MAP7 family proteins are microtubule-tethered allosteric activators of kinesin-1

Kinesin-1 is responsible for microtubule-based transport of numerous cellular cargoes. Here, we explored the regulation of kinesin-1 by MAP7/ensconsin family proteins. We found that all four mammalian MAP7 family members bound to kinesin-1, and MAP7, MAP7D1 and MAP7D3 acted redundantly to enable kinesin-1-dependent transport in HeLa cells. Microtubule recruitment of the truncated kinesin-1 KIF5B-560, which contains the stalk but not the cargo-binding and autoregulatory regions, was inhibited in cells co-depleted of these three MAP7 proteins. In vitro, purified MAP7 and MAP7D3 increased microtubule landing rate and processivity of KIF5B-560. The same was true for MAP7D3 C-terminus, which weakly bound to microtubules and exchanged rapidly on motile KIF5B-560 motors. A C-terminal MAP7 fragment lacking microtubule affinity increased KIF5B-560 recruitment to microtubules in vitro and in cells, and partially rescued kinesin-1-dependent transport in the absence of full-length MAP7 proteins. We propose that MAP7 proteins are microtubule-tethered kinesin-1 activators, with which the motor transiently interacts as it moves along microtubules. Summary A combination of experiments in cells and in vitro reconstitution assays demonstrated that mammalian MAP7 family proteins act redundantly to activate kinesin-1 and promote its microtubule binding and processivity by transiently associating with the stalk region of the motor.

mitochondria, peroxisomes, ribonucleoprotein particles, MT sliding, cytoplasmic flow in oocytes, neuronal development, centrosome separation and nuclear positioning in muscle cells (Barlan et al., 2013;Metivier et al., 2018;Metzger et al., 2012;Monroy et al., 2018;Sung et al., 2008). Also in mammalian myotubes, MAP7 is needed for proper kinesin-1-dependent nuclear distribution (Metzger et al., 2012), but whether this applies to other cell types and kinesin-1-dependent processes in mammals has not been investigated. It is also unknown whether mammalian MAP7 homologues all behave similarly and whether they show different, overlapping or redundant functions.
There are also important unresolved questions concerning the mechanism of cooperation between kinesin-1 and MAP7. In vitro experiments in fly ovary extracts have shown that the minimal dimeric kinesin-1 fragment does not require ensconsin, but the full-length kinesin-1 is no longer able to productively interact with MTs in the absence of ensconsin (Sung et al., 2008). Recent in vitro reconstitution work with purified proteins demonstrated that MAP7 recruited kinesin-1 to MTs and somewhat decreased motor velocity but had only a mild effect on kinesin-1 run length . Importantly, MAP7 was highly immobile in these assays and was not cotransported with the motor, suggesting that MAP7 affects only the initial recruitment of the kinesin to MTs but has little impact on kinesin-1 movement . However, some observations in flies do not agree with this simple model, as it was shown that the Cterminal fragment of ensconsin, which lacks the MT-binding domain (Sung et al., 2008), significantly rescues kinesin-1-related transport deficiencies in cells lacking ensconsin (Barlan et al., 2013;Metivier et al., 2018). Kinesin-1 is well known to be autoinhibited by its C-terminal cargo-binding domains (Verhey and Hammond, 2009), and it was proposed that ensconsin plays a role in relieving autoinhibition of the kinesin (Barlan et al., 2013). This possibility could be in line with the experiments performed in extracts (Sung et al., 2008), but it was not yet tested with purified proteins.
Here, we explored the relationship between kinesin-1 activity and mammalian MAP7 proteins.
We found that all four mammalian MAP7 family members are able to bind kinesin-1. Three of these proteins, MAP7, MAP7D1 and MAP7D3, are expressed in HeLa cells (Kikuchi et al., 2018;Syred et al., 2013), and only co-depletion of all three MAP7's but not their individual loss affected kinesin-1-dependent distribution of mitochondria. Interestingly, in cells lacking these three MAP7 homologues, a fragment of kinesin-1, KIF5B-560 (Case et al., 1997), which contains the motor domain and the dimerizing stalk with the MAP7-binding site, but not the cargo-binding and autoinhibitory domains, lost its MT association. This result was surprising, because this kinesin-1 fragment is thought to be constitutively active. MT recruitment of KIF5B-560 was rescued not only by full-length MAP7s, but also by their C-terminal domains, which lacked the major MT-binding region. These results were recapitulated in in vitro reconstitution assays with purified proteins, including the C-terminus of MAP7D3, which weakly binds to MTs, and the Cterminus of MAP7, which displays no MT binding. The kinesin-binding region of MAP7 can thus activate the motor, possibly by allosterically stabilizing a conformation favorable for MT interaction.
In agreement with published data, we found that MAP7 was immobile on MTs in vitro . In contrast, MAP7D3 interacted with MTs less tightly, and could be observed moving together with KIF5B-560 motors. In spite of these differences, both MAPs increased not only the recruitment of kinesin-1 to MT but also its processivity. Such an effect can only be explained if the interaction between MAP7s and kinesin-1 is transient. This was indeed confirmed by showing that MAP7D3 C-terminus rapidly exchanges on motile KIF5B-560 motors. Taken together, our data show that MAP7 proteins redundantly regulate kinesin-1dependent transport by acting as MT-tethered activators of this kinesin.

