Occludin Acts as a Dynein Adaptor Regulating Permeability and Collateral Angiogenesis Authors: Xuwen Liu, Enming J Su, Alyssa Dreffs, John P Gillies, Madeline Merlino, Lu Gao, Julian S Peregoff, Cheng-mao Lin, Margaret Elizabeth Ross, Morgan E DeSantis, Daniel A Lawrence, David A Antonetti Published: 2025 ABSTRACT AbstractPrevious studies of the tight junction protein occludin (OCLN) suggest that multiple phosphorylation sites on the carboxy-terminal domain contribute a regulatory role in vascular barrier properties. However, gene deletion studies failed to identify a clear functional role for OCLN, despite multiple phenotypic alterations. Importantly, previous studies targeting exon 3 allowed expression of a splice variant starting at exon 4, (isoform 4) that expresses the full carboxy-terminal tail. Here we show that the OCLN carboxy-terminus forms a complex with the light intermediate chain (LIC) of dynein to link tight junction cargo to the minus end directed motor protein. Mutational analysis revealed S471 phosphorylation promotes binding to the LIC while S490 phosphorylation is required for trafficking. Expressing OCLN S490A mutant prevented endothelial cell proliferation and collateral angiogenesis. Ocln gene deletion targeting exon 5, preventing full length and isoform 4 expression, resulted in embryonic lethality. In summary, OCLN links tight junction cargo to the dynein motor, regulating trafficking in a phosphorylation dependent manner and contributing to both vascular endothelial growth factor (VEGF)-induced vascular permeability and collateral angiogenesis.Classification: Biological Sciences, Cell BiologyKeywords: Tight Junctions, Vascular Permeability, Angiogenesis, Diabetic Retinopathy, Stroke Introduction OCLN was the first transmembrane tight junction protein identified1; however, the function of occludin has remained elusive. OCLN has sequence homology to MARVEL (myelin/lymphocyte (MAL) and related proteins for vesicle trafficking and membrane link) proteins with conserved sequence adjacent to the four-transmembrane regions that are characterized by their involvement in vesicle trafficking2. OCLN is now also recognized as among the TAMP or tight junction associated MARVEL protein family including MARVELD1 or OCLN, MARVELD2 or tricellulin and MARVELD3. OCLN (522aa) consists of a 66aa N-terminal cytoplasmic domain, transmembrane (TM)1 (23aa), extracellular loop (EL)1 (46aa), TM2 (25aa), intracellular loop (10aa), TM3 (25aa), EL2 (48aa), TM4 (22aa) and a long carboxy terminus of 257aa3. The carboxy terminus includes a coiled coil domain (CC residues 413–522) that binds to ZO-14, 25 and 36. This carboxy terminus is heavily phosphorylated in the CC domain and adjacent to the CC (reviewed in3,7). OCLN phosphorylation has been shown to regulate barrier properties. There is evidence for Tyr phosphorylation regulating interaction with the tight junction organizing protein zonula occludens 1 (ZO-1)8 and Ser/Thr phosphorylation regulating epithelial junction integrity9 assembly10, and paracellular permeability through dynamic changes in tight junction protein interactions11. Multiple phosphorylation sites were identified in vascular endothelial cells through mass spectrometry analysis12, two of which are in the CC domain and have been found to have functional significance demonstrated by mutational analysis. Phosphorylation of the S471 site in OCLN is required for post-confluence, size reductive proliferation and for maturation of epithelial cell junctional complex13. Phosphorylation of S490 by PKCβ is stimulated by vascular endothelial growth factor (VEGF) in endothelial cells and regulates permeability14,15. Expression of S490A acts in a dominant manner to block VEGF-induced permeability in endothelial cells14,15 and expression of this mutant in transgenic animals inhibits VEGF- or diabetes-induced retinal vascular permeability preserving visual function16 and reduces stroke induced-permeability and allows late tissue plasminogen activator treatment of stroke without hemorrhagic transformation17. Genetic evidence has clearly demonstrated a role for the claudin proteins in barrier properties. Claudin proteins are a large family of tetraspan tight junction proteins of at least 24 different genes in humans that confer barrier properties dependent on which isoforms are expressed with some claudins acting in barrier formation and others as charge and size selective paracellular pores18–20. Claudin 5 expression is restricted to vascular endothelial cells and has been the only claudin shown to confer barrier properties to the blood-brain barrier through transgenic studies as claudin 5 gene deletion leads to death within 10h after birth associated with flux of molecules <800Da into the brain21,22. Studies of gene deletion of Ocln have been less informative regarding the role of this gene product. Gene deletion through targeting exon 3 with flanking lox (flox) sites revealed mice with a host of phenotypic alterations such as brain calcification, male infertility, and gut hyperplasia of the gastric epithelium and thinning of compact bone but tight junctions and barrier properties appeared intact23,24. However, more recent data reveals that these Oclnfl3/fl3 transgenic mice allowed expression of a splice variant from an internal ribosome entry site at exon 4 called μN or isoform 4 (Iso4) OCLN that includes 14 amino acids in the transmembrane region and the carboxy-terminal tail including phosphorylation sites S471 and S49025. Multiple studies have demonstrated that OCLN contributes to cell proliferation in many cell types. OCLN, phosphorylated at S490, was localized to cell centrosomes and shown to be necessary for VEGF driven EC proliferation in culture26, and in epithelial cells the S490A mutation delays mitotic entry27. Phosphorylation of the S471 site in OCLN is required for post-confluence, size reductive proliferation and for maturation of epithelial cell junctional complex13. Further, Oclnfl3/fl3 deletion contributes to gastric epithelium hyperplasia24 and OCLN was identified as an anti-oncogene28. In humans, various OCLN mutations lead to microcephaly, brain calcification and renal disease29–31. Analysis of cortical neural progenitor cell proliferation reveals OCLN localizes to