No evidence of locus heterogeneity in familial microcephaly with or without chorioretinopathy, lymphedema, or mental retardation syndrome

Sequencing of the 22 coding exons of KIF11 and their flanking splice-sites identified fourteen heterozygous mutations in the 23 index patients (Table 1 and Figure 2). All patients with family history (n = 6) had a mutation. These germline variants included two nucleotide substitutions and one single nucleotide deletion; all result in premature stop codons. Three splice-site alterations were also identified. The changes co-segregated with phenotype in the five families for which we had additional samples (n = 20) (Table 1 and Figure 3).

Among the 17 sporadic patients, we found eight with a mutation in KIF11: one nonsense, four frameshifts and two canonical consensus splice-site changes (Table 1) that we consider pathogenic; and one missense alteration that we regarded as probably pathogenic, but requiring further functional validation (Table 1). DNA from parents was available for 7 of the 8 sporadic patients with a KIF11 mutation. All appeared to be de novo changes (100%) (Figure 3) absent in both parents, representing 41% (n = 7/17) of all our sporadic cases. In total, 30% of all the index cases studied (7/23) had de novo KIF11 mutations.

The mutations detected most likely result in loss of function of KIF11 due to nonsense-mediated mRNA decay or premature truncation of the protein (n = 13). Lymphoblast RNA from a patient with a splice-site alteration (Patient II-10: c.308+1G>T) showed skipping of exon 3 (r.211_308del) (Figure 4A), leading to a change in reading frame (p.Thr71Argfs*8) (Figure 4B). The mutant transcript was expressed at a lower level than the wild-type allele, suggesting mRNA loss by nonsense-mediated mRNA decay. In patient V-12, the splice-variant (c.790-3A>G) generates a new AG-acceptor site-like sequence. In lymphocytes, this led to the insertion of 2 nucleotides from the intronic splice site (r.789_790insAG) into the mRNA (Figure 4C and D), causing a reading frame shift and appearance of a premature stop codon 26 amino acids after (p.Val264Argfs*26). We identified only 1 mutation leading to a heterozygous amino acid substitution: Tyr82Phe (Patient I-10). The substitution occurs at a highly conserved amino acid position in the kinesin motor domain from fish to human being. It is predicted to alter protein function by PolyPhen-2, SIFT and Mutation Taster 2.0. The change is absent in the 1000Genomes, ESP6500, and GoNL databases.

Taking into account all the patients found to carry a mutation in this study (n = 26), as well as the 61 already reported [6-10], microcephaly was present in 79/87 patients (91%), eye anomalies in 63/87 patients (72%), intellectual disability in 58/87 (67%) and lymphedema in 41/87 patients (47%). Average SD for microcephaly was −5,1 (ranging from −1,1 to −9,5), on the basis of 64 patients with published data. None of the clinical signs was fully penetrant. Atypical features including cerebral and cardiac anomalies were present in 48 patients (55%). Taking all index patients with mutations into account (n = 52), sporadic cases represent 42% (22/52). It has been proven in 73% (16/22) of them that the mutation appeared de novo.

The nine KIF11 mutation-negative sporadic patients were included in KIF11 mutation screening due to the presence of signs and symptoms corresponding to MCLMR. However, certain patients showed additional features, not described for MCLMR: hydrops fetalis and lymphangiectasia (1 patient: XX-10), motor delay (2 patients: XVIII-10, XXIII-10), and Madelung deformity (1 patient: XIX-10). These patients probably reflect the clinical and genetic variability of microcephaly syndromes.

We discovered 14 KIF11 mutations, 12 of which are novel, in 26 MCLMR patients (Figure 2). A mutation was detected in all patients with a family history of MCLMR, similar to previously reported series [6-10]. We also discovered a heterozygous germline de novo mutation in 7 sporadic patients and in 1 patient without family history of MCLMR, but for whom no parental DNA was available for testing. Thus, 87,5% of our sporadic patients with a KIF11 mutation had a proven de novo mutation. This represents 41% (7/17) of sporadic index patients in this study. Taking into account all index patients with a KIF11 mutation, sixteen sporadic patients have been reported to have a de novo mutation (7 in the current study, and 9 in previously published works), indicating that KIF11 has a high de novo mutation rate: 16 de novo/52 mutation-positive index patients; i.e. 31% across studies. Several de novo mutations have also been identified in other genes mutated in microcephaly (i.e. CTNNB1 [20] and TUBA1A [21]) and in primary lymphedema, such as SOX18, FOXC2 and VEGFR3 [22-24].