MAP7 family members act redundantly in mitochondrial distribution in HeLa cells
To test whether all four MAP7 family members can potentially act as kinesin-1 regulators, we performed a pull down assay with individual tagged MAP7 proteins and KIF5B-560 and found that all four MAP7 homologues could indeed bind to this kinesin-1 deletion mutant (Fig. 1A, B).
Gene expression analysis at the mRNA and protein level indicated that HeLa cells co-express MAP7, MAP7D1 and MAP7D3 (Kikuchi et al., 2018;Syred et al., 2013), and we confirmed these data by antibody staining (Fig. 1C). In contrast, MAP7D2, which is highly expressed in brain tissue (Niida and Yachie, 2011), was not expressed in HeLa cells. To test if all three MAP7's are required for kinesin-1 function, we initially used the distribution of mitochondria as readout, because, as published previously, it strongly depends on kinesin-1 KIF5B (Tanaka et al., 1998). In the absence of KIF5B, mitochondria were no longer dispersed in the cytoplasm but were instead clustered around the nucleus (Fig. 1D, G, H). Next, we generated HeLa cells lacking each individual MAP7 family member, but observed no defects in the localization of mitochondria ( Fig. 1E-H). In addition, no compensatory effects were found within the MAP7 family when one of the MAPs was depleted (Fig. 1E, F).
We next attempted to generate a stable triple knockout of MAP7, MAP7D1 and MAP7D3, but such cells were not viable. This was unlikely due to the lack of kinesin-1-mediated transport, as KIF5B knockout cells displayed no apparent growth or proliferation defects, and the two other kinesin-1 isoforms, KIF5A and KIF5C, do not seem to be expressed in HeLa cells (Nagaraj et al., 2011). Although MAP7 was shown to be phosphorylated and thus inactivated during mitosis (McHedlishvili et al., 2018), it is possible that MAP7 proteins still contribute to cell division, as ensconsin is known to participate in spindle formation in flies (Gallaud et al., 2014), and MAP7D3 was reported to modulate the recruitment of kinesin-13 to the mitotic spindle (Kwon et al., 2016). In order to remove all three MAP7 homologues simultaneously, we performed siRNAmediated knockdown of MAP7D1 and MAP7D3 in the stable MAP7 knockout line, and this approach resulted in an efficient loss of all three MAP7 family members (Fig. 1F). Depletion of all three MAP7 homologs mimicked the effect of KIF5B knockout, leading to a strong perinuclear accumulation of mitochondria (Fig. 1I, J). This defect was rescued by re-expressing in these cells the individual full length MAP7 proteins (Fig. 1J). Furthermore, mitochondrial distribution was partially rescued by expressing the C-termini of MAP7 and MAP7D1 (Fig. 1J).
Rescue with the MAP7D3 C-terminus was less efficient, because the construct was mostly accumulated in the nucleus, and, as its concentration in the cytoplasm was low, only highly expressing cells showed rescue (Fig. S1A). Importantly, we observed no enrichment of MAP7 Ctermini along MTs (Fig. S1B, mCherry channel). In contrast, the N-terminal domains of all three MAP7 proteins nicely labeled the MT network but completely failed to rescue the distribution of mitochondria in cells lacking all endogenous MAP7s ( Fig. 1J and S1B). We conclude that MAP7 family members act redundantly in KIF5B-dependent mitochondria localization, and that their Ctermini, which display no or only low affinity for MTs, are sufficient to support this function.