centrosomes and deletion of full-length OCLN leads to alterations in mitotic spindle pole organization reducing progenitor self-renewal and increasing apoptosis in both mice and human brain organoids25. Together, these studies provide compelling evidence that OCLN contributes critical activities in regulating cell growth and proliferation, specifically in cells that possess junctional complexes and polarization. In the current study, we provide evidence that the OCLN carboxy terminus acts to link the tight junction complex to the microtubule minus end directed motor dynein, allowing internalization of the junctional complex and paracellular permeability. We find that OCLN also traffics to the centrosome to support cell proliferation. Expression of S490A OCLN in mice prevents post-ischemia collateral vessel growth in brain and retina. OCLN mutational analysis reveals the carboxy terminus is necessary for trafficking to centrosomes in a S490 phosphorylation dependent manner. Further the C-terminus of OCLN co-precipitates with the LIC of dynein. S490A Iso4 of OCLN can act in a dominant negative manner blocking VEGF-induced permeability and proliferation of endothelial cells like full length S490A OCLN. Finally, a new gene deletion model targeting exon 5 (Oclnfl5/fl5), that ablates expression of both full length and Iso4 OCLN, is embryonic lethal associated with blood vessel rupture and hemorrhaging with some animals demonstrating gastroschisis. Results Expression of OCLNS490A impedes collateral vessel growth. We previously showed that expression of OCLNS490A mutant in endothelial cells prevented diabetes induced retinal vascular permeability16 and stroke induced permeability in the brain and prevented late tissue plasminogen activator (tPA)-induced hemorrhagic transformation17 clearly demonstrating a role for this phosphorylation of OCLN in regulation of vascular permeability. To determine whether OCLNS490 phosphorylation was required for collateral vessel angiogenesis in the transgenic animals, we developed a laser induced model of branched retinal vein occlusion (BRVO). Retinal veins were blocked by a brief pulse from a 532nm laser using a Micron III retinal imaging system and collateral vessel growth was monitored in retinal whole mounts at various times after injury. We examined vessel growth in OCLNS490A+/+ mice compared to WTOCLN with expression controlled by a floxed stop site after the CAG promoter (described in16). Inducible PDGFiCre was used for vascular endothelial specific expression. Mice were Oclnfl3/fl3 with deletion of endogenous full length Ocln but maintained Iso4 Ocln after Cre excision and simultaneous expression of S490A or WT transgene. Immunofluorescence staining of the retinal whole mount in littermate control (LM Ctrl) animals on day 7 after BRVO revealed dramatic increase in cell proliferation indicated by Ki-67 positive staining, some co-labeled with CD31 while others were in the retinal parenchyma (Fig. 1A), clearly indicating both endothelial and non-endothelial proliferation. The number of Ki-67 positive cells in superficial plexus was significantly increased in LM Ctrl, Cre+, and WTOCLN groups after BRVO as compared to C57BL/6 control (B6 Ctrl) non-laser group (Fig. 1B), while the increased number of Ki-67 positive cells was significantly inhibited in S490A group after BRVO (Fig. 1B and SI Appendix, Fig. S1A, B and C). The result indicates that endothelial restricted expression of transgenic WTOCLN had no effect on this proliferation but endothelial expression of OCLNS490A mutant prevented the increase in cell proliferation of both endothelial and non-endothelial cells in BRVO. Previous studies suggested the time frame of 2–4 weeks for retinal collateral vessel growth32,33. Therefore, we evaluated the role of OCLNS490A on collateral vessel growth at 4 weeks after BRVO. Immunofluorescence staining of retinal wholemounts with CD31 antibody revealed there were more vessels surrounding the occluded vein in the superficial plexus in LM Ctrl after BRVO as compared to the non-laser group and OCLNS490A inhibited this vessel proliferation (Fig. 1C). Quantitative analysis demonstrated that expression of OCLNS490A mutant completely prevented the BRVO induced increase in capillary collateral vessel area and length observed in the Cre- LM Ctrl (Fig. 1D and E). The proliferation experiments were repeated with endogenous Ocln background, (i.e. without Ocln exon 3 excision). Again, expression of OCLNS490A prevented BRVO-induced cell proliferation almost identical to the studies with exon 3 excision (Fig. 1F and G), indicating the dominant effect of OCLNS490A in the presence of full length and Iso4 Ocln. We also explored the laser-induced choroidal neovascularization (CNV) angiogenesis mouse model and observed a modest but not significant loss of vessel growth with OCLNS490A expression in these vessels (SI Appendix, Fig. S2A and B). It is noteworthy these vessels have limited tight junction expression compared to retinal vessels. Expression of occludin S490A impedes collateral angiogenesis. A. Whole mount retinal images of CD31 and Ki-67 immunostaining. Scale bar: 50 μm. B. Quantification of Ki-67 positive cells. Mouse lines; B6 Ctrl: C57BL/6J; littermate (LM) Ctrl: Pdgfb-iCreER−, Oclnfl3/fl3, OCLNS490A+/+; Cre+: Pdgfb-iCreER+, Oclnfl3/fl3; WT: Pdgfb-iCreER+, Occfl3/fl3, WTOCLN+/+; S490A: Pdgfb-iCreER+, Oclnfl3/fl3, OCLNS490A+/+. C. Confocal images of CD31 stained vessels at 4 weeks after BRVO showing targeted vein (V) and image after removal of V using Photoshop (Adobe). Scale bars: 50 μm. (D and E) Quantitation of vessels using AngioTool. Expression of OCLN S490A reduced endothelial cell proliferation in mice carrying endogenous Ocln at 1 week after BRVO. (F) Representative images, scale bar: 50 μm and (G) quantification. Mouse lines: Cre+: Tek-Cre+, OCLNS490A−/−; S490A: Tek-Cre+, OCLNS490A+/+. (H) Retinal explants with IB4 (green) and Hoechst (blue) immunostaining. 