A role for Rab10 in von Willebrand factor release discovered by an AP-1 interactor screen in C. elegans



    1. INSERM Avenir team ‘Trafic intracellulaire et polarité chez C. elegans’, Rennes
    2. CNRS UMR6061, Rennes
    3. Université de Rennes 1, Institut de Génétique et Développement de Rennes, UEB, Rennes, France
    Search for more papers by this author
    • These authors contributed equally to this work.

  • C. E. F. DYER,

    1. MRC Cell Biology Unit, MRC Laboratory for Molecular Cell Biology and Department of Cell and Developmental Biology, University College London, London
    Search for more papers by this author
    • These authors contributed equally to this work.


    1. MRC Cell Biology Unit, MRC Laboratory for Molecular Cell Biology and Department of Cell and Developmental Biology, University College London, London
    Search for more papers by this author

    1. INSERM Avenir team ‘Trafic intracellulaire et polarité chez C. elegans’, Rennes
    Search for more papers by this author

    1. MRC Cell Biology Unit, MRC Laboratory for Molecular Cell Biology and Department of Neurobiology, Physiology and Pharmacology, University College London, London, UK
    Search for more papers by this author
  • D. F. CUTLER

    1. MRC Cell Biology Unit, MRC Laboratory for Molecular Cell Biology and Department of Cell and Developmental Biology, University College London, London
    Search for more papers by this author

Gegoire Michaux, CNRS UMR6061, 2 avenue du Professeur Léon Bernard, CS34317, 35043 Rennes Cedex, France.
Tel.: +33 2 2323 4775; fax: +33 2 2323 4478.
E-mail: gmichaux@univ-rennes1.fr
Daniel F. Cutler, MRC Laboratory for Molecular Cell Biology, University College London, Gower St, London WC1E 6BT, UK.
Tel.: +44 0207 679 7808; fax: +44 207 679 7805.
E-mail: d.cutler@ucl.ac.uk


Background:  Endothelial von Willebrand factor (VWF) mediates platelet adhesion and acts as a protective chaperone to clotting factor VIII. Rapid release of highly multimerized VWF is particularly effective in promoting hemostasis. To produce this protein, an elaborate biogenesis is required, culminating at the trans-Golgi network (TGN) in storage within secretory granules called Weibel-Palade bodies (WPB). Failure to correctly form these organelles can lead to uncontrolled secretion of low-molecular-weight multimers of VWF. The TGN-associated adaptor AP-1 and its interactors clathrin, aftiphilin and γ-synergin are essential to initial WPB formation at the Golgi apparatus, and thus to VWF storage and secretion. Objectives: To identify new proteins implicated in VWF storage and/or secretion. Methods: A genomewide RNA interference (RNAi) screen was performed in the Nematode C. elegans to identify new AP-1 genetic interactors. Results: The small GTPase Rab10 was found to genetically interact with a partial loss of function of AP-1 in C. elegans. We investigated Rab10 in human primary umbilical vein endothelial cells (HUVECs). We report that Rab10 is enriched at the Golgi apparatus, where WPB are formed, and that in cells where Rab10 expression has been suppressed by siRNA, VWF secretion is altered: the amount of rapidly released VWF was significantly reduced. We also found that Rab8A has a similar function. Conclusion: Rab10 and Rab8A are new cytoplasmic factors implicated in WPB biogenesis that play a role in generating granules that can rapidly respond to secretagogue.


von Willebrand factor (VWF) is a large multidomain highly-multimerized glycoprotein that binds to GpIIb/GpIIIa receptors on platelets, as well as collagen, heparin and factor VIII [1]. It is therefore critically involved in both primary and secondary hemostasis. It is synthesized in both megakaryocytes and endothelial cells, and it is thought that the source of most plasma VWF originates within the latter [2]. In endothelial cells VWF undergoes a complex biogenesis that includes the formation of large multimers, their folding into long proteinacious tubules at the trans-Golgi network (TGN), and finally their storage within huge cigar-shaped secretory granules called Weibel-Palade bodies (WPB) [3,4]. Several components of cellular membrane-trafficking machinery have been implicated in the initial formation of WPB at the TGN [5]. The most important so far are related to the adaptor complex AP-1, the heterotetrameric (gamma, beta, sigma and mu subunits) adaptor that binds to the cytoplasmic tail of cargo proteins, and recruits clathrin and other proteins to the membrane to assist in forming trafficking vesicles [6]. Loss of function of AP-1 or clathrin prevents formation of WPB in human umbilical vein endothelial cells (HUVECs) [7]. On the other hand, loss of the AP-1 binding proteins gamma-synergin and aftiphilin leads to formation of WPB that are apparently normal yet are no longer targeted into the regulated secretory pathway, but instead basally released from the cell [8]. Thus AP-1-dependent events at the TGN can have profound effects on the formation and function of WPB.

