Defective lymphatic valve development and chylothorax in mice with a lymphatic-specific deletion of Connexin43

Abstract

Lymphatic valves (LVs) are cusped luminal structures that permit the movement of lymph in only one direction and are therefore critical for proper lymphatic vessel function. Congenital valve aplasia or agenesis can, in some cases, be a direct cause of lymphatic disease. Knowledge about the molecular mechanisms operating during the development and maintenance of LVs may thus aid in the establishment of novel therapeutic approaches to treat lymphatic disorders. In this study, we examined the role of Connexin43 (Cx43), a gap junction protein expressed in lymphatic endothelial cells (LECs), during valve development. Mouse embryos with a null mutation in Cx43 (Gja1) were previously shown to completely lack mesenteric LVs at embryonic day 18. However, interpreting the phenotype of Cx43^−/− mice was complicated by the fact that global deletion of Cx43 causes perinatal death due to heart defects during embryogenesis. We have now generated a mouse model (Cx43^ΔLEC) with a lymphatic-specific ablation of Cx43 and show that the absence of Cx43 in LECs causes a delay (rather than a complete block) in LV initiation, an increase in immature valves with incomplete leaflet elongation, a reduction in the total number of valves, and altered lymphatic capillary patterning. The physiological consequences of these lymphatic changes were leaky valves, insufficient lymph transport and reflux, and a high incidence of lethal chylothorax. These results demonstrate that the expression of Cx43 is specifically required in LECs for normal development of LVs.

Introduction

The lymphatic system encompasses an ordered network of absorptive and conducting vessels that are essential for maintaining tissue fluid balance, immune function, and the absorption and transport of dietary fat. Defects in the development and function of lymphatic vessels can lead to a number of congenital and acquired pathological conditions, including lymphedema, chylothorax, metabolic disorders, inflammation, and immune dysfunction (Aspelund et al., 2016; Betterman and Harvey, 2016). Critical to vessel function are the lymphatic valves (LVs), cusped luminal structures which ensure that lymph moves in the forward direction only, propelled in part by intrinsic contractions of the lymphatic vascular smooth muscle cells (Bazigou and Mäkinen, 2012). In some instances, congenital valve aplasia or agenesis may be a direct cause of lymphatic disease, as is likely the case with lymphedema-distichiasis syndrome (Kriederman et al., 2003; Petrova et al., 2004) and Emberger syndrome (Kazenwadel et al., 2015a; Sweet et al., 2015). A number of proteins responsible for coordinating LV formation have been identified in recent years, including gene-specific transcription factors, some regulated by disturbed fluid flow, as well as several cell surface receptors and their ligands (Yang and Oliver, 2014; Kazenwadel et al., 2016; Aspelund et al., 2016). Understanding the molecular mechanisms operating during LV development and maintenance will be an important step in establishing novel therapeutic approaches to treat lymphatic disorders.

Connexins (Cxs) are a family of proteins recently shown to be critical for LV formation, but their role during valve development is not well understood. Members of this family (21 members in humans) typically form gap junction intercellular channels, allowing for the direct cell-to-cell transfer of small molecules, including second messengers, but Cxs can also assemble into undocked hemichannels that open transiently to release extracellular signals (Goodenough and Paul, 2009; Evans, 2015). In addition, Cxs may contribute to signaling that is independent from their channel function, by functionally interacting with other vital cellular proteins (Laird, 2010; Zhou and Jiang, 2014; Leo-Macias et al, 2016).

Three Cx family members (Cx37, Cx43, and Cx47) are expressed in murine lymphatic endothelial cells (LECs) and become enriched at LVs, where they are differentially expressed on the upstream and downstream faces of the valve leaflets (Kanady et al., 2011; Kanady and Simon, 2011). Deletion or mutation of these Cx genes results in lymphatic defects in both mice and humans. CX47 gene (GJC2) mutations, for example, have been identified in some families exhibiting dominantly inherited lymphedema (Ferrell et al., 2010). In addition, late-onset lymphedema segregated with affected members in a family with oculodentodigital syndrome, a disorder caused by a mutation in the CX43 gene (GJA1) (Brice et al., 2013). In mice, a null mutation in Cx37 resulted in defective LV development as well as a complete absence of venous valve formation (Kanady et al., 2011; Sabine et al., 2012; Munger et al., 2012). Furthermore, when a Cx37 null mutation was combined with the loss of a single copy of Cx43, the outcome was adult Cx37^−/−Cx43^+/− mice with substantial lymphatic reflux and a high incidence of sudden lethal chylothorax. In an even more severe model, Cx37^−/−Cx43^−/− embryos, lacking both Cxs entirely, exhibited bloody lymph and developed profound lymphedema in utero (Kanady et al., 2011).

There is evidence that Cx43 is necessary for the formation of LVs. Mouse embryos with a null mutation in Cx43 displayed a striking phenotype characterized by the complete absence of mesenteric LVs and an abnormally patterned thoracic duct (Kanady et al., 2011). However, interpretation of the phenotype of mice with a conventional, global knockout of Cx43 was somewhat complicated, first, by the fact that Cx43 is widely expressed during embryogenesis and, second, because Cx43^−/− mice die perinatally due to developmental heart defects (Reaume et al., 1995). Thus, it was difficult to rule out the possibility that some of the lymphatic defects in Cx43^−/− mice could be due to the loss of Cx43 from cell-types other than LECs or that lymphatic defects could in some way be secondary to the cardiac defects.

In this study, we have addressed these issues of specificity by generating a mouse model with a tissue-specific inactivation of the Cx43 gene, using the Cre-lox system (Sauer and Henderson, 1988). In Cx43^ΔLEC mice, Cre expression is driven by the Lyve-1 gene promoter to efficiently and much more specifically ablate Cx43 from LECs. We demonstrate that most Cx43^ΔLEC mice survive to adulthood, thereby circumventing the perinatal lethal heart defects found in Cx43^−/− mice. However, although initially viable, Cx43^ΔLEC mice exhibit lymphatic functional deficits, including valve defects, and are prone to lethal chylothorax.