16 retinal explant cultures were established per eye, 6 eyes/group. Scale bar: 50 μm. (I and J) Quantification by MetaMorph software. LM Ctrl: Pdgfb-iCreER−, Oclnfl3/fl3, OCLNS490A+/+; S490A: Pdgfb-iCreER+, Oclnfl3/fl3, OCLNS490A+/+. (K) Representative images of CD31 stained vessels in peri-infarct areas at 3 weeks after stroke. Scale bar: 50 μm. (L) Quantitative analysis of the images shown in Panel K using AngioTool image software. Four regions in each section were quantified and three sections per mouse were evaluated (12 images per mouse) with n=7 independent mice in each of LM Ctrl and S490A groups. LM Ctrl: Pdgfb-iCreER−, Oclnfl3/fl3, OCLNS490A+/+; S490A: Pdgfb-iCreER+, Oclnfl3/fl3, OCLNS490A+/+. Contra: Contralateral region, Ipsi: Ipsilateral region. (M) MRI imaging of cerebral blood volume. LM Ctrl and S490A mice underwent MCAO and 21 days later, T2-weighted MRI images were obtained from before gadolinium injection through 30 min after injection. (N) Regions of interest from the infarct zone were selected and computed for cerebral blood volume. Mouse lines are the same as K and L. Quantification by ANOVA and Sidak post-test (B, D, E, and L), by student t-test (G, I, J and N). The effect of OCLNS490A expression on vessel growth was supported by an ex vivo retinal explant experiment. As shown in Fig. 1H, vascular outgrowth from retinal fragments was significantly decreased from OCLNS490A mice compared to LM controls. Quantification of isolectin B4 (IB4) stained vessels revealed a decrease in total vessel outgrowth and mean process length with OCLNS490A expression (Fig. 1I and J). These studies reveal the inhibitory effect does not require circulating endothelial precursors as the inhibitory effect of OCLNS490A expression on angiogenesis was observed in the ex vivo model. Lastly, we tested if OCLNS490A had similar inhibitory effect on angiogenesis in the brain after stroke. Stroke was induced by photothrombotic middle cerebral artery occlusion (MCAO) and immunofluorescence staining of brain cross sections with CD31 antibody was completed followed by vessel area quantification. In LM control animals MCAO induced a 2.5-fold increase in vessel area in the peri-infarct region at 21 days after stroke compared to the contralateral region. However, the increased collateral angiogenesis was significantly reduced to 1.7-fold increase in OCLNS490A expressing mice (Fig. 1K and L). To corroborate this finding the relative tissue blood volume, as a measure of blood circulation, in the infarct region was measured after MCAO by MRI imaging (Fig. 1M). LM control mice had significantly more tissue blood volume in the infarcted area at 3 weeks after MCAO compared to the animals expressing OCLNS490A in the vascular endothelium (Fig. 1N). Taken together, these data provide compelling evidence that OCLN phosphorylation on S490 contributes to endothelial cell growth and collateral angiogenesis in the central nervous system (CNS) as expression of the phospho-inhibitory OCLNS490A inhibits angiogenesis in vessels with well-developed tight junctions. OCLN localizes at centrosomes in a membrane vesicle requiring Iso4 and Ser490 phosphorylation. Our previous studies have demonstrated that S490 phosphorylated OCLN localizes to centrosomes throughout mitosis and S490 phosphorylation is required for mitotic entry in epithelial cells34 and VEGF-induced retinal neovascularization26. Here, OCLN centrosomal localization was further explored in U2OS cells by mutational analysis to identify the regions necessary and sufficient for targeting centrosomes. The U2OS cell line has large and clearly observable centrosomes aiding analysis. First, we performed immunofluorescence staining with centrosome marker γ-tubulin antibody and antibodies against OCLN N-terminus, C-terminus and pS490 with all 3 showing clear centrosome co-localization (Fig. 2A). Next, we made a series of epitope-tagged OCLN mutants to determine which domains are necessary for centrosomal localization of OCLN (Fig. 2B). U2OS cells were transfected with OCLN mutants with N-terminal GFP or C-terminal V5 tags. Cells were immunostained for GFP or V5 tags, centrosomal marker γ-tubulin and DNA with Hoechst dye (SI Appendix, Fig. S3A and B, and quantification Fig. 2C and 2D). Both N-terminal GFP tagged (GFP-WT) or C-terminal tagged V5 (WT-V5) forms of OCLN localize to centrosomes. Mutants of the N-terminus through the first and second extracellular loop had no effect on centrosomal localization (Fig. 2C). However, OCLN centrosomal localization was completely blocked by S490A mutation (GFP-S490A) and in the mutants lacking C-terminal coiled-coil domain (GFP-ΔCC). The C-terminal CC domain (GFP-CC) alone failed to lead to centrosomal co-localization (Fig. 2D). The full C-terminal tail, either as N-terminal GFP-tagged or C-terminal V5-tagged, was also not sufficient to yield centrosomal co-localization (Fig. 2D and E). However, when we created a V5 tagged mimic of the alternatively spliced Iso4 OCLN that possesses the full C-terminal tail plus 14 transmembrane amino acids, we observed centrosomal colocalization like V5 WT-OCLN. Finally, we added a myristylation motif to the C-terminal mutant (Myr-Cter-V5) to provide a membrane anchor. The result indicated that Myr-Cter-V5 restored OCLN centrosomal localization to near 50% (Fig. 2E). Collectively, these studies demonstrate that OCLN centrosomal localization requires membrane association, C-terminal coiled-coil domain, and S490 phosphorylation. Further, the Iso4 splice variant, which contains all these features, traffics robustly to centrosomes. Occludin localization at centrosomes requires membrane component and C-terminal coiled-coil including Ser490 phosphorylation. (A) Representative confocal images of U2OS cells immunostained with rabbit anti-N-terminal OCLN antibody, mouse anti-C-terminal OCLN antibody, or rabbit anti-OCLN pS490 antibody (all in red) with centrosomal marker γ-tubulin antibody (green) and Hoechst staining for DNA (blue). Scale bar: 5 μm. (B) Schematic diagram of GFP or V5 epitope-tagged full-length OCLN and mutants. (C-E) Quantification of percent of transfected cells demonstrating co-localization with γ-tubulin at centrosomes. Minimum 30 cells of each condition were analyzed by Chi-Square analysis. Black bars, complete colocalization; open bars, no colocalization. (C) Compared to GFP-WT, **P <0.01, ***P<0.001 and ****P<0.0001. (D) compared to GFP-WT ****P<0.0001. (E) Compared to WT-V5 ****P<0.0001, compared to C-Ter-V5 ####P<0.0001. OCLN Iso4 and S490 phosphorylation regulate VEGF-driven proliferation. We