The FHA domain is essential for autoinhibition of KIF1A/UNC-104 proteins

Various cellular processes, such as cell division, morphogenesis and synapse formation, depend on molecular motors. Kinesins, also known as kinesin superfamily proteins (KIFs), are microtubule-dependent molecular motors that play crucial roles in intracellular transport and cell division (Hirokawa et al., 2009). Kinesins are classified into 14 families based on their structural and functional features (Lawrence et al., 2004). The kinesin-3 family includes eight KIFs in mammals. KIF1A, a member of the kinesin-3 family, is a plus-end directed kinesin specifically expressed in neurons (Okada et al., 1995). KIF1A and UNC-104, its homolog in Caenorhabditis elegans and Drosophila melanogaster, are involved in the axonal transport of various cargoes, such as presynaptic components, dense core vesicles (DCVs) and lysosomes (Hall and Hedgecock, 1991; Otsuka et al., 1991; Zahn et al., 2004; Lo et al., 2011; Guardia et al., 2016). Genetic variants of the KIF1A gene cause a congenital genetic disorder known as KIF1A-associated neurological disorder (KAND) (Boyle et al., 2021; Gabrych et al., 2019). We have shown that both hyperactivation and impaired activity of KIF1A are associated with neurodegeneration (Anazawa et al., 2022; Chiba et al., 2019, 2023). Proper control of KIF1A activity is crucial for maintaining neuronal function; however, the molecular mechanisms that regulate KIF1A remain poorly understood.

Most kinesins are dimeric, with the two motor domains (heads) connected by a coiled-coil domain. Dimeric kinesins alternately use the two heads for processive movement on microtubules (Yildiz et al., 2004). The kinesin-3 family members were thought to be unusual monomeric kinesins as the founding member of the kinesin-3 family, KIF1Bα, and KIF1A were both described as monomers (Nangaku et al., 1994; Okada et al., 1995). Later studies showed that C. elegans UNC-104 (CeUNC-104) and other kinesin-3 members can form dimers on the cargo membranes (Klopfenstein et al., 2002; Tomishige et al., 2002; Soppina et al., 2014). Moreover, mammalian KIF1A has been shown to form dimers in the cytosol (Hammond et al., 2009). Several studies have proposed that most of the kinesin-3 family members undergo a monomer-to-dimer conversion to be activated (Al-Bassam et al., 2003; Lee et al., 2004; Huo et al., 2012; Ren et al., 2018; Wang et al., 2022). Intramolecular interactions of these kinesin-3 proteins mediate their autoinhibition, maintaining the motor in a monomeric state (Al-Bassam et al., 2003; Ren et al., 2018; Wang et al., 2022). Upon release from autoinhibition, monomeric kinesin-3 proteins undergo dimerization, resulting in increased binding and processivity on microtubules compared with the monomeric state (Ren et al., 2018; Kita et al., 2024). Beyond affecting motility, kinesin-3 dimerization plays a significant role in regulating cargo transport, such as DCVs (Hummel and Hoogenraad, 2021). However, the crucial process that initiates the dimerization remains elusive.

As the potential regulator of kinesin-3 dimerization, the autoinhibition of kinesin-3 proteins has been extensively studied. Kinesin-3 members typically possess an N-terminal motor domain (MD) followed by a neck coiled-coil domain (NC), coiled-coil domain 1 (CC1), FHA domain (FHA) and coiled-coil domain 2 (CC2). An early cryo-electron microscopy study using CeUNC-104 proposed that CC1 folds back and associates with NC to prevent dimerization (Al-Bassam et al., 2003). Results consistent with this were shown by subsequent X-ray crystallography using truncated versions of mammalian KIF1A or KIF13B (Huo et al., 2012; Ren et al., 2018). They showed that versions containing MD-NC or CC1-FHA can form dimers, whereas versions containing MD-NC-CC1 predominantly form monomers (Huo et al., 2012; Ren et al., 2018). CC1 has been demonstrated to interact with NC and MD, negatively regulating kinesin-3 motor activity (Ren et al., 2018). In addition to MD, NC and CC1, intramolecular interactions between FHA and CC2 have been implicated in the autoinhibition and motor multimerization of KIF1A (Lee et al., 2004). Those analyses have mainly used only the respective domains, and the role of these domains on the regulation of kinesin-3 proteins, especially in the context of the full-length forms, are not well understood.

