Cytoplasmic dynein is involved in a wide variety of cellular functions. In addition to the initially characterized form (MAP 1C/dynein 1), a second form of cytoplasmic dynein (dynein 2) has been identified and implicated in intraflagellar transport (IFT) in lower eukaryotes and in Golgi organization in vertebrates. In the current study, the primary structure of the full-length dynein 2 heavy chain (HC) was determined from cDNA sequence. The dynein 1 and dynein 2 sequences were similar within the motor region, and around the light intermediate chain (LIC)-binding site within the N-terminal stem region. The dynein 2 HC co-immunoprecipitated with LIC3, a homologue of dynein 1 LICs. Dynein 2 mRNA was abundant in the ependymal layer of the neural tube and in the olfactory epithelium. Antibodies to dynein 2 HC, LIC3 and a component of IFT particles strongly stained the ependymal layer lining the lateral ventricles. Both dynein 2 HC and LIC3 staining was also observed associated with connecting cilia in the retina and within primary cilia of non-neuronal cultured cells. These data support a specific role for dynein 2 in the generation and maintenance of cilia.
Cytoplasmic dynein is a minus end-directed microtubule motor protein ( Paschal and Vallee, 1987). The major conventional form of the protein, MAP (microtubule-associated protein) 1C or dynein 1, consists of several subunits. The 530 kDa heavy chain (HC) is responsible for force production. The intermediate chain (IC: 74 kDa), light intermediate chains (LICs: 53-57 kDa), and light chains (LCs: 8-21 kDa) have been implicated in linking the HC with diverse forms of subcellular cargos ( Purohit et al., 1999; Tai et al., 2001; Yano et al., 2001). This form of cytoplasmic dynein has been implicated in a great number of functions including retrograde axonal transport; the distribution and transport of the Golgi apparatus, endosomes, lysosomes, and nuclei; chromosomal movement; mitotic spindle organization and orientation and neuronal migration ( Vallee and Sheetz, 1996).
A second cytoplasmic dynein HC was identified by PCR-based screening, and referred to as dhc1b ( Gibbons et al., 1994), dlp4 ( Tanaka et al., 1995), dhc2 ( Vaisberg et al., 1996) or dnchc2 ( Vaughan et al., 1996). It was initially found to be upregulated during ciliogenesis in sea urchins ( Gibbons et al., 1994). However, in cultured mammalian cells it was localized to the Golgi apparatus, which could also be dispersed by microinjection of anti-dynein 2 HC antibody ( Vaisberg et al., 1996). Dynein 2 was subsequently implicated in IFT, a phenomenon discovered in the green alga Chlamydomonas ( Kozminski et al., 1993). In IFT, assembled `rafts' of protein complexes (IFT particles) composed of some17 polypeptides are translocated within the space between the plasma membrane and outer doublet microtubules of cilia and flagella ( Cole et al., 1998; Rosenbaum et al., 1999). A mutation in the dynein 2 HC gene interferes with retrograde movement of the IFT particles toward the base of the flagella ( Pazour et al., 1999; Porter et al., 1999). In C. elegans dynein 2 was identified as the CHE-3 gene product ( Wicks et al., 2000). CHE-3 was also implicated in retrograde IFT within sensory cilia ( Signor et al., 1999a). In both systems the heterotrimeric kinesin, kinesin II, is responsible for anterograde transport of IFT particles ( Kozminski et al., 1995; Beech et al., 1996; Muresan et al., 1997; Nonaka et al., 1998; Signor et al., 1999b).
Here we report the full-length primary structure of rat dynein 2 HC. We also report an association of dynein 2 HC with LIC3, a homologue of LIC1 and LIC2. Dynein 2 was most abundant in ciliated epithelia and in the connecting cilia of photoreceptor cells. Immunocytochemistry of cultured cells revealed a clear staining of primary cilia, but no specific association with the Golgi apparatus. These data favor a predominant role for dynein 2 in transport within ciliated structures in the brain and elsewhere, and indicate that the function of dynein 2 is evolutionarily conserved between vertebrates and invertebrates.