next determined whether the dominant S490A mutation was effective in controlling VEGF-driven proliferation in Iso4 OCLN as we previously observed for full length OCLN26. Bovine retinal endothelial cells (BREC) were transfected with plasmids containing OCLN mutants and compared to cells with wild-type OCLN or empty vector. Measures of endothelial proliferation in response to VEGF were carried out by Click-iT EdU DNA synthesis assay in 3D collagen culture. This analysis revealed that transfection of OCLNS490A inhibited VEGF-induced DNA synthesis as reported previously26 (Fig. 3A). In addition, VEGF-induced proliferation was inhibited in S490A Iso4 mutant (Iso4-V5 S490A) (Fig. 3B). OCLN mutants lacking extracellular loop 1 (GFP-ΔNL1) or with additional mutations to loop2 (GFP-ΔNL1L2NHE and GFP-ΔNL1L2NHE S490A) all effectively blocked VEGF-induced proliferation (Fig. 3C). As both GFP-S490A and GFP-ΔNL1 inhibited VEGF-induced DNA synthesis but OCLN was localized to centrosomes in GFP-ΔNL1 but not GFP-S490A transfected cells, cell cycle analysis in 3D collagen culture in the presence of VEGF was performed to determine if these OCLN mutants effected cell cycle progression. Cell cycle analysis demonstrated that OCLN GFP-ΔNL1 expression shortened prophase and prometaphase and prolonged metaphase to telophase as compared to GFP-EV, GFP-WT or GFP-S490A OCLN (Fig. 3D). But expression of OCLNS490A did not alter cell cycle progression. Further apoptosis assay showed that GFP-ΔNL1 mutation induced apoptosis as compared to GFP control, or WTOCLN (Fig. 3E). S490A mutant had a more modest increase in apoptosis compared to GFP-control. Finally, we demonstrated that the Iso4-S490A mutant also blocked VEGF induced permeability in primary endothelial cells as previously reported for full length OCLNS490A14 (Fig. 3F). However, full length or Iso4 OCLN and S490A mutants reveal similar basal permeability and co-localization with ZO-1 (SI Appendix, Fig. S4A–C). Taken together, these studies demonstrate that expressing OCLN Iso4 with a S490A mutation is sufficient to block both VEGF-induced endothelial cell permeability and proliferation and that Iso4 is necessary and sufficient to traffic to centrosomes. Meanwhile, deletions of the N-terminus of OCLN affecting loop1, impacts cell division and lead to apoptosis consistent with the full length Ocln gene deletion and impaired neural progenitor cell division previously reported25. Endothelial cell division and permeability are regulated by occludin Ser490 phosphorylation. (A-C) BREC were transfected with GFP, WT-OCLN (GFP-WT) or mutants and DNA synthesis was measured by Click-iT EdU at 8 h after VEGF treatment. Quantification by ANOVA and Sidak post-test from at least 3 independent experiments with each experiment from 6 individual wells in a 24-well plate yielding n of 1. (A) Expression of S490A mutation in full length OCLN or (B) S490A mutation in Iso4 OCLN, inhibits VEGF driven proliferation as does (C) loss of extracellular loop 1 of OCLN. (D and E) BREC were transfected with GFP, WT or S490A OCLN and GFP-ΔNL1. (D) Cell cycle analysis was performed at 8 h after VEGF treatment. Quantification from 3 independent experiments, each experiment consists of the cells from 2 individual wells in a 24-well plate, n=individual cell examined. Results are expressed as the percentage in each group with Chi-Square analysis. Expressing delta-loop 1 of OCLN altered cell cycle. (E) Detection of apoptotic cells by Click-iT TUNEL assay was measured at 8 h after VEGF treatment. Quantification from 3 independent experiments with n = transfected cells examined with Chi-Square analysis. (F) VEGF-induced permeability was measured in BREC transfected with empty vector, S490A in full length OCLN (GFP-S490A) or S490A in Iso4 (Iso-S490A-V5) and analyzed by ANOVA and Sidak post-test. S490A mutation reduced OCLN dynamics. Previous studies revealed that OCLN-containing vesicles move along microtubes (MTs)35,36 and together with the observed centrosomal localization, suggested that OCLN may traffic towards the minus-end of microtubules on dynein. Immunofluorescence staining of BREC with the antibodies against dynein, OCLN pS490, and α-tubulin demonstrated the co-localization of dynein and pS490 OCLN along microtubules. When cells were treated with Dynapyrazole to stall the dynein motor protein, the amount of co-localization was even greater (Fig. 4A). In addition, live cell trafficking experiments were carried out in U2OS cells transfected with GFP tagged WTOCLN or S490A or the phospho-mimetic S490D mutants using spinning disk microscopy (Fig. 4B). Vesicles that ran longer than 5 seconds with displacement above 0.5 μm were evaluated for distance, displacement and speed. The results indicated that S490A mutation had modest but significant reduction in distance and displacement compared to wild-type OCLN and S490D mutants. (Fig. 4C). However, the largest impact was observed on the percentage of vesicles with displacement above 0.5 μm that was significantly reduced in GFP-S490A transfected cells compared to GFP-WT and GFP-S490D (Fig. 4C and D). These studies demonstrate colocalization of OCLN and dynein and reveal that S490A mutant reduces OCLN trafficking. S490A mutation reduced OCLN dynamics. (A) Immunofluorescence (IF) staining with occludin pS490 (red), dynein (green), and α-tubulin (white) in BREC. Top panel treated with vehicle DMSO. The arrows shown in right zoomed-in panel indicate the colocalization of pS490, dynein, and α-tubulin. BREC treated with 10 μM of Dynapyrazole A for 6h (bottom panels). Scale bars: left panels, 3 μm; right zoomed-in panels, 1 μm. (B) Representative still images showing wildtype GFP-OCLN transfected U2OS cells during live cell imaging experiment and the moving tracks identified by Imaris software for quantification. (C) Quantification of vesicle speed, distance, displacement with each data point representing a vesicle with greater than 0.5 μm displacement. (D) Percent of vesicles with greater than 0.5 μm displacement per video from 2–4 transfections with each data point representing the data from a video. Quantification by ANOVA and Sidak post-test (C and D). OCLN interacts with the LIC of the motor protein dynein. To explore the role of dynein in regulation of