Materials and Methods

Mice

Cx43 (Gja1)-floxed (Cx43^fl/fl) mice (Liao et al., 2001) and Lyve-1^EGFP-hCre mice (Pham et al., 2010) (referred to hereafter as Lyve-1^Cre mice) were obtained from The Jackson Laboratory (JAX strain numbers 008039 and 012601, respectively) and interbred to obtain Lyve-1^Cre;Cx43^fl/fl mice (referred to as Cx43^ΔLEC mice). Wild-type and Cx43-floxed allele PCR was based on the protocol supplied by The Jackson Laboratory. PCR to detect Lyve-1^+/+ and Lyve-1^Cre alleles was performed according to previously established protocols, using primers Lyve-1-SF10, Lyve-1-K1-QR, and Cre-SFnew1 (Pham et al., 2010; and Jason Cyster, personal communication). With some Cx43^ΔLEC breeders, we noted over time that germline deletion, monitored by PCR using primers YL49 and YL50 (Liao et al., 2001), of one or more floxed Cx43 alleles had occurred. This was not surprising given reports of Cre activity in the testis of Lyve-1^Cre mice (www.informatics.jax.org). In some experiments, both Lyve-1^Cre;Cx43^fl/fl mice and Lyve-1^Cre;Cx43^fl/- mice were used, as both genotypes resulted in efficient ablation of Cx43 from LECs. Control mice for experiments were Lyve-1^+/+;Cx43^{fl/fl or fl/+} and Lyve-1^Cre;Cx43^{+/+ or fl/+} littermates or wild-type mice. The University of Arizona and University of Missouri IACUC Committees approved all animal protocols.

Antibodies

Primary antibodies for immunostaining were: rabbit polyclonal antibodies to Collagen IV (ColIV) (ab19808, Abcam), Cx37 (18264) (Simon et al., 2006), Cx43 (C6219, Sigma), Cx47 (364700, Invitrogen), Laminin α5 (Lam α5) (405, a gift from Lydia Sorokin; used on whole mount samples) (Sixt et al., 2001), Prospero homeobox protein 1 (Prox1) (11-002, AngioBio), Prox1 (ab11941, Abcam), α-Smooth Muscle Cell Actin (SMA), Cy3 conjugated (C6198, Sigma); rat monoclonal antibody to Cluster of Differentiation 31 (CD31) (HM1013, Hycult Biotech), Laminin α5 (Lam α5) (4G6 A2 11, a gift from Lydia Sorokin; used on frozen sections) (Sixt et al., 2001), Lymphatic Vessel Endothelial Hyaluronan Receptor 1 (LYVE-1) (14-0443, eBioscience); goat polyclonal antibodies to Integrin α9 (Itga9), (AF3827, R&D Systems) Vascular Endothelial Growth Factor Receptor 3 (VEGFR3) (AF743, R&D Systems); chicken polyclonal antibodies to Laminin (Lam) (ab14055, Abcam). AffiniPure minimal cross reactivity secondary antibodies (conjugated to Alexa 488, Alex 555, Alexa 647, Cy3, Cy5, or Dylight 649) and Alexa 647 Streptavidin were from Jackson Immunoresearch or Invitrogen. Cx37 antibodies (18264) were directly labeled using the Alexa Fluor 555 Protein Labeling Kit (A20174, Invitrogen).

Section immunostaining

Tissue samples were frozen unfixed in Tissue-tek OCT and sectioned at 10 μm. Sections on slides were fixed in acetone at −20 °C for 10 min, blocked in PBS containing 4% fish skin gelatin, 1% donkey serum, 0.25% Triton X-100, and incubated with primary antibodies for 1.5–3 h at room temperature or overnight at 4 °C. Sections were washed with PBS containing 0.25% Triton X-100 and then incubated with secondary antibodies for 30–40 min. When directly conjugated Alexa 555-Cx37 antibodies were used in conjunction with unlabeled Cx43 or Cx47 rabbit primary antibodies, the Cx43 or Cx47 antibodies were incubated on the sections first, followed by fluorescently labeled anti-rabbit secondary antibodies and a 5% normal rabbit serum blocking step, before the Alexa 555-Cx37 antibodies were applied. Sections were mounted either in Mowiol 40–88 (Aldrich) containing DABCO or in Prolong Gold (Life Technologies) and viewed with an Olympus BX51 microscope and Photometrics CoolSnap ES2 camera or with a Zeiss LSM 510 confocal microscope.

Whole-mount immunostaining

Mesentery was fixed in 1% PFA overnight at 4 °C, washed in PBS, permeabilized with PBS containing 0.3% Triton X-100, and then blocked overnight in PBS containing 3% donkey serum and 0.3% Triton X-100. Ear tissue was treated similarly except fixation was for 1 h at room temperature. Primary and secondary antibodies, diluted in PBS containing 0.3% Triton X-100, were sequentially applied to the tissue overnight at 4 °C. Samples were mounted on slides in Citifluor mountant (Electron Microscopy Sciences). For whole-mount Prox1 immunostaining of E18.5 thoracic duct and diaphragm muscle, the procedure was the same except Vectastain Elite ABC kit (Vector Laboratories) secondary, tertiary reagents, and DAB substrate were used.

Lymphangiography with Evans blue dye

5–12 mice per genotype were anesthetized with an intraperitoneal injection of ketamine (50 mg/kg)/xylazine (20 mg/kg) and placed on a warming pad. Evans blue dye (EBD) (1% w/v) was injected intradermally into both hindpaws and a dissecting microscope was used to trace EBD transport into the iliac lymph nodes and efferent lymphatics (Kriederman et al., 2003). Hindlimb skin and mesenteric lymphatic vessels and nodes were examined for abnormal dye reflux. The thoracic cavity was opened and transport of EBD into and along the thoracic duct was evaluated. Intercostal lymphatic vessels adjacent to the thoracic duct were checked for signs of dye reflux. EBD was also injected into the ear skin to evaluate EBD transport in ear lymphatics.

Lymphatic network characteristics

For analysis of capillary networks in the ear, low magnification images (4X objective) of whole-mount LYVE-1 staining were manually segmented and area density determined using Adobe Photoshop CS4 (N=4 for control and Cx43^ΔLEC samples). Skeletonization of the lymphatic network was created using the Skeletonize (2D/3D) plugin of ImageJ. The AnalyzeSkeleton plugin was used to determine length density, mean vessel diameter, and branch point density as previously described (Kanady et al., 2015). Lymphatic collecting vessel patterning was assessed in P7 mesentery using low magnification panoramic images of whole-mount VEGFR3 staining, constructed from a series of images captured with a 4X objective (N=6 mesentery whole-mounts for control and Cx43^ΔLEC samples). The mean diameter of the large radial collecting lymphatics was determined by dividing the area of each lymphatic vessel by its linear length, measured in Photoshop (N=28 control vessels; N=26 Cx43^ΔLEC vessels).

Quantification of LVs

Mesenteries (5–14 mice per genotype) from E18.5 embryos, P4 pups, or P7 pups were cut into 3–4 segments and whole-mount co-immunostained in 24-well dishes for a subset of markers which highlight LVs: Prox1, CD31, Lam α5, ColIV, Itga9, and VEGFR3. The total number of LVs per mesentery was determined by counting the valves in each segment. In some experiments, the percentage of LVs with immature (stages 1–3) and mature (stage 4) leaflet morphology was also determined (Sabine et al., 2012).

