In order to test our hypothesis that SMALED2 associated mutations in the cargo binding domain of BICD2 are associated with interactome changes, it was first critical for us to define the interactome of wild-type BICD2. Although BICD2 has been known to function as a dynein cargo adaptor, only a few proteins have been shown to interact with BICD2 (Hoogenraad and Akhmanova, 2016; Olenick and Holzbaur, 2019). A more comprehensive interactome of BICD2 was defined by Redwine and colleagues using proximity biotin ligation (Redwine et al., 2017). However, the goal of that study was to more broadly define the dynein interactome and potential novel BICD2 interacting proteins were not validated by follow up studies.

To begin our analysis, we generated stable cell lines expressing either RFP tagged mini-TurboID (RFP-mTrbo) or full length BICD2 tagged on the C-terminus with mini-TurboID (BICD2-mTrbo). TurboID is a proximity biotin ligase that was generated by selecting mutants of BioID that displayed much higher levels of activity in comparison to the original enzyme (Branon et al., 2018). mini-TurboID is a smaller version of TurboID, the biotinylation activity of which is more tightly coupled to the addition of exogenous biotin (Branon et al., 2018). HEK293 FLP-In T-Rex cells were used for this experiment because this enabled us to integrate the constructs at the same genomic locus across all cell lines. In addition, expression of the fusion proteins was inducible upon the addition of tetracycline. This prevents any toxic effects that might arise from constitutive expression of SMALED2 alleles of BICD2.

Cells expressing either RFP-mTrbo or BICD2_WT-mTrbo were grown to scale and proximal proteins were labeled using biotin. Biotinylated proteins were purified using streptavidin magnetic beads and identified using mass spectrometry, with the entire experiment being conducted in triplicate. Proteins enriched in the BICD2_WT-mTrbo pellet at least two-fold in comparison to the RFP-mTrbo pellet and having a p value of at least 0.05 were considered potential BICD2 interacting partners (Fig. 1C, C’, Supplemental table 1). This list, which consists of 66 proteins, included several dynein components and known BICD2 interacting proteins such as RANBP2, RANGAP1, and PLK1 (Fig. 1D) (Gallisa-Sune et al., 2023; Splinter et al., 2010). In addition, four components of the HOPS complex were also specifically enriched in the BICD2_WT-mTrbo pellet (Fig. 1D). The HOPS complex consists of six subunits and is involved in the fusion of late endosomes with lysosomes (van der Beek et al., 2019). Consistent with known BICD2 functions, a cellular component GO analysis indicated an enrichment of terms such as “cytoplasmic dynein complex”, “nuclear pore” and “vesicle tethering complex” (Fig. 1E).

Surprisingly, we failed to identify RAB6A in our interactome analysis. RAB6A was the first cargo identified for BICD2 using a yeast-two hybrid screen (Matanis et al., 2002). Previous studies have shown that GTP bound RAB6A is the preferred binding partner for BICD2 (Matanis et al., 2002). Thus, in order to determine whether RAB6A is capable of associating with full length BICD2-mTrbo, we transfected cells expressing either RFP-mTrbo, full length BICD2-mTrbo, or a version of BICD2-mTrbo lacking the CC3 domain along with a GTP locked mutant of RAB6A (GFP-RAB6A Q72L). Consistent with published results, GFP-RAB6A Q72L specifically associated with full length BICD2-mTrbo (Supplemental Fig. 1A). Based on this, we conclude that although a GTP locked version of RAB6A is capable of interacting with BICD2-mTrbo, the interaction between endogenous RAB6A and BICD2 might not be that prevalent under physiological conditions.

Four components of the six member HOPS complex were found to associate with BICD2 in our interactome analysis with VPS41 being the fifth most enriched protein (Fig. 1C’, D, Supplemental table 1). We therefore chose to validate these interactions using co-immunoprecipitation. Strains expressing either RFP-mTrbo, full length BICD2-mTrbo, or the delta CC3 domain construct (BICD2_delCC3-mTrbo) were used for this analysis. The TurboID constructs contain a V5 tag. Thus, V5 trap beads were used to immunoprecipitate the tagged proteins. Bound proteins were eluted and analyzed by western blotting using antibodies against VPS41 (Fig. 2A), VPS18 (Fig. 2B) or VPS16 (Fig. 2B). Consistent with our interactome analysis, all three HOPS components specifically co-immunoprecipitated with full length BICD2-mTrbo (Fig. 2A, B). A similar interaction between BICD2-mTrbo and VPS41 was also detected in cells depleted of endogenous BICD2 (Supplemental Fig. 1B). Thus, we conclude that BICD2 associates with VPS41, VPS18, and VPS16 in vivo and that these interactions require the BICD2 cargo binding domain.