D. melanogaster has been an important model organism in the study of the kinesin superfamily (Saxton et al., 1988; Endow et al., 1990). Drosophila kinesin-1 (also known as Khc), a homolog of KIF5 proteins, has been well-known since the early stages of kinesin research and has been extensively analyzed. The fundamental biochemical properties of kinesin-1 were determined using the Drosophila kinesin-1 protein (McDonald and Goldstein, 1990; Saxton et al., 1991; Hackney, 1995; Coy et al., 1999). Compared with Drosophila kinesin-1, D. melanogaster UNC-104 (DmUNC-104; also known as KIF1A or IMAC) is a relatively recently discovered member of Drosophila kinesins (Pack-Chung et al., 2007). In addition to nonsense mutations, some point mutations that disrupt synapse formation have been identified (Barkus et al., 2008; Kern et al., 2013). Although measuring the biochemical changes caused by those point mutations may help us understand the basic properties of kinesin-3 proteins, the biochemical properties of the DmUNC-104 protein have not been investigated.

Here, we purified and analyzed the activity of full-length DmUNC-104 protein. We show that mutations previously identified in or near the FHA domain induce the dimerization of DmUNC-104. These mutations nearly abolished monomer formation, suggesting they initiated the complete release of DmUNC-104 autoinhibition. In single-molecule assays using a total internal reflection fluorescence (TIRF) microscope, the dimerized mutant DmUNC-104 showed elevated activity compared with the monomeric wild type. Collectively, our data suggest that the FHA domain plays an essential role in the autoinhibition of UNC-104.

Full-length DmUNC-104, tagged with a C-terminal superfolder GFP (sfGFP) and Strep-tag II (DmUNC-104^WT; calculated molecular weight of 220 kDa), was expressed using the baculovirus expression system (Fig. 1A). Following affinity purification, we examined the oligomeric state of the purified protein by size exclusion chromatography (SEC) and mass photometry. DmUNC-104^WT displayed two distinct peaks upon elution (Fig. 1B). An additional peak near the void volume was observed, but no considerable amount of DmUNC-104^WT was detected in this peak. Upon subjecting the peak to agarose gel electrophoresis, bands corresponding to the size of insect ribosomal RNA (∼2 and 4 kbp) were observed (data not shown). The bands disappeared after RNaseI treatment (data not shown), confirming that the additional peak mainly contains contaminated RNA. Therefore, we focused our analysis on the latter two peaks. Mass photometry analysis at 20 nM confirmed that both peaks contained monomers and dimers. The earlier peak (peak-1) consists of particles corresponding to 209±42 kDa and 414±39 kDa (mean±s.d.), accounting for 24% and 54% of the population, respectively (Fig. 1C). The later peak (peak-2) contained particles corresponding to 216±18 kDa and 435±31 kDa (mean±s.d.) with 71% and 10% of the populations, respectively (Fig. 1C). The observed molecular weights align with a monomer and a dimer of DmUNC-104^WT, considering the calculated mass of 220 kDa as a monomer. These results indicate that full-length DmUNC-104 can exist as both monomers and dimers, and the state might be in equilibrium, as both monomers and dimers are present in individual peaks even after separation by SEC. To determine whether the conversion between monomers and dimers is dynamic, we treated DmUNC-104^WT with a crosslinker, BS³. Compared with the untreated sample, the cross-linked protein showed a broader chromatogram but still exhibited peak-1 and peak-2 (Fig. S1A,B). After 24 h from the initial SEC, we analyzed the proteins again by SEC. For the untreated proteins, re-injection of peak-1 or peak-2 showed a peak at the original position and small signals corresponding the other peak, respectively (Fig. S1C,E). By contrast, cross-linked proteins predominantly showed a peak corresponding to the original peak (Fig. S1D,F). These results suggest that the monomer-to-dimer transition of DmUNC-104 is highly dynamic.

