Claudin5a is required for proper inflation of Kupffer's vesicle lumen and organ laterality

Left-right asymmetric organ development is critical to establish a proper body plan of vertebrates. In zebrafish, the Kupffer’s vesicle (KV) is a fluid-filled sac which controls asymmetric organ development, and a properly inflated KV lumen by means of fluid influx is a prerequisite for the asymmetric signal transmission. However, little is known about the components that support the paracellular tightness between the KV luminal epithelial cells to sustain hydrostatic pressure during KV lumen expansion. Here, we identified that the claudin5a (cldn5a) is highly expressed at the apical surface of KV epithelial cells and tightly seals the KV lumen. Downregulation of cldn5a in zebrafish showed a failure in organ laterality that resulted from malformed KV. In addition, accelerated fluid influx into KV by combined treatment of forskolin and 3-isobutyl-1-methylxanthine failed to expand the partially-formed KV lumen in cldn5a morphants. However, malformed KV lumen and defective heart laterality in cldn5a morphants were significantly rescued by exogenous cldn5a mRNA, suggesting that the tightness between the luminal epithelial cells is important for KV lumen formation. Taken together, these findings suggest that cldn5a is required for KV lumen inflation and left-right asymmetric organ development.


Introduction
The internal organs, such as the heart, pancreas, and liver, are asymmetrically located along the midline of the body, even though the external body plane of vertebrates seems symmetric. However, about 1.1 per 10000 new born babies every year show failures of left-right asymmetric organ development, situs inversus totalis or situs ambiguous [1,2]. In addition, one fourth of patients with laterality defects also suffer from primary ciliary dyskinesia, a dysfunction of the cilia in the respiratory tract, sperm cells, or fallopian tube [3,4]. This high correlation between ciliary function and laterality defects enabled the discovery of the left-right organizer (LRO), a transient ciliary organ that controls left-right asymmetric development in vertebrates. The node in mice, gastrocoel roof plate in Xenopus, Hensen's node in chick, and Kupffer's vesicle (KV) in zebrafish were identified as LROs; they show similar, but distinct shapes and sizes, PLOS

Ethics statement
All zebrafish work was carried out in accordance with protocols approved by the Institutional Animal Care and Use Committees of Seoul National University.

Zebrafish
Tuebingen wild-type zebrafish was purchased from the Zebrafish International Resource Center (Oregon, USA), and Tg(sox17:egfp) s870 zebrafish embryos were obtained from Zebrafish International Resource Center through Zebrafish Organogenesis Mutant Bank (Daegu, Korea) [44]. Tg(sox17:egfp) s870 embryos were used for all immunofluorescence experiment except S1 Fig.

Real-time polymerase chain reaction (qPCR)
Target-specific primer pair, cDNA, ROX dye and SyBr Master Mix were prepared for reaction. The reaction conditions were as follows: initial denaturation at 95˚C for 10 min, 40 cycles of denaturation at 95˚C for 15 sec, annealing at 55˚C for 30 sec and elongation at 72˚C for 30 sec. PCR reaction was performed using StepOnePlus real-time PCR system (Applied Biosystems). Primer sequences for PCR amplification are as below: dand5 forward (5'-GCC GTT AGT CAT GTG CCG TT-3') and reverse (5'-CTA TGG GTC AGG ATT GCG GG-3'), eef1a1l1 forward (5'-CTG GAG GCC AGC TCA AAC AT-3') and reverse (5'-ATC AAG AAG AGT AGT ACC GCT AGC ATT AC-3').

Whole mount immunofluorescence
The embryos were fixed with 4% PFA for overnight at 4˚C and dehydrated with methanol and stored at -20˚C. Then the embryos were treated with collagenase I for 10 to 45 min at room temperature depending on the developmental stages. After the collagenase I treatment, the embryos were transferred to the blocking solution (5% bovine serum albumin, 10% goat serum in PBST) and incubated for 3 hours at room temperature. Blocking solution was replaced with primary antibody containing solution and the embryos were incubated at 4˚C overnight. Primary antibodies are anti-Cldn5 (1:50, 35-2500, Invitrogen), anti-ZO-1 (1:50, 339100, Invitrogen) and anti-acetylated tubulin (1:200, T6793, Sigma-Aldrich). A series of washing steps (1% dimethyl sulfoxide [DMSO], 0.5% Triton X-100 in PBST) were performed and the embryos were treated with AF-405, 488, 546 labeled anti-mouse or rabbit IgG in blocking solution for overnight at 4˚C. Then, the embryos were washed with PBDTT (1% DMSO, 0.5% Triton X-100 in PBST) for 10 times. Stained embryos were mounted in glycerol and images were obtained by Zeiss LSM700 confocal microscope with ZEN software.