AP-1 is also implicated in transport between the trans-Golgi network and endosomes, in formation of neuroendocrine secretory granules, in removing missorted proteins from immature secretory granules in neuroendocrine and in pancreatic beta cells, and in basolateral targeting within mammalian epithelial cells [6]. Surprisingly, despite all these complex functions, AP-1 has relatively few known interactors [9,10] compared with the numbers known to associate with AP-2 [11,12], a clathrin adaptor operating at the plasma membrane to support endocytosis by clathrin-coated vesicles [13].

Because AP-1 and its cofactors are essential for WPB formation and correct maturation, we decided to look for new AP-1 interactors. Functionally important membrane-associated protein interactions are difficult to identify by yeast 2-hybrid screens or by co-immunoprecipitation strategies. We therefore decided to take a much broader, fundamentally different functional approach to look for AP-1 genetic interactors, by screening in vivo in the nematode worm C. elegans as a tractable genetic model organism.

C. elegans possesses two AP-1 mu subunits, UNC-101 and APM-1. Loss of function of both mu subunits results in embryonic lethality [14]. However, the loss of function of only UNC-101 results in partial larval lethality and the surviving animals are almost completely paralyzed [15]. To identify other genes required for AP-1 function, we performed a genomewide RNA interference (RNAi) screen, looking for unc-101 enhancers (i.e. genes whose loss leads to an increased severity of the unc-101 phenotype). Using an RNAi library covering more than 85% of the C. elegans genome, we identified 12 genetic interactors, including the small GTPase Rab10. Rab10, together with Rab8 and Rab13, is a member of a subfamily of Rabs similar in sequence to Sec4p, shown in yeast to be essential for post-Golgi trafficking and sorting [16,17]. It has been shown to localize to the trans-Golgi network [16,18], to sorting endosomes targeting cargo to the basolateral membrane [18,19], and to early endosomes [19,20]. Rab8A is reported to have both a similar location and similar role to Rab10: it localizes to the perinuclear region and recycling endosomes [21], and also to the Golgi apparatus [22,23]. In addition to a role in sorting polarized cargo similar to Rab10, Rab8A also has a key role in mediating interactions between polarized cargo and the actin cytoskeleton [24,25]. It has also been shown to directly interact with AP-1B in MDCK cells [21] and to colocalize on mature melanosomes with Rab27a [24], a marker of mature Weibel-Palade bodies [26].

Given the close correlation between AP-1 and Rab10 function in various models, we prioritized the investigation of Rab10 among the other newly identified genetic interactors. We demonstrate that Rab10 is enriched at the Golgi apparatus in HUVECs and that depletion of Rab10 by RNAi in this system changes the pattern of VWF release following secretagogue stimulation. In Rab10-deficient cells, there is a selective loss of the earliest secretory response. We further show that Rab8A has a similar function.

Materials and methods

C. elegans strains and reagents

All C. elegans strains were maintained as published [27]. The genomewide RNAi screen was performed using an RNAi library covering 86% of the C. elegans genome (Source BioScience, Geneservice, Nottingham, UK). The RNAi feeding experiments were performed as published [28]. C. elegans has the advantage that RNAi can be achieved by feeding it on bacteria expressing double stranded RNA of a defined gene. Thus, the RNAi clones can be cheaply maintained and amplified allowing large-scale RNAi screens that are much cheaper than those that can be performed in mammalian cells. The following strains were used in that work: the wild-type N2 strain, PS529 unc-101(sy108), DR1 unc-101(m1) and NL2099 rrf-3(pk1426), which were provided by the Caenorhabditis Genetics Center (University of Minnesota, Minneapolis, MN, USA).