We compared the motility of peak-1 and peak-2 of DmUNC-104^WT by single-molecule assay using a TIRF microscope. Both peaks were examined at 4 nM. Peak-1 and peak-2 showed almost identical properties on the single-molecule assays (Fig. 1D). The velocity was comparable (1.12±0.25 µm/s and 1.12±0.28 µm/s; mean±s.d. for peak-1 and peak-2, respectively) and the run length showed no significant difference (2.99 µm and 3.07 µm; median values for peak-1 and peak-2, respectively) (Fig. 1E,F). When we calculated landing rates of the motors on the microtubules, peak-1 showed a higher landing rate than peak-2 (0.028 and 0.020 molecules/µm/s; median values for peak-1 and peak-2, respectively) (Fig. 1G). However, no significant difference was observed in the landing rates when considering motile molecules out of the total landings, which include both motile and immotile landings (0.0087 and 0.0099 molecules/µm/s; median values for peak-1 and peak-2, respectively) (Fig. 1H). These results suggest that peak-1 and peak-2 contain a similar number of motile motors, but peak-1 contains more immotile motors that can bind to microtubules without processive motion. To determine whether the motile molecules observed in peak-1 and peak-2 are in the same oligomeric state, we compared the brightness of the motile motors. There was no significant difference in the fluorescence intensity of motile molecules between peak-1 and peak-2 (Fig. S2). Previous studies reported very slow velocities for monomeric mouse KIF1A compared with dimerized C. elegans UNC-104 (Okada and Hirokawa, 1999; Tomishige et al., 2002). As our DmUNC-104 motors exhibited velocities of ∼1.1 µm/s, we consider the motile motors observed in peak-1 and peak-2 to represent dimers.

The bristly (bris) mutation was isolated through ethyl methanesulfonate mutagenesis in Drosophila, leading to an increased number of dendritic filopodia (Medina et al., 2006). Subsequent studies identified bris as an allele of unc-104, with the responsible mutation being R561H (R562H in the isoform we utilized; hereafter referred to as R562H) (Kern et al., 2013). The study also revealed that bris causes defective synaptogenesis, possibly owing to impaired axonal transport of presynaptic materials such as Bruchpilot (Brp). The R562 residue is located in the β11-loop of the FHA domain of DmUNC-104. To investigate the effect of the bris mutation on the biochemical properties of DmUNC-104, we introduced the corresponding R562H mutation into DmUNC-104 (Fig. 2A). First, we examined the dimerization state. Unlike DmUNC-104^wt, DmUNC-104^bris was eluted as a single peak in SEC (Fig. 2B), with the elution volume similar to that of peak-1 of DmUNC-104^wt protein. Mass photometry of the peak revealed particles corresponding to 237±85 kDa and 476±58 kDa (mean±s.d.), accounting for 12% and 71% of the population, respectively (Fig. 2C). This confirmed that DmUNC-104^bris contains a small number of monomers but predominantly forms dimers. In single-molecule assays, DmUNC-104^bris showed a higher velocity than that of peak-1 of wild type (1.12±0.25 µm/s and 1.25±0.27 µm/s; mean±s.d. for DmUNC-104^WT, peak-1 and DmUNC-104^bris, respectively) (Fig. 2D,E). No significant differences were observed in the run length (2.99 µm and 3.45 µm; median values for DmUNC-104^{WT, peak-1} and DmUNC-104^bris, respectively) compared with peak-1 of wild type (Fig. 2F). However, we observed higher landing rates in both total landing (0.028 and 0.034 molecules/µm/s; median values for DmUNC-104^{WT, peak-1} and DmUNC-104^bris, respectively) and motile motor landing (0.0087 and 0.015 molecules/µm/s; median values for DmUNC-104^{WT, peak-1} and DmUNC-104^bris, respectively) (Fig. 2G,H). These data suggest that DmUNC-104^bris is biochemically more active than DmUNC-104^WT.

