Nonuniform Microtubular Polarity Established by CHO1/MKLP1 Motor Protein Is Necessary for Process Formation of Podocytes

The generation and initial characterization of conditionally immortalized mouse podocyte clones (hereafter referred to as “podocytes”) has previously been described (Mundel et al., 1997b). Podocytes were maintained in RPMI 1640 medium (Life Technologies, Eggenstein, Germany) supplemented with 10% FCS (Boehringer Mannheim, Mannheim, Germany), 100 U/ml penicillin, and 100 μg/ml streptomycin (Life Technologies) in a humid atmosphere with 5% CO₂. To propagate podocytes, the culture medium was supplemented with 10 U/ml recombinant mouse γ-interferon (Sigma, Munich, Germany) to enhance expression of T-antigen, and cells were cultivated at 33°C (permissive conditions). To induce differentiation, podocytes were cultured on type I–collagen (Biochrom, Berlin, Germany) at 37°C without γ-interferon (nonpermissive conditions). Since it takes at least 10 d to induce differentiation, podocytes maintained for less than 1 wk under nonpermissive conditions are referred to as “differentiating cells” or “process-forming cells,” while those cultivated for 2–3 wk at nonpermissive conditions were termed “differentiated cells” or “process-bearing cells.” In the present study, podocytes of clone 5 (MPC-5; Mundel et al., 1997b) between passage 15 and 25 were used in all experiments.

The polarity of MTs in podocytes was determined by the “hook” decoration method (Heidemann, 1991). MT proteins (C₂S fraction) were enriched from porcine brains by two cycles of temperature-dependent depolymerization and polymerization in the presence of glycerol (Shelanski et al., 1973). Tubulin was purified from the C₂S fraction by column chromatography (Vallee, 1986). Differentiated podocytes that had been cultured for 2–3 wk under nonpermissive conditions were rinsed briefly with PBS and permeabilized by incubation with prewarmed PEMG buffer (0.5 M Pipes, pH 6.9, 1 mM EGTA, 1 mM MgCl₂, and 1 mM GTP) containing 0.1% saponin (Sigma) and 2 mg/ml tubulin for 2 min. Subsequently, permeabilized cells were incubated with PEMG buffer containing 2 mg/ml tubulin, but without saponin, at 37°C for 30 min. Both preparations, C₂S fraction and purified tubulin, gave identical results for hook decoration. Samples were fixed with 4% glutaraldehyde in PBS and further processed for transmission electron microscopy (TEM). Sequences of serial ultrathin sections, including the profiles of centrosomes, were obtained to confirm the proximal–distal axis of examined podocytes and observed by TEM. The polarity of each MT (either plus end– or minus end–distal) was determined by “hook-handedness” as described by Heidemann (1991). Viewed from the cell periphery, MTs with clockwise hooks were judged as plus end–distal, whereas MTs with counterclockwise hooks were judged as minus end–distal. This evaluation was done in two series of ultrathin sections, each containing part of a cell starting at the centrosome and extending into the periphery of a process as reported previously (Takemura et al., 1995). The numbers of MTs identified either as plus end– or as minus end–distal on each electron micrograph were summarized among the sections selected from several independent sequences (Table I).

The characterization of the antibodies against CHO1/MKLP1 motor protein were previously reported (Sellitto and Kuriyama, 1988; Nislow et al., 1992; Kuriyama et al., 1994). Differentiated podocytes were fixed with absolute methanol at −20°C for 5 min and immunostained with polyclonal CHO1/MKLP1 antibody at 1:50 dilution. In some experiments, double labeling immunofluorescence with polyclonal anti-CHO1/MKLP and monoclonal α-tubulin (Oncogene Science, Hamburg, Germany) was carried out. Antigen–antibody complexes were visualized with fluorochrome (Cy2, Cy3)-conjugated secondary antibodies (BioTrend, Cologne, Germany). CHO cells were used as a positive control of immunostaining.

