Generation of Rab-based transgenic lines for in vivo studies of endosome biology in zebrafish


  • Brian S. Clark,

    1. Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, Wisconsin
    Search for more papers by this author
  • Mark Winter,

    1. Department of Electrical Engineering and Computer Science, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin
    Search for more papers by this author
  • Andrew R. Cohen,

    Corresponding author
    1. Department of Electrical Engineering and Computer Science, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin
    • Andrew R. Cohen, Department of Electrical Engineering and Computer Science, University of Wisconsin-Milwaukee, Milwaukee, WI 53211

      Brian A. Link, Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, WI 53226

    Search for more papers by this author
  • Brian A. Link

    Corresponding author
    1. Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, Wisconsin
    • Andrew R. Cohen, Department of Electrical Engineering and Computer Science, University of Wisconsin-Milwaukee, Milwaukee, WI 53211

      Brian A. Link, Department of Cell Biology, Neurobiology and Anatomy, Medical College of Wisconsin, Milwaukee, WI 53226

    Search for more papers by this author


The Rab family of small GTPases function as molecular switches regulating membrane and protein trafficking. Individual Rab isoforms define and are required for specific endosomal compartments. To facilitate in vivo investigation of specific Rab proteins, and endosome biology in general, we have generated transgenic zebrafish lines to mark and manipulate Rab proteins. We also developed software to track and quantify endosome dynamics within time-lapse movies. The established transgenic lines ubiquitously express EGFP fusions of Rab5c (early endosomes), Rab11a (recycling endosomes), and Rab7 (late endosomes) to study localization and dynamics during development. Additionally, we generated UAS-based transgenic lines expressing constitutive active (CA) and dominant-negative (DN) versions for each of these Rab proteins. Predicted localization and functional consequences for each line were verified through a variety of assays, including lipophilic dye uptake and Crumbs2a localization. In summary, we have established a toolset for in vivo analyses of endosome dynamics and functions. Developmental Dynamics 240:2452–2465, 2011. © 2011 Wiley Periodicals, Inc.


Identified as orthologs of the yeast SEC4 and YPT1 proteins (Touchot et al., 1987), the Rab-family (Ras-like from rat brain) of small GTPase proteins are essential regulators of protein and membrane trafficking (Mitra et al., 2011). The SEC/YPT/RAB proteins are conserved from yeast to humans, and, to date, more than 65 Rab genes have been identified in vertebrates with distinct membrane localizations and functions (Zhang et al., 2007). Acting as molecular switches based on oscillations between GTP- and GDP-bound forms, the Rab proteins facilitate communication between membrane-compartmentalized structures in eukaryotic cells. This communication is mediated through interactions between specific Rabs and their effector proteins that together direct precise formation, motility, and tethering of vesicles in and between membrane-bound organelles. By regulating protein internalization, sorting, trafficking, and recycling through endocytic compartments, Rab proteins modulate a multitude of cellular processes including cytoskeleton organization, cellular polarity, and receptor-mediated signaling (Seachrist and Ferguson, 2003; Gould and Lippincott-Schwartz, 2009; Nishimura and Sasaki, 2009; Sorkin and von Zastrow, 2009; Stenmark, 2009). Through studies in a variety of organisms, a high degree of functional conservation for specific Rab protein sub-types across species has been demonstrated. For example, lethality as a result of loss of YPT1, the yeast RAB1 ortholog, was rescued through expression of mouse RAB1a (Haubruck et al., 1989). In this study, we focus on zebrafish Rab5c, Rab11a, and Rab7. These Rab proteins are known to be associated with early, recycling, and late endosomes, respectively, for all organisms in which they have been studied.

RAB5 was first identified from human pheochromocytoma cDNA in a screen for genes with high homology to yeast sec4 DNA (Zahraoui et al., 1989). Initial experiments analyzing RAB5 localization found that it associated with membranes and vesicular structures throughout the cytoplasm. Overexpression of RAB5 led to large vesicle structures that were decorated by RAB5 (Chavrier et al., 1990). Immunoelectron microscopy revealed that RAB5 localized to early endosomes, the cytoplasmic side of plasma membrane patches, and on coated pits. Together these data indicated RAB5 functions in vesicle transport and fusion of membrane/cargo from clathrin-coated vesicles of the plasma membrane to the newly formed early endosomes (Chavrier et al., 1990; Gorvel et al., 1991). RAB5 and early endosomes in general have been implicated in regulating a variety of receptor-mediated signaling pathways. For example, in zebrafish, Rab5c has been connected to the regulation of both the FGF and Wnt-signaling pathways during early embryogenesis, by confining the range of FGF morphogen activity (Scholpp and Brand, 2004; Yu et al., 2009; Nowak et al., 2011) and by mediating Wnt11-dependent endocytosis of E-cadherin (Ulrich et al., 2005; Tay et al., 2010).

RAB11 was first identified from MDCK cells in a screen for transcripts homologous to yeast YPT1/SEC4 (Chavrier et al., 1990). The protein was subsequently purified from rat liver and identified as the 11th member of the Rab family of proteins (Sakurada et al., 1991). Immunofluorescent microscopy showed that RAB11 protein localized to the pericentriolar-recycling compartment, later named the recycling endosome. For example, RAB11 co-localized with the transferrin receptor bound for re-expression at the membrane surface (Ullrich et al., 1996). More recently, RAB11 has not only been implicated in regulating signaling through a variety of mechanisms, but also found to maintain apical adherens junctions (Roeth et al., 2009) and in primary cilia biogenesis (Knodler et al., 2010; Westlake et al., 2011).

RAB7 was originally isolated from a Buffalo rat liver cell line and named BRL-ras due to its significant homology to ras and ras-related genes (Bucci et al., 1988). RAB7 was later identified as a member of the Rab family of proteins (Chavrier et al., 1990). Immuno-electron microscopy and marker colocalization studies determined that Rab7 localized to vesicular structures identified as late endosomes (Chavrier et al., 1990). Subsequent studies have defined a general role for RAB7 in trafficking vesicles with cargo destined for degradation as well as in biogenesis of lysosomes and lyososome-related organelles (Feng et al., 1995). More recently, disruption of RAB7 has been demonstrated in various adult-onset diseases including neuropathy, cancer, and lipid metabolism syndromes (Zhang et al., 2009).