The previous study suggested that the bris mutation is a putative loss-of-function mutation, as the phenotype was exacerbated in transheterozygotes with a null allele (Kern et al., 2013). Our results from SEC and single-molecule assays showed that the bris mutation causes dimerization and activation of DmUNC-104. These results imply that the bris mutation might be a gain-of-function mutation. We have shown that C. elegans is a good model in which to test whether a kinesin-3 mutation leads to loss of function or gain of function in the axonal transport of synaptic materials. This assay takes advantage of the fact that axonal transport of synaptic vesicles is reduced in arl-8 mutants (Anazawa and Niwa, 2022). If the unc-104 mutation results in gain of function, it would work as a suppressor of the arl-8 phenotype, whereas if the unc-104 mutation leads to loss of function it would act as an enhancer of the arl-8 phenotype. Using this system, we analyzed the bris mutation. R562 in DmUNC-104 corresponds to R551 in CeUNC-104 (Fig. 3A). We introduced the bris mutation in the C. elegans unc-104 gene by CRISPR/Cas9 (Arribere et al., 2014; Ghanta et al., 2021) (Fig. S4) and assessed the phenotype of the DA9 neuron. The DA9 neuron is a polarized neuron with distinct regions, including a dendrite, cell body and axon (Klassen and Shen, 2007; Klassen et al., 2010; Wu et al., 2013; Niwa et al., 2016; Higashida and Niwa, 2023) (Fig. 3B). En passant synapses are formed along the axon of the DA9 neuron, visualized by a synaptic vesicle (SV) marker, GFP::RAB-3 (Fig. 3B,C). In arl-8 worms lacking ARL-8, SVs mislocalized to the commissure and proximal asynaptic region as a result of reduced axonal transport (Fig. 3D). In arl-8; unc-104^R551H double-mutant worms, however, the localization of RAB-3 signals was partially restored (Fig. 3E). To confirm the phenotype, we counted the number of GFP::RAB-3 puncta in the commissure (Fig. 3F) and measured the length of the asynaptic region (Fig. 3G). The aberrant localization of synaptic vesicles in arl-8 worms was suppressed by the unc-104^R551H mutation. Moreover, we found that unc-104^R551H could suppress the phenotype of arl-8 in an autosomal dominant manner. These results indicate that the bris mutation in unc-104 might be a gain-of-function mutation, not a loss-of-function mutation.

A previous study has shown that the FHA and CC2 domains form intramolecular interactions by bending at the hinge sequence ‘QGID’, which is well conserved across species (Fig. 4A). It has been suggested that a mutation in the hinge disrupts the association between the FHA and CC2 domains of human KIF1A (Lee et al., 2004). Consistent with this, a mutation in the hinge works as a suppressor of the arl-8 phenotype in C. elegans (Wu et al., 2013). To confirm that the effect of the bris mutation is caused by the release from autoinhibition, we introduced the hinge mutation G618R into DmUNC-104 and analyzed its biochemical behavior (Fig. 4A,B). Similar to DmUNC-104^bris, DmUNC-104^G618R was eluted as a single peak in SEC, with the elution volume corresponding to peak-1 of DmUNC-104^WT (Fig. 4C). Mass photometry of the peak revealed particles distributed almost in a single peak, with a molecular weight of 437±70 kDa (mean±s.d.), accounting for 80% of the population (Fig. 4D). This molecular weight corresponds to the dimer of DmUNC-104^G618R. We further tested the motility of DmUNC-104^G618R by single-molecule assay (Fig. 4E). Compared with peak-1 of DmUNC-104^WT, DmUNC-104^G618R showed slightly higher velocity (1.12±0.25 µm/s and 1.21±0.24 µm/s; mean±s.d. for DmUNC-104^WT, peak-1 and DmUNC-104^G618R, respectively) (Fig. 4F). No significant differences were observed in the run length (2.99 µm and 3.15 µm; median values for DmUNC-104^WT, peak-1 and DmUNC-104^G618R, respectively) and the landing rate (0.028 and 0.030 molecules/µm/s; median values for DmUNC-104^WT, peak-1 and DmUNC-104^G618R, respectively) compared with peak-1 of wild type (Fig. 4G,H). However, we observed higher landing rates in motile motor landing (0.0087 and 0.012 molecules/µm/s; median values for DmUNC-104^WT, peak-1 and DmUNC-104^G618R, respectively) (Fig. 4I). These properties observed in DmUNC-104^G618R are similar to those of DmUNC-104^bris, suggesting that the bris mutation disrupts the autoinhibition of DmUNC-104.

