Mutations in the dynein cargo adaptor BICD2 have been linked to SMALED2 (Koboldt et al., 2020). Mutations in the heavy chain of the dynein motor have also been implicated in a version of this disorder (Chan et al., 2018; Das et al., 2018), suggesting that defects in dynein-mediated transport contribute to its etiology. However, the molecular and cellular mechanisms underlying SMALED2 pathogenesis remain poorly understood. Previous studies have characterized mutations within the first coiled-coil domain of BICD2, a region responsible for interactions with dynein and dynactin. These analyses elegantly demonstrated that mutants such as BICD2_N188T result in dynein hyperactivity (Huynh and Vale, 2017). In addition to these mutants, however, recent studies have identified several SMALED2-associated alleles within the C-terminal cargo-binding domain of BICD2 (Ravenscroft et al., 2016; Synofzik et al., 2014). Given BICD2’s role as a dynein cargo adaptor, these findings raise two important questions: (1) Is dynein hyperactivity a common feature of SMALED2-associated BICD2 mutations? and (2) Do these mutations alter the interactome of BICD2 relative to the wild-type protein? The goal of this study was to address these questions and elucidate potential molecular consequences of SMALED2-associated BICD2 mutations.
BICD2 is one of the best characterized dynein cargo adaptors. However, most studies involving BICD2 have focused on the mechanism by which this adaptor activates dynein for processive motility. Relatively little is known regarding the cargo that is linked to dynein by BICD2. In Drosophila, BicD links the RNA-binding protein Egalitarian (Egl) with dynein for transport of specific mRNAs in the oocyte and embryo (Dienstbier et al., 2009; Goldman et al., 2021; Goldman et al., 2019; Mach and Lehmann, 1997; McClintock et al., 2018). Loss of either BicD or Egl compromises transport of these mRNAs and consequently results in defective oogenesis or embryogenesis. The first definitive cargo identified for mammalian BICD2 was the small GTP-binding protein, RAB6A (Matanis et al., 2002). Despite the ability of BICD2 to directly bind RAB6A, most vesicles containing RAB6A move towards the plus end of microtubules, suggesting that their transport is primarily driven by the Kinesin-1 motor, KIF5B (Grigoriev et al., 2007). Other cargos that have been shown to directly bind BICD2 are RANBP2, a nucleoporin, and Nesprin-2 (SYNE2), a LINC complex component involved in linking dynein and kinesin to the nuclear envelope (Gonçalves et al., 2020; Splinter et al., 2010).
In order to determine whether SMALED2 alleles of BICD2 are associated with interactome changes, it was therefore critical for us to determine the interactome of wild-type BICD2. This was done using the promiscuous biotin ligase miniTurboID (mTrbo). In comparison to an RFP-mTrbo control, BICD2-mTrbo resulted in the biotinylation and purification of numerous known interacting partners including RANBP2, as well as several components of the dynein motor. One interesting group of potentially novel interacting proteins was components of the HOPS complex, a six-subunit complex of proteins involved in endocytic trafficking (Spang, 2016). Four of the six HOPS components were identified in the wild-type BICD2 interactome, with VPS41 being the fifth most enriched protein. However, unlike RANBP2, RAB6A, and NESPRIN-2, all of which are able to bind the isolated BICD2 cargo binding domain (Gonçalves et al., 2020; Matanis et al., 2002; Splinter et al., 2010), the HOPS complex components were only able to bind full-length BICD2. The BICD2 cargo binding domain was therefore necessary but not sufficient for interaction with HOPS components. In addition, contrary to our initial hypothesis that VPS41 was the direct binding partner between BICD2 and the HOPS complex, BICD2 retained its interaction with VPS16 and VPS18 in cells depleted of VPS41. This suggests that BICD2 likely recognizes a domain or motif present in several HOPS proteins. We attempted to use Alphafold2 multimer to predict the relevant domain within HOPS proteins that interact with BICD2. Although Alphafold2 was able to generate a high confidence prediction of the interaction site between BICD2 and RAB6A, consistent with published results (Zhao et al., 2024), it failed to produce a high confidence prediction for the BICD2-HOPS complex interaction (data not shown). Thus, although we were able to validate the in vivo association between BICD2 and VPS41, VPS16, and VPS18, we are not able to conclude whether BICD2 is capable of directly interacting with these proteins. To the best of our knowledge, this is the first example of BICD2 interacting proteins that display this binding characteristic. The ScaC protein from the intracellular pathogen Orientia tsutsugamushi was recently also shown to interact with BICD2, and although the binding site of ScaC was different from that used by RANBP2 or RAB6A, it was still able to interact with the isolated cargo binding domain of BICD2 (Manigrasso et al., 2025).
