Different roles for KIF17 and kinesin II in photoreceptor development and maintenance



Kinesin 2 family members are involved in transport along ciliary microtubules. In Caenorhabditis elegans channel cilia, kinesin II and OSM-3 cooperate along microtubule doublets of the axoneme middle segment, whereas OSM-3 alone works on microtubule singlets to elongate the distal segment. Among sensory cilia, vertebrate photoreceptors share a similar axonemal structure with C. elegans channel cilia, and deficiency in either kinesin II or KIF17, the homologue of OSM-3, results in disruption of photoreceptor organization. However, direct comparison of the two effects is confounded by the use of different species and knockdown strategies in prior studies. Here, we directly compare the effects of dominant-negative kinesin II and KIF17 expression in zebrafish cone photoreceptors. Our data indicate that dominant-negative kinesin II disrupts function at the level of the inner segment and synaptic terminal and results in cell death. In contrast, dominant-negative KIF17 has no obvious effect on inner segment or synaptic organization but has an immediate impact on outer segment assembly. Developmental Dynamics 238:2211–2222, 2009. © 2009 Wiley-Liss, Inc.


Numerous studies have demonstrated that heterotrimeric members of the kinesin 2 family, referred to here as kinesin II, are required for intraflagellar transport (IFT) and ciliogenesis (Kozminski et al.,1995; Morris and Scholey,1997; Cole et al.,1998; Nonaka et al.,1998). Although kinesin II clearly plays a role as an anterograde IFT motor in Chlamydomonas rheinhartii and Caenorhabditis elegans, the involvement of at least one additional ciliary kinesin in C. elegans has led to a more complex view of kinesin's role in ciliogenesis. Some studies suggest that kinesin II might play a role in loading cargo on IFT proteins at the base of the cilium as well as transporting them along the axoneme (Scholey,2008). Furthermore, mutations in two of the subunits (KAP-1, KLP-11) of C. elegans kinesin II have no immediate effect on ciliary axoneme formation in sensory cilia (Snow et al.,2004; Evans et al.,2006). This observation may be explained by the finding that another kinesin 2 family member, OSM-3, compensates for kinesin II loss of function. On the other hand, loss of OSM-3 function leads to failure of distal axoneme elongation (Snow et al.,2004; Evans et al.,2006).

Recent work implicates both kinesin II and KIF17, the vertebrate homologue of OSM-3, in assembly of the photoreceptor outer segment (OS), a modified sensory cilium (Horst et al.,1990). For example, the conditional and photoreceptor-specific knockout in mice of a kinesin II subunit, KIF3A, causes an ectopic accumulation of opsin in the inner segment (IS) that leads to photoreceptor degeneration (Marszalek et al.,2000; Jimeno et al.,2006) while a dominant-negative form of the KIF3B subunit expressed during early development in Xenopus rods causes disrupted photoreceptor organization and cell death (Lin-Jones et al.,2003). In contrast, knock-down of KIF17 in zebrafish photoreceptors disrupts or ablates OS formation with little initial effect on the other segments of the cell (Insinna et al.,2008). Although these studies implicate both kinesin 2 family members in OS formation, direct comparison of their relative roles in photoreceptors is not possible because of the use of different strategies in different species.

To directly compare the relative roles of these two kinesin 2 family members in photoreceptors using similar approaches in the same species, we first produced KIF3B morphants for comparison to our previous study of KIF17 morphants (Insinna et al.,2008). This approach was generally unsuccessful because of early developmental anomalies that prevented formation of a photoreceptor layer. We, therefore, used a late onset, cone-specific promoter (Kennedy et al.,2007) to drive expression of two previously described dominant-negative constructs (Lin-Jones et al.,2003; Chu et al.,2006) of KIF3B (DNKIF3B) and KIF17 (DNKIF17) during development of zebrafish cones. Consistent with a previous report, DNKIF3B overexpression resulted in photoreceptor death (Lin-Jones et al.,2003). However, initial signs of disruption with DNKIF3B were within membrane systems of the IS and in the formation of synaptic ribbons at the synaptic terminal. In contrast, overexpression of DNKIF17 led to immediate OS maintenance defects involving disruption of OS discs. This suggests that, while the two motors have overlapping roles in photoreceptor IFT, kinesin II also performs independent functions in photoreceptors distinct from its role in IFT.