Ears (6 adult mice per genotype) were separated into dorsal and ventral halves and whole-mount co-immunostained for Lam α5, LYVE-1 (expressed in initial lymphatics and pre-collecting lymphatics), and VEGFR3. LVs were counted in eight fields, using the 10x objective, spanning each ear sample and the mean LVs/mm² was recorded. For three ear samples of each genotype, the total number of mature LVs in LYVE-1 negative collecting lymphatics was determined. Total valve length and leaflet cusp height of mature (stage 4) LVs were measured from images using ImagePro Plus software.

Six thoracic ducts from adult Cx43^ΔLEC mice (mean age 9 weeks), including two with ongoing chylothorax, were analyzed for LVs. Following EBD lymphangiography, the thoracic duct was removed, still attached to the aorta, from just above the diaphragm muscle to the region near the top of the heart. The sample was frozen unfixed in Tissue-tek OCT and 10 μm serial sections were collected along its length (mean thoracic duct length sectioned was 9,600 μm per sample). The total number of valves per thoracic duct segment was determined. Selected slides in the series were stained with haematoxylin and eosin (H&E) or immunostained for Prox1 to confirm the identification of the thoracic duct.

Adult valve back-leak test

Popliteal lymphatic collecting vessel segments containing a single valve were dissected from 8-week-old mice, cannulated onto micropipettes and analyzed as previously described (Davis et al., 2012; Scallan et al., 2013). Output pressure, downstream of the valve, was elevated while monitoring pressure, using a servo-nulling micropipette, on the input (upstream) side of the valve. Three replicate tests were performed per sample at low pressure (0.5–10 cm H₂O at a ramp rate of 18 cm H₂O/min). Six Cx43^ΔLEC valves and 4 control (Cx43^fl/fl) valves were analyzed. With Cx43^ΔLEC popliteal lymphatics, shorter or abnormal looking valves were preferentially selected for analysis if present. However, at least one popliteal valve from each mouse was tested, even if all the valves looked normal. Thus, the population of Cx43^ΔLEC valves tested ranged in appearance from abnormally short to normal in length.

In situ imaging of vein valves

Anesthetized mice (N≥7 mice per genotype) were exsanguinated by clipping the right atrium. Brachial veins and superficial caudal epigastric veins were examined for venous valves using a dissecting microscope, as previously described (Munger et al., 2016).

Results

Lymphatic-specific ablation of Cx43

To circumvent the problem of perinatal lethality associated with Cx43 null mice and to address phenotypic specificity, we generated mice with a tissue-specific inactivation of Cx43. Cx43-floxed (Cx43^fl/fl) mice (Liao et al., 2001) were interbred with Lyve-1^Cre mice (Pham et al., 2010) to generate Lyve-1^Cre;Cx43^fl/fl mice (referred to hereafter as Cx43^ΔLEC mice) (Fig. S1A, B). Lyve-1^Cre mice have the Cre recombinase coding sequence inserted into the Lyve-1 locus such that Cre expression is driven by the endogenous Lyve-1 promoter (Pham et al., 2010). LYVE-1 protein is robustly expressed by LECs during normal mouse development, including in the jugular lymph sacs at E12.5, and is also detected at some level in lymph nodes, liver sinusoids, spleen sinuses, blood vessels, endocardium and a subpopulation of macrophages (Banerji et al., 1999; Baluk and McDonald, 2008; Gordon et al., 2008). The Lyve-1^Cre mouse line has been used by others to successfully inactivate, in a lymphatic-targeted manner, the genes encoding sphingosine kinase, VEGFR2, and the chromatin remodeling enzyme CHD4 (Pham et al., 2010; Dellinger et al., 2013; Crosswhite et al., 2016). Dellinger et al. showed that Lyve-1^Cre can be used to delete floxed DNA sequences in LECs that generate collecting lymphatics and their valves, as well as lymphatic capillaries (Dellinger et al., 2013). This strategy works even though collecting lymphatics downregulate LYVE-1 because collecting vessels arise from LYVE-1 positive vessels (Mäkinen et al., 2005; Norrmén et al., 2009). Consistent with these prior studies, immunostaining of P4 or P7 mesenteric collecting lymphatics showed that LEC Cx43 was undetectable in Cx43^ΔLEC mice in contrast to the robust signal observed in control littermates (Fig. 1A–D, S1C–H). Thus, Cx43 was efficiently ablated from the LECs of Cx43^ΔLEC mice.

Figure 1. Cx43 is efficiently eliminated from LECs in Cx43^ΔLEC mice.

Cx43 immunostaining of P7 mesentery sections showed that Cx43 was present in LECs of control Cx43^fl/fl mesentery (A) and (C) but was absent from LECs of Cx43^ΔLEC mice (B) and (D). Transverse sections of lymphatic (ly) vessels are shown in (A) and (B); longitudinal sections are shown in (C) and (D). VEGFR3 staining in (A) and (B) is a marker for LECs. The intense green signal in the arteries (a) in (A) and (B) is due to autofluorescence of the internal elastic lamina. In control Cx43^fl/fl LVs (C), Cx43 was detected on the upstream (u) face of valve leaflets (arrowhead), but was absent from Cx43^ΔLEC valves (D). The expression of Cx37 on the downstream (d) face of valve leaflets in control samples (C) was not affected by the absence of Cx43 in Cx43^ΔLEC valves (D). Lam α5 staining was included in (D) to highlight LV leaflets. a, artery; d, downstream; ly, lymphatic vessel; u, upstream. Scale bars: 20 μm

Cx43 expression was also eliminated in LECs specifically at LVs, as assessed in developing (P4, P7) mesenteric collecting vessel LVs (Figs. 1C, D, S1C–H) and adult thoracic duct valves (Fig. 2A, B) of Cx43^ΔLEC mice. In control valves, Cx43 was detected in LECs on the upstream face of valve leaflets (Figs. 1C, 2A), as previously reported (Kanady et al., 2011) but was absent from Cx43^ΔLEC valves (Figs. 1D, 2B). Cx37 expression, found on the downstream face of control valve leaflets (Figs. 1C, 2A), was not affected by the absence of Cx43 (Fig. 1D, 2B). Interestingly, only very low levels of Cx47 were detected in LECs of thoracic duct valves when Cx43 was eliminated, compared to control valves (Fig. 2C–D′). In a previous study, we showed that Cx47 typically colocalizes with Cx43 in a subset of valve endothelial cells in wild-type thoracic duct (Kanady et al., 2011). Finally, the loss of Cx43 in Cx43^ΔLEC mice was highly cell-type specific, as normal Cx43 expression was observed in venous valves, arterial smooth muscle cells, cardiac myocytes, and adipose tissue (Fig. S2A–D).