In Drosophila, inducible transgenic lines expressing N-terminal fluorescent fusions of wild-type (WT), dominant-negative (DN), and constitutively active (CA) versions for nearly all Rab proteins from that species have been generated to investigate the function of Rab proteins in vivo (Zhang et al., 2007). To date, only limited resources exist for studying Rab-based endosome biology in vertebrate animals (Tolmachova et al., 2004). Most experiments currently depend on analysis in cell culture or through transient expression studies in vivo. Developing zebrafish larvae, however, provide an excellent model system for studying cell biological processes in vivo and dynamically due to the fast ex utero development, transparency, and ease of obtaining embryos. The combinations of advances in gene disruption and transgenic techniques have further enabled sophisticated studies of gene function within zebrafish (Kawakami, 2004, 2005; Kwan et al., 2007).

Here we report the generation and validation of transgenic lines expressing N-terminal fluorescent fusions of proteins marking Rab5c early endosomes, Rab11a recycling endosomes, and Rab7 late endosomes. In addition, we have established and verified conditionally inducible dominant-negative (DN) and constitutively active (CA) Rab versions for use with the GAL4/UAS system. We have also created an automated tracking program enabling rapid quantization of sub-cellular organelle features and dynamics. Overall, this transgenic and computational toolset provides opportunities for studies of dynamic, in vivo endosome biology during development and in the context of various normal and pathological cellular processes.


Generation of EGFP-Rab5c, EGFP-Rab11a, and EGFP-Rab7 Transgenic Lines

To mark and study early, recycling, and late endosomes, we generated constructs encoding N-terminal fluorescent fusions of Rab5c, Rab11a, and Rab7. We chose Rab5c (early), Rab11a (recycling), and Rab7 (late) for analysis based on their established localization and functions in these endosome subtypes and conservation of protein function/localization across evolution (Zhang et al., 2007). Most vertebrates have three paralogs of Drosophila Rab5 (Rab5a–c). However, in zebrafish there are additionally two rab5a genes (rab5aa and rab5ab) (see Supp. Fig. S1a, which is available online). Transcripts of rab5aa, rab5ab, and rab5c are all ubiquitously expressed in zebrafish embryos, unlike rab5b, where expression is limited to the yolk syncytial layer, pronephric duct, and telencephalon (Thisse and Thisse, 2004). Thus, rab5c was chosen for examination in zebrafish due to its ubiquitous expression and to avoid potential complications due to gene duplication (rab5aa and rab5ab). Whereas many vertebrates have a single rab11 gene, zebrafish possess four paralogs: rab11a, rab11a-like, rab11ba, and rab11bb. We selected rab11a as a representative recycling endosome marker because rab11a shows ubiquitous expression, unlike the other rab11 paralogs that show some degree of tissue specificity (Thisse and Thisse, 2004). Additionally, the zebrafish Rab11a protein is the most similar zebrafish paralog to the well-studied Drosophila RAB11 (Supp. Fig. S1a). For rab7, only one homolog has been curated in zebrafish, and this gene was chosen for characterizing late endosomes. Interestingly, other species such as frog, mouse, rat, and human possess two rab7 genes: rab7a and rab7b (Supp. Fig. S1). Protein alignments against the most recent build of the zebrafish genome (Zv8), however, indicated the presence of two hypothetical proteins (LOC 449549 and LOC 431725) with significant homology to Rab7 protein, suggesting that other rab7 paralogs are present in zebrafish (Supp. Fig. S1b).

Transgenic zebrafish lines expressing EGFP-tagged versions of the Rab proteins from h2ax regulatory sequence were generated using standard protocols (Fig. 1a) (Kawakami, 2004, 2005; Kwan et al., 2007). N-terminal fusions were generated as they have been shown to not interfere with protein function in other systems (Zhang et al., 2007). These lines were designated Tg(h2afx:EGFP-Rab5c)mw5, Tg(h2afx:EGFP-Rab11a)mw6, and Tg(h2afx:EGFP-Rab7)mw7 and will subsequently be referred to as EGFP-Rab5c, EGFP-Rab11, and EGFP-Rab7 (Fig. 1b). For each line, transgenic fish showed punctate EGFP expression throughout the embryos. Single-copy, functional transgenes were maintained through outcrosses to wild-type strains. No abnormal phenotypes were noted in transgenic embryos and the fish developed and reproduced normally. To determine if the expression of the EGFP-Rab fusions resulted in compensatory down-regulation of the endogenous transcripts, quantitative RT-PCR was performed on both transgenic and non-transgenic larvae. No differences in endogenous rab5c, rab11a, or rab7 transcript abundance were observed between wild-type and transgenic fish (Supp. Fig. S2). We next attempted to investigate whether EGFP-Rab puncta co-localized with endogenous Rab5c, Rab11, and Rab7 proteins. Unfortunately, we were unable to obtain antibodies specific to the zebrafish Rab proteins, precluding us from performing definitive co-localization studies through fluorescent microscopy or immuno-electron microscopy. We, therefore, used alternative strategies to address whether the EGFP-Rab labeling marked the expected endosome compartments.

Figure 1.

Transgenic construct models and Rab protein alignments. A: Schematic of constructs used to generate the transgenic lines. N-Terminal GFP fusions of wild-type Rab proteins are expressed under control of the ubiquitous h2afx promoter. N-terminal mCherry fusions of mutated Rab proteins are expressed under control of the conditional UAS promoter for tissue-specific and temporal regulation of mutant protein expression using the GAL4/UAS system. Flanking Tol2 sequences faciliate genomic integration. B: Allele designation and common names of EGFP-Rab and mCherry-Rab Mutant transgenic lines. C: Sequence alignments of portions of the GTP-binding pockets of the Rab proteins, indicating high homology between zebrafish, fly, and human. Amino acids highlighted in yellow are mutated to generate either dominant-negative or constitutively active Rab proteins by affecting GTP/GDP affinity (dominant-negative mutations) or GTPase activity (constitutively active mutations). Amino acid sequence is conserved for each set of orthologs except when marked by an asterisk (*). Numbers give amino acid position for each protein.