Another unusual aspect of the BICD2-HOPS complex interaction is that it does not appear to be linked to dynein-mediated trafficking. Depletion of dynein heavy chain resulted in the peripheral distribution of GFP-VPS41 and LAMP1 vesicles, indicative of a reduction in minus end transport, and a net gain in plus end directed transport. By contrast, depletion of BICD2 resulted in the perinuclear accumulation of lysosomal vesicles that were mostly immotile. Interestingly, however, overexpression of BICD2 caused the outward spreading of LAMP1 vesicles, a process that depends on KIF5B (Guardia et al., 2016). Previous studies have shown that BICD2 is also able to interact with KIF5B via a central coiled coil domain (Grigoriev et al., 2007; Hoogenraad and Akhmanova, 2016). A recent report suggests that Drosophila BicD is capable of interacting with and activating the motility of Kinesin-1, the fly homolog of KIF5B (Ali et al., 2025). Consistent with the notion that BICD2 might link late endosomal vesicles with KIF5B, depletion of KIF5B in BICD2 overexpressing cells restored the normal localization of LAMP1 vesicles. Additional studies will be required to determine whether BICD2 is capable of directly interacting with these vesicles and whether these vesicles are directly linked to KIF5B by BICD2.
The motility of LAMP1 vesicles has some similarity to the transport of RAB6A exocytic vesicles. RAB6A vesicles are transported from the area of the Golgi towards the cell periphery in a KIF5B-dependent manner, and loss of either kinesin-1 or dynein results in a sharp reduction in the number of motile particles (Grigoriev et al., 2007). In addition, mutations in BICD2 that compromise binding to RAB6A also result in vesicles that are largely immotile (Zhao et al., 2024). Thus, in the case of LAMP1 and RAB6A vesicles, instead of resulting in an increased rate of minus end transport, loss of BICD2 results in compromised vesicle motility, indicating that coordination between opposite polarity motors is critical for their motility.
As noted earlier, mutations in the CC1 region of BICD2 hyperactivate dynein (Huynh and Vale, 2017). Our findings indicate that this property is also shared by BICD2_R694C and BICD2_R747C, mutations present within the C-terminal cargo binding domain. In the absence of cargo, BICD2 is thought to exist in an inhibited conformation due to intramolecular interactions between the N and C termini of the protein (Figure 1B; Liu et al., 2013; Terawaki et al., 2015; Wharton and Struhl, 1989). Cargo binding to the C-terminus of BICD2 counteracts the intramolecular interaction, enabling N-terminal residues within BICD2 to bind the dynein/dynactin complex (Goldman et al., 2019; Huynh and Vale, 2017; Liu et al., 2013; McClintock et al., 2018; Sladewski et al., 2018). How might mutations in BICD2 result in dynein hyperactivation? One possibility is that these mutations disrupt the autoinhibited state of BICD2, effectively causing BICD2 to be present in a more open and uninhibited conformation that promotes dynein/dynactin binding. Molecular dynamics simulations suggest that the R747C substitution causes a registry shift in the coiled coil, likely destabilizing this domain and thus disrupting the intramolecular interaction between the N and C termini of BICD2 (Cui et al., 2020). Another possibility is that the hyperactivation of dynein results in reduced binding between BICD2 and KIF5B. Our results are consistent with this scenario and suggest that the net effect of dynein hyperactivity results in three molecular changes; reduced intramolecular BICD2 interaction, increased interaction between BICD2 and dynein, and reduced interaction between BICD2 and KIF5B.
In addition to hyperactivating dynein, all three mutations, including BICD2_N188T, alter the BICD2 interactome. This finding was unexpected for BICD2_N188T because this mutation is not within the cargo binding domain. One possible explanation for this phenotype is that BICD2_N188T is present in a more open conformation, and this change affects its binding properties. Another possibility that is not mutually exclusive is that the different binding profile results from the altered localization of BICD2_N188T within the cell. In comparison to wild-type BICD2, we generally observed greater centrosomal enrichment of BICD2_N188T. In comparing the three mutants, the general trend was that more proteins displayed a reduced interaction with the SMALED2 mutants in comparison to wild-type BICD2. Among the three mutants analyzed, BICD2_R747C displayed the most drastically altered interactome. This mutant displayed reduced association with RANBP2, importin beta, and HOPS complex components. Interestingly, this mutant also displayed numerous gain-of-function interactions. For instance, although minimal binding was observed between wild-type BICD2 and GRAMD1A, this protein abundantly interacted with BICD2_R747C. GRAMD1A is involved in non-vesicular transport of accessible cholesterol from the plasma membrane to the ER and is often concentrated at sites of plasma membrane-ER contact (Besprozvannaya et al., 2018; Sandhu et al., 2018). However, in cells expressing BICD2_R747C, this localization pattern was disrupted and GRAMD1A co-localized with BICD2_R747C adjacent to the centrosome.
