Alpha-1 antitrypsin (A1AT) is the main circulating protease inhibitor, protecting the lung parenchyma against proteolytic attacks. Alpha-1 antitrypsin deficiency (AATD) is a common but still largely under-recognized genetic disorder. It predisposes to liver and lung diseases and rarely to granulomatosis with polyangiitis and necrotizing panniculitis [1]. The wild-type allele is called PI*M while the most common deficient alleles are known as PI*S and PI*Z, according to their isoelectrofocusing (IEF) pattern. AATD-associated liver disease, observed for the deficient variants Z, SIiyama and MMalton, can be attributed to intracellular polymerization of the misfolded protein leading to endoplasmic reticulum storage disease. Mild liver storage is observed with the S variant which is probably degraded before secretion [2].

The medical indications for AATD screening were either a pulmonary or hepatic disorder or when a routine protein electrophoresis fortuitously revealed a splitting (with or without decrease) of the α1-globulin fraction at protein electrophoresis. The biochemistry laboratories of the academic medical centers of Lyon and Lille (France) currently investigate AATD by serum immunochemical quantification and IEF of A1AT. In the laboratory of Lyon, IEF is carried out on polyacrylamide gels based on the method previously described [3] with slight modifications of pH gradient (4.2–4.9). In the laboratory of Lille, IEF is performed on agarose gels using commercially available kits and immuno-enzymatic revelation (Sebia, Evry, France) [4]. In both laboratories, A1AT inhibitory activity may also be assessed through the serum elastase inhibitory capacity (SEIC) which relies on the inhibition measurement of the hydrolytic activity of the porcine pancreatic elastase by A1AT on a chromogenic substrate (N-Succinyl-Ala-Ala-Ala-p-nitroanilide). This kinetic spectrophotometric test, adapted from the method previously described by Klumpp and Bieth [5], was developed in close collaboration by the two laboratories so that the results could be comparable [6]. Using the correlation between A1AT concentration and SEIC, a theoretical SEIC can be calculated and compared to the measured SEIC with R being the ratio between the measured SEIC and the expected SEIC. For patients in heterozygosity with a new variant, R below 0.8 is presumptive of a dysfunctional variant.

This combination of techniques is sufficient to characterize up to 95% of A1AT abnormalities, mainly ZZ, SZ and SS phenotypes [1, 6, 7]. For the other cases (i.e. unexplained low A1AT level, unusual IEF pattern or IEF pattern inconsistent with clinical history), Sanger sequencing of the SERPINA1 gene including coding exons, 5′ and 3′ untranslated regions (UTRs) and splice boundaries is performed and can be extended to intronic sequences by Next Generation Sequencing technology [8]. All sequence variations are named according to the Human Genome Variation Society (HGVS) and using the reference transcript NM_000295.4 which includes the 24 residues of the signal peptide.

Over the past 10 years, more than 1200 A1AT genotyping analyses performed in our two centers led to the identification of 22 new variants in 35 patients aged from 7 to 81 years (Table 1 and Fig. 1). It is noteworthy that 4 of them were already cited but neither named nor phenotypically or clinically described [9]. According to their IEF pattern and the birth place of the probands, we named them SRoubaix, WSaint-Avre, M1Lille and M1Lyon. The criteria of the American College of Medical Genetics and Genomics (ACMG) were used to classify these 22 variants as benign, likely benign, of uncertain significance, likely pathogenic, or pathogenic [10]. Since we did not have the possibility to test them in expression vectors like HEK293T/17 or Hepa1–6 cells, the available clinical and biochemical data of A1AT were considered, as well as the results of two in silico pathogenicity predictors, shown to have a sensitivity of 0.75 for SERPINA1 mutations [11]. The first one, namely SIFT for Sorting Intolerant From Tolerant, ranges from 0.00 to 1 and is mainly based on amino-acid conservation scores. A SIFT score between 0 and 0.05 is highly predicting of an affected protein function. The second one, namely PolyPhen-2 HVAR, proposes a prediction confidence score between 0.00 and 1.00 which uses multiple alignment and protein structural data. A PolyPhen-2 score higher than 0.8 is considered as probably damaging. The recently described REVEL (for Rare Exome Variant Ensemble Learner) method [12] was also used since it had been shown to be the most suitable one for the prediction of pathogenic A1AT variants [11]. Briefly, a REVEL score of less than 0.354 is highly predictive of a benign character of the variant whereas a score of more than 0.618 is highly predictive of pathogenicity.