Kinesin II Colocalizes and Associates With KIF17 in Both Mice and Zebrafish

Prior immunofluorescence studies established that KIF17, KIF3A, and KIF3B are localized in the synaptic terminal, the IS and along the axoneme of vertebrate photoreceptors (Beech et al.,1996; Muresan et al.,1997,1999; Whitehead et al.,1999; Insinna et al.,2008). Additionally, the distribution of kinesin superfamily proteins (KIFs) was characterized by immunoelectron microscopy in sunfish photoreceptors using anti-LAGSE, an antibody that recognizes the conserved motor domain of KIFs (Beech et al.,1996). We focused our ultrastructural analysis on the localization of the kinesin II subunit, KAP3, and KIF17 in mouse photoreceptors using a pre-embedding electron microscopy (EM) approach (Maerker et al.,2008). Thick sections of fixed mouse retinae were cracked and subsequently incubated with polyclonal or monoclonal antibodies for KIF17 or KAP3 followed by silver enhancement and osmium fixation before final embedment in resin. As expected, labeling with both antibodies was detected along microtubules of the entire axoneme in the OS (Fig. 1A,B,D,E). Labeling was stronger at the base but was still detected in the most distal part of the OS. Of interest, we were unable to detect signal with either antibody in longitudinal sections of the connecting cilium (CC; Fig. 1A,E). This observation is consistent with previous studies of longitudinal sections labeled with the LAGSE antibody (Beech et al.,1996). However, cross-sections of the CC revealed strong labeling for KAP3 (Fig. 1B) and KIF17 (Fig. 1D), showing that the lack of staining in longitudinal sections is likely due to a barrier to access of antibodies (Maerker et al.,2008). Our results show that KIF17 and kinesin II colocalize in structures that are critical for the intersegmental transport of material through the photoreceptor CC. KAP3 and KIF17 were also found at the periciliary region of the IS and in the vicinity of the basal body (Fig. 1A,E), and in the collar-like extension of the IS adjacent to the CC (Fig. 1E, asterisks). Furthermore, KAP3 was localized along the endoplasmic reticulum (ER) and Golgi membranes as well as tubulovesicular elements within the IS (Fig. 1C).

Figure 1.

KIF17 and kinesin II localization and co-immunoprecipitation. A–C: Immuno-electron microscopy (EM) localization of the KAP3 subunit of kinesin II in mouse rod photoreceptor cells. A: Longitudinal section through the connecting cilium and parts of the outer segment (OS). B: Cross-section through the connecting cilium (CC). C: Section through in the endoplasmic reticululm (ER) –Golgi region of the IS. Symbols (><) indicate position of adherens junctions at the outer limiting membrane. D,E: Immuno-EM localization of KIF17. D: Two slightly oblique sections through the CC. E: Two longitudinal images through the photoreceptor at the level of the CC. Arrow points to labeling in the axoneme of the OS. Asterisks indicate labeling in the collar-like extension of the apical IS. F: Co-immunoprecipitation from mouse retinal extracts of either kinesin II or KIF17 with IFT88. IP antibodies are shown at the top. Western blots were done with antibodies for KIF17, IFT88, and the three kinesin II subunits (KAP3, KIF3B, and KIF3A); antibodies are shown on the left. A mixture of IgGs was used as control for nonspecific binding. G: IP experiment as in F from zebrafish retinal extract; the kinesin II antibody (K2.4) recognizes both KIF3A and KIF3B. Scale bars = 400 nm in A, 50 nm in B, 400 nm in C, 50 nm in D (upper), 100 nm in D (lower), 400 nm in E.

Kinesin II is known to exist in a complex with IFT proteins isolated by immunoprecipitation (Cole et al.,1998; Baker et al.,2003; Qin et al.,2004), and we have recently demonstrated that KIF17 is associated with IFT complexes as well (Insinna et al.,2008). This suggests the existence of IFT complexes associated with both motors. We, therefore, evaluated the possible co-immunoprecipitation (IP) of KIF17 and the three kinesin II subunits from mouse retinal extract (Fig. 1F). Immunoprecipitates were immunoblotted for the three individual subunits of kinesin II (Fig. 1F, left) as well as IFT88 and KIF17 (Fig. 1F, right). KIF3B and KIF3A were found to migrate at ∼ 95 kDa and ∼ 85 kDa, respectively. The KIF17 antibody co-precipitated a subpopulation of kinesin II and, reciprocally, the kinesin II antibody co-precipitated a population of KIF17. The weak intensity of the Western blot for KIF3A in the KIF17 IP compared with the kin II IP (Fig. 1F) was consistent in our experiments. This may be explained by the previous observation that KIF3A exists independent of KIF3B in a complex with synaptic ribbons in photoreceptors (Muresan et al.,1999). Because the kin II IP antibody recognizes both KIF3A and KIF3B, it would likely bring down both ribbon bound KIF3A as well as KIF3A in the heterotrimeric complex. While both the KIF17 and kinesin II antibodies immunoprecipitated roughly equal amounts of IFT88, the relative abundance of KIF3B and KIF17 in the two IPs was different. This suggests that multiple IFT complexes containing either KIF17 or kinesin II are likely to be present in mouse retinae; it is also likely that some complexes contain both motors.

A similar IP was conducted using zebrafish retina (Fig. 1G). However, the individual mouse kinesin II subunit antibodies did not cross-react with the zebrafish proteins. Therefore, we used the general kinesin II antibody K2.4 (Cole et al.,1993) that recognizes both KIF3A and KIF3B in the Western blotting. In the kinesin II IP, two bands at ∼ 80/90 kDa correspond to KIF3A and KIF3B (Fig. 1G). The antibody to KIF17 co-precipitated a small but consistent amount of kinesin II from zebrafish retinae (Fig. 1G). Interestingly, in zebrafish tissue, the anti-KIF17 antibody pulled down mainly KIF3A and a smaller amount of KIF3B. However, this could reflect differences in affinity of the K2.4 antibody for the two subunits in Western blots rather than kinesin complexes deficient in KIF3B. Our previous work in zebrafish indicates that the complex immunoprecipitated with the KIF17 antibody also includes IFT proteins (Insinna et al.,2008).

