Cytoplasmic dynein-driven movement of chromosomes during prophase I of mammalian meiosis is essential for synapsis and genetic exchange. Dynein connects to chromosome telomeres via KASH5 and SUN1 or SUN2, which together span the nuclear envelope. Here, we show that KASH5 promotes dynein motility in vitro, and cytosolic KASH5 inhibits dynein’s interphase functions. KASH5 interacts with a dynein light intermediate chain (DYNC1LI1 or DYNC1LI2) via a conserved helix in the LIC C-terminal, and this region is also needed for dynein’s recruitment to other cellular membranes. KASH5’s N-terminal EF-hands are essential as the interaction with dynein is disrupted by mutation of key calcium-binding residues, although it is not regulated by cellular calcium levels. Dynein can be recruited to KASH5 at the nuclear envelope independently of dynactin, while LIS1 is essential for dynactin incorporation into the KASH5–dynein complex. Altogether, we show that the transmembrane protein KASH5 is an activating adaptor for dynein and shed light on the hierarchy of assembly of KASH5–dynein–dynactin complexes.

To conceive healthy offspring, a paternal sperm and maternal egg must be created, which requires the specialized form of cell division, meiosis, in which maternal and paternal homologs pair in prophase I to allow the swap of genetic material by synapsis, generating genetically distinct haploid daughter cells. This is facilitated by chromosomes attaching to the nuclear envelope (NE), usually via their telomeres (Kim et al., 2022; Rubin et al., 2020). In vertebrates and many other organisms, the chromosomes move along the inner nuclear membrane (INM) to transiently cluster in a “meiotic bouquet” (Fig. S1 A) that brings homologs into close proximity and promotes the formation of the synaptonemal complex in zygotene, enhancing synapsis and recombination (Burke, 2018; Kim et al., 2022; Rubin et al., 2020). In many organisms, these dynamic chromosome movements require force generated by the microtubule motor cytoplasmic dynein-1 (dynein, hereafter) either acting directly on the chromosomes (Horn et al., 2013; Lee et al., 2015; Morimoto et al., 2012; Rog and Dernburg, 2015; Sato et al., 2009; Wynne et al., 2012) or indirectly by driving nuclear movement (Burke, 2018; Lee et al., 2015; Rubin et al., 2020). Crucially, the force must be transmitted from the cytoplasm to the chromosomes on the other side of the NE. This is achieved by linkers of nucleoplasm and cytoplasm (LINC) complexes, which span the NE to physically connect the cytoskeleton and nucleus (Burke, 2018; Sato et al., 2009; Spindler et al., 2019; Fig. S1 A).

LINC complexes consist of SUN (Sad1, Unc-84) domain proteins that span the INM, binding nuclear lamins and chromatin-associated proteins in the nucleoplasm and interacting with KASH (Klarsicht, ANC-1, Syne Homology) domain proteins. KASH proteins have large cytosolic domains that bind cytoskeleton-associated proteins and are anchored in the outer nuclear membrane (ONM) by a C-terminal transmembrane domain (Burke, 2018; Kim et al., 2022; Rubin et al., 2020). Inside the NE lumen, the short C terminal KASH-domain sequence associates with SUN proteins, restricting KASH proteins to the nuclear membrane (Hao and Starr, 2019; Morimoto et al., 2012; Starr and Han, 2002). The meiotic LINC complex can contain SUN1 and SUN2, which have both overlapping and distinct roles in meiosis (Ding et al., 2007; Lei et al., 2009; Link et al., 2014; Schmitt et al., 2007). Although Sun1−/− mice are sterile (Chi et al., 2009), SUN2 can recruit telomeres to the NM and drive bouquet formation (Link et al., 2014; Schmitt et al., 2007) and allow some prophase chromosome movement (Lee et al., 2015). SUN1 and 2 bind to a meiotic telomere complex (Dunce et al., 2018; Shibuya et al., 2014) and to KASH5 in vertebrates or ZYG-12 in Caenorhabditis elegans, which recruit dynein to the ONM (Fig. S1 A; Horn et al., 2013; Lee et al., 2015; Morimoto et al., 2012; Rog and Dernburg, 2015; Sato et al., 2009; Wynne et al., 2012). KASH5 is essential for mammalian meiosis as Kash5−/− mice are completely sterile due to impaired synapsis, accumulation of double-stranded DNA breaks, and resulting meiotic arrest (Horn et al., 2013). Chromosome movement and nuclear rotation in prophase I are also lost (Lee et al., 2015). Human mutations that cause mistargeting of KASH5 to mitochondria lead to male sterility (Bentebbal et al., 2021). KASH5 shares sequence homology with the N-terminal region of the protein encoded by the zebrafish gene futile cycle, which is required for pronuclear migration (Dekens et al., 2003; Lindeman and Pelegri, 2012), another dynein-dependent function that drives the female pronucleus toward the male pronucleus along microtubules nucleated from the male centrosome (Gönczy et al., 1999; Payne et al., 2003; Reinsch and Karsenti, 1997; Robinson et al., 1999). ZYG-12 performs the same role in C. elegans (Malone et al., 2003), but the KASH protein responsible for pronuclear migration in mammals is not yet known.

