MKLP1 requires specific domains for its dendritic targeting

Typical vertebrate neurons extend a single axon and several dendrites. It has been hypothesized that the unique morphological and compositional features of axon and dendrite result from their distinct patterns of microtubule polarity orientation. The microtubules (MT) within axons are uniformly oriented with their plus-ends distal to the cell body, whereas microtubules within dendrites are non-uniformly oriented (Baas and Ahmad, 1993). A wealth of information suggests that the mitotic motor protein MKLP1 is responsible for the non-uniform distribution of microtubules in the dendrite. MKLP1 transports microtubules from the cell body into the developing dendrite with their minus-ends leading, thereby establishing the pattern of non-uniform microtubule polarity in the dendrite (Sharp et al., 1996; Sharp et al., 1997).

Using antibodies against MKLP1, immunofluorescence analysis indicated that the molecule is absent from axons but is enriched in developing dendrites. MKLP1 also exists in fully developed neurons, and depletion of MKLP1 from cultured neurons causes a rapid redistribution of microtubules within dendrites, such that minus-end-distal microtubules are chased back to the cell body, and the dendrites acquire some characteristics of an axon (Yu et al., 2000). These facts suggested that MKLP1 is necessary to maintain the minus-end-distal microtubules in the dendrite of developing and fully developed neuron. Using in-situ hybridization, we found that MKLP1 mRNA is restricted to the cell body in developed hippocampal neurons (our unpublished data); also, gene-chip experiments did not indicate the existence of MKLP1 mRNA in the dendrites of neurons (Eberwine et al., 2002). It seems that the targeting of MKLP1 protein to dendrites is responsible for the maintenance of the non-uniform microtubule polarity in the dendrite.

For membrane proteins, some intrinsic sorting signals, i.e. amino acid (aa) motifs within the molecule, can direct the protein to specific locations in the plasma membrane (Rivera et al., 2003; Gu et al., 2003). However, recently an actin-binding domain of a motor protein (a class III myosin) has been suggested to be important for its localization to filopodia tips in transfected HeLa cells (Erickson et al., 2003). Until now, it is not clear whether some domains exist in MKLP1, that are responsible for the dendritic targeting of this motor protein. MKLP1 has two splice variants: MKLP1 856 and MKLP1 960 [comprising amino acids (aa) 856 and 960, respectively]; the latter has one actin-binding domain more (aa 691-794) than MKLP1 856 (Kuriyama et al., 2002). In mitotic cells, both splice variants possess the motor activity and microtubule-bundling capacity that are required for completion of cytokinesis by organizing midzone microtubules and the electron-dense matrix in the center of the intercellular bridge. This indicates that both of them probably have the ability of transporting minus-end microtubules to dendrites along plus-end microtubules (Matuliene and Kuriyama, 2002).

In a National Center for Biotechnology Information (NCBI) conserved-domain search, putative conserved domains have been found in MKLP1 (Neuwald et al., 1997). The N-terminal part of MKLP1 856 contains conserved kinesin-motor domains, and the central region between amino acids 485-655 may be assembled into a coiled-coil `stalk' of MKLP1 856, whereas amino acids 656-856 are the tail of MKLP1 856. In addition, the MKLP1 have a consensus ATP-binding site (residues 113-117), an SSRSH domain (residues 296-300), and a LAGSE domain (residues 334-340) (Nislow et al., 1992). In addition, an Arf-protein-binding domain (Boman et al., 1999) has been identified in MKLP1 960, overlapping with its actin-binding domain. The present study was undertaken to analyze the molecular domain requirements for the targeting of MKLP1 to dendrites.

To examine the region(s) of MKLP1 necessary for targeting the molecule to dendrites, a series of C- and N-terminal truncation mutants were produced (Fig. 1). Each construct included an enhanced green fluorescent protein (eGFP) tag that can be detected by fluorescence microscopy (Kain et al., 1995; Matuliene and Kuriyama, 2002; Chen et al., 2002). To examine whether the eGFP tag was cleaved during its transfection into cells, HEK293A cells were transfected with each of the constructs individually, and cell lysates were subjected to western blot analysis by using an anti-eGFP antibody. In each case, eGFP immunoreactive bands were detected at the predicted molecular mass for the MKLP1 construct to which they were attached (Fig. 2). The results showed that the tag remained attached to the MKLP1 protein in all cases and was not cleaved intracellularly.