Figure 2. Deletion of Cx43 in Cx43^ΔLEC mice reduces expression of Cx47 in thoracic duct valves.

Cx43, Cx37, and Cx47 immunostaining was performed on adult thoracic duct transverse sections. (A) Similar to mesenteric LVs, in control thoracic duct valves, Cx43 was detected on the upstream (u) face of valve leaflets (arrowhead) and Cx37 was found on the downstream (d) face. (B) Cx43 was absent from Cx43^ΔLEC thoracic duct valves (arrowhead), but expression of Cx37 on the downstream face of valve leaflets was unaffected by the absence of Cx43. (C) Cx47 immunostaining (arrowhead) was prominent in a subset of LECs in control thoracic duct valves, but was very low or undetectable in thoracic duct valves of Cx43^ΔLEC mice (arrowhead) (D). The boxed areas in C and D are shown at higher magnification in C′ and D′, respectively. Note that in (D) the two valve leaflets are closely apposed to each other in the section. Lam α5 staining in (B), (D), (D′) highlights the thoracic duct valve leaflets. VEGFR3 staining in (C) and (C′) is a marker for LECs. ct, connective tissue; d, downstream; u, upstream. Scale bars: 20 μm (A) – (D); 10 μm (C′) and (D′).

Chylothorax and lymphatic dysfunction in Cx43^ΔLEC mice

Unlike mice with a conventional knockout of Cx43, Cx43^ΔLEC mice typically survived to adulthood. However, they often died suddenly of chylothorax, with a milky effusion around the heart and lungs, most likely caused by a leak or disruption of the thoracic duct or one its chyle-containing tributaries (Fig. 3A, B). In addition, abnormal accumulation of adipose tissue was observed around the heart and thoracic vertebral column in 7 out of the 14 Cx43^ΔLEC mice that exhibited chylothorax (Fig. S3). The average age at death of 14 Cx43^ΔLEC mice with chylothorax, collected over a 6 month period, was 2.3±0.5 months (all values are means±SEM; range P4 – 6 months). In total, 33 Cx43^ΔLEC mice were generated in our breeding colony during this 6 month period, so the observed frequency of chylothorax was 42%. However, this is likely an underestimate as Cx43^ΔLEC mice were often sacrificed for analysis well before 6 months of age and some of the sacrificed mice may have eventually developed chylothorax if they had been allowed to age. By comparison, chylothorax was not observed in 130 control mice during the same time period.

Figure 3. Chylothorax and lymphatic dysfunction in Cx43^ΔLEC mice.

The chest cavity of a control mouse (A) and a Cx43^ΔLEC mouse exhibiting symptoms of chylothorax (B) was exposed. The Cx43^ΔLEC mouse showed an accumulation of chyle (asterisk) around the heart (ht) and lungs not seen in the control. Chyle was also noted in the intercostal lymphatics (arrow) of a Cx43^ΔLEC mouse with chylothorax (D), indicative of chylous reflux, but this was not observed in the control sample (C). EBD lymphangiography was performed on wild-type controls (E), (G), Cx43^flox/flox controls (I), and Cx43^ΔLEC mice (F), (H), (J). Dye reflux and increased lateral spread was observed when ears of Cx43^ΔLEC mice (F) were injected with EBD, compared with ears of control mice (E). The injection site is marked by a white circle. Cx43^ΔLEC mice showed EBD reflux into a network of lymphatics in the hindlimb skin when the hindpaw was injected (H) but this reflux did not occur in control mice (G). Following hindpaw injection with EBD, there was little or no filling of the thoracic duct (arrow) in Cx43^ΔLEC mice (J) compared with control mice (I).

Chyle was observed in the intercostal lymphatics of a Cx43^ΔLEC mouse with chylothorax, an indication of chylous reflux (retrograde flow) and insufficient lymph transport (Fig. 3C, D). To further investigate whether lymph transport was deficient in Cx43^ΔLEC mice, we performed Evans blue dye (EBD) lymphangiography by injecting EBD into the dermis of the hindpaws and ear skin (Fig 3E–J). 11 out of 12 Cx43^ΔLEC mice showed abnormal EBD reflux into a network of lymphatics in the hindlimb skin associated with the injected hindpaw (Fig. 3H). All 12 of the injected Cx43^ΔLEC mice displayed EBD filling of the lumbar lymph nodes, however there was little or no filling of the thoracic duct in 10 out 12 mice (Fig. 3J). In addition, dye reflux and increased lateral dye spread was observed in 9 out of 9 ear injections (Fig. 3F). These lymphatic functional deficits were noted in Cx43^ΔLEC mice even if they were not currently exhibiting chylothorax. As controls, 5 Cx43^fl/fl mice were similarly injected with EBD (Fig. 3E, G, I) and only one showed mild hindlimb skin dye reflux on one side. In all 5 Cx43^fl/fl controls, lumbar lymph nodes as well as the thoracic duct filled well with EBD and, in addition, transport in ear lymphatics was normal. These results show that lymphatic vascular function was significantly impaired (P < 0.01, Fisher’s exact test, two-tailed) in Cx43^ΔLEC mice compared with Cx43^fl/fl controls.

Lymphatic vessel patterning in Cx43^ΔLEC mice

Since Cx43 is widely expressed by collecting vessel LECs during lymphatic development, we examined if the absence of LEC Cx43 affected collecting vessel patterning in Cx43^ΔLEC mice, using low magnification panoramic images of whole-mount P7 mesentery VEGFR3 immunostaining (Fig. S4). The overall hierarchical organization of lymphatic vessels in P7 Cx43^ΔLEC mesentery looked similar to controls, with collecting vessels typically following closely the organization of the blood vessels. Collecting vessel branching was similar in Cx43^ΔLEC and control mesentery, with smaller collecting lymphatics near the intestinal wall feeding into larger diameter, radially oriented collecting lymphatics. In addition, there was no significant difference in the mean diameter of the large radial collecting lymphatics in Cx43^ΔLEC and control mesentery (83.5±4.5 μm vs. 80.0±4.4 μm, respectively). Surprisingly, however, there were changes in the patterning of lymphatic capillaries in Cx43^ΔLEC adult ear skin (Fig. S5). Although the total area density of lymphatic capillaries, detected by whole-mount LYVE-1 immunostaining, was similar in Cx43^ΔLEC and control ear (0.34±0.02 vs. 0.33±0.03), capillary length density was significantly higher in Cx43^ΔLEC ears (9.4±0.4 mm⁻¹ vs. 7.4±0.3 mm⁻¹) whereas mean capillary diameter was reduced (34.3±1.6 μm vs. 45.3±2.4 μm). In addition, capillary branch point density was substantially increased in Cx43^ΔLEC ears compared to controls (73.4±4.5 branch points/mm² versus 45.3±1.6 branch points/mm²).