Dynein transports a wide range of cargo to the minus end of microtubules (Reck-Peterson et al., 2018). It is a large 1.6 MDa holoenzyme (Fig. S1 B) comprised of two heavy chains (DHC: DYNC1H1) containing the globular motor domains, which are the sites of ATP hydrolysis. Its cargo binding tail domain contains the intermediate chains (ICs: DYNC1I1 and 2) that bind directly to the DHCs; three light chains that bind to the ICs; and two light intermediate chains (LICs: DYNC1LI1 and 2; Pfister et al., 2005; Reck-Peterson et al., 2018). The LICs bind to DHC via their N terminal GTPase-like domain, which is highly conserved between LICs 1 and 2 (Schroeder et al., 2014). The unstructured carboxy terminus protrudes from the motor complex (Celestino et al., 2019; Lee et al., 2018; Schroeder et al., 2014) and is less homologous between LIC1 and 2 apart from two regions of predicted alpha-helix that link dynein to cargo adaptors (Celestino et al., 2022; Celestino et al., 2019; Kumari et al., 2021b; Lee et al., 2020; Lee et al., 2018).

Dynein requires the multicomponent dynactin complex (Fig. S1 B) for function (Feng et al., 2020; Gill et al., 1991; King et al., 2003; McKenney et al., 2014; Schlager et al., 2014a; Schroer and Sheetz, 1991), although dynein and dynactin alone interact weakly (Baumbach et al., 2017; Chowdhury et al., 2015; Jha et al., 2017; Splinter et al., 2012; Urnavicius et al., 2015). An “activating adaptor” (Fig. S1 B) is needed for optimum motility and force generation in vitro (Belyy et al., 2016; McKenney et al., 2014; Schlager et al., 2014a; for reviews see Canty and Yildiz, 2020; Olenick and Holzbaur, 2019; Reck-Peterson et al., 2018). They strengthen the interaction between dynein and dynactin (Schroeder and Vale, 2016; Splinter et al., 2012) by forming extensive interactions with both components to generate a tripartite dynein–dynactin–adaptor (DDA) complex or complexes with two dyneins per dynactin and an adaptor (D2DA; Chowdhury et al., 2015; Grotjahn et al., 2018; Lau et al., 2021; Lee et al., 2020; Urnavicius et al., 2018; Urnavicius et al., 2015) or two dyneins, one dynactin, and two adaptors (Chaaban and Carter, 2022). Adaptor binding releases the dynein motor domains from the autoinhibited Phi conformation (Torisawa et al., 2014; Zhang et al., 2017), helping to align them for microtubule interaction (Chowdhury et al., 2015; Zhang et al., 2017). Finally, adaptors recruit dynein and dynactin to cargo (e.g., Horgan et al., 2010b; Schroeder and Vale, 2016; Splinter et al., 2012; Wang et al., 2019).