Knock-down of KIF3B Causes Early Embryonic Defects

To knock-down KIF3B expression in developing zebrafish embryos for direct comparison to our prior work on KIF17 knockdown (Insinna et al.,2008), we designed three antisense morpholino oligonucleotides: a translation-blocking morpholino (AKIF3B), a morpholino directed at the exon 2 splice donor site (SpKif3B), and a control morpholino containing the sense sequence of kif3b mRNA. The majority of morphants (∼85%) injected at the one-cell stage with a high dose (500 μM) of translation-blocking or splice-blocking morpholino died at 24 hr after fertilization, whereas controls were normal. Lowering the dose to 250 μM improved the survival rate for both morpholinos to approximately 50%. The translation-blocking morpholino caused a dose-dependent decrease in KIF3B expression (Fig. 2B). The depletion of KIF3B also reduced KIF3A, but KIF17 expression was not affected by the depletion of kinesin II (Fig. 2B). A similar depletion of KIF3B was seen in morphants injected with the splice-blocking morpholino (Fig. 2C), but the parallel reduction in KIF3A was not as dramatic.

Figure 2.

Disruption of kif3B expression by an antisense morpholino oligonucleotide. A: (Upper panel) Images of 3-day-old larvae that survived an injection of 500 μM (AKIF3B 500 μM) or 250 μM (AKIF3B 250 μM) of the translation-blocking morpholino. (Middle panel) A control injected embryo. (Lower panel) Three-day-old larva after injection with 250 μM of a splice-blocking morpholino (SpKIF3B). B: Western blots of whole embryo extracts showing the reduction of KIF3B and KIF3A in AKIF3B morphants at 3 days postfertilization (dpf); anti-KIF17 and anti–γ-tubulin were used as controls. C: Western blot of three separate whole SpKif3B (250 μM) morphant and control embryo extracts showing similar depletion of KIF3B and KIF3A as observed in translation-blocking morphants (B above); anti–γ-tubulin was used as a loading control. D: Semi-thin section of the eye of an AKIF3B (250 μM) morphant at 3 dpf. Insets: Electron microscopy (EM) views of AKIF3B (250 μM) photoreceptors. E: Semi-thin section of the eye of an AKIF3B (500 μM) morphant at 3 dpf. Note the absence of retinal lamination and photoreceptor differentiation. Scale bar = 10 μm in D,E, 2.5 μm in insets.

Although most KIF3B morphants died, those that survived through 3 days postfertilization (dpf) had multiple abnormalities, including heart edema, small eyes, hydrocephalus, and strong body curvature (Fig. 2A, AKIF3B 500 μM). Semi-thin sections of their retinae showed a strong developmental delay phenotype compared with controls with defects in lamination and optic stalk formation and a lack of photoreceptor differentiation (Fig. 2E). The absence of photoreceptors with OS precluded analysis of OS formation in these embryos. We, therefore, studied photoreceptors in embryos injected with either AKIF3B or SpKIF3B at 250 μM (Fig. 2D). Although a greater number of embryos survived with less severe developmental anomalies, light microscopy (Fig. 2D) and ultrastructural analysis (Fig. 2D insets) of their retinae showed that OS formation was indistinguishable from controls. This result, however, does not necessarily suggest a lack of a requirement of kinesin II because at low morpholino concentration sufficient kinesin II may have been present in photoreceptors.

Our results with antisense morpholinos are consistent with conventional gene knockouts in mice for both KIF3B and KIF3A, which showed a requirement in early development before retinal differentiation and photoreceptor formation (Nonaka, et al.,1998; Marszalek et al.,1999). Although toxic effects of the morpholinos cannot be ruled out, the similar effects of both a splice blocking and translation blocking morpholino suggests that the effects were specific. Furthermore, none of these early developmental defects were observed in KIF17 morphants (Insinna et al.,2008). Nonetheless, with the delay in retinal development and lack of photoreceptors in KIF3B morphants, direct comparison of the effects of morpholino knockdown of the two motors on photoreceptor OS formation was not possible.

DNKIF3B in Zebrafish Cones Causes Acute IS and Synaptic Defects

As an alternative strategy for comparison of the roles of kinesin II and KIF17 in OS maintenance, we directed the expression of dominant-negative KIF3B (DNKIF3B) and dominant-negative KIF17 (DNKIF17) to cone photoreceptors using a cone transducin alpha (Ta-CP) promoter, which is known to induce transgene expression in greater than 50% of the cones at 3–5 dpf (Kennedy et al.,2007). The DNKIF3B and DNKIF17 constructs were designed so that the motor domain of each protein was replaced by green fluorescent protein (GFP; Le Bot et al.,1998; Lin-Jones et al.,2003; Chu et al.,2006). Although dimerization behavior, which depends on the coil-coiled stalk is maintained, these proteins are nonfunctional. Prior work using this DNKIF3B construct showed that the nonfunctional KIF3B protein is still able to associate with endogenous kinesin II subunits, which results in the inactivation of the whole complex (Le Bot et al.,1998).