In HEK293A cells, eGFP-tagged MKLP1 was localized at the center of spindle poles during anaphase and at midbodies at the end of mitosis (Fig. 3), as it has been described elsewhere (Matuliene and Kuriyama, 2002). This indicates that the eGFP tag did not alter the function of MKLP1 in the formation of midbody matrix and the completion of cytokinesis in mammalian cells.

Each MKLP1-truncation construct (Fig. 1) was transfected into hippocampal neurons at 7 days in vitro (7 DIV) and visualized by fluorescence microscopy after 16 hours of culture (Figs 4, 5, 6, 7, 8). Dendrites were identified by the presence of MAP2 with a monoclonal MAP2 antibody.

We tested hippocampal neurons at 7 DIV that expressed eGFP-tagged MKLP1 or eGFP alone. In cells that had been transfected with eGFP alone, eGFP was detected in long, thin, MAP2-negative axons as well as in dendrites. However, in cells transfected with eGFP-tagged MKLP1 eGFP was restricted to MAP2-positive soma-dendrites (Fig. 4). The distribution of eGFP-tagged MKLP1 was very similar to that described elsewhere (Yu et al., 2000). Therefore, it seems that eGFP-tagging did not change the dendritic distribution of MKLP1 in 7DIV hippocampal neurons.

It has been reported that the abilities of MKLP1 in regard to MT bundling and MT transport are served by the N-terminal half of the molecule (Nislow et al., 1992). When expressed in Sf9 insect cells, it can induce the formation of dendrite-like processes, although this did not occur when full-length MKLP1 960 was expressed (Kuriyama et al., 1994; Sharp et al., 1996). It seems that the N-terminal half of MKLP1 is targeted to dendrites when MKLP1 is expressed in cultured neurons (Sharp et al., 1996). We therefore expressed the C-terminal-deletion mutants MKLP1(1-456), MKLP1(1-710), MKLP1(1-840) in 7 DIV hippocampal neurons to investigate whether deletion of the C-terminal changes the correct dendritic distribution of MKLP1.

When we expressed eGFP-tagged MKLP1 1-456 (the N-terminal half of the molecule) in 7 DIV hippocampal neurons, to our surprise, the green fluorescence was distributed into MAP2-negative axons. We analyzed the expression protein by western blotting and found no spliced eGFP protein (Fig. 2). Thereafter, we expressed constructs MKLP1(1-711) and MKLP1(1-840) and found that both were also distributed into MAP2-negative axons (Fig. 5).

Even more unexpected was the finding that distribution of the MKLP1(1-840) construct [which only has 16 aa less at its C-terminal than MKLP1(1-856)], was very different to that of MKLP1(1-856) (Fig. 5). Interestingly, in HEK293A cells, MKLP1(1-840) also had a different distribution than MKLP1(1-856), the latter binding more easily to microtubules that appeared in the pseudopod terminals of the cell (Fig. 3). However, in these cells, the distribution of MKLP1(1-710) was very similar to that of MKLP1(1-840) (data not shown).

From our preliminary data we got the impression that the C-terminal of MKLP1 was very important in dendritic targeting. We, therefore, expressed in 7 DIV hippocampal neurons MKLP1(811-856), -(711-856), -(456-856) and -(162-856), all of which contain N-terminal deletions, and looked for their respective distribution. The results showed that MKLP1(811-856) has no distinct distribution in neurons, but MKLP1(711-856) and -(461-856) are restricted to the nucleus. However, about half the neurons that expressed MKLP1(162-856) (55%, n=20 transfected neurons) showed sign of the dendritic targeting associated with MKLP1 (Fig. 6), whereas other constructs were restricted to the nucleus.