Total RNA was isolated from differentiated and undifferentiated podocytes using the RNeasy kit (QIAGEN, Hilden Germany). Reverse transcription (RT) of cDNA and PCR were performed according to standard protocols using the following two primers: GCACACCTGTCAATGTCACC (sense primer; nucleotide number +2213 to +2232 of hamster CHO1/MKLP1 cDNA; Kuriyama et al., 1994) and TGATCTACCCATCTGCTCCC (antisense primer; +2425 to +2444). These two primers are identical between hamster and mouse CHO1/MKLP1. After 40 cycles of amplification, PCR products were analyzed by agarose gel electrophoresis.

Based on the cDNA sequence of CHO1/MKLP1, two antisense (AS) and two corresponding sense (S) oligonucleotides (AS1:CAGGTTTCCTGGGCATCTT, AS2:AGCTTTCGCTGGTTTCATG, S1:AAGATGCCCAGGAAACCTG, and S2:CATGAAACCAGCGAAAGCT) were designed as described previously (Sharp et al., 1997; Yu et al., 1997). AS1 corresponds to the sequence between +19 and +37 of the hamster CHO1/ MKLP1 cDNA, and AS2 corresponds to the sequence between −1 and +18; S1 and S2 are the inverse complements of AS1 and AS2, respectively. Purified phosphorothioated oligonucleotides (Research Genetics Inc., Huntsville, AL) were resolved in serum-free medium and kept at −80°C until use. Podocytes were cultured for 4 d under nonpermissive conditions before they were plated onto collagen-coated coverslips; these postmitotic podocytes begin to develop processes (Mundel et al., 1997b). Starting 1 d after plating, podocytes were cultivated for up to 7 d in the presence of oligonucleotides at 0.5–10 μM by daily renewal of oligonucleotide-containing medium and observed by phase contrast microscopy. Cells were immunostained for CHO1/MKLP1 and MTs or processed for hook decoration. In some experiments, podocytes were further cultured in the absence of oligonucleotides after antisense treatment to confirm the reversibility of the observed effects.

Podocytes cultured for 4 d under nonpermissive conditions were plated onto collagen-coated coverslips. 6 h after plating, cells were treated with 0.1–10 μM taxol (paclitaxel; Sigma) for 3 d and immunostained for MTs and AFs as described above. After taxol treatment, podocytes were further cultured for 3 d in the absence of taxol to permit recovery of MT arrays, followed by immunostaining for α-tubulin and CHO1/MKLP1. Some cells recovering from taxol treatment were treated with 10 μM nocodazole for 2 h before immunostaining.

The MTOC was visualized by observation of MT arrays along the time course during recovery from nocodazole treatment (Takemura et al., 1995). Differentiated podocytes were incubated with 10 μM nocodazole (methyl[5-(2-thienylcarbonyl)-1H-benzeimidazol-2yl]carbamate; Sigma) for 2 h; afterwards, nocodazole was washed out with normal medium. At several time points during the recovery phase (0, 1, 2, 5, 10, 30, 60, and 120 min), cells were fixed with a prewarmed (37°C) MT-stabilizing fixative (0.1 M Pipes, pH 6.9, 5 mM EGTA, 1 mM MgCl₂, 4% polyethylene glycol 6000, 0.5% Triton X-100, 2% paraformaldehyde, and 4% sucrose). This fixative extracts free tubulin subunits but leaves polymerized MTs intact (Okabe and Hirokawa, 1988). Immunostaining of paraformaldehyde-fixed cells was performed as previously described (Mundel et al., 1997b) using α-tubulin mAb and fluorochrome-conjugated phalloidin (Sigma). Some cells were stained with vimentin polyclonal antibody (ProGen, Heidelberg, Germany) to visualize the distribution of IFs. To label centrosomes, podocytes were fixed in the recovery phase after nocodazole treatment with methanol at −20°C for 5 min, rinsed with PBS, and stained with affinity-purified γ-tubulin antiserum raised against a Xenopus γ-tubulin fusion protein (Stearns et al., 1991; Reinsch and Karsenti, 1994; courtesy of S. Reinsch, EMBL, Heidelberg, Germany).