KIF5B-560 binding to MTs in cells depends on MAP7 proteins
To show that the loss of MAP7 proteins has a direct effect on kinesin-1 activity, we next examined the distribution of the dimeric KIF5B-560 truncation mutant that can move along MTs but does not bind to cargo. In control HeLa cells, this construct was distributed along MTs, and in most cells, it showed enhanced accumulation on MTs in cell corners, where MT plus ends are concentrated ( Fig. 2A). Depletion of individual MAP7 family members did not alter this distribution except for the knockout of MAP7D3, in which less KIF5B-560 accumulated at corner MTs ( Fig. 2A, B). In contrast, in cells lacking all three MAP7 proteins, KIF5B-560 showed a diffuse localization (Fig. 2C). Expression of MAP7, MAP7D1 or MAP7D2 in such cells rescued the recruitment of the kinesin to MTs, whereas expression of MAP7D3 led to strong co-accumulation of both constructs in the corners of all transfected cells . Expression of the N-terminal, MT-binding fragments of MAP7 and its homologs could not restore the distribution of KIF5B-560, while significant rescue of MT binding by the kinesin was observed with the C-termini of MAP7 and MAP7D1 ( Fig. 2E and S1B). We conclude that KIF5B-560 has low affinity for cellular MTs in the absence of MAP7, and that this affinity can be increased by diffusely localized kinesin-1-binding MAP7 fragments. The C-terminus of MAP7D3, which was mostly nuclear on its own (Fig. S1A), was retained in the cytoplasm when expressed together with KIF5B-560, and shifted the localization of this kinesin fragment to MTs in cell corners, similar to the full-length MAP7D3 (Fig. 2D, E and S1B).

MAP7D3 but not MAP7 can be redistributed by kinesin-1
To understand why MAP7D3 but not the other MAP7 homologues promotes MT plus-end shifted distribution of KIF5B-560, we next examined the distribution of the endogenous MAP7 and MAP7D3 and found that only MAP7D3 could be efficiently relocalized by KIF5B-560 to cell corners (Fig. 3A, B).
To prove that kinesin-1 can indeed rapidly relocalize MAP7D3, we have set up an optogeneticsbased assay, in which KIF5B-560 could be sequestered in the nucleus and then acutely released from it using a blue light-inducible nuclear export system (Niopek et al., 2016). A KIF5B-560-mCherry, containing NLS sequences, was C-terminally tagged with an engineered domain of Avena sativa phototropin-1, AsLOV2, in which the Jα helix was modified to contain a nuclear export signal. Within 1 to 2 minutes after activation with blue light, KIF5B-560 was efficiently exported from the nucleus (Fig. 3D, E, G, H). MAP7D3, but not MAP7 co-accumulated on MTs in cell corners within a 4-minute timeframe ( Fig. 3D-I, Video 1 and 2). We conclude that KIF5B-560 can indeed acutely relocalize its own positive regulator MAP7D3, but not MAP7, when the kinesin expression is sufficiently high.
To explain why the distribution of MAP7D3 but not that of MAP7 was sensitive to the presence of kinesin-1, we hypothesized that MAP7D3 might be more mobile on MTs. To test this idea, we performed Fluorescence Recovery after Photobleaching (FRAP) experiments with GFP-tagged MAP7 and MAP7D3 and found that the latter indeed exchanged much more rapidly on MTs (Fig. 3J,K). The different turnover rates of the two MAP7 family proteins on MTs, possibly combined with the different affinities to kinesin-1, seem to contribute to their differential relocalization by overexpressed kinesin-1.