Previously identified BICD2 cargo such as RANBP2 and RAB6A have been shown to interact with the isolated CC3 domain of BICD2 (Matanis et al., 2002; Splinter et al., 2010). In order to determine whether the same conditions apply for the HOPS complex components, we repeated the experiment using the same three strains mentioned above in addition to a strain expressing the isolated CC3 domain fused to mTrbo (BICD2_CC3-mTrbo) (Fig. 2C). Unexpectedly, although RANBP2 was able to interact with both full length BICD2 and the isolated CC3 domain, VPS41, VPS18 and VPS16 were only capable of associating with full length BICD2 (Fig. 2D, Supplemental Fig. 1C). Thus, the minimal cargo binding domain of BICD2 is not sufficient for interacting with the HOPS complex components. We also attempted the binding experiment using an extended BICD2 C-terminal construct (residues 442-824), but even this construct failed to efficiently bind the HOPS complex components (data now shown). Finally, we generated a BICD2 construct that lacked the first coiled coil domain (BICD2_delCC1, residues 288-824). Although this construct was able to bind VPS41 above background levels, the binding was reduced in comparison to the full-length protein (Supplemental Fig. 1D). RANBP2 binding was also reduced with this deletion construct (Supplemental Fig. 1D). It is possible that this deletion affects the normal folding of BICD2 or that it results in the protein folding into a locked conformation that inhibits cargo binding. Collectively, our results suggest that although the CC3 domain of BICD2 is required for interaction with the HOPS complex, it is not sufficient.

Among the HOPS components, VPS41 was by far the most enriched in the BICD2 interactome (Fig. 1C, C’). We therefore hypothesized that the interaction between BICD2 and the HOPS complex is mediated by VPS41. Contrary to our hypothesis, however, siRNA mediated depletion of VPS41 had little effect on the ability of BICD2 to interact with VPS16 and VPS18 (Supplemental Fig. 1E). This suggests that BICD2 is able to independently associate with multiple HOPS components, potentially interacting with a common motif or domain present within these proteins. Our inability to map the minimal domain of BICD2 required for interaction with the HOPS complex components precludes our ability to test for their direct interaction. Thus, although our results indicate that BICD2 associates with these proteins in vivo, we cannot conclude whether the interaction is direct or indirect.

We next examined the localization of the HOPS complex in control versus BICD2 depleted HeLa cells. Although antibodies are available that can detect endogenous VPS41, VPS18, and VPS16 by western blot, these antibodies were not suitable for detecting the native protein by immunofluorescence. Thus, we used a GFP-tagged VPS41 construct for this analysis. In HeLa cells, microtubule minus ends are present at the perinuclear centrosome and plus ends extend towards the cortex. As expected, given the role of the HOPS complex in late endosome-lysosome fusion, GFP-VPS41 positive vesicles were enriched in a perinuclear area. Dispersed vesicles and vesicles present closer to the cortex were also observed in cells transfected with a control siRNA (Fig. 3A, D, E). As expected, depletion of Dynein heavy chain (DHC, DYNC1H1) resulted in a reduction of perinuclear clustering and an accumulation of vesicles close to the cell cortex (Fig. 3B, D, E, Supplemental Fig. 2A). Because BICD2 is a cargo adaptor for dynein, we expected that depletion of BICD2 would produce a similar phenotype. Surprisingly however, BICD2 depletion resulted in perinuclear clustering of GFP-VPS41 vesicles (Fig. 3C, D, E, Supplemental Fig. 2A, B). A similar phenotype was also observed using a different siRNA targeting BICD2 (Supplemental Fig. 2B, 2D). Previous studies have shown that disruption of BICD2-mediated dynein cargo trafficking results in fragmentation of the Golgi and defective centrosome to nucleus tethering in cells in G2 phase of the cell cycle (Hoogenraad et al., 2001; Splinter et al., 2010). Consistent with these earlier findings, we also found that the depletion of BICD2 using the above siRNA resulted in Golgi fragmentation and centrosome localization defects in HeLa cells (Supplemental 2E-H).

In addition to the HOPS complex, late endosomes and lysosomes are also positive for LAMP1 and previous studies have shown partial co-localization between V5-tagged VPS41 and LAMP1 vesicles (van der Welle et al., 2021). We therefore used a validated antibody to detect LAMP1 vesicles under these same experimental conditions. Consistent with what was observed for GFP-VPS41, LAMP1 was present in more peripheral vesicles upon DHC depletion and was clustered closer to the nucleus in BICD2 depleted cells (Fig. 3F-H, Supplemental Fig. 2I-K).