Materials and Methods
A random primed rat kidney cDNA library was screened using a PCR fragment of dynein 2 [Dnchc2 ( Vaughan et al., 1996)] as probe, resulting in isolation of a 1.4 kb clone D2-1. Rat testis cDNA clones encompassing the entire ORF of dynein 2 HC were isolated using Marathon cDNA Amplification Kit (Clontech, Palo Alto, CA) and primers having sequence near the 3′ and 5′ ends of the determined cDNA sequence. At least two independent clones were sequenced and compared to avoid cloning artifacts. Discrepancies in the sequence were resolved by additional clones. To clone an IFT particle component, we searched the database for sequences similar to the ngd5 ( Wick et al., 1995) which is reported to be the mammalian orthologue of the Chlamydomonas p52 IFT component ( Cole et al., 1998). A human sequence, CGI-53 [GenBank accession number AF151811 ( Lai et al., 2000)], was found to be 87% identical to NGD5. A PCR product encoding Met(1) to Asn(436) of the CGI-53 was synthesized using human adrenal cDNA (Clontech) as template. The resulting ∼1.3 kb fragment was cloned in the pET23d vector (Novagen, Madison, WI) for bacterial expression. To clone LIC3, a PCR product encompassing the entire ORF of the CGI-60 sequence ( Lai et al., 2000) was synthesized using a human EST (accession number EST183255) as template. The resulting ∼1.1 kb fragment was subcloned in the pET16b vector for bacterial expression.
Rabbit antidynein 2 HC antibody was generated using a bacterially expressed fragment consisting of the N-terminal 269 amino acids purified by SDS-PAGE. Rabbit antibody against amino acids 1776 to 2242 of rat brain dynein 1 HC ( Mikami et al., 1993; Zhang et al., 1993) was generated using bacterially expressed polypeptide. For the anti-LIC3 antibodies, the soluble recombinant protein was purified by Ni-NTA column chromatography (Qiagen, Valencia, CA). For the anti-CGI-53 antibody, the inclusion body was solubilized with 8 M urea and was purified using a Ni-NTA column. Anti LIC1 to detect LIC1 and LIC2 ( Tynan et al., 2000a) was described previously. The K26 monoclonal antibody is directed at an unidentified epitope of the bovine photoreceptor connecting cilium ( Horst et al., 1990).
Anti-dynein IC Ab was purchased from Chemicon, Temecula, CA. Rabbit polyclonal anti-detyrosinated tubulin Ab ( Chapin and Bulinski, 1991) was kindly provided by Steven J. Chapin and J. Chloë Bulinski (Columbia University, New York, NY). Anti-Golgi 58K protein mAb and anti-acetylatedα -tubulin mAb were purchased from Sigma (St Louis, MO). Anti-digoxygenin antibody conjugated with alkaline phosphatase was purchased from Roche Diagnostics, Mannheim, Germany. Cy3-conjugated anti-rabbit IgG for immunohistochemistry or immunocytochemistry were purchased from Roche Molecular Biochemicals, Indianapolis, IN or Jackson ImmunoResearch Laboratories, West Grove, PA, respectively. Alexa 488-conjugated anti-mouse IgG was purchased from Molecular Probes, Eugene, OR.
Cytosolic extracts were prepared from fresh rat testis after homogenization in 50 mM PIPES-NaOH, 50 mM HEPES, pH 7.0, containing 2 mM MgCl2, 1 mM EDTA, 1 mM AEBSF, 2 μg/ml leupeptin, 5 μg/ml soybean trypsin inhibitor, 2 μg/ml aprotinin, and 1 mM benzamidine (extraction buffer) as modified from Paschal et al. ( Paschal et al., 1987). Immunoprecipitation was carried out using Protein A Sepharose Beads (Pharmacia, Piscataway, NJ) at 4°C overnight and analyzed by SDS-PAGE followed by immunoblotting. For sucrose density gradient centrifugation the cytosolic extract was loaded on a 11 ml 5-20% sucrose gradient in the extraction buffer without protease inhibitors and centrifuged for 22 hours at 175,000 g in a Beckman SW41 rotor. Quantitative image processing was done with ImageJ software (Wayne Rasband, NIH).