MAP7 and MAP7D3 control kinesin-1 recruitment to MTs and motor processivity
To get further insight into the similarities and differences in the regulation of kinesin-1 by MAP7 proteins, we set up in vitro reconstitution assays. In contrast to previously published experiments, which employed taxol-stabilized MTs in the absence of free tubulin, we used dynamic MTs that were grown from GMPCPP-stabilized seeds (Bieling et al., 2007). Kinesins, MAPs and MTs were observed by Total Internal Reflection Fluorescence Microscopy (TIRFM), as described previously (van Riel et al., 2017). Using purified MAP7 and MAP7D3 labeled with Alexa 647 coupled to a SNAP-tag (Fig. S2A), we found that MAP7 showed very long static binding events, many of which exceeded our observation time (5 min) (Fig. 4A), in agreement with recently published data . In contrast, MAP7D3 displayed a diffusive behavior, with many short binding events (Fig. 4A). These data are in agreement with the FRAP data, showing that in cells, MAP7D3 is more mobile than MAP7 (Fig. 3J, K). The density of MT labeling was higher with MAP7 than with MAP7D3 at the same protein concentration, indicating that the latter has a lower affinity for MTs (Fig. 4A, B and S3A).
Analyses by mass spectrometry and Western blotting showed that although some co-purification of MAP7, MAP7D1 and MAP7D3 with this kinesin was observed when the protein was washed with a low ionic strength buffer, this contamination was removed when the ionic strength of the washing buffer was increased (Fig. S2B, C). We used such a "high-salt washed" KIF5B-560 preparation for all our experiments. When KIF5B-560 was added to MAP7 or MAP7D3decorated MTs, we observed a strong (up to 23.6 fold) increase in the motor landing frequency compared to the situation with KIF5B-560 alone (Fig. 4C and D), in agreement with published data on MAP7 . The landing frequency of KIF5B-560 increased with higher MAP concentrations and correlated with increasing MT labeling intensity by the particular MAP (Fig. 4B,C and S3A). Furthermore, we found that MAP7D3 but not MAP7 caused a very significant decrease in kinesin velocity (Fig. 4H and S3B). Finally, we found that both MAP7 and MAP7D3 could induce a 2 fold increase in kinesin processivity ( Fig. 4D-G), with some kinesin runs exceeding 10 µm in length. We note that for this quantification, we only took into account the runs, in which we observed both kinesin association and dissociation from the MT. Inclusion of all detected runs suggested that in the presence of MAP7 or MAP7D3, even longer runs could occur (not shown).
The increase of run lengths in the presence of MAP7 and MAP7D3 could be explained by kinesin multimerization or by a model where MAP7 acts as an additional MT attachment point. In these cases the distribution of run lengths is expected to be described by the sum of two or three exponential decays (Klumpp and Lipowsky, 2005). However, the corresponding best fit of distributions at Fig. 4F and G converged to a single exponential decay, suggesting that MAP7 directly affects kinesin's binding/unbinding rate constants, rather than introducing an additional intermediate binding state. Moreover, single molecule analysis of KIF5B-560 moving on MTs showed that the kinesin intensity profiles matched that of a single dimer in assays both with and without MAP7D3 (Fig. S3C). In addition, we performed mixed kinesin assays where GFP-and SNAP(Alexa647)-tagged kinesins were used in a 1:1 ratio. If KIF5B-560 would multimerize in the presence of MAP7 proteins, then one would expect to see a significant fraction of twocolored kinesin tracks per kymograph; however, such events were not observed (Fig. S3D), confirming our observation of KIF5B-560 behaving as a single dimer on MAP7-decorated MTs.
Taken together, these data suggest that the presence of MAP7 alters the state of single kinesin dimers.
Interestingly, MAP7 could only promote kinesin processivity at high concentrations, when MTs were fully decorated, whereas MAP7D3 reduced kinesin detachment from MTs even at low concentrations ( Fig. 4E-G). These data correlated with the observation that MAP7D3, but not MAP7, could move together with KIF5B-560 in vitro (Fig. 4I, J). We conclude that MAP7 proteins can affect not only kinesin landing on MTs, as suggested previously Sung et al., 2008), but also processive kinesin movement along a MT. Co-transport of the MAP with the kinesin can facilitate processive motion, but is not essential, as also a statically bound MAP can exert this effect if its density on MTs is high enough.

