To perform their physiological functions in actin assembly, formins need to be appropriately localized. Despite this critical aspect, remarkably little is known about how they are localized. It is clear for the mDia proteins that the N-terminal region is important for localization (Seth et al., 2006
; Brandt et al., 2007
; Gorelik et al., 2011
). In the recent study of Gorelik et al. (2011)
, a basic domain N-terminal to GBD was found to be important for plasma membrane localization, and the small GTPase Rif could also contribute significantly, implying that these two factors work together for correct localization.
We took this type of analysis one step further in our identification of regions in Bni1p involved in its localization and sought to understand the functions of the identified localization domains. This analysis showed an unexpected complexity by revealing that Bni1p has at least four independent localization domains that can target the protein to the cell cortex in a manner independent of either endogenous Bni1p or filamentous actin. Three of these localization domains reside in the N-terminal 1200 residues of the protein and are independent of an actin cytoskeleton. LD1 (residues 1–333) coincides well with a region homologous to the structurally defined GBD domain. Formins of the Diaphanous family are regulated by binding to active Rho proteins, and this binding occurs through both the GBD region and the DID (Rose et al., 2005
). The need for dimerization probably reflects a low affinity of the isolated domain for the cell cortex, so that only with the higher affinity of a dimer is the interaction clear. Native Bni1p is presumably a dimer, as the DD domain, as well as the FH2 domain, can dimerize (Moseley et al., 2004
; Xu et al., 2004
; Otomo et al., 2005
). The second localization domain, LD2 (residues 334–834), includes the DID, the DD, and a region predicted to form a coiled coil. By virtue of the DD and predicted coiled-coil domain, this region almost certainly dimerizes. This region seems to be the most robust localization region, as the internal deletion constructs retaining it localize well in both SPA2
cells, whereas those lacking it localize poorly, especially in spa2Δ
cells. Finally, LD3 encompassing the SBD has been identified as being important for Bni1p localization (Fujiwara et al., 1998
; Ozaki-Kuroda et al., 2001
). We identified all of these localization domains in cells in which the constructs were overexpressed. The fact that they localized suggests that the factors that localize them to the bud cortex or bud tip are likely to be much more abundant than endogenous Bni1p. In support of this conclusion, attempts to disrupt the function of endogenous Bni1p by expression of N-terminal constructs failed to cause any obvious phenotype. In addition to the three N-terminal localization domains, we found that the complementary C-terminal part LD4 (1231–1963), comprising the FH1, FH2, and C-terminal region, also has a weak localization domain to incipient and small buds. This localization is independent of either endogenous Bni1p or the presence of filamentous actin.
Having identified four localization domains in Bni1p, we sought to investigate their functions. Our original strategy assumed that correct localization is an essential part of formin function, so a deletion analysis of Bni1p regions in a bnr1Δ
cell should reveal the essential localization domain. Accordingly, we made a series of N-terminal deletions in the chromosomal copy of Bni1p and assessed their effect on growth in BNR1
cells. During this analysis, we made the unexpected discovery that yeast can grow remarkably well in the absence of a localized formin, namely when the only formin protein present was the delocalized FH1 plus FH2 domain of either Bni1p or Bnr1p (Gao and Bretscher, 2009
). Despite this result, the N-terminal deletions of Bni1p were very informative. Deleting up to the SBD had little effect on growth, whereas deleting more of the N-terminal region to remove the SBD was deleterious to growth and resulted in quite depolarized cells that, even in the presence of Bnr1p, conferred a more severe phenotype than bni1Δ
. Expression of this region of Bni1p therefore confers a dominant-negative phenotype.
Before this study, the only identified localization mechanisms for Bni1p were provided by Spa2p through the SBD (Fujiwara et al., 1998
; Ozaki-Kuroda et al., 2001
) and potentially through the polarisome-associated Bud6p to the C-terminal region (Ozaki-Kuroda et al., 2001
). Because N-terminal deletion constructs containing the SBD were relatively healthy, whereas those lacking it were dominant negative, we suspected that combining spa2Δ
with our Bni1p constructs containing the SBD should phenocopy those lacking it. spa2Δ
indeed compromised growth of Bni1p constructs with the SBD. It is surprising that the dominant-negative Bni1p constructs that lack the SBD are synthetically lethal with spa2Δ
, indicating that Spa2p performs a function in addition to binding the SBD of Bni1p. One possibility is that Spa2p is still contributing to Bni1p localization through Bud6p, which binds both Spa2p and the C-terminal region of Bni1p (Evangelista et al., 1997
; Sheu et al., 1998
; Moseley and Goode, 2005
Because Bni1p localization is not essential for viability in bnr1Δ cells, we explored the possibility of exploiting a situation in which Bni1p localization is important even in the presence of Bnr1p. Spindle orientation is required for proper nuclear segregation and depends both on the Bni1p/Myo2p/Kar9p pathway and the dynactin/dynein pathway. In the absence of the dynein/dynactin pathway, Bni1p/Myo2p/Kar9p becomes the critical mechanism for spindle orientation. We therefore combined our N-terminal Bni1p deletion series with arp1Δ cells, lacking a component of the dynactin complex. Indeed, we found a modest spindle orientation defect (as determined by benomyl sensitivity and the fraction of mother cells with two or more nuclei) in the mutants even in ARP1 cells that increased as additional localization domains were deleted from the N-terminal region of Bni1p, and these defects were strongly enhanced in arp1Δ cells. Therefore all the identified N-terminal localization domains contribute to the Bni1p-dependent spindle orientation pathway. As described earlier, Bni1p constructs from which all the N-terminal localization domains were deleted had a dominant-negative effect on growth and were found to be inviable with arp1Δ. The dominant-negative effects could be abrogated by either mutating the Bni1p construct to eliminate its actin-nucleating ability or adding the Cdc42-binding CRIB domain from Gic2p to enhance its localization to growth sites. Therefore the dominant-negative effects are due to inappropriate localization of actin assembly that affects both polarized growth and spindle orientation.
The results presented in this study reveal that Bni1p has four localization domains and they contribute cumulatively to Bni1p's function in polarized growth and spindle orientation. It is surprising that so many independent localization domains exist in one protein. However, multiple N-terminal localization domains seem to be a common property of yeast formin proteins, as Bnr1p has two separate localization domains, encompassing residues 1–466, and 466–733 (Gao et al., 2010
). These findings could suggest that the cell needs very careful and intimate control of Bni1p to coordinate its localization, and hence cable assembly, with all the other events going on at sites of growth. The strict regulation of formin proteins could be common for higher organisms, as they often possess multiple formin isoforms (Higgs, 2005
). Alternatively, the many localization domains may be used under other growth conditions, for example, during shmooing, or during the highly polarized growth that occurs during filamentous growth. The finding of four distinct localization domains implies that four different mechanisms may be involved in localizing Bni1p. Identifying these mechanisms is the next critical step in this analysis.