Screening strategy

The major modification to traditional genomewide RNAi screens performed in C. elegans was to add young larvae instead of young adults for the primary screen. The C. elegans life cycle is 3.5 days long, starting with embryonic development followed by four larval stages called L1–L4 and an adult stage. This procedure was adopted because of the paralysis of the unc-101 strain: paralyzed adults move very slowly compared with wild type (WT) worms and they eat all the food in one place before moving while laying embryos. Laid embryos therefore hatch at places devoid of food; being almost paralyzed prevents them from reaching food and developing into L2 larvae. It is therefore very easy to get thousands of synchronized L1 larvae. We scored several phenotypes in the added larvae (P0 generation) and the next generation (F1 generation): arrest of the larval development or sterility of the P0, embryonic or larval lethality of the F1 and slower growth were the main observed phenotypes. More specific phenotypes were also recorded but were always observed in both unc-101 and WT strains. After screening the whole C. elegans genome using the Geneservice library, we selected all the clones that gave a phenotype stronger than the Unc-101 null phenotype (i.e. 2166 genes, corresponding to 12.9% of the tested genes). As a positive control, we found that apm-1(RNAi) induced an arrested phenotype in P0. The bacterial strain containing the vector L4440 was used as a negative control. The L4440 vector was used to clone the C. elegans genes while constructing the Geneservice library. It generates a 200 bp double stranded RNA, which is not related to a C. elegans sequence and can therefore be used as a negative control.

We then performed several secondary screens to select the best candidates. The first step was to repeat the screen in duplicate with the 2166 candidates to eliminate false positives, but using the other null allele unc-101(m1). This allowed us to also eliminate interactions with parasite mutations potentially present in the unc-101(sy108) background. We recapitulated the same phenotype for about 70% of the genes.

We next directly compared the phenotypes for these genes between the WT N2 strain and the unc-101(sy108) strain. To do so, we used the traditional procedure of adding young adult worms and scoring the phenotype in the F1 generation [29]. As an internal positive control we found that apm-1(RNAi) induces a larval lethal phenotype in WT N2 worms, and a synthetic embryonic lethal phenotype in the unc-101(sy108) background; this result has already been reported [14]. This direct comparison between phenotypes in WT and unc-101(sy108) strains led to the exclusion of all the genes having similar phenotypes in both backgrounds and therefore not behaving as genetic enhancers. At that point there were 59 candidates (Table 1). Finally the most stringent test was performed by comparing the phenotypes obtained after RNAi in the unc-101(sy108) and rrf-3(pk1426) backgrounds, the latter displaying hypersensitivity to RNAi [30]. This reduced the number of candidates to a more manageable 12 genes. There could, however, be bona fide unc-101 enhancers in the pool of 47 genes that were eliminated in this last step.

Table 1.   List of the AP-1 genetic interactors
Candidate identityC. elegans nameClosest human homolog
  1. Lines corresponding to proteins implicated in membrane traffic are listed in bold characters.

m110.5DAB-1Disabled 2
f55A12.7APM-1AP-1 mu subunit
f32e10.4IMA-3Importin subunit alpha-4
t22f3.3Glycogen phosphorylase
f28f8.5Not conserved
k09a9.3ENT-2Equilibrative nucleoside transporter 1
c10e2.6Monocarboxylate transporter

The clone directed against c37c3.3 (vps-32.2) is also directed against c56c10.3 (vps-32.1) due to sequence similarity between the two genes. Directly targeting vps-32.1 induces embryonic lethality, while targeting vps-32.2 induces a weak phenotype in WT worms, although this gene has only a very weak expression level. In C. elegans the paralog vps-32.1 has the main function [31]. We therefore presume that targeting vps-32.2 also weakly affects vps-32.1 and that the genetic interaction detected between unc-101 and vps-32.2 reflects an interaction between unc-101 and vps-32.1. We found a direct requirement for AP-1 in VPS-32.1 localization, supporting that hypothesis (our unpublished results).

Tissue culture and transfection

HUVECs were purchased from TCS-Cellworks (Buckingham, UK) and were cultured in HUVEC growth medium, which consists of M199 medium with Earle’s modified salts (GIBCO BRL; Invitrogen, Carslbad, CA, USA), 20% FCS, 10 U mL−1 heparin (Sigma-Aldrich, St Louis, MO, USA) and 30 g mL−1 endothelial growth supplement (Sigma-Aldrich). Cells were seeded onto tissue culture plates or coverslips previously coated with 1% porcine gelatin. Nucleofection using the Amaxa system was used to transfect DNA and siRNA into tissue culture cells, according to the supplier’s instructions.