To characterize the MT arrays of cultured podocytes, we first examined their intracellular distribution and polarity. Differentiated podocytes possess prominent cell processes containing MTs that radiate from the perinuclear region into the cell periphery (Fig. 1, a and b). By applying the hook decoration technique in a series of ultrathin sections, the polarity of MTs in differentiated podocytes was determined. When tracing MTs from their distal end, clockwise hooks show their plus end–distal polarity, whereas counterclockwise hooks indicate their minus end–distal polarity (Heidemann, 1991). The results from two independent sets of ultrathin serial sections are shown in Table I. The analysis of cross-cut MT profiles of 15 sections from the two series revealed that 36% of MTs showed minus end– distal polarity with counterclockwise hooks (Table I). Heidemann (1991) reported that the rate of mismatching between the handedness of hooks and the polarity of MTs is less than 10%, i.e., if all MTs are oriented in a uniform plus end–distal fashion, then the percentage of counterclockwise hooks is less than 10%, e.g., 1–6% in mature neuronal axons where MTs are uniformly oriented (Baas et al., 1989). Therefore, our data show that podocytes contain minus end–distal MTs resulting in the nonuniform (mixed) polarity of MTs. The observed efficiency of ∼80% for hook decoration of MTs (the ratio of hook-decorated MTs among all MTs) in the present study was similar to levels of hook decoration observed in previous reports (Baas et al., 1989; Takemura et al., 1995). The frequency of minus end–distal MTs varied not only among different MT bundles but also within individual MT bundles. A given region of an MT bundle could contain up to 50% minus end–distal MTs, while a different region of the same bundle could contain only a few of them. Along the proximal–distal axis of one cell, the percentage of minus end– distal MTs per section was 43, 31, 20, 40, 33, 0, 17, 10, 20, 43, 50, and 20% from the proximal to distal sites. In other words, the percentage of minus end–distal MTs varied randomly along the proximal–distal axis.

Since the nonuniform polarity of MTs in podocytes could be achieved by motor-dependent assembly, we next examined the expression of CHO1/MKLP1. Differentiated podocytes showed CHO1/MKLP1 immunolabeling in a punctated pattern in the cytoplasm (Fig. 2). Double staining with α-tubulin showed that CHO1/MKLP1 was localized in a punctated pattern on MTs, indicating the association of CHO1/MKLP1 with MTs in podocyte processes (Fig. 2,c). To confirm the immunocytochemical results, we analyzed the expression of CHO1/MKLP1 by RT-PCR using RNA from undifferentiated and differentiated podocytes. In both samples, a PCR product with the expected length of 232 bp was amplified (Fig. 2 d). These findings corroborate the expression of CHO1/MKLP1 in podocytes during proliferation and after differentiation.