Regarding the 16 aa at the C-terminal, we found that they are very important for the non-axonal distribution of MKLPl, as evidenced by comparing the distribution of MKLP1(1-856) and MKLP1(1-840) in cultured hippocampal neurons. We also expressed MKLP1(711-840) and -(461-840), and found that they both lost their nuclear location and diffused into axons and dendrites (Fig. 7). In HEK293A cells, both constructs were diffusely distributed, which is different to MKLP1(1-840), which binds to microtubules (data not shown). We also examined the distribution of the MKLP1 `stalk' (aa 461-710), which was revealed to remain in soma-dendrites. The D456-710 also existed in soma-dendrites, although MKLP1(461-710) distributed to more distal regions of dendrites (Fig. 8). The above mentioned results are summarized in Table 1.

We can therefore draw some general conclusions. (1) Full-length MKLP1(1-856)-eGFP is distributed to soma and dendrites in hippocampal neurons, but not to the axon. The actin-binding domain (aa 691-794) of MKLP1(1-960), which is absent in the MKLP1(1-856), is not necessary for dendritic targeting in neurons. (2) In hippocampal neurons, the distribution of MKLP1 mutants coupled to eGFP showed the following characteristics. First, the N-terminal part of MKLP1, i.e. the `motor' domain of MKLP1, is necessary for dendritic distribution because deletion of the domain prevented dendritic distribution [MKLP1(461-856) and -(711-856)]. Second, the C-terminal part of MKLP1, i.e. the `tail' domain, provides the necessary complement for dendritic targeting because deletion of the `tail' caused the axonal appearance of the constructs [MKLP1(1-456) and -(1-710)]. Furthermore, the integrity of the C-terminal part of MKLP1 is necessary for restricted dendritic distribution, because deletion of 16 aa (841-856) prompts its axonal appearance [MKLP1(1-840)]. Third, the `stalk', i.e. MKLP1(461-710), was not necessary for the dendritic targeting of MKLP1, because the mutant D456-710, which lacks it, was still specifically distributed to the dendrite.

This is the first systematic study that investigated the requirements of intrinsic molecular domains on the targeting of a kinesin molecule, using eGFP-tagged truncated mutatants. Several points of interest are discussed here.

The motor domain of MKLP1 is located at the N-terminus, what then is the role of its C-terminal tail region? We knew that, when MKLP1 960 was expressed at full length, it did not induce the formation of dendrite-like processes in Sf9 cells; however, when expressing a truncated form that did not include a substantial part of the C-terminal region dendrite-like processes were formed (Kuriyama et al., 1994; Sharp et al., 1996). One might therefore conclude that the C-terminus is not necessary for the dendritic targeting of MKLP1 neurons.

Our results, however, indicate that the C-terminus of MKLP1 is very important for its correct distribution in soma-dendrites in cultured hippocampal neuron. The eGFP-tagged full-length MKLP1(1-856) was correctly targeted to dendrites, whereas removal of the C-terminal region – as in constructs MKLP1(1-840), -(1-710) or -(1-456) – resulted in the loss of their dendritic localization. It seems that the C-terminal integrity of MKLP1 is necessary for sorting MKLP1 to dendrites in cultured hippocampal neurons. We first hypothesized the existence of some dendritic targeting signals in the distal part of the C-terminus of MKLP1, when the deletion of 16 aa [MKLP1(841-856)] prompted its axonal appearance. Furthermore, constructs that lack the tail domain or the stalk-tail domain [MKLP1(711-856) and -(461-856), respectively] were restricted to the nucleus, only MKLP1(811-856) appeared also in the axon.