Although we were able to visualize GFP-VPS41 for immunofluorescence experiments using a GFP antibody to boost signal intensity, the native fluorescence of GFP-VPS41 was consistently very low, precluding our ability to examine the motility of these vesicles in living cells. We therefore examined lysosome motility using a SiR lysosome kit. This reagent detects cathepsin D present in mature lysosomes. As expected, although lysosomes were present throughout the cell, they were enriched in a perinuclear area and a large fraction of these particles were motile (Video 1). Depletion of BICD2 caused an even more pronounced perinuclear enrichment of lysosomes, consistent with the localization of GFP-VPS41 and LAMP1. However, in contrast to cells treated with the control siRNA, very few motile particles were detected in BICD2 depleted cells (Video 2; Supplemental Fig. 2L). Thus, depletion of BICD2 results in perinuclear enriched lysosomes that are largely immotile.

The localization of GFP-VPS41 and LAMP1 vesicles in BICD2 depleted cells is similar to the phenotype obtained upon depletion of KIF5B, a plus-end directed Kinesin-1 motor (Guardia et al., 2016). Although BICD2 is primarily regarded as a cargo adaptor for dynein, it has also been shown to interact with KIF5B (Grigoriev et al., 2007). In addition, Drosophila BicD was recently shown to activate the motility of Kinesin-1, the fly ortholog of KIF5B (Ali et al., 2025). In order to more closely examine the role of BICD2 in LAMP1 vesicle localization, we over-expressed either GFP or BICD2 in cells. In contrast to cells over-expressing GFP, LAMP1 vesicles were peripherally scattered in BICD2 over-expressing cells. This phenotype was observed in HeLa and Cos7 cells (Fig. 3I, J, L, M and Supplemental Fig. 2M, M’). Depletion of KIF5B in BICD2 over-expressing cells reverted this phenotype and restored the perinuclear localization of LAMP1 vesicles (Fig. 3K-M, and Supplemental Fig. 2C). These results suggest that BICD2 likely functions to link VPS41 and LAMP1 vesicles with KIF5B for plus-end directed transport. Our studies do not reveal the mechanism of this linkage, however, and further studies will be needed to fully understand the role of BICD2 in this process.

Several SMALED2 associated mutations have been identified in the BICD2 cargo binding domain (Koboldt et al., 2020). For this initial analysis, we chose to examine BICD2_R694C and BICD2_R747C. These mutants, which result in the substitution of arginine for cystine at the indicated residues, were particularly intriguing because the R747C mutation was predicted to disrupt the interaction between BICD2 and RANBP2 (Terawaki et al., 2015). By contrast, previous studies suggest that the R694C mutation results in a higher level of interaction between BICD2 and RANBP2 (Yi et al., 2023). In addition to these mutations in the cargo binding domain, we also chose to analyze an additional mutant, BICD2_N188T (substitution of asparagine for threonine). This mutation is present within the first coiled coil domain and has been shown to result in dynein hyperactivation (Huynh and Vale, 2017). No studies have thus far addressed the global interactome of these mutants and how they might differ from the wild-type protein.

Stable cell lines capable of expressing BICD2_N188T, R694C and R747C with a C-terminal mTrbo tag were generated in HEK293 FLP In T-Rex cells. We first determined whether the cargo binding domain mutants retained their ability to interact with dynein. Wild-type and mutant versions of BICD2-mTrbo, containing a V5 tag, were immunoprecipitated using V5 trap beads. The co-precipitated proteins were analyzed by western blotting using antibodies against Dynactin1/p150 Glued (DCTN1) and Dynein intermediate chain (DYNC1I1). Interestingly, not only did BICD2_R694C and BICD2_R747C retain their interaction with dynein, both mutants generally interacted with dynein and dynactin at a higher level than the wild-type protein (Fig. 4A). Due to experimental variability in the co-immunoprecipitation experiment, the increased rate of binding was only statistically significant for the BICD2_R747C mutant (Supplemental Fig. 3A, B). In this regard, they were similar to the BICD2_N188T mutant which had previously been shown to interact with dynein at a higher level and to hyperactivate the motor (Fig. 4A) (Huynh and Vale, 2017).