To establish a functional role of CHO1/MKLP1 in podocytes, expression of CHO1/MKLP1 was inhibited by AS oligonucleotide treatment of process-forming podocytes (for details see Materials and Methods). Treatment of differentiating cells with AS oligonucleotides for 7 d abolished process formation of podocytes; cell bodies remained small, and only a few short processes developed (Fig. 3,a, compare with Figs. 3,b and 1 a). The effects of AS oligonucleotides were dose dependent; cells treated with 3 μM were smaller in size than those treated with 1 μM. In all subsequent experiments, podocytes were treated with 3 μM oligonucleotides. In AS-treated cells, cytoplasmic expression of CHO1/MKLP1 was absent. Only a perinuclear, nonspecific background staining was found in these cells (Fig. 3,c). In contrast, S-treated cells showed a normal punctated distribution pattern CHO1/MKLP1 (Fig. 3,d, compare with Fig. 2,a). Hook decoration revealed that in AS-treated podocytes, MT polarity was almost uniform (Fig. 3,e). 83% of MTs showed plus end–distal polarity, which is significantly higher than those in control cells (Table I). In contrast, S oligonucleotide–treated cells contained MTs with nonuniform polarity in usual frequency (Fig. 3,f), proving that CHO1/MKLP1 is required for the generation of nonuniform MT polarity in podocytes. The antisense treatment was not accompanied by a reduction of polymerized MTs (data not shown). However, the cytoplasm of AS-treated cells contained less membrane-bounded organelles (Fig. 3,e) as compared with S-treated (Fig. 3,f) or nontreated cells (Fig. 1, c and d). In addition, AS-treated cells sometimes contained abnormally clustered organelles in their cell bodies (data not shown). These observations suggest an impaired transport of membrane bounded in AS-treated podocytes. In some experiments, cells were treated with AS oligonucleotides for 7 d and then further incubated for 7 d in the absence of AS oligonucleotides. In these cells, the effects of AS oligonucleotides were fully reversible; after 4 d without AS oligonucleotides, process formation and cytoplasmic expression of CHO1/MKLP1 were completely restored, and cells were indistinguishable from S-treated or untreated cells (data not shown). Treatment with S oligonucleotides did not interfere with process formation or normal expression of CHO1/MKLP1, with this clearly ruling out possible toxicity of oligonucleotides per se. These observations demonstrate that cytoplasmic expression of CHO1/MKLP1 along MTs is required for normal process formation of podocytes. When proliferating podocytes were treated with AS oligonucleotides, their proliferation rate was markedly reduced. In AS-treated cells, no mitotic spindles were found, while in S-treated or untreated preparations dividing cells with spindle formation were frequently seen (data not shown). This inhibition of podocyte cell division is consistent with previous reports describing CHO1/ MKLP1 as a mitotic motor (Selitto and Kuriyama, 1988; Nislow et al., 1990, 1992).

To study the course of MT rearrangement in podocytes, MT arrays were disrupted and then allowed to recover. The application of taxol to podocytes before process formation induced strong perinuclear clustering of MTs and inhibition of process formation (Fig. 4,a). After removal of taxol, cell bodies increased in size and processes developed. During the recovery phase, strongly labeled MT fragments (probably representing taxol-stabilized remnants) were detected along weakly labeled MTs in the cell body and in processes (Fig. 4,b). The amount of MTs continuously running from the centrosomal region to the cell periphery was much lower in cells recovering from taxol treatment (Fig. 4,b) than in normal podocytes (Fig. 1,b). Treatment of these cells with 10 μM nocodazole for 2 h allowed us to distinguish two populations of MTs: weakly stained MTs susceptible to nocodazole treatment and strongly labeled MT fragments resistant to nocodazole treatment (Fig. 4,c). Such strongly stained MT fragments were preferentially observed after taxol treatment and subsequent nocodazole treatment. These findings indicate that taxol-stabilized MT fragments became resistent to nocodazole treatment (de Brabander et al., 1981), while MTs assembled de novo after taxol treatment were sensitive to nocodazole. Furthermore, during recovery of normal podocyte morphogenesis after taxol removal, MT fragments were redistributed from perinuclear MT clusters to the cell periphery in a punctated pattern along newly assembled MTs. The punctated distribution pattern of CHO1/MKLP1 in these cells was quite similar to that of nocodazole-resistent MT fragments (Fig. 4, b and d). The association of CHO1/MKLP1 with MTs was confirmed in double labeling experiments (data not shown).