OCLN trafficking, we performed co-immunoprecipitation (co-IP) studies in U2OS cells and primary BREC. IP of dynein targeting the intermediate chain (IC) co-precipitated OCLN as detected by both anti-OCLN (C-terminus) and pS490 Abs in U2OS cells (Fig. 5A, SI Appendix, Fig. S5A) and in both confluent and nocodazole arrested, mitotic BREC cells (Fig 5D, SI Appendix, Fig. S5D). Since many cargo adaptors specifically interact with LIC 1 or 2 of dynein37–40, we assessed occludin co-IP for this specific subunit of dynein. An OCLN IP led to co-precipitation of both LIC2 (Fig. 5B and E, SI Appendix, Fig. S5B and E) and LIC1 (Fig. 5C, SI Appendix, Fig. S5C). Occludin interacts with motor protein dynein LIC2. Immunoprecipitation (IP) of U2OS cell lysates with (A) anti-dynein IC or control IgG antibody (Ab). The input lysates (In) and IP eluates were immunoblotted (IB) with anti-dynein, anti-OCLN pS490 and anti-OCLN C-terminal Ab. (B) IP with anti-OCLN C-terminal or IgG Ab and IB with anti-OCLN, or anti-dynein LIC2. (C) IP with anti-OCLN C-terminal or IgG Ab and IB with mouse anti-OCLN, or rabbit anti-dynein LIC1 Ab. (D and E) IP from confluent (C) or mitotic (M) BREC. Mitotic BREC were treated with VEGF (50 ng/ml) at 70% confluence for 30 min then synchronized with nocodazole (200 ng/ml) for 16 h. (D) IP with mouse anti-dynein IC or IgG Ab and IB with mouse anti-dynein IC, OCLN pS490, and OCLN C-terminal Ab. (E) IP with anti-OCLN C-terminal or IgG Ab and IB with anti-OCLN, anti-OCLN pS490, and anti-dynein LIC2 Ab. (F) Schematic diagram of OCLN Iso4 with CC1-like box containing both S471 and S490 sites located in OCLN coiled-coil region. (G) Sequence alignment showing coiled-coil region of OCLN has homology with known dynein activators. Asterix represents known CC1-box homology, blue regions reveal consensus with OCLN. OCLN mutants are highlighted in green. (H) U2OS cells co-transfected with LIC2-HA and GFP tagged wildtype OCLN (GFP-WT) and IP with rat anti-HA mAb. The In and IP eluates were IB with rabbit anti-GFP pAb and rat anti-HA mAb. (I) U2OS cells co-transfected with LIC2-HA and GFP-WT or its mutants and IP with rat anti-HA mAb. The In and IP eluates were IB with rabbit anti-GFP pAb and rat anti-HA mAb. (J) Quantification by ANOVA and Sidak post-test from at least 3 independent experiments. (K) Representative confocal images of U2OS cells transfected with GFP-tagged OCLN showing partial or no co-localization of GFP-OCLN and γ-tubulin as compared to complete colocalization group. Scale bars: 5 μm. (L) Quantification of γ-tubulin and OCLN colocalization by Chi-Square analysis. The degree of colocalization was determined by the percentage of the area of GFP Signal colocalized with γ-tubulin over the area of total GFP signal. Complete colocalization: 70–100%; Partial colocalization: above 0% and <70%; No colocalization: 0%. There are four canonical dynein adaptor families, which are defined by the domain with which they bind to the LIC. Two of the most well studied adaptor families include those with a HOOK domain, for example HOOK microtubule tethering protein and those with a coiled coil 1 (CC1)-box, highlighted in bicaudal D homologue 2 (BICD2). The CC1 family interacts with the LIC via an AAXXG motif (where x denotes any amino acid) in the CC1-box. Mutations of this region can disrupt binding with LIC39. Sequence alignment of OCLN and known dynein CC1-box containing adaptors (adapted from39) revealed homology in OCLN CC domain that we referred to as CC1-like box. OCLN possesses an AAXXY rather than AAXXG motif plus additional novel sites of homology to the previously identified consensus. In all, 8 of 12 CC1-box consensus sites were conserved in OCLN and two additional sites of conservation were noted (Fig. 5F and G). Strikingly, these were found to include the S471 and S490 OCLN phosphorylation sites. To determine which region of OCLN interacts with dynein, capture assays were performed in U2OS cells that were transfected with GFP-tagged OCLN mutants and HA-tagged LIC2. As indicated in Fig. 5H and SI Appendix, Fig. S5F, LIC2-HA pulldown clearly co-precipitated WT-OCLN. Mutation of the 2 conserved alanine residues (A477 and A478) to valine (GFP-AAVV) decreased the LIC2 capture of OCLN by 50% (Fig. 5I and J, SI Appendix, Fig. S5G). Both S490D and S490A mutations had only a modest 25% reduction in LIC2 capture. However, when we mutated 7 conserved sites of the OCLN CC1-like box (the two central Ala to Val (A477V, A478V), S471 and S490 (S471A and S490A) and 3 charged amino acids (E469A, E473A, K488A) a complete loss of binding to LIC2 was observed (Fig. 5I and J). Further, more than 80% reduction of OCLN co-IP in LIC2 capture was found in both S471A and S471D mutants using an LIC2-HA capture (Fig. 5J) or a LIC2-flag capture (SI Appendix, Fig. S6A and B) suggesting phosphorylation of this site per se was needed for complex formation with LIC. We then tested how these mutations altered trafficking to centrosomes. GFP-OCLN co-localization with γ-tubulin was determined in U2OS cells as above. However, we noted partial co-localization and catalogued this as a separate group. The AAVV mutant reduced co-localization by 50% while S471A reduced co-localization by 25% (Fig. 5K and L). Consistently, the S471D mutant reduced co-localization by 40% while S490D reduced co-localization by only 20% (SI Appendix, Fig. S6C). Notably, none of these mutations were as robust in blocking centrosomal co-localization as was S490A suggesting the AAVV and the S471A mutant fail to act in a dominant manner like the S490A mutant. We developed additional capture studies to test whether the coiled coil was necessary or sufficient for binding. U2OS cells were transfected with OCLN GFP-C-Ter mutants and HA-tagged LIC2. As indicated in Fig. 6A, LIC2-HA successfully co-precipitated OCLN C-Ter as well as full length OCLN. However, LIC2-HA failed to co-precipitate occludin CC alone (SI Appendix, Fig. S7). Moreover, the binding to LIC2-HA was reduced by adding a glutathione S-transferase (GST) tag at the N-terminus of OCLN C-terminal (GFP-GST-C-Ter) to promote dimerization. Further, LIC2-HA pulldown still co-precipitated OCLN lacking the CC domain (GFP-ΔCC) (Fig 6B). These studies suggest the carboxy terminus has a second binding