We do not know whether some conventional dendritic targeting signals for membrane proteins also exist in the C-terminal of MKLP1. If it do, one of these constructs [MKLP1(811-856), -(711-856) or -(461-856)] should localize to the dendrites. It has been reported that, a dileucine motif, which functions as a dendritic targeting signal, can redirect the usually axonal voltage-gated K⁺ channel proteins Kv1.3 and Kv1.4 to dendrites by anchoring them to dendritic targeting vesicles (Rivera et al., 2003). We reasoned that the dendritic targeting signals for MKLP1 are not like any known dendritic targeting signals for those membrane proteins. In the case of MKLP1, it might regulate its movement by binding to scaffold proteins, which would steer the kinesin to dendrites, or restrict it at the axon.

Based on the results shown in Fig. 3, we also think that the C-terminus might have a function in regulating MKLP1's microtubule-binding activity and motility, because MKLP1(1-840) and -(1-710) bind easier to microtubules, and accumulate at the periphery of the cell where the plus-ends of microtubules are located. Since MKLP1(811-856) is diffusely distributed when expressed in HEK293A cells (data not shown), the mechanism of axonal distribution of MKLP1(1-840) might differ from that of MKLP1(811-856). MKLP1(1-840) should direct the molecule moving along the microtubule and forward it to the axon terminal, while the MKLP1(811-856) should be just simple diffusion. When comparing of results of MKLP1(1-856) and -(1-840), or MKLP1(711-856) and -(711-840), it can be seen that the integrity of the C-terminus is very important in the dendritic and nuclear targeting of MKLP1 (our unpublished data). It would be very interesting to know whether the domains responsible for nuclear targeting also have the capacity to prohibit a distribution into axons? Although it is only conjecture as yet, the importance of the C-terminus of MKLP1 in its dendritic targeting is obvious.

Conventionally, kinesin and other members of the kinesin family bind ATP and microtubules at specific sites in their conserved motor domain, and use the energy from ATP hydrolysis to produce force and to move along the microtubules. Therefore, the motor domain should provide the basic requirements for MKLP1 targeting. In our experiments, as a strong evidence for the importance of the motor domain of MKLP1 in the correct targeting in cultured hippocampal neurons, the mutants MKLP1(461-856) and -(711-856) without the motor domain were restricted in nucleus. However, MKLP1(162-856) – which has no ATP-binding domain – still partly resumed its dendritic location; and it is not clear why a mutant without the motor domain can still keep such a distribution pattern in dendrites.

As for the `stalk' of MKLP1, we found that only MKLP1(461-710) was distributed into soma-dendrites. At first, we thought that there might be a dendritic targeting signal in MKLP1(461-710). To test this hypothesis, we expressed D456-710, which lacks aa 461-710 and found that D456-710 also existed in soma-dendrites, although only MKLP1(461-710) was distributed more in distal dendrites.

It seems logical to accept that MKLP1 can, owing to its motor and tail domain, correctly be sorted into dendrites of cultured neurons, whereas its stalk domain is does not seem to be involved in this function. But why can the stalk alone also be distributed into dendrites but not to axon? We tentatively postulate that the stalk domain conveys a homodimerizing capability of MKLP1 (Kuriyama et al., 1994). In support of this hypothesis, our yeast two-hybrid analysis showed that MKLP1(461-710) can form heterodimers with MKLP1(1-856) and -(461-840), but MKLP1(711-840) cannot form a heterodimer with MKLP1(1-856) (our unpublished data). It might be that the stalk can dimerize with endogenous MKLP1 and is transported to the dendrites, which could partly explain why MKLP1(461-840) and -(711-840) distributed to axons.

It has been reported that another kinesin (KIF5) can be steered to dendrites as a motor protein for AMPA receptors, by directly interacting with the scaffold protein GluR2 (AMPA receptor subunit)-interacting protein 1 (GRIP1) with its heavy chains (Setou et al., 2002), and that movement of Kinesin-1 can be regulated by another scaffold protein, UNC-16 (Byrd et al., 2001). It is plausible that other scaffold proteins take part in targeting MKLP1 to the dendrites of hippocampal neurons, by directly linking with the MKLP1 tail domain. We have made it clear that the MKLP1 tail is very important for its dendritic localization. Perhaps the next step is to find out which scaffold protein binds the MKLP1 tail and promotes its dendritic targeting.