MAP7D3 C-terminus promotes MT recruitment and processivity of kinesin-1 in spite of having only a low MT affinity
Since our results in cells indicated that the C-terminal parts of MAP7 proteins could rescue mitochondria distribution and KIF5B-560 binding to MTs in cells lacking full length MAP7 proteins, we set out to compare the effect of MAP7D3 and its C-terminus (Ct, see Fig. 1A) on kinesin-1 motility in vitro (Fig. 5A). In agreement with a previous publication (Yadav et al., 2014), MAP7D3-Ct displayed a weak MT binding ( Fig. 5B and S4A): MT labeling intensity with 20 nM MAP7D3-Ct was 19.2 fold lower than with 20 nM full length MAP7D3 ( Fig. 5B and  S4A). In spite of this lower MT affinity, MAP7D3-Ct could efficiently increase KIF5B-560 landing rate, decrease its velocity and promote motor processivity . The effect of MAP7D3-Ct on the landing rate was particularly obvious at 75 nM concentration, as MT labeling at this concentration was still 8.3 fold lower than with 20 nM full length MAP7D3, whereas the KIF5B-560 landing frequency was 8.1 fold higher compared to control (kinesin only) ( Fig. 5B and C). These data argue against the simple model that MAP7D3 acts as MTrecruiting factor for kinesin-1, but cannot exclude that the weak binding of MAP7D3 C-terminus to MTs augments KIF5B-560-MT interaction. These data also show that the addition of a weak MAP module to the kinesin coil is sufficient to make kinesin-1 processive.

MAP7 C-terminus promotes MT recruitment independently of MT binding
To investigate whether MAP7 family proteins can exert an effect on kinesin-1 that is independent of MT tethering, we sought to find a C-terminal fragment of the MAP7 protein that still interacts with kinesin-1 but does not bind to MTs. In a previous publication (Yadav et al., 2014), it has already been argued that the high number of positively charged amino acids in the MAP7D3 Cterminus contributes to MT binding, while MAP7 mostly lacks these residues. Using a MT pelleting assay, we indeed found that the MAP7 C-terminus, used before in cellular experiments ( Fig. 1J, 2E and S1B) does not co-sediment with MTs ( Fig. 6A and S4C). We were also unable to detect the binding of this C-terminal fragment of MAP7 tagged with mCherry to MTs in vitro by fluorescence (Fig. S4D). Nanomolar concentrations of MAP7-Ct had no effect on KIF5B-560 behavior, as described (Metivier et al., 2018). To increase the concentration of the kinesin-binding region of MAP7 in the assay, we generated its shorter version (MAP7-Ct(mini), Fig. 1A), which could be prepared from E.coli at high concentration in an untagged form (Fig. S2A). When added at micromolar concentrations to the assay with KIF5B-560, this protein fragment caused a significant (3.4 fold) increase in the motor landing frequency (Fig. 6B, C), whereas the velocity of the kinesin was only mildly affected (Fig. 6D, E). Strikingly, the increase in motor processivity observed with the MAP7D3 C-terminus was not detected with MAP7 C-terminus (Fig. 6F). It is possible that MAP7 C-terminus dissociates from the kinesin very soon after the motor lands on the MT, and since MAP7-Ct does not concentrate on MTs, the chance of its re-association with the kinesin during a run is low. Altogether, MT landing of kinesin-1 can be increased by MAP7 family proteins independent of their MT interaction, whereas the regulation of kinesin processivity by these MAPs depends on their association with MTs.