The assembly and function of dynein adaptor complexes are promoted by Lissencephaly-1 (LIS1). LIS1 is needed for many dynein functions (Markus et al., 2020) and is mutated in the neurodevelopmental disorder, Type-1 lissencephaly (Reiner et al., 1993). LIS1 binding to the dynein motor domain opens the Phi complex (Elshenawy et al., 2020; Gillies et al., 2022; Htet et al., 2020; Marzo et al., 2020; Qiu et al., 2019) and enhances the formation of DDA complexes (Baumbach et al., 2017; Dix et al., 2013), increasing the frequency, velocity, and duration of dynein movement (Baumbach et al., 2017; Dix et al., 2013; Fenton et al., 2021; Gutierrez et al., 2017). LIS1 also increases the proportion of the two dynein D2DA complexes (Elshenawy et al., 2020; Htet et al., 2020), which generate more force, move faster, and are more processive than DDA complexes (Elshenawy et al., 2019; Sladewski et al., 2018; Urnavicius et al., 2018). Furthermore, LIS1 contributes to the recruitment of dynein, dynactin, and/or adaptors to a wide range of cellular cargoes in interphase and mitosis (Cockell et al., 2004; Dix et al., 2013; Dzhindzhev et al., 2005; Lam et al., 2010; Siller et al., 2005; Sitaram et al., 2012; Splinter et al., 2012; Wang et al., 2013).

Activating adaptor proteins share little sequence homology, but generally contain a long coiled-coil domain, and a site for cargo binding (Lee et al., 2020; McKenney et al., 2014; Redwine et al., 2017; Schlager et al., 2014a; Schlager et al., 2014b; Urnavicius et al., 2015). They bind the C-terminal domain of LICs via at least three distinct types of sequence motif—the CC1 box, the Hook domain, or EF hands (Celestino et al., 2019; Gama et al., 2017; Lee et al., 2020; Lee et al., 2018; Schroeder and Vale, 2016)—and the motility of all three types of complexes is promoted by LIS1 (Htet et al., 2020). Other adaptors may not activate dynein but still link the motor to cargos (Olenick and Holzbaur, 2019; Reck-Peterson et al., 2018). For example, Rab7-interacting lysosomal protein (RILP) recruits dynein and dynactin to Rab7-positive late endosomes/lysosomes (Johansson et al., 2007; Jordens et al., 2001; Scherer et al., 2014; Tan et al., 2011) via the binding of its RILP homology domains to helix 1 of LIC (Celestino et al., 2022), but RILP lacks a long coiled coil. The KASH proteins KASH5 and ZYG-12 are good activating adaptor candidates because they bind dynein and have long coiled coils, with KASH5 having an N-terminal EF hand domain (Horn et al., 2013; Morimoto et al., 2012), whereas ZYG-12 has a Hook domain (Malone et al., 2003). However, unlike other activating adaptors identified so far, they are transmembrane proteins.

Mammalian dynein contains either two LIC1 subunits or two LIC2 subunits (Tynan et al., 2000a), providing the potential for differential interactions with adaptors. However, both LICs bind to the adaptors Rab11-FIP3 (Celestino et al., 2019; Horgan et al., 2010a; Horgan et al., 2010b), RILP (Celestino et al., 2019; Scherer et al., 2014), BICD2, spindly, Hook3, ninein, and TRAK1 (Celestino et al., 2019) via the highly conserved helix 1 in the LIC C-terminus. This shared binding ability suggests that LICs may act redundantly, as has been reported in the endocytic pathway (Horgan et al., 2010a; Horgan et al., 2010b; Tan et al., 2011). However, other endocytic functions may be isoform-specific (Bielli et al., 2001; Hunt et al., 2013; Palmer et al., 2009). Likewise, although LIC1 and 2 act redundantly for some mitotic functions (Jones et al., 2014; Raaijmakers et al., 2013), isoform-specific functions and localizations have been reported in mitosis (Horgan et al., 2011; Mahale et al., 2016a; Mahale et al., 2016b; Palmer et al., 2009; Raaijmakers et al., 2013; Sharma et al., 2020; Sivaram et al., 2009), at the centrosome (Tynan et al., 2000b), during neuronal nuclear migration (Goncalves et al., 2019), and cell migration (Even et al., 2019; Schmoranzer et al., 2009). In addition, mitotic phosphorylation of the LIC1 C-terminal domain may offer temporal control of adaptor selection (Kumari et al., 2021a). The degree of specific and shared functions for LICs is an important issue that is not fully resolved.