Based on our data of the distribution of MKLP1(1-856) and its defined mutants, we postulate that motor and tail regions are both necessary for the correct targeting of MKLP1 to dendrites in hippocampal neurons. It is very probable that the motor provides the necessary power while the tail acts as the steering guide. Targeting of kinesin is different from targeting of membrane-anchoring proteins. Kinesin targeting should be an active process, whereas membrane anchoring proteins can only be passively transported with its vesicles.

Primary hippocampal cultures were prepared from brains of embryonic days 18-19 Sprague-Dawley rats essentially as described (Goslin et al., 1991). After trituration with a glass pipette, neurons were plated on poly-D-lysine-coated cover slips (Sigma) at a density of 1×10⁵ neurons per mm², and were grown in Neurobasal medium supplemented with 0.5 mM L-glutamine and B27 (2 ml per 100 ml). Fluorodeoxyuridine was added to inhibit proliferation of non-neuronal cells. Medium was changed every 4 days. Approximately 80% of cells showed morphology typical of hippocampal pyramidal neurons. HEK 293A cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM) plus 10% fetal calf serum (FCS).

Enhanced green fluorescence protein (eGFP)-tagged MKLP1 and its deletion mutants were prepared as described elsewhere (Deavours et al., 1999). Briefly, the full-length MKLP1 (aa 1-856) were produced by reverse transcription (RT)-PCR amplified from HeLa-cell cDNA with primers 5′-AAAGAGCTCGAAGTTCTAGTTCTTGCTG-3′ (forward) and 5′-AAAGGTACCAGTTCGTGGCTTTTTGCG-3′ (reverse). PCR products were digested with SacI and KpnI, purified, and ligated into the pEGFP-N₃ vector. Deletion fragments encoding aa 1-456 of MKLP1 were PCR-amplified with primer 5′-AAACTCGAGAAGTTCTAGTTCTTGCTGCC-3′ (forward) and 5′-TTAGGATCCAGGCGTTAAACCACATATTG-3′ (reverse) for 1-456. PCR products were digested with XhoI and BamHI, purified, and ligated into the pEGFP-N₃ vector. The MKLP1 mutants MKLP1(1-710) and -(1-840) were recombined fragments of full-length MKLP1 that were digested with (SacI, PstI) and (SacI, BamHI), and inserted into the pEGFP-N₃ vector. The MKLP1 mutant MKLP1 (811-856) was PCR-amplified with forward primer 5′-AAAGAATTCATGGCACCTGCCCAACCAGAT-3′ and reverse primer 5′-TTTGTCGACTCGTGGCTTTTTGCGCTTG-3′. PCR products were digested with EcoRI and SalI, purified, and ligated into the pEGFP-N₃ vector. The MKLP1 mutants MKLP1-(711-856), -(461-856) and -(162-856) were PCR-amplified from full-length MKLP1 with three forward primers 5′-AAACTGCAGGAATGAGATGGGTAGATCATA-3′ 5′-AAACTCGAGGTATGGTTGGAAATGAACC-3′ and 5′-AAAGAGCTCATATGCAGTGTGAGGTTG-3′, and reverse primer 5′-AAAGGTACCAGTTCGTGGCTTTTTGCG-3′. PCR products were digested (with PstI-KpnI, (XhoI-KpnI and SacI-Kpn I, respectively), purified and ligated into the pEGFP-N₃ vector. Construct D456-710 (full-length MKLP1, which lacks aa 461-710) was PCR-amplified with forward primer 5′-AAACTCGAGGTGCTGCAGCTGCTGCAGCTGCTGCAGCTTCTGCAGAAA-3′ and reverse primer 5′-TTTCTGCAGCAGCTGCAGCAGCTGCAGCAGCTGCAGCTAAAGCTTTTT-3′. PCR products were digested with XhoI-PstI and purified. MKLP1(1-856) was digested with XhoI and PstI, DNA fragments of 1.5kb and 5kb were purified and ligated with purified PCR products. MKLP1(461-710) was PCR-amplified from full-length MKLP1 with forward primer 5′-AAACTCGAGGTATGGTTGGAAATGAACC-3′ and reverse primer 5′-AAAGGTACCAGTTCGTGGCTTTTTGCG-3′. PCR products were digested with XhoI-PstI, purified and ligated into the pEGFP-N₃ vector. MKLP1(711-840) was PCR-amplified with full-length MKLP1 with forward primer 5′-AAACTGCAGGAATGAGATGGGTAGATCATA-3′ and reverse primer 5′-AAAGGTACCAGTTCGTGGCTTTTTGCG-3′. PCR products were digested with PstI-BamHI, purified and ligated into the pEGFP-N₃ vector. MKLP1(461-840) was produced from MKLP1(461-856) digested with XhoI-BamHI and ligated into pEGFP-N3.

