Mutations in Human Beta-2b Tubulin Result in Asymmetrical Polymicrogyria

Xavier Hubert Jaglin*^1,2, Karine Poirier*^1,2, Yoann Saillour^1,2, Emmanuelle Buhler³, Guoling Tian⁴, Nadia Bahi-Buisson^1,2,5, Catherine Fallet-Bianco⁶, Françoise Phan-Dinh-Tuy^1,2,#, Xiang Peng Kong⁴, Pascale Bomont⁷, Laëtitia Castelnau-Ptakhine^1,2, Sylvie Odent⁸, Philippe Loget⁹, Manoelle Kossorotoff⁵, Irina Snoeck¹⁰, Ghislaine Plessis¹¹, Philippe Parent¹², Cherif Beldjord¹³, Carlos Cardoso⁷, Alfonso Represa⁷, Jonathan Flint¹⁴, David Anthony Keays¹⁵, Nicholas Justin Cowan^4,®, Jamel Chelly^1,2,®.

Polymicrogyria is a relatively common but poorly understood defect of cortical development characterized by numerous small gyri and a thick disorganized cortical plate lacking normal lamination. We show an association between bilateral asymmetrical polymicrogyria and de novo mutations in a b-tubulin gene, TUBB2B, in four patients and a 27 GW (gestational week) fetus. Neuropathological examination of the fetus revealed an absence of cortical lamination associated with the presence of ectopic neuronal cells in the white matter, and in the leptomeningeal spaces due to breaches in the pial basement membrane. In utero RNAi-based inactivation demonstrates that TUBB2B is required for neuronal migration. We also show that two disease-associated mutations lead to an impaired formation of tubulin heterodimers. These observations, together with previous data, demonstrate that disruption of microtubule-based processes underlies a large spectrum of neuronal migration disorders that includes not only lissencephaly/pachygyria, but also polymicrogyria malformations.

Read Also

The crucial role of the tubulin superfamily in diverse cellular processes¹ and the association of TUBA1A mutations with a broad lissencephaly spectrum^2-4 led us to hypothesize that mutations in other tubulin genes that are highly expressed during CNS development might also result in malformations of cortical development. In a previous screen of agyria/pachygyria patients, we excluded the implication of TUBA1B, TUBA1C and TUBB3³. In this study we report the screening of 3 additional candidate tubulin genes (TUBB2A,B,C) in patients with a wide range of cortical dysgeneses, including polymicrogyria (PMG) syndromes associated with epilepsy and/or neurodevelopmental delay (see Material and Methods section). Although no non-synonymous variations were found either in TUBB2A or in TUBB2C (see alignment in Supplementary Fig. 1), heterozygous missense mutations were found in TUBB2B (Fig. 1a) in four unrelated individuals and one fetus. All mutations, c.514T>C (p.S172P), c.629T>C (p.I210T), c.683T>C (p.L228P), c.793T>C (p.F265L) and c.935C>T (p.T312M) affect residues that are rigidly conserved from yeast to human (Supplementary Fig. 2) and reside in exon 4 (Fig. 1a, Table 1). Consistent with a de novo origin of the TUBB2B mutations, none were found in the parents of affected individuals or in 360 normal controls (see referenced polymorphisms in Supplementary Fig. 1).

Brain MRI sequences revealed that all patients share the presence of a complex brain dysgenesis with bilateral, asymmetrical, and anteriorly predominant polymicrogyria (PMG), fusion of the caudate and putamen with internal capsule hypoplasia, corpus callosum agenesis or dysgenesis and, in most cases, cerebellar and pons atrophy (Fig. 1b-j, Table 1, Supplementary Fig. 3). In addition, neurohistopathological analysis of the fetal brain showed asymmetrical bilateral PMG with the absence of a corpus callosum, several nodular cluster of ectopic neurons in both hemispheres and a disorganization of cortical layering (Fig. 1k-r) including the presence of radial columnar heterotopic neurons in the white matter in the two hemispheres (Fig. 1m). These abnormalities strongly suggest migration defects and perturbations of axon tract formation associated with mutations in TUBB2B. In addition to the typical unlayered polymicrogyric cortex, the analysis of the left hemisphere reveals overmigration of MAP2+ neurons through breaches in the pial basement membrane (BM) (Fig. 1l,n,r and Supplementary Fig. 4). Analysis of radial glial fibers revealed a striking disorganization beyond the pial BM (Fig. 1p). These observations recall the phenotype of mice inactivated for Gpr56⁵, a gene associated with bilateral frontoparietal PMG in humans⁶. The macroscopic aspect and folding of the cerebellum appeared similar to the cerebellum of a control fetus of 27 GW. At the microscopic level, many nodular heterotopia were observed in both cerebellar hemispheres.

To investigate the association between TUBB2B mutations and neuronal migration disorders, we studied the expression of Tubb2b during mouse brain development by in situ hybridization and qRT-PCR, and analyzed the consequences on the cortical neuronal migration of TUBB2B loss of function induced by in utero RNA interference.