The data above suggested that DmUNC-104 is a promising protein to study the monomer-to-dimer conversion. Therefore, we investigated whether disease-associated human mutations affect the dimerization of DmUNC-104. Although most disease-associated mutations are known to have defective effects on KIF1A (Esmaeeli Nieh et al., 2015; Budaitis et al., 2021; Boyle et al., 2021; Lam et al., 2021), our previous work suggested that some mutations are gain-of-function mutations for KIF1A (Chiba et al., 2019). We have demonstrated that the recombinant proteins KIF1A^V8M, KIF1A^A255V and KIF1A^R350G show increased microtubule binding in single-molecule assays (Chiba et al., 2019). However, the precise molecular mechanisms of KIF1A activation induced by these mutations remain totally unknown. To study the effect of these disease-associated mutations on the dimerization of DmUNC-104, we introduced corresponding mutations, V6M, A255V and R348G, into DmUNC-104 (Fig. 5A,B). Our SEC analysis revealed that DmUNC-104^V6M and DmUNC-104^R348G predominantly eluted as a single peak, corresponding to peak-1 of the wild-type protein (Fig. 5C). Although they showed a small peak around the elution volume corresponding to peak-2 of the wild-type protein, it was not as pronounced. In contrast, DmUNC-104^A255V eluted into two peaks similar to those observed for DmUNC-104^WT (Fig. 5C). We measured the molecular weight of the disease-associated mutants after SEC. Mass photometry of the major peak of each mutant confirmed that the V6M and R348G mutants predominantly form dimers, whereas the A255V mutant mainly forms monomers (Fig. S5). These results indicate that, among the disease-associated gain-of-function mutations, at least two mutations may induce dimerization of KIF1A/UNC-104 proteins. Taken together, our data suggest that the dimerization of KIF1A/UNC-104 proteins is regulated by intramolecular interactions involving the FHA domain, which plays an essential role. Aberrant dimerization may, therefore, potentially contribute to the development of human diseases.

We showed that a deletion mutant of C. elegans UNC-104, consisting of the MD-CC1-FHA-CC2 domains, exists in an equilibrium between dimers and monomers (Kita et al., 2024). However, we could not analyze the property of the full-length C. elegans UNC-104 protein because of purification difficulty. In this study, we succeeded in purifying full-length DmUNC-104. Using the full-length DmUNC-104 protein, we now clearly show that autoinhibition release is sufficient to convert the monomeric UNC-104 molecule into dimers. Previous studies have reported interaction between the FHA and CC2 domains (Fig. 6A), suggesting that disruption of this interaction is involved in the activation of KIF1A/UNC-104 proteins (Lee et al., 2004). However, it remained unclear how this disruption affects the overall conformation of full-length KIF1A/UNC-104 proteins. In this study, we analyzed the G618R mutation, which disrupts the autoinhibitory interaction between the FHA and CC2 domains. We showed that G618R almost completely converts monomeric DmUNC-104 into dimers. Furthermore, the bris mutation, which putatively interferes with the FHA–CC2 interaction, also promoted the dimerization of DmUNC-104. Our data suggest that the interaction between the FHA and CC2 domains is the key regulator for releasing the autoinhibition of UNC-104 (Fig. 6A,B). However, the FHA and CC2 domains might not be solely responsible for the autoinhibition. We checked the positions of the disease-associated mutations in the KLP-6 structure. Disease mutations that induce dimerization of DmUNC-104 are located in the MD and do not interact with either the FHA domain or CC2 domain (Fig. S3). Given that these mutations do not seem to affect any known interdomain interactions, it is unclear how they induce dimerization of DmUNC-104. These observations suggest that domains other than the FHA or CC2 may also have strong effects on the autoinhibition of UNC-104.

A recent study revealed the structure of a kinesin-3 member protein, KLP-6 (kinesin-like protein 6 in C. elegans), showing that the entire molecule cooperatively self-folds to form an inactive monomer (Fig. 6A) (Wang et al., 2022). The study showed that the FHA domain interacts not only with CC2 but also with MD and CC1 (Fig. 6A). However, we have shown that the biochemical properties of the CC2 domain are totally different between KLP-6 and UNC-104 (Kita et al., 2024). Therefore, it was not known whether the structure of KLP-6 could be applied to other kinesin-3 family members, including UNC-104. Here, we show that the bris mutation (R562H) induces the dimerization of DmUNC-104. In the KLP-6 structure, R538, which is equivalent to R562 of DmUNC-104, is at the contact site of the FHA domain and CC2 (Fig. S3). Our experimental results using DmUNC-104 are consistent with the expectation, derived from the structure of KLP-6, that the folded structure of KLP-6 is similar to that of UNC-104. The properties of DmUNC-104^bris are similar to those of the hinge mutant DmUNC-104^G618R, for which the FHA–CC2 interaction is predicted to be disrupted (Fig. 6B).