Early, recycling, and late endosomes can be characterized by their distribution in polarized cells and the time-dependent accumulation of cargo in uptake assays. To assist in quantifying the number, size, degree of polarized localization, and dynamic features of the EGFP-Rab-marked vesicles, we developed an automated tracking program. The automated tracking program runs on the Windows operating system as a stand-alone program or from within the MATLAB environment on any operating system. The program segments (delineates) and then tracks each fluorescently labeled endosome throughout the image sequence. After segmentation and tracking, the program generates an Excel spreadsheet with organelle size and position information for each endosome. The whole process takes a few minutes to analyze a 60-frame image sequence. Initial results using the automated tracker were confirmed through manual measurements. The tracking software is available free of charge and can be downloaded from

To begin characterization and investigate the specificity of the three EGFP-Rab lines, we first analyzed the size distribution of puncta marked by each fusion protein. Published electron microscopic studies from various cell types have shown that endosomes range in cross-sectional areas from 0.01–1.3 μm2 (Gorvel et al., 1991; Bucci et al., 1992, 1994; Cataldo et al., 1997). Confocal time-lapse microscopy revealed EGFP-Rab-labeled puncta of various sizes, some actively associating/dissociating with vesicles and others remaining stably complexed throughout much of the time-course (Supp. Movies S1–3). Analysis of stable fluorescent puncta (present in five consecutive frames) within retinal neuroepithelial cells indicated that the cross-sectional size distribution fell within the expected range for endosomal compartments (Fig. 2). We then characterized the degree of polarization of EGFP-Rab puncta for each transgenic line in a variety of polarized cells. Previous studies of endosome subtypes in polarized cells, either cultured MDCK cells or epithelia in vivo, have shown non-uniform distribution along the apical-basal axis. Specifically, polarized cells exhibit enriched localization of early endosomes at the apical surface, as indicated by RAB5 expression (Bucci et al., 1994; Leung et al., 2000). Similarly, RAB11-positive recycling endosomes are known to reside in a pericentriolar region at the apical surface (Ullrich et al., 1995). As RAB7 replaces RAB5 protein with late endosome maturation, localization shifts more centrally (Marois et al., 2006). Examination of transgenic embryos showed differences in the polarization of the EGFP-Rab puncta in neuroepithelial cells from the retina, hindbrain, and otic vesicle (Fig. 3). Ensuing characterization of the bright vesicular structures showed that both Rab5c- and Rab11a-positive endosomes localized towards the apical surface, while EGFP-Rab7-positive vesicles were located more centrally in polarized cells of the developing retina, hindbrain, and ear (Fig. 3A–R). Quantization confirmed the polarized distribution, where EGFP-Rab5c and EGFP-Rab11a were most apical (Fig. 3S–U).

Figure 2.

Tracking and cross-sectional area. Determination of EGFP-Rab-positive endosomes. A: Example of the tracking software over five time-series of an EGFP-Rab5c embryo. Tracked endosomes are automatically circled in red and are given an identification number to correlate numerical data in generated spreadsheets with visual images. Arrows point to endosomes recognized by the program (outlined in green) that were not followed for a minimum of five consecutive frames, and, therefore, no data were obtained from these endosomes and they were not given identification numbers. Insert in panel T=0 is a higher magnification of the boxed region showing numbering and outlining of tracked endosomes. T represents progression of time in seconds between frames. B: Distribution of cross-sectional area measurements determined for each of the EGFP-Rab lines. Most of the tracked endosomes have cross-sectional areas between 0.2 and 0.6 μm2, with very few endosomes larger than 1 μm2. P value is calculated from a one-way ANOVA.

Figure 3.

Expression and polarization of EGFP-Rab fusion proteins in neuroepithelial cells. Representative single-plane, confocal images of 28hpf embryos expressing (A–F) EGFP-Rab5c, (G–L) EGFP-Rab11a, or (M–R) EGFP-Rab7. Distinct fluorescent endosomes can be visualized as bright structures above a general membrane background in B, D, F, H, J, L, N, P, and R. Bright-field images (A, C, E, G, I, K, M, O, and Q) illustrate the structure of the developing epithelium with apical and basal surfaces indicated. Scale bars = 25 μm;. Asterisk in E, K, and Q indicates the presence of the anterior otolith. S–U: Quantification of endosome distance from the apical surface in (S) retinal neuroepithelial cells, and (T) hindbrain epithelial cells, (U) otic vesicle epithelial cells. In each cell type, EGFP-Rab5c, and EGFP-Rab11a, polarization is more towards the apical surface than EGFP-Rab7-positive endosomes. N>200 endosomes from images of 10 embryos across multiple clutches for each genotype and cell type. Mean distance from the apical surface is indicated with error bars indicating SEM. One-way ANOVA results indicate P < 0. 0001 for each cell type, followed by the Tukey's post-test with ***P < 0.001.