Sections of embryos at embryonic day 14.5 (E14.5) and 16.5 (E16.5) showed strong labeling restricted to central and peripheral nervous systems (See Supplementary Fig. 5). Using sagittal sections of the brain at E16.5, we found that expression predominated in the cortical plate and also within a thin layer in the subplate, whereas no signal was observed in the marginal zone and fainter labeling was seen in the ventricular and intermediate zones (Supplementary Figure 5a-e). The strong labelling in the developing cortex subsequently decreases after birth although Tubb2b labeling remains intense in the adult cerebellum, hippocampus and olfactory bulb (Supplementary Fig. 5). Taken together these data suggest that Tubb2b is strongly expressed in postmitotic neurons with dominant expression during neuronal migration and differentiation^7-9, and to a lesser extent in progenitor cells. RT-PCR analysis also showed that Tubb2b is expressed in astro-glial cells and C6 glial cells (data not shown).

We used rat in utero RNAi approach¹⁰ to knock-down Tubb2b expression by about 50 % at embryonic day 15.5 (E15.5), a time coincident with the migration of neurons within the cortex, to mimic the consequence of heterozygous loss of function mutations. We tested different small hairpin RNAs (shRNA) and used two shRNAs targeting either the coding sequence (CDS-sh) or the 3’ untranslated region (3’UTR-sh) that repressed Tubb2b expression in vitro by approximately 60%, while scrambled controls did not destabilize the Tubb2b mRNA (Supplementary Fig. 6a-b). These shRNAs combined with a Red Fluorescent Protein (RFP)-encoding reporter construct were electroporated into progenitor cells located of the ventricular zone (VZ) of E15 rat neocortices. These cells give rise to further young neurons expressing the fluorescent protein as they migrate towards the cortical plate. In E20 brain sections, we observed that neurons electroporated 5 days previously with RFP alone reached the cortical plate as expected (Fig. 2a-b). However in utero expression of 3’UTR-sh induced a significant arrest of cells within the sub-ventricular zone (SVZ)/intermediate zone (IZ) (Fig. 2c-d,j-k; stratum 4: F(3,42)=21.716, p<0.0001; stratum 5: F(3,42)=20.394, p<0.0001). To further validate the specificity of our results, we performed in utero RNAi with the coding sequence CDS-sh and found that it leads to the same migration arrest (Fig. 2e, Supplementary Fig. 7). We also showed that the expression of both scrambled-sh controls did not disrupt migration (Fig. 2f,g). Finally, we performed a rescue experiment in which we co-transfected a bicistronic construct driving the expression of Tubb2b and Green Fluorescent Protein (GFP) as a reporter. Although the expression of this construct alone does not alter migration (Fig. 2h), it significantly suppresses the blocking effect of 3’UTR-sh on migration (Fig. 2i,k; strata 8-10: F(3,42)>23.4, p<0.0001). This RNAi-based approach reinforces the evidence that microtubules act as a critical node during corticogenesis and strongly implicates Tubb2b/TUBB2B in neuronal migration.

To define the functional consequences of the mutations in TUBB2B, we examined the potential implications of each of the mutated residues on the known structure of the b-tubulin polypeptide^11,12 (Fig. 3a). S172 resides in a loop that forms part of the guanosine nucleotide-binding site, which when mutated to a proline (S172P) is predicted to disrupt a hydrogen bond and to destabilize the GTP pocket. L228 and F265 are either in the vicinity of or part of the GTP/GDP binding site, with potential consequences for GTP binding. The remaining mutations appear to be less consequential in terms of the GTP binding and overall stability of the protein. Because they are located on the surface, they may interfere with specific partner interactions.

We then investigated the ability of the b-tubulin mutants to produce functional a/b tubulin heterodimers through the complex chaperone-dependent folding pathway^13,14 (Fig. 3b) by transcription/translation in rabbit reticulocyte lysate¹⁵.  All mutant proteins were translated as efficiently as a wild type control (Fig. 3c). In contrast, analysis under native conditions revealed a range of reaction products (Fig. 3d,e) that could be assigned on the basis of their electrophoretic mobilities^14,16. In the case of TUBB2B mutant proteins, these products frequently differed both quantitatively and qualitatively from the wild type control (See Supplementary Table 1). Most conspicuously, two mutant proteins (p.F265L and p.S172P) completely failed to yield the intermediate corresponding to Tubulin-specific Chaperone A (TBCA)/b-tubulin (Fig. 3d,e). In addition, the yield of native a/b heterodimers produced following a chase with added native tubulin was either slightly reduced (in the case of p.I210T and p.T312M), dramatically reduced (in the case of p.L228P and p.F265L), or undetectable (in the case of p.S172P) (Fig. 3e). (For kinetic analyses, see also Supplementary Results and Supplementary Fig. 8).