All siRNA oligomers were purchased from Qiagen (Valencia, CA, USA), except for Rab10 from Thermo Fisher Scientific Dharmacon (Lafayette, CO, USA). Cells were transfected with 100–200 pmol of siRNA by nucleofection (Amaxa; Lonza, Cologne, Germany) using the manufacturer’s instructions. Typically, a 15-cm Petri dish of cells 70–80% confluent was used for four to six nucleofection reactions. Cells were then plated onto a 9-cm Petri dish and incubated for 2–3 days before being nucleofected again with 100–200 pmol of siRNA, and incubated for another 2–3 days before being processed for immunofluorescence and/or secretion assays. Two rounds of knockdown were performed to ensure an effective reduction in mRNA expression and protein levels.

Expression vector constructs

Rab10-GFP and Rab8A-GFP were both kind gifts from F. Barr (University of Liverpool, Liverpool, UK).

Secretion assays

VWF secretion from HUVECs was analysed as previously described [32]. Cells were rinsed in release medium (M199 medium with Earle’s modified salts (GIBCO BRL), 10 mm Hepes, pH 7.4, and 0.2% BSA); 1 mL of release medium containing 100 ng mL−1 phorbol 12-myristate-13-acetate (PMA) was then added to each well for the required time (10–30 min) at 37 °C. Medium (‘secreted VWF’) was collected after each incubation. The remaining VWF (‘remaining VWF’) was collected by lysing cells at 4 °C for 30 min in lysis medium (release medium containing 0.5% vol/vol Triton X-100, 1 mm EDTA and protease inhibitors (Sigma; Sigma-Aldrich). The comparative amounts of VWF were quantified by enzyme linked immunosorbent assay (ELISA) as described previously [33]. The level of secretion was normalized to the total amount of VWF present in the cells. This was calculated by adding ‘secreted VWF’ to ‘remaining VWF’ to get the total amount of VWF (‘total VWF’); ‘secreted VWF’ is then expressed as a percentage of ‘total VWF’. Constitutively released VWF was not collected and measured in these experiments as it was determined to be of a negligible amount and unaffected by the knockdown. Statistical analysis was performed on these values; however, data were displayed by expressing knockdown values as a percentage of mock values in order to clearly show differences between the two conditions. Experiments were repeated a minimum of three times.


Rabbit polyclonal anti-VWF was obtained from Dako North America (Carpinteria, CA, USA). The mAB 100/3 against AP-1 was from Sigma-Aldrich and the mAB against CD63 from Abcam (Cambridge, UK). Affinity-purified rabbit anti-MyRIP C-terminal antibody was described previously [34]. For immunofluorescence (IF), the rabbit anti-VWF antibody was diluted 1:10 000 in PGAS (see IF section) and for ELISA plate coating it was used at 1:600 in PBS. Sheep anti-TGN46 was purchased from Serotec (Oxford, UK) and was used at 1:100 in PGAS for IF. Donkey AlexFluor-488 and -568 secondary antibodies raised against either rabbit, sheep or mouse were purchased from Invitrogen (Carlsbad, CA, USA) and were used at 1:500 in PGAS. Dilutions are from manufacturers’ concentrations. Rabbit polyclonal Syntaxin 6 antibody was a kind gift of A. Peden (Cambridge Institute for Medical Research, Cambridge, UK).


HUVECs were grown on coverslips previously coated with 1% porcine gelatin. They were fixed with 3% PFA, quenched, and permeabilized with 50 mm NH4Cl and 0.2% saponin. Following blocking in PGAS (PBS, 0.2% gelatin, 0.02% NaN3, and 0.02% saponin), cells were incubated with primary antibodies for 45–60 min, washed in PGAS, and then incubated with Alexa-568 conjugated secondary antibodies for a further 45–60 min. The coverslips were then washed in PGAS and mounted with ProLong antifade reagent (Molecular Probes; Invitrogen). Coverslips were examined and imaged at an ambient temperature through a 63×  oil-immersion lens (NA 1.3) on a Leica TCS SPE confocal system (Leica, Wetzlar, Germany). Adobe Photoshop 9.0.1 and Illustrator 12.0.1 were used to generate figures from digital images.