We next examined the localization of wild-type and mutant BICD2 in HeLa cells. Wild-type BICD2 was localized throughout the cell with a fraction co-localizing with Golgi and centrosomal markers (Fig. 4B and F, Supplemental Fig. 3C, C’, C’’). By contrast all three mutants were enriched at varying degrees within the centrosome (Fig. 4C-F). Notably, the centrosomal enrichment was highly significant for BICD2_R694C and BICD2_R747C (Fig. 4F). These results suggest that the cargo binding domain mutants might hyperactivate dynein. To test this more directly, we monitored the distribution of peroxisomes using a well-characterized tethering assay (Kapitein et al., 2010; Passmore et al., 2021). Peroxisomes are relatively immotile but can be transported towards microtubule minus ends in a dynein dependent manner if they are tethered to an activating adaptor (Fig. 4G). Thus, centrosomal clustering of peroxisomes is an indicator of dynein activity. Peroxisomes were tethered to wild-type or mutant alleles of BICD2. Consistent with the higher level of dynein interaction, peroxisomal clustering was increased in cells expressing all three SMALED2 mutants (Fig. 4H and Supplemental Fig.3D-G). Collectively, these results suggest that SMALED2 mutations within the cargo binding domain of BICD2 are also capable of hyperactivating dynein.

As noted previously, BICD2 has also been shown to interact with the microtubule plus-end directed motor, KIF5B (Grigoriev et al., 2007). We therefore determined whether this interaction was maintained with the mutant alleles of BICD2. Wild-type full-length BICD2 associated with KIF5B at slightly higher levels than the binding control, consistent with published results (Grigoriev et al., 2007). Interestingly, all three mutants displayed a reduced interaction with KIF5B, with the BICD2_R747C mutant being the most severe (Supplemental Fig. 3H, I). Thus, the phenotypic outcome of dynein hyperactivity likely results from the combined effect of increased binding between BICD2 and dynein, and reduced binding between BICD2 and KIF5B.

In contrast to HeLa cells, microtubules are organized differently in neurons. In these cells, microtubule minus ends are localized within the cell body and plus ends extend towards the axon (van Beuningen and Hoogenraad, 2016). Thus, anterograde movement towards the axon tip is driven by a plus end motor, whereas retrograde transport towards the cell body involves dynein. Primary hippocampal neurons were established from E18 rat brains and after 3 days in vitro, plasmids containing either wild-type or mutant alleles of BICD2 were transfected into these cells. Consistent with BICD2 being a cargo adaptor for dynein, most of the signal for the wild-type protein was detected within the cell body (Fig. 5A). However, wild-type BICD2 could also be detected within the axon. A similar pattern was noted for BICD2_N188T (Fig. 5B). By contrast, axonal signal for BICD2_R694C and BICD2_R747C was reduced, reflected as a relative increase in cell body enrichment (Fig. 5C, D). The difference in localization was statistically significant for the R747C mutant (Fig. 5E). These results are consistent with the cargo binding domain mutants resulting in dynein hyperactivation. Huynh and Vale found that expression of certain SMALED2 associated BICD2 mutants, including BICD2_N188T, in hippocampal neurons results in reduced neurite growth (Huynh and Vale, 2017). We obtained similar results and found that this phenotype was also shared by the cargo binding domain mutations, BICD2_R694C and BICD2_R747C (Fig. 5F).

We next determined the interactomes of the three SMALED2 mutations and compared these interactomes to that of wild-type BICD2. As before, the entire experiment was done in triplicate. Proteins that displayed a two-fold increased or decreased interaction with the mutant allele in comparison to the wild-type and had a p value of at least 0.05 were considered significant (Fig. 6A-C, Supplemental tables 2, 3, 4). All three mutations, including N188T, which is present in the CC1 domain of BICD2, displayed an altered interactome. The most dramatic difference, however, was observed for BICD2_R747C (Fig. 6C).

In order to validate the proteomics results, we repeated the experiment in triplicate and analyzed the pellets by western blotting. Consistent with the proteomics data, and with published results using a mutation in mouse BicD1 (Terawaki et al., 2015), BICD2_R747C displayed a reduced interaction with RANBP2 (Fig, 6D, E, Supplemental table 4). BICD2 has been shown to associate indirectly with RANGAP via its direct interaction with RANBP2 (Splinter et al., 2010). Thus, as expected, a reduced level of RANGAP was detected in pellets from BICD2_R747C in comparison to the wild-type (Supplemental table 4). A further analysis of the proteomics results indicated a reduced association of BICD2_R747C with nuclear import receptors such as importin beta (Supplemental table 4). This result was confirmed by repeating the binding experiment and analyzing the eluate by western blotting (Supplemental Fig. 4A). We hypothesized that the association between BICD2 and import receptors is likely indirect and mediated via the direct interaction between BICD2 and RANBP2. Consistent with this notion, the interaction of wild-type BICD2 with importin beta was greatly reduced in cells depleted of RANBP2 (Supplemental Fig. 4C). By contrast, knock-down of RANBP2 had no effect on the BICD2-VPS41 interaction (Supplemental Fig. 4C). Based on these results, we conclude that BICD2_R747C is disrupted for interaction with RANBP2, and consequently also displays a reduced association with additional nuclear import factors.