To examine the mechanism of defective heterodimer assembly of the p.F265L and p.S172P mutant polypeptides in detail, we did reconstituted folding reactions in vitro using purified components^14,16. We observed the generation of TBCD/b-tubulin co-complexes in reactions performed with p.F265L and p.S172P (Fig. 4a) that might be ascribable to the relatively high concentrations of CCT and TBCD in these reactions compared to the more physiological concentrations in reticulocyte lysate. Indeed, when the abundance of TBCD was reduced by a factor of 5 in reactions performed with a constant level of CCT, the yield of the wild type b-tubulin/TBCD co-complex diminished to 25% of the original level, and the F265L b-tubulin/TBCD co-complex was reduced to an undetectable level (Fig. 4c,d). Similar data were obtained for the p.S172P mutation (Supplementary Fig. 9). We conclude that in addition to a dramatic failure of p.F265L and p.S172P CCT-generated folding intermediates to stably interact with TBCA, these mutations also result in a reduced efficiency of intermediate interaction with TBCD (See also Supplementary Results). To assess the competence of mutant heterodimers expressed by transcription/translation to incorporate into microtubules, we analyzed their ability to co-cycle with native brain microtubules in vitro and to co-assemble with microtubules upon transfection into cultured cells. In the cases of p.F265L and p.S172P we observed 1) a very low yield of labeled heterodimers incorporated into microtubules and 2) a further diminution between the first (1) and second (2)  cycles of polymerization/depolymerization, suggesting significant instability (Fig. 5a). Upon heterologous overexpression by transfection in cultured cells, we found that 3 mutants (p.I210T, p.L228P, p.T312M) behaved indistinguishably from the wild-type protein in that they were efficiently incorporated into interphase microtubules (data not shown). In contrast, in the case of the p.S172P and p.F265L mutants, there was scant evidence of incorporation into well-defined microtubules (Fig. 5b) (See also Supplementary Results for assessment of microtubule dynamics). Taken together, these experiments show that p.S172P and p.F265L are significantly compromised in their ability to properly assemble into microtubules in vivo, and are consistent with our in vitro transcription/translation experiments in which neither of these mutant proteins yielded significant amounts of polymerization competent heterodimers (Fig. 3e, 5a).

We further sought to test whether expression of the p.S172P and unrelated p.T312M mutants in vivo could complement the phenotype caused by knock-down of Tubb2b. We electroporated the pCAGIG-p.S172P(or p.T312M)-Tubb2b-IRES-GFP construct either alone or in combination with the 3’UTR-sh and analyzed the position of electroporated cells within the cortex. We found that, although expression of each mutant alone does not massively affect migration (Fig. 5e,g), expression of each mutant in the knock-down context maintains the cells stalled within the SVZ/IZ (Fig. 5f,h) and fails to complement the RNAi effect.

In this study, we implicate mutations in TUBB2B as causative of brain malformations encompassing asymmetrical PMG associated with an unlayered cortex, heterotopic neuronal cells in the white matter and neuronal overmigration through the pial BM. It is worthwhile mentioning that corpus callosum dysgenesis, dysmorphic basal ganglia, cerebellum dysplasia and brainstem hypoplasia are common features to TUBA1A-related agyria/pachygyria and TUBB2B-related PMG. Mutations in GPR56 and SRPX2 genes are also known to be associated with PMG. However the diagnosis of TUBB2B-related PMG could be evoked through assessment of clinical and imaging criteria highlighted in Supplementary Table 2.  

Our in vitro data show that the five newly discovered disease-associated TUBB2B mutations involve a spectrum of tubulin heterodimer assembly defects (summarized in Supplementary Table 1) leading to loss of function in the cases of p.S172P and p.F265L. The two most severe defects observed in vitro (p.S172P and p.F265L) are associated with the most and the least severe clinical phenotypes, respectively, suggesting that there is no simple correlation in the panel of 5 mutated patients. We also establish that loss of function is associated in vivo with defective migration and mislocalization of developing neurons within the cortex, suggesting that TUBB2B-related forms of PMG are primarily due to haploinsufficiency, though dominant negative effect can not be excluded for the p.S172P mutation. For those mutations that have less or no apparent impact on tubulin heterodimer assembly, the functional defect seems likely to concern a subtle effect on either microtubule dynamics, or on the interaction with one or more microtubule interacting proteins (MAPs) that are critical for proper cortical neuronal migration, or both. The cellular consequences of TUBB2B mutations, and the mechanisms by which this results in polymicrogyria, an unlayered cortex, and heterotopic neuronal cells remain unknown. We propose that the neuropathophysiology of TUBB2B-related PMG might result from a combination of both neuronal migration impairment and radial glia dysfunction that lead respectively to ectopic neurons in the white matter and cerebellum, and to pial membrane breaches (see Supplementary Discussion).

In higher eukaryotes, a- and b-tubulins are encoded by a multigene family that are evolutionary conserved¹⁷ among different species and are differentially expressed^18,19. To explain the need for these highly conserved multiple genes, it has been hypothesized that the different isotypes may be required to form specific sets of microtubules that carry out unique functions¹⁸. Though this hypothesis is still a matter of debate, our data showing that mutations in TUBA1A and TUBB2B are associated with different gyral abnormalities argue in favor of specific roles of TUBB2B and TUBA1A during corticogenesis and neuronal migration (see Supplementary Table 2). It is also possible that subtle differences of spatio-temporal profiles of TUBA1A and TUBB2B expression (i.e. populations of interneurons, radial glia cells and astro-glial cells) may account for some of the cortical phenotypic differences.