To verify that the EGFP-Rab proteins mark distinct structures indicative of vesicle progression from early (Rab5c-positive) to recycling (Rab11a-positive) or late (Rab7-positive) endosomes, we performed an in vivo lipophilic dye uptake assay (Fig. 4 and Supp. Fig. S3). Otic vesicle epithelial cells were imaged to examine the time-course of dye co-localization within endosome subtypes. The lumen of otic vesicles of 28-hr post-fertilization (hpf) EGFP-Rab transgenic larvae were injected with FM4-64 lipophilic dye to continuously label membranes of adjacent neuroepithelia. Internalization of the dye in otic neuroepithelia was apparent within minutes, indicating a high degree of endocytosis in the otic vesicle epithelial cells, a phenomenon maintained as the cells differentiate into sensory hair cells (Seiler and Nicolson, 1999). At early time points (two minutes post injection), the FM4-64 dye co-localized to a high degree with vesicles marked by EGFP-Rab5c and EGFP-Rab11a, consistent with early and recycling endosomes (Fig. 4E). For this time-point, very few FM4-64-positive structures overlapped with EGFP-Rab7 expression, suggesting EGFP-Rab7 labeled a “late” endosome subtype. Conversely, 10 min after injection of the lipophilic dye, there was a high degree of co-localization of FM4-64-positive structures with EGFP-Rab11a and EGFP-Rab7, but not EGFP-Rab5c (Fig. 4E). Together, these experiments confirmed that the transgenic lines mark the expected endosome subtypes. Interestingly, there was a high degree of co-localization of the FM4-64 dye with EGFP-Rab11a at both the 2- and 10-min time-points. Consistent with this observation, it was previously noted that RAB11 partially co-localizes with RAB5 (Ullrich et al., 1996) and these markers of early and recycling endosomes can exist on sub-domains of the same vesicular structure (Sonnichsen et al., 2000). Therefore, the high degree of EGFP-Rab11a co-localization with FM4-64 dye at both time-points is suggestive of (1) vesicles transitioning from early to recycling compartments and (2) active recycling of material throughout the duration of the experiment. Overall, these experiments coupled with a wealth of published data suggest that the EGFP-Rab5c, EGFP-Rab11a, and EGFP-Rab7 transgenic lines label early, recycling, and late endosomes, respectively.

Figure 4.

FM 4-64 injections into the developing otic vesicle indicate progression of endosome sub-types. A–C: 50-μm images thresholded for high fluorescence providing an example of FM 4-64 dye injection into the otic vesicle of a 28hpf EGFP-Rab5c embryo, 2 min post-injection. Single-channel images show intracellular puncta positive for the FM 4-64 dye (A), EGFP-Rab5c (B), and merged in C. D: Boxed image in C is magnified. Circles in A–D indicate non-membrane-associated, intracellular FM 4-64-positive vesicles. Arrows indicate examples of vesicles positive for both EGFP-Rab5c and the FM 4-64 dye, while the asterisk (*) indicates an FM 4-64-positive vesicle that is EGFP-Rab5c negative. E: Quantification of the percentage of FM 4-64-positive vesicles that co-localize with EGFP-Rab-positive vesicles at 2 and 10 min post-FM 4-64 dye injection. Numerical values below the x-axis indicate total number of FM 4-64 puncta analyzed from a minimum of 10 embryos. n-values indicate total number of FM 4-64 vesicles positive for EGFP-Rab at each time-point. Statistical Analysis: Fisher's Exact Test. ***P < 0.001; ns, not significant.

Generation of Dominant-Negative and Constitutively Active UAS-Driven Rab Transgenic Lines

To establish tools for examining Rab protein function in developing tissues, constructs encoding dominant-negative (DN) and constitutively active (CA) versions of Rab5c, Rab11a and Rab7 proteins were generated (Fig. 1). For each rab cDNA, mutations were introduced to codons corresponding to critical amino acids previously characterized to confer dominant-negative or constitutively active functions when altered in human and Drosophila Rab proteins (Stenmark et al., 1994a, 1994b; Feng et al., 1995; Meresse et al., 1995; Chen et al., 1998). These amino acid substitutions result in proteins with either reduced GTP affinity (dominant-negative) or reduced GTPase activity (constitutively active). Protein alignments revealed the location of the conserved, critical amino acids in the zebrafish homologs (Fig. 1c). The mutant Rab proteins were generated as N-terminal mCherry fusions and expressed from a UAS promoter, thus providing temporal and tissue-specific expression through use of Gal4 transactivating lines (Asakawa and Kawakami, 2008). Using these constructs, the following transgenic lines were established: Tg(UAS:mCherry-Rab5c S34N)mw33, Tg(UAS:mCherry-Rab5c Q81L)mw34, Tg(UAS:mCherry-Rab11a S25N)mw35, Tg(UAS:mCherry-Rab11a Q70L)mw36, Tg(UAS:mCherry-Rab7 T22N)mw37, and Tg(UAS:mCherry-Rab7 Q67L)mw38, and will subsequently be referred to as Rab5cDN, Rab5cCA, Rab11aDN, Rab11aCA, Rab7DN, and Rab7CA, respectively.

To begin to verify the functional consequences of the zebrafish mutant Rab variants, mCherry-rabDN or mCherry-rabCA mRNA was synthesized for each. The DN and CA versions are predicted to disrupt endogenous Rab functions in a dosage-dependent manner. Indeed, increasing concentrations of injected mRNA resulted in augmented expression of each mutant Rab protein and in larger percentages of embryos displaying defects such as altered body curvature, vascular anomalies, and increased cell death (Fig. 5, data not shown). Embryos injected with higher concentrations for the mCherry-rabDN mRNAs phenocopied body curvature defects described in previous reports of morpholino-based protein knockdown for Rab5c and Rab11a (see Fig. 7B,C, data not shown) (Kalen et al., 2009). Global defects in embryogenesis may be due to disruptions in cell polarity, migration, signaling, or survival as previous reports have defined roles of either Rab5c or Rab11a in these processes (Ulrich et al., 2005; Kalen et al., 2009; Yu et al., 2009; Tay et al., 2010; Nowak et al., 2011). Importantly, morphogenic defects induced by high levels of mCherry-rab5cDN expression were diminished in an EGFP-Rab5c background, but not in EGFP-Rab7 transgenic fish (Fig. 5B,C). Similar results were observed for all transgenic lines with mRNA injection. As a control, injection of mCherry-CAAX mRNA resulted in >83% of embryos displaying WT morphology (115/137) with no embryos displaying moderate or severe phenotypes. These experiments suggest that wild-type EGFP-Rab transgenes are functional and the mutant versions are specific in that each wild-type Rab can abrogate the deleterious effects caused by the corresponding dominant-negative version.