An unexpected finding was that the BICD2_R747C mutant also displayed a reduced interaction with HOPS complex components. The proteomics result indicated a reduced interaction with VPS41, VPS18 and VPS33 (Supplemental table 4). However, the mass spectrometry result indicated that the association of BICD2_R747C with VPS16 was unaffected. In order to validate these results, the experiment was repeated and analyzed by western blotting. Indeed, the association of BICD2_R747C with VPS41 and VPS18 was greatly reduced in comparison to the wild-type protein (Fig. 6D, F and Supplemental Fig. 4C). However, using this approach we consistently observed a significant reduction in the association between BICD2_R747C and VPS16 (Supplemental Fig. 4D). Based on these results, we conclude that this SMALED2 mutant is most likely compromised for interacting with the entire HOPS complex.

In addition to interactions that were lost or reduced, all three mutants were also associated with changes in the positive direction. For instance, the mass spectrometry results revealed that the mutants showed a stronger interaction with the centrosomal protein CSPP1 compared to wild-type BICD2 (Supplemental Tables 2, 3, 4). Validation experiments confirmed this finding, demonstrating that the cargo-binding domain mutants had significantly elevated interactions with CSPP1, exceeding that observed with either wild-type BICD2 or BICD2_N188T (Fig. 6D, G). Since BICD2_R694C and BICD2_R747C are enriched at the centrosome (Fig. 4D-F), their altered localization most likely accounts for their increased association with centrosomal proteins.

Among the three mutants analyzed, BICD2_R747C exhibited the most distinct interactome. One particularly notable interaction was with GRAMD1A, a protein absent from the wild-type BICD2 interactome but highly enriched in the BICD2_R747C pull-down, indicating a gain-of-function interaction. GRAMD1 encodes three isoforms: A, B, and C, with GRAMD1A being highly expressed in the central nervous system. To validate this observation, the experiment was repeated in triplicate and analyzed by western blotting. Consistent with the mass spectrometry result, GRAMD1A showed only background binding to the control, wild-type BICD2, BICD2_N188T, or BICD2_R694C. In contrast, GRAMD1A strongly associates with BICD2_R747C (Fig. 7A, B). The same binding profile was observed in cells depleted of endogenous BICD2 (Supplemental Fig. 5A).

We next examined the localization of GRAMD1A in Cos7 cells co-expressing GRAMD1A-mScarlet3 with wild-type or mutant alleles of BICD2 tagged with mNeon. GRAMD1A localizes to the endoplasmic reticulum (ER) and concentrates at sites of ER-plasma membrane contact (Besprozvannaya et al., 2018). Cos7 cells are typically used for examining the localization of GRAMD1A due to their large and flattened appearance (Besprozvannaya et al., 2018; Naito et al., 2019). As expected, GRAMD1A displayed a typical ER localization pattern in cells expressing wild-type BICD2 or BicD2_N188T (Fig. 7C, D). Much like what was observed in HeLa cells, BICD2_R694C was often enriched in the centrosomal region in Cos7 cells (Fig. 7E, Fig. 4D). The localization of GRAMD1A was essentially unchanged in these cells, although a small amount of the protein co-localized with BICD2_R694C in the centrosomal region (Fig. 7E, G). By contrast, consistent with a gain-of-function interaction, a significant fraction of GRAMD1A localized adjacent to Pericentrin, a centrosomal marker, in cell expressing BICD2_R747C (Fig. 7F, G, Supplemental Fig. 5B-F). This effect was specific for GRAMD1A because a general ER maker remained correctly localized in cells expressing BICD2_R747C (Supplemental Fig. 5G, G’, H, H’).

Collectively, our results suggest that SMALED2 associated mutations in BICD2 hyperactive dynein and result in gain of function and loss of function interactions with cargo. It is therefore possible that cargo trafficking defects that arise because of these phenotypes contribute to the symptoms associated with SMALED2.

The full list of DNA constructs, antibodies and reagents used in this work is found in Supplemental table 5.