The bris mutation was initially characterized as a loss-of-function mutation through genetic assays (Kern et al., 2013). However, our biochemical analyses and genetic experiments in C. elegans consistently suggest that the bris mutation represents a gain-of-function mutation. One possible explanation is that the initially identified bris mutant phenotypes might arise from a gain-of-function behavior of the DmUNC-104 molecule. Another possibility is that the effect of the release of the autoinhibition of KIF1A/UNC-104 proteins may vary among species and cell types. It is possible that the bris mutation affects the binding of other molecules, including cargos, and the impact of altered interactions may be more severe in Drosophila, causing the observed loss-of-function phenotype. This parallels the known case of kinesin-1, in which disrupted autoinhibition leads to a loss-of-function phenotype in Drosophila (Moua et al., 2011). To investigate these possibilities, further genetic studies using Drosophila will be required.

In our previous work, we demonstrated that some disease-associated KIF1A mutations, specifically V8M, A255V and R350G, increase the activity of KIF1A (Chiba et al., 2019). However, the precise molecular mechanism remained to be clarified. In the current study, we found that the V8M and R350G mutations induce the dimerization of DmUNC-104. Considering that disrupted autoinhibition, induced by the bris and hinge mutations, promoted dimerization of DmUNC-104 (Figs 2 and 4), these disease-association mutations would release the autoinhibition of KIF1A and enhance the activity. A similar observation was found in an amyotrophic lateral sclerosis-related mutant of KIF5A, a member of conventional kinesin. The amyotrophic lateral sclerosis-associated mutations also result in disrupted autoinhibition and increased activation of another type of kinesin, KIF5A (Baron et al., 2022; Nakano et al., 2022; Pant et al., 2022; Chiba and Niwa, 2024). Excess motor activity might be a common mechanism among different neurodegenerative diseases. A recent study showed that defective autoinhibition of kinesin-1 decreases dynein-mediated retrograde transport in fungi (Qiu et al., 2023). Given that mutations in dynein or dynactin are implicated in neurodegeneration (Cianfrocco et al., 2015), defective retrograde transport might be a possible mechanism underlying human diseases caused by gain-of-function behavior of kinesins. These perspectives indicate the importance of precise motor regulation in maintaining neuronal function. Understanding the molecular mechanisms underlying endogenous KIF1A/UNC-104 protein dimerization will provide insights for identifying potential therapeutic targets.

We have not determined how a single point mutation can cause dimerization of the entire length of DmUNC104, despite the many intramolecular interactions in the full-length DmUNC-104. The mechanism by which the wild type forms dimers is also unknown. In vitro reconstitutions using previously identified cargo adaptors such as DENN, Liprin-α and ARL-8 are required for a comprehensive understanding (Shin et al., 2003; Niwa et al., 2008; Wu et al., 2013).

C. elegans strains were maintained as described previously (Brenner, 1974). N2 wild-type worms and OP50 feeder bacteria were obtained from the C. elegans genetic center (CGC) (Minneapolis, MN, USA). arl-8(wy271) was described previously (Wu et al., 2013; Niwa et al., 2016). Strains used in this study are described in Table S2.

The unc-104(R551H) mutation was introduced using the Alt-R^® CRISPR/Cas9 system (Integrated DNA Technologies). Alt-R^® CRISPR-Cas9 tracrRNA (1072533), crRNA for unc-104 and Alt-R^®S.p. Cas9 (1081058) were purchased from Integrated DNA Technologies. The target sequence was 5′-ACAGGATCTAGAGTTATTCT-3′ (Fig. S4). The repair template was 5′-ATAAATGGAAAACAAGTGACAACTCCTACTGTATTACACACAGGATCCCACGTTATCCTAGGTGAACATCACGTTTTCCGATATAATGATCCACAGGAAG-3′.

The injection mix was prepared as described with a slight modification (Ghanta et al., 2021). We excluded the rol-6 marker because successful injections should produce mutants with unc phenotypes as a result of deletion mutations of unc-104 caused by repair failures. Three days after the injection, we selected a few plates that contained worms with strong unc phenotypes. Then, unc mutant worms as well as superficially wild-type worms were singled out from these plates. Seven days later, plates that contained worms with the unc-104(R551H) allele were identified by genomic PCR and BamHI (NEB) digestion. Thirty percent of the plates contained either unc-104(R551H) heterozygotes or homozygotes.

