Pathogenic variants causing ABL1 malformation syndrome cluster in a myristoyl-binding pocket and increase tyrosine kinase activity

ABL1 is a proto-oncogene encoding a nonreceptor tyrosine kinase, best known in the somatic BCR-ABL fusion gene associated with chronic myeloid leukaemia. Recently, germline missense variants in ABL1 have been found to cause an autosomal dominant developmental syndrome with congenital heart disease, skeletal malformations and characteristic facies. Here, we describe a series of six new unrelated individuals with heterozygous missense variants in ABL1 (including four novel variants) identified via whole exome sequencing. All the affected individuals in this series recapitulate the phenotype of the ABL1 developmental syndrome and additionally we affirm that hearing impairment is a common feature of the condition. Four of the variants cluster in the myristoyl-binding pocket of ABL1, a region critical for auto-inhibitory regulation of the kinase domain. Bio-informatic analysis of transcript-wide conservation and germline/somatic variation reveals that this pocket region is subject to high missense constraint and evolutionary conservation. Functional work to investigate ABL1 kinase activity in vitro by transient transfection of HEK293T cells with variant ABL1 plasmid constructs revealed increased phosphorylation of ABL1-specific substrates compared to wild-type. The increased tyrosine kinase activity was suppressed by imatinib treatment. This case series of six new patients with germline heterozygous ABL1 missense variants further delineates the phenotypic spectrum of this condition and recognises microcephaly as a common finding. Our analysis supports an ABL1 gain-of-function mechanism due to loss of auto-inhibition, and demonstrates the potential for pharmacological inhibition using imatinib.


Introduction
ABL1 is a proto-oncogene encoding a nonreceptor tyrosine kinase with diverse roles in cytoskeleton remodelling and the DNA damage response [1]. It is best known as part of the somatic BCR-ABL1 fusion gene in the Philadelphia chromosome, associated with chronic myeloid leukaemia (CML) and acute lymphocytic leukaemia (ALL) [2].
ABL1 spans 170 kb of chromosome 9q34.12, comprising 11 exons. It has two isoforms owing to use of alternative first exons. The longer transcript (NM_007313), encodes 19 additional N-terminal residues involved in auto-inhibition of the ABL1 kinase.
Tyr245 lies in the SH2-kinase linker domain of ABL1 essential for the "docking" of the Src homology (SH3) domain in the inactive conformation of the protein [5]. Docking of SH3 to the SH2-kinase linker domain is one of three "linchpins" proposed to hold ABL1 in an inactive closed state, and therefore has an important auto-inhibitory role [6]. Phosphorylation of Tyr245 is necessary for maximal wild-type ABL1 kinase activity; a p.Tyr245Phe substitution reduces ABL1 kinase activity by 50% in vitro [7]. Unexpectedly, Wang et al. found that the c.734A > G p. (Tyr245Cys) variant increased ABL1 kinase activity [3], suggesting a possible gain-of-function effect.
Ala356 lies within the myristoyl-binding pocket of the ABL1 kinase domain. In isoform 1b (NM_007313) a myristoyl group bound to the N-terminal glycine of ABL1 occupies this pocket to stabilise the inactive conformation of the protein [8]. This docking of the myristoyl residue is the second "linchpin" of ABL1 auto-inhibition, which is lost in the BCR-ABL fusion product as the ABL1 Nterminus is truncated. The substitution of Ala356 for the polar amino acid threonine is expected to disrupt important hydrophobic interactions within the pocket. Indeed, the c.1066G > A p.(Ala356Thr) variant has increased kinase activity in vitro [3], consistent with a gain of function due to failure of auto-inhibition.
The developmental significance of ABL1 is illustrated by animal models. Abl 2 mice harbour a targeted insertiondeletion in which exon 5 and part of exon 6 are replaced by the neomycin resistance gene [9]. Homozygotes die soon after birth with thymic and splenic atrophy, lymphopenia and osteoporosis [10]. Abl m1 mice have a large targeted deletion of approximately one third of the ABL1 protein from the c-terminus [11]. Homozygotes have increased perinatal mortality, with defects of spleen, head and eye development [12]. Interestingly, Abl m1 homozygotes of a C57BL/6J background also develop cardiac abnormalities [13]. Heterozygotes of both strains are largely unaffected, suggesting that ABL1 does not display haploinsufficiency, and supporting the possibility that the human germline missense variants act through a gain of function.
In this study, we have collected clinical and molecular details of six patients with deleterious ABL1 variants and have modelled the effects of these variants in vitro. We find that all but one of the variants identified in this cohort cluster in the myristoyl-binding pocket of ABL1, and that these variants increase the tyrosine kinase activity of ABL1 in vitro. These results are consistent with a gain-of-function effect, in which the variants disrupt the crucial autoinhibitory binding of an N-terminal myristoyl group to its binding pocket. We find that variants in this myristoylbinding pocket are a common cause of the ABL1 cardiac and skeletal malformation syndrome.