The stalk of KIF5B-560 inhibits MT interaction
Our finding that MAP7 C-terminus improves the MT landing frequency of KIF5B-560 could potentially be explained if the MAP7-interacting stalk of kinesin-1 partly interferes with MT binding. If this were true, a kinesin-1 truncation lacking this stalk should bind to MTs more efficiently. To test this idea, we generated a shorter KIF5B truncation mutant, KIF5B-370, which lacks the MAP7-binding coil region but still dimerizes via its neck linker ( Fig. 1A and 7A). The concentrations of KIF5B-370 and KIF5B-560 were carefully controlled on a Coomassie blue stained gel (Fig. 7B). In in vitro assays, KIF5B-370 appeared to be a faster kinesin with slightly shorter runs compared to F). Importantly, we observed a 7.8 fold difference in motor landing frequency (Fig. 7C,D), indicating that the presence of the stalk region in KIF5B-560 has a negative effect on its interaction with MTs.
In this study, we have systematically analyzed the impact of mammalian MAP7 family proteins on kinesin-1 transport. We found that having at least one MAP7 homolog was necessary and sufficient to enable kinesin-1-driven distribution of mitochondria in the cytoplasm. These results are fully in agreement with the data showing that MAP7/ensconsin is an essential kinesin-1 cofactor in flies and in mammalian muscle cells (Barlan et al., 2013;Metivier et al., 2018;Metzger et al., 2012;Monroy et al., 2018;Sung et al., 2008). Dependence on MAP7 family members likely applies to many other kinesin-1-dependent processes in mammals, because the core part of kinesin-1, the KIF5B-560 fragment, was not able to bind to MTs efficiently in cells when all MAP7 homologues were absent.
Kinesin-1 function could be rescued to a significant extent by a MAP7 fragment that binds to kinesin, but not to MTs, again in agreement with the data obtained in Drosophila (Barlan et al., 2013;Metivier et al., 2018). KIF5B-560 recruitment to MTs could also be restored by this MAP7 fragment, and this effect could be recapitulated in vitro with purified components. Interestingly, in vitro, a short kinesin-1 version (KIF5B-370), which lacked the MAP7-binding stalk region altogether, interacted with MTs more efficiently than KIF5B-560. These data suggest that the stalk might partially inhibit MT binding, and that this effect could be relieved by the interaction with MAP7. It is possible that the stalk-containing kinesin can adopt conformations that are unfavorable for MT binding, whereas the interaction with MAP7 allosterically stabilizes a conformation that promotes MT engagement (Fig. 7E,F). Diverse regulatory steps have been described for kinesin-1, mostly involving the autoinhibitory C-terminal tail region (reviewed in (Verhey and Hammond, 2009) and also the motor domain (Xu et al., 2012)). Our data add to this complexity by showing that the state or the position of the stalk and its binding partners might directly affect the interactions between the motor domains of kinesin-1 and MTs.
Although a MAP7 fragment lacking MT affinity can make kinesin-1 more active, the presence of the MT binding domain makes this regulation much more efficient and robust, as without it, a very high concentration of MAP7 kinesin-binding domain is required to activate the motor. The presence of a MT-binding site, even a weak one (such as the one in the MAP7D3-Ct), concentrates the kinesin-binding domain of the MAP on MTs and can thus facilitate its interaction with the kinesin (Fig. 7F). Furthermore, MT-bound MAP7 proteins have a significant effect on kinesin processivity, and we excluded the possibility that this was due to kinesin multimerization. We note that in our in vitro assays, kinesin run length is relatively high (close to 2 µm on average) even in the absence of cofactors. One possible reason for this is the fact that we work with dynamic MTs in the presence of soluble tubulin, as opposed to standard assays with taxol-stabilized MTs in the absence of tubulin. We thus observe kinesin behavior on freshly formed MT lattices, which might have fewer defects, and even if damaged, can possibly be repaired by tubulin incorporation (Schaedel et al., 2015).
We found that both a very immobile MAP (MAP7) and a more dynamic and diffusively behaving MAP (MAP7D3) could enhance kinesin-1 processivity, but the latter was able to exert this effect when present on a MT at a lower density. Some affinity of the kinesin-bound MAP fragment to MTs was required to increase motor processivity, and this could suggest that, unlike the landing rate, kinesin processivity is governed not just by the motor conformation but by the presence of additional links to MTs. However, the analysis of the run lengths of KIF5B-560 in the presence of MAP7 proteins showed that their distribution was monoexponential, whereas a significant contribution of an additional MAP-dependent MT bound state would be expected to lead to a distribution corresponding to the sum of two or three exponential decays (Klumpp and Lipowsky, 2005). We thus favor the idea that the interaction with MAP7 alters the kinesin conformation, making it more favorable for MT binding, and this is reflected in both higher landing rates and longer run lengths (Fig. 7E). Since the interaction of MAP7 with the kinesin is transient, its MTunattached version (MAP7-Ct) would dissociate very rapidly after landing, and this fragment thus cannot enhance kinesin processivity. In contrast, MT-tethered full length MAP7 proteins or the MAP7D3 C-terminus have a much higher local concentration, which promotes repeated interactions with the kinesin and its maintenance in a state favorable for MT binding. In this way, the rapid binding-unbinding kinetics, which could be described as kinesin "hopping" from one stationary MAP molecule to another, can allow an immobile MAP7 to counteract kinesin dissociation without strongly affecting motor velocity.
MAP7D3 is different from MAP7 because it binds to MTs less tightly and can be "dragged" with the motor to some extent, which possibly explains why it slows down kinesin movement and why its low concentration, which had only a mild effect on the kinesin landing frequency, was sufficient to increase motor processivity ( Fig. 4 and 7E). However, because of the fast turnover within the MAP-motor complex, MAP7D3 is unlikely to be undergoing large-distance transport by the kinesin. MAP7D3 can be relocalized to MT plus ends by KIF5B-560 quite rapidly, but only when the motor is overexpressed. At the endogenous kinesin-1 expression levels, MAP7D3 is not enriched at MT plus ends, suggesting that the levels of endogenous motor are insufficient to drive MAP7D3 to the cell periphery. Still, it is possible that also the endogenous kinesin-1 causes some redistribution of MAP7D3, thus contributing to the localization of its own positive regulator. It has been suggested that MAP7 and MAP7D1 can undergo MT plus end-directed displacement in migrating cells (Kikuchi et al., 2018), and therefore, the dependence of the localization of these MAPs on the presence of the kinesin deserves further investigation.
An important question is whether the distribution of MAP7 proteins contributes to the welldocumented selectivity of kinesin-1 for specific MT tracks, which appear to correspond to the stable, long-lived MT population (Cai et al., 2009;Farias et al., 2015;Guardia et al., 2016;Hammond et al., 2008;Jacobson et al., 2006;Nakata and Hirokawa, 2003;Tas et al., 2017). The idea that MAP7 proteins can spatially control kinesin-1 activity is supported by our observation that overexpressed KIF5B-560 accumulates in cell corners only in the presence of MAP7D3, which can be shifted by the kinesin towards MT plus ends. In contrast, MAP7, which is stably associated with MTs, does not support this peripheral localization of KIF5B-560. A MAP with a slow turnover, such as MAP7, could in principle predispose kinesin-1 for interacting with more long-lived MTs, on which this MAP would gradually accumulate, a possibility that would be interesting to test.
Finally, it would be interesting to know whether MAP7 proteins exert similar regulatory effects on any other kinesins, and thus whether the mechanism described here is shared by other motors.
Taken together, our data illustrate the complexity of the interplay between the motors and the tracks they use during intracellular transport processes.

