Genetic results are available in tables 1 and 2 and in Additional file 1, supplementary tables S3 and S8 to S11.Clinico-biological data are available in Additional file 1, supplementary tables S5 to S7.
We defined two groups: one with clear phenotype (HS group) and one with unclear phenotype (UH group). The different types of variants identified are summarized in figure 1.
Figure 1: Distribution of mutations types
We identified the probable genetic cause of hemolysis in all HS patients using target genes panel analysis (71 CHA genes) on WES data. Nineteen patients were heterozygous carriers of one mutation in ANK1 (8 patients), SLC4A1 (6 patients), or SPTB (5 patients). All patients had different mutations, among which 13 were novel and 8 had already been reported. In one family with typical HS (P11 and affected relatives), affected members harbored 2 heterozygous probably damaging variations, one in SPTB (stop-gained) and the other in SPTA1 (already-reported missense mutation, previously found in a case of HPP which also harbored the alpha-Lely polymorphism in trans). Such an association has rarely been reported in the literature [15, 16, 43, 44]. In 3 families (P4, P10 and P11), genetic tests could be extended to other affected family members and showed cosegregation of mutations and disease (Figure 2). Two HS patients also had additional mutations in other CHA genes. P1 carried a heterozygous mutation of alpha-globin gene, called Tunis-Bizerte hemoglobin, responsible for an alpha- thalassemia trait . P17 carried a variant of uncertain significance (VUS) in PIEZO1 (p.V860M). In total, 3 out of 20 HS patients (P1, P11 and P17) had variations in two different CHA genes. Patient P2, who had significant iron overload also harbored 2 variants in HFE (H63D and C282Y) which had been previously found during iron overload exploration.
Table 1 should be inserted here if possible
Twenty patients were classified as UH because of discordant or non-contributive clinical features and/or biological tests. Familial segregation study could be performed in 6 families (P25, P27, P31, P33, P36, P40, figure 2).
Twelve UH patients have been fully characterized (P21, P22, P25, P29, P30, P31, P32, P36, P37, P38, P39, and P40) by target gene panel analysis of 71 CHA genes on the WES data. One UH patient (P35) could be fully characterized thanks to WES analysis. Seven other UH patients have not been fully characterized (P23, P24, P26, P27, P28, P33 and P34) despite extended WES analysis. Interestingly, in those patients WES identified two new possible target genes.
Among the 12 fully characterized UH patients, nine harbored mutations in genes encoding membrane proteins. SPTA1 was the most frequently mutated gene (in 6 probands: P21, P22, P29, P36, P37, and P39). Variations in SPTA1 result in HE, HPP or HS depending on variants type and phase. In P37 and P39, genotypes and phenotypic data (clinical features, blood smear) were suggestive of HPP. We could not clearly decipher between HE or HPP for P21, 22 and P29. Other membrane genes with mutations were PIEZO1 (P30 and P36, DHSt), KCNN4 (P31, GARDOS MIM 616689), SPTB (P32, elliptocytosis) and SLC4A1 (P36, SEA ovalocytosis, MIM 166900). Among them, some had original presentation. P30 had myelodysplasia and important hemolysis without any family history. She carried a constitutional (present in blood and hair bulb DNA) heterozygous PIEZO1 variant of uncertain significance (VUS) (p.P376A) and had atypical ektacytometry results. P36 is one of the 3 A/S symptomatic patients explored in our study (P29, P32 and P36). This woman originating from Comoros islands experienced spleen infarct after a long-distance flight and showed the association of a HbS trait, SEA ovalocytosis (SLC4A1 : p.Ala400_Ala408del), a G6PD MATERA A- p.V98M variant at heterozygous state , a homozygous SPTA1 variation (p.E2224D) and a PIEZO1 mutation (p.R457C) already involved in DHSt . Her daughter (P36-1) also presented with hemolytic anemia and harbored the same mutations (the SPTA1 variant being heterozygous). Her blood smear showed anisopoikilocytosis and some stomatocytes. Another relative (P36-2) was explored and carried the same mutations as P36-1. He had a hemolytic anemia and a retinopathy typical of sickle cell disease.
Three fully characterized UH patients had mutations in genes not encoding membrane proteins : SEC23B in P25 with CDA type 2 ; CFH in P38 with aHUS; ATP11C in P40. P25 harbored 2 already described mutations in SEC23B gene at compound heterozygous state (figure 2) allowing correction of diagnosis towards CDA type 2 and not HS as initially suspected.
P38 carried a heterozygous mutation in CFH (p.R53C), which had previously been found in patients with preeclampsia-related SHUa , in complement-related glomerulopathies  and in familial forms of AMD (age-related macular degeneration, MIM 610698) . She also carried a heterozygous PIEZO1 VUS (p.G1416R) but had normal ektacytometry and blood smear. In this case one single mutation in CFH probably explains the entire phenotype (AMD, preeclampsia, hemolysis and altered renal function). P40 carried a likely pathogenic hemizygous variation in ATP11C (p.P812S). This X chromosome variation is recorded in gnomAD in only one heterozygous female (no hemizygous males or homozygous females recorded). He also carried one VUS in ANK1. His blood smear was normal and ektacytometry showed atypical profile with only dehydration and no change in osmotic resistance. Functional testing of flippase activity is in progress.