Figure 5.

Dosage-dependent defects of mCherry-Rab5cDN expression are rescued by expression of the EGFP-Rab5c transgene. A: Representative montaged confocal images of 24hpf embryos showing phenotypic designations of Wild-type (WT), Mild, Moderate, or Severe after injection with mCherry-Rab5cDN RNA. Arrow indicates region of edema and blood pooling posterior to the yolk extension, characteristic of embryos with the “Mild” phenotype label. B: Quantification of relative percentage of embryos displaying phenotypes shown in A after injection of mCherry-Rab5cDN mRNA. C: Table indicating the total number of embryos examined for each category represented in B. Statistical Analysis: Chi-Squared compared to 300 pg RNA injected into WT embryo category, ***P < 0. 0001. Three or more clutches were used for each category.

From a sub-cellular perspective, over-expression of constitutively active Rab proteins has been found to enlarge the endosome in which the Rab protein associates (Marois et al., 2006; Zhang et al., 2007). To examine alterations in size distribution as a consequence of RabCA expression, we injected mCherry-rabCA mRNA into EGFP-Rab transgenic embryos. The automated tracker was again used to determine cross-sectional areas of individual endosomes. The sizes of both EGFP-Rab-positive and mCherry-RabCA-positive structures were analyzed. As expected, expression of the mCherry-RabCA proteins resulted in mCherry-RabCA- or EGFP-Rab-labeled endosomes that were much larger than the EGFP-Rab vesicles in uninjected transgenic embryos (Fig. 6). For example, on average, the size of mCherry-Rab5cCA-labeled endosomes was approximately two-fold larger than EGFP-Rab5c-labeled endosomes from uninjected embryos. Similar increases were observed with Rab11a and Rab7. Interestingly, mCherry-Rab mutant protein expression only partially overlapped with wild-type transgenes (Supp. Fig. S4), consistent with previous observations for localization with atypical GTP/GDP exchange dynamics (Zhang et al., 2007). Together, these experiments are consistent with a gain-of-function for the mCherry-RabCA transgenic lines and loss-of-function for the mCherry-RabDN transgenic lines.

Figure 6.

mCherry-RabCA expression increases the size of corresponding EGFP-Rab-labeled endosomes. Expression of constitutively active mutants in the EGFP-Rab lines increases the size of tracked (A) EGFP-Rab5c, (B) EGFP-Rab11a, and (C) EGFP-Rab7 endosomes from >10 embryos for each genotype. RNAs encoding mCherry-RabCA proteins were injected into corresponding EGFP-Rab embryos (for example, mCherry-Rab5cCA injected into EGFP-Rab5c embryos). Time-lapse images were generated, detecting either green or red fluorescence, and subsequently processed through the tracking software. Endosomes were scored as described under each scatter plot. Individual points represent mean cross-sectional area with average area and SEM indicated. Results of one-way ANOVA, P < 0.0001 for A–C followed by Tukey's post-test. *P < 0.05, ***P < 0.001.

Apical Crumbs2a Localization in Neuroepithelia Depends on Endocytic Recycling

The utility of the UAS transgenic lines expressing mutant Rab proteins was confirmed by crossing with several Gal4 driver lines. The dose-dependence of function disruption was tested by crossing UAS:mCherry-RabDN/CA transgenic fish with carriers of an hsp70:Gal4 transgene. The hsp70 promoter can be induced to varying degrees by altering the timing of mild heat-shocks (Scheer et al., 2002). The resultant double transgenic embryos were, therefore, subjected to varying pulses of heat-shocks. Similar to embryos injected with mRNA encoding Rab-disrupting proteins, graded induction of the RabDN and RabCA transgenes displayed dose-dependent embryonic defects that mimic phenotypes resulting from mRNA injection (Rab11a DN shown in Fig. 7, data not shown). In general, the defects caused by heat-shock induction of the transgenes were mild compared to embryos injected with mRNA, likely owing to the developmental timing and transient nature of protein expression. The consequences of altering particular Rab activities in a tissue-specific manner was evaluated by observing Crumbs protein localization following expression of mutant Rabs specifically in neuroepithelia. Crumbs (Crb) is a single-pass transmembrane protein and is a component of an apical protein complex that regulates cellular polarity (Bulgakova and Knust, 2009). In Drosophila, cell surface levels of CRB protein are increased in larval discs mutant for RAB5, owing to a reduction in the internalization of CRB and other apical membrane proteins (Lu and Bilder, 2005). Expression of Rab5CA results in CRB localization within large internal vesicles (Lu and Bilder, 2005). RAB11 mutant flies or those expressing a Rab11DN, however, have more severely disrupted apical cell junctions and show significant levels of internalized CRB protein, indicating a role of endocytic recycling in proper CRB localization (Roeth et al., 2009). To investigate the specificity of the zebrafish mutant Rab transgenic lines, we assessed the localization of Crumbs2a (Crb2a) protein following expression of mutant Rabs in hindbrain neuroepithelia. Expression of mutant Rabs in hindbrain neuroepithelia was accomplished using regulatory sequence associated with the vsx2 gene. The Tg(vsx2:Gal4vp16)mw39 transgenic line (referred to as vsx2:Gal4) activates UAS transgenes strongly in neuroepithelia of the developing retina and hindbrain (Liu et al., 1994; Rowan and Cepko, 2005; Kimura et al., 2006) and to a lesser extent in other regions of the embryo (Supp. Fig. S5). Using the zs4 monoclonal antibody, which recognizes Crb2a protein (Hsu and Jensen, 2010), immunoreactivity was restricted to the apical surface of neuroepithelial cells. In cells expressing Rab5cCA, an increased amount of Crb2a protein localized to large internal vesicles, consistent with augmented internalization of the apical membrane protein (Fig. 8D–H,X). Like previous reports in Drosophila, Rab11aDN expression resulted in severe disorganization of Crb2a localization (Fig. 8I–M,Y). Hindbrain morphogenesis was also severely disrupted in vsx2:Gal4/UAS:Rab11DN double transgenic embryos (Fig. 8I–M). In these embryos, however, neuroepithelia not expressing the transgenes showed normal morphology and appropriate polarized distribution of Crbs2a (Fig. 8A–C,N–R). Rab7DN expression did not affect Crb2a localization, consistent with the lack of a functional link between late endosomes and Crumbs localization to the apical region of polarized cells (Fig. 8S–W).