Morpholino microinjection
1 ng of cldn5a translation-blocking morpholino 1 (MO 1 ), 2 ng of cldn5a translation-blocking MO 2 which targets 5'-UTR and ATG region, respectively, and 2 ng of standard control MO were injected into the yolk at one-cell stage. For specific knockdown at DFCs, MOs were injected into the yolk at 128 to 512-cell stage (referred to as DFC MO). For yolk-specific injection, MOs were injected into the yolk at sphere to dome stage (referred to as Yolk MO). MOs were purchased from Gene Tools, LLC. Sequences of each MO are as follows: cldn5a MO 1 (5'-GTA CTA AAA GGA GTT TAG AAG TTT G-3'), cldn5a MO 2 (5'-AGG CCA TCG CTT TCT TTT CCC ACT C-3'), and standard control MO (5'-CCT CTT ACC TCA GTT ACA ATT TAT A-3').

Statistical analysis
Measurements of KV lumen area and the number and size of cilia were performed using ZEN application (Zeiss). KV lumen area was obtained by measuring the gross area in maximum intensity projection image using Closed Beizer tool of ZEN software. KV cilia size was obtained by measuring the length in maximum intensity projection image using Line tool of ZEN software. Comparisons between control and cldn5a morphants were analyzed in Prism 5 (GraphPad Software, Inc.), and P-values were calculated by applying the unpaired two-tailed Student's t test.

Cldn5a is expressed in KV
The spatiotemporal expression patterns of cldn5a and cldn5b were examined in zebrafish by in situ hybridization. cldn5a was highly expressed within the developing cerebrum, cerebellum, and rhombomeres (red arrow in Fig 1C), while cldn5b was highly expressed in the dorsal aorta and intersegmental vessels as previously reported (green arrow in Fig 1E and green rectangle in Fig 1F, respectively) [42,43]. Interestingly, it was found that cldn5a is expressed in KV at the 6 ss (red rectangle in Fig 1A). To ascertain the serial expression of cldn5a in KV development, embryos were analyzed by immunofluorescence staining with a specific antibody in transgenic zebrafish Tg(sox17:egfp) s870 expressing green fluorescence proteins in KV lineage cells [44]. Though the anti-mammal Cldn5 antibody captured both Cldn5a and Cldn5b (S1 Cldn5 expression was not significant in sox17:egfp-positive cells at 90% epiboly, when the DFCs migrate collectively ( Fig 1G). From the bud to 13 ss, however, Cldn5 was stably expressed in KV cells, while DFCs undergo apical clustering, lumen formation, lumen expansion and KV degeneration (Fig 1H-1M). Moreover, Cldn5 was highly expressed at the luminal surface of the developing KV ( Fig 1N). In addition, a qPCR analysis indicated that cldn5a mRNA expression increased gradually as KV progenitor cells collectively migrated and underwent KV formation (S3 Fig). These data demonstrate that cldn5a is highly expressed at the luminal surface of KV during KV lumen formation and expansion.

Heart laterality is disrupted in cldn5a morphants
Specific expression of cldn5a in KV lineage cells suggests that cldn5a might influence asymmetric organ development in zebrafish. Thus, the status of the heart, a representative asymmetric organ, was investigated by in situ hybridization of cmlc1 in cldn5a-downregulated embryos. Thus, we investigated the role of cldn5a using 5'-UTR-targeting MO 1 , one of the two MO types. Compared with control morphants, cldn5a morphants exhibited significantly higher rates of the middle (42%) and reversed (18%) form of a heart (Fig 2A, 2B and 2E). To verify that the aberrant heart formation in cldn5a morphants was caused by cldn5a deficiency in KV lineage cells, we injected MO into the yolk at the 128 to 512-cell stage (DFC MO) for the exclusive reduction of cldn5a in DFCs (S2D- S2F Fig) and at the sphere to dome stage (Yolk MO) to confirm the effect of MO in the yolk and yolk syncytial layer. DFC cldn5a morphants showed a high rate of disrupted heart laterality (59%), whereas DFC-specific control, yolk-specific control, and yolk-specific cldn5a morphants showed a very low rate of heart laterality defects ( Fig  2C-2E). Finally, mCherry or mCherry-cldn5a mRNA with cldn5a MO was injected to rescue heart laterality defects in cldn5a morphants. In cldn5a morphants with mCherry, 31% of embryos showed a left-sided normal heart; however, 57% of cldn5a morphants with mCherry-cldn5a exhibited a normal heart status (Fig 2F-2I). Thus, these data demonstrated that cldn5a, which is expressed in KV lineage cells, is functionally required for the left-right asymmetric development of the heart.