In situ hybridization
In situ hybridization of frozen sections of mouse embryo was performed as described previously ( Saito et al., 1996). Briefly, after pretreatment (proteinase K, postfixation, and acetic anhydride), sections were prehybridized for 3 hours and hybridized with digoxygenin-labelled cRNA probes (1 μg/ml) for 15 hours. Probes used for in situ hybridization were: an EcoRI(402)-EcoRI(3534) fragment of dynein 1 cDNA (GenBank accession no. L08505); a SacI(473)-XbaI(4065) fragment of dynein 2 cDNA; a full-length SCG10 cDNA ( Stein et al., 1988). Signals were visualized using anti-digoxygenin antibody conjugated with alkaline phosphatase and NBT/BCIP (BioRad, Hercules, CA). Adjacent sections used for sense and anti-sense cRNA probes of dynein 2 were treated under the same conditions.
For immunohistochemistry of mouse brain, anesthetized mice were perfused intracardially with saline and then with ice-cold 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.3). The brains were postfixed 6-15 hours at 4°C, immersed in 30% sucrose phosphate buffer, and 40 μm thick frozen sections were cut in the coronal plane with a sledge microtome. Sections were collected in 0.05 M phosphate buffered saline (PBS) at pH 7.4, washed and incubated 15-18 hours at 4°C in PBS with 2% normal goat serum (NGS-PBS) and 0.3% Triton X-100 containing diluted primary antisera. Sections were washed with PBS, incubated with Cy3-conjugated anti-rabbit IgG for 60-90 minutes (1:50 in NGS-PBS). Sections were mounted onto gelatin-coated slides, counterstained with bisbenzimide, and placed on coverslips for observation with fluorescence microscopy.
For immunohistochemistry of the retina, fresh mouse or bovine retinal tissue was placed in Tissue Freezing Medium (Triangle Biomedical Sciences, NC) and quickly frozen in liquid nitrogen with or without prior fixation in 4% paraformaldehyde. Primary antibodies were detected with goat anti-rabbit or goat anti-mouse IgG conjugated with Cy3 (Jackson Laboratories) or Alexa 488. Discrimination of signals for K26 versus dynein 2 HC involved use of conjugated anti-mouse and anti-rabbit antibodies, respectively. For discrimination of two monoclonal antibodies (K26 versus tubulin), we labeled with one monoclonal antibody and a fluorescent anti-mouse antibody, and then repeated the procedure for the second monoclonal antibody using a different fluorophore. Images of cells labeled with more than one fluorophore were pseudocolored and merged using Adobe Photoshop.
Normal rat kidney epithelial cells (NRK cells) on coverslips were fixed with 4% paraformaldehyde in 137 mM NaCl, 5 mM KCl, 1.1 mM Na2HPO4, 0.4 mM KH2PO4, 2 mM MgCl2, 10 mM EGTA, 5 mM PIPES, and 5.5 mM glucose, pH 6.1 ( Wheatley and Wang, 1998) at 37°C, permeabilized on ice in the same buffer containing 0.5% Triton X-100, air-dried, then blocked in 0.5% BSA in PBS (blocking buffer) for 1 hour at room temperature. Coverslips were incubated with blocking buffer containing primary rabbit sera (anti-dynein 2 HC, anti-LIC3 or anti-detyrosinated tubulin) and either anti-Golgi 58K protein or anti-acetylated α-tubulin mAb for 1 hour at 37°C, then with Cy3-conjugated anti-rabbit IgG and Alexa 488-conjugated anti-mouse IgG. After two washes in blocking buffer, coverslips were mounted using Antifade Kit (Molecular Probes) for fluorescent microscopy.