Figure 7.

mCherry-Rab11aDN expression mimics morpholino-induced Rab11a loss-of-function. Confocal montaged images of 24–28hpf embryos examining the effect of Rab11a loss of function. A: Control embryo. B: Rab11a morphant (1.6 ng injected). C: Embryo injected with mCherry-Rab11aDN mRNA. D: Transgenic hsp70:Gal4;UAS:mCherry-Rab11aDN embryos after high-dose (30 min) heat-shock at 70% epiboly. E: Transgenic hsp70:Gal4;UAS:mCherry-Rab11aDN embryos after low-dose (10 min) heat-shock at 70% epiboly.

Figure 8.

Crumbs2a localization is altered when either Rab11aDN or Rab5cCA proteins are expressed. Immunolocalization of Crumbs2a to the apical membrane in the hindbrain epithelium of (A–C) 28hpf vsx2:Gal4 control; (D–H) vsx2:Gal4;mCherry-Rab5cCA; (I–R) vsx2:Gal4/mCherry-Rab11aDN; and (S–W) vsx2:Gal4/mCherry-Rab7DN transgenic embryos. Images B,C, E–H, J–M, O–R, and U–W represent higher magnification of the boxed regions of the brightfield images in A, D, I, N, and S, respectively. Boxed regions in H and M are shown as higher magnifications in X and Y, respectively. Arrows in F–H and X indicate Crumbs2a immunostaining localized to large intracellular mCherry-Rab5cCA puncta. Hindbrain morphogenesis (asterisk in J) and Crumbs2a localization is disrupted in mCherry-Rab11aDN-expressing embryo (I–M,Y). Crumbs2a immunostaining is preserved, however, in the forebrain neuroepithelia (N–R), and in regions devoid of Rab11aDN expression as indicated by the arrowheads. S–W: Expression of mCherry-Rab7DN does not appear to affect Crumbs2a localization and expression. Scale bars in A, B, D, E, I, J, N, O, S, and T = 50 μm; scale bars in X and Y = 10 μm.


Endocytosis and vesicle trafficking are fundamental processes that regulate many aspects of development, disease, and cellular homeostasis. Members of the Rab family of small GTPases define and are critical for different sub-types of endosomes and other trafficking organelles. We have generated a collection of transgenic lines based on three Rab proteins to study endosome biology within zebrafish. Transgenic lines include EGFP-Rab5c marking early endosomes, EGFP-Rab11a marking recycling endosomes, and EGFP-Rab7 marking late endosomes. In addition, inducible transgenic lines were generated to express constitutively active and dominant-negative proteins in a temporal and spatial specific manner for the selected Rab genes. We have characterized all of these lines and shown that the wild-type EGFP-Rab proteins were located in a punctate pattern, associated with vesicles of appropriate sizes, and in (neuro)epithelia, polarized in a manner consistent with the endogenous endosomes. Furthermore, a lipophilic dye uptake assay showed a time-dependent progression of FM4-64 endosomes from EGFP-Rab5c vesicles (early) to EGFP-Rab7 (late), indicating that the Rab fusion proteins marked the predicted endosomal compartments. With regard to the function disrupting transgenic lines, early expression of the dominant-negative versions phenocopied morphant embryos and showed dosage-dependent effects. The constitutively active Rab variants resulted in increased vesicle size of the corresponding endosomal compartment. Furthermore, Crumbs2a localization was altered in function-disrupting transgenic lines consistent with previous reports.

Additionally, as part of this study we generated an automated tracking program to facilitate rapid, unbiased segmentation and analysis of fluorescently labeled sub-cellular structures. This tracking program provides the ability to assess both size and position of endosomes as they change over time. For example, we used the program to identify stable endosome structures and to measure their size. However, the program could also be used to measure dynamic features such as vesicle trafficking direction and velocity.

Global overexpression of transgenic Rab mutant proteins, with the exception of Rab11aDN, resulted in specific defects, but few global embryonic phenotypes. Differential dose or spatial requirements for each mutant protein could account for observed differences between transgenic lines in production of embryonic phenotypes. Redundancy may also explain why expression of Rab5c and Rab7 mutant proteins resulted in mild global phenotypes, as 4 paralogs of rab5c (rab5aa, rab5ab, rab5b, and rab5c-like) and two hypothetical proteins with high homology to Rab7 are present in the zebrafish genome. At the cellular level, however, Crbs2a trafficking and localization was disrupted by expression of both mCherry-Rab5CA and mCherry-Rab11aDN expression, indicating Rab protein function was affected cell-autonomously by expression of these transgenes.

The usefulness of the transgenic lines is dictated by the expression levels of the transgenes, a direct consequence of the strength of the h2afx promoter or availability of Gal4 drivers for temporal and tissue-specific regulation of expression. The EGFP-Rab transgenes are driven by a 1-kb fragment of the promoter for H2AX, resulting in ubiquitous expression of the transgenes. As development progresses, however, cells devoid of EGFP-Rab expression become apparent (data not shown), presumably due to h2ax down-regulation as cells become quiescent (Thisse and Thisse, 2004). Additionally, the 10x UAS promoter used to promote dominant-negative and constitutively active Rab protein expression is prone to CpG repeat methylation of the UAS repeats, which can lead to promoter silencing in subsequent generations (Goll et al., 2009; Akitake et al., 2011). Furthermore, different UAS transgenes may not express equivalently in differentiated cell types due to position effect variegation. Therefore, selection of transgenic carriers must be monitored for consistent expression levels across generations and each UAS transgene should be confirmed as expressed in the cells targeted for analysis.