cDNA of Drosophila unc-104 (Dmunc-104) was described previously (Pack-Chung et al., 2007). Dmunc-104 cDNA fragments were amplified by PCR and cloned into the pAcebac1 vector with C-terminal sfGFP and 2×Strep-tag II (Kita et al., 2024) by Gibson assembly (Gibson et al., 2009). G618R, R562H(bris), V6M, A255V and R348G mutations were introduced in the pACEBac1_Dmunc-104::sfGFP::Strep-tag II with the following primers. The sequences of the plasmids were verified by Sanger sequencing. Plasmids used in the study are summarized in Table S1 and have been deposited at Addgene (Addgene IDs: 226775, 226776, 226777, 226778, 226779 and 226780). Sequences of primers used for introducing mutations were: G618R_F, 5′-gcgaattgctcgagaagcaacgcattgatctaaaagc-3′; G618R_R, 5′-gcttttagatcaatgcgttgcttctcgagcaattcgc-3′; R562H_F, 5′-cttaagaccggttctcacgtgatcctcggaaag-3′; R562H_R, 5′-ctttccgaggatcacgtgagaaccggtcttaag-3′; V6M_F, 5′-gcggaattcatgtcgtcggttaagatggcggtgcgagtgcg-3′; V6M_R, 5′-cgcactcgcaccgccatcttaaccgacgacatgaattccgc-3′; A255V_F, 5′-cttggccgggtcggaacgagtagattccactggtgccaaggg-3′; A255V_R, 5′-cccttggcaccagtggaatctactcgttccgacccggccaag-3′; R348G_F, 5′-gcacactgcgctatgcggatggtgccaagcaaattgtttgcaagg-3′; R348G_R, 5′-ccttgcaaacaatttgcttggcaccatccgcatagcgcagtgtgc-3′.

Sf9 cells (Thermo Fisher Scientific) were maintained in Sf900™ II SFM (Thermo Fisher Scientific) at 27°C. DH10Bac (Thermo Fisher Scientific) were transformed with the plasmids to generate bacmid. To prepare baculovirus, 1×10⁶ of Sf9 cells were transferred to each well of a tissue-culture treated 6-well plate. After the cells attached to the bottom of the dishes, ∼5 μg of bacmid were transfected using 5 μl of TransIT^®-Insect transfection reagent (Takara Bio Inc.). Five days after initial transfection, the culture media were spun at 15,000 g for 1 min to obtain the supernatant (P1). For protein expression, 400 ml of Sf9 cells (2×10⁶ cells/ml) were infected with 200 µl of P1 virus and cultured for 65 h at 27°C. Cells were harvested and stocked at −80°C.

Sf9 cells were resuspended in 25 ml of lysis buffer (50 mM HEPES-KOH, 150 mM KCH₃COO, 2 mM MgSO₄, 10% glycerol, pH 7.5) along with 0.1 mM ATP, 1 mM phenylmethylsulfonyl fluoride, 0.1 mM 4-benzenesulfonyl fluoride hydrochloride, 0.1 µM Aprotinin, 5 µM Bestatin, 2 µM E-64, 2 µM Leupeptin and 1 µM Pepstatin A. Then, 0.5% Triton X-100 were added to lyse the cells. After incubating on ice for 10 min, lysates were cleared by centrifugation (50,000 g, 20 min, 4°C) and subjected to affinity chromatography as described below.

Lysate was loaded on 5 ml of Streptactin-XT resin (IBA Lifesciences) and passed through the resin by gravity flow. The resin was washed with 30 ml of lysis buffer. Protein was eluted with 25 ml of elution buffer (50 mM HEPES-KOH, 150 mM KCH₃COO, 2 mM MgSO₄, 10% glycerol, 100 mM D-biotin, pH 7.5). Eluted protein was concentrated using centrifugal filters with MWCO 50K (AS ONE Corporation, 4-2669-05), flash-frozen in liquid nitrogen and stored at −80°C. SEC was performed using an NGC chromatography system (Bio-Rad) equipped with BioSep-SEC-s4000, particle size 5 mm, pore size 500A° and a 7.8 mm ID×600 mm column (Phenomenex) equilibrated in GF150 buffer (25 mM HEPES-KOH, 150 mM KCl, 2 mM MgCl₂, pH 7.2). Fractions were analyzed by SDS-PAGE and visualized using the stain-free protocol (Bio-Rad) with ChemiDoc™ Imaging System (Bio-Rad). Full scans of the entire original gels are shown (Fig. S6). Peak fractions were collected and concentrated using an Amicon Ultra-0.5 centrifugal filter (Merck). Protein concentration was determined by the absorbance of sfGFP at 488 nm using a NanoDrop™ spectrophotometer (Thermo Fisher Scientific). Concentrations were described assuming all molecules were monomers. Proteins at 2-4 µM were flash-frozen in liquid nitrogen and stored at −80°C for further analysis.