Cx43^ΔLEC mice exhibit delayed LV development, reduced numbers of valves, and abnormal leaflet morphology

A previous study showed that LVs were completely absent from Cx43^−/− mesentery lymphatic vessels at E18.5 (Kanady et al., 2011). To determine if valve development was similarly affected in mice with a lymphatic-specific inactivation of Cx43, we performed Prox1 and CD31 whole-mount immunostaining on E18.5 mesentery collected from Cx43^ΔLEC mice (Fig. 4B, D) and littermate controls (Fig. 4A, C). In Cx43^ΔLEC mice, the number of LEC clusters expressing high levels of the Prox1 transcription factor, indicative of developing LVs, was only 3.3% of control values (Fig. 4E). Thus, at E18.5, LV frequency was drastically reduced in Cx43^ΔLEC mice, very much reminiscent of the Cx43^−/− phenotype.

Figure 4. Mesenteric LV number is drastically reduced in Cx43^ΔLEC mice at E18.5.

Prox1 and CD31 whole-mount immunostaining was performed on E18.5 mesentery collected from Cx43^ΔLEC mice (B) and (D) and littermate controls (A) and (C) to detect developing LVs. The number of LEC clusters expressing high levels of Prox1 (arrowheads), indicative of valves, was drastically lower in Cx43^ΔLEC mesenteric lymphatics compared to the controls. Lymphatics are shown at higher magnification in (C) and (D). (E) Quantification of valve counts at E18.5. Controls mesenteries (N=14) had 46.6±3.2 (mean±s.e.m.) valves whereas Cx43^ΔLEC samples (N=7) had 1.57±0.81 valves (3.3% of control). *P<0.0001 versus control (unpaired t test, two-tailed). Scale bars: 100 μm.

Mesentery samples were also examined postnatally, at P4 and P7, to determine if LV development in Cx43^ΔLEC mice was permanently blocked at E18.5 or just delayed (Figs. 5, 6 and S6). Whole-mount immunostaining was done for a number of markers that highlight LVs, including Prox1, CD31, Lam α5, Itga9, Cx37 and VEGFR3, and LV frequency and maturity were evaluated. At both P4 and P7, the total number of valves per mesentery was still greatly reduced compared to controls (29.4% of control at P4; 23.6% at P7) (Fig. 5E, F). Nevertheless, there was an increase in valve number in Cx43^ΔLEC mesenteries at these postnatal stages compared with E18.5, indicating an initial delay in valve development rather than a complete block. However, at P4 and P7, the percentage of LVs that showed mature (stage 4) morphology (Sabine et al., 2012) was reduced in Cx43^ΔLEC samples compared to controls (55.8% of control at P4; 50.3% at P7) (Fig. 5E, F). Higher magnification images of immunostained LVs in P4 and P7 Cx43^ΔLEC and control mesenteries revealed that although stage 4 Cx43^ΔLEC LVs exhibited marker expression qualitatively similar to controls in most cases, many of these valves did not appear fully elongated (Fig. 6). Quantitative image analysis showed that total valve length of mature (stage 4) valves was reduced in Cx43^ΔLEC mesentery compared to controls (65.7% of control at P4; 79.6% at P7) (Fig. 7). Cusp length of the LVs was also reduced (73.3% of control at P7) (Fig. 7). Finally, the recruitment of vascular smooth muscle cells to Cx43^ΔLEC lymphatic vessels, as visualized by α-smooth muscle cell actin immunostaining, appeared normal at P7 (Fig. 6I, J).

Figure 5. Mesenteric LV development is delayed in Cx43^ΔLEC mice.

Prox1 and CD31 (A) and (B) or Laminin α5 (Lam α5) (C) and (D) whole-mount immunostaining was performed on P4 or P7 mesentery collected from Cx43^ΔLEC mice (B) and (D) and littermate controls (A) and (C) to detect LVs (blue arrowheads). At both P4 and P7, the total number of valves per mesentery in Cx43^ΔLEC mice was greatly reduced compared to controls (29.4% of control at P4; 23.6% at P7), however there were more valves present than at E18.5, indicating an initial delay in valve formation. The percentage of LVs that showed mature (stage 4) morphology was also reduced in Cx43^ΔLEC mesentery compared to controls. Quantification of P4 and P7 valves is presented in (E) and (F), respectively. (E) At P4, controls mesenteries (N=6) had 197.2±19.3 valves whereas Cx43^ΔLEC samples (N=7) had 58.0±6.3 valves. 88.9±1.8% of control valves were mature compared to 49.6±2.8% of Cx43^ΔLEC valves. (F) At P7, control mesenteries (N=6) contained 286.0±49.3 valves whereas Cx43^ΔLEC samples (N=5) had 67.6±11.6 valves. 91.8±2.2% of control P7 valves were mature compared to 46.2±5.0% of Cx43^ΔLEC valves. *P<0.05 versus control (unpaired t test, two-tailed). Scale bars: 200 μm.

Figure 6. Mesenteric LVs in Cx43^ΔLEC mice exhibit marker expression that is similar to controls, but valve leaflets are not fully elongated.

Shown are LVs from P4 or P7 Cx43^ΔLEC (B), (D), (F), (H), (J) and Cx43^fl/fl control (A), (C), (E), (G), (I) mesenteries immunostained for several markers that highlight lymphatic vessels and valves. (A) and (B) Prox1; CD31. (C) and (D) Integrin α9 (Itga9); Lam α5; CD31. (E) and (F) VEGFR3; Lam α5. (G) and (H) Collagen IV (ColIV); VEGFR3; CD31. (I) and (J) Lam α5; α-Smooth Muscle Actin (α-SMA); VEGFR3. Expression of these markers appeared similar in control and Cx43^ΔLEC lymphatics, but in many cases the Cx43^ΔLEC valves did not look fully elongated. Scale bars: 50 μm.

Figure 7. Total valve length and cusp length are reduced in Cx43^ΔLEC LVs.

Lam α5; α-SMA; VEGFR3 whole-mount immunostaining of P4 and P7 mesenteric lymphatic vessels was used for image analysis of mature (stage 4) valves in Cx43^ΔLEC and control mice. (A) Three examples of LVs with different orientations within the vessel are shown. Dotted white lines indicate how total valve length was measured in each case. For P7 samples, cusp length was defined as the maximum length of the band of Lam α5 staining (yellow double arrow). (B) At P4, control valves (N=21) had a mean total length of 79.6±3.2 μm whereas Cx43^ΔLEC valves (N=39) were 52.3±3.3 μm (65.7% of control). At P7, control valves (N=63) had a mean total length of 74.5±2.2 μm whereas Cx43^ΔLEC valves (N=41) were 59.3±2.9 μm (79.6% of control). Cusp length in P7 Cx43^ΔLEC valves was 73.3% of controls (27.7±1.4 μm versus 37.8±1.5 μm). *P<0.0001 versus control (unpaired t test, two-tailed). Scale bars: 20 μm.