Potential uses for these transgenic lines include analysis of endocytosis-dependent signaling and disease modeling. With regard to the latter, disease-causing mutations and associations for Rab proteins and Rab adaptor proteins have recently been described for a variety of conditions including cancer and neurodegenerations (Mitra et al., 2011). For example, activating mutations in human RAB7 result in Type2B Charcot-Marie-Tooth Syndrome, a dominantly inherited axonal degeneration (Verhoeven et al., 2003; Houlden et al., 2004; Meggouh et al., 2006; Spinosa et al., 2008; McCray et al., 2010). While the molecular-genetic defects of this disease have been established, the cellular etiology remains unknown. Use of the transgenic lines described here may help in better understanding this and related neurodegenerations in which defects in endosome biology have been implicated. Similarly, the transgenic lines and software tools will aid in better understanding the intricacies of in vivo endosome biology and the role of intracellular trafficking in specific cells and cellular processes.


Zebrafish Maintenance

Zebrafish (Danio rerio) were kept at 28.5°C on a 14-hr light:10-hr dark cycle on an AHAB recirculating filtered water system (Aquatic Habitats, Apopka, FL). Embryos were obtained through natural matings and placed directly into system fish water or used for microinjection. N-phenylthiourea (PTU) was added to the rearing water (0.003%) to inhibit melanin synthesis during larval development and facilitate better fluorescent microscopy.

Zebrafish Transgenic Lines

  • Tg(h2afx:EGFP-Rab5c)mw5 (This Study)

  • Tg(h2afx:EGFP-Rab11a)mw6 (This Study)

  • Tg(h2afx:EGFP-Rab7)mw7 (This Study)

  • Tg(UAS:mCherry-Rab5c S36N)mw33 (This Study)

  • Tg(UAS:mCherry-Rab5c Q81L)mw34 (This Study)

  • Tg(UAS:mCherry-Rab11a S25N)mw35 (This Study)

  • Tg(UAS:mCherry-Rab11a Q70L)mw36 (This Study)

  • Tg(UAS:mCherry-Rab7 T22N)mw37 (This Study)

  • Tg(UAS:mCherry-Rab7 Q67L)mw38 (This Study)

  • Tg(vsx2:Gal4vp16)mw39 (This Study)

  • Tg(hsp70:Gal4)1.5 (Scheer and Campos-Ortega, 1999)

  • Tg(5xUAS:EGFP)zf82 (Asakawa and Kawakami, 2008)

Generation of Transgenic Lines

RT-PCR was performed to generate Gateway® (Invitrogen, Carlsbad, CA) 3′ Entry clones including the full-length cDNAs for rab5c (Accession number NM_201501), rab11a (Accession number NM_001007359), and rab7 (Accession number NM_200928) with Gateway® attB recombination sites attached at the 5′ end using the following primers (gene-specific sequences are listed in lowercase):







For each construct, Gateway® (Invitrogen) recombination was performed using the Tol2kit (Kwan et al., 2007) to generate N-terminal EGFP fusions of the Rab proteins, expressed under control of the quasi-ubiquitous h2afx promoter: Tol2 - h2afx:EGFP-Rab.

To generate point mutations resulting in constitutively active and dominant-negative proteins, the QuickChange® (Stratagene, La Jolla, CA) site-directed mutagenesis was performed using overlapping, complimentary primers against the following sequences (modified nucleotide(s) indicated in lowercase):







Additional Gateway® (Invitrogen) recombination was implemented, generating N-terminal mCherry fusions under control of the inducible UAS promoter for conditional expression using various Gal4 drivers (Scheer and Campos-Ortega, 1999): Tol2 – UAS:mCherry-Rab mutant.

In order to direct tissue-specific expression of the Rab mutants to the retina and developing nervous system (Liu et al., 1994; Rowan and Cepko, 2005; Kimura et al., 2006), a portion of the promoter of the zebrafish ortholog of Chx10, vsx2, was cloned to generate a 5′ entry clone to assemble the Tol2 – vsx2:Gal4vp16 construct using techniques described above. More specifically, genomic DNA was isolated and PCR was executed to isolate 800 bp directly upstream of the ATG start site of vsx2, completely overlapping the 72-bp 5′ UTR sequence. Primers used to isolate this sequence for Gateway® (Invitrogen) recombination are as follows (promoter-specific sequences listed in lowercase):



A stable transgenic line was generated, directing Gal4 expression to the neural retina, developing nervous system, and other tissues (Supp. Fig. S5).

Expression of each construct and generation of the transgenic lines was performed through co-injection of the circular DNA plasmid with Transposase RNA as previously described (Kawakami, 2004, 2005). Single, active inserts were obtained through out-crosses to wild-type strains and confirmed when examinations of the proportion of offspring followed expected Mendelian inheritance. All experiments were conducted on these single-insert strains.

Additional entry clones and Tol2 constructs were generated through Gateway® (Invitrogen) recombination and are available upon request and listed in Supp. Figure S6.

Fluorescent Microscopy

Zebrafish embryos were raised in 0.003% (PTU) (Sigma Aldrich, St. Louis, MO) and examined for transgenic expression of either mCherry or EGFP fusions using a Leica MZFLIII fluorescent dissection microscope. Confocal microscopy was performed using a Nikkon Eclipse E800 confocal microscope on larvae anesthetized in 3-amino benzoic acidethylester (Tricaine) (Sigma Aldrich, St. Louis, MO) and embedded in 1% low-melting agarose in glass-bottomed Petri dishes. Images were generated using the Nikon EZ-C1 viewer (Nikon Instruments, Melville, NY), Adobe Photoshop and Illustrator (Adobe Systems, Inc., San Jose, CA), and Metamorph (Molecular Devices, Inc., Sunnyvale, CA) software.

In Vitro Transcription of mCherry-Rab Mutants

Dominant-negative and constitutively active mutant entry clones were used to generate CMV:mCherry-Rab mutant constructs in order to make sense mRNA of the mutant fusions from the SP6 polymerase promoter contained in the p5E-CMV/SP6 construct (Kwan et al., 2007). RNA was generated from linearized plasmids using the Ambion® mMessage mMachine® SP6 kit protocol (Ambion, Austin, TX), followed by DNase treatment. A poly(A) tail was added using the Ambion® Poly(A) tailing kit.