How the dynein motor interacts with KASH5 to drive the dynamic chromosome movements essential for mammalian meiosis has been poorly understood. Here, we reveal KASH5 to be a dynein-activating adaptor that interacts with the LIC helix 1 region, in agreement with recent work (Agrawal et al., 2022). We show that this region in the LIC is also key for dynein’s function at the Golgi apparatus and throughout the endocytic pathway, with LICs 1 and 2 acting redundantly. Analysis of the hierarchy of adaptor complex assembly reveals that dynein can be recruited to KASH5 independently of dynactin and that LIS1 is essential for full complex assembly. The interaction between KASH5 and dynein is disrupted by mutation of the KASH5 EF-hand domain, although dynein recruitment to KASH5 in cells is calcium-independent. Since defective synapsis can lead to genetic abnormalities and infertility, the characterization of the KASH5–dynein interaction is an important step forward in understanding the complex mechanism of chromosome movement during mammalian meiotic prophase I.

As KASH5 is expressed only in the meiotic germ line, HeLa cells provide a convenient “KASH5-null” background for studying the interaction between dynein and KASH5 (Horn et al., 2013; Morimoto et al., 2012). We generated a stable HeLa cell line in which expression of GFP-KASH5 was induced by doxycycline so that the recruitment of endogenous dynein and dynactin to the nuclei of KASH5-expressing cells could be visualized by immunofluorescence. LIC1 and dynactin p50 were both recruited to KASH5-expressing nuclei (Fig. 1 A), in addition to dynactin p150 and IC (Horn et al., 2013). In contrast, neither dynein nor dynactin was recruited to the nuclei of cells transiently expressing a different KASH protein, nesprin-2α2 (GFP-N2α2). Recruitment of dynein to the nuclear envelope in untransfected cells was rare since it only occurs in late G2-prophase (Baffet et al., 2015; Hu et al., 2013; Salina et al., 2002; Splinter et al., 2012; Splinter et al., 2010), and examples where LIC1 is not seen at the NE can be seen in Fig. 6 B, Fig. 8 C, and Fig. S5 A.

We confirmed the recruitment of dynein and dynactin to KASH5 biochemically. HeLaM cells were transiently transfected with GFP-KASH5ΔK (Fig. 1 B) or GFP-N2α2ΔK, which lack their KASH and transmembrane domains, and so are cytosolic. LIC1, LIC2, IC, and p150 are all associated with GFP-KASH5ΔK and not with GFP-N2α2ΔK (Fig. 1 C). The KASH5 N-terminal 166 amino acids (GFP-KASH5ND), containing the EF-hands, was able to pull down dynein IC and dynactin from cell lysates (Horn et al., 2013), and we found that it was somewhat more efficient than GFP-KASH5ΔK for LIC1 (Fig. 1 D), as seen for IC (Horn et al., 2013).

LIS1 plays a key role in the assembly of DDA and D2DA complexes (reviewed in Canty and Yildiz, 2020; Markus et al., 2020; Olenick and Holzbaur, 2019) and coprecipitates with BICD2N, dynein, and dynactin (Splinter et al., 2012). As reported previously (Horn et al., 2013), GFP-KASH5ΔK pulled down LIS1 as well as dynein and dynactin, but this complex excluded the dynein adaptor BICD2 (Fig. 1 C). Immunofluorescence analysis revealed that endogenous LIS1 was detected at the NE in 99.8% of KASH5-expressing cells (Fig. 1 A and Fig. S1 C), whereas the LIS1 and dynein interactor Nde1 was not, even though it could be detected at the NE in late G2 cells (Fig. 1 A).