We injected embryos with the DNKIF3B or DNKIF17 constructs at the one-cell stage, screened for GFP fluorescence in the eye at 3 dpf and examined retinae by light microscopy and EM at 5 dpf. Consistent with prior work (Kennedy et al.,2007), fluorescence from GFP alone (control) at 5 dpf was detected throughout the eye with areas representing 15–40% of the cones expressing GFP at high levels (Fig. 3A); a similar distribution pattern for both of the GFP-tagged dominant-negative proteins was observed (Fig. 3B,C). At this stage, retinae are composed of ∼ 90% cones with OSs approximately half of their adult length; cone OSs are shorter and less developed at the retinal periphery. Our results (Fig. 3) suggest that the Ta-CP promoter drives protein expression efficiently in the great majority of cones. Rods constitute only approximately 10% of the total number of photoreceptors and are present mainly at the ventral periphery (Branchek and Bremiller,1984; Kljavin,1987).

Figure 3.

Localization of Ta-CP–directed expression of green fluorescent protein (GFP), DNKIF3B, and DNKIF17 and cone opsin localization at 5 days postfertilization (dpf). The Ta-CP promoter up stream of GFP, DNKIF3B, or DNKIF17 was injected at the one-cell stage. A: In Ta-CP/GFP embryos, GFP fluorescence was seen throughout the outer retina and cone opsin was highly localized to the outer segment (OS). B: In Ta-CP/DNKIF3B embryos, transgene expression (GFP, green) was similar to controls, but cone opsin was mislocalized to the perinuclear region and outer plexiform layer (arrow) across the entire retina. Inset: Higher power image showing cone opsin mislocalization to the synaptic layer (arrow). C: In Ta-CP/DNKIF17 embryos transgene expression (GFP, green) was distributed across the entire retina, but was largely restricted to the inner segment (IS; see inset). Cone opsin was in the OS, and there was no evidence of mislocalization. Inset: Higher power image showing lack of mislocalization of cone opsin and strong accumulation of DNKIF17 in the IS. Scale bars = 10 μm. The bar in A applies to panels A–C.

Because cones in the central retina are older and are at more advanced stage of OS formation than those at the periphery at the time of onset of transgene expression, the dominant-negative phenotypes are considered separately for the central and peripheral retina (Figs. 4, 5). In embryos injected with DNKIF3B, we found that 15–30% of the cones were in an apoptotic state as evidenced by both nuclear and cytoplasmic compaction. These dead or dying cones were seen at the retinal center and to a lesser extent at the ventral periphery, which contains the most recently developed cones (Figs. 4A,D, 5B). In contrast, control retinae from embryos injected with the Ta-CP promoter construct driving GFP were devoid of apoptotic cells (data not shown). This suggests that the degeneration and death of developing cones is the result of accumulation of the dominant-negative protein.

Figure 4.

Light microscopy of retinae of control, DNKIF3B, and DNKIF17 embryos at 5 days postfertilization (dpf). A,B: Low power images of semi-thin plastic sections of DNKIF3B (A) and DNKIF17 (B) eyes. Central (center) and peripheral (periph) retinal regions are labeled. C: Higher power image of the outer retina of a control eye; the periphery is on the right (Periph). Cone and rod outer segment (OS) are indicated by arrows. D,E: Higher power images of outer retina of DNKIF3B (D) and DNKIF17 (E) eyes; the periphery is on the right (periph). White arrows in D indicate condensed nuclei and cell bodies of dying cones. White arrows in E indicate cones with very short or missing OS. A normal rod OS is indicated in E. Scale bars = 10 μm. The bar in E applies to C–E.

Figure 5.

Summary of the structural phenotypes observed in DNKIF3B and DNKIF17 embryos. A: In wild-type retinae, cones of the central retina are more developed with longer outer segment (OS) than those at the periphery. CC, connecting cilium; CP, calycal process. B: In embryos expressing DNKIF3B, many cells in the central and peripheral retina exhibit a condensed morphology suggesting an apoptotic state. In the periphery, many uncondensed cones also exhibit inner segment (IS) vesicles and accumulation of dense material while retaining normal OS structure. C: In embryos expressing DNKIF17, cones in the central retina with normal IS exhibit accumulations of vesicles in the proximal OS opposite the cilium, but this phenotype is not seen in all cells. In the periphery, OS do not elongate and are often vesiculated. Cones do not exhibit features associated with excess apoptosis.