Among the 12 fully characterized UH patients, 6 had mutations in several CHA genes:
- P29 HBB S mutation and SPTA1
- P31, KCNN4 and PIEZO1
- P32, HBB S mutation and SPTB
- P36, HBB S mutation , SPTA1, SLC4A1, PIEZO1 and G6PD
- P38, CFH and PIEZO1
- P40, ATP11C and ANK1,
One additional UH patient was characterized thanks to WES analysis. P35 was found to harbor a heterozygous deleterious variation in SH2B3: c.1A>G (initiation codon loss). This result allowed to reconsider diagnosis towards a probable myeloproliferative condition. SH2B3 somatic mutations have been reported in myeloproliferative neoplasms such as primary myelofibrosis . He also carried a heterozygous VUS in the SCN9A gene c.2938G>T (p.A980S) which likely explained the severity of painful crises reported in this patient. SCN9A is involved in neurogenic painful syndromes .
Seven UH patients remained unsolved (P23, P27 and P28) or partially solved (P24, P26, P33 and P34) despite WES extended analysis. P23 carried a heterozygous ALAS2 promoter variation (c.-258C>G), which had previously been reported as a cause of X-linked sideroblastic anemia (MIM 300751) . This variation is present in gnomAD at an allelic frequency of 0.54%, with 39 hemizygous males, which suggests that it is most probably a rare benign polymorphism. No other relevant genetic variation was found. P27 was a female child with major hemolytic anemia at birth and neonatal splenomegaly with thrombocytopenia. Her mother, maternal aunt and maternal grand-father had the same phenotype. She carried a heterozygous variation in CFH (p.Q950H) , which was absent in the mother and the maternal aunt and is therefore not responsible for CHA in this family. No other relevant genetic variation was found. P28 was carrier of two heterozygous variants in two genes involved in CDA: one in SEC23B (p.V426I) and one in CDAN1 (p.P86S). No case of digenic inheritance has yet been reported in CDA. No other relevant genetic variation was found. The other patients had a part of their phenotype explained by WES. In P24, candidate gene variations could be identified thanks to WES. A heterozygous missense variation was found in the gene encoding the RNA-specific Adenosine Deaminase (ADAR) (p.P529L). No ADAR mutation has been associated to date with congenital hemolysis in humans but several studies showed a crucial role for ADAR in mouse erythropoiesis. This patient also carried an apparently homozygous variation of TRPV4 (p.P638L), present in the gnomAD database at an allelic frequency of 0.03% without any homozygotes recorded (over >138000 subjects tested). TRPV4 mutations have been found in skeletal dysplasia, arthropathies and in a familial form of osteonecrosis . The TRPV4 mutation could explain osteonecrosis but not hemolysis. P26 carried HAMP and HFE variations probably explaining iron overload and CD46 variation for susceptibility to hemolysis. HAMP gene is out of our genes panel analysis. P33 harbored G6PD and SPTB variations, which combination cannot explain the severe phenotype. P34 had UH and iron overload, and variations in 3 genes: a homozygous p.H63D variation in HFE, which may contribute to iron overload even though its implication is not totally clear [55,56], a heterozygous ABCG8 VUS and 2 heterozygous ADAMTS13 VUS. ABCG8 mutations are found in AR sitosterolemia and AD xanthelasma . A single heterozygous ABCG8 variation is not sufficient to explain hemolysis. This patient had a normal platelet count and ADAMTS13 enzyme activity testing showed a 39% decrease (suggestive of a heterozygous loss-of-function). Normal ADAMTS13 activity in P34 does not support a diagnosis of AR Thrombotic thrombocytopenic purpura (MIM 274150), and therefore cannot explain hemolysis.
In summary, among n UH patients, 12 patients could be fully characterized thanks to targeted analysis on CHA genes. Eight benefited from extended WES analysis leading to full characterization in one additionnal patient and 7 UH patients not fully characterized but two new potential target genes were identified by WES.
Figure 2: Pedigrees of 3 HS families and 6 UH families
Figure 2 Legend: Black squares: affected males; black circles: affected females; open squares: unaffected males; open circles: unaffected females; arrow: the proband. 2a = P11 family pedigree (HS); 2b = P10 family pedigree (HS); 2c = P4 family pedigree (HS); 2d = P25 family pedigree (CDA2); 2e = P27 family pedigree (UH); 2f : P31 family pedigree (GARDOS); 2g : P33 family pedigree (UH); 2h : P36 family pedigree (multiple association); 2i : P40 family pedigree (ATP11C hemolysis)