We next sought to identify which dynein subunits mediate the KASH5 interaction and the roles played by dynactin and LIS1. We used RNA interference to deplete individual dynein subunits in the GFP-KASH5 cell line (Fig. S2 A) and assessed the effect on recruitment to KASH5. Depletion of dynein IC2 (the only form expressed in HeLa cells [Palmer et al., 2009]) did not prevent dynein recruitment (Fig. 2 A), with 100% of KASH5-positive nuclei being labeled with anti-LIC1 (Fig. S1 C). Cytosolic levels of LIC1 were reduced following IC2 depletion, making the NE pool particularly distinct. This is likely due to a modest reduction in total dynein levels when IC2 is depleted, as seen by immunoblotting with anti-LIC1 and 2 (Fig. S2 A). LIS1 recruitment to KASH5 at the NE was also unaffected by the depletion of IC2 (Fig. 2 A and Fig. S1 C). Strikingly, dynactin was only rarely detected at the NE in IC2-depleted cells using antibodies to p150 or p50 (Fig. 2 A and Fig. S1 C), even though p150 was still readily observed at microtubule plus ends. These data suggest that while the interaction between IC and p150 (Karki and Holzbaur, 1995; King et al., 2003; Vaughan and Vallee, 1995) is not needed for dynein and LIS1 recruitment to KASH5, it is essential for the association of dynactin with KASH5.

Since IC2 depletion reduced total cellular dynactin p150 levels by ∼25% (Fig. S2 A), we used a dominant negative approach as another way of testing the effect of disrupting IC2–p150 interactions on recruitment to KASH5 using overexpression of the coiled coil 1 region of p150 (CC1; King et al., 2003; Quintyne et al., 1999). Unfortunately, transiently transfected constructs would not coexpress with GFP-KASH5 in the inducible GFP-KASH5 cell line. We therefore transiently co-transfected HeLa cells with GFP-KASH5 and myc-SUN2 (to enhance localization of KASH5 to the NE) along with RFP-tagged CC1. Notably, while CC1 expression had no effect on dynein or LIS1 recruitment to KASH5, it prevented dynactin accumulation at the NE (Fig. 2 B), confirming that dynein and LIS1 can associate with KASH5 independently of dynactin. As SUN1 and SUN2 have been shown to play a role in Golgi apparatus position and organization (Hieda et al., 2021), the expression of additional SUN2 could potentially alter Golgi apparatus morphology. However, we did not use coexpression of SUN2 and KASH5 in any assays of organelle positioning.

We used RNAi to investigate the involvement of LICs in the dynein–KASH5 interaction. There was no reduction in the proportion of cells with detectable dynein, dynactin, or LIS1 recruited to KASH5 after depleting LIC2 alone (Fig. 3, B and D). LIC1 depletion also had very little effect on dynactin or LIS1 recruitment in a binary scoring assay but had a variable effect on the dynein intermediate chain, with 80.8 ± 24.3% of cells (n = 3 experiments, ±SD) showing IC signal at the NE (Fig. 3, A and D). However, when both LIC1 and LIC2 were depleted simultaneously, the proportion of cells with detectable dynein and dynactin at the NE was reduced by ∼75% (Fig. 3, C and D). LIS1 recruitment was also reduced, with ∼40% of KASH5 expressing cells showing no LIS1 signal at the NE. We were not able to deplete endogenous LIC1 completely using RNAi (Fig. S2 B), which may explain why in some cells a residual level of dynein, dynactin, and LIS1 remained with KASH5. In addition, GFP–KASH5ΔK pull-downs from LIC1 and 2 depleted HeLa cells contained very little dynein and dynactin (Fig. 3 E). Thus, LIC1 and 2 act redundantly to recruit dynein and LIS1 to KASH5 at the NE, with dynactin being recruited downstream, by a mechanism requiring the interaction between IC and p150.

To test if LIS1 is required for KASH5 to associate with dynein and dynactin, we depleted LIS1 using RNAi. Interestingly, the proportion of cells with LIC1 detectable at the NE was virtually unaffected by LIS1 knockdown (Fig. S1, D and E) whereas p150 recruitment was seen in only 20% of cells (Fig. S1, D and G). Dynein IC detection at KASH5-positive nuclei varied between experiments (Fig. S1, D and F), with a mean of 57.1% of cells positive for IC74 antibody labeling. The discrepancy between the anti-LIC1 and IC scoring most likely reflects the weaker NE labeling seen with the IC74 antibody compared with anti-LIC1 antibodies in control cells. Our interpretation of these data is that LIS1 depletion reduces dynein levels at the NE somewhat to levels that are still detectable by anti-LIC1 but sometimes not by IC74. Altogether, these data reveal that LIS1 is vital for the dynactin complex to be recruited to KASH5 downstream of dynein and suggest that LIS1 may also promote the formation of or stabilize the KASH5–dynein complex.