Table 1 Molecular, biological and clinical characteristics of the 22 new SERPINA1 variants
Full size table
Fig. 1
figure 1

IEF patterns of some frequent and rare A1AT phenotypes (polyacrylamide gels with Coomassie blue staining). 1, 33: M1M3; 2, 15, 18: M1S; 3, 17: PLoyettes S; 4: M3PLoyettes; 5, 31: M1Z; 6, 11, 20: M1M4; 7: M3PSolaize; 8, 10, 19, 21, 24, 27, 28: M1M2; 9:M1M1 12: M2SRoubaix; 13: SWVernaison; 14: M3S; 16: M2P; 22: M2XChristchurch; 23: M1XChristchurch; 25: M2XCuris; 26: M1XChristchurch; 29, 32: GSaint-SorlinM1; 30: IM3

Full size image

Seven new variants were assumed to be Null ones: Q0Lille, Q0Casablanca, Q0Saint-Etienne, Q0Achicourt, Q0Saint-Avold, Q0Amiens and Q0Montluel. They resulted from splice-site, non-sense or frame shift mutations leading to premature stop codons with biosynthesis of truncated proteins or pre-mRNA degradation by the nonsense mediated decay mechanism. Interestingly, the c.288_291del frame shift mutation gives rise to two different SERPINA1 Null variants which are associated with distinct genetic backgrounds: M2 for Q0Casablanca and Z for Q0Lille. The c.559A > T (Q0Saint-Etienne) and c.1237_1239del (Q0Montluel) mutations lead to a premature stop codon while Q0Achicourt, Q0Saint-Avold and Q0Amiens are caused by splicing abnormalities. It is noteworthy that Q0Achicourt and Q0Saint-Avold, found in young patients presenting with emphysema, were both in compound heterozygosity with another deficient SERPINA1 allele (Q0Clayton and Z, respectively).

The M1Cremeaux variant was identified in four members of a same family (two sisters and their sons). The propositus was a 36-year-old woman without any pulmonary or hepatic disorder harboring the M1Cremeaux variant in heterozygosity with the dysfunctional Z variant. A1AT biochemical analysis was prescribed because of low α1-globulin fraction at protein electrophoresis during a hair loss exploration. Despite the absence of any specific clinical impact, M1Cremeaux was considered as a deficient A1AT variant (ACMG class5) for four reasons: (i) the A1AT serum level was significantly decreased (0.23 g/L in heterozygosity with the Z allele and from 0.88 to 1.01 g/L in association with a M1 or M2 allele), (ii) the mutation was located at the beginning of the 5Aβ-strand which is an important region for the protein stability [1] (iii) the pathogenic A1AT King variant affects the same amino-acid (p.His358Asp) [13] and (iv) the SIFT score (0.48) was normal but the PolyPhen-2 and REVEL scores (0.999 and 0.650) were highly predictive of pathogenicity.

The two P variants, PLoyettes and PSolaize, were suspected to be dysfunctional according to their decreased elastase inhibitory activity demonstrated by R values of 0.62 and 0.79, respectively. Sustaining our hypothesis, REVEL, SIFT and PolyPhen-2 scores predicted PLoyettes (0.933, 0 and 1.00, respectively) and PSolaize (0.597, 0 and 0.623, respectively) as deleterious. The Wvernaison variant also harbored a decreased elastase inhibitory activity (R value 0.79) and an IEF pattern with almost undetectable bands; nevertheless, SIFT and PolyPhen-2 scores predicted it as benign (0.08 and 0.432 respectively) but not the REVEL score of 0.638. Moreover, these three variants were identified in patients with an inflammatory status (CRP plasma levels higher than 10 mg/L) that probably led to overestimation of the recorded A1AT levels. They were thus classified as likely pathogenic according to ACMG criteria (class 4).