Materials and Methods
Cell Culture, knockdowns and CRISPR/Cas9 knockouts HeLa (Kyoto), Cos7 and human embryonic kidney 239T (HEK293T) cell lines were cultured in medium that consisted of 45% DMEM, 45% Ham's F10, and 10% fetal calf serum supplemented with penicillin and streptomycin. The cell lines were routinely checked for mycoplasma contamination using LT07-518 Mycoalert assay (Lonza). HeLa and COS7 cells were transfected with plasmids using MAP7D3 knockout was performed according to the protocol described in (Ran et al., 2013

Immunostainings, Western blotting and antibodies
For immunofluorescence cell staining, HeLa cells were fixed in -20°C methanol for 10 min and stained for MAP7, MAP7D1, MAP7D2, MAP7D3 and α-tubulin. In the case of cytochrome c, cells were fixed with 4% PFA in phosphate-buffered saline (PBS) for 10 min. Cells were then permeabilized with 0.15% Triton X-100 in PBS for 2 min; subsequent wash steps were performed in PBS supplemented with 0.05% Tween-20. Epitope blocking and antibody labeling steps were performed in PBS supplemented with 0.05% Tween-20 and 1% BSA. Before mounting in Vectashield mounting medium (Vector Laboratories) slides were washed with 70% and 100% ethanol and air-dried. Total HeLa cell extracts were prepared in RIPA buffer containing 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 0,5% SDS and cOmplete protease inhibitor cocktail (Roche).

Pull down assays
Streptavidin pull down assays were performed from HEK293T cell lysates by coexpressing biotin ligase BirA with mCherry-tagged constructs containing a biotinylation site (BirA substrate sequence: MASGLNDIFEAQKIEWHEGGG) (bait), and a GFP-tagged KIF5B construct (prey).
Constructs were transfected altogether into HEK293 cells using PEI with 24 hrs incubation time for proper protein expression. M-280 Streptavidin Dynabeads (Invitrogen) were blocked in a buffer containing 20 mM Tris pH 7.5, 20% glycerol, 150 mM NaCl, and 10 µg Chicken Egg Albumin followed by three washes with wash buffer containing 20 mM Tris pH 7.5, 150 mM NaCl and 0.1% Triton-X. HEK293T cells were scraped and collected in ice-cold PBS followed by lysis on ice in a buffer containing 20 mM Tris pH 7.5, 150 mM NaCl, 1mM MgCl 2 , 1% Triton X-100, and cOmplete protease inhibitor cocktail (Roche). To separate cell debris, the lysates were pelleted 4 o C for 15 min at 16000 g and 10 % of each lysate was saved as input control. Cell lysates were incubated with pre-blocked streptavidin beads for 60 min at 4 o C followed by five washes with wash buffer containing 20 mM Tris pH 7.5, 150 mM NaCl and 0.1% Triton-X.
Streptavidin beads were pelleted and boiled in 2x Laemmli sample buffer. Protein lysates and pull down of both bait and prey proteins were analyzed by Western blot.