site for complex formation with LIC that does not require the coiled-coil domain. However, complex formation with endogenous OCLN cannot be ruled out. OCLN interacts with LIC2 through the coiled-coil domain and a separate binding site(s). (A) U2OS cells were co-transfected with LIC2-HA and N-terminal GFP tagged wild type OCLN (GFP-WT, arrow), GFP C-terminal OCLN (GFP-C-Ter, arrowhead), or GST forced dimerized C-terminal OCLN (GFP-GST-CTer, empty arrow) and IP with rat anti-HA mAb. The input lysates and IP eluates were IB with anti-GFP, anti-HA, and mouse anti-β-actin Ab. (B) U2OS cells were co-transfected with LIC2-HA and N-terminal GFP tagged wild type OCLN (GFP-WT, arrow) or GFP-WT with coiled-coil region deleted (GFP-ΔCC, empty arrowhead) and IP with rat anti-HA mAb. The input lysates and IP eluates were IB with anti-GFP Ab, anti-HA Ab. (C) U2OS cells were co-transfected with 3xFLAG tagged dynein LIC2 (WT LIC2-FLAG), or LIC2 mutant F432A/F433A (FFAA LIC2-FLAG), or control vector (pCMV-FLAG) and co-transfected with pcDNA5-StrepII-Halo-BicD2-V5-FRB (BiCD2-V5). 3xFLAG tagged LIC2 was captured with anti-FLAG M2 Ab and protein G Sepharose beads then eluted by 3xFLAG peptide. The input lysates and IP eluates were IB with anti-V5 and anti-FLAG M2 Abs. (D) U2OS cells were co-transfected with WT LIC2-FLAG, FFAA LIC2-FLAG, or pCMV-FLAG and co-transfected with GFP or GFP tagged wild type OCLN (GFP-OCLN) for 48 h. Cell lysates were IP for 3xFLAG tagged LIC2 with Protein G Sepharose beads/anti-FLAG M2 mAb and eluted by 3xFLAG peptide. The input lysates and IP eluates were immunoblotted with anti-GFP, anti-FLAG M2 mAb. (E and F) Quantification by Student t-test from six (E) or eight (F) independent experiments. The studies suggested that the dynein LIC interaction with OCLN may be different compared to the other known dynein adaptors. To test this hypothesis, we mutated the conserved hydrophobic residues F447 and F448 to alanine within LIC2 (FFAA LIC2) that was previously reported to significantly reduce LIC1 and BICD2 interaction39. In agreement with the previous study, FFAA LIC2-FLAG failed to co-precipitate BICD2 as compared to its WT LIC2-FLAG control (Fig. 6C and E). In contrast, no significant difference was found between OCLN binding to FFAA LIC2-FLAG or its WT LIC2-FLAG control, indicating a distinct binding mechanism (Fig. 6D and F). Collectively, these data provide compelling evidence that OCLN interacts with LIC2 by the CC1-like box, and a second site in the carboxy terminus. Whether this represents direct interaction or a larger complex remains undetermined at this point. Homozygous deletion of full length and Iso4 Ocln is embryonic lethal and leads to blood vessel ruptures. The data indicates that OCLN carboxy terminus interacts with dynein through the LIC allowing trafficking in a phosphorylation dependent manner. Previous gene deletion targeting exon 3 failed to remove Iso4 Ocln25. Therefore, we created a new mouse targeting exon 5 of Ocln to delete both full length and Iso4 Ocln. Mice were created in the University of Michigan transgenic core by CRISPR Cas9 targeted insertion of a gene cassette carrying LoxP recombination sites flanking exon 5 (Oclnfl5/fl5) and were sequence verified. Oclnfl5/fl5 mice were crossed with mice carrying Ella Cre to induce germline deletion of full length and Iso4 Ocln (Fig. 7A). Reduction of full length 55kDa, and modified 70kDa OCLN was confirmed by western blot analysis using C-terminal (Fig. 7B) and N-terminal (Fig. 7C) targeted antibodies, consistent with partial penetrance of Ella Cre41. Reduction of Iso4 was observed with the C-terminal Ab only, as expected. Germline deletion of Oclnfl5/fl5 was found to be embryonic lethal (Fig. 7D). An analysis of embryos revealed normal Mendelian ratios at E14.5. However, by E18.5 the number of homozygous knock out animals began to decrease and no Cre+ Oclnfl5/fl5 pups were observed at birth. Random blood vessel ruptures leading to hemorrhaging were observed in Cre+ Oclnfl5/fl5 embryos examined. These ruptures occurred in the brain (Fig. 7E, SI Appendix, Fig. S8), liver (Fig. 7F), and umbilical vein and artery (SI Appendix, Fig. S9). In addition, extensive blood accumulation was found in the pericardium. These ruptured vessels were considered the most likely cause of death. Two animals were observed with gastroschisis, with failure of proper midline closure and intestine exiting the cavity (Fig. 7G). The distinct difference from deletion of exon 5 that targets both full length and Iso4 Ocln and deletion of exon 3 targeting only full length Ocln highlights the essential role of the carboxy-terminus/Iso4 in development. Homozygous deletion of Ocln from flex5 is embryonic lethal. (A) Ocln gene schematic showing full length and Iso4 start sites with exon 3 floxed (fl3, green arrows) or the newly created exon 5 floxed (fl5, yellow arrows). Oclnfl5/fl5 mice were crossed with germline Cre (E2A-Cre), Oclnfl5/+ mice to induce deletion of FL and Iso4 Ocln. (B and C) Protein was extracted from E15.5 whole embryos of Cre− (Ctrl) or Cre+ Oclnfl5/fl5 mice followed by IB with (B) anti-C-terminal OCLN Ab, (C) anti-N-terminal OCLN Ab, or anti-β-actin Ab. Known location of full-length and Iso4 mouse OCLN are indicated with the arrows and arrowheads respectively. (D) No Cre+, Oclnfl5/fl5 pups were observed at birth. An analysis of embryos revealed normal Mendelian ratios at E14.5. However, by E18.5 the number of homozygous knockout animals began to decrease, Chi-Square analysis. (E-G) Analysis of the embryos reveals random events of blood vessel rupture leading to blood accumulation. These ruptured vessels were considered a likely cause of death. (E) H&E image showing cranial hemorrhage in E14.5 flex5 KO (sagittal). (F) H&E image showing liver/abdominal hemorrhage in E18.5 flex5 KO (coronal). (G) H&E image showing gastroschisis in E14.5 flex5 KO (sagittal). Discussion The current studies reveal that the OCLN carboxy terminus forms a complex with dynein through the LIC. Mutational analysis suggests that S471 phosphorylation4,12 contributes to binding LIC while S490 phosphorylation was necessary for trafficking as S490A mutations blocked centrosomal localization and reduced live cell trafficking of OCLN. Expression of a construct