Affinity-purified DmUNC-104 was diluted to 15 µM (∼3.3 mg/ml) using lysis buffer and incubated with 2.5 mM BS³ (Thermo Fisher Scientific) for 30 min at room temperature. The reaction was quenched by adding 50 mM Tris-HCl (pH 7.5) and incubating for 15 min at room temperature. The reaction mixture was then subjected to SEC.

For the re-injection experiment, proteins separated by initial SEC were kept at 4°C for 24 h. Top fractions of each peak were combined and concentrated to a quarter of the initial volume using an Amicon Ultra-0.5 centrifugal filter (Merck). The concentrated proteins were then re-injected for a second chromatography.

Proteins were thawed and diluted to a final concentration of 20 nM (as monomers) in GF150 buffer. Mass photometry was performed using a Refeyn OneMP mass photometer (Refeyn, Japan) and Refeyn AcquireMP version 2.3 software, with default parameters set by Refeyn AcquireMP. Bovine serum albumin (BSA) was used as a control to determine the molecular weight. The results were subsequently analyzed using Refeyn DiscoverMP version 2.3, and graphs were prepared to show the distribution of molecular weight.

For preparing microtubules for TIRF assays, fresh pig brains were obtained from the Shibaura Slaughterhouse in Tokyo. Tubulin was purified from the brain as described (Castoldi and Popov, 2003). Tubulin was labeled with Biotin-PEG₂-NHS ester (Tokyo Chemical Industry) and AZDye647 NHS ester (Fluoroprobes) as previously described (Chiba et al., 2022). To polymerize Taxol-stabilized microtubules labeled with biotin and AZDye647, 30 μM unlabeled tubulin, 1.5 μM biotin-labeled tubulin and 1.5 μM AZDye647-labeled tubulin were mixed in BRB80 buffer supplemented with 1 mM GTP and incubated for 15 min at 37°C. Then, an equal amount of BRB80 supplemented with 40 μM Taxol was added and further incubated for 30 min. The solution was loaded on BRB80 supplemented with 30% sucrose and 20 μM Taxol and ultracentrifuged at 100,000 g for 5 min at 30°C. The pellet containing polymerized microtubules was resuspended in BRB80 supplemented with 20 μM Taxol and used for TIRF assays. Glass chambers were prepared by acid washing as previously described (Tan et al., 2018). Glass chambers were coated with PLL-PEG-biotin (SuSoS). Polymerized microtubules were flowed into streptavidin-adsorbed flow chambers and allowed to adhere for 5-10 min. Unbound microtubules were washed away using assay buffer [90 mM HEPES-KOH, 50 mM KCH₃COO, 2 mM Mg(CH₃COO)₂, 1 mM EGTA, 10% glycerol, pH 7.4] supplemented with 0.1 mg/ml biotin–BSA, 0.2 mg/ml kappa-casein, 0.5% Pluronic F127, 2 mM ATP, and an oxygen-scavenging system composed of PCA/PCD/Trolox. Purified motor protein was diluted to indicated concentrations in the assay buffer and flowed into the glass chamber. An ECLIPSE Ti2-E microscope equipped with a CFI Apochromat TIRF 100XC oil objective lens (1.49 NA), an Andor iXion life 897 camera and a Ti2-LAPP illumination system (Nikon) was used to observe single molecule motility. NIS-Elements AR software ver. 5.2 (Nikon) was used to control the system.

Statistical analyses were performed using Graph Pad Prism version 10. Statistical methods are described in the figure legends. Graphs were also prepared using Graph Pad Prism version 10. Alignment of amino acid sequences were performed using Clustal WS with default settings of Jalview (Clustal W and Clustal X version 2.0) (Larkin et al., 2007).

During the preparation of this manuscript, we used ChatGPT in order to check English grammar and improve English writing. After using this tool, we reviewed and edited the content as needed and take full responsibility for the content of the publication.

We thank Thomas Schwarz (Harvard Medical School) for generously providing the cDNA of Drosophila unc-104. We thank Atsushi Nakagawa and Jiye Wang (Osaka University) for help with mass photometry. Some worm strains and OP50 were obtained from the Caenorhabditis Genetics Center. This work was performed under the Collaborative Research Program of Institute for Protein Research, Osaka University (CR-23-02).

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.262017.reviewer-comments.pdf