Real-time PCR

RNA was prepared from knockdown and mock tissue culture cells using QIAshredder (Qiagen) and RNeasy mini kit (Qiagen) according to the manufacturer’s instructions. The RNA was then subjected to reverse transcription using the Super-Script III first-strand cDNA synthesis system (Invitrogen). DNA was next amplified under the following conditions : 95 °C, 10 s, 56 °C, 15 s, 72 °C, 15 s, 40 cycles; amplification was monitored by incorporation of Syber Green (Finnzyme, Espoo, Finland). Real-time PCR was carried out in a Mastercycler Eppendorf Realplex qPCR machine (Eppendorf North America, New York, NY, USA). The C(T) data were normalized with an Actin internal control and analyzed using the 2C(T) method [35]. Quantitect quantitative PCR primers to Rab10 and Rab8A were purchased from Qiagen. Actin quantitative PCR primers were as follows: GCGAGAAGATGACCCAGAT-actinF; TGGTGGTGAAGCTGTAGCC-actinR.


Screening of the C. elegans genome

We performed a genomewide RNAi screen in C. elegans, looking for genes that would enhance the lethality or growth defects of the unc-101(sy108) null mutation. Our rationale was that any gene whose loss from control animals has a mild phenotype on its own, yet leads to increased lethality in UNC-101-deficient animals, most likely encodes a product involved in an AP-1-dependent process. Details of the screening strategy are available in Materials and methods, and Fig. 1 shows a flow diagram of the strategy. We found that RNAi of 12 genes increased lethality in the unc-101 null mutation but not in the WT or rrf-3 RNAi hypersensitive background (Table 1).

Figure 1.

 Summary of the RNAi screen performed in C. elegans. The detailed presentation of the screen can be found in the Materials and methods section. L1 larvae correspond to the youngest larval stage and the phenotypes were scored on the developing larvae (P0 generation) and the next generation (F1 generation). L4 larvae correspond to the last larval stage, close to adulthood; the phenotype was therefore only scored on the next generation (F1 generation).

These 12 genes represent the strongest genetic screen-based candidates for having a role in AP-1 function. Among these genes, only four encode known membrane-trafficking factors, including the second AP-1 mu subunit APM-1. The rab-10 gene encodes a small GTPase involved in polarized trafficking in C. elegans and mammalian epithelial cells [19,20]. The vps-32.2 gene encodes an ESCRT-III protein; Vps32 proteins, together with other ESCRT components, are implicated in formation of intralumenal vesicles within multivesicular bodies (MVB) [36]. The dab-1 gene encodes a clathrin adaptor implicated in endocytosis [37]. Both apm-1 and dab-1 are known unc-101 genetic interactors [14,37]. Given the similarities of AP-1 and Rab10 functions in various models, we decided to further explore the function of Rab10.

Rab10 function in human endothelial cells

We began by investigating Rab10 location within human umbilical vein endothelial cells (HUVECs). Anti-Rab antibodies capable of specific recognition of individual members of the 60 + Rab superfamily members are rare, and so we determined the location of Rab10 in HUVECs following expression of a GFP-tagged version of a Rab10-encoding cDNA. Analysis by light microscopy (Fig. 2) of HUVECs nucleofected with this cDNA revealed that Rab10-GFP strongly (but not exclusively) colocalizes with the TGN marker, TGN46 (Fig. 2 upper panel), consistent with reports of its association with the Golgi apparatus in other cells [16,18]. Because Rab10 does not colocalize with VWF in WPB that have budded from the TGN and moved away from the perinuclear region (Fig. 2 lower panel), the role of Rab10 is highly likely to be in the early stages of WPB formation.

Figure 2.

 Rab10-GFP localizes to the perinuclear region, concentrated at the Golgi apparatus, but does not localize to Weibel-Palade bodies (WPBs). Rab10-GFP was overexpressed in human umbilical vein endothelial cells and cells left for 48 h to allow for turnover of WPBs (normally approximately 24 h) [38]. Cells were then fixed, permeabilized and processed for IF. Single slice confocal images were taken at the point at which the Rab10-GFP labeling was strongest. Labeling of the Golgi apparatus using the trans-Golgi network marker, TGN46, showed pixel overlap with the overexpressed Rab10 protein; however, no overlap was seen between Rab10-GFP and the WPB marker, von Willebrand factor. Scale bars = 25 μm.

To analyze the consequences of altering the early formation of WPB we quantified the regulated release of VWF. We therefore suppressed Rab10 expression by nucleofection of siRNA (Fig. 3A) and left the cells for 48 h to allow for turnover of WPB so that most existing WPB would be formed after introduction of the siRNA [26,38,39]. We then analyzed secretion of VWF by ELISA in HUVECs (Fig. 3B).

Figure 3.