Molecular and biochemical analysis of dynein 2
cDNA cloning of dynein 2 HC was initiated with an isolation of a 1.4 kb clone from a rat kidney cDNA library. This clone was used to obtain overlapping rat testis cDNAs containing an open reading frame encoding 4,306 amino acids by PCR extension. Fig. 1A shows a comparison of the full-length dynein 2 HC with that of dynein 1 HC ( Mikami et al., 1993; Zhang et al., 1993) and with a C. elegans dynein HC gene (F18C12.1, accession number Z75536) which corresponds to the che-3 locus ( Wicks et al., 2000). Rat dynein 2 HC is similar to the CHE-3 throughout, with 42% identical amino acids. Comparison of the dynein 2 HC sequence with the dynein 1 HC revealed clear conservation within the C-terminal two-thirds. Within the N-terminal third, the only well-conserved region (amino acids 533 to 772 in dynein 1 sequence) corresponds to the LIC binding-site and a part of IC binding site [amino acids 649 to 800 and 446 to 701, respectively, from Tynan et al. ( Tynan et al., 2000a)].
In the course of our previous studies of LIC1 and LIC2 function ( Tynan et al., 2000a; Tynan et al., 2000b), we identified a related sequence in the database, CGI60 ( Lai et al., 2000; Mikami et al., 2001), which we have termed LIC3. During the late stages of the current project the same sequence was reported under the name D2LIC ( Grissom et al., 2002). LIC3 is similar to a C. elegans sequence F02D8.3 throughout its length ( Fig. 1B, right box). LIC3 (350 amino acids) is much shorter than LIC2 (497 amino acids) and exhibits only 24% identity. However the Smith-Waterman score ( Pearson, 1991) for LIC3 vs. LIC2 is 200, highly significant (1.6×10-7) relative to the score vs. the shuffled LIC2 sequence. Both shared a P-loop motif in the N-terminus and similarity was observed around this site ( Fig. 1B, left box). In contrast, a P-loop motif was not observed in the C. elegans gene and the corresponding region revealed no sequence similarity.
To test which form of dynein LIC3 might associate with, we compared the behavior of LIC3 with that of dynein 1 and 2. Antibodies prepared against the dynein 2 HC and LIC3 each recognized a single major band of the expected size ( Fig. 2A). By sucrose density gradient centrifugation, dynein 1 HC, the 74 kDa dynein 1 IC, and LICs 1 and 2 sedimented very similarly, showing a sharp peak at ∼20S. Dynein 2 HC and LIC3 showed a broader distribution with a peak in fractions 6 to 8 (∼17S) and faster sedimenting material extending to the bottom of the gradient, perhaps associated with membranous or axonemal material ( Fig. 2A; see below). The profile of LIC3 was similar to that of dynein 2 HC, though a small amount of free LIC3 could be seen at low S-value. A component of the IFT particles (CGI-53) had a single major peak (∼19S) between those of dynein 1 and 2.
We tested further for an interaction between LIC3 and dynein 2 HC by co-immunoprecipitation from testis cytosolic extracts. LIC3 was found at high levels in dynein 2 HC immunoprecipitates ( Fig. 2B, right panel). A low, but detectable level of dynein 2 HC was observed in antiLIC3 immunoprecipitates ( Fig. 2B, left panel), suggesting that the LIC3 antibody may partially interfere with dynein 2 HC-LIC3 interaction.
Cellular and subcellular localization of the dynein 2 complex
To gain insight into the different functions attributed to dynein 2, and in view of its relative abundance in juvenile rodent brain ( Tanaka et al., 1995), we examined its distribution in the nervous system. In newborn and adult mice, the most striking immunohistochemical staining for dynein 2 HC was observed in the ependymal layer lining the lateral ventricles ( Fig. 3A,D). Anti-LIC3 Ab stained the same region ( Fig. 3B), consistent with our biochemical evidence for an association with dynein 2. A low level of staining for both dynein 2 HC and LIC3 was also observed outside the ependymal layer. Prominent staining of CGI-53 was observed in the ependymal layer ( Fig. 3C).
We also studied the distribution of dynein 2 HC mRNA. In developing neural tube dynein 2 HC mRNA was detected most intensely in the ependymal layer ( Fig. 4A,B), and weaker in the mantle layer distinguished with the neuronal marker SCG10 [ Fig. 4D ( Anderson and Axel, 1985)]. Dynein 1 HC mRNA was detected in the mantle layer and in a narrow band lining the surface of the central canal ( Fig. 4C). Therefore dynein 2 mRNA appears to be expressed not only in the ependymal cells but also in the neuronal/glial progenitor cells in this layer.
