The transcription factor NK2 homeobox 1 (NKX2-1), a member of the NKX2 transcription factor family, partially controls the synthesis of surfactant proteins (1). NKX2-1 was initially identified as a nuclear protein binding the promoter of thyroglobulin gene. However, various studies later showed that NKX2-1 is also expressed in lungs and brain (2). Homozygous mice lacking NKX2-1 die shortly after birth, lack lung parenchyma and the thyroid gland, and exhibit defects in the ventral region of the forebrain, including lack of a pituitary gland (3).

The human gene located in 14q13 is encoded by three exons and is characterized by its functional domain, the homeodomain in exon 3, which also contains the nuclear localization signal (NLS) (4). The most abundant and functional transcript is generated from an open reading frame starting in exon 2 and encoding 371 amino acids (5, 6). In humans, autosomal dominant NKX2-1 pathogenic variants are associated with brain-lung-thyroid syndrome (BLTS), but not always: there is a large variability in the phenotype presentation that may exclude any of the three phenotypes in BLTS (7, 8). Both neurological and pulmonary features are responsible for the disease morbidity, but lung disease is mainly responsible for the mortality (9). Various pulmonary phenotypes have been described, including respiratory distress syndrome in term infants and interstitial lung disease (ILD) in children and more recently in adults owing to abnormal lung development and the lack of pulmonary surfactant (8, 10).

Surfactant is a multi-molecular complex consisting of phospholipids (80%–90%) and proteins (10%–15%); 2% to 3% are specific proteins called surfactant protein A, B, C and D (SP-A, SP-B, SP-C and SP-D, respectively) (11). SP-B and SP-C are synthesized by alveolar type II cells as pro-peptides (pro-SP-B and pro-SP-C), then assembled with lipids into bilayer membranes and secreted into the alveolar airspace, where they form a surface film at the air–liquid interface (12, 13). Adenosine triphosphate-binding cassette transporter A3 (ABCA3) transports phospholipids from the cytoplasm into the lamellar bodies where they combine with surfactant proteins B and C (12, 13).

The thyroid involvement due to NKX2-1 pathogenic variant is defined by thyroid dysgenesis with or without morphological abnormality of the thyroid. Various phenotypes can be observed, from congenital hypothyroidism diagnosed in the first days of life to a compensated hypothyroidism not requiring treatment (14). The NKX2-1 thyroid targets are thyroglobulin (TG), thyroperoxidase (TPO) and TSH receptor (TSHR) promoters (15–17). Paired box gene 8 (PAX8) is another transcription factor involved in thyroglobulin promoter regulation. Functional studies showed the synergism of the association of PAX8 and NKX2-1 on the thyroglobulin promoter (14, 17).

The neurological symptoms associated with NKX2-1 pathogenic variants include hypotonia in the first months of life, with motor delay and benign chorea occurring during childhood. The severity of choreic symptoms varies among families and usually stabilizes in adulthood (18). Other types of abnormal movements such as myoclonus, upper-limb dystonia and verbal tics have been described (18). In the embryonic brain, NKX2-1 controls the specification of the GABAergic interneurons and oligodendrocytes that populate the ventral and dorsal telencephalic region acting on the glial fibrillary acidic protein (GFAP) promoter. Loss of NKX2-1 leads to ventral to dorsal re-specification of the pallidum causing loss of GABAergic interneurons and oligodendrocytes in the dorsal telencephalic region (19). Other neurological targets of NKX2-1 have been described. NKX2-1 promotes GABAergic and cholinergic cell fate by inducing LIM homeobox 6 (LHX6) (20, 21). However, no functional test, such as luciferase assay, is available to clearly assess this function.

Here, we identified and functionally characterized two NKX2-1 variants identified in two children.

The two pathogenic variants were frameshift variants that created the same aberrant protein downstream of the 86th amino acid. The variant c.85_86del (NM_003317.3) was carried by a father and his daughter, who had isolated respiratory symptoms (8). The variant c.254dup (NM_003317.3) was identified in a boy who presented with BLTS. To understand why two frameshift pathogenic variants, produce such different phenotypes, we used functional comparative studies of both variants with thyroid- and lung-specific in vitro functional assays.

In the present report, we describe two heterozygous pathogenic variants in the NKX2-1 transcription factor. The first frameshift pathogenic variant, c.85_86del, was carried by a girl who showed ILD during the first year of life. This girl and her father, who carried the same variant, exhibited only lung disease. The second variant, c.254dup, previously described (19), was carried by a boy who had BLTS. The boy had respiratory distress syndrome at birth, but the condition improved progressively without treatment. Congenital hypothyroidism was diagnosed at birth, and the boy also showed choreic movements and psychomotor delay.

These two variants shift the reading frame in the same way after the 86th amino acid, leading to aberrant proteins associated with different BLTS phenotypes. Variants (n = 10) that shift the reading frame in the same way leading to an aberrant protein of 406 amino acids with deletion of the functional homeodomain were previously described (10, 34–36). Seven of the 9 subjects described had BLTS; two of them had thyroid and neurological symptoms, and one presented thyroid and respiratory symptoms.