Subjects and methods
Patients Patients 1, 3 and 5 were identified through genetic variant results returned via the Deciphering Developmental Disorders (DDD) study [14] (Complementary Analysis Project #278; DECIPHER IDs: patient 1 = 304716, patient 3 = 300146, patient 5 = 304918). UK ethical approval for the DDD study has been granted by the Cambridge South Research Ethics Committee (10/H0305/83). Patients 2, 4 and 6 were identified as part of routine clinical practice through clinical genetics services in Australia and New Zealand. Informed consent for publication was obtained for all patients whose clinical details and clinical photographs are included in this report. Ethical approval for the study involving patient 4 was obtained from the New Zealand Health and Disability Ethics Committee (16/STH/3). Clinicians of all patients reported to have ABL1 variants were contacted and requested to make assessments of variant pathogenicity in their patients.

Genetic analysis
For patients 1, 3 and 5, whole exome sequencing of saliva DNA samples was carried out through the DDD study. The DDD sequencing and bioinformatics framework has been previously described [15]. The DDD study identified 23 patients with missense variants in ABL1. Variants deemed pathogenic or likely pathogenic among those patients included in this report were confirmed by Sanger sequencing. The exome sequencing strategies used to identify the variants in patients 2 and 4 have been previously described [16,17]

Analysis of variation and conservation in ABL1
Non-pathogenic missense variants of ABL1 in patients with abnormal phenotypes were collated from the DECIPHER database [14]. ABL1 missense variants in healthy population controls were identified through the gnomAD [19] and EVS [20] databases. PhyloP basewise evolutionary conservation scores [21] for every position in the transcript (NM_007313.2) were obtained through the UCSC table browser [22]. Missense constraint scores for every codon were obtained from the MTR Gene Viewer [23]. Pathogenic variants were analysed by the standard pathogenicity prediction programs PolyPhen [24], SIFT [25], MutationTaster [26], and CADD [27].

Clinical features
The clinical features of our cohort are summarised in Table 1, and representative clinical photographs are given in Fig. 1. All six individuals recapitulate the phenotype of congenital heart disease, skeletal malformations and characteristic facies which had been previously described. Hearing impairment has recently been identified as a common feature of the ABL1 malformation syndrome [4]; four of our cohort exhibit conductive or mixed conductive/sensorineural hearing impairment, which was severe and persistent in one patient. Interestingly, two individuals have tall stature, in contrast to short stature in the majority of cases. Some other phenotypic features are also over-represented in our cohort, including camptodactyly (5/6) and microcephaly (5/6). Others are under-represented or absent, including pectus deformity (1/6), ear abnormalities (1/6), gastro-intestinal disorders (1/6), joint hyper-extensibility (0/6), dental decay (0/6), and genitourinary disorders (1/6) ( Table S1).

Pathogenic ABL1 variants
We identified six individuals with five deleterious de novo missense variants in ABL1. Four of these variants have not Adult head circumference centiles are based on charts produced by Bushby et al [34]. BW birth weight.
been previously described. Our findings are summarised in Fig. 2. Four of the five variants cluster in the myristoylbinding pocket within the kinase domain, which is a critical auto-inhibitory region in ABL1 (Fig. 3).
The molecular characteristics of each variant are shown in Table 2. The nucleotide and amino acid at each position is highly conserved between species. All variants are predicted to be deleterious by multiple pathogenicity prediction programs. None of the variants are present in gnomAD. All novel variants were classified as "pathogenic" or "likely pathogenic" by ACMG criteria [28].