In adult Cx43^ΔLEC mice, the number of valves in mesentery lymphatics remained low. However, despite the delayed and reduced frequency of valve development, there were nevertheless some mature LVs present in adult Cx43^ΔLEC mesentery (Fig. S7).

We also examined adult ear skin by whole-mount immunostaining for Lam α5, VEGFR3, and LYVE-1 (Fig. 8A–H, S8). As with mesentery, the number of LVs was substantially lower in Cx43^ΔLEC ear lymphatics compared to controls (26.8% of control) (Fig. 8I). Control ear samples contained two morphologically distinct types of LVs, depending on whether or not the lymphatic vessel was LYVE-1 positive (Fig. S8). The first type of valve, found in LYVE-1 positive pre-collecting lymphatics of the dermis, was crescent-shaped with short leaflets (Fig. S8A–C); these were also noted at transition points connecting the pre-collecting lymphatics to LYVE-1 negative collecting lymphatics (Fig. S8D–F). The second type of valve, found only within subcutaneous LYVE-1 negative collecting lymphatics, was more elongated or flame-shaped (Figs. 8A, C, E, G S8G). These collecting vessel valves, in particular, were greatly reduced in number in Cx43^ΔLEC ears (10.5% of control) (Fig 8B, I). In addition, we observed several examples of abnormal valve morphology in the collecting lymphatics of Cx43^ΔLEC ears, such as only one valve leaflet being present or misaligned leaflets (Fig. 8D, F and S8H).

Figure 8. Reduced number and abnormal morphology of lymphatic collecting vessel valves in adult Cx43^ΔLEC ear skin.

Lam α5; VEGFR3; LYVE-1 whole-mount immunostaining was performed on wild-type adult control (A), (C), (E), (G) or Cx43^ΔLEC (B), (D), (F), (H) ear skin. Two types of valves were observed in control ears: crescent-shaped valves with short leaflets in LYVE-1 positive pre-collecting lymphatics in the dermis (see Fig. S8 in the Supplemental Material section for examples of this type) and flame-shaped valves with elongated leaflets in LYVE-1 negative subcutaneous collecting lymphatics (A), (C), (E), (G). In these images, the microscope was focused on the collecting lymphatics only; LYVE-1 positive pre-collecting lymphatics are situated in a different focal plane. In (A) and (B), collecting lymphatics are outlined with white dotted lines. The total number of LVs was substantially lower in Cx43^ΔLEC ear lymphatics compared to controls (26.8% of control). Collecting vessel valves (arrowheads) in particular were severely reduced in number in Cx43^ΔLEC (A) versus control (B) ears. In addition, abnormal valve morphology was observed in the Cx43^ΔLEC collecting lymphatics, such as only one valve leaflet being present (F) or misaligned leaflets (D). (I) Quantification of valves in Cx43^ΔLEC and control ears. Total valve density (pre-collecting and collecting vessel valves) was 6.50±0.45 valves/mm² (N=6) in control ears versus 1.74±0.37 valves/mm² (N=6) in Cx43^ΔLEC ears. Control ears had 76.0±21.5 collecting vessel valves/ear compared with only 8.0±3.0 valves/ear in Cx43^ΔLEC mice. *P<0.05 versus control (unpaired t test, two-tailed). Scale bars: 50 μm.

The high incidence of chylothorax in Cx43^ΔLEC mice suggested a potential defect in thoracic duct valve development. To investigate this possibility, 6 adult Cx43^ΔLEC thoracic duct specimens (mean age 9.0 weeks), extending from diaphragm to heart, were serially sectioned in transverse orientation. Three of the Cx43^ΔLEC samples tested had no thoracic duct valves present and the other three samples had only one valve. Using the same methodology, we previously reported that wild-type adult mice of a similar age (mean age 10.6 weeks) typically had 3–4 valves in the superior half of the thoracic duct (Kanady et al., 2011). Thus, when compared to this control group (N=5), Cx43^ΔLEC thoracic ducts contained only 15.6% of the number of valves found in wild-type mice (0.50±0.22 valves/duct vs. 3.20±0.37 valves/duct; P<0.0001; unpaired t test, two-tailed). Besides the valve deficiency, thoracic duct morphology in adult Cx43^ΔLEC mice looked otherwise normal, except for one instance of bilateral thoracic duct. At E18.5, Cx43^ΔLEC thoracic duct gross morphology also appeared similar to controls, when visualized by whole-mount Prox1 immunostaining (Fig. S9).

Consistent with the lymphatic-specific nature of the gene deletion, venous valves, detected both by immunostaining (Fig. S2A) and by in situ imaging of intact vessels (Fig. S10), were unaffected in Cx43^ΔLEC mice.

Functional defects in LVs of adult Cx43^ΔLEC mice

To determine if there were functional consequences to the abnormal valve morphology in Cx43^ΔLEC mice, we performed valve back-leak tests on isolated, cannulated popliteal lymphatic vessel segments from adult mice (Fig. 9). Valves were subjected to elevations of output pressure while monitoring pressure on the input side of the valve. A graph of mean back pressure versus output pressure showed an overall decrease in resistance in Cx43^ΔLEC valves compared to control (Cx43^fl/fl) valves (Fig. 9B). Out of the 6 Cx43^ΔLEC valves tested, 4 showed some degree of leakiness whereas 2 others showed normal function. Leaky Cx43^ΔLEC valves had shorter leaflet cusp lengths compared to control valves. None of the control valves showed evidence of leakiness (N=4). These results indicate a variable defect in valve function in a subset of LVs in adult Cx43^ΔLEC mice.

Figure 9. Functional defects in LVs of adult Cx43^ΔLEC mice.

Valve back-leak tests were performed on isolated, cannulated popliteal lymphatic vessel segments from adult control (Cx43^fl/fl) and Cx43^ΔLEC mice. Examples of isolated valves tested are shown in (A) with arrows indicating the normal direction of lymph flow. (B) In the graph on the left, a summary of the back-leak test data shows the overall decreased resistance of Cx43^ΔLEC valves (N=6; from 5 mice) compared to control valves (N=4; from 4 mice). The graph on the right shows the change in back pressure at an output pressure (P_out) of 10 cm H₂0 plotted versus the leaflet cusp length for each valve tested. Four out of the 6 Cx43^ΔLEC valves showed varying degrees of leakiness and had shorter leaflet cusps than control valves. *P<0.05 versus control (unpaired t test, two-tailed). Scale bars: 50 μm.