Ultrastructural defects due to DNKIF3B expression, largely restricted to the IS, ranged from mild to severe. Mild IS defects generally involved enlargement of the cytoplasmic domain in the mitochondria-rich ellipsoid region (Fig. 6B–D). This was mainly seen at the retinal periphery in cells that maintained normal structure and length of the OS (Figs. 5B, 6B–D). In more severe cases, nuclear and inner segment compaction was seen in cells with normal organization of OS discs (Fig. 6E,F). Finally, accumulation of large vesicles at in the ellipsoid region of the inner segment accompanied by enlargement of the apical IS region was seen in ∼ 10% of the cones (Figs. 5B, 6G–I). In contrast, controls had a tightly compacted IS at the level of the ellipsoid and large vacuoles and compacted cells were not seen (Figs. 5A, 6A). In general, OS maintained a relatively well-organized structure with normal disc stacking in cones showing severe disorganization of the inner segment.

Figure 6.

Accumulation of large vacuoles and dense material in the inner segment (IS) of cones expressing DNKIF3B. A: Electron microscopy (EM) view of cones at 5 days after injection of the Ta-CP driving green fluorescent protein (GFP) alone (control). Note the normal structure of outer segment (OS) discs and IS with accumulation of mitochondria (m) in the ellipsoid region. B,C: EM view of cones at 5 days after injection of the Ta-CP driving DNKIF3B showing mild IS defects. This was primarily enlargement of the cytoplasmic area around the mitochondria. Arrow indicates large vesicle below the connecting cilium (CC). n, photoreceptor nucleus; g, Golgi apparatus. D: EM view of a cross-section through the IS of a cone with a mild IS phenotype involving enlargement of the cytoplasmic area around mitochondria; er, endoplasmic reticulum. The upper arrow indicates accumulation of large vesicles in the ellipsoid region. The lower arrow indicates a Golgi region. E: EM view of a cone in the central retina that has a highly condensed IS between the nucleus (n) and mitochondrial rich (m) ellipsoid region. OS discs are normal. F: Cone with a highly condensed nucleus and IS with normal OS organization. G,H: Examples of cones accumulating vesicles and large vacuoles (arrows) within the IS surrounding the mitochondria. Asterisk in H indicates condensed nucleus of an adjacent cone. Scale bar = 1.5 μm in C, 2.5 μm in F.

To determine whether cone opsin was mislocalized in cells expressing DNKIF3B protein cryostat sections were labeled with anti-cone opsin antibodies and examined for both GFP and opsin fluorescence. In retinae of embryos injected with the Ta-CP/GFP construct, opsin was highly expressed in the cone OS but was not detectible in the inner segment or cone terminals in the outer plexiform layer (Fig. 3A). In contrast, in retinae expressing DNKIF3B, cone opsin was mislocalized and was abundant in cone terminals in the outer plexiform layer across the entire retina (Fig. 3B).

To complete our ultrastructural analysis, we also examined cone synaptic pedicles in control compared with DNKIF17- and DNKIF3B-expressing embryos (Fig. 7). Both control (Fig. 7A–C) and DNKIF17-expressing (Fig. 7D–F) embryos had well-developed cone pedicles with postsynaptic invaginations and synaptic ribbons associated with presynaptic membranes (Fig. 7C,F, insets). However, synaptic ribbons were rarely seen in pedicles of DNKIF3B-expressing embryos (Fig. 7G–I). Instead, pedicles with postsynaptic invaginations (Fig. 7I, inset), but lacking ribbons were seen throughout the retina. In the rare instances when ribbons were detected, they were not associated with the presynaptic membrane (Fig. 7I). These findings suggest that KIF3B plays an additional role in synaptic morphogenesis, consistent with prior work showing that it functions in axonal transport in addition to IFT (Muresan,2000).

Figure 7.

Lack of synaptic ribbons in cone pedicles of DNKIF3B-expressing embryos. AC: Control pedicles showing postsynaptic invaginations (large arrows) and synaptic ribbons (small arrows) associated with presynaptic membranes. DF: DNKIF17 cone pedicles showing postsynaptic invaginations (large arrows) and synaptic ribbons (small arrows) associated with presynaptic membranes. GI: DNKIF3B cone pedicles showing postsynaptic invaginations (large arrows) without synaptic ribbons. Occasionally, ribbons are seen dissociated from presynaptic membranes (small arrow in I). Condensed nuclei and pedicles (asterisks) of dying cells are seen in G. Insets in C, F, and I are enlargements of invaginations to show association of ribbons with the presynaptic membrane. Scale bars = 500 nm in I and in inset.

DNKIF17 in Zebrafish Cones Disrupts OS Elongation and Early Disc Morphogenesis

In contrast to our results with DNKIF3B, EM analysis of cone photoreceptors expressing DNKIF17 (Chu et al.,2006) at 5 dpf revealed little if any evidence of cell death in the photoreceptor layer. However, the late developing cones at the periphery failed to elongate their OS compared with control embryos (Figs. 4E, 5C, 8A–D). Furthermore, in developing cells with short OS, disc edges on the opposite side from the CC were abnormal and vesicle-like structures accumulated in the OS (Fig. 8A–C). This phenotype was also observed to a lesser extent in OS localized at the central retina. In the retinal center, the OS of DNKIF17-expressing retinae were normal in length but many accumulated membrane vesicles, particularly at the OS base (Fig. 8E). This area of the OS is known as the disc-forming region, and our results suggest that KIF17 is involved in the transport of elements essential for disc morphogenesis. In our analysis of DNKIF3B and DNKIF17, the rod photoreceptors, which did not express the dominant-negative protein, had normal structure (Fig. 8F).