We next wanted to ascertain the LIC1 region responsible for the dynein–KASH5 interaction. Assays with purified protein have shown that helix 1 and some preceding amino acids (433–458) formed the minimal LIC1 region needed for interaction with Hooks, BICD2, and Spindly (Lee et al., 2018), with additional interactions between LIC1 and Spindly, BICD2, and Hook3 being seen in helix 2 and the linker region upstream of helix 1 (Celestino et al., 2019). In contrast, RILP required only helix 1 (LIC1 440–455; Celestino et al., 2019). We made a series of hLIC1 constructs (Fig. 4 A): full-length LIC1, a truncation containing helix 1 but not helix 2 (LIC1-CT2: amino acids 1–456) and a construct that terminates shortly after the Ras-like domain and lacks both helices and the linker sequences (LIC1-CT3: amino acids 1–388). Cells depleted of LIC1 and 2 by RNAi were cotransfected with myc-tagged LIC1 constructs and the soluble GFP-KASHND. GFP trap IP revealed that while dynein containing LIC1-FL and LIC1-CT2 was recruited to the KASH5 N-terminal domain, dynein with LIC1-CT3 was not (Fig. 4 B). Previous work has shown that deletion of just helix 1 (N440-T456) or helix 1 point mutants (F447A/F448A and L451A/L452A) prevented binding of the LIC1 C-terminal domain to several dynein adaptors using purified proteins (Celestino et al., 2019). We generated the helix 1 deletion and mutations in GFP-LIC1 (Fig. 4 A) and also a deletion lacking helix 1 plus some linker sequence (S433-S458). These were coexpressed with HA-tagged KASH5ΔK in HeLaM cells depleted of both LICs. Full-length GFP-LIC1 pulled down KASH5ΔK effectively, whereas the LIC1 helix 1 deletions and point mutants did not (Fig. 4 C). We then examined the recruitment of LIC1 to KASH5 at the NE in HeLaM cells coexpressing HA-KASH5, myc-SUN2, and GFP-LIC1 constructs. Both wild-type GFP-LIC1 and the mid-length LIC1-CT2 were recruited to KASH5, although CT2 detection at the NE was variable (Fig. 4, D and E). In contrast, LIC1-CT3, the helix 1 deletions, and helix 1 point mutants were poorly recruited. Altogether, these results demonstrate LIC1 helix 1 is essential for interaction with KASH5.

The importance of LIC helix 1 in cargo binding prompted us to assess the role of the LICs on endogenous membrane cargos whose adaptors have not been fully defined. First, we tested if LIC1 and 2 functioned redundantly in Golgi apparatus morphology and positioning, for which there is conflicting published evidence (Kumari et al., 2021a; Palmer et al., 2009; Tan et al., 2011). RNAi depletion of LIC1 or 2 individually led to the break-up of the Golgi ribbon in ∼60% of HeLaM cells, with the Golgi fragments remaining centrally located (Fig. 5, A–C). In contrast, the depletion of both LICs simultaneously led to complete fragmentation and scattering of the Golgi apparatus (Fig. 5, B and C), demonstrating that the LICs act redundantly in Golgi positioning. This was confirmed by expressing RNAi-resistant LIC1-mKate or LIC2-mKate in cells depleted of both LICs (Fig. 5 D), as either LIC was able to fully restore Golgi apparatus clustering. The endocytic pathway also relies heavily on dynein activity (Flores-Rodriguez et al., 2011; Granger et al., 2014; Reck-Peterson et al., 2018; Wang et al., 2019), which contributes to the sorting of endocytic cargo in the early endosome (Driskell et al., 2007). Depletion of both LICs profoundly altered the distribution of early endosomes, recycling endosomes, and lysosomes (Fig. S3 A), while only minimal effects were seen with single depletions (not shown). Moreover, the expression of either LIC1-mKate or LIC2-GFP restored the position of endocytic organelles (Fig. S3 B), confirming that LICs act redundantly in this context. We investigated if the same effects were observed with dynein recruitment to RILP. Recruitment of dynein and dynactin to HA-RILP-positive late endosomes was significantly reduced when both LICs were depleted together, but not individually (Fig. S4, A–D).