Measurements of Endosome Polarization

Both 100-μm (neural retina and hindbrain) or 50-μm (otic vesicle) confocal images of 28hpf zebrafish retinas, hindbrains, or otic vesicles were obtained. Using the MetaMorph (Molecular Devices, Inc., Sunnyvale, CA) software, apical-to-basal distances of neuroepithelial cells in the central retina were observed. Images were thresholded for brightly fluorescent structures. Distance from the apical surface was measured from the center of the endosome structure to the RPE/apical membrane interface (retinas) or luminal surface (hindbrain and otic vesicle).

Automated Tracking Program

The automated tracking program segments, or delineates, endosomes in each image frame and establishes temporal correspondences (tracks) among segmentation results. Endosome segmentation begins using a scaled Otsu threshold, with scaling parameter α (Otsu, 1979). This thresholding is applied to each image to determine foreground regions. The foreground regions identified by Otsu thresholding are further refined into regions of maximum brightness using the extended maxima transform with height parameter h (Soille, 2003; Gonzalez et al., 2004). A morphological opening smoothes these regions (Soille, 2003). Finally, any regions smaller than a minimum area or larger than a maximum area are discarded. These maximally bright foreground regions are considered to be endosomes. The minimum and maximum acceptable areas, as well as the α and h parameters, are user-controllable parameters that account for imaging variability. After endosomes are segmented in all images, an automated multi-target tracking technique is applied to determine temporal associations among segmentation results (Blackman and Popoli, 1999). This tracking approach uses an optimal single-frame bipartite matching to associate tracks to previously segmented endosomes on a frame-by-frame basis, similar to the neural stem cell tracker described previously (Cohen et al., 2010). The association cost matrix used by the bipartite assignment is calculated using a weighted sum of centroid distance and area variation, encouraging longer tracks by weighting each cost term by inverse track length. After tracking, endosome properties including size and location on a track-by-track basis are written to an Excel spreadsheet for examination and processing.

Determination of endosome size (cross-sectional area) 50-μm confocal time-lapse images of 28hpf zebrafish retinas were obtained for 30 frames (∼1.57 sec between frames). The automated tracking program was used to quantify size, position, and number of fluorescent endosomes. Minimum endosome area of 0.1 μm2 was used in accordance with published endosome sizes (Cataldo et al., 1997; Ganley et al., 2004) and manual examinations of tracking program efficiency, with a maximum area of 6 μm2. Endosomes were identified as stable structures that were tracked for 5 consecutive frames. The average area during tracking was determined for each tracked endosome structure. Ten or more embryos were used for each genotype and time-point.

Otic Placode Injections and FM-Dye Uptake

Twenty-eight-hour post-fertilization (hpf) embryos were anesthetized using Tricaine and immobilized in 1% low-melting agarose in glass-bottomed dishes. Standard microinjection techniques were performed to infuse ∼3 ng FM®4-64 (Molecular Probes) (1 μg/μl in DMSO) directly into the otic placode ventricle (t=0). Images of otic vesicle epithelial cells were taken either at 2 or 10 min post-injection. Co-localization events were determined using MetaMorph (Molecular Devices, Inc., Sunnyvale, CA) imaging analysis software as internalized (not associated with the plasma membrane) puncta with >50% of the cross-sectional area of the FM-dye vesicle as positive for both the FM-dye and EGFP-Rab proteins. Data is an accumulation of 3 or more experiments for each genotype and time-point.


The following morpholino oligonucleotides were synthesized by GeneTools, LLC:

  • Rab11a UTR MO2 5′ – ATGTGAAGG AAAGTGGAGGCGACCC – 3′ (Kalen et al., 2009)

  • Rab5c UTR MO3 5′ – CTTGTCAAAT CCCTGGCTGTCAGAC – 3′ (Kalen et al., 2009)


Standard whole mount immunohistochemistry was performed on 28hpf embryos that were fixed overnight at 4° C in 4% paraformaldehyde. Briefly, embryos were washed 3 times in PBS prior to 1 hr incubation in block (2% sheep serum, 1% TritonX-100, 1% Tween-20 in PBS. Primary antibody incubation occurred overnight in blocking solution at room temperature using the Crb2a/Zs4 antigen: 1:20 University of Oregon Monoclonal Antibody Facility (Hsu and Jensen, 2010). Embryos were then washed three times for 1 hr in 1% Tween-20 in PBS. Antibody detection was performed using a Goat anti-mouse 488 (Molecular Probes) 1:800 in blocking solution overnight at room temperature followed by washes with 1% Tween-20 in PBS.


Quantitative PCR was performed using the iQ™ SYBR® Green Supermix with the iCycler iQ™ detection system from BIO-RAD (Hercules, CA). Samples were run in triplicate and averaged, using ef1α as a reference gene. PCR primers were designed to detect endogenous transcript (detecting the untranslated region [UTR] of the transcript), transgene transcript (detecting the EGFP in the EGFP-Rab transcript), or total Rab transcript (detecting the cDNA sequence of both the endogenous and transgenic transcripts). Primer sequences are as follows:
















  • ef1α R: 3′ – AACACCCAGGCGTAC TTGAA – 3′


We thank Anitha Ponnuswammi and Melissa Reske for assistance with molecular biology and Michael Cliff, William Hudzinski, Thomas Waeltz, and Joseph Hudzinski for zebrafish husbandry. We also thank Drs. Clare Buckley, Jon Clarke, and Brian Perkins for comments and critically reviewing drafts of this manuscript. This work was supported by National Institutes of Health grants T32EY 014536 (B.S.C.), R01NS076709 (A.R.C.), R01EY014167 (B.A.L.), and a National Eye Institute Core Facilities Grant P30EY001931 to the vision research community of the Medical College of Wisconsin.