No KIF11 mutation-negative familial cases were identified; however, about half of the sporadic cases were KIF11 mutation-negative. These may be explained by several hypotheses: (1) Deletions or deep-intronic mutations in KIF11, which were not assessed in the study; (2) Somatic/mosaic KIF11 mutations that are not detectable in the blood by Sanger sequencing: mosaicism as a cause of MCLMR can be investigated by sequencing tissue and blood-DNA using targeted deep sequencing; (3) Existence of a mimicking disorder: many different genes can cause microcephaly, including those implicated in autosomal recessive primary microcephaly and syndromic microcephaly [25], or primary lymphedema [26]. The presence of motor delay in patients XVIII-10 and XIII-10 may suggest Cohen syndrome (OMIM 216550) [27]. The Madelung deformity and hydrops fetalis observed in two other patients may also be indicative of a distinct syndrome, as they have never been described in MCLMR. Eight out of nine of our negative cases had microcephaly and lymphedema; (4) Novel genetic cause for MCLMR in mutation-negative patients. Overall, our results suggest that KIF11 is the major causative gene for MCLMR, with no evidence of locus heterogeneity in familial cases.

The most frequent clinical signs and symptoms in patients with a KIF11 mutation are microcephaly (91%), eye anomalies (72%) and intellectual disability (67%) (Figure 3 and Table 1). MCLMR patients may have a structurally abnormal brain i.e. pachygyria (Table 1: Patients VII-10 & XI-100) [9]. Lymphedema is present in 47% of mutation-positive patients (Table 1).

Although none of the individual symptoms is fully penetrant, the overall penetrance of MCLMR is high (83/87; 95%). Only 4 asymptomatic mutation carriers have been identified amongst 87 individuals tested. Patients in the same family (Family II, IV, V, VI, and XI), with the same mutation may however have different phenotypes, demonstrating variable expressivity.

The 44 mutations identified so far in KIF11 (nine missense, twenty-five premature stop codons and ten splice-site changes) are scattered across the length of gene (Figure 2). There are no clear phenotypic differences between missense mutation carriers and those with premature stop codon mutations. As the most 5’ premature truncation codon occurs at the very 5’ end (amino acid 47), and lymphoblasts had a significantly lower expression of the mutant allele of patient II-10 (c.308+1G>T), all mutations most likely cause loss of EG5 function.

KIF11 mutations have pleiotropic effects with variable penetrance. One possible explanation could be the occurrence of somatic second-hit mutation(s), as we have proven for glomuvenous malformation [28,29] and mucocutaneous venous malformation [30]. Random somatic changes inevitably occur in the cells of all individuals, and result in a double-hit situation in some cells in patients with a germline mutation. The resulting phenotypes would depend on the types and developmental stages of cells affected.

There are nine amino acid substitutions identified in MCLMR patients. One of them affects the initiating methionine at position 1 (p.Met1Thr) most likely causing non-translation of KIF11. Seven of the remaining eight amino acid substitutions occur in the kinesin motor domain (p.Tyr82Phe, p.Phe144Leu, p.Arg221Gly, p.Arg234Cys, p.Ser235Cys, p.His244Tyr, and p.Glu344Lys) and the last one in the BimC Box (p.Arg944Cys), implicated in multimerization (Figure 2). Both domains have been shown to be important for KIF11 function during mitosis [11]. The kinesin motor domain contains amino acids important for tertiary structure. Arg234 makes a salt bridge with Glu270 to close the active site essential for ATP hydrolysis [31]. Substitution of Arg234 for a cysteine likely perturbs closure of the pocket and ATP hydrolysis. The impact of Ser235Cys is likely similar due to its vicinity to Arg234. In addition, the serine-to-cysteine change would affect polarity and thus modify the stabilization of a molecule of water in the nucleotide-binding pocket. Moreover, in the 3D structure (PDB ID: 3hqd), Arg221 is located in front of Ser235. Substitution for glycine could influence the structure of the nucleotide-binding pocket. Tyr82, Phe144 and Glu344 are located in an alpha-helix and His244 in a central beta-sheet in the kinesin domain. It is unclear whether these mutations, and the BimC Box mutation, affect stability, structure and/or function.