 Rab10 knockdown in human umbilical vein endothelial cells (HUVECs) results in a reduction in phorbol 12-myristate-13-acetate (PMA)-induced release of von Willebrand factor (VWF) that is more pronounced during the earliest phase of VWF release. HUVECs were transfected with siRNA against Rab10 on days 1 and 3. From 20 experiments, knockdown efficiency as determined by qPCR was on average > 85% (A). Secretion assays were carried out between days 5 and 6. It was determined that constitutive secretion of VWF was not affected by the knockdown (data not shown) so only PMA-induced release was examined in these experiments. VWF levels, both released and total present in the cell, were assayed by ELISA. Released VWF was expressed as a percentage of total VWF present in the cell (lysate + releaseate). Data are shown as knockdown (grey bars) normalized to mock (white bars), where mock is set at 100%. Rab10 knockdown results in a reduction in stimulated VWF release compared with mock (B). This percentage reduction in release in knockdown cells is greatest in the early phase of release (10 min), suggesting a difference in release dynamics in VWF release from Weibel-Palade bodies. Data are from four replicates. A ratio t-test was performed: P values; 10 min (0.017), 15 min (0.109), 20 min (0.098), 30 min (0.046). One star indicates that the P value is < 0.05 but not < 0.01.

Nucleofected cells were stimulated with PMA for 10, 15, 20 and 30 min to stimulate WPB exocytosis and VWF release, and media collected for analysis by ELISA. Values for secreted VWF were first normalized against total VWF. Release data from knockdown cells were then expressed as a fraction of release from mock-treated cells (Fig. 3B). These data showed a reduction in secretion from Rab10-deficient cells. The greatest effect of Rab10 depletion on secretion occurs at the early time points, with a diminution of effect over time, clearly implying that Rab10 plays a role in regulating the earliest response of WPB to secretagogue stimulation.

Does this alteration in release kinetics reflect a change in maturation or number of WPB? Other modulations of the Golgi-associated machinery that acts with AP-1 have caused a fall in the multimeric state of VWF released following HUVEC activation [8]. We therefore analyzed this parameter and found no significant alteration of the multimeric state of released VWF in Rab10-deficient cells (Fig. 4A). However, quantification of WPB by light microscopy within WT and Rab10-deficient cells reveals a small but significant increase in numbers of WPB per cell in the Rab10-deficient population (Fig. 4B; P < 0.03).

Figure 4.

 Effect of losing Rab10 on maturation and numbers of Weibel-Palade bodies (WPBs). (A) Rab10 knockdown has no effect on multimerization of von Willebrand factor (VWF) within WPBs. Human umbilical vein endothelial cells (HUVECs) underwent two rounds of knockdown by siRNA of Rab10 on days 1 and 3. Two to three days after the second round, a secretion assay was then performed. Releasate following phorbol 12-myristate-13-acetate (PMA) stimulation for either 10, 15, 20 or 30 min was then concentrated 20-fold and loaded onto a 1.4% agarose multimer gel. A representative gel indicates that both high- and low-molecular-weight multimers are present in both knockdown and control lanes with no obvious differences in band density. (B). Knockdown of Rab10 results in a slight but significant increase in numbers of WPBs present in the cell. HUVECs underwent two rounds of siRNA-mediated reduction of Rab10 mRNA levels Two to three days after the second round, cells were fixed, permeabilized, and prepared for immunofluorescence analysis. Samples were labeled with antibodies against VWF and the average number of WPBs per cell was calculated. Data indicate that Rab10 knockdown results in an increase in WPBs per cell, perhaps due to a reduction in their release rate. Analysis was carried out by counting all distinguishable WPBs in 20–25 cells from knockdown or control cells. Knockdown of Rab10 mRNA levels as determined by qPCR was 95%. A t-test was performed; P = 0.03.

Finally, is the recruitment of AP-1 itself, or the recruitment of other WPB components affected by loss of Rab10? Analysis by immunofluorescent microscopy of the Golgi association of AP-1 and the recruitment of CD63 (lost in AP-3 deficient cells [40]) or of MyRIP (recruited late in biogenesis and affecting exocytosis by anchoring the WPB to actin [41]) all appear to be normal (Fig. S1). Thus, slightly higher numbers of apparently normal WPB, containing normally multimerized VWF, are found in Rab10-deficient HUVECs.

Rab8A – a partner of Rab10?