Synthesized DNA fragments were generated by either Twist biosciences or Genewiz/Azenta. All final plasmids used in this work were sequenced by Plasmidsaurus. Plasmid sequences as well as detailed cloning strategies will be provided upon request.

DNA encoding wild-type BICD2, generated by gene synthesis, was cloned into the Kpn1 and Xho1 sites of the pCDNA-FRT-TO vector (Thermo Fisher Scientific). Next, a fragment encoding mTrbo was cloned downstream of BICD2_wild-type using High fidelity assembly (NEB). A similar strategy using a synthesized fragment was used to clone mRFP-mTrbo into the pCDNA-FRT-TO vector. BICD2 mutants fused to mTrbo were generated by swapping in a synthesized fragment containing the desired mutation into the appropriate site of the BICD2_wild-type-mTrbo construct. The cDNA for wild-type BICD2 as well as the SMALED2 mutants contained silent mutation that made them resistant to bicd2 siRNA-1. The BICD2 truncations were produced by cloning either PCR generated, or gene synthesized fragments into the appropriate vector. Stable cell lines were created using these FRT constructs and the pOG44 vector (Thermo Fisher Scientific). For localization studies, PCR sub-cloning was used to move the appropriate wild-type or BICD2 mutans into the pLVX-Neo vector (Takara Bio). A fragment encoding mNeon-green along with the V5 epitope tag was cloned downstream of wild-type or mutant BICD2. The GFP-VPS41 construct was generated by cloning a synthesized fragment encoding human VPS41 into the Kpn1 and Not1 sites of the GFP-BICD2 plasmid (Addgene plasmid #161626; (Bonet-Ponce et al., 2020)). This removed the BICD2 cDNA from this vector and inserted the cDNA for VPS41 in its place. For the dynein activity experiment described in Fig. 4, a construct containing the PEX3 peroxisome targeting sequenced fused to mGreenLantern and the FRB dimerization sequence was cloned into the pLVX neo vector (Takara). The BICD2_wild type and mutant construct described above were modified by replacing the mTrbo sequence with the FKBP dimerization domain. The GRAMD1A-mScarlet3 construct was generated by cloning fragments corresponding to human GRAMD1A and mScarlet3 into the pLVX neo vector.

HeLa and Cos7 cells were obtained from ATCC and were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Thermo Fisher Scientific). The HEK Flp-In T-REx 293 Cell Line was obtained from Thermo Fisher Scientific and was cultured according to the instructions provided by the manufacturer. Stable cells were generated by culturing the transfected HEK Flp-In T-REx 293 in 100ug/ml Hygromycin B (Gibco) for 12 to 14 days, with fresh media changes every 3 days. The E18 rat hippocampus culturing kit was obtained from Transnetyx. Primary neurons were prepared according to the instructions provided by the manufacturer. 40,000 cells were seeded onto poly Lysine coated coverslips (Neuvitro Corporation) in each well of a 12-well plate. The cells were transfected after 3 to 4 days in culture using Lipofectomine 2000 (Thermo Fisher scientific).

Stable HEK Flp-In T-REx 293 expressing mGreenLantern tagged peroxisomes were generated by infection with lentivirus. Lentiviral particles were produced in HEK293T cells following a previously published protocol (Wei et al., 2024). Culture medium containing the viral particles were passed through a 0.2 micron filter to remove cell debris. Next, the HEK Flp-In T-REx 293 cells were infected with freshly prepared viral particles for 48 hours. After infection, puromycin (1ug/mL) was added to the culture medium to select for stable cell lines. Uniformly expressing mGreenLantern clones were selected and used to integrate BICD2_wt or mutant containing the FKBP dimerization motif at the FRT locus. Stables cells containing the integrated BICD2 constructs were selected using Hygromycin B as described above.

The Qiagen Effectene reagent was used for transfecting DNA into HeLa, HEK Flp-In T-REx 293, and Cos7 cells using the directions provided by the manufacturer. For transfecting cells in 6 well dishes, 0.4ug of DNA was used along with 10ul of Effectene. Primary neurons grown on glass coverslips in a 12 well dish were transfected using Lipofectamine 2000 (Thermo Fisher Scientific). 0.5-0.8ug of DNA and 2ul of Lipofectamine 2000 was used in each transfection. Expression of the transgenes was induced using 0.3 ug/ml of Doxycycline 24 hours after transfection. The cells were fixed and processed for immunofluorescence the following day. Lipofectamine RNAimax (Thermo Fisher Scientific) was used for the transfection of siRNA according to the instructions provided by the manufacturer.