Benign ABL1 variation
Whereas pathogenic heterozygous germline ABL1 variants cluster within and adjacent to the ABL1 kinase domain, non-pathogenic ABL1 variants in the DECIPHER cohort largely lie outside this region (Fig. 2a). Benign germline variation among gnomAD participants is found in every domain of ABL1 but is relatively scarce within the kinase domain (Fig. 2a). Mean PhyloP scores are significantly higher in the kinase domain and the SH3/2 domains than the rest of the transcript, while MTR scores are correspondingly lower in these regions (Fig. 2b,  Supplementary Table 2). Codons and individual bases within the kinase domain are therefore prone to greater missense constraint and evolutionary conservation than other positions in the transcript.
Somatic ABL1 missense variants associated with imatinib resistance in BCR-ABL leukaemias cluster exclusively within the kinase domain (Fig. 2a). One residue (Ala452) is associated both with pathogenic variation in the germline, and imatinib resistance as a somatic variant.

In vitro ABL1 assay
To investigate ABL1 kinase activity in vitro, HEK293T cells were transfected with plasmid constructs encoding wild-type or variant ABL1 cDNA. Cell lysates were assayed for phosphorylation of ABL1-specific substrates by immunoblotting (Fig. 4). Phosphorylation of ABL1-Tyr245 and STAT5B were substantially increased in lysates transfected with the c.1066G > A p.(Ala356Thr) construct, consistent with previous reports [3]. Phosphorylated ABL1 and STAT5B were also increased for the c.1354G > A p.(Ala452Thr), c.1574T > C p.(Val525Ala), and c.1582G > A p.(Glu528Lys) constructs (Fig. 4a). These results are consistent with gain of ABL1 tyrosine kinase activity due to loss of auto-inhibition by myristoyl binding.
No evidence of ABL1 activation was seen for the negative control, c.881A > G p.(Glu294Gly). This variant was identified through the DDD study in a patient with a likely pathogenic de novo variant in another gene.
The c.731T > C p.(Val244Ala) construct caused an increase in phosphorylation of STAT5B and of overall tyrosine phosphorylation, with reduced autophosphorylation at Tyr245 compared to wild-type. Alteration of Val244 to alanine therefore appears to result in loss of phosphorylation of the adjacent Tyr245 residue. This result is surprising, as phosphorylation of Tyr245 is thought to potentiate ABL1 kinase activity [7]. However, the previously described pathogenic c.734A > G p.(Tyr245Cys) variant also abolishes phosphorylation at this site [3]. Furthermore, other variants within the SH2-catalytic domain linker region have previously been shown to cause ABL1 to adopt an active conformation by disrupting the inhibitory interaction between the SH3 and catalytic domains [7,29]. Finally, as expected, imatinib abolished phosphorylation of ABL1-Tyr245 and STAT5B in all of the constructs (Fig. 4b).