Discussion

Using a Lyve-1^Cre knock-in allele to drive Cx43 deletion in LECs avoided the perinatal lethality associated with Cx43^−/− mice and allowed a more direct test of LEC Cx43 necessity for lymphatic development. The absence of LEC Cx43 caused a delay in LV initiation, an increase in immature valves with incomplete elongation of LV leaflets, a reduction in total valve number, and altered lymphatic capillary patterning. The physiological consequences were insufficient lymph transport, reflux, leaky valves, and chylothorax. Our studies focused on valve development, but further analysis is required to rule out possible additional roles of Cx43 in lymphatic vessel contractility, in the conduction of contractile waves, and in vessel integrity.

Chylothorax in Cx43^ΔLEC mice

Chylothorax in Cx43^ΔLEC mice most likely results from overburdening and rupture of the thoracic duct or intercostal lymphatics due to the deficiency in valves, including those in the thoracic duct. The chylothorax phenotype is similar to Cx37^−/−Cx43^+/− mice which also exhibit a severe reduction in thoracic duct valves (Kanady et al., 2011). In contrast, chylothorax is extremely rare in Cx37^−/− mice (1 case out of ~300 mice), which have a milder deficiency in thoracic duct valves (Kanady et al., 2011). Thus, chylothorax correlates with the severity of thoracic duct valve deficiency in these mouse models, and Cx43 may have a unique role in the formation of thoracic duct valves. An additional feature of Cx43^ΔLEC mice was abnormal adipose accumulation in the thoracic cavity. Adipose buildup has been reported in other mouse models with lymphatic vascular dysfunction, and may be caused by an adipogenic stimulus, such as free fatty acids, present in lymph that has leaked into the tissue (Harvey et al., 2005; Harvey, 2008; Escobedo et al., 2016). Dysfunctional lymph transport could also potentially contribute to the process of adipose accumulation (Escobedo et al., 2016).

Cx43 and lymphatic patterning

The altered capillary patterning in Cx43^ΔLEC ear skin, with smaller diameter vessels and increased branching, was surprising since Cx43 was not detected in the LYVE-1 positive lymphatic capillaries of the adult ear (Kanady et al., 2011). However, it remains possible that Cx43 is expressed by developing lymphatic capillary LECs during embryogenesis and is down-regulated postnatally. Alternatively, the valve deficiency in collecting vessels could have secondary effects on capillary patterning by altering upstream flow patterns.

Cxs and the early stages of valve development

Cx43^ΔLEC mice revealed a delay in LV formation in the absence of LEC Cx43, rather than complete valve agenesis as first suggested by Cx43^−/− embryos (Kanady et al., 2011). Lymphatic-specific deletion of Cx43 allowed an examination of both embryonic and postnatal stages, whereas Cx43^−/− mice die around birth due to heart defects (Reaume et al., 1995). A limitation of the present study is that although LEC Cx43 was not detected in Cx43^ΔLEC mice at P4 or P7, it is difficult to rule out the possibility that delayed LV formation in Cx43^ΔLEC mice, rather than complete valve agenesis, is due to LEC Cx43 expression at an earlier developmental stage. However, the Lyve-1 promoter driving Cx43 gene inactivation is widely expressed in LECs at early stages of lymphatic vessel development, before valve formation occurs. In the developing mesentery, LYVE-1 expression in the lymphatic capillaries begins around E14 and by E16 almost all of the LECs express high levels of LYVE-1 (Norrmén et al., 2009). Using the same Lyve-1^Cre allele, Crosswhite et. al showed that CHD4 is efficiently deleted from Lyve-1^Cre; Chd4^fl/fl LECs by E14.5 (Crosswhite et al., 2016). LV formation in the mesenteric collecting lymphatics, which are derived from the LYVE-1 positive lymphatic capillaries, does not begin until E16. Thus, in Cx43^ΔLEC mice, Cx43 should be deleted in mesenteric LECs before LV development begins.

The signals that induce valve formation in lymphatic vessels are still being determined, but both extrinsic factors, such as fluid shear stress, and intrinsic genetic programs are likely involved (Sabine and Petrova, 2014; Kazenwadel et al., 2015a; Sweet et al., 2015). Previous studies showed that Cx37 plays a critical role in the early stages of valve formation (Kanady et al., 2011; Sabine et al., 2012; Geng et al., 2015). Like Cx37, Cx43 is highly expressed in LECs in embryonic lymphatic vessels (Kanady et al., 2011). However, in contrast to Cx37, which in cultured LECs is jointly regulated by oscillating shear stress and the forkhead transcription factor Foxc2, Cx43 expression is repressed by oscillating shear stress and is not dependent on Foxc2, indicating that the two family members are regulated by different signaling pathways (Sabine et al., 2012). During embryogenesis, there is widespread early expression of Cxs in the lymphatic as well as venous endothelium (Kanady et al., 2011; Munger et al., 2016). As development proceeds, Cx37 and Cx43 become highly enriched at valves but are differentially expressed on the two sides of the valve leaflet, with Cx43 present on the upstream face of the valve leaflet and Cx37 on the downstream face (Kanady et al., 2011; Munger et al., 2016).

Cx43 and valve leaflet elongation

The short LV leaflets in Cx43^ΔLEC mice indicates that Cx43 is required for full leaflet elongation. A similar role for Cx43 may occur during venous valve development (Munger et al., 2016). In cardiac neural crest cells and embryonic cortical neurons, Cx43 is required for directed cell migration, raising the possibility that Cx43 could play an analogous role during valve formation (Kameritsch et al., 2011; Matsuuchi and Naus, 2012; Kotini and Mayor, 2015). Valve leaflet formation involves the elongation, reorientation and coordinated migration of valve-forming endothelial cells (Tatin et al., 2013). Experiments with mouse embryonic fibroblasts suggest that Cx43 plays a role in directed migration by regulating the tubulin cytoskeleton and cell polarity in a channel independent fashion (Francis et al., 2011). Cx43 also interacts with the actin-binding ZO1 protein and associates with other actin-binding proteins (Toyofuku et al., 1998; Giepmans and Moolenaar, 1998; Ambrosi et al., 2016). The Sema3A/Nrp1/PlexinA1 pathway is required for normal LV elongation, and a common downstream target of this pathway is the organization of the actin cytoskeleton (Jurisic et al., 2012; Bouvrée et al., 2012). Moreover, when cardiac neural crest cells were isolated from Cx43^−/− embryos, their cell processes failed to retract normally in response to semaphorin 3A (Xu et al., 2006). Thus, one possibility is that Cx43 influences cell migration during valve development via protein-protein interactions with signaling components that affect the cytoskeleton.