To improve our understanding of the molecular basis of this heterogenous phenotype, we first performed an in vitro study to compare the NKX2-1^DUP and NKX2-1^DEL mutant proteins. Indeed, western blots showed that both had higher molecular weight than NKX2-1^WT, the weight of NKX2-1^DUP being much higher than NKX2-1^DEL in both Nthy and A549 cells. These discrepancies could be explained by the functional motifs within the amino acid sequence of the NKX2-1^wt protein that were affected by the pathogenic variants (Figure 1B). NKX2-1 posttranslational modifications, such as phosphorylations, acetylations, and redox states could alter their molecular weight, mobility in electrophoresis assays and transcriptional activity (37). Among the five SP motifs present in the NKX2-1^wt protein, three were absent in mutants, these latter being important in the phosphorylation state of the protein and its cellular-specific transcriptional activity (38). One putative new SP motif appeared at the Cter of both mutants and could also impact their phosphorylation state and thus their activity. Additionally, three cysteines involved in the dimerization of the wild-type protein (39) disappeared in both mutants, whereas four other cysteines appeared which could disturb the dimerization of the mutants and/or their redox state (40), and consequently affect the molecular weight. We also showed that despite the loss of the native NLS, the two mutant proteins could reach the nucleus, according to putative NLSs that could be present in both mutant sequences (Supplementary Figure S1). Indeed, in both cell lines, the mutants were in cytoplasm but also in the nucleus, where they formed aggregates instead of the diffuse appearance seen with the wild-type protein. The cell nucleus contains distinct classes of subnuclear structures (cajal bodies, splicing speckles, paraspeckles). Many nuclear proteins are known to interact dynamically with one or another of these bodies, and disruption of the specific organization of nuclear proteins can result in defects in cell functions and may cause molecular disease (41). Further experiments to explain the subnuclear re-localization of the two mutants in one or several of these specific nuclear structures are needed and would help refine the cartography of functional domains within the NKX2-1 protein sequence.

Because the mutant proteins can reach the nucleus, they could activate tissue-specific promoters such as thyroglobulin, a specific thyroid gene. Of note, unlike NKX2-1^DUP, NKX2-1^DEL was able to transactivate thyroglobulin but required a PAX8 cofactor for its activity. PAX8 and NKX2-1 are both expressed only in the thyroid, and PAX8 has a synergic effect with NKX2-1 to transactivate the thyroglobulin promoter (14, 17). Despite its highly modified amino acid, NKX2-1^DEL protein could transactivate the thyroglobulin promoter with PAX8, which suggests that PAX8 is essential for the thyrocyte-specific promoter activation of thyroglobulin (42–44). Both mutants have lost their homeodomain and have equivalent sequence sizes. However, they differ significantly from amino-acids 30 to 88 in their sequences, suggesting that this part of the Nterm region could be of interest for a selective binding of the PAX8 cofactor and putative other factors, such as TAZ. Although the two mutants we studied have both lost the LPPY motif (illustrated in Figure 1B), which may be involved in the binding of TAZ protein (45), a mechanism involving both TAZ and PAX8 within this Nterm region of NKX2-1 could be possible. Interestingly, Moya et al. (36) showed that the TAZ cofactor could bind to the 74 first amino-acids of the NKX2-1 protein, maybe through pS-P in the Nterm region. Thus, further experiments focusing on binding of both TAZ and PAX8 on the 30-to-74 first amino acids of NKX2-1 could help to decipher their roles in the tissue-specificity of NKX2-1. Synergy with PAX8 can overcome the functional defect of a specific NKX2-1 variant on the thyroglobulin promoter (14). PAX8 may recruit NKX2-1 via protein–protein interaction without NKX2-1 binding to DNA (46, 47). For the present variants, we hypothesize that the loss of the NKX2-1^DEL homeodomain was associated with the creation of a new interaction domain with PAX8, thus allowing for thyroglobulin promoter transactivation.

To gain insight into the mechanism of the pulmonary phenotype of our two subjects, we assessed transactivation of specific surfactant gene promoters, including SFTPA, SFTPB, SFTPC and ABCA3, in two cell lines. All these proteins are essential for surfactant homeostasis and normal lung development (48–53). Both mutant proteins resulted in decreased luciferase activity suggesting reduced transcriptional activity of the SFTPA, SFTPB, SFTPC and ABCA3 promoters, which is consistent with the pulmonary phenotype of our two subjects. In the 2 index children, the lung phenotype differed as one child presented with severe respiratory distress with rapid improvement with the first weeks, while the other experienced a progressive respiratory failure at 7 months evolving to chILD. Moya et al. have studied 2 children with BLTS, but one with absence of lung involvement. The authors demonstrated that the coactivator TAZ/WWTR1 (transcriptional coactivator with PDZ-binding motif/WW domain-containing transcription regulator protein (1) was able to restore NKX2-1 transactivation suggesting a potential role of TAZ in the development of the lung disease (36). Alternatively, we and others have shown that decreased surfactant protein levels may be associated with a less severe phenotype as compared with complete deficiency (54). The distinct respiratory phenotypes (chILD vs. resolved neonatal RDS) might be related to the differences in the level transcription of SP-B and SP-C which were significantly decreased in NKX2-1^DUP compared to NKX2-1^DEL. Indeed SP-B and SP-C protein levels are crucial in the adaptation of the newborn at extrauterine conditions (55) as illustrated by SP-B deficiency causing severe neonatal respiratory distress and death within the first few months of life (56) whereas chILD is mainly triggered by surfactant metabolism disruption due to abnormal proSP-C protein together with decrease ABCA3, SP-C and SP-B levels (57).

In conclusion, here we explain the difference in phenotype in two subjects carrying two frameshifts NKX2-1 pathogenic variants, c.85_86del and c.254dup. The two variants (dup + 1nt/del-2nt) produce an aberrant protein of 407 and 406 amino acids identical from the 86th amino acid lacking the functional homeodomain and the nuclear localization signal. Unexpectedly, 3D conformation prediction of the mutant protein assumed the presence of a nuclear localization signal, a bipartite sequence, which was experimentally confirmed by confocal microscopy showing both mutant proteins localized in the nucleus and cytoplasm. However, only the c.175_176del NKX2-1 mutant could transactivate the thyroglobulin promoter because of the PAX8 cofactor. Other functional studies are warranted to further elucidate the molecular mechanisms involved in NKX2-1 variants associated with heterogeneous phenotypes.