Given the similar localization and role of Rab8A and Rab10, together with reports of these Rabs acting together or in parallel to control polarized secretion [18], we decided to analyze its localization within HUVECs. Expression of a GFP-Rab8A construct shows a very similar distribution within HUVECs to that of Rab10, associating with the Golgi apparatus and not colocalizing with budded WPB (Fig. 5). We also analyzed its role in VWF release: siRNA-mediated reduction of expression of Rab8A (Fig. 6A) gives an effect similar to that caused by reducing Rab10, when measured 10 min after secretagogue addition (Fig. 6B).

Figure 5.

 Rab8A has the same cellular localization as Rab10. Overexpressed Rab8A-GFP localizes to the perinuclear region, where it is concentrated at the Golgi apparatus, but does not localize to Weibel-Palade bodies (WPBs). Rab8A-GFP was overexpressed in human umbilical vein endothelial cells and cells left for 48 h to allow for turnover of WPBs (normally approximately 24 h). Cells were then fixed, permeabilized and processed for immunofluorescence. Labeling of the Golgi apparatus using the trans-Golgi network marker, TGN46, showed pixel overlap with the overexpressed Rab8A protein; however, no overlap was seen with Rab8A-GFP and the WPB marker, von Willebrand factor. Single confocal sections were taken at the point where the Rab8A-GFP labeling was brightest and analyzed. Scale bars = 25 μm.

Figure 6.

 Knockdown of Rab8A has a similar effect on von Willebrand factor (VWF) secretion as Rab10. Human umbilical vein endothelial cells were nucleofected with siRNA directed either at Rab10, Rab8A, or both. Average knockdown over four experiments was >70% (A). Following two rounds of knockdown, secretion assays and ELISAs were performed 2–3 days after the second round knockdown following 10 min of stimulation (B). Quantity of VWF released was expressed as a percentage of total VWF present in the cell upon which value statistics were performed. Data are displayed as knockdown values normalized to mock, which has been set at 100% to clearly show the reduction in release in knockdown cells. Both Rab10 and Rab8A knockdown results in a 30–40% reduction in phorbol 12-myristate-13-acetate (PMA)-induced VWF release, with the double knockdown giving a slightly greater reduction in release. As for the single Rab10 knockdown experiment, a paired t-test was performed on the log10 of raw numbers where P (compared with mock) = 0.04 (Rab8A), 0.008 (Rab10), 0.004 (double). n = 5. One star indicates that the P value is < 0.05 but not < 0.01, two stars indicates a P value < 0.01.

Rab10 and Rab8A knockdown trigger a similar phenotype. To test whether these two proteins are likely to act in the same or in parallel pathways, we performed the double knockdown of Rab10 plus Rab8A (Fig. 6A). The double knockdown generates a slightly greater but not statistically significant reduction in stimulated release at 10 min when compared with either single knockdown (Fig. 6C). We therefore conclude that Rab10 and Rab8A probably control the same events during WPB biogenesis and work in the same pathway. Analysis by QPCR of the effects of siRNAs against Rab10, Rab8A or Rab8B on the mRNA encoding other members of this related trio reveal some cross-protein effects. However, because the greatest effect that we observe is a 26% reduction of Rab10 mRNA caused by the Rab8A-directed siRNA, we do not believe that this would have significant functional consequences, and will not complicate our analyses.


We have performed a genomewide RNAi screen looking for genes whose loss increases the defects associated with a null mutation in the unc-101 AP-1 mu subunit in the nematode C. elegans. We identified 12 AP-1 genetic interactors. The identification of the trafficking components Rab10, Vps32, apm-1 and dab-1, the latter two of which were already known to behave as unc-101 enhancers, validated our screening method. These data support the idea [42] that a screen for factors involved in endothelial function in man can be carried out in a nematode that does not even have a circulatory system. They are also consistent with the view that the basic cellular machinery underpinning membrane trafficking and organelle biogenesis is very highly conserved, and that cell-type specific cargo evolved to deal with complex physiologies, present in higher animals, exploits this universal machinery in novel ways. The increasing amount of data from screens carried out in a variety of genetically tractable organisms is thus likely to represent an opportunity for data mining to uncover further elements of the cellular machinery controlling WPB formation and function.