At 7 days in vitro (7 DIV), hippocampal neurons were transfected with eGFP-tagged MKLP1 or other MKLP1-deletion mutants using the calcium phosphate method. DNA (1.5 μg per coverslip) was mixed with 2 M CaCl₂ and added to an equal volume of 2×HEPES-buffered saline [NaCl 274 mM, KCl 10 mM, Na₂HPO₄.7H₂O 1.4 mM, D-glucose 15 mM, HEPES (free acid) 42 mM pH 7.1]. The DNA mixture was incubated for 10 minutes at room temperature and added drop-wise (80 μl/ml) to the neurons at 37°C in 5% CO₂ until a fine precipitate formed (after about 25 minutes). Neurons were washed twice with DMEM and returned to the original conditioned media for the duration of the experiment. Approximately 1% of neurons were transfected with this technique. Cells were fixed for imaging 4 days after transfection, at 10 DIV. HEK293A cells were transfected with eGFP-fusion constructs with the same calcium phosphate method as used for neurons, except that the time of exposure to the precipitate was 5 hours. The plates were scraped 24 hours after transfection and prepared for western blotting.

For immunocytochemistry, neurons were fixed in culture with 4% paraformaldehyde in PBS for 25 minutes and were permeabilized in 0.3% Triton X-100 for 5 minutes. After a 2-hour incubation with 10% bovine serum albumin (BSA) to block nonspecific staining, the cultures were incubated for 1 hour with primary antibody containing 2% BSA. Mouse anti-MAP2a,b monoclonal antibody (mAb) (Neomarker) was used to label dendrites of neurons. Primary antibody was used at a 1:200 dilution. After washing off the primary antibody, neurons were incubated with rhodamine-tagged goat anti-mouse secondary antibodies (Sigma) at a concentration of 1:500. Transfected HEK293A cells were fixed with 4% paraformaldehyde in PBS for 25 minutes, washed three times with PBS, and incubated with Hoechst 33258 dye (100ng/ml) for 15 minutes at 37°C.

Thirty μl aliquots of HEK293A cell homogenates were separated by 15% SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked for 1 hour in TBS (50 mM Tris-HCl, 150 mM NaCl pH 7.4) containing 5% (w/v) nonfat dried milk and 0.1% Tween-20, and then incubated with rabbit polyclonal Ab against GFP (sc-8334, Santa Cruz). After washing with TBS Tween-20, membranes were incubated with goat anti-rabbit horseradish peroxidase-conjugated secondary Abs (1:2500 dilution, Pierce, Rockford, IL). Protein signals were visualized using enhanced chemiluminescence (ECL) development (Pharmacia, Piscataway, NJ).

Detection of fluorescent proteins was done with a fluorescent Microscope system: Olympus IX70, cooled CCD camera (Olympus), Lambda 10-2 (Shutter Instrument), controlled by software (Metarmorph imaging system). Labeled neurons were imaged using a 20×/0.40 Ph1 objective. Identical acquisition parameters were used for each construct. Images were processed and labeled using Adobe Photoshop software.

This work was supported by grants from The National Natural Science Foundation of China (Number 30270451, 30500151).