Signal transfer from KV to organ primordia is disrupted in DFC cldn5a morphants
To validate how defective heart laterality in cldn5a morphants resulted from the specific downregulation of cldn5a in KV lineage cells, we performed in situ hybridization of spaw, the nodalrelated gene, which propagates through the left lateral plate mesoderm (LPM) from KV and determines the laterality of organs [28,29]. DFC cldn5a morphants showed bilateral (51%) and right-sided (7%) spaw expression, while only 2% of the DFC control morphants showed bilateral spaw expression (Fig 3A-3C). Next, the expression of dand5, which is expressed around the KV and acts as a molecular barrier of spaw, was investigated [26,27]. dand5 is expressed bilaterally around KV at 6 ss and is subsequently predominant on the right [27]. Expression patterns of dand5 at 6 ss in DFC control morphants were mostly normal with a horseshoe shape; however, most DFC cldn5a morphants showed abnormal (89%) dand5 expression (Fig 3D-3F). In addition, dand5 expression was not predominant on the right side of KV in most of DFC cldn5a morphants (Fig 3G-3I). Furthermore, quantitative analysis of dand5 revealed the significant reduction in DFC cldn5a morphants (0.313 ± 0.018, Fig 3J). Thus, these data showed that the disrupted heart laterality is correlated with aberrant expression of asymmetric signals.

Ciliogenesis in DFC cldn5a morphants
Ciliogenesis is an important event in asymmetric signal transmission. Motile cilia generate unidirectional fluid flow, and the leftward signal is activated by intracellular Ca 2+ release  [22][23][24][25]. In addition, it was reported that cilia-driven fluid flow is important for dand5 expression [27,45]. Thus, we investigated the status of cilia in DFC cldn5a morphants by immunostaining of acetylated tubulin. The numbers of cilia were 62.12 ± 3.18 and 31.57 ± 3.62 per embryo in DFC control and cldn5a morphants, respectively (Fig 4A-4C). Although the number of cilia decreased in DFC cldn5a morphants, the average length of cilia in DFC cldn5a morphants (4.30 ± 0.050 μm) was comparable to that of the DFC control morphants (4.24 ± 0.075 μm) (Fig 4D). Since a common feature of the downregulation of ciliogenesis factors in the KV was a shortened cilia length as well as a decreased cilia number [17,[46][47][48][49], it is suspected that cldn5a might not be directly involved in ciliogenesis. In this regard, we further identified the reduction of sox17:egfp-positive KV lineage cells in DFC cldn5a morphants from the bud to 10 ss (S5D-S5N Fig). The number of sox17:egfp-positive cells in DFC cldn5a morphants at the bud stage was comparable to that of DFC control morphants. In addition, though the difference of the number of sox17:egfp-positive KV cells between the DFC control and cldn5a morphants started to increase at 3 ss, it did not exceed 25% until 10 ss. This was still less than the reduced rate of cilia in DFC cldn5a morphants (49%) (Fig 4C). Instead, single plain images showed that some sox17:egfp-positive cells did not participate in KV lumen formation and that the cilia were only localized near the small KV lumen in DFC cldn5a morphants ( Fig  4A" and 4B"). Thus, these data suggest that the decreased number of cilia within smaller KV lumen might influence the expression and asymmetry of dand5.