While caused by a non-sense mutation, A1AT GSaint-Sorlin (c.1252A > T; p.Lys418*) was ranged as variant of uncertain significance (class 3) since the A1AT biochemical data were normal. As the premature stop codon is located on the very last triplet of the gene, the final protein lacks only one amino-acid and it seems to have no consequence on its synthesis or functional activity. Conversely, the M1Rouen variant was also ranged in class 3 and not considered as benign or likely benign because: (i) it appears at very low allelic frequencies in databases (ExAC and Topmed: 0.0012%), (ii) a pathogenic variant on the same amino-acid (namely, the I variant p.Arg63Cys) has been described and (iii) we could not get any serum sample to assess A1AT quantification and SEIC. In detail, the SIFT and PolyPhen-2 algorithms classify the I variant as deleterious (0 and 1, respectively) while they are contradictory for the M1Rouen variant (0.04 and 0.185, respectively). A border-line R ratio of 0.8 was obtained for an asymptomatic 34 -year -old woman harboring the WSaint -Avre variant in heterozygosity with the dysfunctional Z variant. According to its low frequency in databases (ExAC: 0.0032%) and to its SIFT and PolyPhen-2 scores (1 and 0.000 respectively), WSaint -Avre was also ranged in class 3 of ACMG classification.

The remaining eight variants were classified as likely benign (class 2) because in silico algorithms predicted no impact on gene product and the A1AT quantitation and SEIC measures revealed no abnormality.

Very interestingly, we also identified during the course of this study two SERPINA1 deficient variants that were very recently described: Trento (p.Glu99Val) [14] and SDonosti (p.Ser38Phe) [15]. The Trento variant showed compromised conformational stability after secretion from the hepatocyte [14]. In our cohort, this variant was present in heterozygosity with the MMalton variant in a 42-year-old man with a low A1AT level (0.85 g/L) presenting with hepatic fibrosis. The SDonosti variant was shown to form intra-cellular polymers that prevent its secretion from the hepatocytes. We identified the SDonosti variant in two unrelated individuals (in heterozygosity with the M1 variant and with the S variant, respectively): (i) a 64-year-old woman suffering from emphysema (A1AT level = 1.21 g/L but inflammatory status not known) and (ii) a 41-year-old man suffering from hemochromatosis (A1AT level = 0.80 g/L).

In conclusion, this study highlights the importance of the whole SERPINA1 gene sequencing (and not only the specific research of the Z and S variants) to explain some AATD clinical and biological pictures. Among these 22 new A1AT variants, a significant percentage of severely deficient ones (class 5) was observed (36.4%): Seven Q0 alleles and one deficient M1 allele (M1Cremeaux). Three variants (PLoyettes, PSolaize and WVernaison) could be classified as dysfunctional variants (class 4) mainly because of their reduced elastase inhibitory activity. Three variants (M1Rouen, GSaint -Sorlin and WSaint -Avre) were classified as variants of uncertain significance (Class 3) and the eight remaining ones as likely benign (Class 2). To note, we fortuitously observed that the IEF pattern of the SRoubaix variant depended on the migration medium: W-like on polyacrylamide gels (Lyon) and S-like on agarose gels (Lille) (Additional file 1: Figure S1). Since all patients carrying the SRoubaix variant were of North African origin, we highly speculate that this variant might correspond to the ‘old’ W3Constantine described in 1977 by Khitri [16]. The recent meta-analysis by Silva et al., completed by the present data, represents the most up-to-date list of SERPINA1 variants available so far.