Figure 8.

Disruption of the outer segment (OS) structure in cones expressing DNKIF17. AD: Electron microscopy (EM) views of cones at the retinal periphery at 5 days after injection of Ta-CP driving DNKIF17. Note disrupted OS structure with failure to complete disc edges as shown by arrows. E: EM view of a cone in the central retina accumulating vesicular membranes within the base of the OS (arrows) opposite the cilium (CC). F: Normal rod photoreceptor. Scale bars = 2.5 μm in C (applies to A–C).

In contrast to cones expressing the DNKIF3B construct (Fig. 3B), little if any cone opsin mislocalization was detected in cells expressing DNKIF17 (Fig. 3C). This suggests that, despite anomalous disc morphogenesis within the OS, cone opsin is still delivered to the OS segment normally. Finally, as mentioned above the development of cone synaptic terminals was normal (Fig. 7D–F) in photoreceptors exhibiting disrupted OS formation.


Our immuno-EM analysis of the anti-KAP3 and anti-KIF17 labeling in mouse photoreceptors confirms previous studies conducted in fish photoreceptors using anti-KIF3A and anti-KIF17 antibodies (Beech et al.,1996; Insinna et al.,2008). Indeed, both kinesin II and KIF17 are associated with microtubule doublets of the CC and axonemal microtubules that extend distally into the OS. In addition, both kinesins were found in the periciliary compartment of the IS, the basal body region and the collar-like IS extension, which is thought to be the pathway for the intersegmental transport and the site for IFT complex assembly (Tam et al.,2000; Deane et al.,2001; Maerker et al.,2008). Furthermore, our IP data suggest the existence of protein complexes containing the two kinesins and IFT proteins in both mouse and zebrafish photoreceptors (Baker et al.,2003; Insinna et al.,2008). Additionally, our finding that anti-KIF17 and kinesin II antibodies co-precipitated similar amounts of IFT88, but differentially co-precipitated kinesin II or KIF17 suggests the existence of IFT complexes containing either KIF17 or kinesin II. Indirect evidence for such complexes has been found in C. elegans AWB cilia where the KIF17 homologue OSM-3 has been suggested to be associated with an IFT complex independent of kinesin II (Mukhopadhyay et al.,2007).

Combined, our localization and IP data are consistent with both cooperative and independent functions of the two kinesins within the OS. In C. elegans channel cilia, it has been proposed that kinesin II and OSM-3 behave cooperatively to build the middle (proximal) segment of the axoneme, but that OSM-3 alone is required for the distal segment extension (Snow et al.,2004; Evans et al.,2006; Pan et al.,2006). Although a similar dichotomy of function is possible along the axoneme of zebrafish photoreceptors, the situation is more complicated, because both KIF17 morphants and those expressing DNKIF17 exhibit severe OS ablation or a strong proximal OS defect that is not seen in DNKIF3B-expressing cells. This suggests that, in photoreceptors, KIF17 plays a nonredundant role in proximal OS assembly, possibly by transporting cargo essential for this process.

An important finding in this analysis is cone opsin mislocalization in cells expressing DNKIF3B, but not in those expressing DNKIF17 (see Fig. 3). The fact that opsin localizes to the OS of cones exhibiting obvious disc morphogenesis defects due to DNKIF17 suggests that KIF17 is important for disc morphogenesis, but is not essential for high fidelity targeting of cone opsin. Although KIF17 may be essential in the process of disc assembly or for the delivery of cargo necessary for disc assembly, these findings suggest that the cargo may not be cone opsin. Instead, our results are consistent with previous data suggesting that kinesin II is necessary for proper trafficking of opsin into the OS (Marszalek et al.,2000; Lin-Jones et al.,2003; Jimeno et al.,2006). A potential inconsistency is our previous finding that cone opsin was mislocalized in cones of KIF17 morphants. This discrepancy may be related to the fact that OS formation was severely ablated in KIF17 morphants at a time when opsin synthesis persisted, while DNKIF17 resulted in a milder phenotype that involved disorganization of discs.

Our results with morpholino knockdown of KIF3B are consistent with a role for this motor in early embryonic development (Nonaka, et al.,1998; Marszalek et al.,1999). Kinesin II is a canonical IFT motor required for ciliogenesis, and cilia have been shown to contribute to the transport of signals essential for early embryonic development (Caspary et al.,2007; Haycraft et al.,2007; Rohatgi et al.,2007; Corbit et al.,2008; Nielsen et al.,2008). Many of the early defects may have resulted from failed kinesin II–dependent ciliogenesis or other cellular functions of kinesin II. Although we cannot rigorously rule out toxic effects, the two KIF3B morpholinos, developed here are potentially useful tools for analysis of those early developmental defects.