For small scale experiments (Figure 2, 6, 7, Supplemental figures 1 and 4), HEK Flp-In T-REx 293 cells expressing the desired constructs by Tet-based induction (1ug/ml, 24 hours), were incubated with biotin (500uM) for 40 minutes. The cells were then harvested, and lysates were prepared by resuspending the cells in RIPA buffer (50 mM Tris-Cl [pH 7.5], 150 mM NaCl, 1% NP-40, 1 mM EDTA) containing a Halt Protease inhibitor cocktail (Thermo Fisher Scientific). 1mg of total protein was used in the binding experiment with 15ul of High-Capacity Streptavidin Agarose beads (Thermo Fisher Scientific) in RIPA buffer. The binding was performed overnight at 4⁰C. The samples were washed four times using RIPA buffer, bound proteins were eluted in Laemmli buffer and analyzed by western blotting. All western blot images were acquired on a Bio Rad ChemiDoc MP.

For proteomics experiments, the same cells were grown in 10cm dishes and 5mg of total protein was used in each purification. Biotionylated proteins were purified using 70ul of Streptavidin magnetic beads (Thermo Fisher Scientific), incubated at 4⁰C overnight. The following day, the samples were extensively washed using 1ml of the following: 3 times with RIPA buffer, 3 times with high salt RIPA buffer (50 mM Tris-Cl [pH 7.5], 1M NaCl, 1% NP-40, 1 mM EDTA), 3 times with RIPA buffer and 4 times with PBS. The entire experiment was done in triplicate. After the final wash, the beads were resuspended in 70ul of PBS and shipped to the Emory Integrated Proteomics Core on dry ice.

The mass spectrometry was performed at the Emory Integrated Proteomics Core (RRID:SCR_023530). On bead digestion: A published protocol was followed for on bead digestion of proteins (Soucek et al., 2016). A digestion buffer containing 50 mM NH₄HCO₃ was added to the beads. The mixture was then incubated with 1 mM dithiothreitol (DTT) at room temperature for 30 minutes, followed by the addition of 5 mM iodoacetimide (IAA). This mixture was incubated at room temperature for an additional 30 minutes in the dark. Proteins were digested with 1 µg of lysyl endopeptidase (Wako) at room temperature overnight and were further digested overnight at room temperature with 1 µg trypsin (Promega). The resulting peptides were desalted using an HLB column (Waters) and were dried under vacuum.

LC-MS/MS: The data acquisition by LC-MS/MS was adapted from a published procedure (Seyfried et al., 2017). Derived peptides were resuspended in the loading buffer (0.1% trifluoroacetic acid, TFA) and were separated on a Water’s Charged Surface Hybrid (CSH) column (150 µm internal diameter (ID) x 15 cm; particle size: 1.7 µm). The samples were run on an EVOSEP liquid chromatography system using the 15 samples per day preset gradient and were monitored on a Q-Exactive Plus Hybrid Quadrupole-Orbitrap Mass Spectrometer (ThermoFisher Scientific). The mass spectrometer cycle was programmed to collect one full MS scan followed by 20 data dependent MS/MS scans. The MS scans (400-1600 m/z range, 3 x 10⁶ AGC target, 100 ms maximum ion time) were collected at a resolution of 70,000 at m/z 200 in profile mode. The HCD MS/MS spectra (1.6 m/z isolation width, 28% collision energy, 1 x 10⁵ AGC target, 100 ms maximum ion time) were acquired at a resolution of 17,500 at m/z 200. Dynamic exclusion was set to exclude previously sequenced precursor ions for 30 seconds. Precursor ions with +1, and +7, +8 or higher charge states were excluded from sequencing.

MaxQuant: Label-free quantification analysis was adapted from a published procedure (Seyfried, Dammer et al. 2017). Spectra were searched using the search engine Andromeda, integrated into MaxQuant, against 2022 human UniProtKB/Swiss-Prot database (20,387 target sequences). Methionine oxidation (+15.9949 Da), asparagine and glutamine deamidation (+0.9840 Da), and protein N-terminal acetylation (+42.0106 Da) were variable modifications (up to 5 allowed per peptide); cysteine was assigned as a fixed carbamidomethyl modification (+57.0215 Da). Only fully tryptic peptides were considered with up to 2 missed cleavages in the database search. A precursor mass tolerance of ±20 ppm was applied prior to mass accuracy calibration and ±4.5 ppm after internal MaxQuant calibration. Other search settings included a maximum peptide mass of 6,000 Da, a minimum peptide length of 6 residues, 0.05 Da tolerance for orbitrap and 0.6 Da tolerance for ion trap MS/MS scans. The false discovery rate (FDR) for peptide spectral matches, proteins, and site decoy fraction were all set to 1 percent.

