 0.05, and the SNP predisposes to disease. This level of power would not have been achieved in the siblings. Rs2118177 was protective against COPD in minor allele homozygotes (CC) conferring an OR of 0.28 of disease (95% CI 0.11-0.73, P = 0.01), consistent with the C allele association in the siblings. The CC genotype occurred in 10.5% of subjects with COPD, and 19% of those without. Rs8179181 did not show any disease associations in the replication data set.

DISCUSSION

This study investigated AATD sibling pairs for phenotypic differences, and provides preliminary evidence that such differences may be accounted for by genes that modify specific features. We acknowledge that ascertainment bias may have influenced the clinical results; many subjects with AATD remain undiagnosed and restricted testing for the condition influences the detection rate and types of patients diagnosed ([28]). Thus we must accept some selection bias in both of our study populations whilst acknowledging that they are fully representative of the patients who present in respiratory practice. There could also be some selection bias in the non-index siblings, despite our practice to offer testing to all siblings of index cases, since those who experience symptoms may accept testing more readily than those who are symptom-free. This may explain the high incidence of disease in the non-index siblings in our study −59.5% had an FEV₁ < 80%predicted, considerably greater than the 29% of siblings of usual COPD patients ([29]).

An important observation is that no correlation was seen in FEV₁ between siblings even after adjustment for personal smoking history. This discordance was also seen in the pairs of non-index siblings who would be affected less by ascertainment bias. Conversely the observed correlations for KCO and UZVI after adjustment for the main environmental influence (cigarette smoke exposure) suggest a genetic factor influencing this feature, but not FEV₁, that is shared between siblings, thus highlighting possible differences in the phenotypic influence of modifier genes. Many studies have shown genetic loci that may influence spirometric measures and lung function change in usual COPD, summarised elsewhere ([30]). The existence of genetic modifiers influencing the risk of COPD development in AATD has also been shown recently in a family based study of 378 PiZZ individuals ([5]), although the panel of SNPs tested was different from those in the current study.

We have previously shown that FEV₁ change is associated with basal emphysema, while KCO is associated with upper zone disease ([31]). The correlation seen between siblings for KCO and UZVI, and lack of correlation for LZVI and FEV₁ concurs with this. Two potential causes for the difference in concordance seen could be drawn. First, it may be that the zone of emphysema is dictated by genetic background, and it is this that then affects lung function. Alternatively it may be that alternate subsets of genes drive parenchymal and airway disease. The former hypothesis is better supported by current evidence, demonstrating several genetic influences on emphysema distribution in usual COPD ([32]). Since the existing evidence supports genetic influences on upper zone disease, we sought associations of both this and KCO in the siblings, despite the correlations seen in our data set. It is perhaps not surprising that we were unable to prove any associations for the SNPs we tested with KCO or UZVI given the clinical correlations in our group.

In diseases where a number of genes influence outcome, as is likely to be the case with COPD, each polymorphism may confer only a small increased risk of disease, thus large numbers are needed to detect differences between populations. In this respect our study is limited, and underpowered in the sibling work. This means that we cannot exclude an effect in all SNPs, but does not diminish the significance of our result for rs2118177. The level of P-value to determine significance in genetic studies is controversial, varying dependent on the method of correction used. Standard correction methods such as the Bonferroni are recognised as too conservative for genetic studies ([33]), hence we adopted the well recognised strategy of replication in a an independent data set. This confirmed an association of rs2118177 with FEV₁, but the association with emphysema zone for rs8179181 did not replicate in the non-sibling data set and hence is unlikely to be a true finding.

rs2118177 is an intronic SNP of no known function that tags the promoter region of the gene in HapMap ([34]), though coverage here is relatively sparse. There is no prior evidence supporting disease associations with this SNP. Evidence from this study alone would be insufficient to justify the expense of fine mapping the haplotype block to find the actual disease associated variation, but should encourage further study of the role of surfactant proteins in emphysema and AATD, since a genetic association implies a role for the protein product in pathogenesis. Surfactant proteins may be of importance in COPD because of their role in control of surface tension. Mathematical models of emphysema suggest that low levels of surfactant would influence airspace variability and lung recoil ([35]). Low levels of SFTPB have been demonstrated in the lungs of hamsters with emphysema, going some way to support this hypothesis ([36]). Furthermore, surfactant proteins may interact with AAT to decrease anti-elastase activity ([37]), which could be critical in AATD as anti-elastase activity is already profoundly reduced.

Despite the limitations of our study we have been able to demonstrate the existence of a genetic modifier in the region of SFTPB in AATD. Furthermore our results show segregation of SNP effects on lung function. A larger study of highly characterised discordant sibling pairs examining a comprehensive set of candidate genes could provide more definitive evidence of the importance of genetic modifiers in AATD.

The authors wish to acknowledge the team of respiratory physiologists and nursing staff for their help with clinical data collection. Some of the results of this paper were obtained by using the software package S.A.G.E. which is supported by a U.S. public health service resource grant (RR03655) from the National Centre for Research Resources. The authors are supported by an unrestricted grant from Talecris Biotherapeutics and by the Wellcome Trust, neither of whom were involved in study design or manuscript preparation