The GRAMD1 family consists of three isoforms: GRAMD1A, GRAMD1B, and GRAMD1C. Interestingly, our interactome analysis only identified GRAMD1A as a gain-of-function interaction partner with BICD2_R747C. It is unclear whether GRAMD1B and GRAMD1C also interact with BICD2_R747C. However, given that GRAMD1 proteins can form hetero oligomers (Naito et al., 2019), the BICD2_R747C-induced mislocalization of GRAMD1A could potentially affect the distribution of other GRAMD1 isoforms as well. The GRAMD1 proteins function to sense excess accessible cholesterol in the plasma membrane and to mediate the transport of this cholesterol to the ER. This reduces the rate of new cholesterol synthesis by the ER, enabling the cell to maintain cholesterol homeostasis (Sandhu et al., 2018). It will be interesting to determine whether endogenous GRAMD1A is mislocalized in motor neurons of SMALED2 patients with the BICD2_R747C mutation, and if this results in an expanded accessible pool of cholesterol at the plasma membrane.
A recent study by Yi and colleagues examined the effect of the BICD2_R694C mutation on cargo binding (Yi et al., 2023). Using in vitro experiments, they found that this mutation enhanced RANBP2 binding while having no effect on NESPRIN-2 binding (Yi et al., 2023). Our results using full-length BICD2 are consistent with this finding. We also observed slightly higher binding of BICD2_R694C to RANBP2. However, due to experimental variability, the increase was not statistically significant. The authors also examined cargo binding using a BICD2 double mutant (F743I/R747C). Consistent with our results, this mutant displayed greatly reduced binding to RANBP2, but bound NESPRIN-2 at a much higher level than the wild-type protein (Yi et al., 2023). NESPRIN-2 was not identified as an interacting partner in our study for the wild-type protein or the BICD2_R747C mutant, possibly due to its low expression level in HEK293 cells. Nevertheless, these findings, along with our interactome analysis, indicate that mutations in the cargo binding domain of BICD2 can result in loss- and gain-of-function interactions.
In conclusion, our study is the first to comprehensively examine the interactome of wild-type BICD2 and to identify changes that occur in SMALED2 linked mutant alleles of BICD2. We find that not only are mutations within the cargo binding domain associated with interactome changes, but these mutations are also capable of hyperactivating dynein. Some limitations of this study are worth noting. In the current study, we chose to determine the BICD2 interactome in HEK FLP-In cells (embryonic kidney cells). These cells were chosen because they enabled us to precisely integrate wild-type and mutant alleles of BICD2 at a specific locus. It also enabled us to expand cultures of these cells to levels that were sufficient for proteomic analysis. However, the main cell type affected in patients with SMALED2 is motor neurons. Primary motor neurons are harder to culture to scale and to genetically manipulate to express the desired wild-type or mutant BICD2 transgenes. Thus, although motor neurons were not used in our study, the next significant challenge will be to perform these types of experiments using motor neurons. In addition, although our study identified interactome changes between wild-type and mutant alleles of BICD2, we cannot conclude whether these changes are causative for the symptoms associated with SMALED2. Patients diagnosed with this disorder display a range of symptoms, from relatively mild to more severe (Frasquet et al., 2020; Koboldt et al., 2020). Even patients with the same genetic mutation can display a range of phenotypes (Storbeck et al., 2017). Furthermore, disease symptoms can result from one or two interactome changes that are critical for the health of motor neurons. Alternatively, symptoms might also be caused by many small changes in the interactome that cumulatively affect the health of motor neurons. Lastly, because SMALED2 is an autosomal dominant disorder, patients express wild-type and mutant versions of BICD2 in the same cell. Thus, to accurately model this disorder, studies will need to be conducted in motor neurons that are genetically edited to express disease-associated mutations in a heterozygous state.