Protein Purification
All KIF5B Eluate was then digested by Ulp1 overnight at 4°C while dialyzed against 50mM phosphate buffer pH 8.0 with a 6 kDa cut-off membrane (Spectrum Laboratories). Protein was loaded on a POROS® 20HS (Thermo Fischer Scientific) column in the same dialysis buffer, using ÄKTA® purifier (GE Healthcare) for cation exchange chromatography. Protein was eluted by a 0-100% were then concentrated and exchanged against 25 mM HEPES buffer pH 7.5 with 75 mM KCl, 75 mM NaCl and 10 mM DTT using a Vivaspin column (cut-off: 6 kDa). Concentration was confirmed with an ND-100 spectrophotometer (Nanodrop Technologies). Purity was confirmed by SDS-PAGE and protein was aliquoted and stored at -80°C.

Mass spectrometry
After streptavidin purification, beads were resuspended in 20 µl of Laemmli sample buffer (Biorad) and supernatants were loaded on a 4-12% gradient Criterion XT Bis-Tris precast gel Matrix Science, UK) against a Swiss-Prot database (taxonomy human). Carbamidomethylation of cysteines was set as a fixed modification and oxidation of methionine was set as a variable modification. Trypsin was specified as enzyme and up to two miss cleavages were allowed. Data filtering was performed using percolator, resulting in 1% false discovery rate (FDR). Additional filters were; search engine rank 1 peptides and ion score >20.

In vitro Reconstitution Assays
MT seeds were prepared by incubating 20 μM porcine tubulin mix containing 70% unlabeled,

Image acquisition and processing
Fixed cells were imaged with a Nikon Eclipse 80i upright fluorescence microscope equipped with For quantifying mitochondria in different HeLa cell knockdown/knockout conditions we classified mitochondria as "clustered" when ~80% of the cytochrome c signal was localized in a dense cluster around the nucleus, all other localization patterns with more spread mitochondria were classified as "spread". K560-GFP was classified as localized on corner MTs when clear enhancement of fluorescent signal was seen at peripheral MTs near the cell cortex over MTs that are localized in between the cortex and the cell center.
FRAP measurements were performed by bleaching a 10 x 10 μm square region in a cytoplasmic region between the nucleus and cell cortex followed by 8.5 min imaging with a frame interval of 3 sec. Mean fluorescence intensities were measured from a 4 x 4 μm square region within the original photobleached region to avoid analyzing non-bleached MTs that could slide into the analyzed region. The mean intensity of this region was double corrected for background fluorescence and photobleaching (Phair et al., 2004).
Optogenetic experiments with blue light-inducible K560-LEXY kinesin were performed using spinning disk microscopy. Acquisitions were done with a frame interval of 5 sec after sequential exposure with green light 561 nm laser (to image kinesin) followed by blue light 491 nm laser (to image MAPs and activate K560-LEXY simultaneously). Exposure times of ~ 1 sec per interval with the 491 nm laser were sufficient to actively export optogenetic motors from the nucleus. For measuring fluorescence intensity changes at cell corners, a maximum intensity projection over time of the K560-LEXY channel was made using ImageJ, followed by Gaussian blurring and thresholding to select cell corners to analyze. Mean fluorescence values for GFP-MAP7/MAP7D3 and K560-LEXY were obtained from the same cell corners over time, background subtracted and normalized to the mean fluorescence in that region at T = 0 min.
Changes in mean fluorescence intensity were plotted per cell corner.

Statistical analysis
Statistical significance was analyzed either using the Mann-Whitney U test or Student's t test, as indicated in figure legends. For the t test, data distribution was assumed to be normal, but this was not formally tested. Statistical significance was determined using GraphPad Prism software (version 7.04). Fitting of run lengths with sum of two or three exponential decays was performed on the raw data using maximum likelihood estimation method implemented in mle function of MATLAB R2011b (The MathWorks, Natick, 2011).