Cldn5a is required for proper inflation of KV lumen
We observed that the KV lumen was small in DFC cldn5a morphants; accordingly, we investigated the role of cldn5a in KV lumen formation and expansion. First, sox17 (a marker for DFC specification [50]) promoter-induced EGFP was stained at the 75% epiboly stage, when the DFCs collectively migrate towards the vegetal pole. In both DFC control and DFC cldn5a morphants, each DFC cluster was normally maintained without fragmentation ( Fig  3A-3C), indicating that DFC specification was not affected by cldn5a. Next, we measured the KV lumen area over time. Fortunately, the localization of ZO-1 at the luminal surface of the KV lumen was not affected by the loss of cldn5a (S6 Fig). Thus, we measured the area enclosed by ZO-1 in Z-stack images by maximum intensity projection. However, since the morphology of the ZO-1-positive KV lumen in DFC cldn5a morphants was a distorted circle, we used the Closed Beizer tool in ZEN software which automatically calculates the enclosed area, rather than obtaining KV max , which multiplies the length of the longest radius. The KV lumen areas of DFC cldn5a morphants and DFC control morphants were 2599 ± 207.3 μm 2 and 5798 ± 315.1 μm 2 at 6 ss, 3286 ± 159.4 μm 2 and 7157 ± 281.0 μm 2 at 8 ss, and 3526 ± 348.9 μm 2 and 7515 ± 446.0 μm 2 at 10 ss, respectively (Fig 5). Considering the KV lumen area and the number of cells in the KV together (S5 Fig), these data suggest that the KV lumen was not fully inflated in DFC cldn5a morphants. In addition, we injected mCherry or mCherry-cldn5a mRNA with cldn5a morpholino to rescue the KV lumen area in cldn5a morphants. Compared with cldn5a morphants with mCherry (lumen area: 2183 ± 176.3 μm 2 ), cldn5a morphants with mCherry-cldn5a exhibited recovery of the KV lumen area (4216 ± 274.2 μm 2 ) (S7 Fig). Claudins are major constituents of tight junction strands and previous papers have reported the size-selective loosening of blood-brain barrier in cldn5 knockout mice and an underinflated ventricular lumen in cldn5a morphant zebrafish [37,42]. Thus, we assumed that paracellular permeability was increased and fluid influx probably leaked through the intercellular space between the KV epithelial cells. Thus, recovery of the malformed KV lumen size in DFC cldn5a morphants by increasing the fluid influx was verified. From 90% epiboly to 6 ss, zebrafish embryos were treated with a mixture of 10 μM forskolin and 40 μM IBMX which increases  fluid secretion through Cftr by elevating the intracellular levels of cAMP by activating adenylyl cyclase and inhibiting phosphodiesterase activity, respectively [14,18,19]. Combined treatment with forskolin and IBMX successfully increased the KV lumen area in DFC control morphants (7020 ± 423.0 versus 5410 ± 307.9 μm 2 , Fig 6A, 6B and 6E). However, the combined treatment failed to increase the KV lumen area in DFC cldn5a morphants (2117 ± 275.0 versus 2266 ± 347.3 μm 2 , Fig 6C-6E), indicating that the reinforced fluid influx leaked through the intercellular space between weakly-adhered KV epithelial cells in the absence of cldn5a. Thus, we suggest that cldn5a is required for KV lumen inflation by supporting paracellular junctions within KV epithelial cells.

