MOLECULAR IMPLICATIONS OF AFADIN AND AURORA-A¿MEDIATED PHOSPHORYLATION OF NUMA IN SPINDLE ORIENTATION

Oriented cell divisions contribute to tissue morphogenesis and homeostasis. Planar divisions occurring with the spindle within the epithelial plane enlarge sheets and tubules, while asymmetric cell divisions with the spindle aligned to the apico-basal polarity axis sustain differentiation programs. Several pathways have been involved in establishing correct spindle orientation, both in cultured cells and in vivo. Most of these pathways impinge on the evolutionarily conserved Gαi/LGN/NuMA complexes that orient the spindle by generating pulling forces on astral microtubules (MTs), via direct interaction of NuMA with the MT-motors Dynein/Dynactin. My PhD projects focused on the molecular mechanisms underlying the spindle orientation function of Afadin, and on the relevance of NuMA phosphorylation by Aurora-A for spindle orientation. During planar cell divisions, Gαi/LGN/NuMA assemblies are restricted to the lateral cortex, for molecular reasons that are still unclear. Studies conducted during this thesis indicate that LGN interacts directly with the junctional and F-actin binding protein Afadin, and define the TPR domain of LGN (hereon LGNTPR) and a C-terminal peptide of Afadin (AfadinPEPT) as the minimal interacting regions retaining micromolar binding affinity. The crystal structure of the LGNTPR-AfadinPEPT fusion protein shows that the AfadinPEPT threads along the LGNTPR superhelix with opposite chain directionality, similarly to what observed for LGN in complex with other ligands, including NuMA. Consistently, we provided evidence that Afadin competes with NuMA for binding to LGN. Afadin knock-down in HeLa cells leads to reduced LGN cortical levels, and unexpectedly also to complete loss of cortical NuMA and Dynein/Dynactin, and hence spindle misorientation. Importantly, we discovered that Afadin interacts concomitantly with F-actin and LGN in vitro. Furthermore, we showed that loss of Afadin impairs correct cystogenesis of Caco-2 cells, suggesting that it plays essential functions in epithelial planar cell divisions. Altogether our data suggest a model whereby in metaphase Afadin mediates cortical recruitment of Dynein/Dynactin, by targeting LGN at the lateral cortex via direct and concomitant interaction with LGN and with cortical F-actin. Later, LGN engages with NuMA and Dynein/Dynactin to exert pulling forces on the mitotic spindle. Thus, Afadin represents the first described mechanical anchor between the acto-myosin cell cortex and the Dynein/Dynactin MT-motors. Besides being spatially regulated, the cortical recruitment of Gαi/LGN/NuMA is timely controlled by mitotic kinases coordinating spindle orientation with mitotic progression. It was reported that the activity of the mitotic kinase Aurora-A is required for correct spindle orientation in human cells in culture, and that NuMA is among its phosphorylation targets. However, whether NuMA is phosphorylated directly by Aurora-A and how molecularly its kinase activity affects spindle orientation was still unknown when we started our studies. Analyses in HeLa and RPE-1 cells revealed that, in metaphase, depletion or inhibition of Aurora-A leads to aberrant accumulation of NuMA at the spindle poles and loss from the cortex, despite LGN localizes normally at the cortex. FRAP experiments revealed that Aurora-A governs the dynamic exchange between the cytoplasmic and the spindle pole-localized pools of NuMA. Our experiments in vitro and in cells showed that Aurora-A phosphorylates directly three serine residues on the C-terminus of NuMA, and mutation of Ser1969 into alanine recapitulates the aberrant polar accumulation of NuMA and the spindle orientation defects observed upon Aurora-A inhibition. Thus we concluded that phosphorylation on Ser1969 of NuMA by Aurora-A controls NuMA distribution between the spindle poles and the overlying cortex, and allows proper spindle orientation. Intriguingly, Ser1969 lies within a previously characterized microtubule (MT)-binding domain. However, in vitro co-sedimentation and bundling assays revealed that the binding affinity of NuMA for MTs is unaltered by Aurora-A-mediated phosphorylation, suggesting that unphosphorylated NuMA accumulates at spindle poles via a receptor other than MTs. Most interestingly, with our experiments we also identified a new MT-binding domain of NuMA positioned downstream of the LGN binding motif. This result is consistent with our finding that NuMA can simultaneously interact with LGN and MTs. Based on these data, we propose that in metaphase the MT-binding activity of NuMA may contribute to anchor astral MT +TIPs at cortical sites together with LGN.