To determine whether MT nucleation also plays a role in the arrangement of minus end–distal MTs, MT nucleation sites were visualized in podocytes. Treatment with nocodazole for 2 h caused complete disassembly of MTs in differentiated podocytes (Fig. 5, a and b). In nocodazole-treated cells, tubulin staining was diffuse around nuclei or almost absent (Fig. 5,a, compare with Fig. 1,b) since free tubulin subunits had been extracted by detergent treatment during fixation (for details see Materials and Methods). After complete depolymerization of MTs with nocodazole, recovery of MT arrays started immediately after removal of nocodazole. Nucleation of MTs was first observed at γ-tubulin–containing perinuclear foci (Fig. 5, c–e) representing centrosomes (Stearns et al., 1991; Baas and Joshi, 1992). These results show that the γ-tubulin–expressing centrosome is the major MTOC in podocytes. Within 5 min after removal of nocodazole, MTs nucleated in an asterisk-like pattern from centrosomes (Fig. 5, c–e). MTs continued to elongate into the periphery over the next 30 min, and after 60 min almost the entire cytoplasm was filled with MTs (Fig. 5, f and g). At 1–2 h of recovery, MTs were aggregated into discrete bundles, leaving in between regions largely devoid of MTs (Fig. 5,h). Such regions of MT-depleted cytoplasm were also found in nontreated control cells (Fig. 1,b). During the early stages of recovery, in addition to MTs emerging from centrosomes, short MT fragments were observed in cell bodies and processes (Fig. 5,e). During the later stages of recovery, all MTs emerged from the centrosome (Fig. 5, g and h).

To determine the contribution of other cytoskeletal elements to MT organization, we examined the distribution of AFs and IFs during elongation of MT arrays. The distribution of AFs was not affected by nocodazole treatment. During recovery from nocodazole treatment, no spatial relationship between MT nucleation sites and AF foci was found (data not shown). In contrast, vimentin IFs lost their fibrillar distribution and collapsed in the perinuclear region after nocodazole treatment. The reassembly of MTs was accompanied by a restoration of the normal IF distribution pattern (data not shown). These findings suggest an interaction between MTs and IFs as reported for other cells (Virtanen et al., 1980; Celis et al., 1984). Moreover, time course analysis revealed that the recovery of MT arrays preceded that of IFs. Therefore, we conclude that neither AFs nor IFs are important for the recovery of MT arrays after nocodazole treatment.

The present study shows that podocytes exhibit a nonuniform MT polarity. Since direct formation of minus end– distal MTs is not possible at centrosomes (Pereira and Schiebel, 1997), two hypotheses were proposed for establishing the nonuniform MT-polarity (Baas, 1997; Kobayashi and Mundel, 1998): (a) short plus end–distal MTs are released from the centrosome and transported in antiparallel orientation into the cell periphery along preexisting plus end–distal MTs; or (b) minus end–distal MTs are assembled at MTOCs other than the centrosome. The data from the present study support the first hypothesis. Suppression of CHO1/MKLP1 expression with AS oligonucleotides showed that CHO1/MKLP1 is necessary for the development of nonuniform polarity in podocytes (Fig. 3). CHO1/MKLP1 is located in a punctated pattern along MTs in podocytes (Fig. 2). This pattern is similar to the distribution of CHO1/MKLP1 in neuronal dendrites, where transport of minus end–distal MTs was also described (Sharp et al., 1997). These data suggest that minus end– distal MTs are transported by CHO1/MKLP1 motor proteins along plus end–distal MTs in podocytes. The alternative hypothesis, i.e., the presence of MTOCs other than the centrosome, which plays a crucial role in the assembly of MTs (Joshi et al., 1992; Moritz et al., 1995), does not receive any support from the present study since γ-tubulin was exclusively found at centrosomes (Fig. 5). Apart from a transient centrosome-independent reassembly of MTs during the early stages of recovery from nocodazole treatment, reassembly of MTs exclusively occurred at centrosomes. (Fig. 5). The transient reassembly of MTs probably represented polymerization without any MTOC activity. As suggested by previous works, after complete depolymerization of MTs, the concentration of free tubulin subunits may be higher than the critical assembly concentration, allowing MTs to spontaneously assemble in the absence of MTOCs (Stearns et al., 1991). Therefore, it is reasonable to conclude that the γ-tubulin–expressing centrosome represents the only regular MT assembly site in podocytes. In neuronal cells, the centrosome is the major MTOC, as shown by the same experimental approach used in the present study (Baas and Joshi, 1992; Yu et al., 1993). Although it has recently become clear that MTs can be polymerized by centrosome-free nucleation (Hyman and Karsenti, 1998), the generation of minus end–distal MTs in podocytes by centrosome-free nucleation was ruled out by the AS experiments. These experiments clearly showed that the mixed polarity of MTs, which is independent of MT nucleation (Baas, 1997), depends on the expression of CHO1/MKLP1 (Fig. 3).