We also examined the expression of dynein 2 HC ( Fig. 4E), dynein 1 HC ( Fig. 4F) and SCG10 ( Fig. 4G) mRNA in the olfactory epithelium. At higher magnification, the distribution of the dynein HC species differed ( Fig. 4H). Dynein 2 HC mRNA was clearly detected not only in the receptor cells but also in the basal cells containing precursors of the receptor cells. Dynein 1 HC mRNA was also detected in the receptor cells and apical region of the supporting cells, but was absent in the basal cells. The SCG10 mRNA was detected in basal half of the receptor cells layer, in accordance with recent finding of restricted expression in immature and olfactory marker protein-negative cells ( Pellier-Monnin et al., 2001). These data show that dynein 2 HC is expressed in ciliated cell species in the brain and also in neuronal precursors.
We examined the bovine retina by immunohistochemistry to determine the association of dynein 2 HC with the immotile connecting cilia. A substantial level of traffic is thought to occur through these structures, with an important potential role for dynein 2 in the retrograde direction ( Rosenbaum et al., 1999). Dynein 2 HC staining was prominent on the proximal side of the connecting cilium, the structural link between the inner and outer segments ( Fig. 5A,E), often extending further into the inner segment ( Fig. 5D). LIC3 also showed increased staining at the proximal side of the connecting cilium ( Fig. 5B,F). Staining within the connecting cilia was punctate, which appears consistent with an association with IFT particles (yellow overlap). Staining sometimes extended into the outer segment ( Fig. 5F), where anti-acetylated α-tubulin staining revealed the axonemal microtubules to extend ( Fig. 5C).
We also conducted immunofluorescence microscopy of NRK cells, which contain primary cilia ( Wheatley and Wang, 1998). This structure (arrows in Fig. 5G-L) was readily detected with anti-detyrosinated ( Fig. 5G) and anti-acetylated α-tubulin ( Fig. 5H) Abs, which preferably stain stable microtubules ( Chapin and Bulinski, 1991; Wheatley, 1995). Anti dynein 2 HC ( Fig. 5I) and anti LIC3 ( Fig. 5K) Abs also stained the primary cilia, but in a punctate manner. Outside the primary cilia, dynein 2 HC was detected diffusely throughout the cytoplasm at low levels, but clear colocalization with Golgi 58K protein ( Bashour and Bloom, 1998) was not observed ( Fig. 5M,N).
Dynein 2 structure
The comparison of dynein 2 HC sequence to dynein 1 HC suggests that these genes diverged relatively early in evolution. The N-terminal stem region of dynein 2, which is the most distinctive part of the dynein HCs in general, is clearly divergent from that of dynein 1, except for the middle part of this region ( Fig. 1A). The conserved region includes the LIC binding site within dynein1, consistent with the presence of this type of accessory subunit in dynein 2. LIC3 itself also represents a relatively divergent member of the LIC class of polypeptides. Despite the lower extent of conservation however, vertebrate LIC3, like LIC1 and 2, contains an N-terminal P-loop motif characteristic of nucleotidases, and also exhibits similarity to known ATPases in the surrounding sequence ( Gill et al., 1994; Hughes et al., 1995). It is not known whether LICs are enzymatically active, and the P-loop element is absent in the closest C. elegans LIC sequence ( Fig. 1B). Therefore nucleotidase activity may not be required at least in lower organisms.
The dynein 2 complex exhibited a peak at ∼17S [ Fig. 2A ( Vaisberg et al., 1996)], an S-value lower than that of the dynein 1 complex, suggesting that the dynein 2 could be single-headed or partly dissociate during purification. That LIC3 is a subunit of dynein 2 is indicated by the efficiency of cosedimentation ( Fig. 2A) and co-immunoprecipitation ( Fig. 2B), and colocalization with dynein 2 HC (Figs 3, 5). That LIC3 is physiologically related to dynein 2 is also suggested by recent evidence that the C. elegans homologue of LIC3 (F02D8.3) and the IFT particle subunit OSM-6 [39% identical to CGI53 ( Collet et al., 1998)] are regulated by a common transcription factor DAF-19 ( Swoboda et al., 2000).