Our use of targeted expression of DNKIF3B and DNKIF17 in 5-day-old zebrafish cone photoreceptors allowed us to analyze the immediate effects of DN expression in cones that had already begun to develop their OS. This analysis was also facilitated by separate analysis of the early developing OS of the central retina with the less mature, late developing OS in the peripheral retina (Fig. 5). Surprisingly, cones expressing DNKIF3B started to die within 48 hr after fluorescence was initially detected at 3 dpf, whereas cones expressing DNKIF17 exhibited little if any photoreceptor cell death. Many DNKIF3B-expressing photoreceptor IS accumulated large vacuoles and dense material while retaining normal OS structure. This along with the strong mislocalization of cone opsin suggests that trafficking defects within the IS contribute to the photoreceptor degeneration seen with DNKIF3B. This interpretation is also supported by previous studies showing that depletion of kinesin II in mouse photoreceptors leads to ectopic accumulation of opsin and membranous material in the IS (Marszalek et al.,2000).

The inner segment defects described above could reflect a role for kinesin II in ER–Golgi trafficking as reported previously (Le Bot et al.,1998). Likewise, the lack of development of synaptic ribbons in cone pedicles could reflect an additional role for kinesin II in axonal transport and synaptic development. Although cone pedicles with invaginations from adjacent horizontal and bipolar cells were present in DNKIF3B-expressing embryos, synaptic ribbons were rarely seen. Impaired formation of ribbon synapses is consistent with prior work (Muresan,2000) indicating a role for kinesin II in axonal transport, and raises the possibility that the kinesin II motor transports cargo to the pedicle. It has also been reported that KIF3A is a component of photoreceptor synaptic ribbons (Muresan et al.,1999). Because overexpressed DNKIF3B would be expected bind KIF3A to form nonfunctional complexes, DNKIF3B could also directly affect ribbon formation.

In a previous study in Xenopus (Lin-Jones et al.,2003), the rod opsin promoter was used to drive expression of DNKIF3B at a stage before rod OS formation. This resulted in a dramatic phenotype in which only a very restricted number of rods survived in tadpoles expressing the transgene. This prevented an extensive analysis of defects occurring before cell death. In contrast, our characterization of early ultrastructural defects was made possible by the use of the late onset cone transducin promoter (Kennedy et al.,2007). Although the late onset of transgene expression may explain our success in detecting early IS anomalies, we cannot rule out a somewhat different role for kinesin II in rods (Lin-Jones et al.,2003) compared with cones (this study).

Compared with DNKIF3B, late developing cones at the periphery expressing DNKIF17 had shorter OS (Fig. 4C). This suggests that KIF17 is required for OS elongation. Furthermore, in the central retina, cones accumulate vesicles at the base of the OS but did not show IS defects. The failure to assemble discs at the base of the OS suggests that one or more components essential for disc assembly requires KIF17. Presumably, those components were still transported normally in DNKIF3B-expressing photoreceptors with early signs of degeneration. Reciprocally, the disruption of KIF17 trafficking did not seem to affect the targeting of cone opsin to the OS (Fig. 3).

Our data suggesting different roles for kinesin II and KIF17 in OS assembly may be related to earlier studies suggesting independent sorting pathways for rhodopsin and Peripherin-2/Rom-1, two OS membrane proteins essential for disc formation (Fariss et al.,1997; Lee et al.,2006). Further studies should determine whether peripherin-2 and ROM-1 are potential IFT cargo transported by KIF17 because defective transport of these disc rim components could account for the proximal accumulation of OS vesicles observed in our study (Fig. 7C). Consistent with this idea, a plausible explanation for the difference in DNKIF17 and DNKIF3B phenotypes could be that kinesin II and KIF17 take different trafficking routes in the IS or are involved in transport of different cargo.



The anti-kinesin II antibodies include a monoclonal antibody K2.4 (Cole et al.,1993) obtained from Covance (Richmond, CA) and three monoclonal anti-KIF3A, anti-KIF3B, and anti-KAP3a antibodies purchased from BD Biosciences (San Jose, CA). Polyclonal anti-KIF17 was obtained from Abcam (Cambridge, MA). Anti-cone opsin antibodies were gifts from Dr. T.S. Vihtelic at the University of Notre Dame (Vihtelic et al.,1999).


Immunocytochemistry was performed as described previously (Insinna et al.,2008) in 5-day-old embryos expressing Ta-CP/GFP, Ta-CP/DNKIF3B, and Ta-CP/DNKIF17 constructs. Double-label confocal projections of GFP (green) and cone opsin (red) were acquired using a Leica TCS-SP2 confocal system.