mimicking Iso4 splice variant of OCLN with the S490A mutation, blocked VEGF-induced cell proliferation and vascular permeability. Importantly, a novel gene deletion of Ocln targeting exon 5 that deleted both full length and Iso4 Ocln, was embryonic lethal. Previous studies targeting exon 3 were not lethal, revealing the requirement for the carboxy-terminus of OCLN in development. Expression of S490A OCLN mice appears to allow normal development but restricts VEGF induced permeability16,17 and impairs collateral angiogenesis in the brain and retina. We propose that the OCLN carboxy terminus acts to link a vesicle of tight junction cargo to the dynein motor to allow trafficking from the cell border to the interior, leading to tight junction (TJ) strand breaks and vascular permeability. Further, OCLN trafficking to the centrosome appears essential for VEGF driven proliferation and collateral vessel formation. Previous studies have explored TJ protein trafficking. Early studies revealed PDGF induced endocytosis and TJ strand breaks in Madin-Darby canine kidney (MDCK) cells42. OCLN was found to traffic on microtubules as observed by live cell imaging and dynein inhibitors led to accumulation of OCLN in the cytoplasm, suggesting a role for dynein trafficking of OCLN to the plasma membrane35. In epithelial cells, Rab13 and junctional Rab13-binding protein (JRAB), also called molecule interacting with CasL like 2 (MICAL-L2), was required for OCLN and other tight junction protein recycling and barrier formation43,44. Studies also demonstrated junctional trafficking in endothelial cells in response to VEGF that was dependent on OCLN S490 phosphorylation downstream of PKCβ14,15. These studies demonstrated that OCLN may be ubiquitinated downstream of PKCβ phosphorylation and proteins with a ubiquitin interacting motif could chaperon OCLN. However, no evidence was found for ubiquitination of OCLN in co-precipitates with LIC and the Iso4 form of OCLN is lacking the N-terminal cytoplasmic region that interacts with the E3 ligase ITCH45. It is common for transmembrane proteins to possess both non-ubiquitin and ubiquitin-dependent trafficking pathways and OCLN appears to traffic in both pathways. Dynein motor proteins drive minus end directed transport on microtubules. Cytoplasmic dynein-1 is a 1.4MDa complex made of a heavy chain, intermediate chain, LIC and 3 different light chains. All subunits of dynein are present as dimers in a motor40. Approximately 40 kinesin motors drive trafficking to the plus end of microtubules each with various cargo specificities, but dynein alone primarily drives minus end directed microtubule trafficking. Dynein achieves cargo specificity through specific adaptors that link cargo to dynein, in part by binding to the LIC40. The canonical adaptors also act as bona-fide activators as demonstrated by in vitro processive trafficking using purified proteins. To activate dynein, adaptors form a complex with dynein and the co-factor dynactin. In this complex, dynein assumes a conformation that enables processive movement toward the minus end of microtubules and cargo transport37. Activators generally possess a long coiled-coil domain with an LIC interaction site. Currently there are four established families. Two of the most well studied activator families are the Hook domain and the CC1-box domain containing binding motifs39. Most activators possess a second spindly motif binding site, which mediates binding to dynactin46. The CC1-box or Hook domain on the activator binds to an intrinsically disordered region of LIC that forms a transient helix upon binding thus allowing a range of activators to bind LIC38. Strikingly, OCLN possesses homology to the CC1-box within the C-terminal coiled coil domain and this region resides between the two turns containing the S471 and S490 phosphorylation sites. However, the canonical AAXXG motif was instead AAXXY in OCLN and thus we identify it as CC1-like box. Mutational analysis of this region revealed it is indeed necessary for full LIC binding. However, F447 and F448 mutated to alanine within LIC (FFAA LIC2) that abrogates binding to known canonical dynein adaptors38,47, did not alter OCLN binding. Thus, OCLN with a CC1-like box may be more similar to the adaptor Rab7 interacting lysosomal protein (RILP) that binds LIC directly but also binds the homotypic fusion and protein sorting (HOPS) complex linked to dynein48 or NuMA that has a hook domain and CC1-like box domain and interacts with LIC49. The short coiled-coil domain of OCLN and the ability to bind LIC mutated at F447 and F448 suggests OCLN may not act directly to activate processive motility like established dynein activators40. Here, we establish that the coiled-coil domain of OCLN provides a phosphoregulatory role for LIC interaction and dynein trafficking. The strong effect of S471A or D on inhibiting binding suggests that phosphorylation of this site promotes OCLN binding to LIC that could not be substituted by the phosphomimic. The S471 phosphorylation regulates binding to ZO-1 directly4 and contributes to LIC binding either directly or indirectly. S490 phosphorylation was shown here to impact OCLN trafficking and S490A can act in a dominant negative manner. Clearly the possibility exists for competition of OCLN binding to ZO-1 or LIC for stable junction complex versus dynein trafficking which can be regulated by VEGF signal transduction. The studies here are consistent with previous research revealing a role for the N-terminus of OCLN contributing to proliferation. Deletion of full length but not Iso4 Ocln from floxed exon 3 mice contributes to microcephaly and in organoids impaired spindle pole integrity25 and human OCLN mutations lead to microcephaly29–31. Here, deletion through the first extracellular loop (μNL1) led to reduced VEGF driven proliferation and increased apoptosis. However, S490A mutants completely blocked trafficking to centrosomes, prevented VEGF driven proliferation in cells but with limited apoptosis, and blocked collateral angiogenesis in brain and retina, strongly suggesting a required role for OCLN trafficking to centrosomes for endothelial proliferation. A recent study using exon 3 floxed mice revealed that loss of full length OCLN, but not Iso4, retarded