To examine if the LIC1 region 388–456 that contains helix 1 is needed for Golgi apparatus and endocytic organelle positioning, we used our knock-down and rescue approach (Fig. 5 E, Fig. S3 C and D, and Fig. S4 E). Both full-length GFP-LIC1 and GFP-LIC1-CT2 rescued the Golgi apparatus, early endosome, and lysosome clustering. In contrast, GFP-LIC1-CT3 did not, with EEA1-positive early endosomes (Fig. S3 C) and LAMP1-positive late endosomes/lysosomes (Fig. S3 D) remaining localized in the cell periphery, and the Golgi apparatus was fragmented and scattered (Fig. 5 E). Moreover, GFP-LIC1-CT3 did not interact with RILP, whereas GFP-LIC1-CT2 or full-length GFP-LIC1 were robustly recruited to RILP-positive organelles (Fig. S4 E).

Altogether, these data provide clear in cellulo evidence that the LIC adaptor binding helix 1 is needed for dynein’s interaction with KASH5 and RILP and is also crucial for dynein’s function on the Golgi apparatus, early endosomes, and lysosomes. Furthermore, we demonstrate that LIC1 and LIC2 act redundantly in these situations.

KASH5 shares key properties with other activating dynein adaptors: interaction with LIC helix 1; ability to recruit dynactin; and presence of an N-terminal dynein binding domain (containing EF-hands) followed by an extended coiled coil (Horn et al., 2013). In addition, KASH5’s biological function in telomere clustering in the prophase of meiosis I (Horn et al., 2013; Lee et al., 2015) strongly suggests that it recruits active dynein. When expressed out of its meiotic context, in HeLa cells, we quite often observed clusters of KASH5 and dynein around discrete points close to or on top of the nucleus (red arrowheads and arrows in Fig. 1 A), suggestive of clustering around the centrosome via active dynein. In addition, the asymmetric distribution of KASH5 in the NE with enrichment toward the MTOC has been noted (Horn et al., 2013).

Overexpression of activating dynein adaptor proteins lacking their cargo binding domain disrupts dynein functions by sequestering dynein and preventing its binding to endogenous adaptors (e.g., Hoogenraad et al., 2001; Hoogenraad et al., 2003; Horgan et al., 2010a; Horgan et al., 2010b; Splinter et al., 2012). As KASH5 interacts with LIC helix 1, we hypothesized that it would compete with other adaptors for binding to dynein. Indeed, overexpression of cytosolic GFP-KASH5ΔK in HeLaM cells resulted in complete fragmentation of the Golgi apparatus and redistribution of lysosomes to the cell periphery; phenotypes indicative of a loss of dynein function (Fig. 6 A). In contrast, the expression of a cytoplasmic KASH5 construct lacking its N terminal dynein binding domain (Fig. 1 B, GFP-KASH5ΔNDΔK [Horn et al., 2013]) had no effect on Golgi apparatus or lysosome distribution and morphology (Fig. 6 A). The effects of GFP-KASH5ΔK are therefore due to KASH5 binding and sequestering dynein, preventing its recruitment to other membrane cargoes.

We next determined if KASH5 competes with established dynein adaptors for dynein binding. To test this possibility, we cotransfected HeLaM cells with HA-tagged full-length KASH5, myc-tagged SUN2, and dominant negative versions of the activating adaptors BICD2 (GFP-BICD2N [Hoogenraad et al., 2001; Splinter et al., 2012]) and Rab11-FIP3 (GFP-Rab11-FIP3 I737E, which retains its LIC binding domain but is unable to interact with Rab11 [Wilson et al., 2005]). Endogenous dynein was recruited to KASH5 at the NE in control cells, but this was prevented when dominant-negative dynein adaptors BICD2N and Rab11-FIP3 I737E were expressed (Fig. 6 B). We investigated if the same was true for dynein recruitment to RILP, even though RILP is not thought to be able to activate dynein/dynactin motility (Lee et al., 2020; Reck-Peterson et al., 2018). Dynein was recruited to RILP-positive organelles following overexpression of HA-tagged RILP (Fig. 6 C). However, when GFP-BICD2N or GFP-Rab11-FIP3 I737E were coexpressed, recruitment of dynein LIC1 to RILP was abrogated (Fig. 6 C). Taken together, these findings show that established dynein adaptor proteins compete with KASH5 and RILP for dynein binding.