CHO1/MKLP1 is a member of the kinesin superfamily of motor proteins as shown by sequence analysis and in vitro motility analysis (Nislow et al., 1992). When overexpressed, CHO1/MKLP1 induces bundling of MTs in a nonuniform orientation, indicating that CHO1/MKLP1 is associated with MTs rather than with membranous organelles (Kuriyama et al., 1994; Sharp et al., 1996). During mitosis, CHO1/MKLP1 causes antiparallel sliding between nonuniformly arranged MTs, accounting for the elongation of mitotic spindles (Nislow et al., 1992). Actually, AS oligonucleotides for CHO1/MKLP1 inhibited cell division of proliferating podocytes (the present study) and neuroblastoma cells (Yu et al., 1997). In addition, reduced expression of CHO1/MKLP1 is accompanied by a uniform plus end–distal microtubular polarity in neuronal cells (Sharp et al., 1997; Yu et al., 1997). These findings suggest that CHO1/MKLP1 acts as a motor protein transporting minus end–distal MTs in dendrites. Our data in podocytes provide further evidence for a role of CHO1/MKLP1 as a transporter of minus end–distal MTs. These minus end– distal MTs are short and colocalize with CHO1/MKLP1 in a punctated pattern along MTs. The MT transport hypothesis is further supported by the similar distribution of taxol-stabilized MT fragments and CHO1/MKLP1 during recovery from taxol treatment (Fig. 4), strongly suggesting that taxol-stabilized MT fragments are transported along MTs by CHO1/MKLP1.

The present study clearly demonstrated the redistribution of taxol-stabilized MT fragments into the cell periphery. Interestingly, process formation of podocytes was restored after removal of taxol, although it had previously been reported that taxol remained in the cytoplasm even after removal from the culture medium (Jordan et al., 1996). Thus, taxol retained on MTs may continue to stabilize MT fragments in taxol-treated podocytes. The inability to demonstrate short MT fragments without taxol treatment (Figs. 1 b and 5, g and h) may simply be due to the fact that immunofluorescence microscopy is not sensitive enough to discriminate between short transported MT fragments and long stationary MTs.

The hypothesis that MTs are transported as preformed segments was controversial with many investigators who claimed that only the transport of tubulin subunits contributes to axonal flow of MTs (see for example Baas and Brown, 1997; Hirokawa et al., 1997). In axons, MTs are uniformly oriented in a plus end–distal fashion (Heidemann et al., 1981, 1984; Baas et al., 1988), indicating that axons lack a mechanism to arrange MTs in minus end–distal fashion and consistent with the observation that axons do not contain CHO1/MKLP1 (Sharp et al., 1997,b). Therefore, organization of the microtubular system in axons is fundamentally different from that in podocytes and dendrites. Interestingly, however, Ahmad et al. (1998) recently reported that motor-driven transport of MTs is necessary for axonal elongation.