Discussion
We describe six additional patients with a skeletal and cardiac malformation syndrome caused by heterozygous missense variants in ABL1, which in five cases are confirmed as de novo. All but one of these variants cluster in a myristoyl-binding pocket within the kinase domain, suggesting a gain-of-function effect due to loss of autoinhibition. We present functional data consistent with this mechanism and reaffirm that the clinical phenotype of the ABL1 syndrome includes conductive hearing loss.
A distinctive feature of the variants we describe is their close spatial relationship to one another in the threedimensional crystal structure of the protein, and specifically their position within the myristoyl-binding pocket. This close spatial relationship is not immediately apparent from the position of these variants in the ABL1 transcript, and yet suggests a self-evident mechanism by which they can exert a deleterious gain-of-function effect. We predict that other amino-acid substitutions within the myristoyl-binding pocket, particularly those which disrupt hydrophobic interactions or introduce bulky amino acids, will also be deleterious in the germline. We also predict that missense variation or in-frame deletion of the N-terminal glycine of ABL1, which carries the myristoyl modification, will be deleterious.
We also describe a novel c.731T > C p.(Val244Ala) variant which lies in the SH2-kinase linker domain, immediately adjacent to a previously described c.734A > G p.(Tyr245Cys) variant. Other variants in this linker domain are known to disrupt the inhibitory docking of the SH3 domain to the SH2-kinase linker domain, and thereby constitutively activate the ABL1 kinase [7]. Notably, the c.731T > C p.(Val244Ala) variant we describe causes reduced phosphorylation of ABL1-Tyr245 in vitro. Phosphorylation of Tyr245 is necessary for maximal activation of the wild-type ABL1 [7], yet the c.731T > C p. (Val244Ala) and c.734A > G p.(Tyr245Cys) variants must activate ABL1 independently of the Tyr245 phosphorylation status. Recently, an in-frame deletion of c.434_436del p.(Ser145del) has also been associated with the ABL1 developmental syndrome [30], but no functional work has yet been performed to characterise the effect of this variant on ABL1 kinase activity.
ABL1 is best known as a proto-oncogene. In CML and other haematological malignancies, a somatic translocation between chromosomes 9 and 22 produces the Philadelphia chromosome [31], carrying a BCR-ABL fusion gene. As the N-terminus of ABL1 is lost in the fusion product, the autoinhibitory binding of the myristoyl group to the kinase domain is abolished, and ABL1 gains constitutive tyrosine kinase activity which drives cellular proliferation [32].
Tyrosine kinase inhibitors (TKIs) specific to ABL1, such as imatinib, are the mainstay of treatment for CML. However, resistance to TKI therapy is strongly associated with somatic missense variants in the ABL1 kinase domain, particularly in the ATP-binding loop (P loop) and at TKIspecific binding sites [33].
It is noteworthy that none of the activating germline variants we describe have been associated with somatic TKI resistance. If, as we argue, germline variants in the kinase domain sterically hinder myristoyl binding, they will functionally mimic the loss of the ABL1 N-terminus in BCR-ABL. We therefore expect that TKIs effective against BCR- ABL should be similarly effective against these variant proteins. Indeed, TKIs may potentially in some way be therapeutically beneficial for the ABL1 developmental syndrome or its complications, for example in limiting aortic root dilatation or reducing the tendency for dilatation to occur. However, given that this condition appears to affect embryonic and fetal development, any more complete therapeutic effect would require as early treatment as possible and entail long-term therapy, potentially risking adverse drug effects. While imatinib is used to treat paediatric CML cases, it is expected to be teratogenic in pregnancy. Further work is therefore required to ascertain whether therapeutic scope exists for use of imatinib or similar TKIs in this condition.
It is also noteworthy that activating somatic missense variants in ABL1 have not been found to independently cause haematological malignancy, although both isoforms of ABL1 are ubiquitously expressed. Disruption of myristoyl binding alone may not activate ABL1 sufficiently to drive malignancy. Two other "linchpins" of ABL1 autoinhibtion (the docking of the SH3 domain to a polyproline helix in the SH2-kinase linker, and an N-terminal "brace" over the SH3-SH2 unit) may prevent its excessive activation [6]. Indeed, both the N-terminal "brace" and the myristoyl group are lost in the BCR-ABL fusion. It is not clear whether the activating germline variants we describe can act as driver variants in malignancy. No patients with the ABL1 skeletal and cardiac malformation syndrome described here or elsewhere are reported to have haematological malignancy, but longitudinal follow up of these patients will be required to better determine this.
From a clinical perspective, we believe this condition to be a phenotypically distinctive and recognisable syndrome based on affected individuals' dysmorphology and associated clinical features. Phenotypes such as skeletal malformations, aortic root dilatation and pneumothorax point towards an overlap with genetic connective tissue disorders and we recommend that ABL1 be borne in mind in such cases. The high prevalence of hearing impairment makes audiological assessment advisable. Furthermore, assessment of the aortic root diameter at the time of diagnosis may also be appropriate in individuals found to have pathogenic ABL1 variants. Evidence is currently lacking as to the true risk of aortic aneurysm and dissection in this condition. We are not aware of any affected individuals having had rapid progressive aortic dilatation requiring surgical intervention and this may therefore suggest a more indolent course. However, a precautionary approach of ongoing aortic root screening similar to that used for Marfan syndrome may be appropriate until such time as more accurate natural history data are available.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.