Dynein 2 function
Our data favor a role for dynein 2 in IFT in vertebrates. High levels of expression were observed in the ependymal cells (Figs 3, 4) which bear motile cilia. Clear staining of immotile connecting cilia ( Fig. 5A,B,D-F) and primary cilia ( Fig. 5I,K) was also observed. These observations indicate that the dynein 2 complex is generally present in axonemal structures, and argue against a potential role for dynein 2 in axonemal bending motility. The punctate staining we observe within cilia is consistent with images from nonvertebrates ( Pazour et al., 1999; Wicks et al., 2000), and suggests an association with IFT rafts, which appear to represent collections of IFT particles. The presence of dynein 2 in photoreceptor connecting cilia suggests a role in retrograde transport from the outer to the inner segment. Recent work has, indeed, demonstrated rapid and massive light-induced retrograde translocation of transducin between these compartments ( Sokolov et al., 2002). In addition, because we also detect dynein 2 in the inner segment, it is possible that this form of cytoplasmic dynein, like dynein 1 ( Tai et al., 2001), also participates in apical sorting of rhodopsin to the base of the connecting cilium.
We did not observe clear evidence for Golgi staining with anti-dynein 2 or anti-LIC3 antibody even in cells where the antibody clearly recognized cilia ( Fig. 5M,N). We do not understand the basis for the difference with other studies ( Vaisberg et al., 1996; Grissom et al., 2002), which potentially could reflect differences in sample preparation. Alternatively, the reported Golgi localization ( Vaisberg et al., 1996; Grissom et al., 2002) and apical staining in cultured tracheal epithelial cells ( Criswell et al., 1996) could be related to the accumulation of dynein 2 at the base of axonemal structures seen in the retina ( Fig. 5A,D,E) or Chlamydomonas ( Pazour et al., 1999).
A role for dynein 2 in undifferentiated cells still remains a possibility. Dynein 2 mRNA was more abundant in newborn than in adult murine brain ( Tanaka et al., 1995), suggesting a potential role during development. We observed dynein 2 mRNA within the neuroblasts of the developing neural tube ( Fig. 4A) and in the basal cells of olfactory epithelium ( Fig. 4E,H). What the function of dynein 2 might be in these cells is uncertain. It could participate in Golgi organization, some novel form of intracellular protein transport, or early mitotic events. It is possible that some of the staining in the developing layers could be of epithelial cells prior to ciliation. In support of this possibility, expression of dynein 2 mRNA has been reported in tracheal epithelial cells prior to induction of ciliogenesis ( Criswell et al., 1996).
LIC3 now emerges as a potential candidate for cargo binding during IFT. The dynein 1 ICs and LICs were absent in dynein 2 ( Fig. 2A), as were at least two dynein LCs, LC8 and Tctex-1 (our unpublished observations). LIC1 has been implicated in cargo binding during centrosomal assembly, binding to pericentrin ( Tynan et al., 2000b), which migrates along microtubules toward the centrosome in a dynein-dependent manner ( Young et al., 2000). By analogy, LIC3 might interact with a component of the IFT particles to mediate retrograde transport. Whether this is the case, and how the interaction with cargo is regulated remain important questions for future study.
The database accession number for dynein 2 HC sequence is AB041881. This work was supported by National Institutes of Health Grants GM47434 (R.B.V.) and by the Muscular Dystrophy Association (A.M.). Parts of this study were performed in the Molecular Neurobiology Laboratory of RIKEN (The Institute of Physical and Chemical Research) under instruction of Katsuhiko Mikoshiba. We thank Mitsunori Fukuda, Curtis G. Wilkerson, Sally P. Wheatley, Denis Dujardin and Thomas A. Schoenfeld for helpful discussion; and Stephen M. King for anti-LC8 and anti-Tctex-1 Abs.
- Accepted September 12, 2002.
- © The Company of Biologists Limited 2002