The present study suggests a mechanism for the generation of the nonuniform MT polarity in podocytes (summarized in Fig. 6). First, plus end–distal MTs, which serve as rails for MT transport, nucleate at the centrosome. Second, short MT fragments are assembled at the centrosome and then released from the centrosome (Fig. 6, 1 and 2). Baas (1997) proposed the MT-severing protein katanin (McNally et al. 1996) as a candidate to release MT fragments from the centrosome. If the released MT fragments are very short, their orientation may be reversed by thermodynamic movements. Third, minus end–distal MT fragments are connected to plus end–distal MTs via CHO1/ MKLP1 and then transported into developing processes (Fig. 6, 3 and 4). The results from the AS experiments clearly show that the cytoplasmic expression of CHO1/ MKLP1 is crucial for process formation of podocytes (Fig. 3). Similar effects of antisense treatment were recently described in neuronal dendrites (Sharp et al., 1997; Yu et al., 1997). These findings strongly suggest that minus end–distal MTs play a crucial role in process formation. Generally, process formation is dependent on a continuous flow supplying growing processes with cytoskeletal elements as well as membrane precursors. Minus end–distal MTs may support the transport of elements that are carried only by minus end–directed motors into the cell periphery, e.g., components of the Golgi apparatus in dendrites (Baas et al., 1989), since distinct motor proteins selectively carry specific cargos (Hirokawa, 1997). The problem arising from this concept is that minus end–distal MTs do not form continuous rails but are arranged in a discontinuous pattern along plus end–distal MTs. Therefore, an alternative model would be that transported minus end–distal MT fragments are connected with other cytoskeletal and/or membranous elements without the interposition of another motor protein. The transport of these elements into elongating processes may then be achieved together with minus end–distal MT fragments transported by CHO1/ MKLP1.

Several transport studies clearly showed that membrane traffic in neuronal dendrites, but not in axons, is driven by a machinery similar to the “basolateral” membrane traffic in polarized epithelia (de Hoop and Dotti, 1993). The present study revealed the role of nonuniform MT polarity established by CHO1/MKLP1 in process formation of podocytes, similar to that reported for dendrites (Sharp et al., 1997). It is tempting to speculate that the development of podocyte processes and neuronal dendrites is driven by the “basolateral” transport machinery depending on nonuniformly oriented MTs. Membrane traffic of neuronal dendrites and of epithelial basolateral membranes is regulated by the small GTPase Rab8 (Huber et al., 1993a,b). Transfection of a hyperactive mutant of Rab8 induces process formation in fibroblasts (Peränen et al., 1996), suggesting that Rab8 promotes process formation via regulation of membrane traffic in process-bearing cells. Rab6, another small GTPase, can interact with a kinesin-like motor protein with a high sequence homology to CHO1/MKLP1 (Echard et al., 1998). Therefore, small GPTases may regulate process formation via their interaction with CHO1/MKLP1, although the underlying mechanism remains to be elucidated.

In summary, we have shown that expression of CHO1/ MKLP1 motor and nonuniform polarity of MTs are necessary for process formation of podocytes, indicating that podocytes and dendrites share the same mechanism of process formation. In contrast, the uniform MT polarity in axons is achieved by dynein and dynactin-driven transport of plus end–distal MTs (Ahmad et al., 1998). The present study emphasizes that the transport of MT fragments by motor proteins is a general mechanism for process formation in diverse cell types. Since MTs disassemble at minus ends without protection by capping proteins (Keating et al., 1997; Pereira and Schiebel, 1997), the identification of factors preventing disassembly of MT fragments during their transport awaits clarification.

We would like to thank Dr. Sigrid Reinsch (EMBL, Heidelberg, Germany) for providing the γ-tubulin antibody and Dr. Kai Simons (EMBL) for critical reading of the manuscript. We also thank Hiltraud Hosser and Alexandra Zeller for excellent technical assistance as well as Rolf Nonnenmacher for graphic work.

AF

actin filament

AS

antisense

IF

intermediate filament

MT

microtubule

MTOC

microtubule-organizing center

RT-PCR

reverse transcription PCR

S

sense

TEM

transmission electron microscopy