Several studies have shown that Disabled-2 and its C. elegans ortholog DAB-1 have a role in early steps of endocytosis [37,43–45], none of which suggest a role in WPB formation or function. The role of VPS-32 in C. elegans has also recently been studied: it was shown that VPS-32 loss of function induces an embryonic arrest during elongation and a role was proposed for VPS-32 in autophagy and endocytosis [31]. The canonical Vps32 function in the context of the ESCRT-III complex is in forming intraluminal vesicles within late endosomes [46], again suggesting that this protein is unlikely to be involved in WPB formation and function. Because AP-1 has several very different functions, the genetic interactions with Vps32 or DAB-1 probably reflect functional interactions in the endocytic route. Because the small GTPase Rab10 has been localized to the TGN [16,18], and implicated in basolateral trafficking [19,20] like AP-1B [47], a genetic interaction between these two factors is consistent. The relationship between sorting proteins for apical vs. basolateral routes, and sorting for delivery to the regulated secretory pathway in mammalian cells, is unclear, and the polarity of regulated secretion from endothelial cells is also not well understood. However, because apical-basolateral sorting can occur at the TGN, where WPB are formed, or in closely juxtaposed recycling endosomes, we investigated a potential role for this GTPase in forming WPB.

Using siRNA-mediated ablation of Rab10 expression in HUVECs, we have now found clear evidence of a role for Rab10 in controlling regulated exocytosis from WPB. Importantly, not only does the loss of Rab10 reduce secretagogue-responsiveness, but the induced change is selective. We find that when early response (10 min after activation) is analyzed, a loss of 40% of normal release occurs, whereas at 30 min, the difference, while still statistically significant, is much lower, at around 10%. Assays of VWF release have shown a quasi-linear increase in released VWF over 60 min [32], but this new dataset clearly implies that there is a machinery involved that can differentially affect early and late events. Whether, as in neuro/neuroendocrine or the pancreatic β-cell systems, there are readily-released and reserve pools with all the associated complexity of exocytic machineries is unclear, but now that their presence is suggested, experiments will surely follow. We also find a small increase in the numbers of WPB within HUVECs when Rab 10 is ablated (Fig. 4B). We suspect that this arises from the effect of our observed fall in secretion, accumulated over time.

Rab8 has previously been found to act together with Rab10 [18], and we report that this second small GTPase is also involved in regulating the release of VWF. Exactly how, and through which effectors these two regulatory proteins act is now an open question, but the lack of any additive effect of the double Rab10/Rab8 knockdown indicates that they are likely to act on the same pathway. Our localization data suggest that they operate at the level of the Golgi apparatus, largely because this is where their cellular concentration is highest. However, while data obtained by expressing tagged Rabs has been proven over time to give correct answers, this is not yet a definitive answer. Further, the localization is not exclusively at the Golgi apparatus, and the peri-Golgi and more peripheral staining seen in Fig. 2 and 5 may reflect the presence of these proteins on recycling pathways, where they are also reported to act, and where they could function to return proteins to the Golgi apparatus for incorporation into newly-forming WPB. That we see no colocalization of Rabs 10 and 8 with formed WPB, unlike in melanocytes, where Rab8A colocalizes with Rab27A on melanosomes [24], is also consistent with the hypothesis that they play an early role in formation of these organelles. One possibility is that they act in the delivery of a component of the exocytic machinery present on WPB, which functions to modulate the WPB response to secretagogue. We note that integral membrane proteins involved in exocytosis such as SNAREs will require recycling after use. However, because only a very few WPB components have so far been identified [5], we have no candidates as yet.

Very few modifiers of von Willebrand’s disease have been identified, although it is estimated that two-thirds of the loci implicated in the genetic determination of plasma VWF levels have not been identified [48]. It is likely that either polymorphisms or mutations affecting the machinery implicated in the formation of WPB have a role in VWD etiology. This machinery includes the adaptor complex AP-1 and its interactors [7,8], as well as the new function we report here of Rab10 and Rab8. Neither map to loci implicated in the genetic regulation of murine plasma VWF levels. However, by modulating the fast release of VWF upon bleeding, they could have a role in mild cases of VWD, which could be predicted to induce a slow response of the clotting system upon injury due to a delay in the formation of a hemostatic plug.


We are grateful to M. O’Connor for helpful comments, and her statistical expertise. We thank G. Lesa, and J. Quintin for help and reagents. We thank the C. elegans Genetic Center for strains. This work was supported by an Avenir starting grant from the Institut National de la Santé et de la Recherche Médicale and a grant from Région Bretagne to G. Michaux and the Medical Research Council (U12260000200001) to D. F. Cutler.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.