Discussion
In the present study, we identified that cldn5a was highly expressed at the luminal surface of KV epithelial cells and sealed the paracellular spaces as a functional component of tight junction during KV lumen formation. DFC cluster, maintained by cadherin-based adherens junctions [12], actively protrudes towards the vegetal pole showing filopodia and adheres to the overlying enveloping layer until polarization and detachment from the enveloping layer to form the rosette structure [9,11]. Then, the rosette structure is apically clustered and exhibits a centered lumen, called KV. Our results revealed that cldn5a was expressed from the bud stages and was abundantly localized at the luminal surface of KV when the DFC detached from the enveloping layer and underwent apical clustering and lumen formation (Fig 1). In addition, sox17:egfp-positive cells in DFC cldn5a morphants at 75% epiboly were grouped together without fragmentation and the number of sox17:egfp-positive cells at the bud stage was similar ( S5  Fig). Thus, we assumed that cldn5a plays a role in sealing the KV luminal surface, but not in maintaining DFC cluster, nor in DFC-enveloping layer adhesion. KV lumen is expanded by fluid influx through the Cftr channel, and the proper size of the KV lumen is important for left-right asymmetric development [14,15,18,35]. It was reported that not only the reduced size of KV lumen, but also over-inflation of the KV lumen by reinforced activation of the Cftr channel disrupted regional cell shape changes of KV epithelial cells and caused defective left-right asymmetry [14,18,19,35]. In this regard, cldn5a downregulation resulted in a decreased KV lumen area (Fig 5). Moreover, combined treatment with forskolin and IBMX, which increases the fluid influx through the Cftr channel, was ineffective to restore the KV lumen size in DFC cldn5a morphants (Fig 6). It suggests that the fluid leaked through the defective intercellular space between the KV epithelial cells in DFC cldn5a morphants, although a more direct way such as dextran injection into KV to probe the loss of paracellular tightness in DFC cldn5a morphants is needed. Nevertheless, the incomplete, but partially formed KV lumen in DFC cldn5a morphants still persisted in the absence of paracellular tightness. We suspect that regional cell shape changes by extracellular matrix accumulation and rock2b-mediated cytoskeletal rearrangement led to hollow spaces between the cells [19][20][21]. In addition, normal localization of ZO-1, the cytoplasmic adaptor protein for the tight junction component, at the luminal surface in DFC cldn5a morphants implies the presence of another tight junction protein, which sustains the partially formed lumen between the weakly adhered cells.
As a ciliated organ, the number of motile monocilia within the KV is important for the functional signal generation. More than 30 cilia are needed to drive sufficient force of fluid flow and induce the early left-right markers, dand5 and spaw [45,51]. In addition, the adequate length of cilia is important for left-right asymmetric development, as longer or shorter cilia lengths disrupt organ laterality by regulating the cilia beating frequency [27,45,52]. Although ciliogenesis occurs by complicated mechanisms, including FGF, Hedgehog, Notch, and nc-Wnt signaling pathways [53], Navis et al. [14] suggested that the generation of cilia is not related to lumen expansion. In their model, cftr mutant, cilia were generated within the center of the KV lumen even though the luminal space was undetectable [14]. In our case, a decreased number, but normal length of cilia was located near the under-inflated lumen in DFC cldn5a morphants (Fig 4). This discrepancy could be interpreted as ciliogenesis is related to apical-basal polarity, rather than sufficient space. Considering that neighboring KV epithelial cells lose their paracellular tightness, we suspect that a part of nonlumen-facing KV cells might lose their apical-basal polarity in DFC cldn5a morphants. However, the relevance between lumen formation and the potency of ciliogenesis needs further elucidation.
As a consequence of the reduced number of cilia within the defective KV lumen, dand5 was aberrantly expressed in DFC cldn5a morphants (Fig 3), indicating KV dysfunction. Besides, the expression of dand5 was decreased in DFC cldn5a morphants (Fig 3J), though cilia-driven fluid dynamics were inversely related to dand5 expression [45]. We suspect that this might be the combined result of defective KV with high permeability and a decreased number of cilia. One possible explanation is that the initial transcription of dand5 might be reduced due to the reduction of KV lumen-facing cells. Another one is that leaking fluid might affect the degradation of dand5. In any case, the expression of spaw was not fully restricted to the left LPM showing bilateral or right-sided patterns, as a consequence of the aberrant expression of dand5, and the heart laterality was disrupted in DFC cldn5a morphants (Figs 2 and 3).
The current study utilized morpholino-based approach to determine the role of cldn5a in zebrafish. Two types of cldn5a morphants showed similar phenotypes of disrupted heart laterality and exogenous RNA rescued the cldn5a morpholino-induced phenotypes. Several recent papers reported that genome-engineered mutants failed to recapitulate the morpholinoinduced phenotypes [54][55][56]. Although genetic compensation might be induced in mutant embryos [57], morpholino-based studies should be carefully considered whether their phenotypes are driven by off-target effects. In this regard, we conducted CRISPR/Cas9-mediated cldn5a mutation and analyzed whether heart laterality might be affected in cldn5a crispants [58]. First, we constructed two types of cldn5a-targeting guide RNA (gRNA) chimera which directs Cas9 to forward and reverse strands of the genomic cldn5a, respectively. gRNA2, which targets reverse strand, successfully mutated cldn5a. Moreover, gRNA2-mediated cldn5a crispants showed high rate (61%) of disrupted heart laterality (S8 Fig). Thus, together with the morpholino-based studies, these data further support the role of cldn5a in organ laterality determination.
Taken together, we identified the functional tight junction component, cldn5a, in KV, and its role in proper inflation of KV lumen and left-right asymmetric organ development. Thus, these results advance our understanding of KV lumen formation and organ laterality.