To test directly if KASH5 could act as an activating adaptor and form motile complexes with dynein and dynactin, we used single molecule in vitro motility assays consisting of purified bacterially expressed KASH5 truncations (Fig. 7 A) mixed with purified fluorescently labeled dynein, dynactin, and LIS1. We generated three KASH5 constructs encoding amino acids 1–407, 1–460, and 1–507, all of which contain the N-terminal EF-hand dynein binding domain plus the predicted coiled-coil region (amino acids 166–350), plus a variable amount of the C-terminal region. SEC-MALS analysis showed that all three proteins formed dimers (Fig. 7 B). Of these, KASH51-460 was the most active in motility assays using dynein containing a heavy chain mutant that cannot form the inhibited Phi conformation (Zhang et al., 2017; Fig. 7 C). We next used assays with wild-type purified dynein, which when combined with dynactin and LIS1 alone displayed very little motility (Fig. 7 D). The inclusion of KASH51-460 promoted processive dynein movements (Fig. 7, D and E), although a lesser number than seen with purified Hook31-522 (Urnavicius et al., 2018). KASH5 EF-hands (amino acids 1–115), the coiled coils (amino acids 155–349), or a combination of EF-hands and coiled coils alone (amino acids 1–349) were not enough to activate dynein/dynactin (Fig. 7 F). Importantly, the velocity of processive dynein movements was the same for KASH51-460 and Hook31-522 (Fig. 7 E). KASH5, a transmembrane protein, is therefore an activating dynein adaptor.

A common feature of several dynein adaptor proteins is the presence of an N terminal pair of EF-hands in the dynein-binding domain, as seen in Rab11-FIP3, CRACR2a, Rab45/RASEF, and ninein (Celestino et al., 2019; Lee et al., 2020; Reck-Peterson et al., 2018; Wang et al., 2019). KASH5 contains two putative EF hands, extending from amino acids 36–103, which form the bulk of the dynein binding domain (Fig. 1, B and D; and Fig. 4 C). Sequence alignment of the KASH5 EF hands with CRACR2a, Rab45, FIP3, and ninein revealed that KASH5, like FIP3 and ninein, lacks some key consensus calcium-binding amino acids (at positions X,Y,Z,-X,-Y,-Z), with these changes being consistent across species (Fig. 8 A). EF-hand 1 is particularly divergent, with only the residue in position X (Grabarek, 2006) matching the consensus, and with the key position -Z being a glutamine or histidine instead of a glutamate residue. While EF-hand 2 has consensus amino acids in position X, other residues either do not conform or vary between species. For example, in non-rodent KASH5 EF-hand 2, there is aspartate in place of glutamate at -Z, which can result in magnesium binding (Grabarek, 2006), as suggested for FIP3 (Lee et al., 2020).

This analysis suggested that KASH5 might not be calcium regulated. To determine if the KASH5 and Rab11–FIP3 interaction with dynein in cells required calcium or not, we expressed GFP-tagged KASH5 with myc-SUN2 in HeLaM cells, or GFP-Rab11-FIP3 in Vero cells, and labeled for endogenous LIC1. In control DMSO-treated cells, there was robust recruitment of dynein to KASH5 at the NE and to Rab11-FIP3-positive recycling endosomes (Fig. S5, A and B). This recruitment was not affected by treating cells with the cell-permeable calcium chelator, BAPTA-AM, for 2 h to deplete intracellular calcium. Furthermore, endogenous LIC1 was present at equal levels when GFP-KASH5-ND was pulled down from cell lysates in the presence or absence of EGTA (Fig. S5 C). This demonstrates that the KASH5–dynein interaction does not require calcium and confirms that the Rab11–FIP3–dynein interaction in cells is calcium-independent, as reported for in vitro assays (Lee et al., 2020).

We wanted to test the importance of KASH5’s EF-hands in the dynein interaction, even if calcium was not required. The fue mutation in the zebrafish KASH5 homolog,