Cx43^ΔLEC mice versus Cx43^−/− mice

Despite similar LV deficiencies at E18.5, there were differences in the phenotype of Cx43^ΔLEC mice versus Cx43^−/− mice. In Cx43^−/− embryos, the thoracic duct was grossly erratic in caliber at E18.5, with blind-ended outcroppings and bifurcated segments, and the lymphatic network on the diaphragm muscle was greatly reduced (Kanady et al., 2011). These defects were not observed in Cx43^ΔLEC embryos (Fig. S5), however, suggesting either a very early requirement for Cx43 in LECs or their precursors, before the Lyve-1 promoter is active, or that the Cx43^−/− -specific defects were due to the absence of Cx43 from a non-LEC cell-type. Cx43 is known to be expressed by mesenchymal stem cells, which can affect lymphangiogenesis (Valiunas et al., 2004; Buttler et al., 2013; Maertens et al., 2014), as well as macrophages and lymphocytes (Glass et al., 2015), which have been implicated in pathological lymphangiogenesis (Betterman and Harvey, 2016). In addition, Cx43 expression in thymic regulatory T cell precursors enhances the production of Foxp3⁺ regulatory T cells (Kuczma et al. 2011), a population of cells which were recently shown to modulate lymphedema and promote lymphatic vessel function (Gousopoulos et al., 2016). During development, lymphangiogenesis in the diaphragm is regulated by macrophages, which are closely associated with the sprouting tips of lymphatic vessels (Ochsenbein et al., 2016). However, macrophages associated with lymphangiogenesis are typically LYVE-1 positive (Kim et al., 2007; Ochsenbein et al., 2016) and therefore Cx43 in those macrophages may be deleted in Cx43^ΔLEC mice, but this remains to be determined. Cx43 could also play a role in chemokine signaling during lymphatic vessel development, with this pathway being altered in Cx43^−/− mice but not Cx43^ΔLEC mice. In zebrafish, chemokine signaling is necessary for thoracic duct patterning (Cha et al., 2012). The relevant LECs in zebrafish express chemokine receptors whereas the cells secreting the chemokines lie outside the developing vessel. Moreover, there is evidence for Cx43 involvement in chemokine signaling, as CXCL12 secretion by bone marrow stromal cells is mediated by Cx43 and Cx45 gap junctions (Schahnovitz et al., 2011). In radial glial cells, CXCL12 and Cx43 colocalize at glio-vascular contacts, and it has been suggested that release of CXCL12 could be regulated by the activation of undocked Cx43 hemichannels that function as release channels (Errede et al., 2014). Thus, in addition to the direct role that Cx43 plays in LECs, it may also influence lymphatic development indirectly via its function in a non-LEC cell-type.

Ear lymphatic valves

Two lymphatic plexuses are generated in the ear by postnatal remodeling, a primary plexus of LYVE-1 down-regulated collecting lymphatics in the subcutaneous layer and a secondary plexus of LYVE-1 positive initial lymphatics in the dermis (Dellinger et al., 2008). We noted two types of LVs in the adult ear: elongated, flame-shaped valves in LYVE-1 negative collecting lymphatics and short crescent-shaped valves in LYVE-1 positive pre-collectors. Intravital staining for perlecan was previously used to describe asymmetric valves in pre-collectors in mouse ear in contrast to bicuspid valves in collecting lymphatics (Kilarski et al., 2014). However, the combination of Lam α5, VEGFR3, and LYVE-1 immunostaining used in our study is particularly effective in visualizing differences in valve morphology. In mesentery, LVs are found in LYVE-1 down-regulated lymphatics. However, valves have been previously noted in LYVE-1 positive pre-collecting ear lymphatics (Lutter et al., 2012), in the LYVE-1 positive lymphatics of the corneal limbus (Truong et al., 2011), and in new corneal lymphatic vessels formed during inflammation (Truong et al., 2011). Our results emphasize the need to consider both varieties of ear LVs when characterizing mouse models, as one valve type may be more or less affected by a genetic modification. In Cx43^ΔLEC mice, the absence of Cx43 in LECs more heavily affected the development of valves in the LYVE-1 negative collecting lymphatics than in the pre-collectors.

Cx mutations and lymphedema

The phenotype of Cx43^−/− mice first suggested a role for Cx43 in lymphatic development (Kanady et al., 2011). Subsequently, a family with oculodentodigital syndrome (ODD) provided evidence that a mutation in the CX43 (GJA1) gene can be an isolated cause of lymphedema in humans (Brice et al., 2013). ODD is a rare, autosomal dominant congenital condition caused by mutations in CX43 that affect the face, eyes, teeth, and digits (Laird, 2014). It remains to be seen if other families with ODD also exhibit lymphatic dysfunction or if this is an infrequent association. The CX43 mutation associated with lymphedema causes an amino acid change in the second extracellular loop, known to be important in docking of hemichannels. One possibility is that the ODD CX43 mutation acts in a dominant negative fashion rather than loss of function (Brice et al., 2013). It has been suggested that mutations in the CX47 (GJC2) gene identified in some families with inherited lymphedema might be dominant negative (Ferrell et al., 2010). Significantly, Cx47 colocalizes with Cx43 on the upstream side of LV leaflets (Kanady et al., 2011). A dominant negative mutation in Cx47 or Cx43 might therefore affect the other coexpressed Cx, particularly if they co-oligomerize. In this study, we noted that Cx47 levels were greatly reduced in thoracic duct valves when Cx43 was eliminated from LECs, suggesting that the expression, assembly, or targeting of Cx47 may depend on Cx43. Furthermore, a Cx47 null mutation alone does not cause LV defects or lymphedema in mice, but a deficiency in both Cx47 and Cx43 does lead to mild lymphedema in some embryos, consistent with a dominant negative model (Munger et al., 2016). Thus, our study not only demonstrates that LEC Cx43 is required for normal development of LVs, it also raises the possibility that a dominant negative Cx47 mutation associated with inherited lymphedema, if it inhibits Cx43, might disrupt lymphatic function by negatively affecting LV development and function.

Highlights.

• A lymphatic-specific deletion of Cx43 was made using a Lyve-1 Cre allele.

• Cx43^ΔLEC mice exhibit insufficient lymph transport, leaky valves, and chylothorax.

• Cx43^ΔLEC mice show delayed lymphatic valve initiation and reduced valve frequency.

• Cx43^ΔLEC mice display immature valves with incomplete elongation of valve leaflets.

• Cx43 is required in lymphatic endothelial cells for normal valve development.