Conventional Electron Microscopy

Zebrafish embryos (N = 6 for each treatment group) fixed in 2% paraformaldehyde and 2% glutaraldehyde were prepared for conventional EM in the Medical College of Wisconsin core Electron Microscopy Facility as described previously (Pazour et al.,2002). For light microscopy, retinae (N = 6 for each treatment) of morphants or transgene-expressing embryos were cut in 1-μm sections rostrocaudally until the central retina was reached. Sections were stained with toluidine blue and examined under an epifluorescence microscope (Nikon, TE3000). Thin sections prepared from these blocks were stained with lead citrate and uranyl acetate and viewed using a Hitachi 600 or JEOL JEM 2100 electron microscope.


For immuno-EM we applied a recently introduced pre-embedding labeling technique (Maerker et al.,2008). Eyes of mature C57BL/6J mice were fixed in 4% buffered paraformaldehyde, infiltrated with 30% buffered sucrose and cracked by cycles of freezing in liquid nitrogen and thawing at 37°C. Fifty-micrometer-thick Vibratome sections were incubated with primary antibodies and biotinylated secondary antibodies (Vector Laboratories), and visualized by a Vectastain ABC-Kit (Vector Laboratories). After silver enhancement, post-fixation, and dehydration, specimens were embedded in araldite. Ultrathin sections were analyzed in a Tecnai 12 BioTwin transmission electron microscope (FEI, Eindhoven, The Netherlands).

Immunoprecipitations and Western Blot

Retinae from C57Bl/6 adult mice or dark-adapted adult zebrafish were dissected, homogenized, and sonicated in IP lysis buffer (1% NP-40, 250 mM NaCl, 20 mM Tris HCl pH 6.8, 5% glycerol, 0.02% sodium azide) with a protease inhibitor cocktail (Roche, Mannheim, Germany). Extracts were centrifuged at 20,000 × g for 20 min at 4 °C. Antibodies were incubated on a rotator at 4°C for 45 min with 50 μl of protein G-Sepharose 4 Fast Flow (Amersham Biosciences) and phosphate buffered saline (PBS). Beads were washed with PBS, extract was added to each bead/antibody mixture and incubated for 2 hr at 4°C. Beads were then washed three times in IP buffer, and proteins were eluted from the beads by boiling in 2× Laemmli buffer. Western blotting was performed as described previously (Insinna et al.,2008).

Morpholino Design

The zebrafish kif3b mRNA was predicted in sequence (ZFIN database) based on protein homology to the mouse KIF3B (NP_032470) using tblastp. The translation-blocking, splice-blocking, and control morpholino oligonucleotides used to knock-down zebrafish KIF3B were ordered from Gene Tools (Corvallis, OR). Sequences of the KIF3B morpholinos were as follows: aKIF3B: 5′-GTTGGCCTGCTGGATTCTCCGAATG-3′; spKIF3B: 5′-ATGGCCTGAGAGTA-ATGATAGGGTT-3′; control spKIF3B: 5′-AACCCTATCATTACTCTCTCAG-GCCAT-3′. Morpholinos were injected using a nanoliter 2000 microinjector (World Precision Instruments, Inc) in one- to two-cell stage wild-type TuAB embryos, at varying concentrations with an injection volume of 4.6 nl/embryo. Injected embryos were kept in fish water containing 0.003% 1-phenyl-2-thiourea (PTU) to inhibit development of pigmentation. Western blotting was performed to verify the level of KIF3B protein knock-down.

DNKIF3B and DNKIF17 Constructs

All clones were created by site-specific recombinational cloning using Invitrogen's MultiSite Gateway Technology. A polymerase chain reaction (PCR) fragment containing GFP and the Xenopus KIF3B stalk and tail (Le Bot et al.,1998; Lin-Jones et al.,2003) and site-specific attB recombination sites was used in the BP recombination reaction to create the DNKIF3B entry clone. A PCR fragment containing a mouse dominant-negative KIF17 transgene with the motor domain replaced by GFP (Chu et al.,2006) and site-specific attB recombination sites was used in the BP recombination reaction to create the DNKIF17 entry clone. A PCR fragment containing the zebrafish cone transducin α promoter (Ta-CP) (Kennedy et al.,2007) and site-specific attB recombination sites was used in the BP recombination reaction to create the Ta-CP entry clone. The Ta-CP and DNKIF3B entry clones were used in the LR recombination reaction with pcDNA 6.2/V5-pL-DEST vector (Invitrogen, Carlsbad, CA) to create the Ta-CP-DNKIF3B expression clone used in the experiments. The Ta-CP and DNKIF17 entry clones were used in the LR recombination reaction with pcDNA 6.2/V5-pL-DEST vector to create the TaCP-DNKIF17 expression clones used in the experiments.


We thank Dr. Don Arnold for the DNKIF17 (USC, Los Angeles, CA), Dr. Beth Burnside (UC Berkeley, CA) for the DNKIF3B construct, and Dr. Breandan Kennedy (UCD, Ireland) for the Ta-CP promoter. We also thank Dr. Brian Link for access to the Medical College of Wisconsin zebrafish facility, and Clive Wells for assistance with EM imaging. J.C.B. was funded by a NIH grant and a NIH core grant for Vision Research, and Medical College of Wisconsin Research Development Funds. C.I. was supported by NIH NRSA Training grant.