Quantification settings were as follows: re-quantify with a second peak finding attempt after protein identification has completed; match MS1 peaks between runs; a 0.7 min retention time match window was used after an alignment function was found with a 20-minute RT search space. Quantitation of proteins was performed using summed peptide intensities given by MaxQuant. The quantitation method only considered razor plus unique peptides for protein level quantitation.

HEK293 FLP-in T-Rex cells expressing the desired constructs by Tet-based induction (1ug/ml, 24 hours), were harvested and lysates were prepared using RIPA buffer. 1mg of total protein was used in the binding experiment with 15ul of V5-Trap agarose beads (ChromoTek, ProteinTech) in RIPA buffer. The binding was performed at 4⁰C for 2 hours. The samples were then washed four times using RIPA buffer, bound proteins were eluted in Laemmli buffer and analyzed by western blotting using the indicated antibodies.

Cells (HeLa, HEK293 FLP In T-Rex, Cos7, and primary rat neurons) adhered to glass coverslips were fixed using either 4% formaldehyde for 5 minutes at room temperature (Figures 5, 7, Supplemental figure 3, 4) or with methanol at -20⁰C for 10 minutes (Figures 3, 4, Supplemental figure 2). After fixation, cells were permeabilized by incubating with PBST (PBS containing 0.1% TritonX-100) for 5 minutes. Next the samples were blocked for 1 hour at room temperature using 5% normal goat serum (Thermo Fisher Scientific). Cells were incubated overnight at 4⁰C with primary antibody in blocking solution. The next day, the coverslips were washed three times with PBS. The cells were then incubated with secondary antibody diluted in blocking solution for 1 hour at room temperature. Following this, the cells were washed three times with PBS, stained with DAPI (Thermo Fisher Scientific), and mounted onto slides using Prolong Diamond antifade reagent (Thermo Fisher Scientific).

All imaging experiments were performed at the Augusta University Cell Imaging Core (RRID:SCR_026799). Fixed images were captured on either an inverted Leica Stellaris confocal microscope or an inverted Zeiss LSM780 equipped with Airyscan. Live imaging of SiR labeled lysosomes was performed on an inverted Nikon AXR confocal microscope equipped with the NSPARC detector. Images were processed for presentation using Fiji, Adobe Photoshop, and Adobe Illustrator.

The clustering phenotype of GFP-VPS41 and LAMP1 vesicles (Fig. 3) was quantified using Imaris 10.2. The nucleus was defined using the “surface” feature of Imaris and the vesicles were defined using the “spots” feature. The percentage of vesicles that were less than or equal to 10 microns from the nucleus and the percentage of vesicles that were present a distance greater than 10 microns from the nucleus was determined using object to object classification and the filtering feature in Imaris. The quantification of lysosome motility in live cells (Supplemental figure 2D) was also performed in Imaris. Vesicles were defined using the “spots” feature and were tracked over time. Motile vesicles were defined as those that displayed movement over at least a 3 second interval and had at least 0.1microns of displacement over the course of the imaging experiment. This enabled us to quantify the percent of motile particles per cell and the velocity of these vesicles. The centrosome enrichment phenotype shown in Fig. 4B-E was quantified by determining the ratio of the mean BICD2 signal in the entire cell vs the mean intensity in the centrosomal area demarcated by Pericentrin. This was done using Imaris. The peroxisome clustering phenotype (Fig. 4H) was quantified using the “cells” module of Imaris. Within this module, the nucleus was defined using the DAPI channel and the cell boundaries were determined using phalloidin staining of F-actin. Next, the peroxisomes were defined using the “spots” feature built into this module. This enabled us to calculate the average distance of the peroxisomes to the nucleus on a cell-by-cell basis. The cell body enrichment of BICD2 signal (Fig. 5E) was determined using the “filaments” module in Imaris. This was used to define the cell body and the axon. The mean intensity of signal in the cell body was compared to the mean intensity of signal in the axon. The co-localization between GRAMD1A and BICD2 (Fig.7) was determined using the “co-localization” module in Imaris. Axon length (Fig. 5F) was quantified using the Simple neurite tracer plugin for FIJI. Densitometry for western blot images captured on the BioRad Chemidoc MP were analyzed using the BioRad Image Lab software. The binding level was compared to the amount of the respective protein detected in the BICD2_wt lane. Graphing of data and statistical analysis was performed using Graphpad Prism10.