mammary gland branching due to reduced cell proliferation associated with OCLN localization at centrosomes50. The investigators revealed OCLN interacts with the RAB11 family interacting protein 5 (FIP5) confirming previous studies identifying OCLN interaction with various endocytosis proteins including RAB11FIP1,2, 5, RAB-5b, RAB-7a, RAB-10 and RAB-23 and vesicle associated membrane proteins 2 and 551. The current and previous studies establish a role for the N-terminus of OCLN in cell proliferation and support OCLN vesicle trafficking leading to centrosomal localization. Here we report for the first time, deletion of both full length and Iso4 Ocln with floxed exon 5. Deletion using a germline Cre led to embryonic lethality with vessel hemorrhaging as a likely cause of death. It was noted in the pathology report that points of hemorrhage appeared at regions of extensive branching. The cause of the midline opening and gastroschises is not known. However, gastroschisis is associated with improper development or rupture of the membrane covering the umbilical canal and umbilical ring52,53. Multiple animals were observed with umbilical artery or vein rupture that could potentially lead to gastroschisis. Other gene deletions also may lead to embryonic vessel hemorrhaging. Endothelial targeted gene deletion of β-catenin allowed normal vasculogenesis and angiogenesis but led to irregular vascular patterning and random vessel hemorrhaging observed at E10.554. Further analysis revealed complete lack of capillary formation in the CNS with improperly formed endothelial aggregates and hemorrhage55. Mutations in polycystin (PC or PKD) 1 or 2 lead to autosomal dominant polycystic kidney disease56. However, mutations of PKD1 also lead to life threatening vascular abnormalities and mouse embryos homozygous for PKD1 mutation (PKD1L) die by embryonic day 15.5 due to edema, vascular leaks and ruptured blood vessels57. Study Limitations: The Oclnfl5/fl5 E2a-Cre+ mice are embryonic lethal. However, vessel rupture only accounts for loss of half of the mice by E18.5, and it is not known why the remaining mice die. However, the mice likely die during parturition or shortly after birth and the cause will be addressed in future studies. Further, future studies will directly test the role of Oclnfl5/fl5 deletion targeted to vascular endothelial cells and the response to angiogenic signals as well as vessel fragility. Finally, it is striking that the OCLN S490A mutant provides such a powerful inhibition on collateral angiogenesis but no observed changes in developmental angiogenesis, suggesting a unique mechanistic regulation. In conclusion, the OCLN carboxy terminus acts as a non-canonical dynein adaptor linking vesicle cargo to this microtubule minus end directed motor protein through the LIC subunit. The interaction of OCLN to LIC depends on at least 2 regions with one a CC1-like box that exists in the OCLN coiled coil and is flanked by two previously identified phosphorylation sites, S471 found necessary for proper LIC binding and S490 required for regulation of trafficking. OCLN links tight junction containing vesicles to dynein leading to endocytosis and formation of gaps and increased permeability downstream of VEGF signaling while OCLN trafficking to centrosomes is necessary for proper VEGF driven endothelial collateral angiogenesis. Materials and Methods The details on Materials and Methods are included in SI Appendix with all reagents used in SI Appendix Table S1. All animal experiments conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the guidelines established by the Institutional Animal Care & Use Committee of the University of Michigan. SI Appendix includes methods for: common cell and molecular biology techniques including endothelial and U2OS cell culture, transfection, immunofluorescence staining and microscopy, cell proliferation and apoptosis assays, in vitro permeability assay, live cell imaging, and capture assays with references. Animal studies are detailed for middle cerebral artery occlusion, cerebral blood volume measurement with references provided. This study developed the branched retinal vein occlusion model also detailed in SI Appendix. Briefly, branched retinal veins were blocked by a brief pulse from a 532nm laser using a Micron III retinal imaging system and fluorescein injection used to ensure inhibition of blood flow. Oclnfl5/fl5 mice were created using Crispr Cas technology at the University of Michigan Transgenic Core. Supplementary Material Significance Loss of the blood-brain and blood-retinal barrier contributes to a variety of diseases of the central nervous system such as stroke and brain tumors, and blinding eye diseases like diabetic retinopathy. Here, we demonstrate that the carboxy-terminus of the tight junction protein occludin binds to the motor protein dynein regulating barrier properties in a phosphorylation dependent manner by allowing vesicular trafficking of the junctional complex. Further, trafficking to the centrosomes regulates growth factor induced proliferation of the blood vessel cells. Finally, a novel gene deletion demonstrates the necessity of this occludin and dynein interaction in embryonic development. Acknowledgements We thank Stephen I. Lentz PhD for microscopy support; Boxin Zhang for MCAO model support; Ingrid L. Bergin VMD, MS, DACLAM, DACVP and Yao Lee DVM, MS, PhD, DACVP of the ULAM Pathology Core, RRID:SCR 018823. Grant support: NIH EY012021(DAA), S10OD028612 (DAA), HL055374 (DL, DAA) Vision Research Center P30EY007003, Michigan Diabetes Research Center P30DK020572, Research to Prevent Blindness (DAA). Preprint servers The manuscript has been deposited as a preprint to bioRxiv preprint server with the details as below: MS ID#: BIORXIV/2025/659326 https://www.biorxiv.org/content/10.1101/2025.06.12.659326v1 The copyright holder for this preprint is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license. Footnotes Competing interests D.A.A has acted as a scientific consultant for Eyebiotech, Maze Therapeutics and Venrock. Data, Materials, and Software Availability. All study data are included in the manuscript. References Associated Data Supplementary Materials Data Availability Statement All study data are included in the manuscript.