In vitro mutagenesis (Pro131Ala, Thr926Ala and Ser1033Ala) showed that the N-tail and the BimC domain are important for intra-cellular localization of the protein [32]. Thr926 of the BimC domain is phosphorylated by CDK1 (Cyclin-dependent kinase 1), whereas Ser1033 is partially phosphorylated by NEK6 (Never in mitosis gene a -related kinase 6). Phosphorylation of these sites is important for the mitotic function of EG5 [33,34]. Depletion of KIF11 causes mitotic arrest. This can be rescued by re-expression of the protein, but not by a Thr926Ala mutant, which lacks the ability to interact with microtubules [33]. The Arg944Cys mutation likely causes dysfunction of this same domain.

EG5 is a plus end-directed microtubule motor. It contributes to the establishment and maintenance of bipolar spindles during mitosis and meiosis that push the poles apart [35]. This is required for the separation of duplicated centrosomes. EG5 is considered as a slow kinesin, which slows down the spindle separation during mitosis. It is dispersed in the cytoplasm during the interphase, regroups at the spindle poles during prophase and stays alongside the spindle in metaphase. In mice, Eg5 is expressed in proliferative tissues. Disruption by homozygous deletion of Kif11 in mice resulted in embryonic lethality due to defective implantation in the uterine wall [36]. Microcephaly, ocular anomalies and œdema were not reported in heterozygous embryos. Morpholino-based knock-down of Kif11 also led to embryonic lethality, or severe œdema and circulation defects [37]. Inhibition of EG5 by dimethylenastron (DMN) or ispinesib diminished endothelial cell proliferation (HUVEC, hCMEC/D3 and LEC) and embryonic angiogenesis [37].

EG5 disruption is implicated in neuronal development and the phenotype observed in MCLMR adds KIF11 to the long list of genes implicated in intellectual disability. Inhibition of EG5 in cultures of synaptic neurons led to faster growth of axons, with an abnormal maturation [38]. Intellectual disability due to EG5 dysfunction could thus be explained by neuronal inability to grow and/or to connect correctly. The etiopathogenesis of the eye anomalies could be due to defective neurogenesis or angiogenesis.

EG5 could cause lymphedema by a microtubule-dependent function in trafficking the major lymphangiogenic tyrosine kinase receptor, VEGFR3, in endocytotic vesicles [39]. On the other hand, it could be that the endothelial cell function of EG5 is independent of the kinesin-motor domain, as is the case for KIF26A (kinesin family member 26A), KIF4 (kinesin family member 4A/B) and MAP2 (microtubule-associated protein 2) [40-42]. In fact, EG5 plays also a role in protein translation [43]. The expression of KIF11 in LECs is regulated by FOXC2, a major lymphangiogenic transcription factor [44]. It is however not known if MCLMR patients have lymphatic hyperplasia as observed in FOXC2-mutated patients (lymph and venous reflux and failure or absence of lymphatic and venous valves) or hypoplasia as seen in VEGFR3-mutation caused Nonne-Milroy lymphedema [45,46]. Lymphoscintigraph of one MCLMR patient showed no significant main-tract filling suggesting a peripheral lymphatic-vessel dysfunction [6].

KIF11 mutations are identified in all familial cases of MCLMR. De novo mutation rate is high for KIF11: 16 out of 22 (73%) sporadic cases. Mutations most likely cause loss of EG5 function. There is no genotype-phenotype correlation. The phenotype of the KIF11 mutation-negative sporadic patients may be due to non-detected germline or mosaic KIF11 mutations and/or defects in (a) gene(s) causing (a) mimicking disorder(s), as microcephaly and primary lymphedema are genetically highly heterogeneous. A better elucidation of the pathophysiological mechanisms is needed to enable development of targeted therapies for MCLMR.